Vous êtes sur la page 1sur 413

Contents

1 INTRODUCTION 2 BASIC QUESTIONS IN UNIT OPERATIONS 2.1 Macroscopic mass balance . . . . . . . . . 2.2 Macroscopic energy balance . . . . . . . . 2.3 Macroscopic momentum balance . . . . . . 2.4 Heat balance . . . . . . . . . . . . . . . . . 2.5 Dimensional analysis . . . . . . . . . . . . 2.6 Conclusion . . . . . . . . . . . . . . . . . Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 4 4 5 5 6 6 8 9 10 10 11 12 13 14 16 16 17 18 18 18 19 19 21 22 25 26 26 27 29 30 30 31 32

3 REVIEW OF MACROSCOPIC BALANCES 3.1 Mass balance . . . . . . . . . . . . . . . . . . . 3.2 Forces on a converging angle tting . . . . . . . 3.3 Friction losses in pumping . . . . . . . . . . . . 3.4 Torque needed for centrifugal pump . . . . . . . 3.5 Heat exchanger design . . . . . . . . . . . . . . 3.6 Correlations for heat transfer coefcients . . . . . 3.7 Correction for diffusion in a stagnant lm . . . . 3.8 Conclusion . . . . . . . . . . . . . . . . . . . . Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . 3.A Introduction . . . . . . . . . . . . . . . . . . . . 3.B Spatial specication of a vector . . . . . . . . . . 3.C Specication of a vector in terms of Unit vectors 3.D Coordinate independence of vectors . . . . . . . 3.E Vectors and laws of nature . . . . . . . . . . . . 3.F Vector algebra . . . . . . . . . . . . . . . . . . 4 THE CONTINUUM MODEL 4.1 The continuum model . . . . . . . . . . . . . 4.1.1 Point quantities and volume averaging 4.1.2 Point quantities and area averaging . 4.2 Derivatives of point quantities . . . . . . . . . . . . . . . .

5 THE TRANSPORT THEOREM 5.1 Fluid domains: Imagined . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.2 A system in a continuum of uid . . . . . . . . . . . . . . . . . . . . . . . . . 5.3 The Reynolds Transport Theorem . . . . . . . . . . . . . . . . . . . . . . . . i

5.4 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33 Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34 5.A Velocity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34 6 PHYSICAL LAWS & FLUID SYSTEMS 6.1 Control volume and system . . . . . . . . . . . . . 6.2 The law of conservation of mass . . . . . . . . . . 6.3 Newtons second law or Linear momentum balance 6.4 The law of conservation of angular momentum . . 7 WHAT IS A FLUID? 7.1 What is a uid? . . . . . . 7.1.1 Behaviour of solids 7.1.2 Behaviour of uids 7.2 Forces on uid elements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36 36 38 39 39 41 41 41 43 43 45 45 46 46 47 47 48 48 49 49 49 51 51 53 55 56 58 58 59 59 61 62 63 63 63 65 66 66 67 68

8 STRESS TENSOR 8.1 Direction of pressure . . . . . . . 8.2 Stress tensor . . . . . . . . . . . . 8.2.1 Area as a vector . . . . . . 8.2.2 Tensor and directions . . . 8.3 Cauchys theorem . . . . . . . . . 8.3.1 Surface forces on a system 8.4 Pressure tensor . . . . . . . . . . Appendix . . . . . . . . . . . . . . . . . . 8.A Tensors as dyadic products . . . . 8.B Tensor algebra . . . . . . . . . . .

9 SURFACE FORCES & CONSTITUTIVE RELATIONSHIPS 9.1 Strain rate . . . . . . . . . . . . . . . . . . . . . . . . . . . 9.2 The Newton Stokes law of viscosity . . . . . . . . . . . . . 9.3 Molecular interpretation of the stress tensor . . . . . . . . . 9.4 Momentum diffusivity . . . . . . . . . . . . . . . . . . . . Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9.A Vector calculus . . . . . . . . . . . . . . . . . . . . . . . . 9.B Tensors and frames of reference . . . . . . . . . . . . . . . 9.B.1 Metric tensor . . . . . . . . . . . . . . . . . . . . . 9.B.2 Differentiation . . . . . . . . . . . . . . . . . . . . 9.B.3 Second order tensors . . . . . . . . . . . . . . . . . 9.B.4 Principal coordinates . . . . . . . . . . . . . . . . . 9.B.5 Invariants . . . . . . . . . . . . . . . . . . . . . . . 9.B.6 Frame indifferent equations . . . . . . . . . . . . . 10 PHYSICS OF FLUID FLOW 10.1 Pressure ows and drag ows . . . . . . . . . . 10.2 Development of ow due to drag forces . . . . 10.3 Development of ow due to pressure gradients . 10.4 Eulerian and Lagrangian description . . . . . . ii . . . . . . . . . . . . . . . . . . . . . . . . . . . .

10.5 10.6 10.7 10.8

The no slip boundary condition Stress boundary conditions . . Incompressibility of uids . . Laminar and turbulent ows .

. . . .

. . . .

. . . .

. . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

68 70 70 71 73 73 74 74 75 77 77 77 78 79 80 81 82 82 83 84 84 86 86 86 87 87 88 89 89 90 90 91 92 92 92 93 93 93 94 95 96 96 97 97 97

11 VISUALIZATION OF FLOW 11.1 Stream lines . . . . . . . . . . . . . . 11.2 Flow in a converging diverging section 11.3 Flow past a sphere . . . . . . . . . . . 11.4 Flow in a pipe . . . . . . . . . . . . .

12 SHELL BALANCE OF MOMENTUM 12.1 Laminar lm owing down an inclined plane . . . 12.1.1 Problem statement . . . . . . . . . . . . . 12.1.2 Simplications and assumptions . . . . . . 12.1.3 Mass balance . . . . . . . . . . . . . . . . 12.1.4 Momentum balance . . . . . . . . . . . . . 12.1.5 Combining rate law with balance law . . . 12.1.6 Scaling . . . . . . . . . . . . . . . . . . . 12.1.7 Boundary conditions . . . . . . . . . . . . 12.1.8 Volumetric ow rate . . . . . . . . . . . . 12.1.9 Looking back . . . . . . . . . . . . . . . . 12.1.10 Flow transition . . . . . . . . . . . . . . . 12.2 Laminar ow in a pipe . . . . . . . . . . . . . . . 12.2.1 Problem statement . . . . . . . . . . . . . 12.2.2 Simplications and assumptions . . . . . . 12.2.3 Mass conservation . . . . . . . . . . . . . 12.2.4 Momentum balance . . . . . . . . . . . . . 12.2.5 Rate law combined with balance law . . . . 12.2.6 Scaling . . . . . . . . . . . . . . . . . . . 12.2.7 Boundary conditions . . . . . . . . . . . . 12.2.8 Volumetric ow rate . . . . . . . . . . . . 12.2.9 Looking back . . . . . . . . . . . . . . . . 12.2.10 Friction factor . . . . . . . . . . . . . . . . 12.2.11 Flow transition . . . . . . . . . . . . . . . 12.3 Stratied ow of two immiscible uids in a channel 12.3.1 Problem statement . . . . . . . . . . . . . 12.3.2 Simplications and assumptions . . . . . . 12.3.3 Mass conservation . . . . . . . . . . . . . 12.3.4 Momentum balance . . . . . . . . . . . . . 12.3.5 Rate law combined with balance law . . . . 12.3.6 Scaling . . . . . . . . . . . . . . . . . . . 12.3.7 Boundary conditions . . . . . . . . . . . . 12.3.8 Volumetric ow rates . . . . . . . . . . . . 12.3.9 Looking back . . . . . . . . . . . . . . . . 12.3.10 Flow transitions . . . . . . . . . . . . . . . 12.4 Conclusion . . . . . . . . . . . . . . . . . . . . . iii

Appendix . . . . . . . . . . . . . . . . . . . . . 12.A Need for different coordinates . . . . . 12.B Unit vectors in curvilinear coordinates . 12.B.1 Scale factors . . . . . . . . . . 12.C Vector algebra in curvilinear coordinates 12.D Tensor algebra in curvilinear coordinates

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

99 99 100 101 104 105 107 107 108 108 109 109 111 112 114 114 115 115 118 119 120 123 123 123 124 125 125 126 126 126 127 127 128 130 130 131 131 131 131 132 132 133 134 134 134

13 EQUATIONS OF MOTION 13.1 Equation of continuity . . . . . . . . . . . . . . . . . . 13.1.1 Incompressible uids . . . . . . . . . . . . . . . 13.2 Interpretation of divergence of a vector . . . . . . . . . . 13.3 Substantial derivative . . . . . . . . . . . . . . . . . . . 13.4 Cauchys equation of motion . . . . . . . . . . . . . . . 13.5 Navier Stokes equation . . . . . . . . . . . . . . . . . . 13.6 Stream function . . . . . . . . . . . . . . . . . . . . . . 13.6.1 Axisymmetric ows . . . . . . . . . . . . . . . 13.6.2 Boundary conditions and stream function . . . . Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13.A Equations of uid motion in different coordinate systems 13.B Vector calculus in curvilinear coordinates . . . . . . . . 13.C Tensor calculus . . . . . . . . . . . . . . . . . . . . . . 13.D Tensor calculus in curvilinear coordinates . . . . . . . . 14 VISCOUS FLOWS 14.1 Old wine in new bottles . . . . . . . . . . 14.1.1 Flow down an inclined plane . . . 14.1.2 Flow in a pipe . . . . . . . . . . . 14.2 New wine in old bottles . . . . . . . . . . 14.3 Unsteady viscous ow . . . . . . . . . . 14.3.1 Problem statement . . . . . . . . 14.3.2 Simplications and assumptions . 14.3.3 Equation of continuity . . . . . . 14.3.4 Navier Stokes equations . . . . . 14.3.5 Initial and boundary conditions . 14.3.6 Similarity solution . . . . . . . . 14.4 Start up of ow in a pipe . . . . . . . . . 14.4.1 Simplications and assumptions . 14.4.2 Problem statement . . . . . . . . 14.4.3 Equation of continuity . . . . . . 14.4.4 Navier Stokes equation . . . . . . 14.4.5 Scaling . . . . . . . . . . . . . . 14.4.6 Initial and boundary conditions . 14.4.7 Solution by separation of variables 14.4.8 Looking back . . . . . . . . . . . 14.5 Slow ow past a sphere . . . . . . . . . . 14.5.1 Assumptions and simplications . 14.5.2 Problem statement . . . . . . . . iv . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

14.5.3 Scaling . . . . . . . . . . . . . . . . . . . 14.5.4 Equation for the stream function . . . . . . 14.5.5 Boundary conditions . . . . . . . . . . . . 14.5.6 Solution of the partial differential equation 14.5.7 Drag force . . . . . . . . . . . . . . . . . 14.5.8 Form and friction drag . . . . . . . . . . . 14.5.9 Looking back . . . . . . . . . . . . . . . . 14.5.10 Flow transitions . . . . . . . . . . . . . . . Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . 14.A Separation of variables . . . . . . . . . . . . . . . 14.B Sturm-Liouville problems . . . . . . . . . . . . . . 14.B.1 Regular Sturm-Liouville problem . . . . . 14.B.2 Singular Sturm-Liouville problem . . . . .

. . . . . . . . . . . . .

. . . . . . . . . . . . .

. . . . . . . . . . . . .

. . . . . . . . . . . . .

. . . . . . . . . . . . .

. . . . . . . . . . . . .

. . . . . . . . . . . . .

. . . . . . . . . . . . .

. . . . . . . . . . . . .

. . . . . . . . . . . . .

. . . . . . . . . . . . .

. . . . . . . . . . . . .

. . . . . . . . . . . . .

. . . . . . . . . . . . .

. . . . . . . . . . . . .

135 136 136 137 138 139 139 139 140 140 141 142 142 153 153 155 159 160 160 161 162 163 164 164 165 165 165 167 167 169 170 171 171 172 172 173 173 174 174 175 176 176 177 178

15 APPROXIMATIONS IN FLUID MECHANICS 15.1 Pseudo steady state approximation . . . . . . . . . . . . . . . . . . . . . . . . 15.2 Lubrication approximation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15.3 Solution by expansions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16 HIGH REYNOLDS NUMBER LIMIT & d ALEMBERTS PARADOX limit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16.1 High 16.2 Elimination of the no-slip boundary condition . . . . . . . . . . . . . . 16.3 Importance of irrotational ows . . . . . . . . . . . . . . . . . . . . . Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17 MORE PHYSICS OF FLUID FLOW 17.1 Vorticity . . . . . . . . . . . . . . . . . 17.2 Why vorticity? . . . . . . . . . . . . . 17.2.1 Vorticity and viscous forces . . 17.2.2 Vorticity and energy extraction . 17.2.3 Vortices in ow past bodies . . 17.3 Vorticity and kinematics . . . . . . . . 17.3.1 Shearing and straining motions . 17.3.2 Rotational motion . . . . . . . 17.4 Vorticity equation . . . . . . . . . . . . 17.4.1 Diffusion of vorticity . . . . . . 17.4.2 Amplication of vorticity . . . 17.4.3 Conservation of vorticity . . . . 17.4.4 Generation of vorticity . . . . . 17.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

18 IRROTATIONAL FLOWS 18.1 Velocity potential . . . . . . . . . . . . . 18.2 Determination of pressure . . . . . . . . . 18.3 Irrotational ow past a cylinder . . . . . . 18.3.1 Boundary conditions . . . . . . . 18.3.2 Solution by separation of variables 18.3.3 Determination of pressure . . . . v

18.3.4 Drag force on the cylinder . . . 18.4 Velocity potential and complex variables 18.4.1 A few important planar ows . . 18.5 Stream function in potential ows . . . 18.6 Solutions based on the operator . . . 19 BOUNDARY LAYER THEORY

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

179 180 181 183 184 188 189 190 195 196 197 197 198 198 199 199 200 200 200 201 202 202 203 204 204 205 206 206 206 208 208 210 211 211 212 213 214 214 214 215 216 216 217 219

20 STABILITY AND FLOW TRANSITIONS 20.1 Instabilities due to density/gravity . . . . . . . . . . . . . . . . . . . . . . . . 20.2 Shear layer instability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20.3 Inexion points and instability . . . . . . . . . . . . . . . . . . . . . . . . . . 21 TURBULENCE 21.1 Phenomenological description . . . . . . . . . . . . . . . . . . . . . . . 21.2 Characteristics of turbulent ow . . . . . . . . . . . . . . . . . . . . . . 21.2.1 Indeterministic nature of turbulent ows . . . . . . . . . . . . . . 21.2.2 Enhanced diffusive power of turbulence . . . . . . . . . . . . . . 21.3 Postulated mechanisms and possible approaches . . . . . . . . . . . . . . 21.4 Statistical theory of turbulence . . . . . . . . . . . . . . . . . . . . . . . 21.5 Statistical description . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21.5.1 Probability distribution . . . . . . . . . . . . . . . . . . . . . . . 21.5.2 Measurement of probability distribution . . . . . . . . . . . . . . 21.5.3 Fluctuations around the average . . . . . . . . . . . . . . . . . . 21.6 Reynolds averaged Navier Stokes equations (RANS) . . . . . . . . . . . 21.6.1 Closure problem . . . . . . . . . . . . . . . . . . . . . . . . . . 21.6.2 Reynolds stresses . . . . . . . . . . . . . . . . . . . . . . . . . . 21.6.3 Experimental measurements . . . . . . . . . . . . . . . . . . . . 21.6.4 Spectrum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21.7 Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21.8 Kolmogorovs theory of energy cascade . . . . . . . . . . . . . . . . . . 21.8.1 Homogeneous and isotropic turbulence . . . . . . . . . . . . . . 21.8.2 Generation of many length scales . . . . . . . . . . . . . . . . . 21.8.3 Kolmogorov length scale . . . . . . . . . . . . . . . . . . . . . . 21.8.4 Velocity uctuations and length scale . . . . . . . . . . . . . . . 21.8.5 Prediction of spectrum of velocity correlation . . . . . . . . . . . 21.9 k Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21.9.1 Equation for k . . . . . . . . . . . . . . . . . . . . . . . . . . . 21.9.2 Equation for dissipation . . . . . . . . . . . . . . . . . . . . . . Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21.A Diffusion & random movements . . . . . . . . . . . . . . . . . . . . . . 21.A.1 Characteristics of random motion . . . . . . . . . . . . . . . . . 21.A.2 Random motion and concentration distribution . . . . . . . . . . 21.A.3 Characteristics of the random motion and the diffusion coefcient 21.A.4 Langevin equation for diffusion . . . . . . . . . . . . . . . . . . 21.A.5 Dispersion & Diffusion . . . . . . . . . . . . . . . . . . . . . . . 21.A.6 Taylors theory of turbulent dispersion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

vi

22 MACROSCOPIC BALANCES AND FIELD EQUATIONS 22.1 Averaged eld equations . . . . . . . . . . . . . . . . . 22.1.1 Mass balance . . . . . . . . . . . . . . . . . . . 22.1.2 Momentum balance . . . . . . . . . . . . . . . . 22.1.3 Mechanical energy balance . . . . . . . . . . . . 22.1.4 Equation of mechanical energy . . . . . . . . . . 22.1.5 Macroscopic mechanical energy balance . . . . . 22.2 Dimensional analysis . . . . . . . . . . . . . . . . . . . 22.2.1 Pressure drop in a tube of varying cross section . 22.2.2 Drag on a sphere due to oscillating ow . . . . . 22.2.3 Scale up problems . . . . . . . . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . .

223 223 223 224 226 226 227 228 228 230 232 233 233 234 234 234 235 235 236 237 237 238

23 FIRST LAW OF THERMODYNAMICS & HEAT BALANCE 23.1 First law of thermodynamics . . . . . . . . . . . . . . . . . 23.2 First law and closed systems . . . . . . . . . . . . . . . . . 23.2.1 Rate of change of energy . . . . . . . . . . . . . . . 23.2.2 Rate of work done . . . . . . . . . . . . . . . . . . 23.2.3 Rate of heat supplied . . . . . . . . . . . . . . . . . 23.3 Equation of mechanical energy balance . . . . . . . . . . . 23.4 Thermal energy balance equation . . . . . . . . . . . . . . . 23.5 Heat balance over a control volume . . . . . . . . . . . . . . 23.6 Heat ux through boundaries . . . . . . . . . . . . . . . . . 23.7 Connection to temperature . . . . . . . . . . . . . . . . . .

24 MECHANISMS OF HEAT TRANSFER & CONSTITUTIVE RELATIONSHIPS 239 24.1 Heat Flux Vector . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 239 24.1.1 Direction of heat ow . . . . . . . . . . . . . . . . . . . . . . . . . . 239 24.2 Heat Conduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 240 24.3 Fouriers law of heat conduction . . . . . . . . . . . . . . . . . . . . . . . . . 240 24.3.1 Direction of heat ux vector . . . . . . . . . . . . . . . . . . . . . . . 240 24.4 Radiation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 241 24.5 convection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 242 24.6 Classication of heat transfer problems . . . . . . . . . . . . . . . . . . . . . 242 24.7 Similarities and differences between momentum & heat transfer . . . . . . . . 243 25 PHYSICS OF HEAT TRANSFER 25.1 Thermodynamic equilibrium at the interface . . . . . . . . . . . 25.2 Development of temperature prole: Constant temperature case 25.3 Boundary conditions . . . . . . . . . . . . . . . . . . . . . . . 25.4 Development of temperature prole: Constant ux case . . . . . 25.5 Onset of natural convection . . . . . . . . . . . . . . . . . . . . 25.6 Convective heat transfer . . . . . . . . . . . . . . . . . . . . . . 25.7 Solution of heat transfer problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 244 244 245 246 247 248 248 249

26 SHELL BALANCES IN HEAT TRANSFER 251 26.1 Heat transfer in a series of slabs . . . . . . . . . . . . . . . . . . . . . . . . . 251 26.1.1 Statement of problem . . . . . . . . . . . . . . . . . . . . . . . . . . . 251 26.1.2 Simplications and assumptions . . . . . . . . . . . . . . . . . . . . . 252 vii

26.2

26.3 26.4

26.5 26.6

26.7

26.1.3 Heat balance over a control volume . 26.1.4 Combining rate law with heat balance 26.1.5 Scaling . . . . . . . . . . . . . . . . 26.1.6 Boundary conditions . . . . . . . . . 26.1.7 Looking back . . . . . . . . . . . . . Temperature distribution in a heated wire . . 26.2.1 Problem statement . . . . . . . . . . 26.2.2 Simplications and assumptions . . . 26.2.3 Heat balance over a control volume . 26.2.4 Combining rate law with heat balance 26.2.5 Scaling . . . . . . . . . . . . . . . . 26.2.6 Boundary conditions . . . . . . . . . 26.2.7 Temperature prole . . . . . . . . . . 26.2.8 Looking back . . . . . . . . . . . . . Heat transfer transfer coefcient . . . . . . . 26.3.1 Bulk temperature . . . . . . . . . . . 26.3.2 Heat transfer coefcient . . . . . . . Heat transfer in laminar forced convection . . 26.4.1 Problem statement . . . . . . . . . . 26.4.2 Heat balance over a control volume . 26.4.3 Coupling to uid mechanics . . . . . 26.4.4 Simplications and assumptions . . . 26.4.5 Combining rate law with balance . . 26.4.6 Scaling . . . . . . . . . . . . . . . . 26.4.7 Boundary conditions . . . . . . . . . 26.4.8 Solution of the equation . . . . . . . 26.4.9 Looking back . . . . . . . . . . . . . 26.4.10 Bulk temperature . . . . . . . . . . . 26.4.11 Heat transfer coefcient . . . . . . . 26.4.12 Conclusion . . . . . . . . . . . . . . Convective heat transfer to slabs . . . . . . . Temperature proles in a bearing . . . . . . . 26.6.1 Problem statement . . . . . . . . . . 26.6.2 Simplications and approximations . 26.6.3 Heat balance . . . . . . . . . . . . . 26.6.4 combining rate law with heat balance 26.6.5 Boundary conditions . . . . . . . . . 26.6.6 Scaling . . . . . . . . . . . . . . . . 26.6.7 Temperature proles . . . . . . . . . 26.6.8 Looking back . . . . . . . . . . . . . Fully developed temperature prole . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

252 253 253 253 255 255 255 255 256 257 257 258 259 259 260 260 261 262 262 262 263 264 265 265 266 267 268 268 269 269 270 270 271 271 272 273 273 273 274 274 275 282 282 282 283 284

27 CONSERVATION EQUATIONS FOR NON-ISOTHERMAL SYSTEMS 27.1 Balances on a system . . . . . . . . . . . . . . . . . . . . . . . . . . . 27.2 Thermal energy balance on a system . . . . . . . . . . . . . . . . . . . 27.2.1 Divergence theorem . . . . . . . . . . . . . . . . . . . . . . . 27.3 Application to arbitrary system . . . . . . . . . . . . . . . . . . . . . . viii

27.4 Temperature form . . . . . . . . . . . . . . . . . . . . . . . 27.4.1 Constitutive equation . . . . . . . . . . . . . . . . . 27.5 Boundary conditions . . . . . . . . . . . . . . . . . . . . . 27.6 Coordinate systems . . . . . . . . . . . . . . . . . . . . . . 27.7 Solution methodology . . . . . . . . . . . . . . . . . . . . . Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27.A Equations of energy balance in different coordinate systems .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . .

284 285 285 285 285 287 287 288 288 288 289 289 290 290 292 292 294 300 302 317

28 Solution of thermal energy balance equation 28.1 Old wine in new bottles . . . . . . . . . . . . . . . . . . . . . . . . . . . 28.1.1 Heat transfer through slabs in series . . . . . . . . . . . . . . . . 28.1.2 Temperature distribution in a heated wire . . . . . . . . . . . . . 28.1.3 Heat transfer in laminar forced convection . . . . . . . . . . . . . 28.2 New wine in old bottles . . . . . . . . . . . . . . . . . . . . . . . . . . . 28.2.1 Viscous heating in Couette ow . . . . . . . . . . . . . . . . . . 28.2.2 Solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28.2.3 Heat trasnfer in a radial ow catalytic reactor . . . . . . . . . . . 28.2.4 Leveque solution: convective heat transfer for short contact times 28.2.5 Heat trasnfer to a semi-innite slab . . . . . . . . . . . . . . . . 28.2.6 Heat conduction in an innite slab . . . . . . . . . . . . . . . . . 29 THERMAL BOOUNDARY LAYER 30 NATURAL CONVECTION 30.1 Boussinesq approximation . . . . . . . . . . . . . . . . . . . 30.2 Fully developed temperature and velocity prole . . . . . . . 30.2.1 Equation of continuity . . . . . . . . . . . . . . . . . 30.2.2 Equation of motion . . . . . . . . . . . . . . . . . . . 30.2.3 Coefcient of expansion . . . . . . . . . . . . . . . . 30.2.4 Thermal energy balance . . . . . . . . . . . . . . . . 30.2.5 Fully developed temperature prole . . . . . . . . . . 30.2.6 Equation of motion . . . . . . . . . . . . . . . . . . . 30.2.7 Scaling and velocity prole . . . . . . . . . . . . . . 30.2.8 Grashoff number . . . . . . . . . . . . . . . . . . . . 30.2.9 Looking back . . . . . . . . . . . . . . . . . . . . . . 30.2.10 Transitions . . . . . . . . . . . . . . . . . . . . . . . 30.3 Boundary layer formed due to natural convection . . . . . . . 30.3.1 Inuence of Prandtl number . . . . . . . . . . . . . . 30.4 Scaling in natural convection boundary layers . . . . . . . . . 30.4.1 Estimate of buoyancy force . . . . . . . . . . . . . . . 30.4.2 Prandtl number is unity . . . . . . . . . . . . . . . . . 30.4.3 Prandtl number 1 . . . . . . . . . . . . . . . . . . 30.4.4 Prandtl number 1 . . . . . . . . . . . . . . . . . . 30.4.5 Heat ux in natural convection . . . . . . . . . . . . . 30.5 Similarity solution of natural convection on a vertical at plate 30.5.1 Looking back . . . . . . . . . . . . . . . . . . . . . . 30.5.2 Transitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . .

320 320 321 321 322 322 322 323 323 324 324 324 325 325 326 326 327 327 328 329 329 329 332 332

ix

31 PHYSICAL LAWS FOR MULTICOMPONENT SYSTEMS 31.1 Multicomponent body . . . . . . . . . . . . . . . . . . . . 31.1.1 Mass average velocity . . . . . . . . . . . . . . . 31.2 Physical laws applied to a multicomponent body . . . . . . 31.3 Law of conservation of mass of species . . . . . . . . . . 31.3.1 Consistency with balance of total mass . . . . . . 31.4 Momentum balance for a multicomponent body . . . . . . 31.5 Birds eye view . . . . . . . . . . . . . . . . . . . . . . . 32 MECHANISM OF DIFFUSION 32.1 Diffusive motion and diffusive ux . . . . . . . . 32.2 Diffusive ux vector . . . . . . . . . . . . . . . 32.2.1 Direction of mass transfer . . . . . . . . 32.3 Mass transfer by diffusion . . . . . . . . . . . . 32.3.1 Alternative denitions of diffusive uxes 32.3.2 Molar average velocity . . . . . . . . . . 32.4 convection . . . . . . . . . . . . . . . . . . . . . 32.5 Multicomponent systems and other mechanisms . 33 PHYSICS OF MASS TRANSFER 33.1 Equilibrium at the interface . . . . . . . . . . 33.2 Development of concentration proles . . . . 33.2.1 Mass transfer with chemical reactions 33.3 Boundary conditions . . . . . . . . . . . . . 33.3.1 Concentration conditions . . . . . . . 33.3.2 Flux conditions . . . . . . . . . . . . 33.3.3 Mass transfer coefcient . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

334 335 335 336 337 339 340 341 342 342 343 343 344 345 345 346 346 348 348 348 349 350 350 351 353 356 356 357 359 359 361 362 362 362 362 363 364 364 365 366 366 366 366 367

34 SPECIAL FEATURES OF MASS TRANSFER PROBLEMS 34.1 Bulk velocity due to diffusion . . . . . . . . . . . . . . . 34.2 Bulk velocity and equation of motion . . . . . . . . . . . 34.3 Dilute solutions . . . . . . . . . . . . . . . . . . . . . . . 34.4 Molar form of mass balance . . . . . . . . . . . . . . . . 34.5 Classication of mass transfer problems . . . . . . . . . . 35 SHELL BALANCES IN PURE DIFFUSION PROBLEMS 35.1 Equimolar counter diffusion . . . . . . . . . . . . . . . 35.1.1 Problem statement . . . . . . . . . . . . . . . . 35.1.2 Simplications and approximations . . . . . . . 35.1.3 Balance of mass of species
. . . . . . . . . . 35.1.4 Combining balance with constitutive equation . . 35.1.5 Boundary conditions . . . . . . . . . . . . . . . 35.1.6 Scaling . . . . . . . . . . . . . . . . . . . . . . 35.1.7 Looking back . . . . . . . . . . . . . . . . . . . 35.1.8 Similarity to heat transfer . . . . . . . . . . . . . 35.2 Vaporization of solute into an insoluble gas . . . . . . . 35.2.1 Problem statement . . . . . . . . . . . . . . . . 35.2.2 Simplications and assumptions . . . . . . . . . x . . . . . . . . . . . .

35.2.3 Balance of species . . . . . . . . . . . . . . . 35.2.4 Combining balance with constitutive equation . 35.2.5 Boundary conditions . . . . . . . . . . . . . . 35.2.6 Mass balance equation . . . . . . . . . . . . . 35.2.7 Scaling . . . . . . . . . . . . . . . . . . . . . 35.2.8 Solution . . . . . . . . . . . . . . . . . . . . . 35.2.9 How is stagnant? . . . . . . . . . . . . . . . 35.2.10 Dilute solutions . . . . . . . . . . . . . . . . . 35.2.11 Looking back . . . . . . . . . . . . . . . . . . 35.2.12 Similarity to conduction heat transfer problems 35.3 Catalytic reaction . . . . . . . . . . . . . . . . . . . . 35.3.1 Problem statement . . . . . . . . . . . . . . . 35.3.2 Simplications and assumptions . . . . . . . . 35.3.3 Balance of species . . . . . . . . . . . . . . . 35.3.4 Combining balance with constitutive equation . 35.3.5 Boundary conditions . . . . . . . . . . . . . . 35.3.6 Mass balance equation . . . . . . . . . . . . . 35.3.7 Scaling . . . . . . . . . . . . . . . . . . . . . 35.3.8 Solution . . . . . . . . . . . . . . . . . . . . . 35.3.9 Dilute solutions . . . . . . . . . . . . . . . . . 35.3.10 Looking back . . . . . . . . . . . . . . . . . . 35.4 Summary and nal comments . . . . . . . . . . . . . 36 EQUATIONS OF SPECIES MASS BALANCE 37 DIFFUSION BOUNDARY LAYER 38 ADVANCED TOPICS IN MASS TRANSPORT

. . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . .

368 368 369 369 370 371 371 372 372 373 373 373 374 374 374 375 376 376 377 378 379 379 389 393 395

xi

QUOTABLE QUOTES
The development of general ability for independent thinking and judgement should always be placed foremost, not the acquisition of special knowledge. If a person masters the fundamentals of his subject and has learned to think and work independently, he will someday find his way and besides will be better able to adapt himself to progress and changes than the person whose training principally consists in acquiring of detailed knowledge. Einstein

The Simpler the better" is a good rule for an engineer in search of a theory. He is interested in truth of course; but it is the truth of an artist, which consists in emphasizing essentials and stimulating the imagination, not in accumulating detail without regard to relevance.

Launder and Spalding in "Lectures in Mathematical Models of Turbulence", p 23.

This problem of closing gaps and making practical way is Engineering. It requires problems, although there are no new basic long way to go from basic principles to a

the thing work in the most serious study of design principles....., there is a practical and economic design.

Feynaman v II, (16-8) (Quoted out of context!)

Chapter 1 INTRODUCTION
An engine in general can be dened is a device made to achieve some purpose: an engine of change. In a broad sense, a steam turbine changes heat into power. Petrochemical plants change crude oil into other chemicals. Engineering as a discipline then concerns itself with carrying out processes needed to achieve the purpose. In chemical plants for example, one is interested in various kinds of changes: changing the temperature in a reactor or altering the concentrations of species in a stream or pumping (changing the position) products into storage tanks and so on. What tells us if the changes are feasible or not? Thermodynamics considers closed systems and species conditions under which their state will continue i.e., remains unlatered, or they exist in equilibrium. Consider one such criterion:

  

where , is the change in the entropy between any new state and the current state of the system, and  and  are the temperature and internal energy respectively. If the above criterion is satied, the system is in equilibrium and therefore will not move from its current state 1 . Thermodynamics also implies that if the above condition is not satised, the system will move away from its current state, say to some new state. The criterion can be applied to the new state as well. If the criterion is not satised, the state will change till the criterion is satied, i.e., till equilibrium is reached 2 . For example, thermodynamics can predict from the above criterion that when two bodies are brought into contact heat ows from the hotter body to the colder body and that the temperatures of both bodies are equal when equilibrium is attained. Another equivalent condition of equilibrium states that if for a particular change from the current state of the system3 .  !#"$ % 



the change is feasible4 . This particular form of the equilibrium criterion predicts that when two solutions containing a solute at different concentrations are brought into contact, mass of solute transfers from regions of higher concentration to those of lower concentrations so that
Of course, it has to be kept in mind that the values of and are xed. Remember that we are talking about closed systems. Thus, the above argument will not be true if the system is continually forced to move in some way. 3 It is being assumed that extraction of any forms of useful work other than mechanical is absent. 4 Conversely, the system is in equilibrium if
2 1

&

'

(*),+-

.0/ 13254
1

chemical potential, and hence concentration in this example, becomes equal everywhere in the nal mixture. Processes involving changes in the state of a system, similar to those considered in the two examples, form the basis of Unit Operations. The criteria discussed above can be understood in a more useful way. In a system, there could different parts or regions, which are not in the same state (e.g., not having the same  or  be ), and when change takes place, the entire system reaches an identical, and minimal (or maximal) state where no further change is possible. In all such processes involving a change, a difference characterizing the nonequivalence of two regions is identiable 5 . In specic instances, it can be related to easily measurable quantities. In heat transfer it is the difference in temperature. In mass transfer, it is the difference in the chemical potential, which often is equivalent to difference in concentration. Thus, due to the spatial variation of these characteristic differences, commonly referred to as driving forces, transport of heat and mass occurs. Although it is not apparent from the commonly-used texts on Unit Operations, the ow of uids can be identied with transfer of momentum. The three processes of momentum, heat and mass transfer are commonly referred to as transport processes, and constitute the knowledge base of Unit Operations. In this sense, understanding transport processes is, in principle, equivalent to understanding the fundamentals of all Unit Operations 6 . The molecular mechanisms underlying the three transport processes turn out to be similar. As a result, the transport processes themselves show analogous (but not identical) behavior on macroscopic level as well. These are the commonly referred to analogies between momentum, heat and mass transfer. For this reason, it is advantageous to study them together. The present lectures are an introduction to the study of the transport processes. In thermodynamics, the differences in functions, e.g. internal energy, between two states are calculated by imagining a reversible process that takes the system from one state to another. In such a process, changes occur in innitesimal increments; correspondingly, the driving forces are also innitesimal in magnitude. It is apparent that the rate of such a process is also innitesimally small. Reversible processes have the maximum efciency since they cause no increase in entropy of the universe. The reversible path, though it is the most efcient one, has to be abandoned if the process is to be carried out at a rate that is practically useful. It is expected that the rate of a process increases when the magnitude of the driving force is increased. However any process carried out with a nite driving force is irreversible. Such processes are commonly referred to as irreversible or dissipative processes7 . As the processes occur at a nite rate, they are also known as rate processes. The rate processes involving tranport of momentum, heat and mass are also referred to as tranport processes or transport phenomena. One of the main questions that need to be answered in the study of tranport processes is: What is the dependence of the rate of a transport process on the driving force? The relationships between the rate of a process and the driving force are referred to as constitutive relationships or rate laws. These form one of the topics to be studied in these lectures.
Very often, it is the free energy difference. Please do note the accent on in principle. The peculiarities and differences between various equipment and the detailed knowledge required to design and operate them are important. Thus it is said that physics forms a basis for understanding chemistry but that does not mean knowing physics means knowing chemistry. Understanding from a fundamental view point helps in general, and in fact, in appreciating the differences as well. 7 An irreversible process is accompanied by a nite free energy change which could have been used to produce useful work had a reversible path been followed. In this sense the process has dissipated the driving potential. In momentum transfer (and to a smaller extent, in mass transfer), heat is produced as the nal result of this dissipative nature of the irreversible process. Heat transfer results in an increase in entropy or in dissipation of the organized nature of thermal energy.
6 5

Any process is aimed at carrying out a change. It is expected that if the extent of change to be brought about is known, the rate law would provide an estimate of the time required to accomplish the change. Consider a simple example. Suppose 10 kg of water are available at $  7 6 87 C. Call this the input state. Suppose we desire to raise its temperature to 9 C. Call this the output state. The extent of change in thermal energy required is the difference in the thermal energy of the output and input states, and in the present example, it is approximately equal to 1600 kJ. If we are able to supply heat at a constant rate (an unrealistic rate law as it turns out) of 1 kW, it would require 1600 : to do the job. Real processes are very complex: they may be batch, semibatch or continuous; they may operate under steady-state or unsteady conditions. Hence an accurate accounting is required to assess the extent of desired change. Balance laws provide the accounting procedures that accurately characterize the extent of change required to reach the output state from the input state. Formulating the balance laws in general and simplifying them to specic examples forms a major portion of our lectures. The simple example above showed how balance laws are combined with constitutive relationships to provide answers to practical questions. Several examples, more complex than the above but still simple, of combining balance laws with constitutive relationships will be considered in this course. These form the stepping stones to develop a physical feel for problems, which is essential to develop an ability to analyze more complex problems.

Chapter 2 BASIC QUESTIONS IN UNIT OPERATIONS


Mass and energy balances are very familiar to chemical engineers. Procedures to make these balances are given in many undergraduate books on Unit Operations, and are sometimes derived using some intuitive arguments. Making such balances is intimately connected with the subject of Transport Processes. We start with the laws governing balances, or balance laws for short, as written in Unit Operations books, and raise questions about their origins. These questions would concretely illustrate the need for a fundamental outlook and provide motivation for studying Transport Processes.

2.1 Macroscopic mass balance


Consider any equipment through which material is owing. It could be a storage vessel (see Figure 2.1). We draw a boundary to mark the control volume, i.e., the domain of interest. We
Output Feed Flow rate = m in Flow rate = mout

Figure 2.1: Mass balance around an equipment. Control volume is marked by dotted line. usually write the following mass balance equation for the control volume: Rate at which Rate of Rate of input of Rate of output of mass is accumulation of = mass into the mass from the + mass in the generated in the equipment equipment system equipment If we are interested in making a mass balance at steady state1 , the accumulation term is zero. Now we assert that, according to the law of conservation of mass, mass can neither be
1

A process is said to be at steady state if none of the variables characterizing it change with time. Same is

created nor destroyed. Hence the last term is equal to zero. Thus for the storage vessel

; =@?!A <C ; BEDGF <>

However, the law of conservation of mass is valid for a system2 , and the above is an intuitive generalization of the law to a control volume or to an open system. How is the generalization done precisely?

2.2 Macroscopic energy balance


Another commonly used equation in uid mechanics is the Bernoulli equation. It is commonly written as Work done by the control volume This is a very important and useful equation. Why is it valid? All of us are familiar with the law of conservation of energy or the rst law of thermodynamics. As with the previous example, it is also valid for closed systems only. An intuitive generalization of the rst law is seen in books on thermodynamics, where losses are attributed to the irreversible nature of the process. Those generalizations also have a term dealing with the rate of supply of heat. Does it mean that the Bernoulli equation is a version of the rst law of thermodynamics for isothermal systems where heat is not being supplied? May we separate heat and work terms like that? What should be done for nonisothermal systems? Further, Unit Operations books give expressions for losses. I  For example, when a uid ows at an average velocity in a pipe of length W and diameter , Frictional losses

HJI$K H I$K J A LNMO PQMSRUT8V =@? L MNOPCMSRUT8V BEDGF M 

W Head lost due to friction AYX[Z L 

I$K

where Z is the friction factor. This has its origins in Newtons laws, and some books do contain a derivation. Why does the Bernoulli equation itself not follow from Newtons laws? Or does it? How do we know that losses are given by this expression? Almost all Unit Operations books show that velocity varies with position in a tube i.e., a velocity prole exists in a tube, and that it is parabolic in laminar ow. Then why should we use the square of the average velocity and not the average of the square of velocity in the above expression?

2.3 Macroscopic momentum balance


Here is another typical problem that is solved in text books. Find the force F required to keep a pipe tting in place when uid enters it at a velocity \ =@? (see Figure 2.2). We write
the case with an equilibrium state, but it is distinguished from steady state by the fact that it has no potential for change. 2 We use the following nomenclature throughout these notes. We use the word system to mean the same as closed system in Thermodynamics. Thus a set of objects of the world selected for observation is referred to as a system. The system, by denition then, always consists of the same set of objects. Hence objects are not exchanged between the system and surroundings. The law of conservation of mass states that the mass of a closed system, or system according to our nomenclature, is constant.

Vout

Vin F

Figure 2.2: Momentum balance around a tting.

a momentum balance, keeping in mind that momentum is a vector. The balance, however, is similar to the mass balance we wrote earlier. For steady state it is written as Rate of output of Rate of input of Sum of forces acting momentum from the = momentum into the + on the equipment equipment equipment Newtons second law is valid for a closed system, i.e., for a system with constant mass, and the above equation is once again an intuitive generalization of Newtons law to open systems. How is it done? Often, in Unit Operations, problems in uid ow are solved using energy balance, e.g., power requirements for pumping a uid. Why do we not apply momentum balances to such and, indeed all, problems involving the ow of uids? After all Newtons laws are the basis of motion!

2.4 Heat balance


Finally, consider a typical heat transfer problem solved in text books: transfer through a composite slab. Here we use the following conservation law: Rate at which Rate of Rate of input of Rate of output of heat is accumulation of = heat into the thin heat from the + generated in the energy in a thin slice slice slice slice We simplify this for steady state and, in the absence of heat generation, show that temperature proles are linear provided thermal conductivities are independent of temperature. Again, is it not strange that the rst law connected with energy does not talk about heat and work separately and yet we talk of heat balance? Although this may be justied in conduction problems, where work terms are absent, we also use heat balances in other equipment where ow occurs. When is the use of heat balances justied?

2.5 Dimensional analysis


Dimensional analysis is widely used in Unit Operations. To apply this type of analysis, all the variables important to the problem have to be guessed at rst. It is expected that the variables are related to each other through some equation, which of course is not known, and that is why 6

dimensional analysis is being attempted. However, such an equation must be dimensionally consistent, i.e., the left and right hand sides must have the same units. The units that commonly appear in chemical engineering problems are mass, length, time, and temperature. If four suitable variables are selected from the set guessed, all the other variables of the set can be rendered dimensionless in terms of the selected variables. Thus a set of dimensionless groups are arrived at. The equation that relates all the guessed variables is equivalent to a relationship between the dimensionless groups. The equation that relates the dimensionless groups to each other is still not known. However, this procedure has two advantages. Firstly, we get an idea of how to plan experiments since we can often assign meaning to dimensionless groups. For example, we interpret the Reynolds number to be the ratio of inertial to viscous forces. Secondly, it combines the effect of several variables and hence each of the variables need not be investigated separately for its effect. Thus, if we want to ascertain the effect of velocity and viscosity on the pressure drop in a pipe, analysis tells us that it is sufcient if we investigate the effect of the group ] I P$^_ dimensional . Let us take an example. Suppose that we are interested in calculating the drag force ` on a sphere while a liquid ows past it with a velocity a . In applying dimensional analysis, we have to guess the variables that are important. We would expect that the frictional forces exerted by _ the liquid would depend upon its viscosity , and upon the area of contact between the sphere ] and the liquid. The latter depends upon the diameter of the sphere . Hence we expect ` to ] _ be related to a , and through some complex equation:

` Ab adc ] c _ 

The particular function is not known. Force has dimensions of mass multiplied by acceleration, or length divided by the square of time. We now select the variables from those that we thought were important to represent the dimensions of mass, length and time, and write force in terms of these. We can select diameter to represent length: ] Length e Velocity has units of length divided by time. Let us then use diameter and velocity to represent the units of time: ] Time e Viscosity has units of mass divided by the product of length and time. Therefore we can use viscosity to represent the unit of mass:

] K _ Mass e a

Hence

The unknown equation that was written earlier for ` can now be written as

`fe

Mass. Length/square of Time

e _] a

] c_  b  ` a c A _] a a ] _
7

However, the left hand side is dimensionless i.e., it has no units. The right hand side also must have no dimensions and the only possibility is that it is a constant 3 . Thus

_ ]` a A

constant

which, you would recognize, is Stokes law. Of course the constant is not known. The constant _ ] and must be determined experimentally. However it is easily seen that the effects of c a need not be investigated separately, and hence the extent of experimentation required to determine the constant is reduced considerably. You also know that this is not the complete story. We know that this is true only for creeping ow. So where did we go wrong? The error crept in while making the list of variables that are important in the problem. Suppose we thought that the inertia or the acceleration of the uid P as it goes around the sphere is also important. Then we must include the density of the uid , in the list. Now we can replace the dimensions of density as follows:

P e
and we get another ratio4 :

Mass/cube of Length

e a ]

Pa ] _ _ ]` a A
Function of

which is of course the Reynolds number. Dimensional analysis would indicate that we should expect a relation like P ]

a  _

Reynolds number is the ratio of inertial to viscous forces. When the ow is slow, inertial forces will be small and Reynolds number will be small. Hence, based on the previous analysis, the right hand side can be expected to become a constant as Reynolds number becomes small. Notice how the meaning assigned to Reynolds number helps us in arriving at this conclusion. One question would arise in our minds. How do we guess the right number and the correct variables? Further, how do we assign meaning to the dimensionless groups?

2.6 Conclusion
Several basic questions have been raised about the material learnt in Unit Operations. Those can be answered only after a study of Transport Processes. In these lectures, we study transport processes, and connect it to the treatment to the material learnt in Undergraduate courses.

See Appendix for this longer footnote. Different groups arise depending upon the variables chosen to represent the basic dimensions. The Buckingham-Pi theorem tells us how many independent groups will be obtained from the kind of analysis carried out.
4

Appendix : Dimensional Analysis


One way of understanding the argument of dimensional analysis is as follows. Suppose one set of people are using MKS system of units. Another set of people, as in USA and UK, could be using FPS system. We expect that any system of units must obey the following restriction. If two masses <hg and < K , are weighed by the followers of the two systems, they should nd ^ the same value for the ratio <ig < K . Similar restriction follows for lengths and time. This is possible only if the two units are related to each other by a constant value. Thus a one meter long scale has 3.28 feet in it. It is convenient to think of change of units as change of scale used to make measurement, and the conversion factor as the scale factor. Thus we have four fundamental scale factors for each system of units: one each for mass, length, time, and temperature. Notice that the ratios we considered had no units or they were dimensionless, and are independent of unit of measurement. We come across several quantities other than mass, length and time. e.g., force. Units of such quantities are derived from the fundamental units, e.g., Newtons are related to j R c < and : . The scale factors to convert units of such derived quantities from one system to another are not independent and can be obtained from the three fundamental scale factors. It follows that units of derived quantities in different systems are also related to each other by a constant scale factor, and that ratios of derived quantities will also be independent of the system of ^ units. Thus if two forces, ` g and ` K , were measured in two different units, the ratio ` g ` K will be dimensionless and is independent of the system of units. One more important result also follows. Any quantity, which is formed by several quantities but which is dimensionless, also will have values independent of the system of units. In other words, dimensionless quantities are independent of scale of measurement. ^_ ] a . This quantity is dimensionless, and hence must be indeNow consider the ratio ` ] _  pendent of the scale of measurement. Now consider the function b adc c . The numerical ] and _ will be different in different system of units. When these values are values of ac ] _ ^a ] _ substituted, the numerical value of b may be different. However, b adc c must be independent system of units, and hence must be equal to a constant. Further, a quantity adc ] c of  ] _ ^ _ b g K clk ck cnmnmom a like , where k g ck K comnmnm are dimensionless groups, must be independent of the system of units and hence, at best, be a function of k g ck K comnmnm as these are independent of the system of units.

Chapter 3 REVIEW OF MACROSCOPIC BALANCES


Review a few examples from Unit Operations, from Mass conservation and thermal energy balances. A review of vectors is given in the appendix. In this chapter, we review solutions of a few problems illustrating the typical approach of Unit Operations, and raise more questions to motivate study of transport processes. The examples will cover mass conservation and thermal energy balances. Additionally, some aspects of correlations also will be discussed. In this notes, vector and tensor notation will be used, and the material will be covered in appendicies of chapters as needed. A review of vectors is given in the appendix of this chapter.

3.1 Mass balance


Consider a centrifugal pump. It sucks liquid from it center and throws it out at its periphery. Viewed from a frame of reference moving with the vane, the uid going out at the periphery is expected to leave almost tangentially to the vanes if the spacing between the vanes is small 1 . Hence, viewed from a stationary frame of reference, uid also has a velocity corresponding to the rotation of the impeller. At the inlet it is expected to enter almost radially. When it comes into contact with the impeller it also acquires velocity corresponding to the rotation of the impeller 2 . We want to relate the velocity at the periphery to that at the inlet value. Refer to Figure 3.1. Take a stationary control volume that encloses the impeller but just outside it as shown in the gure3 . Hence, at steady state, application of the law of conservation of mass gives the result that the rate of input of mass into the control volume minus the rate of out put of mass from the control volume is equal to zero. The two surfaces of the control volume are at p K and p g . Let qsr and qut be the unit vectors in the p and v direction4 . Thus, if we assume that
Refer to discussion in McCabe and Smith The resultant velocity is shown in Figure 3.1 as . If the resultant velocity is not tangential to the vane, then the entry of the uid is not smooth and this leads to losses. 3 The dashed lines show the stationary control volume. They are supposed to coincide with the entry to the and periphery of the impeller. For the sake of clarity, they are separated. 4 We are assuming that you are familiar with hnit vectors in cylindrical coordinates. These however are reviewed in appendix of the chapter on Shell balance of momentum.
2 1

wyx / zE{}|

10

V2

Angle between tangent to vane and tangent to the impeller is 2 r

2 r

V 1

1 V 1,rel r 1

Figure 3.1: Velocity vectors in a centrifugal pump. On the right, a cross section is shown.

the velocity is uniform over the entire surface, we get

where g and K are the velocities of the uid with respect to a stationary frame of reference, and  is the thickness of the impeller. We assumed that the uid leaves tangentially to the vane as viewed from a frame of reference moving with the impeller. Hence, K A qutp K M K where K is a vector whose direction is parallel to the tangent to the vane. Noting that the dot product of unit vectors in the radial and angular directions is zero, the above equation of mass conservation can then be rewritten as

 A L~ k p K qsrnm K L k~p g  s q rom g

where a is the volumetric pumping rate. Thus we have succeeded in relating the outlet and inlet velocities the pumping rate. We also note that the rst term in the equation is equal to  L k~p K I K8, : to 8 ^ L k . Hence

L k~p K qyrnm K A L kp g  I d g A a;

; ; I K A a a A L k~p KGG, : k ^8L  L ~ k p K :nE

We will return to this later when we relate the outlet velocity to the power consumed by the pump.

3.2 Forces on a converging angle tting


As a uid ows through a converging angle tting, the direction of motion of the uid changes; and hence its acceleration is not zero. This is in addition to the change in the magnitude of the velocity itself. Thus the tting has to exert a force on the uid to create the acceleration and a force has to be exerted on the tting to keep it in place5 . Given the velocity with which the uid
5

f can come from, for e.g., the pipe wall connected to the rest of the pipe line.

11

enters, we want to calculate the force required to keep the tting in place. Let the inlet and outlet areas of the tting be
g and
K respectively. First we have to relate the outlet velocity to the inlet velocity. Using the control volume shown in Figure 3.2, and applying the law of
v2
2

Angle made by outlet to horizontal is

v1
1

/4 f

Figure 3.2: Force required to keep a tting in place. conservation of mass, we get

I I
g gdA
K K

Now we need a momentum balance to compute the various forces on the control volume. The forces acting on the control volume, apart from are the pressure forces. They act normal 6 to the two areas and by convention are compressive when pressure is positive. Momentum balance therefore gives

 A

6 6 g
l g K L M  M M P
g I gK P
K I KK L M  O O

where we have used Cartesian coordinates to express the various vectors on the control surfaces. Notice that we can not solve this equation unless we nd a relation between g and K . O O It is customary to obtain the additional relation by assuming that there are no frictional losses in the tting or by using some correlation to calculate the frictional losses. If we assume that losses are negligible, the energy balance gives,

6IK g 6IK K L g M OP A L K M OP

Therefore, the force needed to keep the tting in place could be worked out. There are several unsatisfactory aspects here. Since we know that Newtons laws contain all the necessary information about motion in general, how do we explain the need to use energy balance? Notice also that square of velocity appears. If there is a velocity prole, should the square of the mean velocity be used? Or should it be mean of the square of velocity? we have to account for it. What is the correction for the prole? It is these kind of questions that make a more fundamental approach necessary.

3.3 Friction losses in pumping


Consider the following typical problem solved in texts on Unit Operations. A piping system is shown in Figure 3.3. It is desired to nd the power required to maintain a velocity in it. Since
6

We will discuss this in greater detail when we deal with tensors

12

2 H 1

Figure 3.3: Piping system.

the area of the tank from which the uid is being pumped (point 1) is large, the velocity there is taken to be zero. The tank is open and the uid is exiting into ambient, hence the pressures at points 1 and 2 are the same. Engineering Bernoulli equation is then applied:

6IK L K MRs A

Rate of work done by pump Head lost due to friction

where the rst term on the right hand side is per unit mass ow rate. The second term on the right hand side is related to the conversion of mechanical energy into heat due to friction. Hence the power required for pumping is given by

I K K H 6 I KK P A Pumping power required


L S M RUM

Head lost due to friction V

Thus if friction losses can be calculated, the power requirements can be estimated. The friction losses in the pipe are calculated using the following formula:

I$KK W A Head lost due to friction X Z  L  where W is the total length of pipe, is its diameter. The friction factor Z depends on the Reynolds number , and it is read from charts. For laminar ow it is given by 6 A Z

Where does this formula come from? In some books it is shown that it can be obtained after deriving the parabolic velocity prole for ducts of circular cross-section. What happens for ducts of other cross-sections? Why do we not have such a formula for turbulent ow? Once again a fundamental approach helps us to answer these questions.

3.4 Torque needed for centrifugal pump


Earlier, we have considered a mass balance around a centrifugal pump. In this section we recapitulate a typical calculation of the power required to maintain a certain pumping rate. 13

When the impeller of the pump is rotated, the uid adjacent to the impeller acquires the velocity of the impeller since the uid sticks to the impeller and acquires the same tangential velocity as the impeller.7 The uid therefore rotates with the impeller and acquires radially directed acceleration. Since no force exists to counter this acceleration, the uid continues to accelerate in the radial direction, i.e., moves out radially8 . The torque exerted by the pump on the uid will be in the T direction, a direction perpendicular to the plane of paper. Let it be denoted by ~ . By applying an angular momentum balance on the control volume shown in Figure 3.1, we get

I g I PL qs M $ k p g g rp s q r g  A L ~ k p K K rp K qsr K 

We discussed earlier the entry condition of uid. If we assume the uid enters radially only, the second term on the left hand side is equal to zero. From the earlier discussion, we have K A qutp K M K . It is more convenient to write

K @m qyt  M qsr K @m qsr  m K A qst

I K o8 I K m@qyt A k  A Kd,8 m L k~p KG I K rqy p K M I Kd,8  . Substituting these result, the torque can be calculated to be equal to I Since we related K to the pumping rate earlier by using mass balance, it can be substituted to obtain ; I a  L P A qy k~p K K r  qup K M L kp K or ; ; a P  A a p K M L k~p K The power required to maintain the ow is given by ~ .
But There are many assumptions made in the above and we raise a few questions. We said that the uid elements adjacent to the impeller wall acquire the velocity of the impeller and transmit it to the other uid elements. What is the mechanism of this? Is it due to the viscous nature of uid? Does it mean that we can not pump inviscid uids using centrifugal pumps? Further, viscous forces decrease the fraction of work done by the pump that goes into increasing the kinetic energy of the uid. How much is the fraction?

3.5 Heat exchanger design


Consider the all too familiar problem of determining the length of a double-pipe heat exchanger required to effect a certain amount of heat exchange. Let quantities with subscript refer to cold uid and those with subscript refer to hot uid. Let be the ow rate. We consider a thin slice of a heat exchanger as shown in Figure 3.4 By mass balance over this section we get

3 3 y A 

We shall later see that this is referred to as the no slip boundary condition. We have given a scenario that ignores vanes. If vanes are present, they push the uid at direction normal to their surface. This also causes pumping action and is discussed by Batchelor in his book. See page 397.
8

14

Hot T h2

Cold
T c1

T c2

z
T h1

Figure 3.4: Double pipe heat exchanger. or 3 is constant. Similarly we can show that is also constant. Balancing thermal energy 9 over only the cold uid we get

3 % E  3 % E  y A  k  T    where we have used the usual symbols. Dividing by T and taking the limit as T   % 3  T A k    3# % E   3 % E  y Dividing by T and taking the limit as T!   % 3  T M 3 % n  3n % n  y A  M ,    % 3  T A

Notice that this is a combination of balance law with rate law where the overall heat transfer coefcient is being used to describe the rate of the process. A thermal energy balance over both the streams put together yields

3 %  K   A 3 %  K   The above can be used to solve for  in terms of  , and the result substituted in the differential equation derived for  . The differential equation can then be integrated to get the famous formula 3# %  K   A k  W E     is log mean temperature difference. It can be used to nd the length of the pipe where  required to heat the cold uid from a temperature of  g to  K . We described this briey to
illustrate the balancing procedure. The temperature used in the above equation is an average value. What kind of average should we use? Is it an arithmetic average? Does it depend upon velocity prole? We need a more fundamental understanding to answer these questions. Consider a shell-and-tube heat exchanger. We use the following formula to relate the heat ; load to the area necessary for the required heat transfer rate:

This can be integrated along with the boundary conditions to get

  E ; A 
`  B  ?
15

We will not reiterate the questions raised earlier about the separate thermal and total energy balances.

where ` is a correction factor. In a shell-and-tube heat exchanger, the ow of the tube-side and shell-side uids is neither counter current nor cocurrent. The factor ` takes this into account. Of course its value is read off charts. How are the charts prepared? In general, how can the effect of any kind of convection be taken into account?

3.6 Correlations for heat transfer coefcients


We raise a different set of questions in this section. Consider convective heat transfer in laminar ow in a tube. Here we imagine that a uid is owing through a hot pipe. Suppose the velocity prole is parabolic when the uid enters the pipe. The following expression can then be used for predicting the heat transfer coefcient, :

H  p V #  j W W _ V  p

where we have used the usual symbols. If the ow is plug-like, the heat transfer coefcient has to be calculated from   H IP

j W

 #

Notice the differences in the effect of velocity and viscosity. Are these correct? If they are, why do these differences arise? Yet another example is the correlation for laminar ow natural convection past a vertical hot plate:

Here the heat transfer coefcient is dependent on the driving force, in contrast to the forced convection heat transfer. This is to be expected since the ow itself is created by the temperature difference. But, why is there a dependence on the 0.25 power of temperature difference, i.e., why is the power not equal to 0.5?

  K  

3.7 Correction for diffusion in a stagnant lm


Consider gas
dissolving from its mixture in gas into a solvent. Suppose is insoluble in solvent. In gas absorber design, the following expression is used for calculating the mass ux s across the gas-liquid interface: !

s all chemical engineers. Notice that in case of heat transfer where the notation is familiar to the driving force, i.e., difference, was used directly to calculate the ux. Here, temperature u  , is being corrected by the factor d to account for the the driving force, fact that gas is stagnant since it can not dissolve in the solvent. Suppose we have a three component mixture instead of binary: gases
c and . Suppose both and are insoluble. Is the above expression valid if we use the sum of partial pressures of and together instead of that of alone?
16

AN5BE u

3.8 Conclusion
We briey reviewed he balancing procedures used in Unit Operations. At the same time we pointed out the kind of questions that need clarication. The objective of a course in Transport Processes is to show the path that leads to answers to such questions. EXERCISES P _ 1. a. A Newtonian incompressible uid (density = , viscosity= ) is draining through a tube ] (diameter = , length= W ) connected to the bottom of a conical tank. The initial level of liquid ] . Find the time required to drain the tank completely assuming (i) in the tank is B pseudo-steady state (ii) all other losses except those in the tube are negligible, and (iii) the ow is laminar. 1. b. In part (a) the rate of fall of height is not constant. Find the axisymmetric shape of a tank that will give constant rate of fall in the level of liquid. This is known as the Egyptian clock.
H Shape ?

Part a

Part b

17

Appendix : Vectors and Scalars 3.A Introduction


We encounter many physical quantities in the present lectures: temperature, concentration, velocity, force, pressure etc. It is necessary to have an understanding of their mathematical character. There are several quantities which have magnitude but no sense of direction in space. Such quantities are called scalars. Temperature, concentration, density are some examples. There are quantities which have a magnitude as well as a direction in space. Such quantities are called vectors. Velocity, and force are examples of such quantities. It helps to imagine a vector to be an arrow. The length of the arrow is its magnitude. Its direction is the direction in space in which the tip of the arrow points. For future discussion, it helps to note that the direction of the vector is independent of the way we look at it. Imagine you have by some trick reduced your self into a point10 and are in a very tiny sphere located in space. If you view the vector from that sphere from any angle, it will look the same to you, i.e., its magnitude and of course its direction will be the same.

3.B Spatial specication of a vector


We need a language to describe vector quantities. The main problem is: How do we describe orientation?. If we imagine a vector to be a pointed line, we can characterize it unambiguously only if we are able to uniquely specify its two end points. The method to achieve this was provided by the famous French mathematician Descartes.11 Suppose we wish to characterize a vector. The starting point of the vector can form a reference point, called the origin. As we are dealing with three-dimensional space, we construct three non coplanar and mutually perpendicular12 straight lines starting from the origin. To honor Descartes such axes are called Cartesian axes. Normally they are referred to as c c T axes. They may also be referred to as = c A 6 c L c respectively.13 We also refer to the planes on which only and (or g and K ) vary, as the (or g K ) plane or as the T (or ) plane. We now take the end point of the vector in space. On the T plane passing through the point, we draw a line to intersect the T axis. We measure the distance from the origin to the point of intersection. This distance is referred to as the T (or ) component of the vector. In a similar manner we can draw lines on plane to intersect the axis to get the component of the vector, and on plane to intersect the axis to get the component. It can be proved from geometry that no other point can
It is called sookshma sarira in Hindu mythology French are not good at spelling. His name is pronounced as Daycart. One obsession with Descartes was to nd an absolutely truthful statement. It is reported that he once locked himself in a stove, and thought and thought about what could be denitely true. He came out and pronounced: I think, therefore I am (or, equivalently, I think, therefore I exist). This led to the philosophy of existentialism. It is no wonder that Descartes, obsessed with correctness, came up with the idea of unambiguously specifying location of points in space. 12 It is not necessary that they be mutually perpendicular. When the three directions are mutually perpendicular, the system chosen is referred to as an orthogonal coordinate system. We also use right handed coordinate systems. This means the direction of and are such that if you rotate the four ngers of the right hand from positive axis to positive axis, the thumb will point in the direction of positive axis. 13 This is a convenience and not an advance. For interesting comments, see: An unnished dialogue with G.I.Taylor G.K.Batchelor, J. Fluid Mech., 70, p 625, 1970.
11 10

$

18

z coordinate of point P
x y plane or z plane

z direction

Unit vector in

y z plane or x plane

Point P Unit vector in y direction y

coordinate of point P x

Unit vector in x direction

Figure 3.5: Cartesian coordinate system have the same components; hence the vector is uniquely specied. Thus three numbers, c and T e.g., (10, 5, -2) refer to some specic vector and another vector will not have the same components unless it is identical to this one. If the vector is a line segment joining point P in space to some chosen origin, the components are also commonly referred to as coordinates of P. The line segment directed from origin to P is referred to as the position vector.

3.C Specication of a vector in terms of Unit vectors


There is a more important representation of vectors that is often used. The three dimensional space in which we live can be characterized by three unit vectors: one for each dimension. The unit vectors as their name implies have unit length. Their direction is given by the tangent to each of the lines obtained by increasing one coordinate at a time. The unit vectors tangent to the Cartesian axes are commonly denoted by i, j, k. Then any vector can be represented by the sum of three vectors, each of which is the unit vector multiplied by the corresponding component: I0 I0 I

A M MS Let us demonstrate the use of the notation. If we denote the three Cartesian unit vectors I = indicial 6 = A by , and the components by , c L c , the same representation can be more compactly
written as Notice that the components of e g are (1, 0, 0).

I A = = =

3.D Coordinate independence of vectors


One question would occur to us. There are two degrees of freedom available in constructing the three mutually perpendicular lines about which we just talked. Suppose we constructed coordinates c c T . Another observer can construct and lines such that they make an angle with the and lines chosen by us. Similarly, it is also possible to have and T axes making 19

an angle with the and T axes chosen by us. Therefore, the components obtained in each of these systems will be different for the same vector. So how do two different observers agree that they are talking about the same vector? This is done as follows. Suppose each observer takes the three unit vectors from the others system. Each can work out the components of the others unit vectors in his system. This information can be used to calculate the components of any vector in the others system from the values of his components since any vector is a linear combination of three unit vectors. If there is a match, the vector being referred to by both is the same. Consider the details of the operation from the view-point of the observer in the unprimed system. The unit vectors of the unprimed observer are e = , while those of the primed observer 6 c L c be the three components of the e= unit vector in are e= . See Figure 3.6. Let = mmm*m A
x 3 x
1

e1 e e 3 e e 2 e 2 3

x
1

Coordinate system of unprimed observer

Coordinate system of primed observer

Figure 3.6: Coordinate systems of primed and unprimed observers and their unit vectors. the unprimed observers system. There will be nine such quantities since both and can take on values 1 to 3. In the more familiar notation, they can be listed as c cnmnmom c . Thus the unprimed observer would express the unit vectors of the primed observer as

= A =

In the more familiar c

A M M $ c and after the primed observer shows Thus the unprimed observer can measure him the unit vector . Similarly he can measure components of the other unit vectors. Now A Z  6 c L c The primed observer would say his components are Z j A A Z
20 we consider the operation of determining whether they are talking about the same vector or not. Take any vector f. The unprimed observer would nd the components of the vector to be 6 L Z c# A c c . Thus . Hence

cT

notation, the unit vector in direction can be expressed as

The unprimed observer can substitute for e in terms of his unit vectors:

A Z
and after rearrangement

A Z Z 3A Z

Thus if they are talking about the same vector, the unprimed observer should nd that

Similar procedure can be followed by the primed observer I = regarding vectors referred to by the unprimedI observer. Thus, any vector with components in the unprimed system must have components in the unprimed system given by

I =A = I A 6 c m n o m  m g

(A3.1)

The relationships between components of a vector in two different systems are referred to as rules of transformation14 . In fact the from of the equations is taken as the denition of a vector. It facilitates generalization to the view point of frames of reference. The two coordinate systems constructed by the two observers may be thought of as two frames of reference with respect to which measurements are being made. Vectors are mathematically dened as those, whose components in different frames of reference follow the above rules of transformation. Viewed this way, a vector becomes independent of the particular coordinate system chosen since the components of a vector in different frames of reference are related to each other in a unique way.

3.E Vectors and laws of nature


Physical laws have to be independent of the coordinate system chosen. Such laws then have to be expressed as relationships between vectors, so that the relationship is independent of the coordinate systems (e.g., Newtons second law relates the force vector to the acceleration vector). Thus it is natural to describe motion in terms of vectors. These ideas can be extended further to formulate laws in different frames of reference. Thus one can relate vectors in one frame of reference to that in another (e.g., velocity vectors in two frames that move with respect to each other) and test whether physical laws (e.g., Newtons laws) are frame independent or specify requirements to make laws frame independent (e.g., material behaviour). The ideas discussed above are presented in a delightfully in Feynman, Vol.I, Cheater 11.
You can not but help but notice the similarity between this and matrix operations in linear algebra. Mathematics has this mysterious unity of concepts arising in different contexts and that is part of its beauty.
14

21

3.F

Vector algebra

The unit vectors of the rectangular Cartesian coordinate system are orthogonal to each other. The orthonormality of the unit vectors is written conveniently using the Kronecker delta is known as the Kronecker delta and takes a value of 1 if A and of zero otherwise. We have already discussed that any vector is a linear combination of three vectorial components and the three unit vectors:

= m A =

A =8= =

Using the delta function and the above representation, the dot product of two vectors can be computed as follows:

m A A A A

E = =  y , m = [=  = = 8=  = = 8=  =

It may be noted that the simultaneous appearance of a repeated index and summation over all indices is typical of the dot product. We can use a special case of the above formula to nd a component of a vector. For this we let one of the vectors be the unit vector corresponding to the component we wish to nd:

8=YA A A A

= m m = = 8=

Another illuminating way of expressing a vector follows from the above representation:

A m A

m 

If one of the vectors is a unit vector, i.e., a vector of unit length, then,

= = 8 = :v 8?

(A3.2) (A3.3) (A3.4)

where v is the angle between the two vectors. Thus, we can interpret m as the component of in the direction. Notice that whenever an index is summed the symbol given to it is irrelevant. Thus

A 8 = =A =
22

When operations between vectors are carried out using indicial notation, it is good practice to use different symbols for summed indices of different vectors. See the confusion caused if this precaution is not taken:

m = A = 8= = m = A = 8= = = A = 8= A yg M K M . It is dened as Another useful symbol is the permutation symbol = = A 6 6 if }$j forms an even permutation of 123 A if }$j forms an odd permutation of 123 A 
if two or more indices are equal

8=A

A permutation is dened as an exchange of the neighbouring indices. Thus 312 needs to undergo two permutations to be brought into 123: rst by interchanging 3 with 1 (to get 132), and then by interchanging 3 with 2 (to get 123). Thus g K is equal to 1. The cross product of two vectors can be dened in terms of the permutation symbol:

h A = = 8= 

Examples Now let us show the ease with which vector identities can be proved with the help of the new indicial notation.   Show that h m A

= 8=   m = A = = 8=  G A = = [=  ,  While carrying summation over i and k we will get terms like  = 8= M = 8= . But by =  A =A = . Thus the net sum over i and k is zero for all permutation rules = A  h m A . values of j. Thus
Consider another example involving unit vectors. Note that a unit vector can be represented

 m A

as

=A = =

Hence the cross product between two unit vectors can be worked out

h ? A A

= = ? ?

(A3.5)

Clearly, the cross product is zero if <A and is equal to where < A is determined by the permutations required to get < j into the order 123. Exercises 23

A
j

and the sign

1. Prove the following identities.   A m m (i)    A m  m  m ~ m  (ii) m 2. Construct a unit normal vector to the plane ABC, where the points A,B and C have coordinates (1,0,0),(0,2,0) and (0,0,0.5) respectively. (Hint: Any vector lying on a plane is tangential to that plane.) 3. In the above problem two vectors, exactly opposite in direction, can t the bill. What can they represent?(Hint: Is area a vector?)

24

Chapter 4 THE CONTINUUM MODEL


Matter as a continuous medium Properties as point quantities Area and volume averages One possible way of solving problems in transport processes is to apply Newtons laws to a collection of molecules. Such a procedure is known as molecular dynamics simulation. To get the ideas clear, consider a single component system. We can write an equation of motion for each molecule provided we know the forces experienced by them. All molecules move under the combined inuence of various external forces, such as gravitational eld. In addition, each molecule exerts an attractive or repulsive force on the others depending upon the distance between them. The forces are attractive when the molecules are far apart. When they come very close, their electron clouds overlap and the forces become repulsive. Lennard Jones potential is one representation of such intermolecular forces between a pair of molecules in non-polar systems. Let x= beF the position of each F molecule in some reference frame. Let F = (r= ) be the molecule on the i molecule, where we have explicitly indicated that force exerted by it depends upon theF distances between the two molecules, p = . Thus we can write Newtons second law for the i molecule:

K = <  x K A

F=

p =  6K

There will be innumerable equations of course since there are molecules in each mole, which translates to that many molecules in just 18 cm of water! Clearly the computational difculties are unsurmountable at this time. Hence we need an alternative model and that is the continuum model1 .
We mention a couple of interesting things here. Suppose we make a change of variable from to . The equations above remain unaltered. Hence the equations can not tell whether the system is moving forward or backward in time. In this sense the system of equations is reversible. Irreversibility has to be introduced into the system through a hypothesis and that is one of the puzzling aspects of this approach. Irreversibility is often introduced through the assumption of molecular chaos, i.e., randomizing the outcomes of collisions between molecules. Semi-empirical methods, referred to as Lattice Boltzmann methods, based on this kind of molecular approach are being increasingly and widely used to solve problems of uid mechanics.
1

 

25

4.1 The continuum model


Continuum model proposes that, in view of the large number of molecules present in a small volume, the discrete nature of matter can be ignored and matter can be treated as continuous. As space is continuous, continuum model implies that matter is also present at every point 2 in space. Matter has properties. Hence, in the continuum model, we can assign values to a property at every point in space. Thus we refer to temperature and velocity etc. at a point. Such quantities are referred to as point quantities or elds. Models where we assign a value to quantities such as temperature, velocity etc at every point in space are called continuum models 3 . We want to make explicit the assumptions behind description of point quantities. Temperature and velocity are terms which refer to the state of matter. Temperature is proportional to the mean kinetic energy per unit mass. Similarly, velocity is the mean momentum per unit mass. In other words they can be related to only material present. A point on the other hand has neither volume nor mass: it is a mathematical idealization. So what does it mean to assign a value of temperature and velocity etc. to a point?

4.1.1 Point quantities and volume averaging


The continuum model proposes that a limiting process can be used to dene point quantities unambiguously. Suppose we are able to determine positions and velocities of all molecules in a region of interest at some time. Now consider any point, P. Consider concentric spheres of decreasing radius p centered around the point P. Since the positions of all molecules are known,
Spheres of decreasing radius centered around P Range of molecular dimensions ~ 1 nm m V r Microscopic range ~ 0.1 m

P r
= r lim 0

Range of macroscopic dimensions


m V

Figure 4.1: Graph showing the limiting process to obtain a volume average. it is possible to determine the total mass, < , of all the molecules contained in each spheres. ^ Since the volume of each of the sphere is known, we can determine the ratio < a in each of P them. Continuum model proposes that the point density can be dened as

= <>= < A (4.1) r a a ^ We can extract this limit by plotting the values of < a against p . A possible graph is shown in ^ Figure 4.1. Clearly, when p approaches the molecular dimensions, the ratio < a will uctuate P A

r 

 

a great deal, and might not even be continuous. It is therefore not shown. Continuum model
2

Points do not occupy any volume and so cannot contain matter. Other concepts, density in this particular instance, have to be introduced to mimic reality. That is why this is termed a model 3 The name arises because they treat matter composed of discrete entities (molecules) as a continuous medium or continuum.

26

hypothesizes that there is a sufciently broad range of size, larger than molecular dimensions ^ and smaller than the macroscopic dimensions, where the ratio < a will be constant, and that can be taken as the limit. The expected behaviour is shown in Figure 4.1 and the dotted line shows the value accepted as the result of the limiting process. It is important to note that the radius is decreased around a selected point, and hence the limit evaluated can be uniquely assigned to that point. By an analogous process we can easily evaluate other point quantities:

 A
and

  k   

k r

I =K ^8L    = <>= = <> = K I r  <a L r  <a

(4.2)

A A

is the molecular weight of the uid, and is the universal gas constant. Thus the where continuum model hypothesizes that, although matter is discrete in nature, unambiguous point quantities varying in space with time can be dened, even though matter is discrete in nature. l  l  l  P Thus we can dene c c T ,  c c T , v c c T etc. Such point quantities that depend on time and location are referred to as elds, e.g., velocity eld. Transport processes deal with these elds and try to predict how these elds behave under a variety of conditions using the known laws of physics.

r   = =<><>= = = r  <a r  <a


(4.3)

4.1.2 Point quantities and area averaging


Each of the above limits can be thought of as volume averaged quantities 4 . For example, in the case of mass, we determined the average mass per unit volume for different sizes of spheres, and density is the limit of this averaged quantity as the size decreased to zero. Another type of limit we come across is the limit obtained by averaging over an area. Suppose we are interested in a quantity that depends upon the area, such as the mass of molecules crossing a surface per unit time. We expect this quantity to depend upon both the area and the surface chosen. Thus if we consider a domain that is cylindrical in cross section and of some length, the mass of molecules crossing a plane will depend upon whether we chose the circular planes or the curved cylindrical surface. Further, it might even change from point to point on the surface. Thus we could think of quantities per unit area. Such limits are is obtained by considering what happens on an area
in the limit of the area going to zero. Consider an area . Since the area is small, it can be considered to be planar. We could ask the question: How much of mass crosses the plane per unit time and per unit area? Only those
Volume averaged quantities are also dened in the context of porous media. Though the process is the same, it is concerned with deriving volume averaged equations.
4

27

molecules having a velocity component normal5 to the plane will cross it (Refer to Figure 4.2). F = = The normal component of the velocity is m where is the velocity vector of molecule

v.n t

Figure 4.2: Limiting process in area averaging. The rings show the successively smaller areas while the direction of the normal vector is kept constant. and is the unit vector normal to the plane. Consider a small time . All the interval molecules will not travel in the direction of for the entire duration of since some of them  will collide with others and change direction. However in the limit of , all of them will = move in the direction of v and cross the plane. Thus, if we draw a box of length m where v is the volume averaged velocity vector, as shown in Figure4.2, all the molecules contained in  that box will cross the plane in time in the limit of . The area averaged mass ux across the plane is given by

! !

Average mass ux

A A A

Thus, in the limit of and


tending to zero, the number of molecules crossing the plane P per unit time per unit area is given by m . Here it is important to note that in the limiting process of area going to zero, the direction of is kept constant. In other words, to obtain the limit, the area is being shrunk in a particular way, i.e., by keeping the direction of its unit normal constant. Note that the above formula can P be broken up as the dot product of two vectors: the ux vector and the normal vector to a plane of differential area. Therefore, once the normal vector is constructed and the point values P of and are known, the mass crossing any differential area per unit time and unit area can P be calculated and is given by m . Later we will see more of this type of decomposition and limiting processes on area.
As in any other language, English has the delightful and some times annoying feature of having several meanings for the same word. It is this that lends color to language, allows poets to invoke many images with the same words, and of course allows jester to pun. Normal here is not being used in its normal sense. Normal here means perpendicular. Is it the normal behaviour of Indians to be normal to each other?
5

= <>= m
'      ! "$# F&%  P
! m !
     "P$# F&% 
m

(4.4)

28

4.2 Derivatives of point quantities


In future we will be interested in the temporal and spatial changes of the point quantities we have just dened. Derivatives are dened as the limit of the ratio of change in the quantity to the corresponding change in the independent variable. Since the point quantities have been unambiguously dened in terms of volume averaging, it is not difcult to dene derivatives with respect to spatial variations. In continuum models, it is assumed that such spatial derivatives exist. Now let us consider the time derivative. As matter is composed of molecules, there is a time scale of dynamics of molecules. In the case of gases, it could be the mean time interval between collisions. Therefore, there is an intrinsic time scale in which quantities could be changing abruptly. It is thus possible that mathematical difculties exist in determining time derivatives by actually taking a limit as time interval between observations goes to zero. Recall that point quantities were dened as averages over nite volume, that the volume over which they are being averaged is expected to be large enough to contain large number of molecules, and that the limiting process has actually not been taken all the way to zero volume. In considering the limiting process here as well, the continuum model hypothesizes that, even though the limit of the ratio of change in any point quantity to the time interval in which the change occurred may not exist as the time interval actually goes to zero, a range of sufciently small but nite time intervals exist over which the ratio remains constant. The constant value obtained is taken to be the time derivative.

29

Chapter 5 THE TRANSPORT THEOREM


We consider the differences in treating fields in solids and fluids. We then discuss the Reynolds Transport Theorem which allows us to deal with these differences. Material on kinematics is presented in the Appendix. Though advanced in nature, it will clarify some concepts. But today he only saw one of the rivers secrets.... He saw that the water continually owed and owed and yet it was always there; it was always the same and yet every moment it was new. From Siddhartha by Hesse Many of the questions we encountered in the previous chapters were concerned with the following basic difculty: the laws of physics are stated in terms of closed systems but we need to apply these to to open systems, i.e., equipment through which uids ow in and out. This difculty is resolved and presented in a very elegant way in the article by Truesdell and Toupin in Handbuch der Physik volume 3/1. Here we give an intuitive treatment to provide a physical picture. The basic difculty really has two parts. The rst is to nd a way of dening a system in a continuum where, unlike in solids, no obvious boundary exists demarcating the set of objects of interest from the rest of the world. The second is to devise a way to relate the results obtained by applying physical laws to a system to the performance of equipment of interest. The rst part is discussed in this chapter, and the second part is taken up in the next one.

5.1 Fluid domains: Imagined


The subject matter of transport processes is concerned mainly with uids. 1 Unlike solids, uids (e.g., water and air) do not possess distinct and identiable shapes. When a bullet travels in air, it is clear where its center of mass is since its boundaries are easily identied. This is true
1

Transport of heat and mass in solids is also of interest.

30

of the entire bullet as well as of any arbitrary part of the bullet that we choose. The boundary of the part we chose also remains identiable even when the bullet is in motion. In contrast, uid or a part of it, does not present such a clear boundary. Look at a river. Except for oating solid bodies, it is difcult to clearly demarcate a part of the uid and to follow it. Suppose, through some device, we managed at some time to mark all the matter in a domain in a river and make it visible. Then the domain and its contents become identiable, and their course at subsequent times can be followed. If we managed to do this, we would nd that, unlike in solids, the shape of the marked domain can change as it2 ows. This is because the surface of the domain has the same velocity as of the uid on the surface and hence the shape of the domain will constantly change. The surface of the domain constitutes a moving boundary as shown in Figure 5.1.
v

A travelling bullet. The rectangle with the dotted boundary is the system. Its shape and volume do not change while the bullet moves.

v v v v v v v v v v v v

v v v

v v

A marked and travelling fluid packet and a sub-part of it. Notice both change shape and the changes can be different. The velocities at different points are different.

Figure 5.1: Closed systems in solids and uids.

5.2 A system in a continuum of uid


Consider an imaginary domain marked in a uid. Now, an object can cross a moving boundary only if it has a relative velocity with respect to the boundary. Hence there is one important property of the domain so marked. As its surface has the same velocity as the uid there, no uid can cross its boundary. Thus this domain, which we have marked marked in imagination, constitutes a closed system3 . Clearly we can apply all the laws of physics to the contents of
In making this statement, the small changes that occur in the shape of a solid due to elasticity are being ignored as they usually are negligible. 3 For simplicity we will some times refer to such a marked domain as a system.
2

31

this imaginary domain, whose boundaries are moving with the velocity of the uid. There are other important differences between a closed system of a uid and a part of a moving solid. All the uid inside a system may not travel with the same velocity but the solid moves like a rigid body. The density of a uid may vary from point to point in the system while such is not the case, to a very good approximation at least, for solids. Hence while dealing with uid systems we must account for variation of properties from point to point in the systems.

5.3 The Reynolds Transport Theorem


Application of the laws of physics to a closed system often needs calculation a property of the entire system. For example, to apply the law of conservation of mass, we need to calculate the total mass contained in the system. For a solid, we easily calculate it since its volume remains constant. It might not be so for a closed system of uid since density in it can change from point to point, and its volume and shape can also change. Thus we need to know how to compute the properties of a system whose boundary is moving. Almost always, these properties are obtained by integrating a quantity, dened per unit volume, over the volume of the system, e.g., integration, over the volume occupied by the system. Thus integration of density gives the mass of the system, of the product of density and velocity gives the momentum of the system dened per unit volume. Let it be a function of both time and spatial etc. Let be any entity l  A location. Thus c . Then the total quantity of that entity in the system, , is given by

( ( *)

where a is the volume occupied by the system at time . Many physical laws of interest in irreversible processes are about the rate of change of some property. For example, the law of conservation of mass states that the rate of change of mass in a closed system is zero. Similarly, Newtons second law states that the rate of change of momentum of a closed system is equal to the sum of forces acting on it. Thus, to apply these laws, we need to know how to evaluate the derivative of a quantity, such as , with respect to time. In the instances of interest to us, is an integral over a domain whose boundaries are changing with time. Mathematically it is equivalent to differentiating an integral whose limits are a function of time. All of us are familiar with the Leibniz theorem which tells us how to differentiate an integral with respect to a variable when its limits are a function of the variable:

0/ * l

+ * l A ., - " F&% ( *) c ll a

This is a theorem for an integral in one dimension, if we consider to be a space dimension. Let us examine what the various terms on the right signify. One is familiar with integration over immobile domains and there only the rst term of the l above equation would appear. The rst c term therefore represents change in time since changes with time inside the domain.  ^  ^ We can interpret Z and R as the velocities of the upper and lower boundaries of the system. The last two terms represent corrections to the rst term since the system is occupying or vacating locations in physical space where properties can be different.

21 " F&% & % " 2 1  c l   * l l  Z 5 l l  R l l   A , ,  c 3 M 3 Z c  3 R c  " F&%3 " F&%.4 4 F

(5.1)

32

For our purposes we need a generalization of the Leibniz theorem in three dimensions. This is referred to as the Reynolds Transport Theorem. It is given by

where

is the outward normal to the system domain. The outward normal to a volume is

5) c l  , 5 ) l  *) c l
( , A ,  6- ( c a  m 7 ( M .- " F&% 4 - " F&% " F&% 4 
n v n v

(5.2)

v n

Vs (t) A s (t) is the total area of the system

Figure 5.2: The Reynolds Transport Theorem. dened as the normal that points away from the volume. See Figure 5.2. One important aspect of the above theorem should be noted. The theorem relates the rate of change of any extensive property of the entire system to the point quantities both in the domain of the volume currently occupied by the system, and on the surface of that volume.

5.4 Conclusions
Our discussion is not yet complete. We wanted to calculate the change in properties of an open system. Hence one more step is needed to relate the properties of a system to those of an open system. We consider this in the next chapter.

33

Appendix : Kinematics
Kinematics is the subject concerned with the description of motion. Here we assume that space is a continuum and time ows continuously. The objective of kinematics is then to describe the history of motion of the various material particles occupying space. Let us identify  6 L B B A = a set of particles at time by identifying their locations c c c mmm . Thus the value of = uniquely identies one particle since only one particle can occupy a point in space at any given time. Some times, = are referred to as lmaterial coordinates. If we follow the path of  = these particles, we can note the locations c of each of these particles as a function of time. Thus we have a description of the locations of all the particles at all times or a description of the entire motion. Nominally we can write that

58

*)

Note that and are independent coordinates. If the above functional relationship is available, we can specify the location of a particle at all times, once = of that particle is specied. The entire process described above can be thought of as mapping from = space onto space. Each particle is allowed lto occupy only one position. This implies that the above mapping is unique, i.e., only one = c exists for a given = and vice versa. Therefore, we can write

) =A ) = 58 = c l 8 = A 8 = *) = c l

*8

In other words the mapping can be inverted.

5.A Velocity
From the above functional relationship, other interesting kinematical quantities can be determined. For example, we can dene derivatives of properties of the particle as we follow it. When we follow a particle through the space, the locations occupied by it have to be such that = is constant. The derivative obtained by following a particle can then be dened as:

H   V =

 8

For example velocity of a particle is the rate of change of its position as we follow it:

] = H  = = A ]  V =

)  8

EXERCISES

p 1. Consider . At some instant B let all the particles  steady ow in a tube, pcvUc on 8 A = A plane be marked. Thus let the locations of these particles be given by T
a. What are the locations x (specify pcv ) occupied by these particles at a later time t? b . Show graphically the paths c. What the shape taken by the surface containing all the marked particles? d. Using the denitions given, What is the velocity vector given by ? 34

 

2. Let pcv represent the coordinates of a set of material particles at any time B . The material particles occupy positions given by

e. What is the acceleration vector given by ?

F F A K F F A g

9 " ;: =<>%

9 " ;: =<?%

a. Plot the trajectories of the particles with the material coordinates (1,0), (0,1), (1,1) and (-1,1). b. Find the velocity eld c. Is the ow steady from the view point of an observer stationary in the laboratory frame? d. Is the acceleration of particles zero? What about the net force on the particles? e. Do you see any contradictions to your answers to parts c and d? f. What ow do you think the ow described above might represent?

35

Chapter 6 PHYSICAL LAWS & FLUID SYSTEMS


We discuss the application of the Reynolds Transport Theorem to open systems. We derive the basic equations of shell balances. We begin our discussion with a familiar example of how laws of nature are applied. The dynamics of rigid bodies is well developed and familiar to all of us. Suppose we want to calculate the path of a ball. It can be calculated if the velocity of the ball is known since by denition  where and are the position and the velocity of the ball. The momentum of the ball is equal to the mass of the ball times the velocity, v. We apply Newtons second law to calculate the velocity. 

) 

 Momentum of the ball A




Forces acting on it

(6.1)

which can be used once the forces are identied. We would like to derive similar equations for uids also which can be used to calculate the velocity in a uid as a function of forces acting, and also the path a uid particle follows as it ows into and out of an equipment. The Reynolds Transport Theorem tells us how to calculate the rate of change with time of some quantity for a system composed of a uid. Thus, when we consider the dynamics of uids, we want to calculate a term similar to the one on the left hand side of eq.6.1. But before that, we would like to show that the conventional momentum balance used in books on Unit Operations follows from the results obtained by applying the transport theorem. We want to do this in a way that is applicable to any arbitrarily chosen quantity, not just momentum, and show that the intuitive generalizations of balance laws used in Unit Operations indeed arise from an application of transport theorem to a system of uid.

6.1 Control volume and system


Consider a volume in space of interest to us. We restrict ourselves to the cases when this volume is stationary, i.e., to a stationary control volume, referred to as CV for short. It could be part of an equipment through which a uid is owing. Thus, even though the control volume is stationary, uid could be moving through it. At a given instant of time, this control volume is lled with a packet of uid. This packet comes from somewhere and goes somewhere else 36

after some time. At the instant of time being considered, however, the boundaries of the control volume and this uid packet coincide (see Figure 6.1).
At an instant of time = t t The control volume is shown by a circle. The system is the shaded object. It is occupying part of the control volume At an instant of time = t The system now occupies the entire control volume

At an instant of time = t + t The system has left part of the control volume

The sysetm has left the control volume

Figure 6.1: System moving in and out of control volume of interest.

5) c l   , *) l 5 ) l  ( A , ,   - ( c  - " F&%@4 a M - " F&% m7( c


" F&% 5) 4 l A , ( c  a M , m7( *) c ll
 4 4 where a and
are the volume and the area of the control volume. Since the control volume is stationary its boundaries are also stationary, the differential with time can be taken outside the integral. Hence the rst term can be rewritten as ,  ( *) c l   a A  ,  ( *) c ll a 4 4 Thus we are able to relate the rate of change of total quantity + of the system at any instant to the distribution of ( in the control volume and its boundaries at that instant. The equation obtained by substituting this relationship into the transport theorem can be referred to as the 
transport theorem for stationary control volumes:

Thus, if we choose that particular packet of uid as the system, at the instant of time when the system occupies the control volume, we have

 * ) l  A , A .- ( c  ,  ( 5) c ll a M , m7( *) c ll


" F&% 
37

(6.2)

6.2 The law of conservation of mass


Now we are ready to show how to apply the laws of nature to control volumes. We take as an example the law of conservation of mass. The total mass of a system is given by Total mass of a system

, - P 5) c ll a " F&%


A   A 

According to the law of conservation of mass, the rate of change of mass of a system is zero. Hence we have Rate of change of mass of a system

or

 A  P c l  F

By applying eq. 6.2 to replace the right hand side, we have

 A   P c l a m P c ll
M 

A , *)

 ., - " &% 5) ,

 ., - " F&% P 5) c l a

*)

Consider the application of this to the ow of a uid in a tube of circular cross-section (see Figure 6.2). The rst term is obviously the rate of change of mass in the control volume i.e.,
r

- z

z=0

z=L

Control volume is shown by dashed lines. Outward unit normal vectors are shown. The flow is from left to right.

Figure 6.2: Mass balance on a control volume for ow in a pipe. the rate of accumulation of mass. As the pipe wall is non-porous, the surface integral vanishes  and on the cylindrical surface. Only the surface integrals on the circular faces at T A T A W are not zero. These surface integrals give the volumetric ow rates. The word inlet implies that uid enters the CV. Hence, there the velocity and the outward normal to the control surface are in opposite directions. Similarly, the outward normal and velocity will be parallel to each other at outlet. Thus the sum of the two area integrals is equal to the ow rate through the outlet or output ow rate minus the ow rate into the input or input ow rate. Hence, the above equation reduces to the familiar rule we have always applied: Rate of output of mass Rate of accumulation of = Rate of input of mass to mass in the CV the CV from the CV 38

6.3 Newtons second law or Linear momentum balance


Newtons second law states that the rate of change of momentum of a system is equal to the sum of forces acting on it. Momentum is a vector. Thus, considering the component of the momentum, Newtons second law can be written as Sum of x component of forces acting on the system

We now let i.e., eq. 6.2

 P c l I c ll a A   F P I , and apply transport theorem, be the momentum per unit volume,i.e., A A I m P c l c ll

component of the sum of forces acting on the system

 ., - " &% *) *) (  0 I 5 ) l  5 ) l   P , , A  c c a M

Rate of change of x momentum of a system

*) *)

Sum of Rate of Rate of output Rate of input of accumulation of component of = momentum to of momentum + momentum in forces acting on the CV from the CV the CV the CV It is this form that is widely used in books, in: Transport Phenomena by Bird, Stewart and Lightfoot, for example. The above equation can be interpreted as an equation for balance of linear momentum if we imagine that forces generate linear momentum just as chemical reactions generate mass of products while of course consuming reactants.

We can interpret these terms as we did those in the law of conservation of mass. The rst term on the right hand side represents the rate of accumulation of momentum. The second term represents the rate at which momentum leaves the control volume minus the rate at which momentum enters the system. Further, at the instant under consideration, the contents of the system and the control volume are identical and hence the forces acting on the contents of the system and CV are the same. Thus

6.4 The law of conservation of angular momentum


The law states that the rate of change of angular momentum of a system is equal to the torque acting on the system, and is the analog of Newtons second law for rotating systems. Select an origin and let be the position vector from the origin of any point in the control volume. Then is the angular momentum at the point specied by the position vector. Hence for a CV we can write

where is the unit vector in the direction. This is the equivalent of Newtons second law.

component of the sum torques acting on the CV

 l ~  l l    P P c l~ c l m 
A c c m a  m of  M 

 , B B

*) B

Exercise 39

The law of conservation of energy can be stated as follows: the rate of change of the sum of internal and kinetic energies of a system is equal to the rate at which work is done on the system minus the rate at which heat is supplied to the system. Derive a form for the conservation of energy that can be applied to a CV. You can use the symbol (you guessed it right!) for the heat ux vector. You do not know enough to calculate the rate of work done on the system. State it in words.

40

Chapter 7 WHAT IS A FLUID?


The differences between fluids and solids are discussed. Application of Newtons law needs specification of forces acting on fluid elements. Classification and specification of forces on fluid elements is considered. Specification of forces requires new type of quantities: tensors. Let us briey recall from the previous chapter6 the result of application of Newtons second law: Rate of Sum of Rate of input of Rate of output accumulation of = component of momentum to of momentum + forces momentum in acting on the CV from the CV the CV the CV If we know the velocity vectors inside the control volume and at its surfaces, all but the last term in the above equation can be calculated. Thus in order to apply the linear momentum balance, we have to learn to specify forces acting on a uid element. This is the subject of the present chapter. We are interested in uids and so it is better to begin with a denition of a uid.

7.1 What is a uid?


All of us are qualitatively familiar with uids and solids, and the differences between their behaviour. These differences are used as the basis to dene a uid. They also serve to understand the differences in the methods employed to describe response of uids and solids to applied forces.

7.1.1 Behaviour of solids


Consider a cube of solid. It retains its shape, and does not deform unless a force is applied. We could apply forces perpendicular, commonly referred to as normal forces, to any of the six faces. It will result in either compressive or extensional action on the solid. This is shown in Figure 7.1, where all faces except abfe and cdhg are under compression. This kind of force makes the solid expand perpendicular to the stress-free faces i.e.,abfe and cdhg, and contract along the other directions. 41

b g

Figure 7.1: Compressive stresses on a cube of solid.

We could also have forces acting tangentially on one or more of the six faces. These are referred to as shear forces and they tend to shear the solid.1 The forces applied on the surface deform or strain the surface elements, and the strained elements, in turn, transmit stresses to the elements adjacent to them. Thus the action of the forces applied on the surface spreads to the entire solid. As a result, the whole solid deforms. If the magnitude of the forces is not very large, the solid takes on a new shape in equilibrium with the forces applied. It retains this shape as long as the forces are applied, and when the forces are removed, the body goes back to the shape it had when no forces were applied. This is of course how an elastic solid behaves. Hookes law describes the relationship between the applied stresses and the resultant strains. An example of this is shown in Figure 7.2. Consider a solid placed between two
a b

Figure 7.2: Deformation of an elastic solid. concentric cylinders and bonded to the cylindrical surfaces as shown in part a of the gure. Four lines are drawn on the material. Now attach a weight to a string and wrap it around the outer cylinder. Let the weight hang down while holding the inner cylinder so that it cannot move. The downward action of the weight exerts a shear stress on the solid. The solid rotates clockwise and takes a nal equilibrium position as shown in part b of the gure. The lines are now elongated as the material is under strain. Notice that the lines are deformed not just at the outer surface, but throughout the material. In this conguration, the entire solid experiences
1

In principle, the forces could change in magnitude as well as direction from point to point on the surfaces.

42

a shear force. If the weight is removed the solid would recover from the strain and would go back to the state shown in part a of the gure.

7.1.2 Behaviour of uids


Now consider the behaviour of a uid in contrast to that of a solid. Firstly, a uid cannot hold a shape by itself. It has to be placed in a bounding container. The uid takes the shape of the container in which it is placed. This is the essential difference between a solid and a uid. Once placed in a container, the uid can be compressed just like a solid but only by applying normal forces to the uid. The volume of the uid changes and a new equilibrium is established. The uid experiences a uniform pressure in all directions or an isotropic normal force. The specic volume of the uid under such a state can be calculated by an equation of thermodynamic state. When the forces are removed, the uid regains its old volume. Under the inuence of isotropic normal forces, the liquid behaves as a solid. Now consider how a uid behaves under shear. Fill the space between two concentric cylinders with a uid, and conduct an experiment similar to the one on the solid described above. It is our experience that the weight continues to go down and never comes to halt, unlike with a solid. This fact is used to dene a uid as a material which deforms continuously under the action of a shear force. The uid continues to deform or ows in response to continued application of shear forces. When we remove the weight, the uid never fully goes back to its original position. The exact behaviour depends upon whether the uid is purely viscous, viscoelastic, or a more complex uid. If the uid is purely viscous, it will not recover any of the deformation. When a liquid is poured into a container, it experiences shear forces till it conforms to the shape of the container, and then only normal forces are left. At that juncture it reaches mechanical equilibrium, and all motion dies.

7.2 Forces on uid elements


In chapter6, we showed how the motion of uids can be understood by applying Newtons laws in the same way the motion of solids could be understood. As was mentioned in the beginning of this chapter, to execute this plan, we need the ability to describe the forces on every point in any arbitrary volume of uid, and also on its surface. The forces arise due to the action of gravity on molecules, intermolecular forces between molecules etc. Thus, the methodology to describe forces in a continuum has to be evolved. The boundary of the system evolves and moves in a smooth or continuous manner. Hence, the force elds themselves will have to be continuous. The force elds then must conform to the limiting processes described in chapter on the Continuum Model. Science is knowledge that can be veried by observation2 and codied into as small a set of rules as is possible. An approximate description of the scientic method is as follows. A basis for understanding an observed phenomenon is proposed. This is referred to as a hypothesis. Values of quantities that can be observed in an experiment are predicted based on the hypothesis. If the predictions agree with the observations, the hypothesis survives, till it meets
There is no Truth in science. There is only truth which keeps changing, it is hoped convergently, depending on observations! The observations are used to dene categories and abstracted into rules followed by categories. The rules are usually referred to as laws.
2

43

a Nemesis. In uid mechanics, it is hypothesized that forces acting on any arbitrary part of a continuum can be broadly categorized into two classes. One class is referred to as body forces and the other is referred to as surface forces or contact forces. Body forces act on every part of matter, however small its volume. The total force acting on a uid element due to body forces is obviously proportional to the volume of the element. An example of this is the gravitational force exerted by earth on uid elements. As the distance between the centers of mass of the earth and the uid element changes, the gravitational force varies. For all practical purposes, however, the distance is nearly equal to the radius of earth, and hence can be considered as constant. Thus, the force exerted by the earth on uid elements is proportional to the mass or volume of the uid element. Therefore the gravitational force is a body force. Another example is the electrical force acting on ions in a uid. Here the electrical eld and the magnitude of the charge in a uid element determine the force. The charge in a uid element is given by the product of volume and charge density, and hence the electrical forces can be thought of as body forces. The volume averaging that was discussed in chapter 4 can be used to dene the body forces as continuous quantities and can be used in the continuum description of uid motion. Matter is held together due to attractive forces between molecules. A force has to be exerted if we take any body and wish to move one part of it relative to other parts. Thus every part of a body exerts an attractive force on the other parts. The attractive force exerted by a molecule decreases very rapidly with distance. For this reason, as a limiting process, these forces can be considered to act on a surface, a mathematical construct of zero thickness, that separates any two parts of the body 3 . In other words, if an arbitrary surface is used to divide any uid body into two parts, it is a good approximation to think that each part exerts a force on that surface. These forces are referred to as surface forces or contact forces and are obviously proportional to the surface area that divides a uid body into two parts. It is therefore better to characterize them by dening a force per unit area, similar to the way we characterized body forces per unit volume. The surface force per unit area is called stress vector. From our earlier discussion on area averaging in chapter 4, the stress vector can be expected to depend upon the direction of the plane chosen for dening the point quantity. The stress vector varies spatially too. Thus it is a more complex quantity and more difcult to specify than the body forces. The stress vector is especially fundamental to our discussion, and consequently the next chapter to is devoted to a discussion of how it is calculated. It is related to a new type of quantities: tensors

When the thickness of a uid layer approaches about 100 nanometers, this approximation breaks down. Thin uid lms on solids have many technological applications, and uid mechanics of such lms has to account for the violation of this assumption. Much literature is available on the methods used to handle this.

44

Chapter 8 STRESS TENSOR


We consider quantities that have different directions at the same point, for example, pressure. We discuss the stress vector defined in the previous chapter. The stress tensor is then defined. Area averaging was discussed in chapter 4 and it was shown there that area averaged quantities depend upon the direction of the plane over which area averaging is done. If this logic is extended, if the quantity to be averaged is a vector, the averaged quantity should have a sense of two directions: one each for the plane and the vector. Are such quantities mere mathematical curiosities or are they encountered in uid mechanics? Consider an example that shows that area averaged quantities are in fact very commonly encountered.

8.1 Direction of pressure


Everyone is familiar with pressure. Pressure is normally treated as a scalar though it is force per unit area. Several things are implicit in this usage. Pressure is stated to be an isotropic and normal force per unit area. This means that it acts in a direction normal to any plane in a uid. This is shown in Figure 8.1. Thus the force due to pressure a point can have several

Enlarge Point P

Direction of the pressure force is normal to any plane chosen

Figure 8.1: Force due to isotropic pressure. directions. The stress vector discussed in chapter 7 is similar in character: it can have different 45

directions depending on the direction of the plane selected to pass through a point. Since any point has an innite number of planes of different directions passing through it, an innite number of stress vectors exist at any point. Cauchy proved an elegant theorem to show that all these innite quantities can be described using only nine quantities collectively referred to as the stress tensor.

8.2 Stress tensor


The mathematical nature of the stress tensor can be intuitively grasped as follows. A vector is a quantity that has both magnitude and direction. At any given point, in principle, a vector can take on an innite number of directions. Thus, there exist innite directions for the plane passing through a point where the stress vector is to be specied. In addition, the direction of the force in itself can assume an innitely large number of directions. The stress vector is thus an entity which contains the doubly innite possibilities. However, all the innite directions of a vector can be described by representing any vector as a linear combination of the three unit vectors1 . Just as three unit vectors are sufcient to describe the innite possibilities of the direction of a vector, the nine (3x3) quantities of a tensor have the potential to describe the doubly innite possibilities of the stress vector. The tensor represented by the collection of nine quantities has two kinds of direction stored in it. In the case of the stress tensor they are the direction of the plane and the direction of the force acting on the plane. Thus it is common to refer to a scalar as a tensor of zeroth order, a vector as a tensor of rst order, and the stress tensor as a tensor of second order. The concept can be generalized to have tensors of higher order, although such generalization is not needed at the level of the present course.

8.2.1 Area as a vector


Intuitively, it is clear that a plane can have two directions: the directions into which its top and bottom surfaces point. Therefore a planar area can be thought of as a vector. Since the area is not a directed line segment, the way to specify its direction is not obvious. The direction of a planar area is dened by adopting the following convention. Construct a unit normal vector to the area pointing away from it, and take the direction of the vector to be the direction of the area. This is shown in part a of Figure 8.2, where normals are constructed to the area EFGH in two directions. According to the convention adopted, n refers to the top surface while n to the bottom surface. Thus any planar area has a direction given by the unit normal vector pointing away from the surface, and the normal itself can be described by the three unit vectors. Hence, in Cartesian coordinates, the area of any plane has c c and T components. This shown in part b of Figure 8.2. Consider the top side of the plane ABC, i.e., that side which points away from the origin O. Then, area of the plane OBC is its component since OBC is the projection of ABC onto the plane perpendicular to the axis. Similarly, the and T components are given by areas OAC and OAB respectively.
1

The number three arises is because our world is three-dimensional.

46

z C H E -n A x n F G o B y

Figure 8.2: (a) The top surface has n direction while the bottom surface has -n direction. (b) The components of the area ABC are considered in the text .

8.2.2 Tensor and directions


As mentioned earlier, the stress vector can be thought of as having two parts to its directionality: one is the direction of the plane on which it is acting and the other is the direction of the force itself. Both these can be represented by three unit vectors each. The plane has three components. On each of the component planes, the force has three components. Thus it seems reasonable that the stress vector can be described by assigning nine components in all. These nine components collectively are referred to as the stress tensor. If the stress tensor is 2 represented by , the nine components can be listed in the Cartesian coordinate system as c cnmomnmoc c . The rst subscript represents the direction of the area and the second describing the direction of the force. Just as with vectors, a tensor also can be expressed as a linear combination as follows:

E E

E DE

D A E M E M

momnm M 5M

8.3 Cauchys theorem


As mentioned earlier, the stress vector on any plane can be calculated given the stress tensor and the direction of the plane. The Cauchys theorem shows how to do this. Here, the theorem is stated without proof. Let be the normal vector to a plane. The stress vector on that plane, denoted by , is given by

D E M E M E  EE l M EE M EE  (8.1) M M The components of the stress tensor are point quantities; hence, if they are known, the forces on any small area can be found. Note that . D instead of D . is used since the rst subscript of the stress tensor is assigned to the area.
A A m M M
As far as possible, we try to use the following notation: bold upper-case letters or Greek letters for tensors, bold lower-case letters for vectors, and the corresponding letters in regular font for components.
2

FHG

FIG

47

Note that the component of the stress vector is given by M M ; the other components are similarly obtained from the above expression. The components of the stress tensor can be given a physical interpretation. Suppose we consider an small innitesimally A  T A plane, i.e., one to which the vector is normal. Then for that plane . Thus is the component of the stress vector acting on T plane. By similar reasoning and are the and T components of the stress vector acting on the T plane.

EE

8.3.1 Surface forces on a system


We mentioned in chapter 6 that in order to apply Newtons second law, it is necessary to calculate the forces acting on a uid system or, equivalently on a control volume. In chapter 7, the forces acting on a system have been classied as body forces and surface forces. Here, calculation of surface forces exerted on any arbitrary differential area of a uid system or a CV by the surroundings is of interest to us. But the surface force is given by the stress vector multiplied by the differential area. Hence, Cauchys theorem suggests that the surface forces can be calculated if the stress tensor at every point on the surface is known. Thus, in order to calculate the stress vector, a unit normal vector is constructed pointing away from the system, and into the surroundings. The stress vector is computed using the Cauchys theorem, i.e., n. . A convention needs to be adopted to specify whether it is the surroundings or the contents of the control volume that exert the force so calculated. In this notes, the following convention is adopted: n. is the force exerted by the surroundings on the surface of the system or, equivalently on the surface of the CV, per unit area. Suppose that a different convention is adopted, i.e., n. is the force, per unit area, exerted by the system on the surface. Since the physical result has to be identical, the components of the stress tensor in such a convention will be of the opposite sign to that obtained by using the convention adopted in this notes. Some books do use such a convention, notable amongst them being the text by Bird, Stewart and Lightfoot.

8.4 Pressure tensor


Before we leave this chapter, consider again the isotropic nature of pressure. Suppose it is desired to construct a stress tensor that gives rise to a stress vector that is normal to any plane constructed through a point. Such a stress tensor will always give rise to a stress vector that will be like pressure3 , and hence must satisfy

m A O
We know such a tensor: the identity tensor I. Thus pressure corresponds to the stress tensor -pI. This will be used often in future.

JD

The following convention is being used: pressure is compressive when its numerical value is positive.

48

Appendix: Tensor algebra 8.A Tensors as dyadic products


The idea of tensors arises quite naturally, mathematically speaking, if a new product of two vectors, which is neither a dot product nor a cross product, is considered. The expression vw has not been assigned any meaning in our discussion on vectors so far. Such a product is referred to as a dyadic product of vectors, and is dened as

= =

=I = I = =

(A8.2) (A8.3)

The order of the two vectors that constitute the dyadic product is important. The product of two unit vectors is referred to as a unit dyad. Note that the unit dyads = and = are different, and this is why the order in which the product is taken, i.e., vw or wv, is important. Thus there are nine quantities = , and each of these is multiplied by F the corresponding component of the dyadic product. It should also be noted that the component is in general different F from the [ component. The unit dyads play the same role in tensor analysis that the unit vectors played in vector analysis. The notion of dyads can be generalized to tensors because all that matters is the entity which has nine unit dyads multiplied by the corresponding components. Such an entity is called a tensor of second order, and we will shortly explain why its order is two. In general, therefore, a second-order tensor has nine components and we can write it as

= = 

A second-order tensor can also be displayed as an array:

associate it only with its components. Unfortunately, unlike in the case of vectors, there is no easy way of visualizing it. The dyadic product of two vectors has been dened as a second-order tensor. The stress tensor, which is of second order as you now see, has two senses of direction. This also meshes with the idea that it is a dyadic product of two vectors. This notion of number of directions can be generalized. A vector can be thought of as a tensor of rst order since it has only one direction. A scalar as a tensor of zeroth order since it has no direction at all. Clearly we can have tensors of order greater than two.

gg  gK  g  #  K g  K#K  K  g  K  # Once again it is better to think of the a tensor as an entity by itself and not necessarily

MN

OP

8.B Tensor algebra


The algebra and calculus of tensors is analogous to of vectors of course after remembering the order of the unit vectors that compose the unit dyads. Indicial notation once again facilitates 49

the evaluation of various operations. Thus

L m

A A A

. The dot product of a vector and a tensor resulted in a vector. In general the dot product gives rise to a tensor whose order is less than the sum of the orders of the two entities involved in the product by two. Thus a dot product of two second-order tensors results in a second-order tensor. A useful quantity is the unit tensor, which is dened as

L L A Note that in general m

I  m = = = I =  = =

= =   = I

= =

A dot product between a tensor and the unit tensor returns the original tensor. This is easily veried. EXERCISES (i) ~m A m (ii) : = .( . )

R
R

Prove the following  identities

50

Chapter 9 SURFACE FORCES & CONSTITUTIVE RELATIONSHIPS


We discuss the idea of constitutive relationships. We then consider the example of fluid reponse to stress by using solid mechanical behaviour as the basis. The Newton-Stokes law which proposes that stress is linearly proportional to strain rate is introduced. Viscoisty of fluid is the proportionality constant. Stress in fluid is interpreted as momentum flux, and we show that an expression can be derived for viscosity from kinetic theory. The linear momentum balance equation contains surface forces and velocity. Thus, it can not be solved unless a relationship is found between the surface forces and either the velocity or some other quantities related to velocity. The objective of this chapter is to develop this relationship. Apart from the mathematical difculty that forces a search for the relationship, there are physical grounds to expect such a relationship to exist. These physical ideas are the central theme of this chapter. When a uid ows stresses are created in it; or equivalently, we have to apply stresses to make it ow. Hence we expect that the stresses created would depend upon the characteristics of ow, e.g., its velocity. For generating the same velocity in a given apparatus, we also expect that some uids would need application of a smaller force as opposed to others. In common parlance, they are referred to as less or more viscous uids. The relationship between the characteristics of ow and the stresses is known as a constitutive relationship 1 . The science of formulating constitutive relationships relevant to the ow of uids is known as Rheology 2

9.1 Strain rate


The stresses developed in a solid are related to the strain imposed on it, and Hookes law is an example of the relationship between the two. However, a uid is a material which deforms continuously in response to an applied stress. Thus, it is not the deformation, but the rate of
The relationship is expected to depend upon the nature or the constitution of the uid. Hence the name. Greek philosopher Heraclitus proposed that change is the only permanent thing and said that All things are . The word , owing. He wrote (English of course was not yet born during his time) in Greek: pronounced as Rheo, is the root-word of name of the science that describes ow: Rheology.
2 1

SUT VIWXTZY\[^]

Y\[_]

51

deformation, that determines the stresses created in response to ow. Hence we have to dene a quantity that is characteristic of the rate of deformation. We do this by developing the concept of strain, a quantity that is familiar. In the continuum treatment, the stresses are continuous, and hence, quantities that represent rate of deformation must also be continuous or point quantities. Strain is the relative increase in length. It is also a point quantity and, in a solid under stress, the strains at different locations can be different. A point quantity that reects strain is dened as follows. Consider the simple case where only a normal force is exerted in the direction on a sample. Consider two neighboring points, A and B, on the axis in the sample. Let the vector pointing from point A to point B be . On application of the normal stress in the direction, the points A and B move to a new location. In relation to point A, however, point B moves in the direction (because stress was applied in the direction alone) by the vector , as shown in Figure 9.1. The point quantity reecting strain is dened as the ratio ^ in ^ . Strain under shearing action may be similarly dened the limit tending to zero i.e.,

Vector after application of force

4 4

Deformation vector given by j


x

Deformation vector given by i

Vector after application of shear force in y direction Vector before application of force Elongational strain
y Shearing strain

Figure 9.1: Strain in elongational and shearing motions. by examining the motion resulting on application of a force in the direction. Again consider the vector separating the points A and B. In relation to A, point B now moves by . The ^ . In the theory of elasticity, point quantity corresponding to shearing strain is given by it is these strains that are related to stresses, e.g., through Hookes law. In case of uids, we are interested in the rate of strain and not in strain itself. Consider the example above of elongational strain to see how strain rate can be dened. If the strain occurs in time , then a quantity that reects the strain rate is the ratio

'

4 4

H`

! 9
6

A point quantity corresponding to this, can be dened by taking the limit of this ratio as and tend to zero. 6 The limit with respect to time, however, is given by velocity between points A and B. Hence

 

 !9 4 4
I0

I , the

component of the relative

 

I A
52

is representative of strain rate. Similarly, for the simple example considered earlier on shear, the quantity I is representative of shear strain rate. Thus the rate of strain is related to the relative velocity between neighboring points. It is easy to see that strain is actually a tensor since the relative displacement is a vector and it can vary in three dimensions. The same applies to the velocity gradient also. Neighboring points can have relative velocities in three directions and the relative velocities can also vary in three directions. It turns out that the velocity gradient tensor does not always reect the relative motion between points, i.e., a velocity gradient can exist while relative motion is absent. Consider the simple example of a cylinder of uid undergoing rotational motion as if it were a rigid body. It is obvious that in this motion, points do not move with respect to each other or that relative movement is absent. Let I the cylinder rotate with an angular velocity . Then its velocity in the v direction is given by t A p . It is easy to see that the partial derivative of the v component of the velocity with respect to p is not zero. Thus, even though the velocity gradient tensor has non-zero components, relative motion may be absent. However, stresses can exist only if there is relative movement. Hence, it is necessary to nd a quantity that reects relative motion. Without going into further mathematical details, we give the following result. The tensor given by I I describes rigid-body type of rotation. Thus it does not contribute to the generation of relative motion and hence to stresses. The tensor that contributes to relative motion is given by

4 4

I= M

4= 4
I

This quantity is a tensor and plays a signicant role in determining the stresses generated in a owing uid. The tensor referred to as the rate of deformation or strain rate tensor, and is dened as I 6 H I=

4 M 4 =V Before we leave this section, one last comment is4 in order. It is easy to see that 4 I= I 4 4= 4 sign,4 has only three nonzero components. It also is antisymmetric and hence, if we ignore the turns out that the three components are given by the three components of the curl of velocity vector, b . The vector b also plays an important role in understanding uid motion and is given a name: vorticity. We will return to it later.

4= 4

A L

9.2 The Newton Stokes law of viscosity


Hookes law relates the strains caused when a stress is applied to a solid body. It proposes that the strain and stress tensors are linearly proportional to each other. A similar assumption that 53

the stresses generated and the strain rates are linearly proportional to each other gives rise to the Newton-Stokes law of viscosity. Consider a situation where uid is not owing and is hence in equilibrium. Then, under characterizes the state isothermal conditions, pressure of a single component uid. Here the stress tensor is given by where the superscript indicates the state of equilibrium. When O a uid ows, additional stresses are generated. For convenience, the stress is still thought of to have an isotropic part and is written as

where is referred to as the deviatoric part of the stress tensor. The superscript is removed from the rst term to indicate that the isotropic pressure can be different from the equilibrium pressure. The quantity is still called pressure, although some books refer to it as hydrodyO namic pressure. It is dened for moving uids as one third of the mean normal stress at any point: 6 This implies that = =@= = 0. When a uid ows, the stresses generated are expected to be proportional to the rate of strain . If, in analogy with Hookes law, it is assumed that the deviatoric stresses are linearly proportional to the strain rate tensor, the Newton-Stokes law of viscosity:

A M O

Q R

Ic

=@=A = O

A L_

= @= =

is obtained3 . The last term on the right hand side is needed to make = =@= = 0.4 Almost all uids obey the Newton-Stokes constitutive relationship. Fluids which obey this relationship are referred to as Newtonian uids. Those which do not obey this relationship are known as non-Newtonian uids. Polymers and their solutions are a prominent class of non-Newtonian uids. Suspensions are another class of uids which can be non-Newtonian, e.g., paints and emulsions. I Examine this relationship further to understand it better. Consider a simple ow where only is nonzero and is a function of alone. For this ow, the Newton-Stokes law gives

Ic

Consider a plane at  A shown in Figure 9.2. According to our notation, the force exerted onas by elements at5 A the plane is m , and for this ow, it is equal to i. Thus, is the component of this force. We expect it to be directed in the M  direction for the velocity are traveling faster than gradient shown in Figure 9.2. This is because the elements at A  the elements at A , for the velocity gradient shown in the Figure 9.2, and hence the elements

l A A _   I

.D

3 Note the sign convention here: the sign in front of viscosity is +. As mentioned earlier some authors use a negative sign. 4 Note the sign convention here: the sign in front of viscosity is +. As mentioned earlier some authors use a negative sign. 5 on the plane. However, in view of the fact that they It is the force exerted by all the elements above are short ranged, all is often is abridged by at.

ed g4 f

54

Velocity in x direction Stress exerted by elements at y=0+ on the y plane

x Stress exerted by elements at y=0on the y plane

Figure 9.2: Stresses in a one-dimensional ow.

A in the positive direction. The Newton-Stokes law does drag at A the elements at _  I ^  is positive since is positive, and is positive. Thus the Newton-Stokes show that law is consistent, as it should be, with our intuitive expectation. It is the positive sign appearing in the proportionality between the viscous stress tensor and the strain rate tensor that ensures the consistency. There is more to the ug K sign and we will return to this later. Stress is measured in Newton/m . Velocity gradient has units of s . The unit of viscosity is therefore kg/(m s). The unit of viscosity in c.g.s. units is g/(cm s), also known as poise 6 . One hundredth of a poise is referred to as (Only one chance to guess this!) a centipoise, cp. Most liquids commonly encountered have a viscosity of 1 cp. In contrast, most gases have a viscosity of the order of 0.001 cp. _~^P . This Another quantity of interest in uid mechanics is the ratio of viscosity to density, K ^ is known as kinematic viscosity and given the symbol . It has units of < : , the same as diffusivitys. The c.g.s. unit is called Stokes. The signicance of the identity of units with diffusivity will be seen in the next section.

9.3 Molecular interpretation of the stress tensor


Earlier it was indicated that surface forces are molecular in nature. These are the manifestation, on the macroscopic level, of the attractive forces between molecules and the momentum exchange occurring when molecules from slower-moving and faster-moving regions collide due to random motion. Imagine two trains moving parallel to each other7 at different speeds. They are loaded with coal. Suppose workers from train A shovel coal at some rate so as to fall into compartments of train B. The workers from train B reciprocate and shovel coal at the same rate into A. Thus the masses of both trains remain the same, but there is an exchange of momentum between them. Coal from the slower train, when it falls into the faster train, decelerates the faster train. Similarly, coal falling from the faster train into the slower train accelerates the latter. Thus there is a force exerted by each train on the other. The basis or origin of the force is the exchange of momentum. This force arising out of momentum exchange is the analog of the surface forces. As molecules move past each other, an exchange of momentum occurs either through collisions
The unit is named after Poiseulle, pronounced Pwajuah. He is the doctor who established the relationship between pressure drop and average velocity in laminar ow, obviously in connection with ow in blood vessels. 7 This portion is paraphrased from a book by Sears on the kinetic theory of gases.
6

55

or under the inuence of intermolecular forces, and gives rise to surface forces. Thus the stress tensor has amolecular interpretation: it represents the momentum ux across any interface. In particular, is the momentum ux in the negative direction8 . In gases the momentum ux is dominated by collisional exchange. Simple dimensional arguments can used to arrive at an expression for momentum ux in gases. Let be the typical distance molecules travel before they collide9 and transfer their momentum. move I Molecules ^ 10 in the with a characteristic velocity , . Suppose there is a velocity gradient uid.  plane Refer to Figure 9.2. All molecules lying between A and A will cross the A without suffering a collision. Thus they collide and exchange momentum after they cross the A  plane. An estimate of relative of these molecules with respect to molecules with I velocity ^ . Therefore, on colliding, whom they collide is given by they exchange momentum  A equal to the mass of all the molecules crossing the plane multiplied by the relative I8 ^ P velocity. Thus momentum exchanged per unit area is given by ( )( ). This exchange ^ . Thus momentum would have occurred in time taken by the molecules to cross the plane:  I P ^ , and hence ux is therefore given by

4 4

I 4 4 c P 2 ji4 4 This in turns gives a valuable result: _ P jki As mentioned earlier, i can be identied with the mean free path and j with the root mean square velocity. If these are substituted into the above expression, the following result is obtained: _ mlUn ^Ho K

jki

4 4

i i ij 4 4

where m and d are the mass and diameter of the molecules, T is the temperature and k is the Boltzmann constant. It is interesting to note that the viscosity of gases therefore increases with temperature.

9.4 Momentum diffusivity


The above derivation has important features, and it is worthwhile to understand it thoroughly. The central idea here is that exchange of momentum occurs due to random motions of any entities. Momentum spreads out into space or diffuses due to random motion. Hence diffusive motion of molecules is the underlying mechanism of momentum transfer and momentum ux is manifest as viscous stresses on macroscopic scale. It is customary to assume that the diffusive
8 The negative sign is there to make the stress vector interpretation, the constitutive relationship used in this notes, and the momentum ux interpretation, consistent between themselves. If we had used a negative sign in our constitutive relationship, would be the momentum ux in the direction. Such a convention has some advantages (though we do not use it and instead have opted for the more widely used convention): for one, the momentum ux, as expected in analogy with diffusion, is from regions of high velocity to those at lower velocities. 9 It is commonly referred to as the mean-free-path. 10 Gas molecules move with a velocity corresponding to random movement as well as the macroscopic motion. However, the former is usually very large compared to the latter except in hypersonic ows. The characteristic velocity can therefore be computed from the temperature, which is related to the mean kinetic energy of molecules.

W@p@q

56

ux of a quantity is proportional to the concentration of that quantity. Hence, the momentum ux can be written as H P I  P I  where

c A h 4

Thus or can be thought of as momentum diffusivity. The above derivation has important features, and it is worth while to understand it thoroughly. The central idea is the spreading or diffusion of a quantity due to random motions of any entities that act as carriers of that quantity. For example, energy, and mass could diffuse due to random motion of molecules and these two are of specic importance to this course. If random motion can be characterized by a length scale and a velocity scale, the product of these two quantities is a good estimate of the diffusion coefcient of any quantity.

_^0P

4h A

57

Appendix 9.A Vector calculus


There exist many situations where a specic physical quantity varies from place to place. The variation itself could depend on the direction in which the change is measured. Thus, if two = , the difference in the value of a dependent scalar variable (such as points are separated by the temperature T) between the two points is given by

 A =

It can be recognized as a dot product from the appearance of sum over index. It a= repeated = = , and another is the dot product points from one point to the other, F of the vector that ^ = . The latter vector is called the gradient of a scalar and vector whose component is  can be written as

 = =

4 4

The idea of a vector can be used in some generalized sense and think of = as the component of a vector. However this vector in itself does not represent any physical quantity. It acquires meaning only when it is applied to or operated upon a physical quantity, such as F ^ = temperature. Such a quantity is referred to as an operator. Thus, is the component of the gradient operator . The gradient operator applied to a scalar results in a vector, which is referred to as the gradient of a scalar, e.g.,  . It is possible to construct not only a dot product of the gradient operator and another vector, but also a cross product between the two. The dot product results in a scalar, and is called the divergence of a vector. The cross product generates a vector, and is called the curl of a vector. In the indicial notation these can be written as

 A = =  =

4 4

4 4

b
Prove m

m A A

= =

4=

I=

The ease with which vector identities can be proved with the help of indicial notation can once again be demonstrated by the following examples. Example

4 =

 A hm m
Left hand side

Jb

Jb r

A A

= = m~ I = I  = m = =

58

The unit vectors do not vary in space, and hence can be treated as constants 11 . They can therefore be taken outside the differential operator.

everywhere. Such an exchange would not alter the term both the indices are being summed over all possible values. Continuing with the derivation Left hand side

= H I Left hand side = = M H I A = = M I = H I A = = = VSM = In obtaining the second term, the index j was exchanged with A

4 4

4 4

=V

I =V

I = = V

The presence of sum over a repeated index ( j ) suggests that the above terms can be written as a dot product between two vectors. Now consider the right hand side. Right hand side

4 4

=V = I

I = = V

A A A

m = = =

I m = = I = = = I = H I H I = = V = = = V =

4 4

4 4

4 4

= = =

Exercises Prove the following identities (i) (ii) m (iii) m (iv) (v) (vi) m

bb

bb b s b ( b Jb b b b b

:n  A Q:  :  M h   A   m m h  A (divergence of a curl is zero)  A  (curl gradient of is zero)  A    A Kg m  m m 

bb

b b bb Jb

9.B Tensors and frames of reference


9.B.1 Metric tensor
Earlier, we referred to different observers using different frames of reference. To facilitate easy understanding, let us assume that Cartesian coordinates are used by both the observers. The
11

This is true only of unit vectors in Cartesian coordinates as will be seen later.

59

unit vectors of the unprimed observer are e = , while those of the primed observer are e = . Let the origins of both observers be the same. Any position vector x can be written as

= =A =

The components of the two systems can be related to each other by taking a dot product with unit vectors of one system. For example

A = m = = = R = = R m m

where

It is easily seen that g is a tensor. It is referred to as metric tensor. In a similar manner

3A = m = = = R = =

The above is the rule of transformation between the two frames of reference. In fact the above can be taken as the denition of a vector: The components of a vector must satisfy the above rule of transformation. Only those quantities that satisfy the rule of transformation are vectors. The rule of transformation is shown in matrix notation:

) Aut ) c

") ) u A t and

An important property of the metric tensor is obtained by repeated appliation of the trans" formation rule: Hence Writing the above explicitly

) Avt ) Avtkt ) t0t " A q t :ug Avt "

= R = A R =

In a matrix notation, the above also can be written as

Thus the metric tensor is orthogonal to itself. One more property of the transformation rule is also important. The transformation rule can be applied to any vector. Thus

M A = R = = M =  A = R = =
60

and

Thus it follows that

Similarly

u A = = R M A = R = = M = 

and

A = R = = R = 3A

Thus

g does not depend upon spatial location for Cartesian coordinates. For more complex coordinate systems, such as curvilinear systems, or those that have been obtained by deforming Cartesian system, g is not a constant.

9.B.2 Differentiation
Consider the spatial variation of a scalar variable, like temperature,  . There can be three ^ = c A 6 c L c or from a different observers view point  ^ c 8 A 6 c L c . derivatives  Do these three quantities form a vector? One has to verify if the transformation rule is satised by the components in the different frames. From chain rule of differentiation

4 4

4 4

4 4 4 A 4 4 4 4 They do satisfy the transformation rule and hence the three derivatives form the three components of a vector. The vector is dened as the gradient vector. It is customarily written as . b  Let us consider one more derivative that is important. Suppose an object is marked and we follow its position as a function of  time. The three components of its position vector can     ^ 6 L ^ = A A be differentiated with time to get cu c c or c [ 6 c L c . Are these three
quantities components of a vector? One has to check if these obey the transformation rule:

 A

=  = = R =  =

  A

A
If g is not a function of time,

= =  =   R = =  =

A 4 A 4

 = A =  = R 

61

and hence the three quantities are the components of a vector. It is the velocity vector of a particle: 

= A =  =

If g is a function of time, the frames of reference are perhaps moving with respect to each other and more complex denitions of time derivative may have to be adopted to make it a vector. It is these considerations that go into formulating rehological equations of state.

9.B.3 Second order tensors


It is easy I = ^ to see that components of velocity can be a function of position. Hence quantities can be evaluated. What kind of quantities they are? Intuitively, it appears that like they may require two transformations: one for velocity and one for position. There seems to be an order here. A scalar has no sense of direction. It can therefore be dened as a quantity that follows the following transformation rule:

4 4

 A

It is of zeroth order since the metric tensor appears to the zeroth power. A vector is dened by g

) Aut )

ut 

Now, the order to which the metric tensor appears in the transformation rule is equal to unity. Thus from a scalar to a vector, the order has gone up by one. This notion can be generalized to dene rules of transformation where quantities of higher order of the metric tensor are used. All such quantities which obey prescribed rules of transformation are called tensors. A scalar is a tensor of zeroth order. A vector is a rst order tensor. The second order tensor can be dened analogous to a vector. Its components must obey the following rule of transformation:

c = A R
A

R = R

It can be written in matrix notation as

As was seen earlier, the second order tensor can also be expressed as a dyad:

" w t t R R A

The order of the unit vectors I = ^ is very important since they go with the indices which have order. Let us return to . Are these the components of a second order tensor?

4 I 4 4 A 4R R = 4 4 4 form the components of a second where we have assumed g to be a constant. Hence, these order tensor. The same second order tensor could be written in dyadic form as I b A 4 4

62

4 4

I= A

I=

9.B.4 Principal coordinates


In uid mechanics and rheology, often one looks for a coordinate transformation that makes ow look simpler. Thus the velocity gradient tensor has nine components. But more insight may be gained if a coordinate system is used where it appears that there are only three components. Several uid mechanics books will call this as a transformation where the ow is seen as a combination of rotation and dilatation. This section deals with the mathematical principles behind such a transformation. " Consider a real symmetric matrix A. Then it is possible to nd an orthogonal matrix Q such that Q AQ is diagonal. If" a second order tensor is symmetric, then, it is possible to nd a matrix to diagonalize it: Q Q is diagonal. It was shown that g is an orthogonal matrix. Thus if we let Q= g, then, is diagonal in the transformed coordinate system. Such coordinates are some times referred to as principal coordinates, e.g., in the context of stress tensor. Similarly, the velocity gradient tensor is written as a sum of skew-symmetric and a symmetric tensor. The latter is the shear rate tensor while the former refers to rotation of the uid elements. The shear rate tensor can be diagonalized and therefore, in the transformed coordinates, the ow is dilatational.

9.B.5 Invariants
An interesting question to ask is whether there exist some quantities, that can be obtained from components of a tensors, but whose whose values do not change during transformation of frames of reference. Their importance is clear. Suppose there is a physical property. Its value can not change when different frames of reference are used. This section is about mathematics of such quantities. Suppose certain scalar quantities can be formed from the components of a tensor. Since they are scalars, their values are unaffected by change in frame of reference. They are referred to as invariants of the tensor from which they have been formed. Consider the following example.

c = = A
A A

R = R =

and hence

= = A =

R=R=

The invariants play an important role in formulating rheological models.

9.B.6 Frame indifferent equations


Consider an equation involving vectors and tensors:

A 3m
63

This equation is equivalent to three equations:

I =A A

c = R m

We want to know if this equation is valid in a different frame of reference also, i.e., does

also hold? Now,

I = A A A A A A

= I R R = R = R R ? ? R ? ? ? = R R R R   ? ?  = ? ? =

which is the same as

The equation is unaltered by the transformation or is frame invariant. This is the advantage in writing equations in vector form. Some equations may not be frame invariant since the transformation may involve g which can be time dependent or space dependent. Earlier on we mentioned that physical laws are expected to be frame invariant. Such a property is ensured by writing the laws in terms of vectors.

R m

64

Chapter 10 PHYSICS OF FLUID FLOW


We describe what happens when forces are applied to a fluid. We discuss what happens at the boundaries between fluids and solids. We discuss the terminology used in fluid mechanics. ...the river had something special to tell him, something which he did not know, something which still awaited him. ... It seemed to him that whoever understood this river and its secrets, would understand much more, many secrets, all secrets. From Siddhartha by Hesse Transport of heat and mass are intimately linked to ow of uids. Fluid motion plays an important role in many phenomena you observe around, and even inside, you. Flow and oods of rivers, rain and monsoon, movement of water to the leaves at the top of trees as tall as 100 m, ow of blood in your body and, through it, tranport of oxygen, drugs, nutrition etc to cells, the list is endless. Mathematics plays an important role in quantifying these phenomena though it is still incomplete, and in some instances, far from being satisfactory. But, mathematics is only a tool, and an understanding of ow phenomena in a physical way, at least partially, is very important. In this chapter we discuss a few physical ideas that allow us to get a feel for what happens when a uid ows. The subject dealing with uids not in motion is referred to as uid statics. Here, pressure and gravity are the only forces that act on the uid. As the uid is not moving, the net force on any uid element must be zero. Hence the pressure forces must balance the weight of the uid. This is a rather simple situation and is not considered any further in this notes. Consider how a solid body moves. If it is static, a force has to be applied to make it move. On application of force, it acquires acceleration, and hence velocity, and begins to move. As it moves new forces such as friction may arise. It accelerates as long as all the forces are not balanced. A situation may occur when the forces balance, i.e., the net force is zero. When this happens the body moves with a constant velocity. Thus, when a solid ball is dropped into a uid, its weight is usually greater than the buoyancy force exerted by the surrounding uid. Hence it begins to accelerate and acquires a velocity in the direction of gravity. As it begins to move, the uid exerts a frictional force which increases as the velocity of the ball 65

increases. This frictional force continues to increase till the frictional force and the buoyancy force together equal the weight. The body continues to move with whatever velocity it had acquired when the net force became zero. This velocity is what we refer to as terminal velocity. This kind of description, an account of cause and effect, that we seek in the solution of problems in tranport processes. Often, problems are complex, and it is not possible to tell such a story to a complete or even satisfactory extent. However, the broad features can still be understood using some basic principles. It is these that we discuss below.

10.1 Pressure ows and drag ows


As with solids, if a packet of uid is static, a force has to be applied to move it. Generally a force can be applied in two ways: by using a pump to apply pressure, or by using a solid to drag it. It is observed that when a uid is in contact with a solid and the solid moves with a velocity v, the uid immediately adjacent to the solid, i.e., the layer of uid in contact with the solid also moves with the same velocity1 v. This is called the no-slip condition. The no-slip condition is valid for uid-uid interfaces also. The uid layers in immediate contact with the solid cannot slip past the solid, i.e., there is no relative velocity between the solid and the uid at the surface where they are in contact. Thus, if a solid plate is placed in a uid and dragged through it, it creates motion in the uid. Motion is often created by a combination of pressure forces and the effects of no-slip condition. For short, the latter effects will be called as being the effect of drag forces.

10.2 Development of ow due to drag forces


Consider a channel between two large plates, shown in Figure 10.1, lled with a uid. Let the uid be static. Suppose at some time, the top plate is dragged with a constant velocity a in the direction while the bottom plate is kept stationary. The uid next to the plate also

{z yx

Increasing time y x y Velocity Note velocity is Vx at the top plate and zero at the bottom plate

Fluid

Fluid as well as plates are stationary

Bottom plate is stationary

Figure 10.1: Development of Couette ow. must move with the same velocity according to the no-slip condition. As soon as a layer of uid adjacent to the plate begins to move, because away from the plate is not yet in I ^ the uid A  . lHence motion, a velocity gradient is generated, i.e., the moving uid drags 2 the next layer by exerting a force in the direction , i.e., . Hence the next layer begins
See the later section on boundary conditions for more comments. Focus on the physics at the moment and do not worry about why it is in greater detail later.
2 1

4 4

W q?p

and not

W p>q

. We will deal with these

66

to accelerate and move. The motion spreads in this manner towards the bottom plate. As the bottom plate is stationary, by the no-slip condition, the layer next to it cannot move. Thus the uid layer remains stationary there and continues to exert a retarding force on the layers above it. The effects of the retarding force of the bottom plate also spreads upwards. Thus each layer experiences an accelerating force from the layers above and a retarding force from the layers below. At some point in time, the two forces become equal in magnitude and hence the net force on an element vanishes. From that time, the velocity of the uid elements remains constant. The ows where the velocity and other variables do not change with time are called steady ows, while if they change, the ows are called unsteady ows. In the example just A while it is a at considered, the velocity is not the same at all points of the uid: it is  . The spatial dependence of velocity is called a velocity prole. In the example zero at A considered, the velocity prole is non-uniform, i.e., the velocity is a function of the location in the uid. A schematic of the development of the velocity prole with time is shown in Figure 10.1. The ow between two parallel plates where one plate is moved with a constant velocity while the other remains stationary is referred to as Couette ow. I[ A As will be shown later, the velocity prole is linear in Couette ow, i.e., a ^ .

10.3 Development of ow due to pressure gradients is caused by applying a pressure gradient: a uid Consider another example where the ow entering a pipe of length W with uniform velocity a and pressure B . Suppose the pressure B as shown in Figure 10.2. If this experiment at the end of the pipe is maintained at
Pressure here is P o
r z

Pressure here is P

Fluid entering with uniform velcocity V

Devloping region

Fully developed region

Developing velocity profile

|~} |~} |~} |~} |~} |~} ~} | ~} | ~} | ~} | ~} | ~} | ~|~| } }} } } } } }  }  }  }  }  

Forces on a small annular cylindrical element. It experiences an decelerating force on its outer periphery and a accelerating force on its inner periphery. Pressure forces are also accelerating

Figure 10.2: Flow of a uid in a pipe due to pressure gradient. goes on for a long time, steady state can be expected to prevail in the pipe. Since the pressure decreases down the tube, there is a pressure gradient. This gradient pushes the uid in the direction of decreasing pressure. Hence the uid elements entering the pipe will accelerate under the inuence of the pressure gradient, and their velocity increases. By the no-slip condition, however, the uid layer in contact with the pipe wall has to be stationary. Hence this layer exerts a decelerating force on the adjoining elements towards the interior of the pipe. Hence the elements entering the pipe at a radial location closer to the wall do not accelerate as much 67

as the elements entering near the center. The slower-moving elements near the wall exert a decelerating force on the faster moving elements, and this effect spreads across the radius of the tube till it reaches the center. A velocity gradient is therefore established. Thus each annular element, apart from the accelerating force due to pressure gradient, experiences a decelerating force on its face towards the wall and an accelerating force on its face towards the center. All the while the uid elements are also moving forward. Thus, the uid elements, as they move along the length of the pipe, acquire different velocities depending on their distance from the wall, i.e., the axial velocity is a function of the radial position of the element in the pipe. Such ows, where velocity changes in the direction of ow, are referred to as developing ows. If the uid elements travel sufciently far into the tube, the accelerating and decelerating forces may be expected to balance each other, i.e., the net force on the element is zero. From that location onward the velocity of the uid element cannot change but remains constant as the uid element travels down the tube. Such a ow, where the velocity does not change in the direction of ow, is called a fully-developed ow. If the ow is laminar, the prole is the well-known parabolic velocity prole. In the experiment just described, the velocity of each element is expected to depend on the radial location. Further it is also expected to change along the length of the tube for some length, and then remain unchanged with the axial location downstream. The length needed for the velocity prole to become fully developed is referred to as the entry length. This is shown in Figure 10.2. The ows where velocities are functions of two space dimensions are referred to as two-dimensional ows; where the velocities are a function of only one dimension, the ows are referred to as one-dimensional ows. In the example just considered, the developing ow is two-dimensional while the fully developed ow is one-dimensional.

10.4 Eulerian and Lagrangian description


In the description of development of pressure ow, it was argued that uid elements accelerated in the developing region. Notice that the ow itself is steady. Thus, to an observer sitting in a laboratory, i.e., from what is commonly known as the Eulerian view-point, nothing changes with time. However, an observer moving with the uid element would nd that the uid elements accelerate till they reach the fully developed region. A frame moving with the uid is called a Lagrangian frame. Observers in this frame are said to have a Lagrangian viewpoint. As predictions of uid mechanics theory are to be tested against experiments, Eulerian description is the most widely used. However, Lagrangian view offers a convenient platform for developing theories of some phenomena.

10.5 The no slip boundary condition


As shown earlier, the no-slip condition plays an important role in determining the characteristics of ow. The conditions which prevail at the boundary of two phases are referred to as boundary conditions. Boundary conditions come from nature 3 . They are based on observation. Thus, formal expressions of boundary conditions are needed. Consider a boundary between two phases as shown in Figure 10.3 The no-slip condition states that there can be no relative velocity between the two phases at the interface. Hence the two velocities at the
3

A quote from Prof. Amundson.

68

Phase II

Phase I

,1

Figure 10.3: Boundary conditions.

boundary are equal:

It should be noted that the above equation is a vector equation and hence is equivalent to three equations; one for each component. It is customary to write the three components of the no-slip condition along the normal and the tangential vectors to the phase boundary. At any point of the phase boundary we can construct a unit normal vector, and to emphasize that this normal is on a phase boundary, denote it by n . A unit tangent vector to the phase boundary can be constructed. In fact, an innite number of them can be constructed. However, once one of them is selected, only one more vector perpendicular g  K to it and  g the normal  K vector can be constructed. Select any one pair and denote it by t and t . n , t and t together form a right-handed orthogonal system. The three components of the no-slip boundary condition can be obtained by using this coordinate system. A dot product of the no-slip condition with the three mutually orthogonal unit vectors gives:  

 
A

   A The no-slip condition also implies the following equations: F\  g m   A  F\ K m  A


nm

Consider once again the two examples of ow development to examine the way to write the three components of the no-slip boundary condition. In the Couette ow the following results are obtained by applying no slip condition:

I A I A   at A and I A  I0  at A and A a at A I A I A I A  r t
at p

Similarly for the ow in a pipe, the following form the three conditions: where is the radius of the pipe. The no-slip condition as stated above assumes that mass transfer does not occur across the interface. Mass transfer occurs only in multicomponent systems. If mass transfer occurs, the phase boundary can have a velocity different from the velocity of species crossing the boundary. This situation is dealt with later when multicomponent systems are considered. There are interesting and important conditions where no slip condition does not seem to be obeyed. One such case is the movement of a three phase contact line. Imagine a drop of water put on a clean glass sheet. As soon as it is placed, there is a line where three phases, air, water and solid glass, are in contact. The drop begins to spread and eventually forms a thin lm. It thus covers new area which it did not occupy. To do this, water must slip past the solid surface. This process is called wetting. Reconciling the no slip condition and the wetting ows is still a problem being researched by many. 69

10.6 Stress boundary conditions


The boundary conditions can also involve stresses. They can however be derived. Refer to Figure 10.3. Consider an innitesimally small area
on the interface. Denote the total stress tensors in the two phases by and respectively. The forces exerted by the two phases
on the interface can be calculated from these two stress tensors. They are given by .

and . . Hence the net force acting on the interface4 is .( ). The mass of an interface is, however, zero. Hence the interface cannot sustain a net force on it. Therefore

 D .

D.

D . D  Au
om

D . D

De

The above is a vector equation and hence is equivalent to three equations. They can be obtained   g by taking a dot product of it with the three mutually orthogonal vectors n , t and t K . The resulting equations are

As an example consider a two-phase interface that coincides with the   g K t = j and t = k and the equations above become

 D  JD
F\ JD
F\ Ec E

nm [m  A om 8m  g A om 8m K A

JD. JD.
F\ JD.
F\

om [m  om [m  g om [m K T

(10.1) (10.2) (10.3) plane. Hence, n = i, (10.4) (10.5) (10.6)

O M A A A

c ? ? E  E 
O M

It is with the use of boundary conditions, constitutive relationships, and the physical laws that details of motion of uids can be determined.

10.7 Incompressibility of uids


As a uid ows, there are variations in pressure in the eld. In response to such variations, the density of the uid may change. For some uids the variations in density are small since their density is not very sensitive to pressure. Liquids cannot be compressed very much and hence belong to the class of uids where density changes are very small for large changes in pressure. In dealing with such uids, considerable simplications in equations can be achieved by assuming that the density remains constant, i.e., that they are incompressible. Usually such an assumption is valid as long as ratio of the velocity of uid to the velocity of sound in that uid is very small. The ratio is called the Mach number, . Thus incompressibility is a good assumption as long as In unsteady ows, in addition to the above condition, the time period of motion must be large 5 . It should be mentioned that most ows encountered in chemical engineering can be treated as incompressible.
Here the effects of surface tension are being neglected. If the interface is curved, a force due to surface tension has to be added. 5 See Batchelor for more details.
4

70

10.8 Laminar and turbulent ows


A complex feature of uid ow is that the patterns of ow can change change drastically with minor changes in the conditions of experiment,i.e., in the operating variables. The ow pattern appears to have undergone a transition from one state to a completely different one due to small changes in the operating variables. The situation is like the change from liquid to vapor state when temperature is increased a little bit above the boiling point6 . Such transitions in uid mechanics are referred to as ow transitions. From the theory of dimensional analysis, it may be expected that experimental conditions where transition occurs may be combined into dimensionless groups, or parameters for short, and that the parameters are loaded with physical signicance. If such parameters can be idented, it may be possible to predict that a ow transition will occur when the numerical value of the parameters change from some critical values. An outstanding problem in uid mechanics is discovering such parameters and predicting the critical values at which ow transitions occur and ascertaining the path or the route followed to achieve the transition. Perhaps the most important parameter or dimensionless group is the Reynolds number , the product of the length scale of the experiment and a characteristic velocity of the experiment, divided by the kinematic viscosity of the uid:

At low Reynolds numbers, ow is usually orderly and uid particles follow regular and welldened paths. Such ows are called laminar ows. When the Reynolds number is increased, laminar ow can become unstable. The ow can undergo a transition giving rise to another a more complex but still orderly ow. As the Reynolds number is further increased, no ow may be stable and the ow may not have any order. In such a disorderly ow, the uid particles follow complex and time-dependent paths, even when the experimental conditions are kept constant. Such a ow is called turbulent ow. At present it is not clear whether laminar ow must undergo innite number of transitions before giving birth to turbulence or a nite number of transitions are enough. An example of development of turbulence is ow in a jet of water issuing from an open tap. Suppose the ow rate from the tap is small. Up to small distances from the exit of the tap, the surface of the jet is smooth, the particle paths are straight and uid elements coming out at a radial location do not change their radial position. It is in this sense that uid ows as laminae and hence this type of ow is called laminar ow. Further down however, ripples develop on the surface of the jet and the ow is wavy. If the ow rate is increased or as the Reynolds number increases, the wavy nature develops at shorter distances from the exit of the tap. Even further down, the orderly wavy ow yields to a more complex ow where no order is visible or to turbulent ow. It is well known that the pipe ow we discussed earlier becomes unstable when the Reynolds number is increased beyond a critical value. Turbulent ows are very difcult to analyse and most of this notes will be concerned with laminar ows. A study of these, however, is important in developing a physical feel for ows. EXERCISES 1. For the situations given below, specify the relevant boundary conditions.
Fluid mechanics is not an equilibrium phenomena and hence the ow transition being discussed cannot be taken to be a phase transition.
6

igj h

71

a. A rectangular cavity, W A (extending to innity in the direction perpendicular to the paper), is lled with an incompressible Newtonian liquid. The top plate is being moved in the M direction with a constant velocity a . Sketch only stream lines; do so also for the case W . ] b. A thin circular disc (radius= ) is rotating in a beaker (diameter= ) of liquid (height= ) with a constant rotational speed . Assume that the diameter of the rod holding the disc is zero and that the liquid-gas interface is at.


W Liquid y L Part a

H x


R z D Part b

72

Chapter 11 VISUALIZATION OF FLOW


We discuss the paths followed by fluid particles Stream lines are introduced. All important problems in uid mechanics are complex and it is very difcult to solve them exactly. Physical insight is necessary to make reasonable simplications so as to obtain solution to the complex problems. One practice helpful in developing physical insight is to imagine how things happen. One such good practice is to imagine how uid particles move in a ow eld. They must accelerate if they experience forces and hence it is good to think about the various forces acting on them. Other principles that inuences ow are the mass conservation, and the incompressibility assumption. For example, if a uid ows through a duct which narrows down, it has to accelerate. A more difcult thing is to imagine the simultaneous effect of the forces and the mass conservation principle. Let us take a few examples to make these ideas clear.

11.1 Stream lines


Consider a ow eld. Starting from a given point A, a line can be drawn such that it is parallel to the velocity vector at A. Taking a point B on this line, but very close to point A, a line parallel to the velocity vector at B can be drawn. Another point C, very close to B, can be taken on this line, and a line parallel to velocity vector at B can be drawn. A line is obtained if this process is repeated, and all the points are connected. Such a line is called stream line. Obviously, the velocity vectors at every point on a stream line are tangent to it. Flow elds can be visualized by drawing stream lines. If the ow is steady, the stream lines passing through a given point do not change with time. If a uid contains very light tracer particles, e.g., smoke, they faithfully follow the path of uid itself. Suppose two snap shots of a ow eld at two different instants separated only by a short time interval are available. The path followed by the uid during the small time interval can be traced by joining the two successive positions of the same particle. The path would of course be tangent to the velocity of the particles. The path followed by uid can be traced if several snap shots of the ow eld are taken and the locations of the same particle at successive time intervals are joined. Equvalently, a photograph made with long exposure time will show streaks, which are the paths followed by uid particles during the time of exposure. Suppose the ow is steady. The paths followed by successive uid elements that pass through a given point will not change with time. The velocity vectors are parallel to these 73

paths. Therefore, under steady conditions, the stream line passing through a point and the path followed by any uid particle that passes through the same point are identical. Stream lines in steady ows can be made visible by injecting tracer particles and taking photographs for long exposure times.

11.2 Flow in a converging diverging section

Figure 11.1: Stream lines in a converging diverging section Consider steady ow through a converging diverging section. By the no-slip condition, velocity of the uid at the walls must be zero. In particular, the component of the velocity normal to the wall must be zero. Hence, particles can be expected to travel inward and follow the contour of the wall. The stream lines are then expected to converge in the converging section and diverge in the diverging section as shown in part (a) Figure 11.1. At steady state, the mass ow rate at any cross section must remain constant from the principle of conservation of mass. Hence the velocity would increase towards the throat from the entrance and decrease from the throat to the exit. The uid elements accelerate in the converging section and decelerate in the diverging section. If body forces are negligible, momentum balance suggests that pressure forces must exist to balance the acceleration created by changes in the velocity of the uid elements. In particular, the pressure at the throat would be less than at the entrance so that the pressure gradient will accelerate the uid element in the axial direction. For the same reason, the pressure forces will oppose the ow after the throat so that the uid elements can decelerate. If viscous forces are absent, the above description is equivalent to saying that the work done by pressure forces is converted into increase in kinetic energy of the uid in the section before throat and that, in the diverging section, kinetic energy decreases to overcome the adverse pressure gradient in the section beyond the throat. Here an interesting thing can happen. Due to the no-slip condition, the elements near the walls experience drag force that opposes ow as well as the favourable pressure gradient. Due to this the elements near the wall will gain less kinetic energy than those in the center of the channel. However, they will sense the same pressure force opposing their ow in the diverging section. As a result they may not be able to overcome the opposing forces while sticking to the wall. Hence they separate from the wall and a back circulating ow or an eddy structure develops in the diverging section. This is depicted in part (b) of Figure 11.1. Which situation amongst those shown in part (a) or (b) prevails depends upon the Reynolds number and the angle of the throat.

11.3 Flow past a sphere


Consider uniform steady ow past a stationary sphere. Far away from the sphere, its effects will not be felt and the ow will be uniform. However, those uid elements that approach the 74

sphere will have to bend around and go past the sphere due to the no-slip condition. Hence they will accelerate till they cross the equatorial plane of the sphere. After that, they will converge back into the rear of the sphere, and hence they will decelerate. Here the situation is similar to that described in ow through the converging-diverging nozzle. Hence at large Reynolds numbers, separation of ow can occur. The average stream line pattern expected is shown in Figure 11.2.

Figure 11.2: Stream lines in ow past a sphere

11.4 Flow in a pipe


Consider steady ow in a pipe under the inuence of constant pressure gradient. Suppose that the uid enters the pipe with uniform velocity. In the entry portion of the pipe, the uid elements towards the center accelerate under the inuence of the pressure gradient, while those near the wall decelerate due to drag created from the no-slip condition. In other words, at some axial location inside the pipe, the velocity at a radial location near the wall will be less than that at the entrance at the same radial location. Consider an annular control volume near the wall as shown in Figure 11.3. As the velocity in the axial direction at entrance is greater than that inside the pipe, there the volumetric ow rate in the axial direction through the control volume must be lower than at the entrance. But as the liquid is incompressible, the same volume entering the control volume must ow out of the control volume. Hence there must be a radial ow. This makes sense since. Fluid elements in the center do not experience drag but only the accelerating inuence of the pressure gradient. Hence uid from the wall regions has to move to the center to account for increase in the axial velocity as the uid elements in the central zone. Thus both radial and axial velocities must be there in the entry zone and the radial velocity must vanish when the velocity prole becomes fully developed. The expected pattern of stream lines is shown in Figure 11.3.

Flow towards the center to compensate for theslowing down near the wall Expected strream lines

Figure 11.3: Stream lines in a pipe including the entry zone

Exercises 1. Plot stream lines for the following cases: 75

a. Fluid entering a channel and owing due to pressure gradient. b. Flow past a weir. 2. For the situations given below, sketch the expected stream lines. , (extending to innity in the direction perpendicular to the a. A rectangular cavity, W A board) lled with an incompressible Newtonian liquid. The top plate is being moved in the M direction with a constant velocity a . Sketch only stream lines also for the case W . ] b. A circular thin disc (radius= ) is rotating in a beaker (diameter= ) of liquid (height= ) with a constant rotational speed . Assume that the diameter of the rod holding the disc is zero and that the liquid-gas interface is at.


W Liquid y L Part a

H x


R z D Part b

76

Chapter 12 SHELL BALANCE OF MOMENTUM


We consider examples of shell balances: flow of a film falling down an inclined plane; flow in a pipe; flow of two stratified fluids in a channel. In any practically important ow, the characteristics of the ow generally depend upon both time and spatial coordinates. Such situations are complex and the mathematical equations that describe the problem exactly often can not be solved analytically. However, any physical insight that can be gained by solving simpler but similar problems is valuable. Hence, in tackling complex ows, the following procedure is used. First we try to visualize what is happening in the situation being considered. Then we try to see whether, while retaining the physically important aspects, any simplications can be effected so that a solution can be obtained. Such a solution can be used as a rst approximation to the more complex ow. It can also form the basis for either numerical solution or some type of series solution of the complex ow. We illustrate this procedure in this chapter by considering a few examples of simple ows where the equations governing the motion of a uid can be derived in a lucid manner that reveals the underlying physics, and solved. These examples, while not conforming faithfully to any practical situation, are mathematically tractable. They form a rogues gallery, showing characteristics that are typical, and thus offer insights into more complicated situations whose details cannot be so easily deciphered.

12.1 Laminar lm owing down an inclined plane


12.1.1 Problem statement
Consider an incompressible liquid contained in a tank. The tank has a slit opening at the bottom. The opening leads onto an inclined plane as shown in Figure 12.1. Let the tank width, the dimension in the direction perpendicular to the paper, be very large compared to the thickness of the lm. The uid ows out of the tank and down the plate because of the head available. If the uid has been owing out of the tank for a long time, we expect steady state to be established.1 We expect both the ow rate and the lm thickness to increase if the opening
The tank may empty as time progresses, and steady state may not be attained. The rate of inow into the tank is such as to keep the the level constant.
1

77

Inflow of fluid to keep level constant vx Fluid A B Width of the strip is x

Direction of gravity x

y
Expand

yx yx
g cos C
y+

Solid wall

vx

Figure 12.1: A lm falling down an inclined plane.

is made larger. We can therefore dene our problem: calculate the ow rate as a function of the thickness of the lm at steady state. The rheology of the uid is required to solve this problem. The uid will be assumed to be Newtonian.

12.1.2 Simplications and assumptions


The geometry is ideally suited to Cartesian coordinates. Since the tank width is very large compared to the thickness of the lm, we may perhaps assume that the tank and the inclined plane extend to innity in the T direction. As the uid comes out of the tank it is exposed to the atmosphere and hence the lm surface is at atmospheric pressure. As the uid ows out, it continues to experience a force due to gravity that accelerates it down the inclined plane. The elements adjacent to the inclined plane, however, have to be stationary due to no-slip condition. Hence velocity gradients and consequently viscous drag forces are generated. These spread from the surface of the inclined plane toward the free surface of the lm. 2 Thus, after some distance down stream of the slit, the drag forces can balance the gravitational forces. From then on, acceleration becomes zero, and the ow becomes fully developed. We are interested in nding the fully-developed velocity prole in the uid. In other words, we are visualizing a situation where the gravitational and frictional forces are balanced. We assumed that the tank and the inclined plate are innitely wide in the T direction. Since the tank dimensions are large compared to the lm dimensions, it is unlikely that any signicant viscous stresses will develop in the tank. Thus, uid can be expected to ow out of the slit with a velocity dependent only on the height below the interface in the tank, e.g., according to Toricellis law. Thus, uid elements exit from the slit opening at a condition independent of T . Further, the inclined plane is also innitely wide. As uid ows onto it in a condition independent of T , we can expect this state of independence to continue as the uid ows down the plane. Thus, we do not expect any quantities to depend upon the T coordinate.
The gas-liquid interface is normally referred to as free surface. The reasons for this will shortly become clear.
2

78

12.1.3 Mass balance


Our rst task is to ascertain I which components of the velocity are nonzero. As the ow is I in the direction, clearly is not zero. We do not expect any cross-ow and hence we expect to be zero. We are looking for a solution in the fully-developed region and hence, by denition,

The above implies that the rate at which uid is owing in the direction remains constant. Since there is no input or output of uid in the direction, we also do not expect any ow in the direction. This is proved mathematically as follows. Consider the thin shell of thickness and a strip of length as shown on the left of Figure 12.1. We balance mass over this differential control volume:

I0  A

 A   P c l a m P c ll
M 

A , *)

*)

The control surfaces are the pairs of parallel surfaces AB and CD, and BC and AD. The normal vectors to these are i, and i, and j and j, respectively. Hence we get

as well as that I the ow is at steady state, the rst term is zero. Further, since the ow is fully developed, is not a function of . Hence the second term is also zero. Dividing the above equation by and taking the limit as the differential lengths go to zero, we get

I P0 I I I0  A  P P M y  M P M P y l M P We are considering that the liquid is incompressible, and so is constant. Both because this

Physically the limit being evaluated is the limit as I the volume of the element goes to zero. Integrating the differential equation, we nd that is at best a function ofI8 . Since no uid enters through the solid plane, the no-slip boundary condition implies that must be zero at I0  A for all values of . Hence must be zero everywhere. Thus we have the following results: I A I A  I I0  and A (12.2) Mass conservation has been used as the rst step, and it is the rst step very often. As we shall see later, this is known as using the equation of continuity. Note that the limiting process of a balance over a differential control volume or a a shell balance gave rise to a differential equation. Given the necessary boundary conditions, differential equations permit us to determine, at all points, the dependent variable or eld variable, in this case the velocity. In other words we were able to derive eld equations: equations that determine the eld variables. This feature, deriving differential equations by evaluating the limits of balances as the volume of the differential control volume goes to zero, is the constant refrain of this chapter. This is the technique we use for specic cases in this chapter. As we shall see later, we can derive general equations that can be suitably simplied for specic cases.

I  A

(12.1)

79

12.1.4 Momentum balance


Now we derive the equation of motion by applying Newtons second law to the system. Then we obtain a vector equation and hence it has three components. Since the T direction, however, is of no consequence, we have effectively only I two components. We consider the component rst since it should be easier to handle as is zero. We use the following form which was derived in chapter 6.

Rate of Sum of Rate of output Rate of input of accumulation of component of of momentum + to momentum in = momentum forces acting on the CV from the CV the CV the CV The left hand side is zero at steady state. The rst and second terms on the right hand side are , we construct zero since the velocity is zero. To calculate the force acting on the face at M the outward normal (in this case the  dot product of the normal and the total stress j) and take tensor. The stress vector is i+ j + k. Since we need the we M  component, I0 O take the dot product of the stress vector with . As is zero, it j. It is found to be O M follows from the Newton-Stokes law that is zero. Hence the component of the force per is O . The sign, understandably, is negative since pressure is unit area on the face at M compressive in our convention, and hence will act in negative direction on a plane to which j is the normal. We perform similar calculations for the control surface at . The body force is equal to the mass of the control volume multiplied by the component of the body force in direction. All these results can be substituted in the momentum balance to obtain

From the discussion of boundary conditions in chapter 10, either velocity and or stress continuity conditions can be applied. There is no general rule to tell us which of these boundary conditions can be applied to get a result that is useful. However, a feel is developed to nd the useful ones by experience, and it is here that solving simple problems is of help. In the present problem, a condition to determine the pressure is needed. Hence, the boundary condition that requires normal components of the stress be continuous can be expected to be useful. Application of this condition at A will give no useful result since that requires knowledge A of the normal stress in the wall. At the boundary at , however, the pressure in the gas phase is atmospheric3 , and application of the condition gives the following result valid at all F values of :
3

 A y P R 5 M O O Dividing by and taking the limit, we get the component of the momentum balance or the component equation of motion:  A (12.3) O P R This is the well-known equation of hydrostatics. We can easily integrate this with respect to to get A P R M  O where is an integration constant that can be a function of since we integrated with respect to . We have to apply boundary conditions to determine the integration constants.

We neglect hydrostatic pressure variation in the gas since density of gases is small.

A M O
80

In writing the above, it was assumed that the gas above the uid is not in such a motion as to generate normal viscous stresses there.4 This is a good assumption here, but in general such assumptions have to be veried. This assumption allows us to equate the normal stress in the gas phase to atmospheric pressure. Further, from the Newton-Stokes law

Hence we get the result that

Using this boundary condition, we get

A O A O

Thus, pressure is not a function of and the variation is simply hydrostatic. This is justiable since there is no motion or acceleration in the direction. I[ Now we go on to apply the momentum balance. Taking into account the fact that is not a function of , we get

I  A L_ A F for all at A

  M PR

l P R ,8 y y > M M O O Taking the usual limit after dividing by , and noting that is not a function of , we get O  P o A M R This equation is the component of the equation of motion. It reects our understanding that  A

the ow characteristics are determined by the balance between the viscous and the body forces. proceeding to solve this equation, let us examine the terms obtained after dividing Before by . The rst term is y

The numerator is the -component of the net viscous forces acting on the element. The denominator is the volume of the element. Thus this term signies the net viscous forces per unit volume acting on the element. Similar analysis shows that the other term represented the body forces per unit volume.

12.1.5 Combining rate law with balance law


So far we balanced momentum. In the beginning of these lectures we emphasized that the balance laws have to be combined with the rate laws. Thus, to nd the velocities, we have to use the rate or constitutive law in the momentum balance. The constitutive law relates the viscous stresses to velocity gradients. Substitution of the rate law into the balance equation then gives a differential equation for velocity. For this ow the Newton-Stokes law gives

 I0 A _ 

Substituting into the momentum balance we get

 K I PR o  A K _
81

(12.4)

We will return to this point later.

12.1.6 Scaling
Dimensionless groups are familiar to all chemical engineers. It is very often advisable to work with variables which are dimensionless. Thus quantities are scaled with respect to some quantity which has the same dimensions. To reap the full benets of the scaling, such a quantity must be an approximate estimate of the variable which is to be nondimensionalized. Thus a suitably nondimensionalized variable has a numerical value of the order of unity 5 . Often it may be necessary to guess groupings of physical parameters of the problem which give estimates of the dependent variables. Such guesses may have to be modied at a later stage if some defects are found. Experience gained in solving idealized problems, like the ones being considered in this chapter, is of value in making guesses of scales. We can nondimensionalize the coordinate by a characteristic length of the problem. The obvious quantity that can be used is the lm thickness. Obviously the nondimensionalized coordinate ranges between 0 and 1 and hence is of the order of unity. There is no such apparent quantity for velocity, and it has to be guessed. We know that the velocity is a result of the balance between gravitational and the viscous forces. We have shown while deriving forces ^ represents the momentum balance that viscous forces per unit volume. Equivalently, _  KI ^  K also represents the same. Let I be the characteristic quantity we I are looking for. We can therefore estimate that the viscous forces per unit volume in terms of are given by

4 4

I K _

Such an estimate follows from the fact that we have a characteristic length and from a simple P interpretation of a derivative. The body forces per unit volume are given by R o . Equating these two, we get an estimate of the characteristic velocity:

Thus we dene the non-dimensional quantities

I0 I A A I c KI K M 6 A 

KP o R _ 

Hence, in terms of the dimensionless variables, the equation of motion becomes

12.1.7 Boundary conditions


Since this is a second-order differential equation, two boundary conditions are needed to solve it and determine the velocity prole. This becomes clear when we integrate the above differential equation: 6

K I A L M g M K
or equivalently

Thus two conditions are needed to determine the two constants appearing in the above equation. The no-slip at the wall gives one boundary condition:

I A 

I A 

at

A 

This procedure has relevance to numerical methods of solution also.

82

Applying this gives K = 0. We need one more condition. It can be seen that the velocity continuity condition at A , though valid, is not of much help since we do not know the ow in the gas phase. The other possibility is to apply the component of the stress boundary A condition at to get Neither is this of much help since we do not know the stresses in the gas phase! Here we use l physical intuition. is the drag exerted by the gas on the interface. We say that it is drag since we expect the gas far away from the liquid lm ( ) to be stationary. However gases are a thousand times less viscous than liquids. Hence we expect this drag to be nearly zero 6 . Thus we have the following boundary condition

c /

c / A c

 I0 A _  A  I  A 

In terms of non-dimensional quantities this works out to be at

at

A 6

We apply this condition and nally get

I A 6 6 6 K L

>

The relationship between the velocity and the coordinates is referred to as the velocity prole. The above velocity prole is parabolic in terms of distance from the interface.

12.1.8 Volumetric ow rate


In terms of dimensional quantities the velocity prole is given by

K I0 A K P R ,8 6 6 L_ 

>

(12.5)

Note a few important things about the velocity prole. Firstly, velocity is zero near the wall, as expected from the no-slip condition. Secondly, since the gas liquid interface experiences no viscous drag, we expect the velocity to be maximum there, and the velocity prole conrms this. Since the gas phase exerts no drag on the liquid, the liquid is free of shear and such shear-free interfaces are referred to as free interfaces. Typically, gas-liquid interfaces are free interfaces wherever the gas phase is stationary far away from the interface. We are now ready to calculate the ow rate as a function of the lm thickness, which was our original objective. Since the velocity is not a function of , the ow rate is constant at any value of and is calculated as follows. Consider volume of Figure 12.1.
the differential control
The volume of uid owing through any area is given by v.n where n is the unit normal to the area. Hence the volume of uid owing at , per unit width of the channel, through the
Stress depends on viscosity and velocity gradients, and we still have to determine that the velocity gradients are not very large in the gas phase. We will get you to resolve this issue in a problem in a later section.
6

83

little area is given by with respect to :

I

. Hence the total ow rate

; A

I  P A R _,8

is given by integrating the above

We can dene an average velocity so that when it is multiplied by the area we get the ow rate. In fact, it is this average velocity that we use in Unit Operations after assuming that uid ows everywhere with the same velocity , i.e., that the ow is plug ow. As the area for ow per unit width of the channel is , the average velocity is given by

PR K, 8 _ I Note that the guess of characteristic velocity is three times that of the average velocity and I0 A
two times that of the maximum velocity. Not a bad guess at all! The velocity prole is often conveniently written in terms of the average velocity:

I0 A 6 I m9
12.1.9 Looking back

6 6 K

Equation 12.5 does make qualitative sense. Since the ow is driven by gravity, when all the other properties are the same, we expect a uid that is more viscous to ow less fast. Similarly, since the ow is being driven by gravity, when all the other properties are the same, we expect a uid which is denser to ow faster. The particular features cannot be guessed like that. e.g., the maximum velocity is 1.5 times the average velocity, ow rate depends on the cube of the lm thickness. These form the bonus for our effort. Notice also that the maximum velocity is 1.5 times the average velocity, as expected since some elements move slowly.

12.1.10 Flow transition


We assumed that ow is one-dimensional and that the equations of motion gave the solution which we derived. In this solution, the uid elements have only one non-zero component of velocity, and, as they move, their velocity remains constant. Thus their paths are steady and smooth lines, in this particular example, the paths are also straight lines. A ow with smooth path lines is a laminar ow. The mathematical solution obtained shows that laminar ow is possible. Hence it should exist under suitable conditions, and should be observable experimentally. It is possible, however, that the assumptions are not valid for some conditions of experiment and, under those conditions, laminar ow predicted by the solution would not be observed. In the present case also, as the ow rate increases, or as the Reynolds number increases, the ow becomes unstable and waves develop on the surface. When waves grow, velocity in the direction is also non-zero. As we did not allow for such a possibility, our

84

solution is not capable of predicting it. Clearly, the solution is not valid at high Reynolds numbers. We dene the Reynolds number as

25 and waves develop when 25 1000. The ow The ow is stable when becomes turbulent at Reynolds numbers higher than 1000. It turns out that linear stability analysis can be used to predict that falling lms do become unstable and that wavy ow develops. We considered this simple ow in great detail so that all considerations in solving for velocity proles could be easily brought out. The other problems in this chapter also deal with simple ows. In discussing those, some details will be skipped and it is expected that the reader will think about them and ll the gaps.
EXERCISES 1. The lm thickness is shown to decrease with length in Figure 12.1. Is it accurate or is it an artistic license? 2. Explain the directions stress terms in Figure 12.1, i.e., explain why the stress marked on the points in the direction while the stress on the face at points in the on the face at M negative direction. 3. Ihdnag Associates are proposing to clean up Ulsoor lake using a device shown in the accompanying gure. It is a at belt (similar to a conveyor belt) that picks up a thin lm of liquid of thickness as it moves through the liquid. The belt is coated with a catalyst. The liquid lm, while it is carried by the belt, is exposed to the UV component of sunlight and hence the pollutants in it are oxidized. To make the exposure time large, the length of the belt is large compared to . To process large volumes of liquid, the width of the belt (in the direction perpendicular to the paper) is also large. a. Calculate the ow rate of the liquid, per unit width, carried by the belt. Flow may be assumed to be laminar. b. Calculate the minimum power requirements, per unit width, for the device.

I _ P

Direction of gravity Liquid falls back into tank Liquid level in tank

Angle =

Liquid taken up

Figure 12.2: Figure for problem 3

85

12.2 Laminar ow in a pipe


12.2.1 Problem statement
Consider steady laminar ow of a Newtonian and incompressible uid in a pipe. We assume that the uid enters the uid with uniform velocity and pressure. Let the pressure at the exit end of the tube be lower. If this ow has been occurring over a long time, we expect steady state to prevail. As discussed earlier in chapter 10 this ow is driven by a pressure gradient. A fully developed ow, where uid elements move with constant velocity, can exist when the viscous forces and pressure forces balance each other. We therefore expect the ow rate to be dependent on the pressure gradient. Our objective then is to calculate the steady ow rate as a function of the pressure gradient in the fully-developed region. As with the ow of a uid down an inclined plane discussed in the previous section, here also, the ow develops fully at some distance from the entry. We also expect disturbances at the exit, and hence the axial velocity in the exit region may change with distance. These are known as the entry and exit effects. Thus, experimentally speaking, the fully developed ow is observable only in the central section of a long tube.

12.2.2 Simplications and assumptions


The pipe is shown in Figure 12.3. We use the cylindrical coordinate system as the pipe boundaries are naturally described in this system.

R r z=0 lies in the fully developed flow flow zone

p z rz g

rz z

z Direction of gravity

p z+ z

Figure 12.3: Fully-developed laminar ow in a pipe. We want to rst I decide which are the nonzero velocity components. We, of course, expect the axial velocity, , to be nonzero since pressure gradient is in that direction. We assume that the pipe is perfectly circular and smooth. Thus there is no reason to expect the pipe walls to induce any angular velocity. Further, the uid is entering the pipe with a uniform axial velocity, and does not have any swirling motion. Hence, we expect that the angular velocity,

86

the circumference of the pipe are indistinguishable. Thus, we do not expect velocities to vary in the v direction. Similarly, as the ow is fully developed in the axial direction, the velocities will not vary in the T direction. We can summarize the postulated dependencies of ow as follows:

t , is zero every where in the pipe7. As the pipe is smooth and perfectly circular, all points on A c v A 

12.2.3 Mass conservation

4T

Consider the differential control volume shown in Figure 12.3. It is annular with length T , inner radius p and outer radius p M p . Note that, the cross-section is a full circle, and such a control volume is not suitable for analyzing v variation. But, this is suitable for the ow postulated. Mass balance over the control volume gives

I I I I T P L ~ k p T r~ r#yr  M P L k~p T r~ ryr  M L kp p L k~p p y A  I As the ow is fully developed, p  at T and TM T are the p same,  and the last two terms cancel out. Dividing by T , and taking the limit a T , p I r  A  p
This equation can be integrated to nd

4I

The left hand side is a function of p while the right hand side is a function of T , and hence the only solution is that both sides should be equal to a constant. However, by no-slip condition, the p component of the velocity should be zero at the pipe wall:

p r A Z T 

p A I  Hence the constant is equal to zero and we nd that r A


at

I A  r

in the entire ow eld.

12.2.4 Momentum balance


The momentum balance in the v and p directions involves control volumes with curved surfaces and it is necessary to take a control volume that is differential in the angular direction also. It is cumbersome to do this in general even though it can be done for the present simple ow. Instead, the general equations to be derived later in chapter 13, are used directly in such instances. Hence, we do not consider momentum balance in these two directions now, and will simply state the results:

Op  Hence A O O T . Now we consider the momentum balance in the T


Rate of accumulation of T momentum in the CV

 A

4 4

Ov A

Rate of input of = T momentum to the CV

Sum of T Rate of output component of of T momentum + forces acting on from the CV the CV

(12.6) direction:

If the uid enters with a swirl, will it sustain or decay?

87

The term on the left zero at steady state hand I K . The  rst two terms on the I right L k~phand p , side  IUK is o  ,   P L P side are given by and kp p y , respectively. Since T is not a function of T , the two terms cancel each other. Hence, the momentum balance states that the sum of T component forces acting on the control volume is zero. This gives

c 4 is due to a balance between viscous 4 that the ow This equation, as was stated earlier, suggests 4 4 forces, body forces and pressure forces. We integrate this once to bring out the physical aspects better:  A p P p c r# g OT L M R L M M p 4  where g is a constant of integration. 4 As p , the shear stress tends to innity, unless g is zero. There must be a physically realistic solution for the problem posed, and none of the
physical quantities are expected to blow up (i.e., not go to innity). Hence,  A Thus we see that the viscous shear stress goes to zero at p :

 A L ~ k p p o O O y  M L k~p T r  r#yr L k~p T r  r M L k~p p l T P R where R is the T component of the acceleration due to gravity. For the choice of the direction of T coordinate shown in the gure, R is equal to R . Dividing by p and T , taking the usual limit, and rearranging the resulting equation slightly, we get 6   A P O T M RM p p p r#

c r# A

must be zero.

at p

A 

and is maximum at the wall. This is expected since the wall is the source of friction. The above equation suggests, however, that the shear stress is linear with p , and that could not have been guessed. There is an interesting point to be noted here. p = 0 is a line and not a physical surface, and yet a boundary condition is needed. This situation is typical of axisymmetric ows. In the present ow, since velocities do not depend upon v direction, the ow is symmetric around the T axis, and hence, the ow is axisymmetric. In such ows, usually, the distance from the axis, like p here, is used as a coordinate, and a boundary condition is needed for p also. The requirement that nothing should blow up at the origin, is one boundary condition that is commonly employed. The ow is symmetric around the axis and, often, this also implies that the derivative with respect to the distance from the axis should be zero. This could be used as a boundary condition. Returning to the equation of T momentum balance: it can be put into a simpler-looking form by taking advantage of the fact that there is no free-interface to impose any specic pressure boundary condition. Hence we can combine the rst two terms on the left hand side by dening P RUT . The equation then simplies to a new variable A

 A

4 4

pL M r# T

12.2.5 Rate law combined with balance law


The above is the momentum balance and we have to substitute the rate law or the constitutive relationship into it. Substituting, we get

I 6 p p A _ T L

88

(12.7)

Note that p A at p A . As discussed earlier, this condition, instead of the requirement that all physical quantities must be bounded, could have formed a boundary condition.

I ^

12.2.6 Scaling
We now try to estimate characteristic quantities to nondimensionalize the equation. It can be shown that the pressure gradient is constant when the ow is fully developed. Thus

P T WO M R W where is the pressure drop over length W . Now we can select to be the characteristic O hence p A p ^ . Following the argument given in the previous section length and I A K _ W I I ^I . Substituting this, we get and A I 6 A p Lp After integrating, we get I A 6 K p M K X

12.2.7 Boundary conditions


We have the no-slip boundary condition at the wall:

I A 

at p

or in dimensionless quantities

Hence

This is the well known parabolic velocity prole. The velocity prole and the stress prole are plotted in Figure 12.4.

I A 6 6 K p X

I A 

at p

A 6

Stress

Velocity

Figure 12.4: Stress and velocity proles in fully developed laminar ow in a pipe.

89

12.2.8 Volumetric ow rate


In terms of dimensional quantities, the velocity prole is given by

I A K _ X W

K p

We are now ready to calculate the volumetric ow rate. Considering the differential annular control volume, the ow rate through the annular area is given by

Hence the volumetric ow rate

I qum L k~p p A L kp p

; A

is given by

Note the sensitive dependence on the radius of the This is the famous Hagen-Poiseulle law. pipe. At constant pressure drop, if the radius of the pipe decreases only by 10%, the ow rate decreases by 40%. That perhaps explains why doctor Poiseulle was interested in this relationship. We dene the average velocity as the volumetric ow rate divided by the area. It is given by K
The velocity prole is usually written in terms of the average velocity prole:

, L kp I  p A k  _ W 

I A _ W

I A L I

6 p K

12.2.9 Looking back


Let us see what we can expect from our physical understanding of the problem. (1) The velocity is zero near the wall by the no-slip condition and the frictional forces are zero at the center. (See the stress prole.) Therefore we expect that the velocity should increase from the wall toward the center and reach a maximum at the center. (2) As the ow is driven by pressure gradient, the average velocity, and hence the volumetric ow rate, should increase when pressure drop increases. (3) When length of the pipe is increased while keeping the pressure drop constant, the magnitude of the force driving the ow, the pressure gradient, decreases, and hence, the ow rate decreases. (4) The drag forces which oppose ow increase when viscosity is increased, and hence the ow rate should decrease with increasing viscosity. (5) The pressure gradient acts on the cross-sectional area of the pipe while the drag forces act on the circumference of the pipe. The former is proportional to the square of the radius while the latter is proportional to the rst power of radius. Hence, while keeping other things constant, as the radius of the tube is increased, the velocity gradient should increase to keep the pressure and drag forces in balance. Thus, if radius is increased, the ow rate should increase. All the above expectations are conrmed by the equation we derived relating the pressure drop to the average velocity. The details could not have been guessed, e.g., velocity prole should be parabolic, fourth power dependence of ow rate on pipe radius. Another detail worth noticing is that he increase in ow rate with the increase in pressure drop is linear: this is typical of laminar ows. 90

12.2.10 Friction factor


The results of investigations of the kind just described are usually summarized in terms of drag coefcients for ow past a body or in terms of friction factors for ow through a duct. In either case, the denition is the same. They are dened using the following logic. We expect the drag force ` to increase with the area
on which friction is exerted. Further we expect the drag force to depend on the characteristic relative velocity ar between the solid surface and the uid. Later on we shall see that in turbulent ows, which are by far the most common, the inertial forces play an important role, and the drag force is proportional to the product of the density and the square of the relative velocity, or

K `
Par Hence it is customary to dene friction factor, Z , for ows through ducts as P a rK ` Z
L and the drag coefcient, , for ow past bodies as P a rK `
L

A k W and the drag force is k W where is the viscous shear For pipe ow
8 stress at the wall. From the equation we derived for the viscous stress, we know that

1 c

Pa K ` r A Lk W Z L ] Rewriting in terms of pipe diameter and rearranging, A ]W P a r K L XZ ] where is the diameter of the pipe.9
But

A L W A L W

Let us derive one more formula which is familiar to all. The above equation can be rearranged to get ] L

We can choose the characteristic relative velocity to be the average velocity. If we substitute our results of velocity prole we get

Z A XW

Pa K r

where is the Reynolds number. In Unit Operations books we nd a log-log plot of Z vs. The line with a slope of 1 corresponds to the above equation.
8 9

6n 6 ] _I P n A Z 'z

We use the magnitude so that the coefcients will have positive values. This formula should ring a bell in your mind if is equated to the average velocity.

91

12.2.11 Flow transition


We assumed that the only non-zero velocity is in the axial direction and this is true only in laminar ow. It is well known that laminar ow in a pipe is observed to give way to turbulent exceeds 2100. As it turns out, this value of the critical is also known to ow when increase in experiments where disturbances such as mechanical vibrations etc are reduced. as high as 20,000. One of the outstanding problems in Laminar ow has been observed at uid mechanics is to predict the observed critical value of the Reynolds number at which the transition occurs. We repeat our earlier comments on lm ow. Existence of a solution for laminar ow shows that under some conditions, namely, at low , it is possible to experimentally observe the velocity proles and verify the theoretical predictions, and that the solution is not valid at high . EXERCISES 1. Using the fact that the ow is fully developed show that the pressure gradient is constant ^ or that pressure falls linearly with length. This justies the use of W in Unit Operations O instead of pressure gradient. 2. How is it that we solved a rst-order differential equation here to get velocity prole while we solved a second-order differential equation in the previous section?

12.3 Stratied ow of two immiscible uids in a channel


12.3.1 Problem statement
We consider two uids owing adjacent to each other. The geometry is shown in Figure 12.5. Since the two uids are present as two layers, they are said to be stratied. We apply a pressure
y x bII Fluid 2 x= 0 marks the beginning of fully developed flow zone Interface between the fluids y yx x yx

bI Fluid 1

Figure 12.5: Stratied ow of two immiscible uids. gradient in the direction. If the uids have been owing in the channel for a long time, we expect steady state to prevail. Assume that the channel is innitely wide in the T direction, i.e., in a direction perpendicular to the paper. For the same reasons advanced in the discussion on pipe ow, we expect the ow here as well to be fully developed when the pressure forces balance the frictional viscous forces. We want to calculate the steady ow rates of the two uids as a function of the pressure gradient in the fully developed region. We assume that both uids are incompressible and Newtonian.

92

12.3.2 Simplications and assumptions


We select the Cartesian system because it is most suited to specify boundary conditions in this problem. We wish to determine which components of the velocity vector are I A nonzero.  , and As with the falling lm we do not expect any ow in the T direction and hence that velocities do not depend upon T . Since we are interested in the fully-developed region, velocities are also not functions of . The arguments given are valid for both uids. We use 6 L subscript and for uids 1 and 2 respectively. The simplications can be summarized as

4 4

I=  A c

4 4

I=  A c T

A 6cL

12.3.3 Mass conservation

Consider the differential control volume shown in Figure 12.5. Mass balance at steady state for such a differential control volume in uid 1 gives

each other. The usual limiting process gives

 A P g I g  I g  y  P g I g  I g  y  M I As the ow is fully developed, is not a function of and hence the rst two terms cancel

A similar procedure gives the following equation for uid 2:

4 4

4 4

I g   A I K   A

These equations can be integrated to show that

I g  A g  I K  A K  Z c Z

and from no-slip I  condition, I  the  components of the velocities of the two uids are zero at the walls. Hence, g A K A .

12.3.4 Momentum balance


We can balance momentum in the and T directions. We assume that gravity is in the direction, i.e., R A R . Since there is no ow in these two directions, it is easy to show that momentum balance in these two directions yields

 A

for both phases. This indicates that pressure variation in the direction is hydrostatic and that pressure does not vary in the T direction. Hence pressure in both uids is a function of only and . Since pressure variation in both uids is hydrostatic, we get

OT A O P R

gA

g P g R c O

and

K A

K P K R O

93

Note that g and g are functions of , and boundary conditions are needed to deterO As these O mine them. appear in the equations for pressure, it might be fruitful to consider the component stress balance at the interface between the two uids. Since the velocity is zero, l is zero in both uids. Hence the stress-continuity boundary condition gives

Thus, we get

Sum of Rate of Rate of input of Rate of output accumulation of = component of momentum to of momentum + momentum in forces acting on the CV from the CV the CV the CV The term on the I left is zero for The rst two terms on the right side K side K state. I steady   hand P are given by g g and P g g   y respectively and they cancel each other since the ow is fully developed. As before, since the ow is fully developed, i.e., the uids are not accelerating, the sum of the forces acting on the control volume must be zero. Thus, -momentum balance yields

Note that B is a function of . It should be so since the ow is driven by the pressure gradient O in the direction. Now consider a control volume in uid 1 and make a momentum balance in the direction:

gdA

B P gR c O

gdA

at

A 

and,

K A

B P KR O

c g  l 4 4 4 Substituting for g , the above simplies to 4 O  g c  A  B O M 4 4 and is equal to the pressure value at Once again, we emphasize that B is a function of alone O the interface. Finally, momentum balance in uid 2 gives c K   A  B O M 4 4  B ^ the These equations suggest that the viscous forces opposing  ow are balanced by the pressure
 A g O M
12.3.5 Rate law combined with balance law

The usual limiting process gives

 g  g    A g g y y M O O

is a constant, and the stress varies forces. It can be shown that for fully developed ow, O linearly with . This is similar to what was found in the case of ow in a pipe.

The constitutive relationships have to be substituted into the above equations before the velocity proles can be derived. Since both uids are Newtonian, the constitutive relationship for both has the same form  I

c l A

_  m

94

Since is not a function of , the derivative is not a partial one. Substituting this into the momentum balance  B  KI = 

 A  O M _ =  K c A 6 c L  ^ As mentioned earlier, the pressure gradient B is a constant since the ow is fully develO oped. Hence we replace it by the pressure drop divided by the length over which the pressure drop occurs:  B  O A W O
Hence the momentum balance equations become

 K I =  6 A K _ = WO c

A 6cL

12.3.6 Scaling

As the ow is pressure-driven, we can use the above equation to get characteristic quantities. The two uids have different characteristic lengths,  g and K . Thus we dene the nondimensional distances in the two phases, g and K , using the respective lengths in the two phases. Similarly we have two different characteristic velocities in the two phases given by

I g  A  Kg _ g W O I K  A  KK _ K W O I =A I I =   =

These quantities can be used to dene two non-dimensional velocities in the two uids:

where subscript has been omitted in the non-dimensional quantity for simplicity. Using these denitions we get identical equations for the two uids:

We can integrate these equations to obtain

I gA 6 gK L M g g M K

 KI=  =K A 6 c

A 6cL

I K A 6 KK L M K M The four constants have to be determined using four boundary conditions.


and

95

12.3.7 Boundary conditions


We have the no-slip boundary conditions at the walls of the channel. These give

I g  A  I gA 

In terms of non-dimensional quantities the boundary conditions are at

at

A g gdA 6

and

I K  A  IK A 

at at

A K

and

K A 6

We need two more boundary conditions. As mentioned earlier in the discussion on ow down an inclined plane, we have to look for boundary conditions at all boundaries and nd those that give results that can be used. Here, the only other conditions that we can think of at the walls are about stress continuity. However, knowledge of stresses in the walls is required to apply them, and hence they are not useful. Thus we turn to boundary at the uid-uid interface where no-slip as well as stress balance conditions can be applied. The no-slip condition at the interface implies that all three components of the velocities in the two uids have to be continuous. As only the component is nonzero, we have

I g  A I K 

at

A 

We also have the stress boundary condition. We have already used the normal stress boundary condition. The component of the stress boundary condition gives

c g  l A c K  l at A  In terms of non-dimensional quantities these boundary conditions translate to K _ g I g A  KK I K ?   g _ K & and  I g  I K  K  g  A  g  K 


I g  A WO I K  A WO  Kg L0_ g  KK L0_ K

These two along with the two obtained from the no-slip condition at the walls form the boundary conditions needed to solve for the four constants. The nal dimensional velocity proles are given by

_ g  g K _ g M _ K  g g_ K M

K M g_ K  M _ K K g

K K_ K V g g   g_  g H K  KK _  K K V  g _ K H

g K M g K M

 Kg _ K _  Kg _ K _

K 6 g g g K 6 g K K

12.3.8 Volumetric ow rates

The volumetric ow rates (per unit width) of both the uids can be calculated as follows:

; gdA

, m7 
96

;K A
They are given by

; gdA L _ Kg g ; K A L _ KK K

K g M g  K _ g K M WO H _ g  KK _ K Kg M _ O g K M W

, m  : H _ K _ K

M _ X _  g  g K V K g M _ X _  K g K V K g

12.3.9 Looking back


We see expected behavior in the above equations: decreasing velocities and ow rates with decreasing pressure gradients and increasing viscosities. An important way of checking the results is to look at the limiting behavior of the solutions, where possible. One such here is the limit of the solution as the viscosity of one phase goes to innity while the other parameters are kept constant. We expect the viscous liquid to move more slowly compared to the other since they both experience the same pressure gradient. Hence we expect, in the limit, the very viscous uid to be stagnant or to behave like a solid. Hence the velocity of the less viscous . Further, the velocity prole must be parabolic. Let us check to uid must go to zero at A see if the solution does show this. Thus

  

K g H I K  A  KK 6 H 6 M V  L _ V M O  K  K K K W

A 

K I g  A  Kg H W O L_ g  g  g V  This, as expected, shows that velocity is zero both at A and A  g


and in the same limit

prole is parabolic.

and that the velocity

12.3.10 Flow transitions


Pressure-driven ow of a single uid in a channel becomes turbulent when the Reynolds number based on channel height exceeds 2000. The stratied ow is also subject to other insta bilities because of surface tension and large shear rates that may prevail at the interface. These two effects give rise to waves on the interface. That is how we see waves on the water surface when wind blows over a lake. EXERCISES 1. Can we use the ideas of the present problem to justify the stress-free boundary condition used in the falling lm problem?

12.4 Conclusion
Three prototype problems were discussed in detail in this chapter. They illustrate (i) physical arguments, (ii) shell balances to derive equations of mass conservation and momentum, (iii) 97

combining these with rate laws to derive equation of motion, (iv) deriving boundary conditions, (v) solving the equation of motion to derive the velocity proles, and (vi) rationalizing the results with expectations. The ow instabilities relevant to each ow have also been mentioned. These steps, when systematically practiced, will help greatly in gaining physical insight into ows.

98

Appendix: Curvilinear coordinate systems 12.A Need for different coordinates


The rectangular coordinates are familiar to everyone and are easy to use as we have seen. Quite often, however, they are not the best choice for solving problems. It is known that in describing physical situations through solution of differential equations, boundary conditions have to be specied. Consider the familiar problem of determining the velocity prole in a tube when a uid ows in the axial direction in laminar ow. Here we have to specify the velocity of the uid at the tube wall to be zero. If rectangular coordinates are used to solve this problem, the boundary conditions are specied as follows:

where R is the radius of the tube. Clearly I this is not convenient. It would be more convenient if the condition could be specied as = 0 when radius is equal to R. In other words it would be ideal if a single coordinate (radius) could be used to describe the surface on which the boundary condition is to be specied. It is often possible to construct coordinate systems to do this. The spherical and cylindrical coordinate systems are two examples of the many possible systems10 . They are shown below:
z

I A 

for all points

g c K 

which satisfy

K K K gM K A

Point P r z plane

r plane

y Unit vector in the z direction Unit vector in the direction r

Unit vector in the r direction

Figure 12.6: Cylindrical coordinates:

pcv$c T   p

c   v  L kdc   T

EXERCISES 1. Why does v not go beyond k in spherical coordinates ? 2. (i) What surfaces are obtained in cylindrical coordinates when (a) kept constant, (b) v is 5 r is p  and v g  v K m (ii) kept constant, (c) z is kept constant, and (d) z is kept constant but What surfaces are obtained in spherical coordinates when (a) r is kept constant, (b) v is kept    p m constant (c) is kept constant, and (d) v is kept constant but 3 Which coordinate system is best suited for solving problems where boundary conditions are to be specied for the domains described: (i) a domain contained between plates perpendicular to the plane of the paper; (ii) a duct with the shape of a circular sector; (iii) an annular space between two concentric cylinders; (vi) an annular space between two concentric cones.

10

Interested readers may refer to the book on Field Theory by Moon and Spencer.

99

z r Unit vector in r direction

Point P

Unit vector in direction

Figure 12.7: Spherical coordinates:

pcv$c   p

6( 

c   v  kdc  

 Lk

12.B Unit vectors in curvilinear coordinates


Any coordinate system should be able to deal with vectors, vector algebra, and vector differential operators. These operations in rectangular coordinates have already been dealt with. The equivalent expressions in other coordinate systems are therefore needed. These can be derived  6 L = A in a general way as follows. We reserve the symbols  c c for the rectangular Cartesian system. Thus, the usual rectangular coordinates c c T are respectively equivalent to g c K cl . The unit vectors in the rectangular Cartesian system are denoted by = c A 6 c L c . Thus the commonly used i, j, and k are respectively equivalent to g c K c . It may be recalled that the unit vectors = were shown to be the unit tangent vectors to the = line obtained by varying only while keeping the other two coordinates constant, and that the unit normal vectors to the surface obtained by keeping = constant are the same as the unit vectors = . We can use either of these interpretations to dene and relate the unit vectors in any curvilinear coordinate system to the unit vectors of the rectangular Cartesian coordinate system. A 6 c L c  . For example gdA p , K A v , We denote general curvilinear coordinates by A T can represent the cylindrical coordinate system. Every point in space can be represented by coordinates or rectangular Cartesian coordinates. Thus are uniquely related curvilinear = to and vice versa11 . For the cylindrical coordinates, this requirement can be stated as

Let q represent the unit vectors in the curvilinear coordinates. If the above relationships between = and are given, the unit vectors in the curvilinear coordinate system can be constructed as follows. If any single is kept constant a surface is obtained by varying the other two coordinates. This surface is referred to as the surface. By analogy with the rectangular Cartesian coordinates, the corresponding unit vector q is therefore the unit normal vector to this surface. It is also the unit tangent vector to the line obtained by varying only while keeping the other two coordinates constant, the line being referred to as the line. See Figure 12.8. Note in particular that q g at different positions are not parallel to each other. This implies that q in curvilinear coordinate systems are not constant vectors. It is not necessary that all of them are not constant. This is in contrast to the unit vectors in Cartesian coordinates.

A =

or

=A =  lc A 6 c L c

11

The relationships must be unique and invertible.

100

y 1 = constant lines. 1

increases from inside to outside 2 1 1

Ellipses are y 2 = constant lines. y increases outward


2

Figure 12.8: General coordinate lines. For clarity shown for only two dimensions.

Consider a surface on which some quantity is constant, e.g., an isothermal surface. This can be represented as  =  A constant If any two points are separated by by

= , the temperature difference between those points is given A A =

or q x, is tangential to the isothermal surface. The dot product between this vector and the gradient of temperature is equal to zero, hence   must be normal to the surface. This result can be generalized. Given any function  = ,  represents a vector normal to the surface on which  remains constant. Applying this idea to the relationships dening the curvilinear coordinates, is normal to the surface and hence is parallel to the unit vector q . Hence

= =  = m A   mq If the two points lie on the isotherm, then  is zero and we have   m@q A  Since the two points are close to each other, the vector

 = =

)4 4

bb

A A

b b

= = = K H = =V

12.B.1 Scale factors

If we travel in the direction of q for a short distance , then only changes. Let us de note this change by . The curvilinear coordinates need not have units of length at all (e.g.,

4 4

101

H A vector is formed when we travel along the direction of q such that changes by little . The little vector is expressed as q in curvilinear coordinates. Let = be the Cartesian coordinates of the same vector. Then, A b  m = = = = , are tangential to the  line or normal But the little vector q , and equivalently to the surface. Hence b  and = = are parallel to each other. Hence A b  = = But = = is the length of the little vector and is equal to . Hence A H
b By comparison with the equation dening , we see that 6 A b or K H V A 6 K = 4 = 4 for the unit vector is nothing but the recipThus the denominator in the expression we derived rocal of the scale factor. Hence, = = q A b A = 4 4 We can now rationalize this equation. A unit vector is basically a ratio of two lengths: one corresponding to a length in a particular direction and the other are its components along the Cartesian coordinates. In curvilinear coordinates, changes in the coordinate do not represent changes in length. The scale factors allow calculation of changes in length when coordinates are changed, since the quantity is the change in length when  is changed by a small amount. It is for this reason that they appear in the above equation. The relationships above can be used if  are known explicitly as a function of = . When = are given as an function of  , the following procedure is better suited. explicit Consider the vector . This is a little vector in the direction of q . It can also be q = = = = rectangular Cartesian coordinates represented by where are the changes in the corresponding to the changes A . See Figure 12.9. The = can be computed from the rela A
102

in cylindrical coordinates). To obtain distances traversed when a curvilinear coordinate is changed, we need to dene scale factors which convert the change in the curvilinear coordinate variable to a length dimension. Thus, if is the distance traversed when a curvilinear coordinate is changed by , the scale factor is dened as

y vector
x
2

x2 vector x1 vector

Figure 12.9: Representation of a differential vector in curvilinear and Cartesian coordinates. Only two dimensions are shown for clarity. tionship = c changed. Hence

 A 6 c L c . In forming the little vector tangential to the = is given by =A = A 6 c L c

line, only

has

Therefore

If the new coordinates dened are orthogonal to each other, the following equation must satised:

4 = 4 A 4 4 The expression above can be alternatively interpreted as follows. q is the unit tangent vector = = = to the line. If is a little vector tangent to the  line, and is the length of the little vector, then 6 A = = q = 4 4 is the unit tangent vector.
A q mq A

4 q q

= = = 6 = =

We again emphasize the fact that the unit vectors in curvilinear coordinates are not constant. The expressions also indicate that the unit vectors direction depends upon the locations since

4 4 4 4 4 4 ^ can be position dependent. and = This raises complexities in summing vectors and in nding 4 4 the components of the Laplacian operator, gradient of vector etc. The expressions for the unit vectors in cylindrical and spherical coordinates should be examined thoroughly to
understand this point clearly. 103

q A = =

H 6 = V

EXAMPLE 1. In spherical coordinates we have the relationships that specify Cartesian coordinates as functions of the spherical coordinates as well as the reverse. They are

 c K A p v c A p o8 v gA p v o and K K K  A Kg c p A g M K M c o v A K K K g M K M From these relationships we can see that if we travel p in the p direction, the distance covered is equal 6 . If we change v by a small quantity v while keeping to p itself. Hence yr A is given by p v . Hence t is equal to all other coordinates constant, the distance covered while keeping all other coordinates constant, by a small quantity p . Finally if we change the distance covered is given by p v d . Hence is equal to p v . Summarizing sr A 6 c yt A pc A p v

  ( H

H (

 ( (

The expressions above can also be derived using the formulae given earlier for the scale factors. Now using the expressions derived for the scale factors we can express the unit vectors of the spherical coordinates in terms of the unit vectors of the rectangular Cartesian coordinates:

(   g g M o 8 v   (   K K M , 8 v   ( g Mo8  v K ( v A  q ( M ( We can easily invert them using the formula =A qsrnm =  qsr M qytnm =  qst M q m =  q2 Thus g A  v o (   qUr M o8 v ,8 (   qyt  (  q K A  v   ( qsr M  o8 v  ( qyt M ,8 ( q A ,8 v qsr  v qyt
qr A y qyt A v o8 o8 v ,8 

12.C Vector algebra in curvilinear coordinates


Vector algebra in curvilinear coordinates is done in exactly the same way as in rectangular Cartesian coordinates, i.e., by expressing the vector in terms of its components and using the orthonormality relations between the unit vectors. A more complex example of vector algebra is taking dot products between vectors expressed in different coordinate systems. Consider, for example, the conversion of components of a vector from one coordinate system to another. Any vector may be expressed as follows:

I A = = =A I = q m
or

Therefore

I =A

I A

q
I

I= =  m q =

EXERCISES 104

1. The cylindrical coordinates are specied by pcv$c T p o v$c K A p vUcl A T , a. nd the scale factors, ; b. express qyrncGqstncqu as functions of = c. check whether this system is orthogonal; ^8L and at r= 1 and v d. plot the unit vectors at r =1, v A k 2. The spherical coordinates are given by g5A p v ,8

Z g c K cl

where

g A

3. 4.

5. 6.

A kJm c K A p v c A p o v$c a. check whether the system is orthogonal; 6 cv A k ^8L m A k ^L and at p A 6 cv A k ^ c A  m b. plot the unit vectors at p A = . Express = in terms of q X In the text, q were expressed in terms of I I I A vector v has g c K c as components in the rectangular Cartesian system. Calculate the components in the cylindrical and spherical coordinate systems. Express these in terms of the corresponding coordinate variables. = =8=A q Hint for 3 and 4: Any vector can be expressed as A Find derivatives of q with respect to in cylindrical and spherical coordinates.  Any surface can be described by an equation of the form of A Z g g c K . Construct a unit normal to such a surface in the rectangular Cartesian, cylindrical and

J(

 ( 6 (

spherical coordinates.

12.D Tensor algebra in curvilinear coordinates


Following the idea of a dyadic product, the second-order tensor in curvilinear coordinates can be dened as

The algebraic operations on tensors in curvilinear coordinates can be carried out exactly in the same way as was done in Cartesian coordinates. The algebraic operations can be used to evaluate components of one tensor in one coordinate system if they are given in any other system following the examples given for vectors. Consider the following example.

q q

= =3A q q   The relationship between  and  = can be determined as follows. Taking a double dot product with , we get  A = =  = A q q   A mq ~ m@q   Since the conversions between q and = are known, the dot products between the unit vectors

in the curvilinear and Cartesian systems are known, we can determine the above relationships completely. 105

EXAMPLE Express 

A A A

in terms of only spherical coordinates.

 gK

 K  K rr v ( ,8 ( M r#t  v o ( K o8 v  ( M r  v ,8 ( K K M tr ,8 v o K (  v  ( M t#t ,8 v o8 (  ( M t o8 v ,8 (Q}r  v  (  t o8 v  (Q}  ( ,8 (


g mq ~  K m q  

EXERCISES

1. Express  in terms of  = . 2. Prove that, if  = is symmetric, then 

is also symmetric

106

Chapter 13 EQUATIONS OF MOTION


We derive the general equations of continuity and motion. A quantity called substantial derivative appears there and we discuss its significance. We interpret the various terms that appear in the equation of motion. In chapter 12, the equation of mass conservation and equation of motion were derived for different situations. The procedure followed in each case was identical. A small control volume was chosen, mass and momentum balances were made, and differential equations were obtained in the limit of the volume of the control volume goes to zero. The procedure can be generalized, and partial differential equations can be derived by evaluating the limit of the momentum and mass balance on a three dimensional differential control volume as its volume goes to zero. This is the objective of the present chapter. The partial differential equations can be used as a starting point and simplied for the specic case in mind.

13.1 Equation of continuity

Consider a rectangular parallelopiped of dimensions c c and T as shown in Figure 13.1. Application of the law of conservation of mass to this stationary control volume yields
H D z y x z E x C F y D p G H C

zz

zy zx
G

Figure 13.1: A small three dimensional control volume

4 4

P I T A l P   I M P

I P  y s I  P  y U
107

TM

I  P I  s l P y T

Divide the above equation by T , which is the volume of the control volume, and take the limit as it goes to zero. The following result is then obtained:

mass.

P I  P I  A bm P  A P I  (13.1) 4 4 4 4T 4 of continuity 4 . It is4 equivalent to the law of conservation of This is referred to as 4 the equation P

13.1.1 Incompressible uids


If we assume that a uid is incompressible, then reduces to

This form of the equation is used very often since most uids behave like incompressible uids at the velocities encountered in common ows.

bm

A 

is a constant. The equation of continuity (13.2)

13.2 Interpretation of divergence of a vector P  P I8  that The term m has a nice physical interpretation. Consider the term T appears in the mass balance on the differential control volume. T is the area of the plane AEHD of Figure 13.1 and i is the outward normal to it. Hence the term is equal to m
on that plane where
is equal into of the control volume through I  to T . It is equal to the ow P that face. Similarly, y T is equal to m
on plane CBFG. Similar analysis on other terms shows they are all equal to m
on the other four planes. Hence the sum of all these terms, which appear on the right hand side of the mass balance on the differential P control volume, is equal to the net ow into of the control volume. The sum of m
 P m
, where the integral is being evaluated over the on all the planes can be written as surface area of the differential control volume. However, by divergence theorem this is equal to l   P m a where the integral is being evaluated over the volume of the differential control P  volume. However, since the volume of the control volume is small, we can assume that m is nearly over the P constant  P region  occupied by the control volume and hence we can equate it to m a . Thus m can be interpreted as the net ow rate of mass out of the control

sb b b b volume per unit volume of the control volume. Viewed in this way, the equation of continuity
makes physical sense. The left hand side is equal to rate of accumulation of mass per unit volume. The right hand side is equal to the net ow rate of mass into the control volume per unit volume of the control volume. They both should of course be equal according to law of conservation of mass. P Using the same logic, the divergence of any vector ux of a quantity (note that is the  mass ux) is equal to the net output of that quantity per unit volume. For example, m , where is concentration of some species, will be the net out ux of that species per unit  volume of a differential control volume centered around the point where m is being evaluated.

108

13.3 Substantial derivative

M mJb P A P bm 4 4 The left hand side can be rewritten as H 4 M mJb V P ^ M mJb can be also be interpreted. 4 The operator Suppose, changes in some quantity, ( are to be measured while with the velocity of the uid . The total change will have several 4 because 4 moving contributions the quantity can change with time and also with position. In a small time I interval the position being made would have moved by a I at which the measurements aredue vector given by . Hence the change in the quantity to change is position alone will be I l  given by Taylors series lexpansion and is equal to change in ( due to change in mJb( . The l  I l  I ^ ^ time is given by ( . Thus the total change is given by ( M mJb(2 . Hence, the rate of change of 4( as4 measured by an observer moving with the uid is given by 4 4 ( M m$bs( ] ] ( 4] ^ ] ^ 4 is a derivative with time obtained by an observer Hence the operator M m$b moving the uid. This is4 called substantial time derivative or substantial derivative for short. 4 Note the difference between the two derivatives with respect to time. The partial deriva $ P ^ tive, , represents rate of change in density with time at a xed location. The substantial P is the rate of change in density with time as measured by an observer moving derivative of 4 4 with the velocity of the uid.
1 2

The equation of continuity can be rewritten as follows:

13.4 Cauchys equation of motion


Consider a momentum balance on the same differential control volume shown in Figure 13.1. For convenience, only the surface forces acting on the plane CGHD are shown. Forces act on the other faces also. Balance of component of momentum can be written as follows. Rate of accumulation of momentum A The net rate of input of is given by
1

4 4 faces AEHD and CBFG momentum into the control volume through P II  P II  y T Y
w Y 
109

I P 0 T

An example of this is measurement of changes in the temperature of water in a river while sitting in a boat which coasts along the river. 2 Suppose the observer moves with a velocity v , and he measures the rate of change of some quantity, say density. He would nd it to be equal to

The operator

U.  w

is called the total derivative.

The net rate of input of momentum into the control volume through faces ABCD and EFGH is given by I0,I0  I0oI0  Finally the net rate of input of momentum into the control volume through faces ABFE and DCGH is given by I I0  I I  The component of the surface forces acting on the same six faces are respectively given by

>

and

The component of the body force is given by the momentum balance Rate of accumulation of momentum in the CV Rate of input of = momentum to the CV

O M c  y O M c  T c O M c l  y O M c  T c O M c  y P O M c  s
R T

. We substitute these terms into

c c c 4 (13.3) b O M M M T M P R 4 4 4 4 4 4 4 4 The above can be4 recast in the following form c c c I I   P P A 4 M bm 4 O M 4 M 4 M 4 T M P R (13.4) 4 of the above equation is equal 4 to4 4 4 The left hand side P ] I I P  P bm M ] 4 M 4 of continuity. Thus the component of equation of The rst combination is zero by equation motion can be written as I l P ] A O M c M c M c M PR (13.5) 4 4 4 4 T 4 component 4 4 of Cauchys 4 equation 4 of motion. No assumptions The above is referred to as the P I  A m P I 
were made in deriving the above equation, and only Newtons second law of motion was used. Hence, it is obeyed by all uids. The various terms can be given physical interpretation. The left hand side is product of density, or mass per unit volume, and the substantial derivative of the component of velocity. The latter is the component of acceleration measured by an observer moving with the uid. Hence the left side is the product of mass per unit volume times acceleration. Therefore, from 110

divide by T and take the limit as the volume of the control volume tends to zero. The following equation is then obtained:

Sum of Rate of output component of of momentum + forces acting on from the CV the CV

Newtons second law, it must equal the net forces acting per unit volume. The body force term is obviously equal to the component of the body force acting per unit volume since R is the component of the body force acting per unit mass. Pressure acts always normal to the area. Hence component of force due to pressure on the entire control volume can arise only from the planes that form the boundaries of the control volume. Hence, the pressure term in the equation of motion arises out of taking the limit of

as the volume of the control volume goes to zero. Hence, O is the net component of the pressure force acting per unit volume. Consider the viscous stress terms. For a small differential control volume,

y s T O O T

c M c l M c V a A 4 4 4T a 4 differential 4 control 4 volume. But by divergence theorem the right where is the volume of the hand side is equal to , m 3 R m 

H
i.e., an integral over the entire area of the differential control volume. But the integrand is the component of the viscous forces exerted on the surface of the control volume by the surroundings. Hence m ( .i) is equal to the net component of the viscous forces acting on the uid per unit volume3 . Hence the left hand side of the Cauchys equation of motion is then equal to the sum of component of the various forces acting per unit volume. This is of course how it ought to be since the right hand side is the the product of mass time acceleration per unit volume. Following the same procedure outlined above, the and T components of Cauchys equation can be derived. If these equations are multiplied by the respective unit vectors and added, the vector form of the equation of motion is obtained. It is given by

4 4 bm 3 R m l a

b R

P ]] A P m M M O

b R t

(13.6)

Note that the above is a vector equation and hence is equivalent to three equations. The three components of equation of motion contain several unknowns: six components of the stress tensor4 , three components of velocity, and pressure. The Cauchys equation of motion cannot be solved to nd the velocity proles. More information has to be given and that concerns the rheological behavior of uids. Most uids, as mentioned in chapter 9, obey Newton-Stokes law.

13.5 Navier Stokes equation


If the Newton Stokes constitutive relationship is substituted into the Cauchys equation of motion, the equation of motion is obtained for a Newtonian uid. If it is further assumed that the
One can alternatively interpret it as the net ux of the component of momentum into a differential volume per unit volume. 4 The stress tensor is symmetric for most uids we deal with.
3

111

uid is incompressible, then m can be set to zero. The Newton-Stokes constitutive relationship reduces for incompressible uids to

A L0_

The equation of motion for a Newtonian and incompressible uid is called Navier Stokes equation. Skipping the algebra, it is displayed below.

P ]] A _ K M P M O

(13.7)

It is a vector equation, it is equivalent to three equations. This equation has to be solved simultaneously with the equation of continuity. Note that there are three components of velocity and pressure, a total of four quantities, to be determined and there are four equations, three equations of motion and the equation of continuity. Thus the system of equations is complete: there are four unknowns and four equations. Any problem regarding the ow of an incompressible Newtonian uid can be solved if the above equation can be solved simultaneously with the equation of continuity m . The way rivers ow, the way the weather changes, the way a cricket ball swings, the way amoeba swim, every phenomenon, is contained in the Navier Stokes equation. As Feynman said regarding electromagnetic theory, Solve the Maxwells equations. That is all there is to it. However solving the Navier Stokes equations is a very difcult task as we will see later on. Hence to get experience and intuition, a few simple examples are taken up in the next chapter.

13.6 Stream function


One general simplication used to solve equations of motion for Newtonian incompressible uids is however discussed here. As mentioned above, the Navier Stokes equations have to be solved simultaneously with the equation of continuity or four equations have to solved simultaneously. Considerable simplication is achieved if the ow is two dimensional, i.e., there are only two nonzero components of velocities which vary as a function of only two spatial coordinates5 . Even for two dimensional ows, three equations have to be solved simultaneously. For such ows however the idea of stream lines can be used to reduce the number of equations to be solved from three to one. In chapter 11, streamlines have been dened as those to which the velocity vectors are tangential to. This denition can be given a mathematical description. Consider ow in and I directions. Hence, =0. The uid paths will be lines on plane. The and coordinates are related on any line. A constant value of a function of and describes a line. This idea is utilized in the following way. Dene a stream function, c . Each streamline is described by the stream function, , being equal to a constant. Different streamlines are obtained by changing the constant. However, the gradient of the stream function is perpendicular to the streamlines6 . But the velocity vector is tangential to the streamline. Hence,

m
5 6

Jbm

A I0

The ow could be a function of time. Recall our method of constructing unit vectors in curvilinear coordinates in chapter 12.

I0

4 4

A 

112

This restriction can be satised by dening the following relationships:

4 b K M 4 4 b K 4 4 b K A hb 4 is solved 4 4 to nd , the 4 velocity 4 proles are fully determined. Note Thus, if this single equation that the order of the differential equation for is greater than for the Navier Stokes equation. The discussion of the extra boundary conditions needed will be taken up at the relevant places. The stream function has great utility in visualizing ow elds. Refer to Figure 13.2. Consider two streamlines. Let the difference in the stream functions be . Consider two
8
P2 y O y P1 x

need not be solved! More reduction in the number to be solved can be achieved by indulging in some 7 algebraic gymnastics. Differentiate the component of Navier Stokes equation with respect to , the component of Navier Stokes equation with respect to , and subtract one from the other. The following equation is then obtained:

4 4 denitions 4 since There is a signicant advantage with these K K I I A  A M M 4 4 4 4 4 continuity 4 is identically 4 4 4 satised. 4 Hence, at least one equation In other words the equation of

I0 I c

Figure 13.2: Relation between ow rate and the stream function points g and K , one each on the streamlines. Let the two points be separated by and . g K Thus the vector from to is given by Mi . The area9 vector of this line g K , i.e., n .
,pointing is given by . Hence ow rate across this line is given by Flow rate across the line

A
7 8

I0 I

Rare in this notes. As it is said: there is no free lunch any where, and the reduction of the number of equations is achieved at the cost of increasing the order of equation to be solved. Nevertheless, it is a signicant saving in complexity of solving that it is worth to use the procedure. 9 The area is per unit height in the direction.

4 4

113

Since the velocity vectors are tangential to the streamline, uid cannot cross the surfaces marked by the streamlines. Hence, is equal to the rate of volumetric ow between the streamlines10 . This idea is utilized in visual display of ows. If streamlines are plotted at some constant interval of stream function, the density of streamlines reects the ow rate. Where the streamlines are densely packed, there is more ow. Additionally, arrows are marked on the streamlines to indicate the direction of ow. A map of streamlines gives an excellent visual display of ow phenomena. The stream function is widely used in computations and also in presenting results. The ow in the plane is called a planar ow. In this type of ow, there is no change in the direction perpendicular to the plane. Such ow can also be described in cylindrical coordinates, and stream function is dened as follows. Planar ow in cylindrical coordinates

I A  I A 6 I c r p v c t A

13.6.1 Axisymmetric ows

4p 4

Consider ow around a sphere. The ow can be expected to be symmetric around the axis parallel to the direction of ow far away from the sphere. If T axis is that direction, only the p and v components of the velocity are non-zero and they depend upon those coordinates only. It also is a two dimensional ow but is not a planar ow. Such ows are referred to axisymmetric ows. The stream function has to be dened suitably. The stream function dened in axisymmetric ows is some times referred to as Stokes stream function. The denitions are given below. Axisymmetric ow in cylindrical coordinates

I A  I A 6 I A 6 t c p p c r p T

Axisymmetric ow in spherical coordinates

and Lightfoot.

 4 4 of The corresponding equations of motion in terms

6 I A  I A I A 6 c r pK v v c t p v p

 4 are listed 4in the text by Bird, Stewart

13.6.2 Boundary conditions and stream function


Velocities are obtained from the stream function by differentiation. As a result, when the Navier Stokes equations, which are differential equations for velocities, are converted into differential equations for the stream function, the order of the equations increases by one. The number of boundary conditions available however are enough for solving Navier Stokes equations only. Thus, one needs more boundary conditions. These are dened by taking advantage of the fact that the relationship between velocities and stream function is not altered if constants are added to it. Thus, any number of constants can be added according to convenience. The choice depends on the problem being solved. Specic examples will be considered later on.
10

The sign convention adopted determines the ow direction and is arbitrary.

114

Appendix 13.A Equations of uid motion in different coordinate systems


Often, we have to solve uid mechanics problems involving complex geometries: spherical and cylindrical etc. We should therefore have Navier Stokes equations in other coordinate systems also. For convenience we list the three components of Equation of continuity, Navier Stokes equations, and stress-strain rate tensor relationships in Cartesian, Cylindrical and Spherical coordinate systems. Equations of uid mechanics for an incompressible Newtonian Fluid Equation of continuity in different coordinate systems Rectangular coordinates

4 4 4 4 b Cylindrical coordinates 4 4I 4 4 ] P P I P P I P t A ] b m A M r p M p v M T A P 4 4 4 4 Spherical coordinates 4 4 4 4 I ] P P I P It P P A ] b H m M r p M p v M p  v ( A P 4 4 4 4 4 4 4 4 A P p6K


T
Component:

] P ] A

H I I P I P I P A P P A M M T m M M

H 6 I  6 I P p p p r M p vt M

4 4

I0 M

4 4

V T

4 4

V T

  4 4 4 Newtonian4 Fluid Navier Stokes equations for an incompressible


H I I I K I I I0 I _ A M M T V O M M K M

6 6 I p KI r I t v  V M p v v M p v p

 4 ( 4

Rectangular coordinates

4 4 4 4 Component: H I I I0 P 4 M 4 M 4 4 Component: H I I I M M P 4 4 4 4
P

H I0

H KGI0 I I I I0 A _ M T V O M K M

4 4 4 4

4 4 4 4 4 4

KI0 K M

4 4 4 4 4 4

KI PR K S V M T KGI0 PR K V M T KI VSM P R K T

H KI I I I I A M _ K M V M O T T

4 4

KGI0 K M KI M K

115

Stress strain rate relationships

c =
A
Cylindrical coordinates

H I= M

4 4

A A = V c c T c c T

p P

Component:

H I

KI r M PRr 4 K 4 4 4 4 4 4T 4 II 4 I4 4 4 4 v Component: I I H I I I t M r t M t t M r t M I tV P 4 4 p p 4 v 6 p _ 4 H T6 I  6 K I t L I r K I t P 4 4 A 4 p Ov M p 4 p p p t VSM p K v K M p K v M T K M R t 4 4 4 4 4 4 4 4 4 4I 4 KI 4 KI T Component: H I I I I I I I H 6 P M r M t M V A O M _ 6 P 4 4 p p 4 v 4 T 4 T p 4p p 4 p V M p K 4 v K M 4 T K M R 4 rate4 relationships 4 4 4 4 4 4 4 Stress strain H I H 6 I H I I c rr A L_ 4 p r V c t#t A L_ p 4 v t M p r V c A L_ 4 T V 4c A c A _ 4 I t 6 I r 4 r#t tr p 4p p M p 4 v I 6 I t _ A A 4 c t c t# 4 T M p 4 v 4 I I _ A A 4 c r# c #r 4 p M 4 T4r 4 4

4 4

I I I I$K I r M Ir r M t r t M I rV p p v p TH 6 6 KIr L It p I r  A O M _ VSM p K v K p K v M p p p p

Spherical coordinates

  4 4 4 4 Component: 4 4 4 4 $ I K I K H I I I I I I I r M r r M t r M r t M V P  4 4 p p 4 v _ Hp K I v 4 ( L I pL I t 4 4 A Op 4 M b r 4 p K r p K v 4 4 4 4 116 b

H K H H K 6 6 K 6 v V M K K KV pK p p pV M pK v v v p v

4(

I L I L PRr o  K K t v S V M p p v

 4 ( 4

v P

Component:

H I

I I I I I r M r p M p v M p  v ( M p M t p , v V 4 4 6 4 _ H K I4 I I L L o v I t r 4 A 4 p  v 4 O( M b 4 p K  K v M p K  v ( M p K  K v ( V M P R 4 4 4 4 4 4 Stress strain rate relationships c rr A L_ H 4 I p r V  c t#t A L_ H p6 4 I v t M I p r V  c A L_ H p 6 v 4 I ( M I p r M I t op  v V 4 4 4 I I 6 c rt A c tr A _ p 4 p p t M p 4 v r I 6 Ir _ A A 4 4 c r c r p 4p p M p  v 4 ( I 6 It  v _ c t A c , t A 4 p 4 v  v M p 4  v 4 ( 4 4 P H I I I I t I

4 4

I I I I I I I t M Ir t M t t M t M r t p p v 6 p v H p L KI A O M _ t M pK p v

4 4

4( 4b

4 4

IK,  v I p r K v p V

 v I t K LK o V M PRt v p Kv

Component:

 4 ( 4

117

13.B

Vector calculus in curvilinear coordinates

As can be seen from the tables just presented, many complex terms appear in the equations of motion. They arise from differentiating vectors in curvilinear coordinates. Our concern in this appendix is to dene differentiation with respect to curvilinear coordinates, and determine expressions for various operators like the gradient, curl etc. The complications arise here since the direction of the unit vectors in the curvilinear coordinates vary with the location in space. The gradient operator in the rectangular Cartesian coordinates has been dened as

b b
=

= = =

The above still denes the gradient operator but expressions, equivalent to the right hand side in the above equation, have to be found in curvilinear coordinates. By using the chain rule, the above equation can be rewritten as

4 4

V =

However, in chapter 12, the following relationship was derived for the unit vectors in the curvilinear coordinates:

ator is given by

4 = 4 expression for the gradient operIf this is substituted into the previous equation, the following ator is obtained: 6 H b q 4 V 4 For example in spherical coordinates 6 6 A b qyr 4 p M qyt p 4 v M q p  v 4 ( 4 operators4 can be found as 4 follows. The Laplacian operExpressions for more complex vector
q A = =

4 V

qsr p M s q t p v M q p v V m qsr p M qyt p v M q p v V Clearly, derivatives of q with have to be evaluated. These can be found as indicated in one

4  4 ( 4 4  4( 4 4 4 4 4 4 of the earlier exercises. They are given for spherical coordinates by A  qUr A qst qsr A  vq y q r 4p A  4v A 4( A 4 4 p qut 4 4 v qst qs4r 4 ( qyt o vq 4 4 p q2 A  4 4 v q A  4 ( 4 q A  vq2 o vq 4 4 4
A

b H mJb

118

Substituting these results,

b Jb

m A
EXERCISES

H K H H K 6 H 6 v VSM K K KV pV p pV M pK v vV v p v

1. Find an expression for operator in cylindrical coordinates. The nal expression should contain only and q . K 2. Find an expression for the Laplacian operator in cylindrical coordinates. 3. Find an expression for in cylindrical and spherical coordinates.

4(

One has to be very in using the Laplacian operator on a vector. The Laplacian of a Kcareful  or vector is dened as m  . Thus a component of the Laplacian operator has to be obtained by the following procedure:

b $b

b K q m b K

The computation is tedious since it requires several cross derivatives. In spherical coordinates

spherical and cylindrical coordinates and tabulated in a very convenient form in the text by Bird, Stewart and Lightfoot. EXERCISE Find

42 r t 4 v 4 2 4 2 4 4 2 4  A  qst 4 4  42 ( 4 4 42 ( 4 K  t M q ,8 v t A    4 4 K ( 4k ( 42 ( 4 ( 2 4   4 4 4 v M qst o v  4 4 ( k 42k(  424 ( 4 4 Using the expression derived for the Laplacian and the above expressions, the p component of the Laplacian of a vector can be computed to be L r L t o v L t L K K K A A b r qsrnm b b rd p K p K p K 4k v p K  v 42 ( 4 been evaluated 4 for the Components of the commonly encountered vector operators have
A

4 K v 44 v K 4 4 ( 4 K( K 4 4

qyt r M qyt v t Sqsr t K K r r t qsr v K qsr r M qst v S qsr v M qyt v K t Sqyt t 8 v M q r v8q M qyt t M q t o qUr v M q v r M q v r r qsr v M qyt  v8qyt K  t o  , 8 , 8 K L qUr v t qUr v M qst v v M q M q K K M qUr v M q o8 v

qsr v r M qyt v r M qsr r M K qsr K r M

42 442 424 ( 4( 4 4

K  r

in cylindrical coordinates

13.C Tensor calculus


The calculus of tensors also follows the same lines as vector calculus. Consider some operations of tensor calculus. The divergence of a tensor is dened as the dot product of the 119

gradient operator and the tensor. Consider evaluation in Cartesian coordinates. It can be easily evaluated as before using the indicial notation.

L bm

= = = =

= m  =  =  = =

Note that while the divergence of a vector is a scalar, the divergence of a tensor gives rise to a vector. This is consistent with the rule mentioned earlier. With the concept of the dyadic product between two vectors, which produces a tensor, a new calculus operation can be performed on a vector to obtain a second order tensor:

b
EXERCISES Prove the following  identities m (i) m  A m M A 5: (ii) m :  A m m  (iii) m m

I= = =

Note the ordering of the indices carefully for this operation as it is a special case. Note the differences in the way indices have been ordered in this operation and the previous operations.

b b K Q L b Jb L L b  b b b

13.D Tensor calculus in curvilinear coordinates


The extension of these ideas to calculus of tensors in curvilinear coordinates can be carried out since it involves only operations upon unit vectors. A few examples are given in tensor calculus as illustration. Consider the gradient of a vector. Following the way this operation was dened in rectangular Cartesian coordinates,

qI q 4 4 is put at the end to preserve the order in which Note that q is not to be differentiated, but the unit vectors appear in the dyad to be consistent with the way we the unit dyad was used in dening b v in Cartesian coordinates. The following example makes this clear.

EXAMPLE Find the vp and pv components of nition

in spherical coordinates. Now from the above de-

120

v. Thus, it is equal to the sum of all terms of the right hand side of the above equation which are multiplied by the dyad qutqUr . The rst term clearly yields

I 6 I q q M q q V v p component is equal to the double dot product between qr3qut , and The 

4 4 4p 4
I I t

6 It A r p s

The second term has the following structure.

Hence

q qsr yr 4 p 4
I

A A 

HI

qr M It qt M I r y p p

4 qkp V 4

qsr

p 4 To nd the pv component we have to look for the q rqyt component of the original equation. 4 The rst term clearly gives 6 Ir 6 Ir A yt 4 v p 4 v The second term has the following structure 4 4 HI I qyr M I t q t M I qk V qyt q qyt A r (A13.8) v v v p 4 4 4 yt 4 v 4 A I rGqyt 4 I tqyr M 4 [ qyt 4
Thus this term contributes

 A tr

pt

I p r qytGqyt p qUrqyt p t

. Hence

 A 6 H Ir It rt p # v V p

Once again, various calculus operations on tensors commonly encountered in spherical and cylindrical coordinates have been tabulated in a very convenient form by Bird, Stewart and Lightfoot.

L 1. Given that

EXERCISES

bs inL cylindrical coordinates b m in cylindrical coordinates 3. Find the r component of


2. Find v0p component of 121

is symmetric express 

in cylindrical coordinates.

L 4. Find the ( component of bm in spherical coordinates L 5. Find an expression for b in cylindrical coordinates a K _ the 6. For answering following questions you should use the following tensor relationship  Q _ R = - + bm a
I A L 6 p KK I A a c r a b. Suppose that in cylindrical coordinates it is given that  c I t A  . Construct c. For the equations given in b, construct the R tensor if eq. 3.6 - 9 of Bird, Stewart and Lightfoot is valid
a. What is the pv component of in spherical coordinates

122

Chapter 14 VISCOUS FLOWS


Use of equations of motion are illustrated with the following examples Unsteady viscous flow similarity solution (BSL 4.1) Unsteady flow in a pipe (BSL 4.1) Creeping flow around a sphere (BSL 4.2) It would help if you read the appendix on partial differential equations We have derived the equations that govern ow of uids. We also said that specic equations relevant to any situation can be derived from these general equations of motion and mass balance. We discuss how these equations are used in the present chapter.

14.1 Old wine in new bottles


We have already solved three uid ow problems in an earlier chapter. Now we want to show that we get the same equations that were solved when we use the general equations for two problems.

14.1.1 Flow down an inclined plane


Refer to g.12.1. that T component of velocity is zero and that the ow is fully I We ^ postulated developed, i.e., is zero. Substituting these into the equation of continuity we get

4 4

which is same as eq.12.1. Thus, the conclusions regarding the nonzero velocities as represented by eq.12.2 follow: I A I A  I I0  and A Now let us turn to the momentum balance or the Navier Stokes equations. First consider the component. Substituting the above simplications we get

I  A

4 4 4 4 4 4 4 4 4 out terms 4 are zero 4 for 4 4 4 4 4 The crossed the following reasons: I0  A because of steady state 4 4 123
P

H I

I I I0 M M

I0 I M

H KI I K I K I _ PR A K K K V V O M M M M T T

The other terms on the left hand side are equal to zero because

The viscous force terms are zero because =0. The equation then is identical with eq.12.3. From this it follows that is not a function of . Now let us turn to the component equation O of motion.

I0 A I A  I

and

I  A

will nd that uid elements are not accelerating. The fact that the left hand side of the equation of motion, which is equal to the acceleration of uid elements as measured by an observer moving with the uid, is zero indicates the same.

4 4 4 4 4 4 4 4 4 out terms 4 are zero 4 for 4 4 4 4 4 The crossed the following reasons: I0  A because of steady state 4 The other terms on the left hand 4 side are equal to zeroI0 because I0 A I A  A  and 4 The viscous terms crossed out are zero because 4 I I0 A A O A  4 T 4 4 The resulting equation is 4  _ 4 K I0 4 P A K M R 4 same as eq.12.4. We note that since the ow4 is fully developed, the component of velocity does not change with . Other velocities are zero. Hence an observer traveling with the uid
P

H I0

I I0 I M M

I I M

H K I0 K I I _ A K M K M T V O M

K I PR K S V M T

14.1.2 Flow in a pipe


We were looking steady fully developed I A I ow in the axial direction. Hence P$^ =0 and I ^ at T incompressible =0. We also postulated that r t =0. We also I ^ postulated that ow is symmetric around the T axis or ow is axisymmetric and hence t v =0. If we substitute all this into the equation of continuity in cylindrical coordinates, we get the trivial result that 0=0! It of course means that our assumptions are consistent with each other. Now turning to the p and v components of equation of motion, we nd that the entire left hand side of these ^ are zero because of steady state. The other terms equation is zero. The terms involving are zero because of I

4 4

4 4

T 4 ^ v = ^ p =0 and because of axisymmetric nature of velocities. 4 Hence these equations show that O again O which 4 4 4 I ^ agrees with eq.12.6. Finally we turn to the T component equation of motion.4 Once is zero because of steady state. Further because of 4 4 I A I A I A  r t 4T 1244

I A I A r t

4 4

4 4

A 

the entire left hand side of the equation is zero. Once again this implies that acceleration of uid elements as observed in a frame of reference moving with the uid is zero. This is consistent since an observer moving with the uid will nd acceleration to be zero in fully developed ow and when the other velocities are zero. Once the right hand side simplies to

6 H I  A _ OT M p p p p V M P R

If we integrate this once we get

is zero since stress is zero at the center of the pipe. Then the resulting equation is identical to eq.12.7. Exercise: Show that the equations of continuity and motion give the same equations derived using the shell momentum balance equations for problems worked out in the second chapter of the text by Bird, Stewart and Lightfoot.

I p _ A O T L M  p M P R pL I ^ where is a constant of integration. It must be zero since p

4 4

(14.1)

4 4

4 4

14.2 New wine in old bottles


Now we consider some new problems. The strategy we follow is generally the same. We pose a physical problem, and idealize it to capture the essential features. The equation of continuity is then applied to this idealized picture to draw conclusions about the nonzero velocities or velocity dependences etc. Then we move on to the relevant components of equation of motion, i.e., Navier Stokes equations. We obtain equations by simplifying Navier Stokes equations using our idealization and the conclusions drawn from equation of continuity. Then we get the relevant boundary conditions. We then nd the velocity proles and draw appropriate conclusions. The above procedure is so enticing and in a way misleading. It is misleading since very few problems can be solved in this way. However as mentioned earlier, we gain intuition from these examples and hence they play a central role in our education.

14.3 Unsteady viscous ow


Imagine a boat being dragged into a river. We want to know how much force needs to be exerted to drag it because of the viscous resistance offered by the uid. We might expect it to vary with time: large initially and decreasing with time. This is because the acceleration required to change the velocity of the uid from its values to the new values set up due to the motion of the boat will be large initially and will progressively decrease. It is this kind of phenomena we want to capture. We expect it to depend upon the velocity with which it is being dragged. So we sharpen the problem and ask how much force is needed to drag it at a constant velocity a . It also depends upon the velocity of the uid. We sort of expect it to depend upon the relative velocity between the uid and the boat. So let us assume that the uid is stationary before the boat begins to move. Boat has a very complex shape and we will get into mathematical trouble very soon. We need a simple shape. Let us assume it to be a at plate. Of course we will lose all the effects 125

that may be caused by the curvature of the boat walls, and they could be (and are!) important. However at least this problem will capture part of the viscous effects. We simplify this further and say that the plate extends to innity in all directions! We can only calculate the drag force per unit area since we lost the niteness of the size of the plate1 . It would still some idea of the kind of viscous resistance we can expect. We could think of a plate being dragged in a uid layer of nite thickness. In the same spirit of idealization as adopted so far, we can assume that the uid is also innite in extent. Once again we may hope that it will give some idea about what will happen when uid layer thickness is nite.

14.3.1 Problem statement


In summary, we want to calculate the drag force per unit area required to drag an innitely large at sheet at a constant velocity in a Newtonian incompressible uid of innite extent. Cartesian coordinate system is ideally suited to solve the problem. We locate the plate at = 0. See g.14.1.

1.0 ~ v V

Plate moves in x direction with constant velocity V

0.05 = y/ 4 t

Figure 14.1: A plate being dragged at a constant velocity, V

14.3.2 Simplications and assumptions


As the plate is being dragged I in direction, we assume that only component of velocity the A I A  . The plate extends to innity in the T direction but exists in the uid, and hence the uid senses identical boundary condition from the moving plate for all values of T . Hence, we do not expect the nonzero velocity to depend upon T . Let us assume that the gravity is in the direction of T .

14.3.3 Equation of continuity


The equation of continuity then simplies to

Idealized problems of this type have other uses also. They can serve as bench marks for testing home made numerical codes.

I0  A
126

Hence,

I0

is only a function of .

14.3.4 Navier Stokes equations


of Stokes equation by using the result of We simplify the and T components I A the I Navier  A equation of continuity and the fact that . We get

 A O A O T M P R
Hence The component equation of motion simplies to

A c l M P R!T O I KI0 A O M _ K
(14.2)

4 4

4 4 4 4 the equation 4 of4 continuity, we get If we differentiate the eq.14.2, and use
Integrating this once, we get

plate spreading into uid, and every point on the plate is identical. Hence, we do not expect any variation in pressure in the direction. Further, we do not see any obstacles in the ow path which can create accelerations, and hence, on this count also we do not expect any variation to come in pressure in the direction. We therefore do not expect any pressure gradient to ] develop in the direction. Hence we equate the constant to 0. Equation ref.14.2 simplies to K I0 I

l O A
T c l P However we have shown earlier that A c M R5T . Hence by substituting this into the O above equation we nd  A
T c l As left hand side is a function of while the right hand side is a function l of T , this implies that both sides must be independent of and T , and can be equated to . However, when l we move far away from the plate, we do not expect pressure to vary with time. Hence ] ] ^ A , and hence pressure should must be a constant. Let it be . This would imply that O increase linearly with . The ow is being set up by the effects of drag force exerted by the

4 4

A  K O

4 4

14.3.5

A _ K 4 4 4 4 Initial and boundary conditions P I A  A  c


127

(14.3)

As the uid is initially stationary, the initial condition is given by

(14.4)

We have the no slip boundary condition at the plate. Hence

I A adc A 

(14.5)

We can not use protably any other condition at the plate. The stress condition will in fact allow us to calculate the drag force required to keep the plate in motion with a constant velocity. We anticipate that the effects of the drag of the plate, occurring by diffusion, will take longer and longer times to affect the velocity as we move farther and farther away from the plate. In the limit of innite distance from the plate we expect the velocity to be not altered at all from its initial condition. Hence I0 A  (14.6) c The velocity prole can then be determined by solving eq.14.3 with the above initial condition and boundary conditions.

14.3.6 Similarity solution

At this juncture we normally look for characteristic quantities. It is obvious that a can be used I0 for nondimensionalizing velocity. Let

However we have no characteristic quantities for time and length. The uid extends to innity and hence the problem does not have an intrinsic length scale either. There is no boundary placed at some distance, which can form an estimate of length scale, where the momentum diffusing can be absorbed. Similarly, there is no opposing force to counter the effects of drag being imposed by the moving plate to estimate a scale for velocity gradient, and from which a length scale can be estimated. For the same reasons, we see that there is no scope for reaching a steady state, and thus, we do not have an estimate of the time scale either. Such problems are special and can give a solution known as similarity solution. In such problems where some characteristic quantities are missing, one can make them up using other properties and independent variables of the problem. In this sense K ^ different independent variables become similar. In this problem, we see that the ratio has units of time and hence we can think K^ of a combined variable . If a similarity solution exists, then the velocity would be a function of only the combined variable and not and separately. In this sense, the two variables are similar in their effect on the velocity. The combined variable is referred to as similarity variable. Theorems are not available to predict as to when a similarity solution exists and how to nd it. It is more of an intuitive idea, one tries it out and checks if such a solution exists or not. Since it is being proposed that the dependent variable is a function of a combination of independent variables or only the similarity variable, it is obvious that the partial differential equation for the dependent variable must be reducible to depend on only the similarity variable or to an ordinary differential equation, and that the original independent variables must not appear in the nal ordinary differential equation. That is the rst check on the hypothesis: check if the equation can be reduced to depend exclusively on the similarity variable. We have previously postulated the form of the variable through physical arguments. This is also a procedure which involves guess work and one may have to try different forms in general. Let us dene the similarity variable

` A K Xh
128

where the factor 4 is put for making the resulting differential equation look pretty! A few steps of the algebra are as follows:

4 A ` 4 A` 4I  I 4` 6  I Similarly A U` A A` 4 4 Xh This can be continued and it is left 4 to you to 4 show that eq.(14.3) reduces to KI I  ` L A` K M A` A
A 

I

I A L 

(14.7)

(14.8)

(14.9)

So the hypothesis has passed the rst test where the partial differential equation has reduced to an ordinary differential equation, and the original independent variables do not appear in the equation. We face a new problem now. Since we had a partial differential equation, we had enough boundary and initial conditions, a total of three, to solve the problem. Now we have only an ordinary differential equation of second order and we need only two conditions in terms of . Thus the second test is that the boundary and initial conditions specied for the original problem must reduce without any contradiction into the required number. In the present problem, the boundary condition at =0 reduces to

I A a The specication of =0 and The conditions at =0 and

reduce to one I single value of similarity variable, i.e., =0. are the same: =0. Hence both these conditions collapse into one and reduce the total number available by one. Thus the hypothesis has passed the second test too. The solution to this problem is given by

at

A 

or

I A 6

at

A 

I0 A 6 a k

s

where erf is the error function dened as

s A L

: ,

T

A sketch of the resulting velocity prole is shown in g.14.1. As expected the non^ dimensional velocity depends only upon the combination and not separately on the X two independent variables and . As can be seen from the gure, the velocity increases with time at a xed value of as momentum diffuses from the plate or, equivalently, as the effects of the drag of the plate penetrate deeper and deeper into the uid. Similarly, we have to go farther and farther from the plate as time progresses to nd zones where the effects of the drag of the plate are not felt. One therefore % denes two quantities: penetration time and penetration distance. Penediffusion or drag force to be tration time, , is the time required for the effects of momentum % felt at a specied distance. Similarly penetration depth, is the depth up to which the effects of momentum diffusion or drag force are felt by a specied time. We of course have to specify

129

what we mean by the term: the effects of momentum diffusion or drag force are felt. It is arbitrary and we can say that the effects are not felt when the velocity I8   at the location and time m 98a . The absolute numspecied is less than 5% of that imposed by the boundary, i.e., ber we calculate will therefore be arbitrary but the numbers I A   will allow us to compare different situations. Thus, effects are deemed to be felt when m 98a . In our problem, the value of I0 I xed once the value of is xed and we xed = 0.95. Hence

l%

and

Kinematic viscosity is diffusion coefcient for momentum. Thus, penetration depth should increase, and penetration time decrease with an increase in kinematic viscosity. These are conrmed by the above. Suppose we had a uid of thickness W above the plate and we did the experiment. We can then K ^ guess that the effect of drag forces will not be felt up to the thickness W for a time less than and for times much smaller than this, the solution we derived can be used, even though W the uid layer is not innite in extent. Thus we see that ideas of penetration time and depth have enormous value in thinking about limiting solutions. We have to exert a drag force to keep the plate moving at a constant velocity. The drag force will change with time. We expect that the effort to accelerate all uid initially at rest will be large. As time progresses, since some elements have picked up some velocity, the effort needed will become smaller. The drag force will have to be exerted in the direction, and opposite to the frictional force exerted by the uid l on the plate. Hence the component of the drag force per unit area is given by m .j = . Thus

c h
%

Drag force to be exerted per unit area

A A

Drag force should be larger for more viscous uids, and should increase with the speed with which the plate is dragged. The above equation satises our expectations.

l c A _ a k h

I0 _

14.4 Start up of ow in a pipe


Consider the unsteady ow set up in a tube when a pressure gradient is suddenly applied when the uid is initially at rest. We know the steady fully developed velocity prole is parabolic. We might be interested in knowing the time period one has to wait for the velocity prole to be established. Since we are considering that pressure gradient is known, we might be interested in the ow rates obtained during the transient.

14.4.1 Simplications and assumptions


In this problem also, similar to what was done in the case of steady ow, we will examine the middle section of a long tube so that we can neglect the entry and exit I effects. When the exit and end effects are neglected, we can expect that the axial velocity, , would be not be a function of T . Thus, I

4 130

A It A  v

14.4.2 Problem statement


We are thus postulating that

I A l u a pc , and the problem is to nd the function.

14.4.3 Equation of continuity


We are considering incompressible uids and hence the equation of continuity will simplify to

p I r  A  c p p

In view of the impervious nature of the wall, this implies that

or

I p r A

Constant

I A  r

14.4.4 Navier Stokes equation


The p and v components of equation of motion can be used to show that pressure does not depend upon p and v . The T component of equation of motion simplies to

h4 4 4 4 problem, 4 we assumed 4 that the direction of gravity and where, like in the case of steady state the ow direction coincide. As there is no boundary where pressure is xed, we can combine the gravitational potential and the pressure term into A O P RUT
The equation then simplies to

A P 6 O MRM T

KGI K T

6 H I 4 4 h p 4 p Ip 4 p V 4 the axial4 variation4 of 4 is being We are examining the case where assumed to be zero. Thus differentiation of the above equation with T establishes that K K A  c or A constant A W W 4T 4T 4 equation to be solved 4 is then The partial differential H I I A M h 6 V p 4 W p 4p 4 p 4 4 4 14.4.5 Scaling I A P6 T M
In this problem we do have a characteristic length scale . Hence we can nondimensionalize p with the radius. This problem, as with the previous one, is one of percolation of viscous effects in response to the forcing of ow by pressure gradient. Hence we expect the diffusion time to be relevant the relevant scale to nondimensionalize time. We can formulate a diffusion time scale using the concepts of the previous problem by letting desired penetration depth to be . Thus we select the characteristic time scale A K ^ the and nondimensionalize with . Though apparently a velocity scale is not present, we do have one. The steady state velocity is

131

a convenient scale since the velocity must increase from the zero value A to the nal K ^ _ value. Thus we can select the maximum of the steady state velocity prole, a X W to be the appropriate velocity scale. Thus we have the following non-dimensional quantities:

A h K  The partial differential equation then takes the following form: I 6 H I A X M p 4p p 4 p V 4 4 4 4 14.4.6 Initial and boundary conditions
I A X_W I K p A p

(14.10)

The uid was initially at rest. Hence the initial condition is given by

I 8 A  pc

or in terms of dimensionless variables

I [ A  p$c I 6 8 A  c

The boundary condition at the circumference of the pipe is given by no slip condition:

I [ A  c

or in terms of dimensionless variables

There is really no other boundary. The origin or p =0 is an articial boundary which does not constitute a physical boundary. Thus we have to use some condition at p =0. If we look back at the way we solved the steady state problem, it turns out that we used that velocity gradients should be nite at the center of the tube. In cylindrical and spherical coordinates we encounter this situation often where we need a condition at the origin of the radial coordinate. The most commonly used condition is to require that the dependent variable or its derivatives be nite at p =0. Here also we can require therefore that

I  l c 

is nite

14.4.7 Solution by separation of variables

Since there is a steady solution, we can postulate the solution to be of the type

I I I  p$ c  A p  M D pUc l I I where is the steady solution and D is the unsteady solution. Hence, the steady part of the solution satises the following ordinary differential equation: 6  H  I  A (14.11) X M  p  V
p p p
Substituting the assumed form of solution into eq.14.10, and using eq.14.11, we nd that the unsteady part of the solution must obey the following partial differential equation: (14.12)

I D 6 H I D A 4 p 4p p 4 p V 4 4 4
132

I 6 A  I I I  c and hence D 6 c 8 A  [ is nite and hence D  c l is nite velocity. Hence the initial condition requires that There is no initial condition for the steady I D [ A I  pc p by the parabolic velocity prole: The steady solution is obviously given I A 6 K p part of the solution is given by Hence the initial condition for the unsteady I D 8 A 6 K pUc p is given The solution is given in terms and by of Bessels functions  B ? p : F I D l A   p$c
where B and g are Bessel functions of zeroth and rst order. order Bessel function:  

We have to now work out the boundary and initial conditions for both parts of the solution from the boundary conditions specied. We know that the no slip boundary condition is satised by the steady solution. Similarly the steady velocity is nite at the origin. Hence

? g ?g ?p

are the roots of the zeroth

B ? A

14.4.8 Looking back


We have the solution as two parts. The unsteady solution as well as steady solution are both positive. They both are equal in magnitude initially. Thus the initial velocity is zero. As time progresses the unsteady part decays exponentially. Thus the steady velocity prole is also established exponentially. The exponential term can be written as

h We expect that the response of the uid to the imposed pressure gradient would be fast if viscous resistance is small or if is large and h is small. The expression for the time constant conrms this expectation. The unsteady part will decay to about 10% of its original value when   L c c 9Um X m or when m X . This gives a rough idea of time required for steady state to be established. : K For a pipe of radius 0.05 m and liquids whose kinematic viscosity is of the order of 10 m /s, the time constant is of the order of 1000 seconds or 15 minutes. A few velocity
9Um X
 

Thus, can interpret this as follows. There are innite number of time constants given by K ^ we ?K  . The solution corresponding to very small time constant decays very fast. Hence the surviving unsteady solution corresponds to the largest time constant which corresponds to the rst root, g . It is equal to 2.405. Thus the time constant for establishing the steady velocity prole is given by K

: F A : " % F

proles at different time intervals are shown in g.14.2. The ow rate will also approach the steady state value exponentially, the response time being of the same order as calculated just now. 133

Center of tube 1.0 v z ~ t =0.5 ~ t =0.2 ~ t =0.05 1.0 0 r/ R 1.0

Vmax

0.5

Figure 14.2: Development of unsteady ow in a tube

14.5 Slow ow past a sphere


Imagine a stationary sphere. A uid ows past it. Here we consider steady ow. The uid approaches the sphere with a uniform velocity. We therefore expect the uid to have the same uniform velocity far away from the sphere. This is a typical problem which considers ow past a body. One hopes that an understanding of behavior of ow past isolated objects which can be used to build understanding of ow past assemblages of solids e.g., packed beds. Since we are interested in an isolated sphere, we can consider the uid to extend to innity. The situation is shown in g.14.3

14.5.1 Assumptions and simplications


It is convenient to use spherical coordinates. We assume that the sphere is perfectly spherical. Thus we do not expect any variation in the direction of azimuthal angle, . Thus we expect the velocities to be dependent on p and v . Further, since the ow far away is uniform and does not have anyI swirl or azimuthal angular velocity, we do Inot expect I it to be created. Hence we assume that =0. The only non zero velocities are then r and t .

14.5.2 Problem statement


We wish to nd the velocity prole in slow uniform ow past and according I a stationary I sphere   to our assumptions, this means to determine the functions r pcv and t pcv . Slow ows are 6 . also referred to as creeping ows and this implies that

134

r V z V 8

14.5.3 Scaling
A very important simplication in this problem comes from the slow ow condition. The slow is small. We will use this in effecting simplications in the equations ow implies that of motion that govern the ow. Here a scale for both length and velocity are available. Thus we select and a as scales for the length and velocity respectively. We need to select a scale for pressure. A reference pressure far away from the sphere will be available. Let it be 2 .However we are interested in scaling the pressure differences caused by ow. The pressure Odifferences could be due to acceleration of uid or due to viscous friction. In fast ows we do expect acceleration to play an important role and we could nondimensionalize the pressure differences with respect to inertial forces. As the ow is slow, we expect viscous forces to dominate and we should therefore scale the pressure differences with respect to viscous forces. _ ^ . Thus it seems appropriate to dene a An estimate of viscous stresses is given by a non-dimensional pressure in the following way

In addition to this we have the following dimensionless velocities and length:

K I I t  v  A  K p r M  p 4p p v 4v The dimensionless equations of vector be as follows: will 4 motion in the 4 form m b  A b M b K O


 
2

When we substitute all these into the equations, we get the following. The equation of continuity is given by 6 6

We are assuming that gravity does not play a role here. Otherwise could take the pressure on the equatorial plane of the sphere but far away from the sphere as the reference pressure.

Figure 14.3: Uniform ow past an isolated sphere

I A r I A It p A p r a t a

 K A O _ O a O

135

Hence in the limit of , we can neglect the left hand side. This is equivalent to neglecting inertial forces in the entire ow eld3 . This is a major simplication and as a result we obtain a linear equation. The problem involves a two dimensional ow and hence is ideally suited for solving using a stream function.

14.5.4 Equation for the stream function

 4  4 4 k 4 equation of continuity since it is automatically satised by the use of stream function. Further,
we eliminate the pressure by following the procedure described in the section where the stream function was discussed. Skipping a lot of algebra, the equation for stream function turns out to be K H 6 K

6 I A I A 6 r pK v v c t p v p ^ K a  . We need not worry about the We can dene a dimensionless stream function, as

The stream function used here is the Stokes stream function as the ow is axisymmetric. The velocities are related to the axisymmetric stream function as follows

14.5.5

4p K M p K 4 Boundary conditions

 v

A 4v  v 4v V 4 4 0

(14.13)

The boundary conditions are as follows. Firstly we have the no slip boundary condition on the surface of the sphere. I A I A  t r c at p A Then we have the boundary condition far away from the sphere

A a I A r m@qsra

It turns out that this is not a uniformly valid approximation. While it is true near the surface of the sphere, it is invalid far away from the sphere no matter how slow the ow is. Oseen (See the text by Batchelor) found a more uniform approximation by replacing the inertial terms by 

0 A ak ,8 v p and I A t $mqytGak A ak  vYp stream function. Thus Now these conditions have to be written in terms of the dimensionless we have 6 A o 8 As p v p K  v v 4 and 6 4  A v p  v p As p 4 ( - 4 w w0
3

where e is the unit vector in the T direction. First we need to convert these boundary conditions into velocity components in the spherical coordinates. Thus we have

136

Both these conditions are equivalent to one single boundary condition:

6 K K Lp v

as

Let us recall our discussion regarding boundary conditions and stream function in chapter13. As mentioned there, the order of the differential equation for stream function is greater than that of Navier Stokes equations. In this problem, the equation for stream function is fourth order and, in principle, we should have four boundary conditions. However we seem to have only three: one far away from the sphere and two on the surface of the sphere. Once again, as mentioned in chapter13, we could have added a constant to the stream function and the velocity boundary conditions would still be satised. Therefore, the extra constant that needs to be determined could be specied arbitrarily. For example, in the above expression, we have implicitly chosen a value of zero  for this constant at v A . Our choice implies that, for the stream line passing through v A and v A k , the stream function is zero. As we shall see a little later, this stream line also coincides with the surface of the sphere, which corresponds to the no slip or zero velocity surface. Hence, the stream function at any point by itself is also equal to the difference between the stream function at the location and that on the surface of the sphere. The difference in stream functions is also equal to the ow rate between the stream lines, and since we have chosen the no slip surface as the zero stream line surface, it also conveniently describes the actual ow rate between the stream surfaces. Usually this is how the value of the arbitrary constant involving the stream function is chosen.

14.5.6 Solution of the partial differential equation


The equation for the stream function is of fourth order. We present here an intuitive solution. Those who are interested in a more formal approach should consult the book Hydrodynamics by Lamb or Low Reynolds Number Hydrodynamics by Happel and Brenner. In view of the limiting form of the stream function we try the following solution:

 When the assumed form is in the partial differential equation, we get an ordinary p substituted  differential equation for Z : L H  L  A  H 
A Z p  K v
The solution of this fourth order homogeneous differential equation is given by or

Using the boundary condition on the stream function far away from the sphere, we deduce that ] he constant must be zero, and that = -1/2. The other two constants can be evaluated from the no slip boundary condition at the surface of the sphere. After some algebra we nd that the velocities in dimensional form are given by

p  A
K p p K ] p Z H p M M M A p
K M p M p K M ] p V  K v k
6 L H p V M L 6 H

pK pKV

pK pKV Z p

I A r a

8 v , p V

137

and

I A t a 

0
A

6 X

p V X

6 H

v p V

We still have to nd the pressure. Pressure is found as follows. We write

O A

  Op p M Ov v _ qsrnm K  p M p8qstom K v

The resulting equation can be integrated with the boundary condition that O from the sphere. Once again skipping a lot of algebra, it is found that

4 Jb 4 4 4

$b

or

 v  A _ a  H o L pK V O _ A  L a o K v p O O

far away

14.5.7 Drag force


We are now in a position to nd the drag force on the sphere. We expect the drag force exerted by the uid on the sphere, ` , to be only in the T direction. The T component of the drag force is given by

A A

, ,

 

vLk vLk

:vqyrom m :v o v M rr o 8 v v # r t O

JD

c 

where the stresses and the pressure are evaluated at the surface of the sphere. Substituting the Newton Stokes law of viscosity we nd that, at the surface of the sphere,

_ a  A A rr E c r#t L

k  v$c O

_ a O N L

,8 v

Substituting all these results we nd that

A k_ a

which is the famous Stokes law of drag. As mentioned earlier, we put this in the form of drag coefcient, . The drag coefcient for ow past submerged bodies is dened by using the area projected on to the plane perpendicular to the ow direction.

6P K A `
 L a r K a r is a reference velocity. Substituting a r A a c
 A k , we nd that where u L A X

138

where the Reynolds number is dened using the diameter for the length scale. We mentioned that the forces in other directions are expected to be zero. Hence, we should nd that

 A A
We leave it to you to verify it.

, ,

vLk  vLk 


:nEvqsrnm m
 !

D :nEvqsrnm Dm

14.5.8 Form and friction drag


It is often convenient to break up the total drag force into two components. The force due to tangential stresses is referred to as skin friction drag. The rest is form drag and is very _ a and k _ a sensitive to shape. For creeping ow past a sphere, they are given by X k X respectively.

14.5.9 Looking back


We should expect drag to increase with the velocity of the uid, and viscosity. We also expect the drag force to increase with size since the area where frictional forces are exerted increases.

14.5.10 Flow transitions


The above solution will not be valid  m 6 as Reynolds number increases. The solution is valid for creeping ow, i.e.,, till . For higher Reynolds numbers, the various complex phenomena described earlier come into play.

139

Appendix
We recall the method of separation of variables for solving partial differential equations. It is being assumed that you already know it! This appendix gives examples of the type of equations we encountered in our text. For more detail, the reader should refer to standard texts on partial differential equations.

14.A Separation of variables


Several linear partial differential equations will be encountered in the course. We can often solve them by using the technique of separation of variables. The idea is as follows. Suppose we are dealing with the coordinate system g c K c . Consider some independent l  g variable c K c c . Suppose it obeys an equation of the form , it is found that the partial differential equation is made up of sum of ordinary differential equations, " = c ? c l g  and . Let us represent these ordinary differential equations by , one each for Z $" = c " ? c =  where the superscript indicates the F derivative. It often turns = out that the Z original partial differential equation as represented by sum of ordinary differential equations, " = one each for  and can be rearranged to get

= c l A  l#" g g #" K K "  When we substitute the assumed form into the partial differential equation obeyed by

" \%

" \%

 c ? c l A = $ " " ?  c cc A 6 c L c Zg Z =c c = n

" I%

" \%

Since each is a function of a single independent variable, the only way in which the above equation can be satised is that they all must equal a constant. Then each of these differential equations is solved to construct the desired solution. Consider the following partial differential equation in Cartesian coordinates.

Assume that A  tion is equivalent to Dividing by 


%("

l&% "Q $"


%'"

we nd

K 4 4 4  4 4 this 4 we nd that the partial differential equa. Substituting  K K A A   K M   K A K M


% " % "

The left hand side is only a function of while the right hand is a function of and . Hence both must be equal to a constant, say *) . Hence we have

6  6  K% 6 K" A % K M " K  

6  A * )  

and
)

6  K% 6 K" % K M " K
140

The rst equation is a differential equation for be rearranged to get 6 K%

These two are ordinary differential equations that can be solved. We nd the solution that satises the boundary conditions by choosing the constants and ) . Separation of partial differential equations in curvilinear coordinates can be done in a similar way and you should refer to a book to see how it is done.

6 K" %  K M ) "  K Once again while the left hand side is a function of only, the right hand side is a function of . Hence both sides must be equal to another constant, say . Hence we have 6 K" "  K A and 6  K% %  K A h+ ) A

and can be solved. The second equation can

14.B Sturm-Liouville problems


A concept that is invariably associated with the technique of separation of variables is the idea that any arbitrary function can be expanded in terms of a set of complete orthogonal functions. This set is referred to as a (complete) basis set. This is similar to the idea of Fourier series with which you are already familiar: i.e., any arbitrary function of an independent  L variable, , k can be expanded in terms of sine and cosine functions of . In a broad manner it is similar to any arbitrary vector in space being expanded in terms of the basis vectors, i.e., the unit vectors. The main point to note is that such basis functions have to come from solving ordinary differential equations of the type that arise when a partial differential equation is separated. The Sturm- Liouville theory4 gives mathematical conditions, which when satised, assume that a set orthogonal and complete basis functions arise while solving ordinary differential equations along with boundary conditions. Consider a homogeneous differential equation of the form

be the domain of interest. Suppose we need to nd a solution subject to the boundary conditions

A  O   where refers to differentiation with respect to . Let 


,% .) %

p 

 M

 M

(A14.14)

yg %  M K %  A   g %   M K %   A 

(A14.15) (A14.16)

and

The following are the results of theory by Sturm and Liouville.


This is a brief summary culled out from the book on Fourier series and boundary value problems by R.V. Churchill.
4

141

14.B.1 Regular Sturm-Liouville problem


If c/-$cpc and p are real, continuous and O O Sturm-Liouville problem.

as well as -

, the problem is called a regular

1. A regular Sturm-Liouville problem has nontrivial solutions for innite number of real 6 c L cscnmomnm , called eigen values or spectrum of eigen values. values of ) ? c A
%

2. The function ? obtained by solving the differential equation for each eigen value is called an eigen function.

3. The eigen functions corresponding to two distinct eigen values are orthogonal to each other, i.e.,

, O 

&%

% ? ?

4. The eigen functions are unique and complete. 5. The completeness the eigen functions assures that any arbitrary function Z   of  can be expanded in terms of % ? : domain of

in the

 % Z A ? Z? ?
,%

where Z ? are constants. They can be determined using the orthogonal property of the eigen functions:

Z ?!A

Z  K l  
%

14.B.2 Singular Sturm-Liouville problem


If any of the conditions imposed upon c/- and p are not satised, the problem is called a O singular Sturm-Liouville problem. We discuss only one case that is of interest since it arises in curvilinear coordinates where p vanishes either at or at  . In this case also the results stated for the regular Sturm-Liouville problem are valid provided cp and p are continuous in the   O  . However the  closed interval , and - is continuous in the open interval boundary condition at or  have to be dropped or modied as follows: 1. 2. 3.

p  A  , eq.A14.15 is dropped p   A  , eq.A14.16 is dropped p  A p   , the boundary conditions have to be altered to %  A %  



and

 A

An example is the Bessels equation which arises in cylindrical coordinates. When boundary conditions are dropped, that the function must be bounded or be nite is required for physical reasons. This is the case with solutions involving Bessel function, Legendre functions etc.

142

Problems for Chapter 13.


13.1 A sphere of radius is rotating around an axis passing through its center. The uid far away from it is stationary. The ow is laminar and steady. Neglect body forces. (i) Specify a convenient coordinate system that you will use to solve the problem, and draw a gure to show your choice (ii) specify the nonzero velocities, (iii) the coordinates on which the nonzero velocities will depend (iv) simplify the equation of continuity, and (v) specify the boundary conditions. 13.2 A solid sphere is placed at the interface of two immiscible uids such that its equator  while is aligned with the interface. Fluid I occupies the space uid II occupies T  the space T . See the adjoining gure. The sphere is rotating around an axis passing through its center and perpendicular to the interface. The speed of rotation 6 . Prof Yin proposes that is 0 rad/sec and is such that the Reynolds number the following represents the velocity proles

I A
0

p V

 vUc

 v k c I L


A

k L v k U v c V p

Prof Yang says that this is wrong since viscosities seem to play no role. Who do you agree with? Provide support for your conclusions with suitable equations. Use the usual symbols for densities and viscosities. Neglect forces due to gravity. 10 marks
z Fluid I Fluid II r

Figure for problem 13.2. 13.3 A spherical gas bubble of radius B at pressure B is placed in a uid extending B . The to innity in all directions. The uid is at atmospheric pressure gas bubble therefore expands. The uid is very viscous and hence the Reynolds 6 . Gravitational forces are negligible and hence the bubble remains number spherical in shape and its center of mass does not move. We want to use pseudo steady state assumption to relate l the velocity of expansion of the bubble to the pressure at any instant. Let l be the pressure inside the bubble at any time and . Specify (i) the components of velocity expected let its radius be denoted by to be nonzero and the coordinates on which they depend upon, (ii) simplify the equation of continuity and the relevant components of the equations of motion, l (iii) B specify  ^  the boundary conditions needed to solve the problem, (iv) relate to , and (v) l if the bubble is made of ideal gas and the expansion is isothermal, indicate how can be determined.

**

143

13.4 A Newtonian and incompressible uid enters an annulus. The radius of the inner wall is = while that of the outer wall is B . The ow is laminar, fully developed and steady. A constant pressure gradient is applied in the axial direction. Only the inner cylinder is rotating with a constant speed 0 . (i) Specify a convenient coordinate system that you will use to solve the problem, (ii) specify the nonzero velocities, (iii) specify the coordinates on which the velocities of the uid depend. (iv) simplify the equation of continuity and the relevant components of the equation of motion. (v) specify the boundary conditions needed for solving the equations. 13.5 Nylon bers are made by extruding a molten polymer through an orice of radius B . Let us assume that the polymer exits from the orice as a lament of radius B . Let us assume that the velocity of the molten polymer is uniform and is given B by a . After some length, W , the lament is wound onto a roller, and the roller is rotated at a high speed. As a result, the bre will be under tension. Let it be  . As the lament is being drawn, the radius of the lament reduces from the orice to the roller. Suppose we assume that the polymer is Newtonian and incompressible. Further, assume that the drag force exerted by the surrounding air on the lament is negligible. Simplify the relevant equations to determine the T component velocity as a function of the distance from the point where nylon comes out T . 13.6 For the situations given below, and when the ow is steady and laminar, (i) specify components of velocity expected to be nonzero, (ii) specify the relevant boundary conditions, and (iii) sketch the expected stream lines.

a. A rectangular cavity, W A , (extending to innity in the direction perpendicular to the board) lled with an incompressible Newtonian liquid. The top plate is being moved in the M direction with a constant velocity a . Sketch only stream lines also for the case W .

b. A circular thin disc (radius= ) is rotating in a beaker (diameter= ) of liquid (height= ) with a constant rotational speed . Assume that the diameter of the rod holding the disc is zero and that the liquid-gas interface is at.
< =2 < =2 < ;: =2 < =2 < =2 < =2 < =2 < =2 < =2 < 12 < 31 =2 < =< V =2 =2 12 ;:;: 3131 12 12 ;::; 3113 2 1 ;:;: Liquid 12 12313131 y 12 12 W ;:;:;: 12313131 12 ;:;: 12 3131 12 ;:;: 12 31 2 1 ;:;:92 8 92 8 92 8 92 8 92 8 92 8 92 8 12 8 3131 98 x 92
L Part a

42 5454 7676 42 42 5454 7676 442 2 4 6776 42 55454 42 7676 42 4 5 R 42 7676 42545454 7676 H42 42 5454 42 7676 42 4 5 z 42 76 4254?2 >2 >54 ?2 > ?2 > ?2 > ?2 > ?2 > ?2 > ?2 > 76 ?>
D Part b

Figure for problem 13.6.

13.7 For the situations given below, do the following. In all cases assume that ow is laminar. DO NOT SOLVE THE RESULTING EQUATIONS. (i) Specify a convenient coordinate system that you will use to solve the problem, (ii) specify the nonzero velocities, (iii) the coordinates on which the nonzero velocities will depend (iv) simplify the equation of continuity and the relevant equations of motion, and (v) specify the boundary conditions. 144

a) A uid is owing through a long concentric annulus. The radius of the inner wall is = while that of the outer wall is B . The applied in the axial direction. Only the inner cylinder is rotating with a constant speed 0 . b) A solid sphere is placed at the origin. The sphere is held such that it neither rotates Far away from the sphere, the uid is in a simple shear I nor moves. ; ow: A @ . The ow is in steady state.

c) A uid is placed in a bucket of radius up to a height . The uid and the bucket are stationary. Suddenly, the bucket is rotated around its axis with constant rotational speed 0 . We are interested in unsteady state. Assume that the interface is nearly at.
y Fluid Fluid Inner cylinder rotates with a constant speed. Figure for part (a) Figure for part (b) z Velocity profile far away

H 2R Figure for part (c)

Figure for problem 13.7.

13.8 A Newtonian and incompressible uid is owing under steady conditions through a rectangular channel with sudden decrease in cross section as shown in the accompanying fgiure. The channel can be assumed to extend to innity in the direction  L W is g B . A perpendicular to the paper.  Pressure at is B while that at A It is also given that at A , the ow can be assumed to be only in the direction and the gradients in direction are zero. Answer the following questions giving your reasons: i) What components of velocity are expected to be nonzero? tem[ii)] Simplify the equation of continuity. (2 Marks) iii) Simplify the relevant components of the Navier-Stokes equations. (4 Marks) iv Specify the boundary conditions required to solve the problem. (12 Marks)
y 2B o x L L 2B 1

Figure for problem 13.8.

13.9 A cone and plate geometry, which is used in rheometers for measuring the viscosity of uids, consists of a bottom at plate and a top plate in the shape of a cone. The angle between the conical surface and the at plate is v , as shown in the 145

gure. The uid is placed between these two plates, and the top plate is rotated at a constant angular velocity 0 , and the viscosity is inferred from the torque on the top plate. Analyse the ow in this geometry using a spherical coordinate system whose origin is at the point where the cone meets the bottom at plate as shown in the gure. Consider the steady ow of an incompressible Newtonian uid in this geometry.

DEB DEB DD EB DD EB DD EB DD EB DD EB DD CB AA EDED CB AA CB AA CB AA CB AA CB AA CB AA CB AA CACA EB EB DEDBEB DEB D EB D EB D EB D EB D EB D EB D CB A ED CB A CB A CB A CB A CB A CB A CB A CA EB EB EB EB EB EB CB CB CB CB CB CB CB CB


Origin of coordinate system

Cone Fluid Plate

Figure for problem 13.9.

1. First, sketch the conguration and coordinate system, and dene the boundaries of the ow domain in terms of the three coordinates. 2. What are the boundary conditions for the three components of the velocity on the bounding surfaces of the cone and the bottom at plate? 3. Write down the mass balance equation for an incompressible uid. Which component of the velocity is non-zero in this ow? Which coordinates does the non-zero component of the velocity depend upon? 4. Which components of the stress tensor are non-zero? Write down the simplest expressions for these. 5. Write down the momentum balance equation for the non-zero component of the velocity in its simplest form for a unidirectional ow. What is the value of the pressure gradient in this equation? 6. Write down the simplest forms of the momentum equations for the other two components of the velocity which are zero, and determine the pressure gradients in these directions. 13.10 Consider a lm composed of two immiscible uids owing down a plane inclined at an angle v to the horizontal. The lower uid has a thickness g and density and P _ viscosity g and g , while the upper uid has thickness K and density and viscosity P K and _ K respectively. Assume the thicknesses of the two lms are uniform. a) What are the governing equations for the uid ow? b) What are the boundary conditions at the lower solid surface, the upper free surface and the interface between the two uids? c) Determine the velocity proles by solving the equations. 13.11 A Newtonian incompressible uid is owing  in the direction in an innitely wide ^  . Simultaneously, channel due to a constant pressure gradient, the top plate

146

Figure for problem 13.11.

of the channel is being moved in the gure.

direction with a constant velocity,

. See shown

a) Mark the various forces acting on a thin strip of dimensions in the gure. The channel is horizontally placed.

and

b) Simplify the relevant equations needed to determine the steady and fully developed velocity prole. c) State the boundary conditions. d) Stetch expected velocity proles without solving the equations for  ^  the 0, A 0, and  ^  0. Explain your sketches.

 ^ O

13.12 Two immiscible uids, A and B, are placed in a very long cylinder, with the denser uid at the bottom. See gure. Initially both the uids are stationary. Then the cylinder is rotated at a constant rotational speed 0 . The speed is very very small.

Fluid A z

Ha

Fluid B

Hb r

Figure for problem 13.12.

a) Mark the various forces on an annular cylindrical element shown. b) Simplify the relevant equations needed to nd the unsteady angular velocity prole. 147

c) State the boundary and initial conditions. d) What is the expected velocity prole at steady state. 13.13 A very thin plate of length W is falling vertically down in a Newtonian and incompressible uid. It is falling with a constant velocity a , i.e., it has attained its terminal velocity. The uid itself is contained in a closed container. See the gure. The container is very wide in direction, and is also long compared to its thickness and the plate, i.e., NW . Let the mass of the plate be . We are interested in nding the terminal velocity as a function of . Let us proceed as follows:

2W

L Gravity z x H

Figure for problem 13.13.

ii) Do you expect that a net ow will exist in T direction? Think carefully.

i) What are the boundary conditions on the plate and the walls at

iii) Suppose you set up the problem with coordinates moving with the plate, i.e., with a constant velocity a in the T direction. Neglect end effects and take note of the fact that the container is very wide in the direction. What are the non-zero velocities and on which coordinates they depend upon? iv) Based on the above answers, simplify the equation of continuity and the relevant components of the equation of motion. vi) Solve the equations and relate a v) Restate the boundary conditions. to

13.14 A cylinder with conical top and bottom surfaces and weight is falling at a steady velocity, a B , under the inuence of gravity in a very long tube lled with Newtonian incompressible uid. The length of the cylinder, W , is very large compared to the gap between the cylinder and the tube, and hence the drag on the cylindrical surface is very large compared to that on the conical surfaces. Dervie a formula to relate cW and the physical properties of the uid to a B .

148

Hint: It may be helpful to use a frame moving with the cylinder.


R

Gravity

Figure for problem 13.14.

13.15 Consider the space between two cones as shown in the gure. An incompressible Newtonian uid enters the annular space at pressure g as shown. The pressure at g . Assume steady state has been reached. We are interested in the outlet is K the middle ow region, i.e., neglecting end and entry effects.

O Figure for problem 13.15.

2. On which coordinates (pcv$c ) do the nonzero velocities depend upon? 3. Simplify the equation of continuity and the relevant components of equation of motion. 4. Specify the boundary conditions. 5. Solve the equations to nd the velocity proles. 13.16 Consider the space between two cones as shown in the gure. It is lled with an incompressible Newtonian uid. The inner cone is rotating around the axis shown with an angular velocity 0 . The outer cone is stationary. The speed of rotation is small. Use spherical coordinates. Neglect end effects, and forces due to gravity. Answer the following questions assuming steady state has been reached. 149

1. Which components of the velocity are expected to be nonzero?

6(

Figure for problem 13.16.

a)

1. Which components of the velocity are expected to be nonzero? 2. On which coordinates (pcvUc ) do the nonzero velocities depend upon? 3. Simplify the equation of continuity and the relevant components of equation of motion. 4. Specify the boundary conditions.

6(

b) Is there a possibility that this conguration can be used as a pump, that is uid will ow through it without imposing an external pressure drop? 13.17 A thin lm of uid is placed on a disk of radius . See g.17The disk is now rotated at a constant angular velocity, . The speed is such that the ow is laminar. Due to centrifugal forces the liquid ows out and the thickness of the lm decreases. We want to calculate the rate of thinning as a function of the angular velocity. We want to make the following assumptions.
z

Fluid

Figure for problem 13.17.

1. The thinning is slow since viscous forces are large. Hence pseudo steady state can be assumed, i.e., we solve the equations assuming that the lm thickness is constant. 2. The lm can be assumed to be parallel to the disk surface at all times. 3. The pressure above the liquid is atmospheric. 5. The v component of velocity is given by Answer the following questions. 150 4. Neglect the effect of gravity.

I A t p

b Simplify the p component equation of motion keeping in view the pseudo steady state assumption. c Specify the boundary conditions. d Solve the equations to obtain the radial velocity of the uid. e Determine the ow rate at the edge of the disk. f Relate the above to the thinning rate. 13.18 A thin lm of uid is placed on a porous disk of radius . See gure. The disk is now rotated at a constant angular velocity, . The speed is such that the ow is laminar, and the Reynolds number is very small. Due to centrifugal forces the liquid is thrown out radially, and the thickness of the lm should decrease. However, uid is being supplied through the bottom of the disk at a constant velocity a and it is observed that the lm thickness remains constant at . We want to analyse this ow I and  relate  a to . Flow is expected to be axisymmetric. It is postulated that

a What are the nonzero components of the velocity? Which of them can be neglected in view of the pseudo steady state assumption?

r A p Z T

ii) Prof.Ihdnag says that t A p every where in the lm for slow ows. Do you agree or disagree and why? iii) Make estimates of all relevant characteristic length scales and velocity scales. iv) Simplify the p component equation of motion. Think carefully about the terms on the left hand side and the pressure gradient in the radial direction. v) What are the boundary conditions needed to solve the above equation?

i) Take the control volume shown in the gure, and make a mass balance. Make  a guess for p .

vi) Outline a method to relate a to .

z R

Fluid Film thickness = r

Porous disk Fluid supplied at constant velocity V

Figure for problem 13.18. 13.19 Consider fully developed, steady, laminar ow in T direction in duct of equilateral triangular cross section. See Attached gure. The uid is Newtonian and both density and viscosity can be assumed to be constant. a) Which components of the velocity vector are nonzero? What are they functions of?

b) Simplify the equation of continuity, and the T component of NSE. 151

0.5

Figure for problem 13.19.

0.5

d) Specify the boundary conditions needed to solve the equation of the T component of NSE.

c) Is the pressure gradient in the T direction a constant?

152

Chapter 15 APPROXIMATIONS IN FLUID MECHANICS


We discuss two commonly used techniques in fluid mechanics: pseudo-steady state and lubrication approximation. We have discussed solutions of laminar ow problems. As could be seen, it becomes harder to nd solutions as the dimensionality of the problem increases. More complexity can be expected as the Reynolds number increases and the equations become non-linear. The complete equations of motion are very hard to solve. Hence techniques are developed to nd approximate solutions, we discuss two common ones. Exact solutions can be found for the specic problems considered. They however form prototypes useful to understand the technique.

15.1 Pseudo steady state approximation


In this type of approximation, solution to an unsteady problem is approximated by solution to a suitable steady state problem. Such an approximation can be resorted to when there are two or more time scales involved in the problem being solved. In the most simple case, there will be two time scales. Usually, the rst corresponds to the cause for the unsteadiness in the ow, and the second is related to the response of the uid to the cause. Consider ow in a channel under the inuence of a pressure gradient that varies with time. Pressure drop is the driving force for ow, and since it changes with time, ow is unsteady. Let the time scale on which the driving force is changing be . The uid response is indicated by a change in the velocity proles. Let the time scale of the response be r . The rst time scale, , can be thought of as the time scale of forcing the process, or a time scale imposed upon the process. The second one, r , can be thought of as the response of the process, and this depends upon the properties of the uid, the geometry of the equipment and so on. They are independent of each other and can have different values. If r , then the uid is able to rapidly adjust itself to the changing environment. Under such conditions, the velocity proles would be close to the steady state values corresponding to the instantaneous values of pressure drop. This is the essence of the pseudo-steady state approximation. Let us illustrate the implementation of the approximation by an example. in Consider a channel innitely wide in the T direction and whose walls separated by direction. Let the pressure decrease in the direction. We want to determine the unsteady velocity prole when the pressure drop per unit length is a function of time. For the sake of

c1

c1

c1

153

illustration, let us assume that the pressure gradient increases linearly with time. To focus on the nature of the pseudo-steady state approximation, let us neglect the end effects. Then, only the component of velocity will be non-zero. WhenIthis A postulated I0  velocity prole is substituted in the equation of continuity, we will nd that c T . This in turn implies that the pressure gradient is not a function of or T coordinates. Hence, combining this conclusion with the given dependence on time, we obtain:

Only the component of the equation of motion need be considered. After substitution of the postulated velocity dependence, the component of the equation of motion simplies to


6 M O A I

 c1

4 4 4 c 1 is the natural scale to nondimensionalize 4 K4 ^time 4 changes occur on that scale. is since _ the natural length scale. We could use
as a velocity scale. Pressure drop can be nondimensionalized with
. The non-dimensional component equation of motion will then read K K I 0 I A 6 M l  M K hc1 4 4 Thus we have the following ratio 4 K6 4 h c1
P

K I0 A O M _ K

c r A 6 M l M K c1 4 4 If the response time scale is very small compared to the4 time scale of forcing the system, the 4 term on the left hand side of the system is negligible K I and the equation becomes 6 l  M K A  M 4 4 nd the velocity prole. The above velocity This can be solved with the boundary conditions to prole is the steady prole corresponding to the instantaneous pressure drop applied. It is as if the uid is adjusting so fast that it is reaching steady state as soon as the pressure drop changes. ^ 6 Thus, in general, pseudo-steady state approximation is valid when c r c 1 .
EXERCISES 1. Look at problem 7B.10. When is pseudo steady state approximation valid?
1

appearing in the equation. The velocity prole adjusts itself to changes pressure drop by momentum K ^ diffusion. The time scale of this response is momentum diffusion time scale, and is the diffusion time scale. It is the response time of the system or r in the nomenclature we used earlier. But, it is identical with the rst term in the above ratio. Thus the equation can be rewritten as I K I0

Several instances of the use of pseudo steady state1 approximation exist in chemical engineering. A notable example is that used in chemical kinetics involving free radical reactions.

It is also referred to as quasi steady state approximation.

154

2. Two circular plates of radius are placed parallel to each other with a gap B . Now the two plates are squeezed toward each other by applying a constant force ` normal to the disks. See Figure 15.1. The gap between the disks then reduces as the uid ows out. Calculate the rate of approach of the plates using pseudo-steady state approximation and assuming that the ow is laminar. When can the pseudo-steady state approximation be used? Here the effects of gravity can be neglected. This problem is same as 3C.1 of text.

F z R F

do

Figure 15.1: Squeeze ow between disks 3. Thin uid lms are spread on disks by centrifugal coating, e.g., in making computer drives. Here a blob of liquid is placed on a disk of radius and the disk is spun around its axis with a constant rotational speed . See Figure 15.2. Due to centrifugal action, the liquid ows out at the edges of the disk and forms a lm. As long as the disk is rotated, the liquid ows out, and the thickness of the lm reduces. Assume that the surface of the lm remains parallel to the surface of the disk. Calculate the rate of thinning of the lm using pseudo-steady state approximation and assuming that the ow is laminar. When can the pseudo-steady state approximation be used? Here the effects of gravity can be neglected. 4. Solve problem 2D.2 of the text by BSL.

15.2 Lubrication approximation


Another commonly used approximation is the lubrication approximation. In this approximation, an attempt is made to reduce a two dimensional ow by a one dimensional ow. Consider ow in a very long channel whose depth is changing slowly along its length. Had depth been constant, in view of the long length of the channel, the ow might have been considered fully developed, and that only one component of the velocity is nonzero. However, because of the variation in depth, the cross section available for ow changes, and the velocity changes in the direction of ow. Hence, the ow has to be two dimensional in order to satisfy the equation of continuity. An interesting question to ask is whether a one dimensional approximation can be employed, i.e., the velocity every where in the duct can be considered to correspond to the 155

2R h(t) Top surface of the fluuid

Figure 15.2: Thinning of a uid lm on a rotating disk

ow in a duct of uniform cross section with the value being equal to the local cross section and the local driving force. Such problems are usually encountered in lubrication. Consider ow in a journal bearing, see the left side of Figure 15.3. A uid lls the gap between eccentric cylinders, and the inner cylinder is rotated with a constant speed. If the gap is small compared to the radius of the cylinders, and the eccentricity is not large, the geometry is like a channel with slowly varying cross section. This ow is interesting since it generates force perpendicular to the ow direction. This force can support the weight of the inner cylinder and hence keep the thin lubricating lm from being squeezed out.

L
y

Figure 15.3: Illustration of lubrication approximation. To keep the mathematics simple and to illustrate the point, we analyze drag ow between a 156

x Minimum gap = ho h = ho + x tan

The top surface moves with velocity V in the x direction

plane and a plate inclined to the plane, shown on the right side of Figure 15.3. The pressure in the uid at the ends of the plate is the same. The top plate is moved in the direction, and drags liquid along with it. The volumetric ow rate of liquid must be constant at all cross sections to satisfy mass conservation. Hence the component of velocity must change in the direction as cross sectional area changes in that direction. This of course implies that component of the velocity must be nonzero. The lubrication approximation aims to neglect the component of the velocity or to approximate the ow with ow between parallel plates. If the lubrication approximation is valid, the equation of continuity implies that

HI I I0 I H KGI0 KI P 0 _ A V O M K M K V M Usually lubrication approximation is used when the ow is slow or when Reynolds number is ^ 6 , and the left hand side can be equated to zero. We will examine small. Hence, B a _ a ^ KB while this in greater detail later. The rst of the viscous force terms is of the order of _ ^ W B  or _ a ^ W K . Hence, amongst the viscous force the second one is of the order of a  terms, we can neglect the second term in comparison with the rst. Thus, we have K I0  A _ O M K _ ^ If we differentiated this with respect to , the second term would be of the order of a  B , 6 and K ^ thisK might not be small for small gaps and viscous uids even though  . Hence, , which is similar in order of magnitude is also not negligible. This implies KB that the _ ^ O pressure gradient in the direction is not a constant and is of the order of a . If this was ^ is a constant, it will be equal to not recognized, there would be a contradiction, since if O zero since the pressure in the uid at the ends of the plate is equal. Now we can examine the component of the equation of motion. The left hand side of it also can be neglected since the Reynolds number is small. Then, we have H KGI0 KGI0  A _ O M K M K V Clearly, the second term of the viscous force _ ^ K terms is negligible compared to the rst one. The other term is of the order of a  B . Thus the component K of the pressure gradient is ^ was estimated to be of the order of _ a ^ B . This is much larger than also of this order. ^ , and we can O treat pressure as a function of only . Thus, the component equation of O motion reduces to K I  A  _ O M K

From the geometry of ow, can be estimated to be of the order of a  B . The characteristic value of component can then be estimated from the equation of continuity to be a  . Thus we intuitively expect that both the terms in the equation of continuity would 6 be very small when  S . Consider the component of the equation of motion.

4 4

I ^

I I  A

4 4

4 4 4 4

4 Integrating this equation with respect to y and enforcing 4 the no slip boundary conditions, we get
I0 A K  H K a L_ O K V
157

4 4

4 4

where is the gap between the plates at any location and is given by A B M  . The pressure gradient is not known, and hence we can not determine the velocity prole. This situation has arisen since we did not solve the equation of continuity. Hence, the mass conservation must be enforced. Obviously, it can not be enforced in an exact way since the equation of continuity was ignored. Thus, we must enforce in some approximate way so that the pressure gradient can be made consistent with it. This can be done in an overall way as follows. The volumetric ow rate can be obtained by integrating the velocity prole and is given by  must be constant by mass conservation, and hence the pressure gradient must be such that can remain constant. Therefore, the above equation can be treated as an ordinary differential equation for pressure: ;

; A a L 6 L_  O

a 6  L _ A  O LK ; Note that is not known. However we have two boundary conditions for pressure: it is is equation can be integrated with the two equal at A and A W . The above differential boundary conditions to get B ; A a B M where is the gap at A W . Substituting this back into the differential equation, and inte A grating, pressure can be found as a function of . If the pressure at be B , we have O _ a B o  B A  K B M  O O
As mentioned earlier, the lubrication ow generates a force normal to the ow direction. The normal force, ` ? , exerted on the top can be calculated and for small angles it is just  toplate A W . The normal force is not equal to zero and is equal the integral of B from A O O given by _ H

` ?A  K a

B LB V B M

This is the lubricating force that can support the weight of the bearing. Let us examine the validity I of the I0 ^ approximation made. The acceleration of uid elements or the inertial force is given by . Neglect of this term in comparison with the viscous forces is the approximation made. It is justied only if the ratio of the inertial forces to the viscous forces is small. Nominally, the ratio is given by

Pa K  KB A  B a B _ a

4 4

Thus lubrication approximation would be a good approximation if the above ratio is small or for low Reynolds numbers and when the gap changes gradually.

158

15.3 Solution by expansions


Though not covered in this notes, another important technique is to nd solutions by nding corrections to an approximate solution. It is reminiscent of Taylors series. Several problems have a parameter on which the solution depends. Reynolds number is one such, and  the technique of solution by expansion proceeds to nd a solution in the limit of , and determines the corrections as power series in . In case of ow past a sphere, for example, one might attempt to nd

I  A I B   I g   r p cvUc r pcv M r pcv mnmnm IB  and I g is the rst correction etc. As it turns out this approach where r is the limit as r
works in many problems. The approach fails when the problem is singular, and boundary layer theory and, yes, Stokes or creeping ow past a sphere are the well known examples of this. See the appendix to chapter 16.

EXERCISES 1. Consider a ow in a tube of conical cross section. Solve it by using the lubrication approximation. Derive a criteria for validity of the approximation. Solve it exactly by using spherical coordinates. Verify the derived criteria for validity of the approximation. 2. Solve problem 2C.5 of the the text, BSL. 3. Consider the eccentric cylinders shown in Figure 15.3.The radius of the outer cylinder is .  Point F is the center of the inner cylinder. The center of the outer cylinder is located at A c A . Find the lift on the inner cylinder. The gap at all locations is small compared to the radius of the cylinders.

159

Chapter 16 HIGH REYNOLDS NUMBER LIMIT & d ALEMBERTS PARADOX


We nondimensionalize the Navier-Stokes equations and look at the high Reynolds number limit. We show that this limit gives rise to equations that lose friction terms altogether. These equations correspond to irrotational flows, and we demonstrate their importance. We examined several solutions to low Reynolds number ows, and understood the characteristics of ows dominated by frictional or viscous forces. Now, we turn our attention to ows at high Reynolds numbers. Earlier, in chapter 10, we discussed qualitatively the changes to be expected when the Reynolds number is increased. Such ows are of interest both because very complex phenomena are encountered and because most practical ows occur at high Reynolds numbers. We earlier saw how scaling can help us in reducing equations to dimensionless form and also in seeking solutions at various limits of the dimensionless of parameters appearing in the equations. We apply this technique to examine ows at high Reynolds numbers. We assume that there exist a characteristic length , a characteristic velocity , and characteristic time scale . We can then nondimensionalize the Navier-Stokes equations. Since we are interested in ows at high Reynolds numbers or where inertial forces are important, we can nondimenK P sionalize pressure with inertial forces, i.e., . Let the dimensionless velocity, pressure, coordinates, and time be represented by , , , and . It is easy to show that the Navier-Stokes O equations become 6

j ) i c M m b A b M b K M i R K t j 4 j R O where is the Reynolds number 4 and R is the magnitude of the body force.
tion and we obtain

(16.1)

16.1 High GIH limit In the limit of high , the terms representing viscous forces drop out entirely from the equa-

i c M m b A b M i R K t j 4 O j R 4 160

(16.2)

This equation is very important and is referred to as the Euler1 equation. There are many odd and incongruent aspects of the Euler equation. 1. The viscous force terms, which contain the highest order of spatial derivative, have vanished. Thus we can expect to be unable to satisfy all the spatial boundary conditions. 2. Since viscous frictional terms are lost, uid motion predicted by the Euler equation will be conservative in character. By this we mean that the irreversible conversion of mechanical energy into heat is absent and hence mechanical energy is conserved. Thus, these solutions will not represent real ows in the entire ow domain. 3. Despite the two lacunae mentioned above, as will be explained later, solving the Euler equation is still important. Solutions to the Euler equation are found in an indirect way, as if entering through a back door. We discuss these ideas in detail a little later, but present a sketch here. It will be shown in the next chapter that K if can be written as the A gradient of a scalar potential , i.e., as , then, = 0 for incompressible Newtonian uids, i.e., the term representing viscous forces in the Navier Stokes equations vanishes identically. If velocity can be written as the gradient of a scalar, it follows from vector identities that Q , which is called vorticity, becomes identically zero. In the next chapter we will show that vorticity is related to the tendency of uid elements to rotate. Flows where vorticity is zero are referred to as irrotational ows. Thus, K irrota , vanish tional ows of incompressible Newtonian uids, where viscous forces i.e., 2 identically , are solutions of the Euler equation. Irrotational ows can be determined by solving for the velocity potential. Thus, solutions to the Euler equation are found in an indirect way by solving for the velocity potential. Finally, the pressure found by substituting the velocity into the Euler equation given below in dimensional form:

b(

16.2

t J b b 4 4 Elimination of the no-slip boundary condition


M m  A P6 M O

(16.3)

The elimination of viscous terms in the limit of high Reynolds number, and the consequent inability to satisfy all the boundary conditions implies that some boundary conditions have to be left out in nding solutions to the Euler equation. Consider ow past a stationary sphere. We need to satisfy two velocity boundary conditions on the surface of the sphere. Firstly, as the sphere is solid, uid can not penetrate the surface. Hence, the component of the velocity normal to the sphere must be equated to zero. Secondly, by the no-slip condition, the tangential velocity also must be zero. Let the velocity of the uid far away from the sphere be a and its diameter ] be . If we use the nondimensionalization scheme outlined above, then, the limiting form of ] ^ the equation of motion as a permits only one of the two boundary condition to be satised. If the velocity component normal to the sphere is not set to zero, uid will ow into the solid, and out of the expected ow domain. This would mean loss of uid from the domain of ow and mass conservation would not be satised. Hence, continuity of normal components of velocity is enforced. This means that the no-slip condition on tangential velocity

} h

Euler is pronounced oiler. All ows where viscous forces are zero are not irrotational. A classical example is that of rigid body rotation: here the vorticity is constant and hence viscous forces are zero.
2

161

has to be violated. This of course will imply that the sphere does not exert any friction on the uid. This is consistent with the expectation that solutions to the Euler equation of motion are inviscid. The conclusion that a uid moving past a stationary sphere does not experience drag also implies that the drag experienced by a sphere moving in an otherwise stationary uid is also zero. However, we know from Stokes solution, i.e., at low Reynolds numbers, that the drag is not zero. For both these limits to be consistent, this implies that as Reynolds number increases, perhaps beyond some value, (this is equivalent to the sphere moving faster) drag force should decrease or that the sphere should encounter less resistance as its speed increases! This, however, is not true and experiments do show that the resistance increases as the sphere moves with greater velocity. We seem to have an absurd theoretical result and yet no mistake in our analysis can be spotted. That solutions of the Euler equation of motion always show drag forces to be identically equal to zero was proved by a Frenchman of the name dAlembert. 3 The absurdity of the result was referred to as dAlemberts paradox. The paradox was resolved by Prandtl, the pioneer of Boundary Layer theory. The argument is complex and we will come to it after an intervening chapter. Briey however, Prandtl pointed out that the mistake committed is in the selection of the length scale. For now it is enough to know the central idea of the boundary layer theory. It hypothesizes that at high Reynolds numbers, viscous forces are concentrated in a thin layer near the stationary walls and are negligible outside this layer. Thus, the proper length scale relevant to the calculation of drag is the thickness of this layer rather than some other length scale even though the latter might be relevant ] elsewhere. Thus, in the example of the ow past a sphere, is not the correct length scale for calculating ows near the surface. The problem of selection of scales is more general and is discussed a little more in the appendix on singular perturbation problems.

16.3 Importance of irrotational ows


Feynman said that solutions to the Euler equation describe ow of dry water because these results will never match experimental results and hence uninteresting! If solutions of the Euler equation give absurd results, why solve it? As mentioned earlier in chapter 15, it is not possible to nd solutions to the Navier-Stokes equation in complex geometries and at high Reynolds numbers. However, it is possible that different limiting, and simpler, forms of the NavierStokes equation are valid in different zones of the ow domain. In such instances, it would be protable to solve the limiting forms in the appropriate domains, and patch them together to construct the solution over the entire domain. Thus, the Euler equation of motion is not valid inside the boundary layer, and hence not useful in calculating drag forces. But it is valid outside the boundary layer and would describe ow there. The solution to the Euler equation can be patched with the solution inside the boundary layer to construct the entire picture. Therefore it is important to know how to solve the Euler equation and we consider this in the next chapter before we turn to boundary layer theory.

Refer back to our comment about the French. To confound us, this name is pronounced Dalmbert, the way it is spelled!

162

Appendix: Singular perturbations


Problems, where a limiting procedure of simplication of the starting equations results such contradictions, are referred to as singular perturbation problems. Solutions to the Navier is a parameter and the Navier-Stokes equation are Stokes equation depend upon . Here referred to as parametric equations. Asymptotic solutions to such equations are obtained normally by expanding the dependent variable as a power series in the parameter or a perturbation series solution. Problems where this procedure succeeds are referred to as regular perturbation ows, the obvious choice is to expand dependent variables as power problems. In high u g . This method will fail under the present circumstances because the rst term of series in ug  , does not the series solution, i.e., the solution of the equation obtained in the limit of exist. Thus, the regular perturbation series fails to provide a solution. Problems where regular perturbation method fails are referred to as singular perturbation problems. Usually, this is a direct consequence of the the characteristic quantities selected for scaling. Though some scales may be the obvious choices, they might not be correct scales over the entire ow domain, as pointed out in our reference to the concept of boundary layer. Creeping ow past a stationary sphere is another example of singular perturbations. Re fer to the footnote in chapter 13 where the scales selected, viz and a , do not produce a uniformly valid approximation.

163

Chapter 17 MORE PHYSICS OF FLUID FLOW


Definition of vorticity. Interpretation of vorticity. Importance of vorticity. I break my promise made in the previous chapter to tell you about solving Eulers equation, very important as it is! We have to undertake a detour and learn about vorticity before we proceed to nd solutions of Euler equation. Very often in science, entities which have only indirect physical existence are used. Despite it they play a very central role in understanding science. Electrical eld vector is one such. Feynman wrote that the electrical eld vectors appear in his mind like little oating arrows which disappear when looked at closely! Yet the central role played by electrical eld in electromagnetic theory is apparent. Vorticity is one such concept which is as elusive as electrical eld, but is important in uid mechanics as electrical eld is in electromagnetic theory. A lot of this material is based on the material from the excellent book Natural Aerodynamics by R.S. Scorer, (pages 49 to 82 in the 1958 edition) and the primary standard text book by Batchelor (pages 264 to 282). You are strongly advised to refer to these since only rudiments are offered here.

17.1 Vorticity
Vorticity J is dened as the tendency for a uid element to rotate i.e., change its orientation. It is obviously related velocity. To get a feel for this, consider a little uid L andtoL its angular element with sides . Let the T axis pass through its center. Suppose its center of mass is not moving but it rotates around T axis with angular velocity . Where will be the uid element located after an innitesimal time interval ? See g.17.1. Let us examine where the corner points will move to. The velocity of any point due to the angular velocity is given by  where is the position vector of the point. The position vector of point D is given by i j . Hence the velocity of point D is given by j + i . Thus during time point D will move in the positive direction and positive direction by distances equal to the respective velocity multiplied by time . Let us denote by and by . Then point D will move to D as shown in g.17.1. Similarly it is easily shown that points A, B and C move with the same magnitude of distances along the and axes but their directions being as shown in g.17.1. The shape thus remains the same but the orientation has changed. The element has rotated. The angle between the original position of any line, e.g., OA and OA,

'

164

A B B 2 y O y x D C C 2 x D x A

Direction of rotation

Figure 17.1: Motion due to pure rotation . Thus if the local angular velocity is not zero, uid elements have a would be equal to tendency to rotate. But vorticity is the tendency of uid elements to rotate and hence is related to their angular velocity.

'

17.2 Why vorticity?


In this section we describe phenomena, where the role of vorticity is seen, to motivate the study of vorticity. We simply describe and do not explain the events. The idea is to show the important role played by vorticity.

17.2.1 Vorticity and viscous forces


If a uid is in uniform motion, uid elements do not slip past others. Hence we expect viscous frictional forces to be zero. Further, if a body of uid is has constant vorticity 1 , all elements will rotate with the same angular velocity around some axis. Hence a uid with constant vorticity rotates like a rigid body. In such a situation also, uid elements do not slip past others even though the velocity is not constant. Once again we expect viscous frictional forces to be zero. Thus one would guess that vorticity distribution would play an important role in determining viscous forces. An interesting problem arises here. Suppose motion starts from rest. Then vorticity is zero. We know that viscous forces are experienced as ow occurs. Hence vorticity must be generated and distributed non-uniformly in the ow eld, and a study of these is obviously important.

17.2.2 Vorticity and energy extraction


Consider a cylindrical uid element rotating around its axis with an angular velocity k . It is referred to as a vortex lament. Suppose by some means we do some work on it and stretch it. Let its length increase from its value W B by a small increment W . Its radius would decrease and hence its angular velocity has to increase by conservation of angular momentum. This is the well known example quoted by Physics text books as to how dancers increase their rotational speed by drawing their arms inward! See g.17.2. Let the radius change from its
1 We have not yet dened an equation for vorticity. But it will be shown a little later that vorticity is proportional to the angular velocity of uid elements. Thus constant vorticity is same as constant angular velocity.

165

1 1 2 2 > 1

Vortex filament Radius r1

Stretched vortex filament, radius r2

Figure 17.2: Increase in vorticity by stretching value p

kind of analysis carried out above, this change is shown to be proportional to . Thus by doing work on the vortex lament and stretching it, we were able to increase its kinetic energy. Hence if a vortex lament gets caught in motion on a length scale larger than it and gets stretched, it will extract energy from the surrounding uid motion. Such a mechanism is very important in turbulence. We require one more concept a little later and we will show it here as it is easily derived from the above ideas. Suppose the increase in the length of the vortex occurred in time . Then the rate of change in the vorticity of the lament is given by

p . By mass conservation, it is given by B K A p A p W Constant c or c p L W B W K Angular momentum per unit volume is proportional to p . By angular momentum conservation2 , the angular velocity will change from its value B by given by A L B p A B W pB W B K K The kinetic energy per unit volume is proportional to p . The kinetic energy of the rotating element, per unit volume, will also change since angular velocity has changed. Following the W B
after stretching by

Hence we can calculate the rate of change of vorticity of the vortex element being stretched, in a direction parallel to the direction of the vorticity vector, from the above discussion. It is given by I

B W W B Suppose two points A and B lying along say axis are separated by a small distance W B . Suppose both points move along the axis but with different velocities. The relative component velocity of point B with respect to point A is then given by I W B During a small time interval , the distance between the two points will change by W A I W B

4 !

Angular momentum is conserved only when viscous forces are absent. The effect that is being described will still be there, but the estimates may not be quantitative.

166

Putting it in words, the fractional increase in vorticity is therefore proportional to the stretching rate of a vortex lament.

17.2.3 Vortices in ow past bodies


A very interesting phenomena is observed in almost all ows past bluff bodies (as opposed to long and gently curved or slender bodies). Consider ow past a long cylinder. The ow is perpendicular to the axis of the cylinder. At low Reynolds number the ow is almost symmetric as shown in g.17.3. (You should look photographs. They are presented in many books on uid mechanics, for example, in Batchelor, after page 352.) As Reynolds number increases, the ow
Flow from left to right

MN MN MM N MM NMNM N N MNM N MN MM N MM N MM NNMM N N NM N N N M M MN NM N N NM MM N M N M N NN MN MN MMM NNMMNM N M M NM N N MM N MM N MM NNMM NNMM N N N N N


Re < 0.1

KL KL KK L KK LKLK L L KLK L KL KK L KK L KK LLKK L L LK L L L K K KL LK L L LK KK L K L K L LL KL KL KKK LLKKLK L K K LK L L KK L KK L KK LLKK LLKK L L L L L


Re ~40

O O O O OOO O O P PP P P P P PP

Re ~55

Q QR RR RR QRQ Q QR Q RQRQ

Re ~ 100

Figure 17.3: Formation and shedding of vortices in ow past a cylinder separates and circulatory ow sets in behind the cylinder3 . The circulating packets of uid are vortices and are stationary, i.e., stay behind the cylinder. As Reynolds number increases, the region behind the cylinder, called a wake, shows instability and begins to oscillate 4 . It turns out that the vorticity is generated by the cylinder and is concentrated in the wake region. It is like a thin sheet and is appropriately called a vortex sheet. The vortex sheet becomes unstable at high Reynolds numbers. At still higher Reynolds numbers, the vortices are swept from the cylinder or shed into the uid stream. However they form again behind the cylinder. Thus the process of formation of vortices and their shedding continues. The uid down stream of the cylinder thus contains these vortices which are shed and owing down. It is like a street with vortices and is called von Karman vortex street. One expects the vorticity and its distribution to play an important role in understanding many of these phenomena, some of which are not fully understood even now.

17.3 Vorticity and kinematics


Any motion of particles can be thought of as a combination of translation, distortion and rotation. The idea of translation is easy. It is the simple physical movement without change of shape, like the translation of a rigid body. It is easily represented by a constant velocity vector.
3 4

We have referred to this in our rst chapter on physics of uid motion. This is made visible by either injecting a dye or hydrogen bubbles.

167

Uniform motion of a uid is of not much interest and hence relative motion occupies our attention. Kinematics 5 of uids is a study of this. Any relative motion of one uid particle with respect to another can be thought of as a combination of distortion and rotation. Distortion refers to change in shape. For example a square can become a rectangle. See g.17.4. It is
y

Distortion without rotation

Rotation without distortion

Distortion with rotation

Figure 17.4: Deformation of uid elements obvious that when the square element has become a rectangle, no rotation occurred. Suppose a body of uid rotates around an axis with a constant angular velocity. This is no different from a rigid body rotating around an axis. This kind of motion does not result in any change is shape or distortion. The uid then is undergoing rotation without distortion. This is also shown in the g.17.4. Finally in the case where the square has been distorted, its length along only the axis has increased. It is a combination of rotation accompanied by distortion since we see that the length of the bounding lines originally lying along the axis has increased and they have also rotated in clockwise direction. Now we want to show mathematically that any relative uid motion can be described as a sum of rotation and distortion. Motion where all particles are moving with the same velocity is referred to as uniform motion. Relative motion is absent in it. Hence relative motion can be present only when velocity gradient tensor: I ^ = gradients exist. Let us therefore examine components of the velocity = = . The relative displacement of points separated by a vector = during time interval is then given by

! 4 4

I= 6 H I =V M L = V The rst term is already known to us. It is the component of the strain rate tensor . Let us denote the second term by a tensor 0 = . The relative displacement of points during time interval is then given by

We rewrite the velocity gradient tensor as

4
I

4 4

A L =

6 H I= M

I = =

4 4

4 4

= M =

= s =

We examine the relative motion due to each of these parts. The total relative motion is of course the sum of these three. To keep ideas in the center, let us consider two dimensional ows, i.e, # g g # K K only c c g K and 0 g K are not equal to zero.
Kinematics in general is description of motion by itself, i.e., specifying position of particles in terms of velocity and acceleration.
5

168

17.3.1 Shearing and straining motions


Let us consider the kind of relative motion generated by strain rate tensor. It has two types of terms: =@= and = In the rst kind, the direction I ^ of the velocity and the direction in which its change is being considered are the same, i.e., = = . The relative motion due to it is given by

4 4

I = == = =

Consider only one term to get ideas clear. A typical term is element . Consider an I8 ^ 6 L and L the with dimensions as shown in g.17.5 We consider motion only due to .
y y B A B x x A
2 x

4 4

I8 ^

D B A y
O

B B y

D C C

4 4

C
A

D x C C

x A
A

Movement due to dv x / dx

Movement due to dvx /dy + dv y / dx

Movement due to dvy /dy x- dv x /dy

Figure 17.5: Relative motion due to components of strain rate tensor Further let us assume for simplicity that is a function of only. The vector between point C and A is given by 2i . The relative motion of point C with respect to point A due to this term is given by I Thus, after time , the point A will move by distance relative to point A, by distance I

I L L

4 4

4 4

! !

I l

! while point C will move,

This is shown in g.17.5. I Thus I the points, will increase ^ the original distance between . Since we assumed that is a function of only, point D according to the value of will I0 ^ also move by the same amount with respect to point B. Thus I the element will elongate if is positive. This is pure extension. Thus terms like = ^ = contribute to pure strain of elements. I ^  6^8L , I ^ Now consider relative motion due to a term like . Since we are M I A  considering two dimensional ows, we have and dependence on T coordinate is zero. Hence from our general formula, we nd that the relative displacements of points separated by relative position vector M is given by

4 4

4 4

4 4 4 4 !

4 4

relative displacement

A A

that of

q . 6 p
TS TS

TS will be nonzero if is not zero due to continuity equation. Its effect, which will be similar to , has to be added to get the total effect.

p 6

I0 6 H I L M M

y V M y

'

169

and shown in the second frame of Consider a square differential element with sides g.17.5. Let us locate the origin at the center of the differential element. We consider point O to be our reference and gure out where the other points move with respect to it. Thus we have to nd out the relative velocity of the various points with respect to point O. Consider point C. Its relative vector is given by . Hence the it will move relative to point O by distance y M Thus the point will move in the negative direction and positive
proportional to . Similarly point B will move by

'

direction by equal amounts

direction by equal amounts proportional to . Similarly it is easy to work out that point A will move in the negative and directions, while point D will move in the positive and directions by the same amounts. Thus the new shape is formed by elongation along the AD diagonal and contraction along the BC diagonal as shown in g.17.5. Thus the square has become a rhombus and this is equivalent to motion due to pure shear.7 Thus the strain rate tensor produces shearing and straining motions. We expect all these motions to create relative motion between elements and hence cause viscous forces.

y This point will then move in the positive direction and negative

17.3.2 Rotational motion

and dependence on T coordinate is zero. Thus we are looking at the relative displacement due to

I= V I  We are considering two dimensional ows, we have A

Now consider motion due to the second group of terms in the expression for relative velocity

6 H I L =

6 H I0 L

From our general formula, the relative displacement of points separated by relative position vector M is given by relative displacement

I V

A A

I 6 H I M s L V u 0 M 

It is helpful at this juncture to look back and notice the similarity between this formula and the formula we derived for displacement due to rotational velocity at the beginning of this chapter: L k r. Now consider the square element with side shown in the third frame of g.17.5.
Note that while line AB has rotated clockwise, line AC has rotated counterclockwise. Hence the net rotation is zero.
7

170

s M Thus the point will move in the positive and directions by an amount proportional to 0 . Point B will move by y 0 Thus it will move in the negative and directions by an amount proportional to 0 . Point A will move in the positive direction and negative direction while point D will move in the negative direction and positive direction, all by the equal amounts proportional to 0 .
0

We once again consider of points with respect to point O. The vector of point C 5 movement with respect to O is . Hence the relative displacement is given by

'

'

The result is rotation around an axis perpendicular to the paper ad around point O as shown in g.17.5. Thus the antisymmetric part of the velocity gradient tensor creates rotational motion. To summarize, the relative motion of points is a sum of pure strain, pure shear and rotation. Of these, rotation is represented by antisymmetric part of the velocity gradient tensor. These terms do not cause relative motion and hence will not contribute to the generation of viscous frictional forces. It is for this reason that the deviatoric stresses were assumed to be proportional to the symmetric part of the velocity gradient tensor, e.g., in Newton Stokes law. Let us get back to the antisymmetric part of the velocity gradient tensor. It must be related to vorticity. The particular term we considered is causing a rotation around the T axis or this term is equal to the T component of the angular velocity. We can therefore write

T component of the angular velocity A L

6 H I0

We could similarly write the other components of the angular velocity also. It is easy to recognize that the left hand side of the above equation is equal to half of the T component of the curl of the velocity vector. Vorticity J is dened as
J

I V

and it is equal to twice the angular velocity vector caused by the velocity gradients in ow elds.

17.4 Vorticity equation


We can use the Navier Stokes equations (i.e., we are considering an incompressible and Newtonian uid) to derive an equation for vorticity. We simply have to take the curl of those equations! Skipping a lot of algebra, we get the following equation:

A J m  M

$b

hb
K

(17.1)

17.4.1 Diffusion of vorticity


We jump a little ahead in order to interpret this equation. Consider the Fouriers law of heat conduction

A j 
171

where q is the heat ux by conduction. Recalling our interpretation of divergence of a vector, K  is therefore, m is the net rate of input of heat by conduction per unit volume. Thus j the net rate of input of heat by conduction. This expression is typical of all diffusive processes, like conduction of heat, diffusion of species etc., and The net rate of input by diffusion per unit volume is given by the transport property multiplied by the divergence of driving force for transport. Driving force for diffusion is the gradient of some quantity, e.g.,  for heat. Now we know how to interpret two terms in the vorticity equation. We already know that the left hand side is equal to the rate of change of vorticity observed when moving with the uid. Further, we have already seen that kinematic viscosity is a diffusivity. Thus the second term on the right hand side is equal to net input rate of vorticity by diffusion. Thus if the rst term on the right hand side is equal to zero, the equation simply is a statement of conservation of vorticity: the rate of accumulation of vorticity is equal to the rate at which it is diffusing into the system. One can get a glimpse of the usefulness of vorticity as a concept since it behaves very much like a diffusing species.

b C

17.4.2 Amplication of vorticity

. Take any Now consider one component of the rst term on the right hand side, e.g., J m point P. From that point draw a vector tangent to the vorticity vector there. Take another point Q a distance : away from P and parallel to the vorticity vector. The unit vector along the vorticity vector is given by J / . Then the vector pointing from P to Q is by :UJ / . I8given Hence I0 the component of the relative velocity between points P and Q is given by ( : J . )/ . Thus I

$b

 I[

Since points P and Q are parallel to the vorticity vector, we can think of it as a tiny vortex
Vorticity vector Q P

J m  I A :

Jb

Figure 17.6: Vorticity amplication due to velocity gradient lament. Thus this term is proportional to the relative velocity between ends of a tiny vortex lament. The relative velocity stretches the lament and hence the vorticity gets amplied, as we have discussed in a previous section. Thus this term represents the rate of amplication of vorticity due to vortex stretching.8

17.4.3 Conservation of vorticity


The whole equation can now be interpreted. Vorticity in a ow eld convected due to velocity elds, created by vortex stretching, and diffuses due to gradients in vorticity. In some places it
Suppose we have two dimensional ow, i.e., only two nonzero components of velocity varying in two spatial dimensions. Then the vorticity vector is perpendicular to the plane containing the velocity vector and hence this term is zero. Thus amplication of vorticity is not possible in two dimensional ows and vorticity is only redistributed by diffusion.
8

172

may get amplied due to vortex stretching. If gradients in vorticity are created in this or some other manner, diffusion tends to even out the gradients.

17.4.4 Generation of vorticity


Now we return to one nal point. Suppose motion starts from rest as almost all motions have to. The rest state has no vorticity. The vorticity equation tells that vorticity is conserved and can only get redistributed. Thus vorticity in ow elds will have to be zero and hence viscous forces will have to be zero if it is not generated some where. Hence vorticity must be generated some where. Motion usually begins by some solid body moving in a uid. By no slip condition the uid immediately next to the solid acquires the same velocity. Consider rst the effect of no slip in the tangential component of velocity. Because of the no slip condition, a step change in tangential velocity is created in the uid. This creates a huge velocity gradient and hence huge vorticity. All this vorticity is contained in a zone near the wall. It is a thin uid lm where large vorticity is concentrated. It is called a vortex sheet. Thus a large gradient in vorticity is also generated. Vorticity then diffuses from the vortex sheet into the uid. The normal component of the velocity sets off a compression wave and only an irrotational ow due to this component is established. Thus it is the tangential velocity that is of importance in creating vorticity.

17.5 Conclusion
We discussed the concept of vorticity. It is related to the rotational motion of uid. The relative motion of points can be thought of as sum of straining, shearing and rotational motions. Of these, rotational motion can be identied with vorticity. For Newtonian uids 9 , viscous forces exist only when vorticity gradients exist. Vorticity is not generated in ow eld. It is either convected, amplied, and redistributed by diffusion. Vorticity is usually generated at a solid wall due to the continuity of tangential components of velocity when ow is started by solids in motion. Flows where vorticity is zero are referred to as irrotational ows, and these ows can then give us solutions to Eulers equation. This is the subject of the next chapter.

The situation is more complex for non Newtonian uids, and viscous forces, i.e., equal to zero for irrotational ows.

V are not identically

173

Chapter 18 IRROTATIONAL FLOWS


We discuss solution of the Euler equation.

Stokes equations is caused by the nonlinearity of the equations. The nonlinearity is of course in the inertial terms, which is present in the Euler equation also. However an ingenious trick is used to obtain solutions to the Eulers equation: dont solve it! Let us elaborate. Suppose functions v which satisfy the following equations are found:

t (16.3) $ b b 4 4 The basic difculty is obtaining solutions to the Navier or to determine irrotational ows.
M m  A P 6 M P O

Let us now turn our attention to solution of the Euler equation (16.3)

b K A c

and

A bm v

Then, for a Newtonian uid, the equation of continuity and also the condition that net viscous forces per unit volume be zero are satised. The discussion on vorticity connecting it to viscous forces showed that irrotational ows satisfy the second condition. Thus, irrotational ows that satisfy the equation of continuity also, i.e, irrotational and incompressible ows, are the ones which satisfy both the conditions. Since velocity is known, pressure can be determined by substituting velocities in the Euler equation, i.e., (16.3). Thus, we now focus our attention on nding solutions to incompressible and irrotational ows.

18.1 Velocity potential


Flows where the tendency for uid elements to rotate is absent are irrotational ows. Hence, irrotational ows obey the following equation:

A 

When velocity can be represented as the gradient of a scalar potential, , the above equation is identically satised. is referred to as the velocity potential and the relation between velocity and potential is normally dened as (18.2) Thus any ow which can be described by a velocity potential is an irrotational ow. 174

(18.1)

b(

Velocity elds in incompressible uids have to also satisfy the equation of continuity:

bm

A 

Substituting for velocity in terms of potential we nd that

b K(

A 

(18.3)

This is the well known Laplace equation, for which many solutions with a variety of boundary conditions have been worked out. Thus by solving the Laplace equation, velocity elds for irrotational and incompressible ows can be determined. Such velocity elds are also known as solenoidal elds. Now let us formally prove that for incompressible uids, the viscous forces are zero when the ow is irrotational. Consider the vector identity:

b b

 A

b b m  b m$b b
K A 

If the ow is irrotational, the left hand side is zero. The rst term on the right hand side is zero for incompressible uids. Hence we obtain

m  A

b $b

M m  A P 6 M (18.4) O is obtained. Thus incompressible and irrotational ows satisfy the Euler equation or the limit . ing form of Navier Stokes equation when

This is of course the term that corresponds to the viscous forces per unit volume. Thus if this is substituted in the Navier Stokes equations, the Euler equation

4 4

$b

18.2 Determination of pressure


The solution of a ow problem is complete only when pressure is also determined. Thus the procedure of determining the pressure, once the velocity potential is known, has to be specied. The inertial term in the equation can be put in terms of velocity potential using the following vector identity: 6

m  A L

$b

m  

For irrotational incompressible ows hence

m  A 6 L

Jb

m  IK 6 A L P O M

t b b b 4 4 Substitute for velocity in 4 terms of velocity 4 potential, and after a little rearrangement, the following equation is obtained: H IK ( b 4 M L M OP V Awt 4 175
M m  A M

Substituting this in the Euler equation, it is found

It can be seen, in principle, that since and hence are known, the above equation can be integrated to nd the pressure. It can be put it in a form that looks familiar. Suppose that the only body force is that due to gravity. The body force can be written as the gradient of gravitational potential energy. Let the direction of T coordinate be opposite to the direction of gravity. Then, where R is the magnitude of the acceleration due to gravity. The following result is obtained when this is substituted in the above equation:

I K ^8L

tA b

RUT

This equation implies that the term inside the brackets does not vary in space or

is just a constant, and This is the Bernoulli equation for unsteady state. At steady state, Z the equation is of course the Bernoulli equation with all the readers are familiar. Hence, once the velocity eld is known, pressure can be determined by using the above equations 1 . Now consider some examples to let the theory take concrete shape.

IK  M L M O P MRUT8V A

IK l M L M O P MRUT A Z

(18.5)

l

18.3 Irrotational ow past a cylinder


Consider steady uniform ow approaching an innitely long circular cylinder. No changes are expected in the direction parallel to the axis of the cylinder and and let that direction be T . See
y

U x

Figure 18.1: Irrotational ow past a cylinder Figure 18.1. Thus the following equation has to solved:

p p p p
18.3.1 Boundary conditions

4 4

(V

6 K  M pK vK A

The above equation has to be solved with the following boundary conditions. Far away from the cylinder, the velocity is given by

(1

In unsteady conditions, W is needed. It is related to the cause forcing the velocity eld. For example it could be varying pressure eld which causes ow and that will allows determination of W .

A 

as

(-

176

This implies that

I  U p A r A  o vc

These equations can be integrated to nd the limiting form of the velocity potential. It is given by A p ,8 v as p The other boundary condition is at the surface I of the cylinder. The no-slip condition with respect to the tangential component of velocity, t , has I to be sacriced. The no-slip condition with respect to the normal component of the velocity, r is chosen as the boundary condition. Hence the boundary condition is I  or

and

I p v A t A  v

as

18.3.2

4p 4 Solution by separation of variables


p Y X v  X  p  and v are found to be 6  H  6 K A    K p V p p p p X  vK M KX A 

r A

A 

The solutions can be obtained by separation of variables. The solution should be of the form

(A 3 After separating variables, the equations for 3


and

The solution to the last equation is


X

A ? o v M ] ? v In view of the periodic nature of the problem in the v direction, the boundary conditions, and

the orthogonal nature of the cosine and sine functions, the solution can be expected to simplify to X

Thus absorbing the constant into

Substituting this,

( A 3 p  ,8 v should satisfy the following ordinary differential equation:


6  H  6   p p p p V pK

A g, 8 v

, the solution should be of the form

The solution to this equation is

A 

A
p M p
177

or

is equal to

(
A A

H H

8 v
p M p V , 8 v p M p V , H

from the boundary condition far away from the cylinder. Thus

Constant can be determined from the other boundary condition that gradient of p A or the surface of the cylinder. This gives

is zero at

A 

K 6 M K V p o8 v p

Note thatI two boundary conditions are sufcient to nd the solution. Thus, the no-slip condition on t could not have been satised. This is in accordance with our earlier discussion. The velocities can be easily found. They are given by

and

I A H 6 KK o r  p V 8 v H K I A 6 t  M pK V v

18.3.3 Determination of pressure


The Bernoulli equation can be used to nd the pressure. For this, the conditions far away from the cylinder are needed. The velocity there is  . Let the pressure far away from the cylinder be . Hence, if effects of gravity are not important, Bernoulli equation can be written as

IK K  LYM O P A  L M OP

Thus, pressure can be calculated from:

pcv  O O A A A

P A L  K I K p cv 

(18.6) (18.7) (18.8)

The value of v.v is needed, which is easily done since the velocities are known:

L m

IK L 6 IK IK L r M t K H6 K K ,8 L L  v M p V  pK

It is interesting to calculate the variation of pressure on the surface of the cylinder. The kinetic energy on the surface of the sphere is given by

P K 6 K X v L

178

Flow direction

+ = = 2 /3 -

=/6 = 0

Figure 18.2: Variation of pressure (in potential ow) on the surface of the cylinder

and the pressure variation on the surface is given by

pcv  O O

P K 6 K X v L

The variation of pressure is shown in Figure 18.2. In particular, note the adverse pressure gradient, i.e., pressure gradient opposed to ow direction, in the rear of the cylinder. If the uid elements near the surface have as much kinetic energy as calculated earlier, they would be able to overcome the adverse pressure gradient, and continue to ow in the direction. However, if the no slip condition is satised, the kinetic energy would be smaller, and hence the elements near the surface would not have enough energy to overcome the adverse pressure gradient. It is for this reason that ow separates.

18.3.4 Drag force on the cylinder


Let us calculate the drag force on the cylinder is zero for a Newtonian uid. The drag force is the force exerted by the uid in the direction of ow. Hence if its magnitude is ` , it is given by the surface integral

7
` A A A K

 v s q rom m  r

,  v O M c rr  qsrom M c r#tqutom  r K ,  v O M c rr  o8 v c r#t  vH r


 

JD

(18.9)

First let us consider the pressure term. Substituting for pressure, we nd

v , 8 A O v A


 O

K  K K  P K    v M  L v o v 6 X v o v Z

179

Since the uid is incompressible and Newtonian, at p

rr A A A

Further, at p

A I r

pK  L _ pK _  X o8 v

L_

4 ( 44 4  4 4

c r#t A
_

H 6 I I r M p tV p p p v

The rst term is zero since

is zero every where on the surface. Hence

c r#t A

_  X v

Hence the contribution to drag force due to the deviatoric stresses is

v 8 rr , v rt v  r

c 

EXERCISES:

_ X  

v o 8 L v

Thus, the drag is zero. This of course is the dAlemberts paradox.

A 1. How would the above solution change if the pressure far away were to be varies as O far B M
? 2. Find the velocity proles for irrotational ow past a sphere when ow Oaway from it corresponds to uniform ow. 3. What ow is represented by (i) A\[ v in cylindrical coordinates. This is called a free vortex or a line vortex with strength equal to [ . (ii) A < p in cylindrical coordinates. This is called a line source with strength equal to < . (iii) A K r in spherical coordinates.  Sketch the stream lines in each of the cases.

( ( u

18.4 Velocity potential and complex variables


Irrotational ow solutions can be found using the theory of complex variables. Recall the denitions of velocities in terms of both stream function and velocity potential:

I0 A A I0 A A An analytic function ` of a complex variable T satises the Cauchy Riemann conditions. Thus if ` A Z M R and T A M , Cauchy Riemann conditions state that Z A R c Z A R

4(

4 4

4 180 4

Thus if we dene

then, Cauchy Riemann conditions are identical with the relationships that connect velocities with the stream and potential functions. Thus any analytic function is a solution to some irrotational incompressible ow. Thus taking any complex analytic function, and working backward to nd the boundary conditions satised by it, one can gure out which ow conguration the stream function corresponds to! One such example is given in the text Bird, Stewart and Lightfoot.

( M ^

18.4.1 A few important planar ows


Before leaving this section, let us examine a few important potential and stream functions. Uniform ow

A  T Hence   and A  M A  M or A Hence I A I0  c and A  The ow corresponds to uniform ow in the positive direction, and the stream lines are straight lines parallel to the axis. The pressure remains constant since the velocity is constant.

Consider the analytic function

Flow from a source or to a sink Consider the analytic function Hence

( ^

where the radius and angle description for the complex variable has been used. Firstly, it is easy to see that stream lines, i.e., lines for which v is constant are radial lines. Thus ow is toward or away from the center. From the velocity potential

< A Lk T < <  M A L k p M Ev or A L k p

and

A < Lv k

Thus ow is radially outward if < is negative, and it is ow from a line source. If < is positive, the the ow to toward a line sink. < is referred to as the strength of the source or sink. The L area though which the ow takes place per unit length of the source or sink is kp , and hence < is equal to the ow rate per unit length. The equation of continuity is derived on the basis of law of conservation of mass. Yet in this ow, there is ow in a single direction, originating from a point2 . Hence the mass has to come from the origin. Hence this ow is referred to as ow from a source. The reverse is true
2

I A < r L~ kp

or

I A  t

For the potential written, the origin of the coordinate system coincides with the location of the source

181

for a sink. Thus we can think that m v=0 every where except at the origin, where it suffers a singularity. Note that the radial velocity decreases away from the center. This has to be so, since the volumetric ow rate generated by the source or sink is constant while the area through which ow is taking place increases with increasing radius. Pressure can be determined from Bernoulli equation. Pressure falls toward the center as velocity increases and attains a constant asymptotic value far away from the center. Flow toward a sink can be used to model ow in atmosphere toward a depression. Flow around a line vortex Consider the analytic function Hence

calculated from the velocity potential

A L< p I k It is easy to see that the stream lines are circles now, and hence only t is not zero. It can be

( M

A L < p M Ev  k

< A L k T
or

A< L v k

and

I A < t L~ kp

The angular velocity decreases away from the center, and it goes to innity at the center. Thus the pressure goes to minus innity at the center and reaches an asymptotically constant value far away from the center. Let us examine the character of this ow a little more. Using the Stokes theorem relating the surface integral of curl of a vector to the line integral of the vector along the path that is bounding the surface:

, bm 
 A , ,

^]

m`_ba

To specialize to the example being considered, take any circular area of radius on the plane, centered around the origin. Then, the left hand side of the above equation is an integral on the circumference of the circle or is equal to

K I  t   v A <

Thus the integral is constant and independent of the radius of the circle chosen. Yet h is zero every where. This has to be interpreted as if all the vorticity is concentrated in a line at the origin, just as all the mass ow came from a line in the previous example. A line where vorticity is concentrated is called a line vortex. This can be visualized as follows. Suppose an innitely long cylindrical uid element is made to rotate around its axis like a rigid body with a constant rotational velocity . The vorticity of this uid element per unit volume is given by k where k is the unit vector along the direction of its axis. In the limit of the radius of the uid element tending to zero while maintaining its vorticity per unit volume constant, a line where vorticity is concentrated or a line vortex is obtained. Thus the velocity eld derived corresponds that caused by a line vortex. Flow around a tornado, a little away from its center, is like the ow around a line vortex.
182

Before leaving this example, one more issue need to be claried. Vorticity refers to the tendency for uid elements to rotate or to change their orientation at a at local level. In the ow being considered, uid elements are going round in circles but they are not changing their orientation. This behavior is shown in Figure 18.3.

Fluid element moving in circular paths without changing orientation.

Fluid elements moving in circular paths and changing their orientation.

Figure 18.3: Distinction between local rotation and global rotation of uid elements.

1. Which ow is represented by (i) A


T , (ii) A
T (iii) A
, (This is known as a doublet. It is as if a source and sink are placed adjacent to each other.) (iv) A  TMdK c T   body. cK . Find the shape of the stream line corresponding to A . This is called half Rankine

EXERCISE:

= (v) A
T M M K

? Sketch the stream lines in each case.

18.5 Stream function in potential ows


The potential and stream functions are the real and imaginary parts of a complex analytic function respectively. It is therefore interesting to nd the partial differential equation satised by the stream function since it is some times more convenient to solve for it rather than the potential function. The stream function identically satises the equation of continuity. Hence, the relationship to be satised by the stream function so that vorticity is zero should give the desired differential equation. As stream function is being used, only two dimensional ows I are being considered. Consider Cartesian coordinates, and planar ow in plane, i.e., = 0 and the ow does not depend upon T coordinate. The only non-vanishing component of vorticity is the T component and hence for the ow to be irrotational, the following equation has to be satised:

 A

Now consider planar ows I but which are more conveniently considered in cylindrical coordinates. Here, once again, = 0 and the ow does not depend upon T coordinate, and the 183

4 4 Hence, the stream function also satises the 4 Laplace equation for irrotational incompressible 4 ows.

The relationship between the stream function and velocities can be substituted into this to obtain K K

 A K

I m

only non-vanishing component of vorticity is the T component. Hence, the following equation must be satised by an irrotational ow:

 A 6 I  6 Ir p p p t p v m

Substituting for the velocities in terms of the stream function, it is found that

6 K  A 6 Hp p p p VSM p K v K

But the right hand side of the equation is equal to the Laplacian in cylindrical coordinates when T dependence is not there. Thus, the stream function also satises the Laplace equation for all planar ows. It turns out the stream function satises a slightly different equation I A for axisymmetric ows. and dependence on v Consider cylindrical coordinates rst. In axisymmetric ows, t direction is absent. Hence, only v component of vorticity exists and for irrotational ow,

4 4

 A t A

Substituting for velocities in terms of Stokes stream function (refer to chapter 13)

 A 6 KK p T M

A new operator , similar to the Laplacian operator, is dened to put this equation in compact form: K K 6

4 p K p 4 p M 4 T K A  K Thus, is equal to the non-zero vorticity 4 component 4 4 in axisymmetric ows. Similarly for axisymmetric ows which are conveniently considered in spherical coordinates, the ( component of velocity is zero and the dependence on ( coordinate is absent. Here the only nonzero component of vorticity is the ( component. This gives the following denition of the opera K
3

r T

H 6 p p p V

tor and the equation obeyed by the stream function:

18.6 Solutions based on the e operator


It must be said that solving for velocity potential or stream function are equivalent procedures. Either can be chosen based on convenience. The Laplacian as well as the operator are linear. Hence sum of several solutions is also a solution of the Laplace equation. Several complicated solutions can be built using this principle. Here this technique of obtaining solutions is illustrated. ^ o 6 o v  . This satises the irrotational incompressConsider the stream function
X k ^ K ibility condition. The stream function gives rise to only radial velocity given by
X k~p . This corresponds to ow to a point sink or a point source depending the sign of
, where the 184

4p 4

K M  K v
p

 4 4 4 4

H 6 A  V v v v

Angle= - r z h r

Stream lines for a source located at the origin

A dipole

Figure 18.4: Source and a dipole in spherical coordinates

sink or the source is located at the origin. See Figure 18.4. If a sink and a source of equal strength
are kept at a distance , and the distance is made to approach zero while keeping the product
is kept constant, the resulting conguration is called a dipole. The product
^ X k is called the dipole strength _ . The stream function corresponding to a source and a o   ^ o 8 , 8 vfn v . See the Figure 18.4. The limit sink separated by distance is given by
X k of this as tends to zero can be obtained as shown below. The limit can be rewritten as

A A

A A

  

A _ v L p K K v p It gives stream lines which look like corresponding to ow around a sphere provided the stream , where is the radius of the sphere. Thus the stream function is set to be zero when p A function is found to be 6 A  L K v L p K K v p

Note that the above procedure is equivalent to differentiation with respect to any coordinate. Derivative of any irrotational incompressible stream function will denitely satisfy the equation corresponding to irrotational incompressible ow, and we could have derived the stream function for a dipole directly by differentiation. Stream functions for irrotational incompressible ows can also be obtained in this way. 6^8L  p K K v . It is easily verConsider spherical coordinates and the stream function ied that this satises the irrotational incompressibility condition. This stream function was encountered earlier: it corresponds to uniform ow in the T direction. A dipole can be combined with uniform ow to get a new stream function corresponding to yet another irrotational incompressible ow: K 6

o8   v
v f[ o Xk g f[ _ X k
_ Xk
K T _ v p

     44

(18.10)

Ah for the tyranny of notation! I dont know what I would do if the potential function was a function of this angle.

185

The pressure can be found using Bernoulli equation. This then corresponds to irrotational incompressible ow past a sphere. EXERCISES:
[ 1. Which ow is represented by v in cylindrical coordinates? Sketch the stream lines. Find the force, normal to the direction of ow, on the body over which the ow is taking place. 2. Consider the axisymmetric stream function in cylindrical coordinates. What ow is repre-

A  6 M r p o8 v

sented by

g K K K A Kg  p ? What ow is represented by A K  p  6

Sketch stream lines for this stream function. This is called a Rankine body. It can represent ow past a bluff body.

KT p K  T M

186

Problems for Chapter 18.


18.1 i) A solid sphere is placed in a uid. The uid is less dense than the solid. Suppose that we assume that ow in the uid is potential ow. Can the solid reach a terminal velocity? Give reasons. ii) Suppose we replace the solid sphere by a gas bubble. The bubble will rise since it is lighter. Let its velocity be denoted by a If an observer moves with the bubble, he will nd the uid to be owing past him with a velocity a . The viscosity of gases is very very small and hence the shear stresses at the bubble-uid interface can be equated to zero. Does potential ow past a sphere that was derived in class (also given in book, see page 149) approximate the ow past the bubble? Give reasons.

iii) A gas bubble is placed in a large amount of uid. We want to nd its response to variations in pressure far away from it. The bubble will expand or contract as the pressure falls or increases. The response will be linear if the pressure amplitude is small. However, at high amplitudes, the bubble experiences greater inertial effects when it contracts. Due to this, the bubble can collapse. This phenomena is referred to as cavitation . We want to derive an equation to l  compute the radius of a bubble, , in response to pressure variations. Let A the pressure far away from the bubble be given by . We M make the following assumptions. (i) Motion is purely radial. (ii) Gas in the bubble is ideal. (iii) The temperature remains constant every where. This will be in correct when the bubble contracts very fast. (iv) Fluid is inviscid.

a) Determine the potential function for purely radial motion. The uid far away from will be stationary. Denote the velocity of the bubble  the ^  bubble A ; . These two conditions should give the boundary wall by conditions needed to completely determine the velocity potential. l in eq. 18.5 in this problem? b) What is Z . What is the pressure in the c) Let the radius of the bubble be B at bubble at any given time in terms of ? and derive an ordinary differential equation for d) Apply eq.18.5 at p A .

This equation was derived by Lord Rayleigh. If viscous effects are included, the resultant equation is referred to as Rayleight - Plesset equation. When either equation is solved, it shows the highly nonlinear response of the bubble.

187

Chapter 19 BOUNDARY LAYER THEORY


We show how the paradox of of no drag at high Reynolds numbers is resolved. Frictional forces are concentrated in a thin layer near walls. The theory makes good predictions of drag based on this idea.

188

Chapter 20 STABILITY AND FLOW TRANSITIONS

We have dealt with the methodology of solving uid mechanics problems. We found that 6 , the problems have well dened solutions, though they may be mathematically when complex. As Reynolds number increases, we have learnt to deal with balance of diffusion and convection of vorticity. As very high Reynolds numbers, further simplications were made by using the concept of boundary layers. Usually an increase in Reynolds number is accompanied by a change in character of ow. Superimposed on the main ow, circulation patterns can appear as Reynolds number is increased. These are called secondary ows. Further increases in Reynolds number can result in a very chaotic ow. This is called turbulent ow. The transition from laminar ows to turbulent ow need not go through the intermediate secondary ow transition. The transition affect the gross phenomena. Consider data obtained for ow of water in a tube by Reynolds shown in Figure 20.1. If we visually inspect the tube and the contents while ow occurs, we will nd no change at all, even through the graph shows a qualitative difference. Reynolds became a pioneer in tracer experimentation by introducing a thin stream of colored uid at some location in the tube. The colored tracer uid made ow visualization possible and showed that the ow was regular at low ow rates while it was chaotic at high ow rates. The rst type of ow is referred to as laminar while the latter is known as turbulent. One can perhaps state that The unsolved problem of classical physics concerns prediction
Plot is on log-log scale

Slope ~ 3 Power Slope ~2

flow rate

Figure 20.1: Flow transition as reected by sudden increase in power consumption as ow rate increases 189

of the characteristics of turbulent ow and, often, the factors that govern the transition from laminar to turbulent regimes. We will return to this idea later. Notice in Figure 20.1 that the change in the pattern of power consumption, as reected by the change in the slope of the curve, is abrupt. That is why it is referred to as a transition. It is reminiscent of phase change. Laminar-turbulent transition has been thought to be, and many might still feel so, similar to many other transitions that occur between several states depending on the stability of the state. In uid ow, examples of such transitions, which have been explained by stability theory, are many. The idea that laminar-turbulent transition is actually a series of transitions is quite prevalent. Thus it is worth while to study how instabilities arise.

Figure 20.2 is unstable. It can be shown by the following argument. Suppose due to random

20.1 Instabilities due to density/gravity density gradient and gravity. This kind of Instabilities can arise due to the combination of instabilities are referred to as Rayleigh Taylor instability1. Take a very simple example. ConP P . We know that the state shown in the sider a stratied layer of two liquids A, B. Let
A B A B A B

A A B

Figure 20.2: Instability due to buoyancy disturbances, which are always present, a dimple as shown in the second frame of the gure is formed. The dimple so formed is now present in an environment of higher density. Thus it will move upward under the inuence of buoyancy force proportional to the volume of the dimple. However as the dimple moves up, its volume increases as shown in the frame below it, and thus the buoyancy force acting on it increases. Thus the dimple keeps on moving upward and gets bigger. The nal result will be the stable state where B is present on top of A. One can contest this argument by saying that the result may depend upon the assumed shape of the dimple formed. Let us consider therefore what happens if the dimple is formed in an exactly opposite way as shown in the third fame of the gure. Now the dimple is in a lighter uid and hence will move downward under the inuence of buoyancy force whose magnitude is proportional to the volume of the dimple. Thus as it moves downward its volume and hence the downward force acting on it increase as shown in the frame below it. Thus downward motion continues till inversion occurs. Another objection can still be raised. We might still reach the conclusion that inversion will occur even if one started with a stable state, i.e., B is present above A. It will be left for you to show that this objection can be overruled and B resting above A can be shown to be stable by the same type of arguments.
1

D.W. Sharp, An overview of Rayleigh Taylor instability, Physica D, 12, 3 (1984)

190

We are not done still! A nal objection can still be raised by those people who know physical chemistry. Consider a curved dimple of the type we assumed will form but in a vessel of small size. Due to interfacial tension a downward force is generated on a dimple protruding upward. This force therefore will oppose the upward buoyancy force and inversion may not occur! However we can claim that inversion will occur if buoyancy force is greater than the force due to interfacial tension. Suppose we assume that the dimple is hemispherical in shape. Then, inversion will occur if
A B A B Down ward force due to interfacial tension

r is the radius of the tube

Figure 20.3: Effect of size of vessel on instability due to buoyancy

or

P P  L k~p ~ k p R P ^ p RK

is the condition of inversion. Note the physical interpretation of the left hand side. It is the ratio of buoyancy force per unit volume to the force due to interfacial tension/volume. It is therefore a dimensionless group, called Bond number, . The numerical value 3 obtained in above equation depends upon the particular shape assumed and hence may not be correct. We can modify our argument as follows. Let be some length scale typical of the disturbance. Then P  Buoyancy force Ag R

Force due to interfacial tension A K P  K g  R P  g C ^ KR K

Hence inversion would occur if or

Ei

Ei

where g c K and are some constants whose values are specic to particular situations and can either be calculated for simple cases or be measured experimentally. The constant is the critical Bond number . Thus inversion occurs if

E i

 P  ^ KR

E i

Conversely if we have a vessel which does not permit disturbances larger than some value such that the Bond number is always less than , inversion will be prevented! Notice the power of the results. Just by thinking in terms of forces, we could reason our way to a very good criterion and it forms a powerful predictive criteria when coupled with 191

some experimentation. Another very important feature of the above is the appearance of nondimensional groups. We have seen that many non-dimensional groups appeared in uid mechanics. They are pregnant with physical signicance and with proper understanding, powerful predictive criteria can be developed. It may also be noticed the way experimental effort is reduced by using non-dimensional groups. Another feature worth noticing is the way a typical estimate of a physical quantity characteristic of the phenomena, in above it happens to be the length scale of the disturbance, was used. We have seen this earlier, and once again, this is a powerful feature of uid mechanics literature, encountered very often. When selection of such typical quantities is done with physical insight, the results obtained are both beautiful and powerful. A nice application of the above instability we just discussed is to falling lms on inclined planes. Note that this type of ow occurs on packing in packed towers, humidication towers,
Length scale of wave isl

g sin g (This is the buoyancy force as discussed earlier

Figure 20.4: Instability in falling lms wetted wall columns etc. Now suppose a wave of length scale develops on the surface. P As shown in the gure, the vertically downward gravitational force R can be resolved into P R ,8 components: normal and parallel to the inclined plane. Their values respectively are P and R . The former tends to pull the wave crest down, i.e., it tends to dampen the wave. The latter destabilizes the ow by making the wave propagate. Thus we can write a stability criteria for wave propagation on thin falling lms:

where some constant. propagate, and is similar to the above result. Exact analysis including viscous and surface tension effects has been performed and shows that thin lm on vertical walls are unstable provided Reynolds number exceeds a critical value. Another related instability is centrifugal instability demonstrated by Taylor. Consider a mass of liquid rotating. In general the angular velocity of elements is a function of radius. If it were a bucket rotating for a long time, then = constant and the whole body will rotate like a rigid body. If however, the uid were placed between concentric cylinders, with outer one rotating with an angular velocity 0 B and inner one stationary, the velocity prole is given by

PR P R o8 ^ , waves will always form and Exact theory shows that if A k

A
0

where is the radius of the outer cylinder and is the ratio of the radii of the inner and the outer cylinders. In both the cases, radial ow is absent and hence centrifugal forces are 192

h ^ p  p ^ h  i h p 6^ h
h

balanced by pressure forces:

Is this ow stable? direction to p K . To get physical ideas clear, let us assume that viscous forces can be neglected, i.e., the ow is inviscid so far as this radial excursions of uid are concerned. Then the element will have a new angular velocity dictated by angular momentum conservation:

K P p A Op Let an element at p g go due to some velocity uctuation in the radial

A A

angular velocity of the element in the new position

K p g K p g  pK

(20.1)

Thus it will experience a centrifugal force equal to

P p K  K A P p K K p g o p g  pK K O p A P p p K 

The pressure gradient it experiences of course would be corresponding to the centrifugal force at the mechanical equilibrium already established: Thus if p K p g , the particle would continue to move out if centrifugal force exceeds the pressure gradient, i.e., g

is a constant and hence the above is not satised. Hence ow is stable. In the example of uid placed between concentric  cylinders,  where the outer cylinder was stationary and inner cylinder was rotating, then p g p K (see the velocity prole given earlier). Hence it is possible that

P p K K p g o p  P pK g Thus, if p K p and if K K p g p g p K the ow is unstable. Now for rigid body rotation,

K p K p K 

p K

K K p g p g p K p K

is satised and ow can be unstable. Look at some data, (qualitatively shown in Figure 20.5) and note the beautiful agreement at low values. Suppose that the ow does become unstable, and elements are thrown outward. When these elements reach the wall, they will have to turn back since they can not go through the wall. This leads to a cell like structure. See Figure 20.6 Note that these well structured ows are still laminar, and are referred to as secondary ows. If the speed difference between the two cylinders is increased further, the secondary ow undergoes more transitions and the ow becomes turbulent. One of the beautiful examples of predictions of linear stability theory ow is the theory of the above described instability, including the structure of cells, developed by G.I. Taylor 2 . Just to indicate the complexity of problems, let us consider a further aspect of the problem. If speed of the inner cylinder is increased further after secondary ow has been attained, a wavy vertical component appears (superimposed on the rotational motion) in the annulus. The speed of these waves is 1/3 of the speed of the inner cylinder. So far no one has been able to predict this!
2

The reference is Phil Trans Roy Soc, A 223, 289 (1923)

193

Unstable 2 1 R 1 Stable If inviscid theory is right, all experimental conditions lying above the dotted line should be unstable

2R 2 2
Figure 20.5: Stability of ow between rotating cylinders

Axis of the concentric cylinders


Figure 20.6: Secondary ow structure: ow between rotating cylinders

194

20.2 Shear layer instability


This type of instability is referred to as Kelvin Helmholtz instability. and it a very ubiquitous instability. Consider two immiscible liquids owing past each other in opposite directions. Under some conditions, usually when velocity is high, the velocity will change direction over a thin zone. The zone over which velocity gradients are large is called a shear layer. Notice that this layer has a tendency to rotate in the direction marked. Suppose that due to some disturbance, the shear layer deforms into a wavy structure. As the uid has to ow around the crests, the ow near the crests will be faster. For the same reason, as the uid has to ow into the troughs, ow near the troughs will be slower. See Figure 20.7 By Bernoullis equation, the

-+ +_ -++

Shear Layer

Deformed shear layer

Figure 20.7: Instability of a shear layer pressure falls when ow becomes faster and vice-versa. Thus the pressure near the crests will be less, and more near the troughs. These zones are marked ++ and ! respectively in the gure. It can be seen that the amplitude of the disturbance will increase due to the pressure differences created and the wavy disturbance will grow in amplitude. If we superimpose the differential velocity of the shear, it produces the structure shown in the lower frames of the Figure 20.7. If we superimpose the tendency to rotate due to the vorticity, it produces a spiraling vortical structures seen. They are approximately represented in Figure 20.7. The vortex structure itself is unstable and it can further breakdown to produce turbulence. If it gets caught in shear in the other dimension, it can extract energy. We took the example of immiscible uids to make the ideas clear. A shear layer can form in a single uid also and the same phenomena can occur. In miscible systems, if shear layers are created, the above mentioned motions are generated and cause mixing. A fast jet issuing out of a nozzle is a good example of a shear layer and the phenomena described are illustrated very nicely in jets. You should look at the very nice photographs reproduced in a book called An album of uid motion by van Dyke. The break down of the vortex can also produce regular patterns, when viewed on large scale. There is a belief that coherent structure like these exist in turbulence. The large scale struc195

ture, small scale mixing are illustrated by the photographs of mixing layers taken by Roshko, displayed in many books on turbulence, e.g., Turbulent Flows by S.B. Pope. In all the photographs of these phenomena, notice the connement of turbulence and disturbances to a zone in the entire ow eld. This is the edge of the boundary layer.

20.3 Inexion points and instability


It can be shown that for inviscid ows it is sufcient for the ow to become unstable if there is an inexion point in the velocity prole. The previous example we considered is a vortex sheet. The shear layer has an inexion point. Other examples of common ows are shown in Figure 20.8 In viscous ows an inexion point in velocity prole may lead to instability

Sheet vortex

Points of inflection Idealized vortex sheet

Flow behind bluff bodies

Wake region Points of inflection

Flow issuing out of a jet Points of inflection

Figure 20.8: Instability of a shear layer since viscosity stabilizes the ow due to dissipation but may also destabilize the ow due to diffusion. The latter effect leads to Tollmein-Schlichting waves in boundary layer ow. Many ows can be thought of as having inexion points. We have considered qualitatively the causes of instability in ows. Quantitative theories can be developed for this.

196

Chapter 21 TURBULENCE
It takes a shrewd fluid dynamicist to avoid turbulence for long Quote from Davidson, Journal of Fluid Mechanics, vol 429, p 410. I had less difficulty in the discovery of the motion of heavenly bodies in spite of their astonishing distances, than in the investigation of the movement of flowing water before our very eyes. Galileo Quoted by R. Narasimha, J. Ind. Inst. Sci., vol 64 (a), pages 1-59, 1983 It remains to call attention to the chief outstanding difficulty of our subject Quote on turbulent flow from Hydrodynamics by Lamb, p 663

21.1 Phenomenological description


It is often observed that the orderly streamlined motion or laminar ow observed at low Reynolds numbers gives way rather suddenly when a critical value of Reynolds number is exceeded, and the motion becomes apparently random. To quote Batchelor . . . some of these motions are such that the velocity at any given time and position in the uid is not found to be the same when it is measured several times under seemingly identical conditions. In these motions, the velocity takes random values which are not determined by the ostensible, or controllable, or macroscopic, data of the ow, although we believe the average properties of the motion are uniquely determined by the data. Fluctuating motions of this kind are said to be turbulent. Let I us make the discussion concrete by taking one quantity, say component of the velocity, . If it is measured at a given location as a function of time, one would nd it to be a strongly varying function of time. It is possible that the average of the observations over a long enough time period is constant. As usual, we would refer to such ows as steady. Consider such a steady ow. A typical velocity record is shown in Figure21.1. It must be emphasized that the uctuations are not due to errors in measurement, but are inherent to the ow itself. Thus turbulent ows are irregular or random, in space and time. 197

Time

Figure 21.1: Typical record of observations of velocity at a given point

21.2 Characteristics of turbulent ow


There are several characteristics of turbulent ows. Let us consider two of these. An objective of theories of turbulence is to predict these.

21.2.1 Indeterministic nature of turbulent ows


As the equations of motion of uids rest upon solid physical foundations, it is expected that such complex and apparently random motion can still be predicted by solving, numerically if not analytically, the Navier Stokes equations (NSE). Though there is no evidence to doubt such a belief, solutions to NSE have not been found at high Reynolds numbers so far due to limitations on speed and size of computers. Further, uniqueness of three dimensional solutions of NSE is yet to be established. We can summarize the situation as follows. Let us look at the situation from a theoretical or analytical view point. Suppose there is a proof of uniqueness. Then it would imply that if the input is precisely known, then turbulent ow is also fully and uniquely determined, and therefore the commonly observations made on turbulent ow are due to some uncontrolled variations into the system through inputs. It would be possible to put this to experimental test only if the input to the system can be controlled to arbitrarily specied accuracy. However, recent studies on dynamical systems, broadly referred to as chaos, indicate that even when the solutions are determined uniquely, they could be very sensitive to initial conditions, and that tiny changes in initial conditions would make the solutions look random or chaotic or turbulent. What ever might be the cause, turbulent ows seem to be out of reach of precise description at present. This comment applies equally to computational efforts also. Let us illustrate the last comment more concretely. Usually turbulence seems to produce variations on ne length scales. It would be of the order of 0.1 mm or even less. Thus, if we want to compute velocities of uid of volume 1 cm , it would require one million grid points. At each grid point we have to keep track of three components of velocity and pressure. Thus we need to store 4 million quantities and manipulate them. As Reynolds number increases, the variations occur on an even ner length scales and hence the computational effort becomes even more difcult. For this reason, exact computations of turbulent ows have not made much progress. Thus, turbulent ows do not appear amenable to exact description and modeling seems to be necessary. 198

21.2.2 Enhanced diffusive power of turbulence


As stated earlier, turbulent ows appear to be irregular, or random, in space and time. The random motions are like the random molecular motions. Hence, just as random molecular motions have diffusive power, random turbulent motions also have diffusive power. We have shown earlier that diffusivity by random motions is given by the product of a characteristic length and velocity. Since these scales of turbulent motion are larger in magnitude than the corresponding ones at molecular level, turbulence causes enhanced rates of diffusion. These aspects are discussed in more detail in the appendix to this chapter. Let us illustrate this with an example. A refrigerator has a tube bank behind it to dissipate heat into the room. It does so by natural convection1 . Suppose the length of the tube bank is 1 < , and that the temperature difference between the tubes and room is about 10 , while the room temperature is of the order of 300 . The force, that accelerates the uid , 6  buoyancy ^ 8[ . The element, per unit mass of it, is of the order of R velocity attained by the uid element as it rises, at the end of its contact with one meter long tube bank, assuming that o     [   ^ 6 ^ 6 viscous forces are negligible, is of the order of or about 0.8 < : . If we R assume the uctuating velocities are about 10% in magnitude, they would be of the order  m  < that ^ : . Similarly if we assume that length scales are also 10% in magnitude, of K ^ it turns out to be of the order of 0.10 < . Thus diffusivity will be of the order of 0.01 < : . This value is K about 100 times greater than the molecular diffusivities, which are of the order of 0.0001 < ^ : . It is this enhanced diffusivity which makes turbulent ow desirable in heat and mass transfer. For the same reason, this also implies more dissipation of mechanical energy.

21.3 Postulated mechanisms and possible approaches


It is natural to ask as to how turbulence arises and how to go about predicting the characteristics of the turbulent ow described. The routes to turbulence are not completely charted. As we briey mentioned in chapter20, some laminar ows have been shown to be unstable. Thus minor perturbations involving some length and time scale will grow to some extent and probably stabilize. The resulting ow would then have the main ow and a ow of a ner length and time scale superimposed on it. As Reynolds number is increased further, the ow can become once again unstable and generating a more complex ow involving one more ner length and time scale . . . and so on. This route was proposed by Landau. The resulting ow would involve many length and time scales and would look turbulent. It is as if, several musicians owning many kinds of instruments, tuned to different frequencies, are playing them simultaneously. However, no one has been able to show this by calculations. In the above approach, ow is imagined to be occurring on several length and time scales, and that ow can be decomposed into, perhaps, innitely many time and length scales. It is possible to think of statistical characterization of the resulting ow, and we will consider this a little later. Another proposed approach is to think that of turbulence as a combination of several nonlinear oscillators. This will give rise to chaos, which is closely connected to the sensitivity to initial conditions we breiy described in the last section. It has been shown that a few such oscillators put together can show ow patterns similar to that of turbulent ows. This approach, referred to as dynamical systems approach, hopes to nd a few signatures which when com1

This material follows that given in Tennekes and Lumley very closely.

199

bined will simulate turbulent ow. Of course this approach is also yet to be established. This approach will not be considered any further in this notes.

21.4 Statistical theory of turbulence


As mentioned earlier, the Navier Stokes equations can not be solved to understand the characteristics of turbulence or that a deterministic approach2 will not be successful. In fact, several other phenomena of science seem to be out of reach of precise theoretical determination. These are referred to as stochastic, and the precise value of a stochastic variable during a given observation seems to be dictated by chance. The statistical theory of turbulence proposes that velocity and pressure elds in turbulent ow be treated as stochastic variables. Thus one does not hope to predict the exact behavior of a stochastic variable but to predict the probability that the variable takes on a value. A familiar example is forecasting the monsoon. One can only predict what is the expected value of the rain fall, what is the reliability of such a prediction and so on. Thus in dealing with turbulent ows, one does not aim at predicting precise time dependence of velocities etc in a given ow situation, but only try to predict the probability with which a value is attained by these quantities. We therefore take a small detour into the general methods of statistical description of any variable before focusing on turbulence.

21.5 Statistical description


In this chapter, we use, as far as possible, the following notation. Italic letters will be used to denote the time dependent variable. Its average value will be denoted by the same symbol but with upper case, and with a bar on top. The instantaneous value of the deviation from the average will be denoted by the same symbol but in Roman script.

21.5.1 Probability distribution


The usual approach adopted in describing stochastic processes is to aim at predicting the probability of occurrence of a variable taking a certain value. For example, in experiments involving fty tosses of a coin, one can ask what is the probability of getting thirty heads. But in turbulence, the variables we deal continuous. For such variables, a I probability distribution I0 I0  is I with y I are dened. Thus is the probability that the value of lies between and ZI  I0 Z I . Thus, the average value of can be computed from M

, Z I  I  I0 I0  is the probability distribution, its integral must be equal to unity: Since Z 6 A, I0 y I Z
I  aj A
2

The motion is completely determined by the equation of motion and boundary conditions. Thus the approach of solving the equation of motion is called the deterministic approach.

200

Another statistical quantity of interest is the variance;

E K I0  A w I0 c

I K 

The statistical properties can be measured and veried I8 against the stochastic models for the probability distributions. For example, one can measure as a function of time and determine the average, and veried against a stochastic model for the same. Steady turbulent ows are simpler to deal with. As stochastic processes are inherently unsteady, a more precise classication of steady state processes is used. Stochastic processes are classied as strictly stationary or weakly stationary. All the statistics of the former process are unaffected if the time is replaced by M . Thus

I c l A I0 c M Z

c

is true for a strictly stationary stochastic process. Thus, we can treat the probability distribution to be independent of time in taking averages etc. Here it will be assumed that a steady turbulent ow is a strictly stationary process.

21.5.2 Measurement of probability distribution

Let us examine how the probability distribution I of any quantity, , can be experimentally determined3 . We rst acquire a time record of I I g quantity over some sufciently long time  . We  then mark a window of width around as shown in Figure I 21.2. We now measure time Ig I all I. = g intervals (marked byI in the gure) for which the value of was between and M I I I The probability that lies between g and g M is equal to sum of all the time intervals when

v1 v

dv

t 0 1

t 2 t 3 Time

t n-1 t n T

Figure 21.2: Experimental determination of probability distribution divided by  or

We can do this for all values that can assume and hence we have the entire probability distribution. Thus if a long time record of any random variable is available, its probability distribution function can be measured.
3

I l  I Z g A " I

 

?= g = 

This material follows that given in Tennekes and Lumley very closely.

201

21.5.3 Fluctuations around the average


It is also seen that the averages that we normally dene coincideI with the denitions that use the probability distribution. For example, from the time record of , we will think of its average to be 6 "

aj A "
j

6 A I l " a  I I= But if we slot the values of into bins , then the above is same as ?g = = I aj A = = "   = is the duration of F where I is the total number of times, and I I I value of fell between = and = M . Thus I I I a j A = = Z =   I  , we get or in the limit of I I l I aj A Z I

If the integral is discretized, we will write it as

   , I * lA

 

* !

  ' " %

" % ' " =%

time interval, when the

where for simplicity we assumed that the range of is 0 to . We have a few other useful quantities which we will now dene. I j Consider the instantaneous value of the deviation of the variable from its average: k a . What is its expected value or average?
kj

A I aj  A

square of the deviations is of course not equal This is of course expected. The average I of the to zero but is equal to the variance of :
k

I , I a j y I A 

K A

In this introduction to statistical methodology, we have placed emphasis on the concepts and also have given an indication of how the quantities of statistical signicance can be measured.
21.6 Reynolds averaged Navier Stokes equations (RANS)
Observe the motion of the water surface, which resembles that of hair, that has two motions: one due to the weight of the shaft, the 202

I a K A ,
j

I o I a j  K  I A

E K I 

other to the shape of the curls; thus, water has eddying motions, one part of which is due to the principal current, the other to the random and reverse motion. Leonardo da Vinci, ca. 1510 It is clear that if we could predict the probability distribution of all possible quantities of interest, we could predict expected outcomes of turbulent ows. However this has not been achieved. Hence, only some averaged quantities are examined. Thus one attempts to solve for quantities of interest with the help of equations of motion. We can derive equations, for steady turbulent ows, which should be obeyed by velocity uctuations by substituting

I =A

= M a j = c Aml M j O

into the various equations, and averaging them. Consider the equation of continuity rst. We get

 A

This of course implies that

4 4
k

M = =

4 4 4

=M a = A =

4 4

a = =

= k A  =

21.6.1 Closure problem

Following the same procedure with Navier Stokes equations, we have

we get

K = = h b M a 4 4 4 4 4 4 4 4 where for simplicity we have neglected the body forces. We can average these equations and
k k j

6 l M   j j = = A P M k M a k M a = M
k

Noting that

4
k

6 j j = j a A M a P = M
k

the above can be rewritten as

Thus in order to solve for the mean velocity, we need information about k k = and their derivatives. If we try to derive an equation for k k = , then we nd that we would need information 203

k = j aj = A 6 j M a P = M

= k A  =

hb

K j= a

hb

K j= a

about k k = k and so on. In other words, we can never derive a set of self contained equation for any uctuation, and would require information about products of uctuation of higher order. This is called a closure problem since the set of equations being derived are never becoming self contained or closed. Thus, we can not even solve for averaged quantities exactly. Hence we need to make some hypothesis or develop models to solve the equations for averaged quantities. This will be considered later.

21.6.2 Reynolds stresses


The equations derived in the previous section are referred to as the Reynolds averaged Navier Stokes equations since these were derived by Reynolds rst. The higher level product term = that arises, viz. k k , can be interpreted. Consider a term like k k . Suppose for the moment that the average velocity in direction is zero, and that the average velocity in the direction is constant. Then v is a uctuation in component velocity itself and hence, due to it, a uid particle will travel in the direction. When it does so, it takes along with it the other components of momentum, namely and T components. In particular, it will carry with it momentum, over and above the average value, corresponding to v . Thus, k k is the averaged transfer of momentum in the direction due to uctuations. These terms therefore represent momentum transfer due to velocity uctuations. Thus they are equivalent to stresses, and are called Reynolds stresses: F Here, we added superscript to indicate that they are turbulence related and will vanish in laminar ows. Hence we can rewrite RANS as

P k k = =

where we have indicated the stresses calculated using the averaged velocities and Newton Stokes law by superscript . Those would correspond to the stresses calculated using Newton Stokes law as if laminar ow corresponding to the average motion prevailed. One of the important ideas arising out of this development is that averages of quantities like k k are very important to theory of turbulence, and need to be modeled to circumvent the closure problem. The interpretation that Reynolds stresses arise due to momentum transfer by random velocity uctuations on macroscopic scale also gives a clue that they can be treated as a diffusive process occurring due to turbulence. This idea is used extensively in turbulence modeling. This will be considered later.

P \j m

$b

j F A \ = M m M m
j

b R b R

21.6.3 Experimental measurements


We have already mentioned the mean and variance as a few experimental quantities amenable to measurement. A few other quantities are also important as indicated by RANS, and also can be measured. We discuss one of them, the velocity correlation. Cross correlation functions Let us examine the Reynolds stress terms as they appear in the RANS a little bit more. A typical term is
k

4 4 204

It can be rewritten as

primed quantity, then, to evaluate the above, we If we denote the values at M by must measure k k as a function of coordinate between from a xed location. Here v and v are two random variables. The average of product of two random variables is called cross correlation. It indicates whether the uctuations in the two variables are related to each other or not. Thus if the velocity uctuations in various directions are independent of each other, then the off diagonal components of Reynolds stresses would be zero 4 . Thus, it is important to know how the velocity uctuations at different locations are correlated. We of course expect them to be uncorrelated as the distance between the two points increases. But the decay with distance separating them is important.
21.6.4 Spectrum
The spatial dependence of any cross correlation is usually expressed in terms of Fourier transforms. Suppose there is a wave of velocity uctuation moving through the eld. Then the correlation of velocity disturbance between two points will also appear as a periodic function of the distance. Thus Fourier transform will help in visualizing the correlation as consisting of uctuations correlated over some length or loosely speaking as an eddy. If the uctuations were totally periodic in space, a nite number of sine or cosine functions will be sufcient to represent the correlation. However, if there is no periodicity, or in other words, if innite distance is also a wave length, then a continuous distribution of wave lengths is needed to represent the cross correlation. In turbulence, there seems to be no apparent periodicity. Hence, the correlation is expressed as5 ,

 

k  y k 

 L  k  ,8 j p  p j k p M k wherep is distance along the axis. Batchelor in his book Homogeneous Turbulence shows  that j can be interpreted the kinetic energy contained in a wave of wave length equal as L ^ to k j . The function j  is called the spectrum since it represents the variation of the amplitude of the energy with wave number, j .
Spectrum is an important experimental quantity. It can be measured by feeding the two velocity signals to an integrator. The average of the integral over several time intervals or a sufciently long interval gives the spectrum. The correlation can be obtained by inverting the spectrum:
k

M p k  A

 o8 l j j$   j

Thus if we measure spectrum over sufciently large range of wave numbers, the correlation function can be obtained from it. An interesting relationship that we will use later on is a
The diagonal elements would not be zero since they are averages of square of quantities viz. v n We should write both sine and cosine functions, but we have just written for simplicity only cosine transform. For real variables, and when the spectrum is symmetric in wave number, only positive wave numbers need be considered, and cosine transform is sufcient.
5 4

205

special form of the above for coordinates:

= 0, and when the cross correlation is independent of spatial


k

K A

l j  j

Thus, in this instance, the kinetic energy contained in the uctuations is equal to the integral of the spectrum, and is location independent. Measurements of spectrum serve to verify theories of turbulence.
21.7 Models
The equations describing turbulent ows are very complex and can not be solved directly. Modeling is then necessary. Two approaches are described here to give a avor of the very extensive work.

21.8 Kolmogorovs theory of energy cascade


Big whirls have little whirls, Which feed on their velocity; And little whirls have lesser whirls, And so on on to viscosity. Richardson (1922) Big fleas have little fleas upon their backs to bite them, and little fleas have lesser fleas, and so on ad infinitum. As Reynolds number increases6 , the effects of viscosity are expected to become unimportant. This implies that a ow once started, and if wall effects are not there, should continue to exist 7 in its state without any supply of power. Thus, if a uid is pumped through a wire mesh, far away down stream from the mesh and also far away from the walls, the uid should continue to have the same kinetic energy. Such an experiment generates turbulence. However, it is observed that turbulence decays away from the mesh due to viscous effects. Kolmogorov resolved this paradox by postulating the cascade theory of turbulent energy. The theory is set in an idealized view of turbulence. We will rst review these concepts.

21.8.1 Homogeneous and isotropic turbulence


As we just now mentioned, the theory has been developed in the context of absence of wall effects. Though it sounds restrictive, the results have wide applicability, and Kolmogorovs
tY s ut in the book Perspectives in Fluid Mechanics This material is a summary of a part of the chapter by o `o prq' edited by Batchelor, Moffat and Worster. 7 Another implication of this is that drag forces should go to zero. As discussed earlier, this paradox was resolved by the boundary layer theory, which postulates that viscous forces are concentrated in a thin layer near the wall.
6

206

theory has proved to be of great utility. In view of the assumed absence of wall effects, certain simplications are made in the theory. The rst approximation that comes to mind is to assume that average characteristics of turbulence i.e., correlations, are independent of the position in space along a direction. Let us examine this in more detail. Consider a correlation between component of velocity at one point and component of velocity at another point, i.e., k v kw . Take any point as reference. In general, the correlation would depend upon the vector separating the two points and position of the reference point. Thus, if the mean value of this correlation does not change when the position of the reference point is changed, the turbulence is said to be homogeneous. Moving the reference point, or the origin, is equivalent to translating a coordinate system without rotation. A second reasonable assumption to make is that turbulence is isotropic: the characteristics of turbulence are independent of direction, i.e., they are insensitive to rotation of coordinate 8 system and reection of the ow on any plane . Consider the correlation between two components of velocity at a point, e.g., k k . Let us rotate or reect the coordinate frame, and measure the velocity uctuations etc with respect to the new orientation of the coordinate system. Let them be represented by u. Once again consider the correlation between and components of the velocities at the point: x x . If the correlation is independent of the rotation of coordinates along with the conguration, i.e., k k A x x , turbulence is called isotropic9 . Consider a simple application of these ideas when the two points are the same. Consider a correlation like k k at any point. If we now rotate the coordinate system around the axis by 180 degrees. Then the old axis has become the new axis. Thus the correlation between and components of the velocity is equal to x x in the new system and would be equal to k k in the old system. If turbulence is isotropic then

k k

A k
k

A x

Hence both must be equal to zero. Thus for isotropic turbulence, we then nd that

A k
k

A  $ A U k /k
k k

Similarly, it can be shown that

K A
k

K A

K
k

Further, following similar arguments about derivatives, it can be shown that the spatial derivatives of these averages is zero. Thus isotropic turbulence is also homogeneous. Consider a more complex situation involving two points P and P. There are three directions involved here: the direction of the velocity vectors and the direction of the vector separating the two points. The two points form a conguration of interest. Let the two points P and P lie at two different locations but along the axis. Examine a correlation like k k , where v is at P and v is at P. Rotate the coordinate system and the conguration around axis by 180 degrees clockwise. Then using arguments similar to what was used before, the correlation between and components of the velocity is equal to x x in the new system and would be equal to k k in the old system. Thus if turbulence is isotropic then
x
8

A x

A k k k

These assumptions sound contradictory to the resolution of the paradox which we just mentioned since if there is dissipation, turbulence must decay in the direction of ow. But the decay can be sufciently slow to allow such an assumption to be made. Similarly, in some local sense, the ow could be isotropic. 9 See the book on Homogeneous turbulence by Batchelor p 40

207

Hence both must be equal to zero. With use of more complex arguments, it can be shown that there are only two types of cross correlations10 that do not vanish:
k

5)

  c M k

5)

and k

5)

  M k

*)

The rst is a cross correlation of velocity components directed along the line joining the two points, or a longitudinal correlation, while the second is a cross correlation of velocity components directed perpendicular to the line joining the two points, or a transverse correlation. In view of the isotropy, it also clear that the cross correlation must depend upon only the distance between the points of measurement:

p 
k

5)

and
y

p 
k

5)

  M p k K k

*)

  M pK k k

5)

It turns out that due to equation of continuity, the two are related. This simplies our considerations a lot since we have to deal with a single characteristic velocity uctuation, and a single function of cross correlation depending upon distance. K Let us also mention that we have only one spectrum corresponding to k f(p ). It is represented by g#g . Thus, experimental measurements on g#g in ow behind a wire mesh were used to test Kolmogorovs theory.

21.8.2 Generation of many length scales


Kolmogorov suggested that rate of energy or power input, to the system is at large length scales as indicated by the large value of Reynolds number based on the length scale of the apparatus. The energy supplied creates motion on large length scales. The motion becomes perhaps unstable and motions of smaller length scales are created. This process occurs without loss of energy since the Reynolds number is large and hence viscous effects are unimportant. However as this process of creation of small length scaled motions continues, motion will be created on a small enough length scale that Reynolds number is of the order of unity and viscous effects become important. At length scales of this order, the energy supplied is dissipated into heat by the action of viscosity. This is schematically shown in Figure 21.3.

21.8.3 Kolmogorov length scale


The size range where viscous effects are not important, but which are smaller than the length scales where energy is being supplied is called inertial subrange. Let be a length scale lying  in the inertial subrange. Let  be an estimate of cross correlation of velocity11 on length scale of the order of . The turn over time or the time required for motion on this length scale ^ to become unstable can be estimated to be  . The instability K ^ is destroyed by viscous effects. The time required for viscous effects to propagate over is . Thus one would estimate the length scale where the two time scales are about equal to be length scale where the inviscid transmission of energy through creation of motion on smaller length scales ends and the motion

i i h

10 11

See the book on Homogeneous turbulence by Batchelor p 40 From now on, we simply refer to this as velocity uctuation.

208

Smaller length scale Large length scale Largest length scale

Figure 21.3: Generation of many length scales by instability

where energy dissipation becomes important. The length scale is called Kolmogorov length scale in honor of Kolmogorov. As is easily worked out, Reynolds number at this length scale is equal to unity: where the symbol  z has been used to denote the velocity uctuation corresponding to the Kolmogorov length scale. The above also gives the following relations. By equating an estimate of rate at which energy dissipated by viscous action to the power being supplied to maintain the ow at steady conditions,

`  h
z

A 6

h `

 z  K K

Combining the above with the idea that Reynolds number at the Kolmogorov length scale is unity, we nd that g

and

 z A

hV h$  g A `h
H

Now it is easy to understand how Kolmogorov resolved the controversy of decay of turbulence behind a wire mesh. The power suppled to ow at the length scales of the wire mesh generates velocity uctuations on ever decreasing smaller length scales. When these length scales reach the a level where Reynolds number is of the order of unity, energy is dissipated into heat through them. The dissipation occurs every where in the uid. Another way to think is that a part of the energy supplied does not go to enhancing ow, but goes into generating these wasteful motions! Kolmogorovs theory is a picture. It explains the controversy 12 . But what more can be said about turbulence based on this story?
Some aspects of science are very interesting. Usually, the rst fact a theory explains is smuggled in as a hypothesis. It is the other aspects that it is able to explain that give credence to it. A nice example is Darwins theory of survival of the ttest. Is it not obvious that the ttest survive? It is the utility of the theory in explaining other features and its consistency with the revolution brought about by the discovery of DNA that lend credence to Darwins theory.
12

209

21.8.4 Velocity uctuations and length scale


One of the objectives of statistical theory of turbulence is to predict the scale and size of uctuations. A concrete quantity is the velocity correlation or the spectrum. Kolmogorovs cascade theory is able to do this. Suppose that the scale at which energy is being supplied, W , is very large. Let us suppose that Reynolds number of the ow is very large. Hence, estimating the characteristic velocity at this scale by the non-dissipative power transmission, we get

Hence we would estimate

range, we can say that the inertial range and the length scale at which power is being supplied are also widely separated for large Reynolds numbers. Hence at very high Reynolds numbers, one does not expect that velocity uctuations corresponding to the length scale on which energy is being supplied,  { , to be a suitable charac teristic velocity uctuation on smaller length scales,  . In other words we would not expect the former to be suitable for nondimensionalizing the latter, or, to use the modern linguo,  does not scale with  { . One would expect the velocity uctuations typical of the inertial subrange to be a more suitable quantity with which  can be scaled. We know that z is one suitable value, since it is a value of velocity uctuation typical of inertial subrange, but which occurs at the least size of the inertial subrange. In a similar manner, we can expect a suitable ^ . Hence we expect non-dimensional length in the inertial size to be

: W Thus, the dissipative length scale and the length scale at which power is being supplied are ` widely separated for large Reynolds numbers. As is the smallest length scale in the inertial

g W W 

h `i  i  to depend upon h since it is a quantity characteristic of inertial However we do not expect ` h range. that , we can conclude that the function ` has to be proportional to i ^ `s g Knowing  so that i is independent of h . Thus, H g g g g gK   i A `V $  i  h i h
or

g H   A  $ ` V

  A `  z

i `H

`i V

Thus Kolmogorovs theory predicts that

The last relationship could have been derived in a different way also. If a motion on the length ^ scale having a characteristic velocity  dies down in a turn over time of  , the power it would have transmitted to the lower length nondissipatively would be equal to its kinetic  scales ^ . But this must also be equal to the power being energy divided by the turn over time:  supplied, , and we recover the above relationship.

g   

Xi

210

21.8.5 Prediction of spectrum of velocity correlation


Since Kolmogorovs theory is able to make a predict the relationship between the velocity uctuation between two points and distance separating them in the inertial range, it can be p  the used to predict the spectrum of f in that range. Thus

 k  o j$p  p p M kL k K p  o  j$p  p kL H H H  p p pV K  z K A ,8 j V ` V k  z K b j y g  g K H b j s V g b y j s where b j is a universal function of j . From arguments similar to those used to derive the dependence of velocity uctuations on length scale, it follows that the spectrum K in the y equilibrium range should depend only upon energy dissipation rate and j . Hence  z b j g#g j  A A

, ,

` h h h ` ` ` ` h must be independent of viscosity. Since `  j y ` : b j s and K : g#g j  j


` `

` `

` `

These has been experimentally veried and is considered as a great success of Kolmogorovs theory.

21.9 k |~} Model


Let us now turn to the engineering models of turbulence. Kolmogorovs theory is a qualitative but valuable picture. It has to be utilized in some more concrete models to predict characteristics of practical turbulent ows. These are done in many ways in different models. Here we give one such model developed largely by Prof. Spalding and his students. Let us recall that any diffusion coefcient is obtained from product of a velocity and length scale typical of a random process. We have also interpreted Reynolds stresses as momentum diffusion due to macroscopic velocity uctuations. Hence, we can use this analogy between molecular diffusion and turbulent diffusion to write

where is turbulent diffusivity of momentum. The product is referred to as eddy viscosity. Kolmogorovs theory gives a big hint and prompts us to write

h " F&%

k k A

F&% H " h

4 4

4 4

P F

h " F&%  i  i
211

h " &%

Thus we can predict Reynolds stresses if we are able to predict the local values of the velocity uctuations and the length scales. Many models attempt to do this and one such model is the k model. The model attempts to derive equations for the kinetic energy of turbulence per unit mass k, and the local power dissipation (due to turbulence) per unit mass, . Thus local values for these quantities are obtained by solving these equations. Then, we write

and

h " &%

There will be proportionality constants. If they are universally valid across many types of ows, the models will be very useful.

  K   F  

ii

i l

21.9.1 Equation for k


Kinetic energy in uctuations is k:

6 A L k K M k K M k K

The equation for k can then be derived by taking a dot product of the equation of motion with the velocity uctuation vector and time averaging it. We get the following equation:

6 l  \ m  A m  m \ m m M P M P m m

Jb l

Jb

is thought of as production of kinetic energy or conversion energy of main ow into kinetic energy contained in turbulence. This analogy is drawn since the term is similar to that coming from vortex stretching. Hence it is modeled as

4=V As can be seen, the equation for k naturally involve complex products involving three velocities 4 4 as the closure problem would indicate. These have to be modeled. For example the rst term on the right hand side, namely, m m$b  \
M 6 F  A m m \ L

where

c =" %

b R " % 

$b

R " &% b

The second term on the right hand side, namely,

bm m

 M OP

looks like turbulent diffusion of kinetic energy and pressure energy. Hence it is modeled as

 M OP k =~A N
212

h E " F&% b l

where we are using a diffusion coefcient for kinetic energy different from that for momentum. As discussed in the previous section, it is postulated that

Thus our nal equation for k is given by

21.9.2 Equation for dissipation

% " H % "  h l A l m$b bm E b M h  4  M 4 V 4 4 4 4


` `  Y   

l & % " h

  A K F A

An equation for is still needed to solve the above equation. The power dissipation per unit mass due to turbulent uctuations is equal to

L = 

4 4

= M

An equation for this can be derived by taking a dot product of the equation for vorticity with the uctuating vorticity and time averaging it. Once again several terms will appear and have to be modeled. Without giving details we write the nal equation

4 4


= M

4V
V =

\m  A m

$b

" F&% h b E b


F H g M = 

&% lh"

4 4

4 4

4 4

K S

The two equations for kinetic energy and dissipation have to be solved simultaneously with the equation of continuity and the three components of equation of motion to complete the calculations. It can be seen that there are ve constants that appear in the equations: c g c K c and  . These ve constants are generally taken to be 0.04, 1.44, 1.92, 1.3 and 1.0 respectively. With these values the model is able to calculate many turbulent ows.

General references

1. A rst course in turbulence by Tennekes and Lumley 2. S. Corrsin, Turbulent ows, American Scientist, vol 49, p300, (1961) 3. W. Liepmann, The rise and fall of ideas in turbulence, American Scientist, vol 67, p221, (1979) 4. LVov and Procaccia, Turbulence: A universal problem. Physics World, p35, August 1996.

213

Appendix: Taylors theory of turbulent dispersion 21.A Diffusion & random movements
Molecules are always in random motion, and diffusion occurs due to it. Net ux in any direction arises since there is a concentration gradient. The objective of this section is to relate diffusion coefcient to the characteristics of the random movement. Consider one dimensional diffusion for simplicity. Randomness of movement implies that molecules will move with equal probability in either direction, as shown in Figure 21.4. It can also be seen from the gure that the ux coming to any plane from zones of high concentration to zones of low concentration will be more than the other way and it is this that causes a net ux from zones of high concentration to zones of low concentration.
Low concentration High concentration

Equal numbers move to left or right

Less flux approaches the plane from this side

More fluxapproaches the plane from this side

P1

P2

Figure 21.4: Net ux occurs despite random movements due to concentration gradient

21.A.1 Characteristics of random motion


If we observed a single particle undergoing random motion, it will drift away from its location13 . Random motion of particles in one dimension with equal probability is equivalent to tossing an unbiased coin. Combinatorial calculations on the tossing of a coin (Binomial trials) can be used to show that the particles drift away from the place they start. This is called diffusional drift. As diffusion is a random phenomena, we can only predict what happens on the average when a collection of particles is observed. Thus we try to derive expressions for averaged values of some quantities of interest. The average is obtained by observing many particles in the collection, or by repeated experiments on a single particle keeping all conditions of experimentation constant. We will use this argument of equivalence often.
This does not mean that some vacant spots are being created. Thinking based on probability can only be applied to a large collection of particles. In such a large collection of particles, equal number will be found in all directions at a given radius around any point. Supercially, the drift of a single particle that moves in either direction with equal probability is surprising since we expect it to oscillate around the place where it started from. However, equal probability of movement along, say direction implies that the mean value of of the particle assigning its initial location as should be zero and it is not necessary that mean of the square of should be zero. It is easy to visualize this in three dimensions. The particle will move on the surface of a sphere with continuously increasing radius, but it will cover both and directions of the three axes with equal probability.
13


214

One quantity of interest is the ratio of the distance a particle moves from some initial location to the time interval it took to move or the drift speed. For example consider a collection of noninteracting particles. Suppose that they move only in one dimension. We could place all of them on a plane at some initial time and observe the distribution of their numbers on either side of the plane as they move randomly. Thus, we could mark the locations of all the particles at the end of some time period of interest. This information can be recorded as the fraction of particles located at different values of (assuming the one dimensional diffusion is occurring along the axis) in an interval . This of course is the probability distribution of the location of the particles along the axis. Using the probability distribution, the mean distance by which the particle have moved or the mean coordinate of the particles or any other average property of the particles can be found. Equivalently, we could place a single particle at some location or origin and mark its location after some time interval of interest as it moves around the origin, and do this repeatedly. by averaging over many observations, the probability distribution From this experiment also, of the location of the particles along the axis can be found. F Let us say we have particles, and after some time period, , let = be the location of the particle. Then the mean coordinate of the particles will be given by

= = g A
We of course expect this to be zero if the probability of movement in the positive or negative direction of axis were to be equal. Another way of saying this is that the probability distribution would be symmetric or it will be only a function of the distance along the axis, :: l l l We can calculate the mean of the square of the distance traveled by the particles:

Z A Z A Z : : : A d =K = g K A

This of course will not be equal to zero. The same also can be written in terms of the probability distribution:

K A

u,

K  : Z : l :

Suppose the probability distribution was determined in the above experiment. Then the number l   of particles found in the interval around is given by . The number concentration, Z   that is number per length along the axis, , is then given by Z . This connection allows us to connect the ideas developed so far to diffusion process. In diffusion theory, the problem equivalent to the experiment we described with particles is diffusion in one dimension of a thin and innitely wide strip of solute placed at origin. Mathematically the concentration proles are described by

21.A.2 Random motion and concentration distribution

4 4

A ]

4 4

215

with the following initial condition and boundary conditions c [ A s  

 c

c l  A


for all

where is the Dirac delta function. The solution to this is well known and is given by

l K c A 6 ]  L k ] O X

Thus we can connect the probability distribution to this solution

 A 6 ] K  L k ] O Z X

21.A.3 Characteristics of the random motion and the diffusion coefcient


The mean of the square of the distance traveled by particles can be determined from the above expression, and it is found to be Thus we have this beautiful result

K A L] ] A L6   K

The diffusion coefcient can therefore be evaluated by observing the rate at which the mean distance of particles, from the place where they were located initially, increases with time. Another equivalent way to evaluate the diffusion coefcient is to observe the distance traveled by a single particle for different intervals of time, evaluate the derivative, and average over the many intervals of time.

21.A.4 Langevin equation for diffusion


Another insight is obtained by looking at the random motions from a different perspective. Brown, a botanist, observed an interesting phenomena under a microscope. He found that lifeless particles suspended in a liquid were moving around. This was surprising and this phenomena is referred to as Brownian motion. It is explained as follows. Particles are buffeted on all sides with equal probability by molecules of the liquid surrounding it. Though such collisions are random, they are not balanced at every instant, but are so on the average. Hence, particles move due to the net momentum transferred from the instantaneously imbalanced collisions. Langevin modeled the unbalanced impact of molecular collisions on the particles as a random force, . The following differential equation was written by him, called Langevin equation in his honor, for one dimensional motion of a Brownian particle,

K  <  K A Z  M

A 9 9

In above Z is the friction coefcient, and hence the rst term on the right hand side represents the friction force exerted by the uid on the particle. If is the velocity of the particle, then the above can be rewritten as 

<  A Z M
216

This equation can be multiplied by the position of the particle and rearranged to get

 <   M Z A

We can integrate this

As is a random force, it will not have any correlation with the location of the particle, . Hence, for sufciently large times, the second term of the integral will zero. Further, for sufciently long times, we expect the velocity of the particle to reach a constant value,  . Then, we have < If we now average over many particles, or for a single particle over several sufciently long time intervals, as the initial velocity is not correlated with the initial position, we get

 F A

F : 1 F ,  : 1 F

M  < H K

M  < V

K  F A  Z

But Hence we have

Einstein estimated the mean velocity with which a particle moves by equating it to the thermal energy, since it is in thermal equilibrium with the surrounding uid:

< K A  Z 6  K ] A L  A ] A <  K Z

(A21.1)

where j is the Boltzmann constant and  is the temperature. Further if one makes the audacious assumption that Z is given by Stokes law for a sphere, we get the following

L<  K A j ] A j _  k

It is known as the Stokes Einstein equation. It works surprisingly well!

21.A.5 Dispersion & Diffusion


We have discussed the random motion or diffusion of Brownian particles. In chemical engineering, conventionally, we talk about diffusion only in the context of molecular solutions. Thus salt diffuses in water due to gradients in its concentration, and in this context, we are dealing with a molecular solution of salt in water. What then is the difference between this and diffusion of Brownian particles? The underlying phenomena are the same, but the nal state is different. Consider smoke or very ne dust particles in air. Clearly they move randomly or diffuse even in still air by Brownian motion. Thus, if blob of smoke is left in a box full of air, smoke 217

particles will diffuse away from their initial location. The blob spreads over larger and larger distances, and eventually, the smoke will be evenly distributed over the box. Suppose the original particles are a tenth of a micron in size. Their size remains the same after uniform dispersion as well since smoke is not soluble in air. Thus, on a scale larger than tenth of a micron, say on a scale of ten microns, the number of smoke particles per unit volume will be the same every where in the sphere. The uniformity is on a scale larger than the size of the particles. On a scale only slightly larger than that of particles, the dispersion will look grainy. The same applies to molecular solution as well. Suppose we did the above experiment by placing a lump of sugar in water. Sugar will dissolve and diffuse away, and the concentration of sugar will be the same at all points in the box. The sameness is on a scale larger than molecular dimensions, and graininess will be present on a scale comparable to molecular dimensions. In general our interest is on a scale much larger than molecular size, and so we take it that the solution is uniform. In this sense, depending on the length scale of interest, smoke or dust particles can be treated as if they are diffusing in a medium. Let us once again consider smoke particles in air. The particles are usually very small and their inertia is therefore small. Thus, they will move with the local velocity of air. If there is turbulence in air, the particles will move randomly according to the velocity uctuations in a turbulent ow eld. This is similar to diffusion. The problem is of immense practical value in the context of atmospheric pollution and it is this problem that Taylor addressed. Before considering Taylors theory, let us make a few comments to complete our discussion on dispersion and diffusion. As we just pointed out, species (or heat) move from one place to another place due to eddy diffusion on a coarse scale. Suppose, the particles placed in the above experiment were soluble. Imagine a blob of a solution of a dye placed in water and that the dye is soluble in water. If the solution is stirred, the blob will be seen to continually break into smaller lumps while they move in random directions. The turbulent velocity uctuations break the big blob up into smaller and smaller pieces. It is as if a collection of small blobs were initially present in the big blob, and the smaller blobs move in a manner very reminiscent of the motion of smoke particles. However, along with this dispersion into small blobs, dissolution and diffusion of dye also occurs to create a molecularly homogeneous state. The larger the length scale is, the larger is the time scale of molecular diffusion. Thus, as long as the blob size is large, molecular diffusion will not be able to smear out the dye on a molecular scale in a reasonable time scale. As molecular diffusion is not effective, the breaking of the blob only enhances the number of regions where concentration gradients are sharp. As the number of regions increase, or as the blob size becomes smaller, diffusion becomes effective, and molecular homogenization takes over. This is similar to the ideas in cascade theory. There, motion is created on smaller and smaller length scales. The processes in the inertial subrange are non-dissipative, while they are dissipative on Kolmogorov length scale. The breaking up of the initial lump into smaller ones can be thought of as a dispersion process, and is the result of movement due to macroscopic velocity uctuations or eddy diffusion. As length scales decrease due to this, a length scale is reached where the concentration gradients become sufciently large, and speed of molecular diffusion becomes comparable with the dispersal process. This length scale is the equivalent of Kolmogorov length scale for diffusion of species. In chemical engineering nomenclature, the dispersal process is referred to as macro mixing while the molecular diffusion process is referred to as micro mixing. It is easy to think of analogous process for dispersal of thermal energy. As it turns out, the dispersion process is generally fast while diffusion leading to molecular homogeneity is slow. Thus, in many instances, it is good to think that the dye blob rst dis218

perses and then dye molecules diffuse from these smaller blobs. Thus the concepts of Taylors dispersion in turbulent ow is of value even in the context of soluble substances.

21.A.6 Taylors theory of turbulent dispersion


Let us return to the equation A21.1 relating diffusivity to the correlation between position and velocity. We have a very important result here. If the correlation between position and velocity is measured, the above expression can be used to calculate the diffusivity. Taylor 14 showed how this concept can be used to obtain turbulent diffusivities. We will jump a little ahead to make the discussion close to Taylors. The equation for species conservation or thermal energy balance can be written in a manner analogous to equation of motion and continuity. If there are no heat sources or chemical reactions, such an equation is simply a conservation equation. Thus we can derive the equation by adding the diffusion term to our continuity equation. Thus, for turbulent ow we will get for species balance

\ m M m A ]

Jb

Jb

The thermal energy balance will be given by

Taylor showed that these can be written as the product of a gradient and a turbulent or eddy diffusivity. He further showed how the concept of correlation can be used to measure the eddy diffusivity. In an ingenious way, typical of Taylor, he thought of the following experiment to focus on dispersion or eddy diffusion alone. Consider non conducting tiny particles suspended in a turbulent uid. For simplicity consider velocity as well as uctuations exist in only one dimension. Thus they move along a line. Then if we consider all particles crossing a plane at a given time, they all would have come from different locations due to the uctuating velocity. See Figure 21.5. Let the initial locations of all the particles reaching the plane at at time be  =. Thermal energy of particles can be dispersed only if they originally had different thermal energies. Hence, when the experiment started there was a temperature gradient in the particles. Let it be linear. Thus, taking the plane at as the reference point, the particle reaching from  = would have had an initial temperature given by

A P j % K P % We therefore have the turbulent uxes given by for species and T for thermal energy. \ m  M m
T

Jb

$b

where  is the mean temperature. The heat ux at time crossing plane at sum of the energy being carried by particles crossing the plane at :

  [     = A =    5  P % k = l  = l =

* 5

is given by

14

Diffusion by continuous movements, Proc. London Math. Soc., ser 2. vol 20, p 196-211, (1921).

219

b1 b2 b3 x axis x

Figure 21.5: Particles reaching a given location at the same time come from different places due to velocity uctuations.

But since l the particles are non conducting, they do not lose any heat as they move around.   [   = A = Hence   . Hence the heat ux at time crossing plane at is given by

[ P % k = l  =  = H  A P % k = l   Q  =   V =  A P % k = lo Q  =   = The term containing   will vanish since it is a constant and v= l is a random quantity. Let % us denote  = by , the distance traveled by the particle. The heat ux due to random motion

P % k = l  = l A =

* *

* *

of particles is therefore given by

 % l &%  P =  = k

* n

The summing over particles of course is the same as the averaging process over the plane F located at . Hence the heat ux in the direction due to random motion of particles, is therefore given by 
-

" F&% A

Hence the eddy diffusivity is given by the same type of expression for molecular diffusion, where is the eddy diffusivity. The difference is that the velocity uctuations are due to turbulence and not due to molecular motion. 220

" F&%

F l A

" &% *

P % k = l&% 
k

5 n

" &%

l&%

Taylor rewrote the above expression in a different way to make it more suitable for measurements. Since all particles have the same position at time , we can write

>  =A
%

=A A

, = * l
k k

Hence

" F&% l * l A *
k %

,%

= l

* , F = 5 l
k

where we explicitly show that and v should belong to the same particle by indicating that the averaging involves the velocity of the same particle but at different times. Thus, diffusivity can be found from the average of the product of the velocity of a particle now and its velocity at earlier times. This is called autocorrelation of velocity. For isotropic and homogeneous turbulence, this is equivalent to measuring velocity at the same point at different times and averaging it. Taylor proposed that this is a good estimate for many turbulent ows as long as the correlation dies reasonably quickly. Taylor calculated dispersion of smoke from stacks in atmosphere using this hypothesis and showed that it works well. It can be seen that Reynolds stresses, turbulent heat uxes and mass uxes of species can all be related to turbulent velocity uctuations in this manner, and this model would suggest that all eddy diffusivities are equal;

where is the thermal diffusivity, and

h " F&% A " F&% A ] = " F&%


] =

is the species diffusion coefcient.

221

Problems for Chapter 21.


21.1 Viscous dissipation has terms like If the ow is steady and turbulent, 4 4 i) what is the time average of this quantity? 4 4 I8 what is the time average of the above quantity multiplied by

I I

ii) what is the time average of the above quantity multiplied by j ?

I

222

Chapter 22 MACROSCOPIC BALANCES AND FIELD EQUATIONS


Now we revisit our discussion about Unit Operations approach. We had a few criticisms. 1. We use various average quantities, like velocity etc in making macroscopic balances while we do know that they vary across the cross section in an inlet or an outlet. What should be the correct average? 2. We use the engineering Bernoullis equation often. Since it can not come from the rst law of thermodynamics, what is its origin? 3. The dimensionless group approach through Buckinghams Pi theorem is used though the difculty with it is in the selection of variables. We hoped that the eld equation approach will throw more light. Does it? Now we want to see where we stand with respect to these1 .

22.1 Averaged eld equations


Here we are interested in the equations obeyed by the averages of various quantities. For this purpose we use the following approach. Consider some domain of interest. Typically it is an equipment, and it is stationary. Thus we have a stationary control volume of interest. We know that the eld equations are valid at every point in the control volume. Hence we average the eld equations over the whole control volume and derive equations for the averaged quantities.

22.1.1 Mass balance


Consider some control volume like the one shown in Figure 22.1. We consider that the control volume has only one outlet and one inlet for simplicity. We take planes at the inlet and outlet, which for convenience are marked by and respectively, as bounding the control volume. The rest of the control volume encloses the uid in the control volume as shown in Figure 22.1. The equation of continuity P
1

This chapter is very close to chapter 7 of BSL.

A bm P  4 4
223

Rotating shaft doing work

Inlet

n1 Plane 2 n2 Outlet

Plane 1

Figure 22.1: A stationary control volume. The control volume is shown by dotted lines.

is valid over the entire domain and hence the above equality would still be valid if it is integrated over the entire control volume. Hence
/$ & / #

where <>F BEF is the total mass contained in the control volume. The term on the right hand side of the integrated mass balance can be rewritten by using the divergence theorem

,  P  a A ,  bm P l a 4 4 boundaries are stationary, the term on the left hand side Since the control volume as well as its can be written as  H  <>F BF  P , A A  A a V
/$ #

,  bm P l a
/$ &

A A A A

where a quantity under the bar ; is the average of that quantity over the area of the the inlet and outlet respectively, and is the average mass ow rate. The velocity vector and the unit normal to the area will be in the opposing directions at the inlet and that explains the signs. At the rest of the control surface, the normal component of the velocity is zero by no slip condition. Hence it does not make any contribution to the integral. Thus the averaged equation or the macroscopic mass balance gives

, - P m U , P m g U M , P m K $ P ;a g  g M; P a K  K gM K


  $ $ 

 <>F BEF ; ;  A g K

which is equivalent to the usual statement that the rate of accumulation of mass is equal to the difference between rates of input and output of mass to the control volume.

22.1.2 Momentum balance


If we apply the same approach using the equation of motion, we get the integral momentum balance. Thus, we integrate

4 b 4

P M m P A O M m M P

b R t

224

over the control volume. Since the control volume as well as its boundaries are stationary, integral of the rst term on the left hand side is equal to

where F BEF is the total momentum in the control volume. Integral of the second term on the left hand side after converting it into a surface integral using divergence theorem gives

 , 

I A   U$  P #

, - m P  $ A , g m P  U M , K m P  U


$ $

It is seen that the average now involves quadratic terms in velocity. This can be written in a simple manner only if we assume that the velocity vector (time smoothed if necessary) and the normal to the area of the plane are parallel. Then the term is equal to

P a K gY g M P a K K K

where is the surface area vector whose direction corresponds to the outward normal to the area. The gradient of pressure when integrated over the control volume can be converted through divergence theorem into an integral over the control surfaces:

/$ & O  a A

, b

, - G O $
$

This term will make a contribution over the inlets, outlets as well as the other surfaces. The former is easily written as where it is clear that the pressure must be averaged over the area of the inlet or outlet. The other contribution can be written as the pressure forces exerted by the walls of the control volume on the uid. Denote it by . The forces due to deviatoric or viscous stresses can not be written simply due to their dependence on velocity. Thus, the integral of m over the control volume is simply represented by . It is easily seen that the force due to gravitational body forces is given by <>F BEF . Thus the integral momentum balance is given by

O g g M O K K

prole is at. Thus these averages are not very different under turbulent ow conditions when the velocity prole is quite at. The above equation can be easily interpreted. The total momentum in the control volume changes since (i) it is convected in or out of control volume, (ii) by the action of pressure and viscous forces on the control volume, and (iii) the action of body forces. If we interpret that the effect of forces is to generate momentum, the above equation falls into the pattern of balance laws widely used in chemical engineering, viz rate of accumulation is equal to net rate of input plus rate of generation. 225

F BEF  A P a K g g M P a K K K g g M K K M M M <>F BEF O O It is seen that the velocity averages are more complex. Before we leave this topic, it is worthK while pointing out that averages like a and a will be not signicantly different if the velocity

*t  A

b R

22.1.3 Mechanical energy balance


We now look at a new aspect. We have talked about drag forces while solving equations of motion. If there is movement corresponding to the exertion of force, work would have been done. It therefore appears interesting to see if work done can be calculated.

22.1.4 Equation of mechanical energy


This can be done by taking a dot product of the equation of motion with velocity. We will use indicial notation for ease of manipulations.

We note that

I= PI= A I= = =

IK = 4 4 4 4 with respect to any 4 coordinate also. Hence the and that similar relation holds4 for derivative I=
" "

4 4

PII= M

I= 6 A L

I =I = A 6 L =

O =
"

4 4

c =

P R =

equation can be written as

6 PIK A L

= M M =

6 PIKI L I = O = M O = = I =  = = P R =I =

4 c = I = V 4 4 R
M Pm

4 H=

I=

This is more compactly written in vector form:

This equation is easily interpreted. The kinetic energy in the control volume changes rstly because of mass carrying kinetic energy enters or leaves the control volume. The pressure forces do ow work. Out of this some goes into storing energy if the material is compressible. This part corresponds to m . The difference goes into increasing the kinetic energy in the control O volume. Similarly viscous forces also do work, called shaft work, and this term corresponds  to m 3m . All of the shaft work can not go into increasing kinetic energy since some of the work done is converted into heat due to irreversible nature of viscous friction. The latter corresponds to . Body forces are conservative and all work done by the body forces increases the kinetic energy.

6 P I K  A 6 m P I K  m  m m 3m  L L M O M O

b R

b R

226

22.1.5 Macroscopic mechanical energy balance


The equation of mechanical energy can be integrated over the control volume to get the macroscopic balance of mechanical energy. We give a simplied version corresponding to incom A ) so that the similarity between the engineering Bernoullis equation pressible uids ( m widely used in Unit operations approach can be seen. For the special case of stationary control volume, the boundaries of the control volume are not moving ; and incompressible uids, the inlet and outlet ow rates must be equal. Let it be denoted by . It itself can be a function of time. Integral of the term on the left hand side of the mechanical energy balance equation is easily evaluated to be  F BEF 6

Clearly F BEF is the total kinetic energy of the control volume. Integration of the rst term on the right hand side of the mechanical energy involves triple velocity product. It can not be simply expressed unless the velocities at the inlet and outlet are parallel to the normals to the respective areas. In that case, the integral of those terms is shown to be




IK c F BEFA L  # P  a

6P  I= 6P  IB 6 ; I= 6 ; IB L a = = I = L a B B I B A L I = L I B

The term involving pressure similarly is equal to

OP

= ;

OP

B ;

if we assume that pressure across the area is constant at both the inlet and the outlet. The term containing the divergence of deviatoric stresses, after applying divergence theorem is equal to

, - U m R  m
$

If we neglect the contribution of normal stresses doing work at the inlet and outlets, this term has contributions only on the other parts of the control surface. On all the control surfaces, except where moving parts are involved, the velocity is zero by no slip condition. Thus the only nonzero term consists of the product of the force exerted by the moving part of the control surface on the uid (in view of the outward normal convention being followed) and the velocity created there. This is ; of course the rate at which work is being done on the control volume by the surroundings, ; . The integral of the term is obviously the losses due to friction, which are denoted by . Let us consider the special case where the body force can be represented as the gradient of a potential2 , . Noting that m = 0 for incompressible uids, the integral of the last term then is equal to

R b

If

+
2

 a A P P / m

, b +

, -um7+ $

is assumed to not vary across the area of an inlet or an outlet, then, the term is equal to

If gravity is the only body force, then opposite to each other.

; + = ; + B d

, where the direction of gravity and the

coordinate are

227

Thus the equation is written in the form

 F BEF 6 ; I = 6 ; I B B ; = ; ; ; = ; B  A L I = L I B M O P O P M M

+ +

At steady state, ow rates at the inlet and the outlet are equal and the accumulation term is zero, and the above simplies to the engineering Bernoullis equation:

where and , respectively, are the ratios of rates of shaft work done and frictional losses to the mass ow rate.

6 I= O P M L I = M =

B 6 IB =A O P M L I B M

+ B M /

22.2 Dimensional analysis


The approach of dimensional analysis based on the eld equations is to formulate the problem in terms of the relevant equations exactly. Then the relevant scaling parameters are chosen 3 . The equations and boundary conditions are then nondimensionalized. The solution of these equations would depend upon the position and the dimensionless groups that appear in the equations. This form can be used to draw conclusions. Let us take a few examples.

22.2.1 Pressure drop in a tube of varying cross section

A B 6 M j T  ) Let us say that the pressure drop over some length W is the quantity of interest when the average velocity is prescribed to be a . See Figure 22.2 The axial velocity can be scaled with a . ) is
Fluid in r z Ro k Ro Fluid out

Consider a tube of varying cross section and for concreteness let it be a sinusoidal function of length along the axis:

Figure 22.2: Tube of varying cross section a length scale relevant to T . We expect the radial velocity to be present since K BK the radial cross section is varying. The change in cross sectional area is represented by j . We can select a
Scientic research can not be automated. There will always the choices and judgments to be made. Scientic approach offers a more reliable method to make the judgments. Thus, in the Unit operations approach also, we had to choose variables that would inuence the process. The eld equations approach tell us the various phenomena at work and hence more insight is available through earlier exercises to make choices.
3

228

velocity scale for radial velocity from this estimate by using the equation of continuity. It will be given by

I A I I A 6 ) Ir c r jK B au au B There are several length scales available here: c)~cW . We can use all these based on our experience. We know that viscous forces depend upon the velocity gradients in the radial direction. Thus we could use B to nondimensionalize the radius. We also expect that the uid
Hence we have will accelerate and decelerate as it ows through the duct. The acceleration will inuence the inertial terms and we have accounted for this by estimating the magnitude of the radial velocity. periodic solutions. We We can select ) to nondimensionalize the axial coordinate anticipating K P need to nondimensionalize the pressure. We can choose a after taking a difference with the inlet value: =

auGj

K B
)

or if the characteristic length on which the radius varies, ) , is large. We can see this reected  6 ^ B by the dimensionless group j ) . Now let us move on to equation of motion. Substituting these, after nondimensionalization the p component equation of motion is given by

6 I  I a 6 I  a j K B I  A A K p 4p p r M 4 T B p 4p p r M T 4 or 4 4 6 4 I  4 K BK I  A p 4p p r M j K 4 T The above equation reects our understanding well since 4 4 we do not anticipate solutions to be different from that of a straight tube if either the variation in the radius is small i.e., j is small
The equation of continuity then gives
) )

A O OK Pa

H 6 I  BK K I r K B I I r I I r A h j r p M T jK B p M B p p p p r V M K TK a 4 4 4 4 4 4 4 of motion 4 becomes 4 4 4 equation The T component4 of the BK K I 6 H I K I I I I I A h j r r p M T T M B p p p p V M K TK au 4 4 4 4 4 4 4 can 4 be 4 as 4 4 4 The boundary conditions written H B I A I A  6 A r at p M j ,8 T[V and
) ) ) ) )

The solution would be of the form:

B I A I pUc T c G c jc ) V c

B I A I r pcT c G c jc ) V c and r
229

A 

at

 T A

and A

B p cT c G U c jc ) V

A a B . Suppose we are interested in pressure drop over some length W . We where see that it can depend upon radius also. So we have to dene what is meant by pressure drop. Suppose we dene it to be the average over the cross section, . Integration of removes the dependence on p and we will get

B A T c G c jc )

The dimensionless pressure drop will be unaltered if the dimensionless groups appearing in the solutions are kept constant. Thus a measurement made in terms of dimensionless quantities on a small scale will be equal to that for the real scale provided the dimensionless groups are kept constant. The dimensional pressure drop is given by

l % A " O /

 K W cGjc  A P u a B O l W  P au  K  B cGjc 

Thus if the dimensionless groups are kept constant, we can write

" / %

"/ / %

This is the basis of scale up of equipment. Often, the dimensionless groups themselves can fall into two classes. In our example we see ^ that the groups j , and B ) are related to shape of the duct or geometric in nature. When this group is constant between the small and big scale equipment, we say that geometric similarity exists. The other type of groups are those which contain conditions of operation. An operating condition in our example is the velocity, and the dimensionless group is Reynolds number. If this group is also kept constant between the small and big scale equipment, we say that kinematic similarity exists.

22.2.2 Drag on a sphere due to oscillating ow


We examined uniform ow past a sphere of radius . Suppose we are interested in knowing the drag force when the ow far away from the sphere is oscillatory:

and a can be selected as the length and velocity scales. Let us also choose inertial forces to nondimensionalize pressure over and above the value far away from the sphere. There is a choice for the time scale. The frequency of oscillation as a time scale. The viscous diffusion time scale is also relevant. One might expect that the ratio of these two time scales would give information about the need to take the oscillatory nature of ow into account as opposed to performing a pseudo steady state analysis. As it turns out it does not matter which is selected since all relevant groups will appear in the end. However, selecting one scale as opposed to the other can offer an insight into the nature of and solution to the problem. Let us select the diffusion time scale for nondimensionalizing the time. Thus

 A a o l

A a A O P a OK

p Z A

A h K

230

6 K I  6 I   A  p K 4p p r M p  v 4p t v 4 since the same length and velocity scales The equation of motion can be 4 written in vector form are being used in both components of the equation of motion 6 6  K M mJb A b M kb 4 The boundary conditions would 4 read p A 6 c3 A  and H K p A o8 h V Thus the solution can be formally written as H K A pUcv$c c h V H and K A p$cv$c c V h
The nondimensional continuity equation would be

Notice that one dimensionless group has come from the boundary condition. The nondimensional groups now are the Reynolds number and the ratio of the time scale of diffusion to the time period of oscillation. The drag force can be computed from the velocity proles. The drag coefcient is its usual nondimensional representation. It is given by

`K P K k 6 a L k  K v m m  v A K K P k L a H _ v A P K  v m P a K 5M a K m V g a r H 6 A L   v v QM m ~V g r The integral is only a function of the dimensionless groups and hence H K A c V A

,  D " % k R ,  ,  R % h

The drag forces can be scaled up exactly the same way as discussed with the previous example.

231

22.2.3 Scale up problems


We have broadly discussed problems involving scale up. The approach where dimensionless groups are kept constant between two different sizes of experimentation is usually practised in a generalized manner. For example, correlations between dimensionless groups are developed, which can be used later. This is what is seen in books on Unit operations with the examples of drag on objects when a uid ows past them or for ow in ducts and so on. Here a plot of a drag coefcient or friction factor is presented against Reynolds number which can be used across a wide variety of sizes. It is feasible to do this only for very commonly used equipment. More specialized equipment have to be treated as special cases. In such cases the approach where dimensionless groups are kept constant between two different sizes of experimentation may not be feasible. For example, it might not be possible to reach the values of dimensionless variables realized on a big scale in an appratus of smaller scale. The well known example is that of design of ships. Here the Froude number and Reynolds number are the relevant dimensionless groups. Let us use lower case letters for small scale and capital letters for hte bigger scale. Thus to keep Froude number constant we need

IK A K a W
while to keep Reynolds number constant, we need

I A W a

if we have to experiment with the same uid. These two are contadictory. If we decide to change the uid, we require the kinematic viscosity of the uid to be used on small scale ^ then to be equal to W  ^ W  . Even for a 1:10 scale up ratio, this number is of the order of 1/30. It is almost impossible to nd uids with such widely different kinematic viscosities. It is here that physical insight into the problem has to play an important role. It is interesting to note that Froude realized that scale up of drag on ships is more accurately realized by giving importance to the number subsequently named in his honor, and ignore the inuence of Reynolds number!

232

Chapter 23 FIRST LAW OF THERMODYNAMICS & HEAT BALANCE


We discuss the application of first law of thermodynamics to control volumes. We then show that under some assumptions, the usual heat balance can be obtained from it. First law of thermodynamics is the most fundamental law dealing with energy, just as Newtons second law is to determining motion of uids and solids. Heat is one form of energy and application of heat or removal of heat is an important aspect of chemical processes. Thus it is necessary to know the effects of such operations. However the only law we have is the rst law concerning the total energy. Temperature is the most common variable used to characterize the effects of addition or removal of heat. Thus, it is often desired that the temperature distribution in an equipment be known. In this chapter we wish to develop rules which permit us to focus on the effects of heat, which is only one form of energy, and also to relate it to the temperature of the system.

23.1 First law of thermodynamics


The rst law of thermodynamics can be thought of as a law of conservation of energy. It states that the change in the internal energy of a closed system is equal to the heat supplied to it minus the work done by it on the surroundings:

 A

The above formulation is applied to changes between two equilibrium states of the system, and internal energy is a property of the material at equilibrium. We are interested in nonequilibrium systems and here other forms of energy are encountered. In particular, kinetic energy1 of matter in motion is of great interest. The rst law is then restated by joining kinetic energy with the internal energy. Thus it is proposed that the change in internal 2 energy and
We are referring to motion at the macroscopic level and not at the molecular level. Internal energy is a thermodynamic function and is not dened for a non-equilibrium process. It is assumed that if the system is not very far from equilibrium, the functions as dened in thermodynamics can be used. Thus, temperature can also be dened, and the relationship between it and the internal energy is the same as that found in
2 1

233

kinetic energy is equal to the heat supplied to it minus the work done by it on the surroundings:

Kinetic energy

 A

If the processes causing the changes occur over a small time interval, the above equation can be divided by the time interval and, in the limit of the time interval vanishing, obtain an equation in terms of derivatives of the left and right hand sides with respect to time. Thus, the rst law of thermodynamics can be reformulated as follows: The rate of change in the internal energy plus the kinetic energy of a closed system is equal to the rate at which heat is supplied to it minus the rate at which work done is by it on the surroundings.

23.2 First law and closed systems


23.2.1 Rate of change of energy
Let  be the internal energy per unit mass of the system. The total internal plus kinetic energy of a closed system is given by 6

,6 - P  " F&%

M L

I K l

The rate of change of this is given by

 ., - P  " F&%

M L

6 I K  a

23.2.2 Rate of work done


Both body and surface forces act on the system. The work done by these has to be calculated.  P Consider the work done by body forces. A little volume a inside the system has a mass a . P a . When it moves with a velocity v, work is done by the body The body force on it is P  forces at a rate given by m a . Hence the rate of work done by the body forces on the entire system can be written as

tt

Rate at which work is done by body forces

, - P t m  a " F&%

The surface forces can the stressltensor. The force exerted by the surround
be calculated from 
. If the velocity on the surface is v , then ings on a little area of the system is m 
O m rate at which work is done by the surface forces on the little area is given by m Ogiven Thus, the rate at which work is done by the surface forces on the entire system is by

A Rate at which work is done by surface forces

,-

" &%

m O F

R  m 

equilibrium systems. What is meant by far from equilibrium is dealt with more precisely in books on irreversible thermodynamics. High speed ow is an example of the exception. It is sufcient to mention that almost all situations normally encountered in chemical engineering are not very far from equilibrium.

234

23.2.3 Rate of heat supplied


Heat can be supplied to the system in two modes. It can be supplied inside the system through out its volume. For example, it can be due to absorption of radiant energy. Another way heat could be supplied is by passage of electricity, i.e., ohmic conversion of electrical energy into heat. Yet another way heat could be supplied, generated is a better description, is by conducting chemical reactions3 . All these processes supplying heat occur through out the volume, and are similar to body forces ; in transfer. Let us denote the rate of heat .momentum The rate at which heat is supplied to the system is supplied per unit mass by the symbol then given by

,. - P ;  a " F&%

The second mode in which heat can be supplied to the system is through its surface. We expect this rate to be proportional to the surface area. Thus we also expect it to depend upon the direction of the area chosen4 . Hence the rate of heat supplied per unit area should be a vector. Let us denote this by . Then the rate at which heat is supplied to the system through its surfaces is given by

, - m C 
" F&%

It can be seen that this is entirely similar to the calculation of surface forces encountered in momentum balance. There, a stress tensor was needed to calculate the force. The tensorial order of force is one less than that of the stress tensor. The rate of supply of heat is a scalar. Thus, to calculate this, we have to use the heat ux vector, a quantity which is one greater in tensorial order than the scalar. Putting all this information together, the mathematical statement of the rst law is therefore given by

, -

" F&%

P ;  a

, - mJC 
M , - m O " F&% " F&%

 6 I K l  m 
 P P A  m a a M   F  M L F

,- t " &%

 , - " &%

After applying the transport theorem we can rewrite it as

(23.1)

, -

23.3 Equation of mechanical energy balance


The above equation is a statement of conservation of energy. As mentioned in the beginning of this chapter, we are interested in thermal energy or heat only. Rate at which work is done
We need not write this separately if we consider that chemical reactions generate several species and carry out the analysis for multicomponent systems, as we shall do in later chapters. For the moment we are restricting ourselves to single component systems. Due to this we have to add this in this ad hoc manner. 4 Recall the area averaging discussed earlier.
3

, - m C 
M , - m O R  m 
" F&% " F&% " F&% P , IK  a A , P P P  t , m a  m M 6M L M 6- " F&%k4 - " F&% " F&% 4
P ;  a

M L

6 I K 

(23.2)

235

is related to application of force and the velocity created in response. Equation of motion describes the response of uid to applied forces. Thus, it should be possible to derive an equation for the work done, and its effects, from the equation of motion. The effect of the work done could be to increase the kinetic energy or increase the gravitational potential energy and so on. An account of the effects of work done can be thought of as mechanical energy balance. The rate of work done by a force is equal to the dot product of the force and the velocity created. Thus, if we obtain the dot product of the equation of motion with velocity, we can derive an equation of mechanical energy balance. The equation of mechanical energy balance has been derived in this manner in chapter 22. It is given by

Let us reinforce the meaning of the last two terms of the above equation. The fourth term on the right hand side is the rate at which work is being done by surface or viscous forces. Viscous forces arise from friction. Hence some of the work done by viscous forces has to result in heat generation. The last term represents the rate of heat generation due to friction. Thus the sum of the fourth and the fth terms is the work done by viscous forces minus the rate at which it is lost into heat. The equation then represents the mechanical energy balance: when work is done, a fraction of it is lost in heat, the rest goes into raising the kinetic energy of the system and or storing energy in the form of a more compressed state. The equation of mechanical energy balance is valid at every point. Thus it is valid in the entire domain of the system. Thus, if we integrate this equation over the entire domain of the system it will still be valid. Carrying this out and using Gausss divergence theorem we get

6 P I K  m 6 P I K  A P m m L L M O M O m M m m

t b

b R R b

balance in chapter 22.

,6 - P t m  a , - O m 
M , - m R3m 
M ., - O bm  a F&% F&% " F&% " F&% " " 6 P I K l 6 PIK ,  , A , (23.3) .- R b a 6- L a M - m L
" F&% " F&% 4 " F&% 4 In fact, this is the form with which we started to derive the equation of mechanical energy

23.4 Thermal energy balance equation


We subtract the equation of mechanical energy balance equation from the equation obtained by applying rst law to a closed system. The resulting equation is about the effects of heat and is the thermal energy balance equation:

, -

" F&%

P ;  a

, - mJC 
 , - O b m  a M  , - R b " F&% " F&% " F&%

a A  F

For incompressible uids, it is more convenient to put the equation in terms of enthalpy. Let be the enthalpy per unit volume. Then

, - P  l a M , - m P  
" &%k4 " F&% 4 (23.4)

A  MNO P
236

After adding suitable terms to the left and right hand sides and use of the divergence theorem, the following equation is obtained

, 6- P ;  a , - mJC 
, .- ]] O  a M , 6- R b

a A 

23.5 Heat balance over a control volume


The balance equations derived for system were converted to those useful for stationary control volumes by applying them to a system which occupies the control volume of interest, e.g., in deriving momentum balance in chapter 6. Following a similar procedure to the thermal energy balance and after some rearrangement, we obtain

, .- P 4 4

 a P m+  M

,-

(23.5)

A , 

;   l  P  a A P P m a m 
M 
M 

, C

, R b

]   a  ] O a

This equation is easily interpreted. The rate of change of enthalpy of a control volume is equal to the rate of net input of enthalpy into the control volume by convection plus the rate at which heat is being generated in the control volume plus the rate at which heat is being supplied through its surfaces plus the rate at which heat is being generated due to viscous dissipation in the control volume minus a small correction due to compressibility effects. Ignoring the last term therefore we can write the heat balance as follows: Rate of genRate of Rate of Net rate Rate of ineration of generation accumuof input of put of heat + heat due + of heat due lation of = enthalpy into + through to chemical to viscous enthalpy in the CV by surfaces reactions and dissipation convection the CV other means It may be noted that this is in the form that has been used in the undergraduate books on Unit Operations. It is this form that we will also use in making balances.

23.6 Heat ux through boundaries


The above balance equation contains q, the heat ux through boundaries. This, as remarked earlier, is similar to the surface forces encountered in momentum balance. Hence we expect that an expression for it is needed before we can solve the resulting balance equation. Why does heat pass through surfaces other than by the physical movement of uid? In chapter 9, we saw that in the presence of velocity gradients, momentum ux (stress tensor) occurred through surfaces by diffusion. We can expect a similar kind of mechanism by which heat is transported through the surfaces. Thus a constitutive relationship is required for the heat ux. The mechanisms that cause heat ux and the constitutive relationships will form the subject of our discussion in the next chapter.

237

23.7 Connection to temperature


In the beginning of the chapter, we said that one of our objectives here is to ascertain the effects of heat transfer on the system and to relate it variations in temperature in the system. The thermal energy balance equation meets the rst objective. The next step is to relate this to the temperature. It was mentioned earlier in the chapter that, as long as the system is not far from equilibrium, it can be assumed that the thermodynamic state functions like, enthalpy etc and state variables like temperature etc can be assumed to exist and relations between them are the same as those obtained under conditions of equilibrium. Hence, for a single component system,

 c  A O

Further, other thermodynamic relationships derived to relate changes in enthalpy to temperature are also valid. Hence

A A

 %   M "  O O %n  M "  O O

Thus, it can be seen how the thermal energy balance can be related to temperature by using the above relationships. Further simplications are made by relating the changes in enthalpy due to changes in pressure to changes in density. We just show it here to once again emphasize the idea that thermodynamic relationships are used in non-equilibrium situations. Thus,


Hence

6   A  M P O 6

4  A 44  4 where we have used a Maxwell relationship to derive the second equation from the rst. Summarizing,  A %  6 6 H  P  (23.6) V M P M %  4  O 4

" A O

P M  " H O 6 6 P M   PV % 6 6 H P P M  V %

238

Chapter 24 MECHANISMS OF HEAT TRANSFER & CONSTITUTIVE RELATIONSHIPS


24.1 Heat Flux Vector
Energy and temperature are concepts well grounded in Thermodynamics. Heat can be thought of as energy in motion. A system absorbs heat from surroundings which are at a temperature greater than itself. Consequently its internal energy increases. Here, heat is said to ow through the boundaries separating the system and surroundings. As heat ows through surfaces, it is natural to assume that the ow rate of heat will be proportional to the area of the boundary. Hence heat ux, i.e., the rate of ow of heat per unit area, will be an important quantity to understand. Any moving quantity will have a direction and hence will be vector. We designate heat ux vector by . We now discuss mechanisms of heat transfer and the evaluation of the heat ux vector.

24.1.1 Direction of heat ow


Suppose two blocks, 1 and 2, of equal mass of same material but at different temperatures are brought into contact with each other at constant pressure, and insulated from the surroundings. Let their initial temperatures be  g and  K . After some time, they will attain thermal equilibrium and will be at the same temperature. The temperature of one block must decrease and the other must increase. We want to know the direction of heat transfer or in other words which of the two blocks will cool. The condition of equilibrium for isolated systems (i.e., constant mass and energy) is    Thus, any process for which

 l8 

is feasible and will occur leading to equilibration. We will use this criterion to evaluate the direction of heat transfer. Suppose a small amount of heat ows from block 1 into block 2. Denote it by - g K . Let the temperature of block 1 will change by  . By conservation of energy, the temperature of block 2 will change by  . The change in entropy of the system for this process is

 A - g K - g K A - g K  g K  M g g K K    
239

. Thus if - g K , then  g  K , and if - g K , then But for this process to occur, g  K  . Hence heat ows from regions of higher temperature to those at a lower temperature.
24.2 Heat Conduction
Conduction is one mechanism by which thermal equilibration occurs when two bodies at different temperatures are brought into contact, and isolated. Matter is composed of molecules which are in random motion. As molecules execute random motions, they collide with each other and exchange energy. It is through this process that energy is conducted from location to location. Thus, heat can be conducted only through matter. Simple models based on kinetic theory give a good picture of how heat is transported through conduction in gases. Heat conduction in solids occurs by more complex mechanisms. These are discussed briey in the text by Bird, Stewart and Lightfoot, Chapter 9.



24.3 Fouriers law of heat conduction


As heat ux is zero when there exists no temperature difference, we expect it to be proportional to the temperature differences. We can therefore think of the temperature difference to be the driving force for heat transfer. As heat ux is a vector, we expect it to be related to a vector connected to temperature difference in space and to the direction of its difference. Intuitively, it can be seen that temperature gradient satises these requirements. Fourier formulated the relationship between the heat ux vector and the temperature gradient, and in his honor, it is known as Fouriers law of heat conduction: In above, j is the thermal conductivity, and is a property of the material. Relationships, between the rate of an irreversible process to appropriate driving forces are referred to as constitutive relationships. Thus Fouriers law is analogous to the Newton-Stokes law of viscosity. The Fouriers law is the constitutive relationship for heat conduction. The relationship depends upon the constitution of the material, e.g., its chemical compositions, physical state etc. All known homogeneous and isotropic materials satisfy Fouriers law of heat conduction. This is in contrast to the situation in momentum transfer, where a large class of materials e.g., polymers, do not satisfy the Newton-Stokes law of viscosity. A comment about the negative sign is in order. Heat should ow from regions of higher temperature to those of lower temperatures, and the negative sign ensures this. Consider the component of the Fouriers law:

A j 

A j  Thus, if the temperature decreases in the positive direction, heat should ow in the positive direction or - should be positive. If temperature decreases in the positive direction, then ^  is negative, then - is positive according to Fouriers law, and that is consistent.
-

24.3.1 Direction of heat ux vector


Here we want to consider further the relationship between the heat ux vector and isotherms is an innitesimal length vector, change in temperature  along to gain some insight. If 240

4 4

the direction pointed by the innitesimal vector is given by

 A $m 

B Jb

Take a reference point on an isothermal surface, and let be the tangent vector to the surface at that point. Consider another point on the surface away from it by a small length : , and along the tangential direction to the isothermal surface. The temperature difference between the two points is given by  

 A

: [m 

But since both the points are on the isothermal surface, i.e., the surface on which temperature remains constant,  = 0. Hence  Hence

F b

Hence, is normal to the isothermal surfaces. In other words, heat ux takes the direction of steepest temperature change.

F8m$b  A Cm
F A 

24.4 Radiation
Radiation is of molecular origin. When a change in any of the rotational, vibrational, and electronic energy levels of a molecule occurs, electromagnetic radiation is emitted. This is the source of radiation from materials. When radiation falls on a materials surface, it is absorbed by molecules and their rotational, vibrational, and electronic energy levels are altered. Due to collisions or other means of molecular interactions, this energy also gets partitioned into translational energy. Thus, the temperature of the material rises. Unlike the conduction mechanism, radiation does not need any medium for it to propagate and carry energy. Molecular states are continually undergoing transitions, and hence a body always emits radiation. Hence, if its temperature has to remain constant, it must receive heat from other sources. Bodies radiate energy ux depending on their temperature. The maximum energy that can be emitted by a body can be derived from laws of thermodynamics, statistical mechanics, and quantum mechanics. A black body (it is really a hypothetical body) is dened as that which radiates energy at the maximum rate at a given temperature. The energy ux emitted by a black body is given by Stefan-Boltzmann law

< , and In above, is the Stefan-Boltzmann constant and is equal to 5.67  is absolute temperature in . Bodies however emit radiation at a lesser rate, though the dependence on temperature remains the same. Due to the strong dependence on temperature, radiation becomes a dominant mechanism of heat transfer at high temperatures and or when other mechanisms of heat transfer are weak. Radiation falling on the surface of a body can be partly reected, and the rest can pass through the body, and while it does so, can be absorbed. We do not consider this topic any more and for details, one can refer to 1 or other standard texts on radiation (Siegel).

1

r A
-

E

6

^ K

See text by Bird, Stewart and Lightfoot, Chapter 15

241

24.5 convection
There is one more way in which heat can be transported from place to place. Fluid can move from one place to another and of course it carries energy with it too. It is this mode of heat transport that is represented by the term

m P 

in the thermal energy balance which was derived in the previous chapter. As uid moves from one place to another, it can encounter environment at different temperatures. Then it will exchange energy with the environment and energy transfer occurs. The intimate exchange leading to spatial homogenization of temperature over length scales of molecular dimensions still has to occur by conduction. Thus if we had uid particles of zero thermal conductivity, they would not exchange energy even when they pass through zones at different temperature due to convection. In this sense there is some difference in the way energy can be transported by convection as opposed to by conduction. As thermal conductivity is not identically zero for any uids, energy transfer by convection simultaneously with conduction. Convection depends upon motion. Hence the energy transfer by convection depends upon ow conditions also. This feature links calculation of the rate of energy transfer with the equations of motion. A uid can move due to the action of an external agency: a pump etc. Heat transfer accompanied by such motion created (or forced) by an external agency is dened as forced convection. Motion can also be created due to a combination of gravity and density differences. The condition where a heavier uid is situated on top2 of a lighter uid is unstable and uids will move in an attempt to attain a more stable conguration. Such an instability can occur when a uid is heated. Depending on where heating is carried out, it is possible that hotter, and hence necessarily less dense uid, is below the colder or more dense uid. Then motion is created naturally when the system attempts to reach a more stable state. Heat transfer accompanied by such motion is dened as natural convection.

24.6 Classication of heat transfer problems


Heat transfer problems can be broadly classied into two categories. If heat is not transferred across the boundary of a control volume by convection, the only means of heat transfer is by conduction. Such problems are referred to as heat conduction problems. Such is the case with heat transfer in solids. If heat transfer occurs in a uid in such a manner that natural convection is avoided and there is no external agency causing motion, only conduction is left as mechanism of heat transfer. Analysis of such problems are similar to heat transfer in solids. As opposed to conduction heat transfer, convection is also present in convective heat transfer. Convection problems can further be classied as natural or forced. More detailed classication can carried out based on whether ow is laminar or turbulent ow, ow is fully developed or developing. Problems typical of these will be considered in the following chapters.
2

Gravity is being assumed to act downward

242

24.7 Similarities and differences between momentum & heat transfer


Conduction mechanism of heat transfer is similar to the mechanism by which momentum is transferred in uid motion. Just as momentum ux depends upon velocity gradients, the heat ux depends upon the temperature gradient. Thus Fouriers law is analogous to Newton-Stokes law. Thus, the Fouriers law of heat conduction is similar to the Newtons law of viscosity. Heat ux is similar to the momentum ux. Velocity is similar to temperature. They are only similar and not identical e.g., heat ux is only a vector while momentum ux is a second order tensor. A major difference is with non-Newtonian materials. Fortunately, no non-Fourier materials have been found. Thus there is no similarity between momentum transfer and heat transfer in this respect. This does not mean that momentum is one up on heat transfer. Heat can be transferred by radiation, and there is no analogue of this kind of mechanism in momentum transfer. Thus, we can expect similarities between the problems of heat transfer and momentum transfer in Newtonian uids. Thinking about analogous problems between the two deepens the understanding of both.

243

Chapter 25 PHYSICS OF HEAT TRANSFER


The development of temperature profiles in various systems is discussed. The boundary conditions needed in solving heat transfer problems are discussed. In this chapter we consider a few simple situations involving heat transfer and focus on describing the development of temperature proles to get a physical feel. A temperature gradient must exist before heat transfer can occur. The temperature gradients usually come into being when two phases at different temperatures are brought into contact. Consider a simple example. Suppose a pan of water is to be heated. This can be done by placing an electrical immersion heater into it. As current is passed in the heater, the surface of the heater becomes hot. The water however is at room temperature. Hence a temperature gradient is setup and heat ows from the hot surface of the heater to water. A description of development of the temperature prole in water is possible when the thermal energy balance equation is solved with suitable boundary conditions. All types of boundary conditions used in the context of heat transfer will be stated a little later but one of them, which is essential to discuss qualitatively the development of temperature proles, is given rst.

25.1 Thermodynamic equilibrium at the interface


A basic hypothesis made in solving heat transfer problems is that the interface between two phases is always at thermodynamic equilibrium. Thus the temperature in the two phases at the interface is the same or the temperature is continuous across the phase boundary. This is one of the boundary conditions for heat transfer problems. Consider once again the example of heating of water by an immersion heater. The boundary condition states that the temperature of the surface of the heater and water adjacent to the heater are at the same temperature. It is worthwhile to notice the analogy between this boundary condition and the boundary condition of continuity of velocity across uid-uid or uid-solid interfaces (refer to Figure 10.3). Thus, referring to Figure 25.1,

e e
 A 
244

Phase II

Phase I

Figure 25.1: Boundary conditions for energy transfer

25.2 Development of temperature prole: Constant temperature case


Let us reconsider the example of heating of water by an immersion heater. For convenience, let us approximate the immersion heater by a plane. Let us also assume that water extends to innity in the direction perpendicular to the plane which approximates the heater. In this problem, soon after the current is switched on, the surface of the heater gets hot. The temperature of the uid adjoining is however unchanged, except right next to the heater surface. The uid temperature at the uid-heater interface, according to the boundary condition just stated, must be the same. Hence, a large temperature gradient is established in the uid phase and heat will be transferred from the heater surface to the uid, the rate of which is dictated by Fouriers law of heat conduction. The heat coming in will raise the temperature of the uid layers adjoining the heater surface. Hence, the temperature increase will percolate deeper into the interior of the uid phase. Thus, heat ux will be established into deeper and deeper regions of the uid. It is easily seen that a part of the heat a thin layer of uid receives goes into raising its temperature and the rest goes to the next layer in the direction of decreasing temperature. This is shown in Figure 25.2. In drawing this plot we assumed that a controller maintains the temperature of the heater constant, and that uid will remain stagnant. We will discuss more of these aspects shortly.
Temperature of heate r and fluid the same at the interface Increasing time Heater Fluid

Figure 25.2: Development of temperature In this problem, the heater is a source of heat. The domain of heat transfer extends to innity and hence a heat sink is not present. Thus, the temperature proles never reach a steady state. Suppose the uid was placed between a hot and a cold plate. In this instance, the cold plate will act as a sink and steady state will be reached when the heat owing from the hot plate is removed from the cold plate. It is worth while to notice the similarities between these problems and the uid mechanics or momentum transfer problems we considered earlier. The boundary condition of constant temperature is similar to the no slip boundary condition, in particular, the constant velocity boundary condition imposed by a moving plate. The problem of a heater in an innitely large uid is then similar to the problem of an innitely wide plate suddenly set in motion in an otherwise quiescent uid of innite extent. The problem of uid in between a hot and a cold plate is equivalent to the Couette ow between parallel plates. 245

From Figure 25.2, it is seen that the temperature gradient decreases with time. Hence the heat ux decreases according to Fouriers law. Thus, as mentioned earlier, a controller is needed to regulate the heat ux to keep the temperature constant. It is possible to imagine that we have a surface which can deliver a specied or programmed heat ux too. Analysis of such problems will need more general boundary conditions, and this is what we will discuss now.

25.3 Boundary conditions


We will now formalize the boundary condition of temperature continuity, and state the other boundary conditions. Refer to Figure 25.1. We let the phase boundary move with a velocity 1 V . The important aspect here is that the velocity of the interface is different from the velocity in either phase at the interface. Such a situation arises, for example, when water vaporizes. We will examine this shortly. Let us consider a mass balance made by an observer sitting on the interface. Let us consider a single component system, and we will generalize this for multicomponent systems, there by permitting reactions, in a latter chapter. Consider a little area
on the interface. The rate at which mass reaches the interface from phase is given by

P \  m

face. This equation can be interpreted. Either side of the above equation is equal to the net ; mass ux crossing the moving boundary, < . The above equation says that the mass ux must be continuous across the interface in the absence of chemical reactions. If it is known that mass ux is zero, then, the velocity component normal to the interface in both the phases is equal to the velocity of the interface, and that is identical to the condition of velocity continuity used in the chapters on uid mechanics. The above condition indicates that the normal components of the velocities in the two phases, relative to the phase boundary are not continuous if there is a ; net ux of mass across the boundary. Suppose vaporization occurs at the interface. Then < is not zero. Then, normal components of the relative velocities will not be equal as the mass ux is continuous while the densities in the two phases are different. In the case of vaporization, the liquid density is thousand times more than that of vapor and hence the relative velocity in the vapor phase will be thousand times more than that in the liquid phase. The tangential velocity boundary conditions will be identical to those derived earlier in the chapter on momentum transfer, i.e., tangential velocities will be continuous. Now let us turn to the boundary conditions relevant to energy transfer. The boundary condition corresponding to the continuity of temperature is given by

 \  m 8
As interface has no capacity, the rate of accumulation of mass must be zero. Hence, P \  m 8 A P  \  m 8 A < ; This constitutes the velocity continuity boundary condition in a direction normal to the interwhile that reaching from phase  is equal to

 A 

We are giving the boundary condition for a moving boundary here. The boundary conditions for momentum transfer were however specied for stationary boundary as it requires more machinery to set them up. Interested reader can refer to the book on Momentum energy and mass transfer in continua by Slattery.


246

(25.1)

We had stress boundary conditions in momentum balance. Conditions analogous to this in heat transfer are boundary conditions pertaining to heat ux. We follow a procedure similar to the one followed with momentum transfer. Consider a little area
on the interface. The rate at which energy reaches the interface from phase is given by


\  P 3 M  m
$\  P M
Y

8 /

while that reaching from phase  is equal to

  m 8u

interface by convection. The left hand side is equal to the net heat ux arriving at the phase boundary by conduction and radiation plus the heat generated per unit area due to reactions. As there is no accumulation, it must equal the net energy ux leaving the interface by convection. Several simplied forms of the above condition are of interest. Suppose that there is no mass transfer across the boundary and that the radiative and reactive terms are zero. Then, the above boundary condition reduces to continuity of normal component of the conductive ux in the two phases:  

 (25.2)  where we have used the boundary condition on the normal component of the velocity. The ; is the net mass ux across the above boundary condition can be interpreted as follows. < interface. Hence the term on the right hand side is equal to the net energy ux leaving the

In above we have accounted for conduction and convection. Additionally there could be radiative heat ux, and there could be heat generated due ;  to reactions occurring on the surface, e.g. catalytic reactions. Let us denote the former by r . We will consider the reaction terms in detail later. For the moment let us denote the heat generated per unit area per unit time due to ;  E r . Since interface has no capacity, it can not accumulate energy. Hence the reactions by energy balance on a small area of interface must be given by

C C  m 8

;  ;  E M r M r A < ;

/ /

This is similar to the continuity of normal stresses in momentum transfer. It is this form that we will encounter most often.

C C m8

25.4 Development of temperature prole: Constant ux case


Let us return to our problem of heating water. Let us reexamine the temperature proles obtained earlier when the temperature of the surface of heater was assumed to remain constant. The temperature of the water increases with time. Hence we expect the temperature gradient to decrease with time. Now the heat ux boundary condition will state that the heat ux in the heater must equal to the heat ux into the water. The former is of course the heat being supplied by electrical heating. The heat ux into the water is given by

where is the coordinate normal to the heater surface. As heater is not moving, is zero. Hence the heat ux is simply given by the conductive ux into water. However, we are expecting that the temperature gradients will decrease with time. Thus the heat ux supplied to the heater to keep the temperature constant must decrease with time. 247

I j  M P % 

I[

Now let us examine the development of temperature proles if the heater is supplied with constant power. The heat ux boundary condition will now tell us that the temperature gradients at the wall must remain constant. The temperature of water will still increase with time due to heating. Hence the temperature of the phase boundary must increase in such a way that the gradient remains constant with time. This is shown in Figure 25.3 Clearly, if temperature
Note increase in surface temperature Increasing time Heater Surface x Fluid

Figure 25.3: Temperature development with constant wall ux of the surface increases beyond the melting point of the heating material, the heater will melt. This is how an electrical fuse works!

25.5 Onset of natural convection


We have assumed in the above discussion, as well as in the discussion on temperature development in constant wall temperature case that the uid will remain stagnant. Whether motion will begin or not depends upon the stability of the static condition of uid in the presence of temperature gradients. If a uid in a container is being heated from the bottom. As the temperature proles are established, the uid at bottom becomes hotter, and hence less dense, than the uid at the top. Such a conguration is unstable. As a result, the uid begins to move to attain a stable conguration, whereby the hotter uid rises and the colder uid descends. This is how natural convection starts. Description of the velocity proles that develop in natural convection is one of the interesting problems in heat transfer and we will discuss this in more detail later on.

25.6 Convective heat transfer


Convection in general increases the heat transfer. The extent of increase in heat transfer depends on many factors, and that can be found only by solving the energy equation. Here we want to qualitatively explain how heat transfer rate increases. Consider a simple case where plugs of uid move past a hot surface. In this way it is similar to the problem we just discussed. To make things easy to understand, let us also assume that the plugs do not exchange heat between each other by conduction. See Figure 25.4. Since they do not exchange heat with each other, the temperature proles in each slab will be identical to those obtained for a plug in contact with a hot surface when the contact time is equal to the coordinate of the plug divided by the velocity of the plug. Thus, they will look like those shown in Figure 25.4. If only conduction was taking place all of the plugs would have been in contact with the hot surface for the same contact time, and hence all of them would have had the same temperature gradients. However since the plugs are moving now, almost all plugs would be in contact with the hot surface for a shorter time, and hence have steeper temperature gradients in them. Thus 248

Fluid flows in the y direction

1 z

3 4 1 2

Temperature

Figure 25.4: Temperature development in plugs of uid moving past a hot surface

heat ux to the plugs would be more. Thus convection leads to a larger heat transfer rate as compared to conduction. Another way of thinking about this is as follows. Fluid next to a hot (or cold) wall is stagnant and it receives heat by conduction alone. Fluid away from the wall moves and hence its contact time with hot elements will be less than if it were stagnant. As a result it will be colder (or hotter) than when it is stagnant. This difference increases with distance from the stagnant wall. As a result, at any given cross section, the temperature gradients have become steeper, and hence, the heat transfer rate increases. The underlying feature is the movement of fresh uid into what might have been stagnant otherwise that accounts for enhanced heat transfer.

25.7 Solution of heat transfer problems


We have discussed the basic features of heat transfer. The next step for us would be to calculate the rate of heat transfer. The procedures are outlined here and examples of the solution are presented in the next chapter. The procedure for setting up heat transfer problems is similar to the ones followed for momentum balance. We learn the basic ideas through shell balances. Here a thermal energy balance is made on a shell, and then the constitutive equation is substituted into it. The resulting differential equation is solved along with the boundary conditions. After that we derive general differential equations that represent the thermal energy balance and which form the starting point to solve the problems of heat transfer. These will be similar to the Navier-Stokes equation. In problems involving convection, it is easily seen that we need velocity proles. Hence, in general, the Navier-Stokes equations have to be solved simultaneously with the equation of thermal energy. Often we can decouple them under fairly general simplifying assumptions. The case where this is not possible is of course where natural convection occurs. Thus discussion of natural convection forms a important special category.

249

Problems for Chapter 25.


25.1 Derive the jump conditions for the following problem at the liquid-solid interface. It would help if you rst sketch the possible temperature proles. (melting temperature) is brought into contact with a A liquid initially at  B  %   cold wall which is maintained at

25.2 If thermal conductivity is a very complicated but nonnegative function of temperature, is the prole shown in the gure possible to be reached in the following two cases? (a) Temperature depends only on . (b) Temperature prole shown is at some time .

T 1

T 2

Figure for problem 25.2.

25.3 The diagram shows the lines to which heat ux vector is tangential every where. The direction of heat ux is also shown. Plot the isotherms and indicate the direction in which the temperature decreases.
y

Figure for problem 25.3. 25.4 A rectangular block is shown with two of its parallel faces maintained at  g while the other two parallel faces are maintained at  K . Sketch (qualitatively) the isotherms at steady state.

2W

T 1

T 2 T 1 2L

Figure for problem 25.4.

250

Chapter 26 SHELL BALANCES IN HEAT TRANSFER


Examples of applying shell balance technique to solve a few problems in heat conduction and convection are considered. These are: Heat transfer in series of slabs Heat transfer in a heated wire Heat transfer in forced convection Heat transfer in a bearing This chapter is similar to the one on use of shell balances in solving uid ow problems. The rst two problems are familiar to all since they are part of the bread and butter of Unit Operations. We then move on to a problem in convection. This example is used to illustrate the concept of heat transfer coefcient, and has to be studied with care since it has many features that are often encountered in convective heat transfer. The last problem illustrates the role of viscous dissipation in determining the temperature distributions. In problems in uid mechanics, velocity prole was usually the rst to be calculated. In heat transfer, it is the temperature prole that has to be determined. The procedure to be followed is similar to what was adopted in solving problems of uid mechanics. Thus, a shell heat balance is made to develop a differential equation with temperature as the dependent variable. Then boundary conditions are specied as dictated by the problem statement. The differential equation is then solved with the boundary conditions to obtain the temperature prole. Problems in uid mechanics could be solved only in simple situations, i.e., where several simplications could be made. Similar is the case with heat transfer. However, considerable physical feel can be developed by solving such problems.

26.1 Heat transfer in a series of slabs


26.1.1 Statement of problem
Consider a stack of slabs of different materials. The surface at one end of the stack is being maintained at a constant temperature,  B while the surface at the other end of the stack is being maintained at a constant temperature  . Let us assume that the other surfaces of the stack are insulated. The geometry of the arrangement is shown in Figure 26.1. This kind of situation prevails when some equipment are insulated, e.g., furnaces, heat exchangers etc. The objective 251

in the present problem is to calculate the heat loss through the stack of slabs at steady state. As we shall shortly see, this requires calculation of the temperature prole.
E A B G H x
X1 x X2

Figure 26.1: Slabs in series. The faces ABDC, EFHG, ABEF, and CDGH are insulated. A small control volume of thickness is shown on the right side.

26.1.2 Simplications and assumptions


Cartesian coordinate system is selected since it is best suited to describe the geometry of slabs. The slabs are solid, and hence convection is absent1 . Thus only conductive heat transfer is important. All faces except those perpendicular to the axis are insulated. Further, the faces
 and ] ` are in a homogeneous condition. Thus, only the heat ux in the direction can be expected to be nonzero. All physical properties are assumed to be constant, in particular, that they do not change due to variation in temperature.

26.1.3 Heat balance over a control volume


We select a control volume of thickness and with area (of the plane) same as the slabs. Let the area be . The heat balance is written as (refer to chapter 23) Rate of genRate of Net rate Rate of ineration of generation of input of put of heat + heat due + + of heat due enthalpy into through to chemical to viscous the CV by surfaces reactions and dissipation convection other means As per the statement of the problem, steady state prevails. Thus, the accumulation term can be set to zero. As there is no ow, the viscous dissipation term is zero. As chemical reactions are not occurring in the system, heat generation due to these effects is also zero. The only control surfaces in the control volume through which heat ows are at and M . Thus, the heat balance yields  Rate of accumulation of = enthalpy in the CV

y A

1 If the layers are made of uid, and if the direction of the temperature gradient is such that natural convection is not generated, results of this section will be applicable there also.

252

Following a procedure identical to that followed in making shell momentum balance, the above equation is divided by and the limit of the resulting equation is evaluated as the differential quantity tend to zero. The following differential equation is then obtained:

-   A

This equation has a very simple interpretation: the heat ux remains constant. This is justied as heat ux has to remain constant in view of conservation of energy, constancy of the area through which heat ows, and since accumulation and generation are absent. Notice the similarity between this equation and the equation of continuity. The similarity is to be expected since both describe the behavior of a conserved quantity.

26.1.4 Combining rate law with heat balance


It is easily seen that the above equation can not be solved unless heat ux can be related to temperature. A similar situation was encountered in solving uid mechanics problems, where progress could not be made till stress tensor was related to velocity. The constitutive relationship, namely the Newton Stokes law, was used there to achieve the link. Here, similar role is played by the Fouriers law of heat conduction. After substituting the Fouriers law, the balance equation then becomes,  H 

j   V A 

As all physical properties were assumed to be we can the thermal conductivity can be removed from the above to obtain the nal form of the heat balance equation:

 K  K A

(26.1)

The above is valid in all the three slabs. Let us denote the temperatures in the three slabs by  , , and  . The scaling here is simple. A characteristic temperature difference,  B  , is available. Hence a non-dimensional temperature difference can be dened in each phase. For example,

26.1.5 Scaling

B   A v B 

A characteristic length is also available in each phase. A different characteristic length can be used in each phase. It is also possible to use the total length of slabs. As can easily be seen, the form of the differential equations would not change. Non-dimensional forms will not be pursued any further since it does not add any more to understanding the process.

26.1.6 Boundary conditions


Three second order differential equations have to be solved and it requires a total of six boundary conditions. The boundary conditions have to be specied at the interfaces of the slabs. 253

There are four interfaces, two at the ends of the stack and two in between the three materials.  A First consider the end at . It is given in the problem statement that the temperature is being maintained constant at  B : Consider next the interface between slab and . Let it be located at g . At this interface, neither the heat ux nor the temperature are specied. This is typical of situation at internal boundaries2 , and is reminiscent of two adjoining uid lms owing down a plate that was considered in chapter 12. There, both types of boundary conditions, namely continuity of velocity as well as the stress, were used. Thus it may be expected that both the boundary conditions pertaining to ux and temperature have to be used here. Now refer to eq.25.1. The boundary condition states that the temperature is continuous across the boundary. Thus,
%

 A B

A 

(26.2)

 A 

(26.3)

The other condition is an equation of heat balance given by eq.25.2. All the interfaces under consideration are stationary and hence \ = 0 for all boundaries. There is no heat generation due to reactions and radiation is absent. Further, velocity is zero since solid phases are being considered. Thus all terms in eq.25.2 are zero except the conduction terms. Thus the boundary condition implies that heat ux is continuous at this boundary:
-

 A  -

After substituting the constitutive relationship into the above, it yields

phase be located at Let the other boundary between ideas gives the following boundary conditions:

 j  A j and @

 

g
%

(26.4)

K . Application of similar
(26.5) (26.6)
%

A Finally, consider the end face of the stack. Let it be located at statement it is being maintained at a constant temperature of  . Hence  A  A
%

A   j  A

 j

 

% A K A % K

. As per problem (26.7)

These six equations constitute the required boundary conditions. The solution to the equations is well known and is given by

the ends of the stack of the slabs. The other boundaries are referred to as internal boundaries.

j % % 5 % g (26.9) K g K g % K j j > % % g M % % % % (26.10) K K K accessible boundaries. In this problem, these are A boundary condition can be imposed only on physically
%

B  A B    B  A  BB    A B  

6 j % g% g 6 j 6 % g M 6 j 6 % g M

  % g % g   % K %  % K

(26.8)

254

The heat ux is easily calculated:


-

where

are easily interpreted. Each slab The results are well known to all chemical engineers and offers a conduction resistance. The resistance is given by the ratio of thickness to the thermal conductivity. Since the slabs are in series, the total resistance of the slabs, , is the sum of all these. The heat ux, analogous to electric current, is given by the ratio of the driving force to the resistance. The driving force is the total temperature difference. By the same analogy, the total temperature difference is the sum of temperature drops across each slab. This feature is easily seen in the above equation. For example, the second % g  ^ equation shows that the temperature is given by the drop in the rst slab, j plus the drop corresponding to the thickness in the second slab. The temperature drop is proportional to the resistance of the slab. The resistance is inversely proportional to the conductivity and directly proportional to the thickness. Thus the temperature drop in a more conducting material will be smaller than in a less conducting material provided, if they both are of the same thickness. Thus, a small thickness of an insulating material is effective, and convenient to use, in preventing heat loss.

A %j g M

B   A
%

j% M K g

jT % K

26.1.7 Looking back


The solution shows that the temperature drops linearly into the slabs. This is a consequence of thermal conductivity being constant and heat ux remaining constant. The latter is a consequence of the fact that area through which heat ows remains constant. These are expected and of course agree with the formulae derived in Unit Operations books.

26.2 Temperature distribution in a heated wire


26.2.1 Problem statement
Consider a long wire through which current is passing. The wire is insulated. The surface of the insulation is being maintained at a constant temperature,  B . The geometry of the arrangement is shown in Figure 26.2. Obviously, this problem is only a prototype of many instances where heat is generated inside a body from where heat has to be removed, e.g., a nuclear reactor where an exothermic reaction takes place. In these problems, the temperature prole inside the solid is of interest, mainly with a view to know the maximum temperature reached inside it. It will be assumed that steady state has been reached.

26.2.2 Simplications and assumptions


Cylindrical coordinates are selected as they are ideally suited to describe the geometry of wire. As steady state is being considered, temperature is not a function of time. As the wire is very long, end effects can be neglected and this implies that the temperature prole is not a function of T or the axial coordinate. As the wire and the insulation are perfectly circular, symmetry 255

Wire Insulation R R

Figure 26.2: Insulated and heated wire. A control volume is shown on the left. around the T axis can be expected and hence temperature will not be a function of v . Thus,  A   p only.

26.2.3 Heat balance over a control volume


A control volume of thickness p and of unit length is selected for making heat balance. Since the temperature is only a function of radius, length does not play any role and in the end the heat loss per unit length is calculated. The heat balance is written as (refer to chapter 23) Rate of genRate of Net rate Rate of ineration of generation of input of put of heat + heat due + of heat due enthalpy into + through to chemical to viscous the CV by surfaces reactions and dissipation convection other means As steady state prevails, the accumulation term is set to zero. As there is no ow, the viscous dissipation term is zero. As no chemical reactions are occurring in the system, heat generation due to these is zero. Heat is generated only due to; the losses. Using the notation adopted . ohmic Let be the current earlier, let the heat generated per unit mass be passing through the wire. Let the electrical conductivity of the material of the wire be j . If is the area of cross material through which current is passing, the resistance of section and is the length of the  ^ the electrical path is given by j . Hence, Rate of accumulation of = enthalpy in the CV

P ; A

or

where is the current density, . As before, effects of radiation are absent and heat ux is only due to conduction. The control surfaces through which heat ows in the control volume are the cylindrical surfaces at 256

K Resistance K A j K P ; A j

and p

p . Thus, the heat balance yields ; L k~ p -,r  r L k~ p -,r  r#yr M L ~ kp p A  6   ; p -r A p p

Following a procedure identical to that followed in the previous section, the above equation is divided by p , and the limit of the resulting equation is evaluated as p tends to zero to obtain

Note that p-,r is proportional to the heat owing in the radial direction. Similarly p p is proportional to the volume of the differential element.Thus the above equation is easily interpreted: the heat owing out from the control volume is equal to heat generated in the control volume, both quantities being calculated on per unit volume basis ; . This equation is valid both in the is zero. Once again, notice the simwire and also in the insulating phase, except that there ilarity between the left hand side of this equation and the equation of continuity in cylindrical coordinates.

26.2.4 Combining rate law with heat balance


We substitute the Fouriers law of heat conduction into the heat balance. Then we have the following equations to solve, where subscript is used for the wire and for the insulation:

P ;

A  A

H  p  p V 6  H = = j p p p p V 6  j p  p

(26.11) (26.12)

where, as before, it was assumed that the thermal conductivities of both materials are independent of temperature.

26.2.5 Scaling

The equations have to be nondimensionalized. The length scale can selected to be . One in the wire and 6 )  in be could also select two length scales, ) the insulation phase. As discussed in the context of the previous problem, it does not change much. Thus is selected to be the length scale, and non-dimensional length is dened as 3

is the temperature at the periphery of the insulation and off hand one might think that it can be used to scale temperature. This however is unphysical. The entire mechanism of heat transfer depends upon temperature differences. Thus, the temperature differences have to be scaled. In the present problem, a characteristic temperature difference has to be found to scale  B the temperature difference   . In the present problem, the total heat to be lost through the insulation is known since it is equal to the total heat generated in the wire. Therefore, it seems
We are graduating to higher levels! Instead of starred English letters for non-dimensional quantities, we have now gone to Greek. The latest trend is to use the same symbols after giving the scales. It does save a lot of trouble to printers, but often creates confusion.
3

B

A p

257

appropriate to build up a characteristic temperature difference from this quantity. The characteristic quantity should be the largest possible, since only then the non-dimensional quantities would be of the order of unity. Such a characteristic temperature difference can be estimated as follows. A temperature difference is needed to lose all the heat being generated. The largest estimate of it would be that which is required to lose the heat through the material of the small  est conductivity. In the present problem it is the insulating material. Thus, if  is the characteristic temperature difference

  = Total heat to be dissipated per unit volume j K or ;   A P K j= The non-dimensional temperature differences can then be dened as =   B  =  =  B  j j A = A v P ; K and v P ; K P ; A

In terms of the non-dimensional quantities, the differential equations take the form:

9 9 9 9 9 9 9 9

6  H v =   V M jj A 6  H v   V A

c c

 
)

(26.13)

 6

26.2.6 Boundary conditions


To solve the above two second order differential equations, four boundary conditions are required. The boundary between the insulation and the wire is an internal boundary. There, drawing from the experience of the previous problem, both the temperature continuity condi tion and the heat balance will have to be used. The temperature of the boundary at p A is being maintained constant at temperature,  B , and that forms the boundary condition there.  A The line p forms another boundary and a similar situation was encountered in solving uid mechanics problems. At such a boundary, either the gradients are set to zero or it is demanded that the variable itself must be bounded. The boundary conditions are summarized below:

 p   j p

It may be mentioned that  must remain nite can be an alternative to the rst equation. In terms of non-dimensional variables

A = A Y A A

A A

 = = j =  p B  

at

p A at p A at p A at p

j )

v   v 

9 v =YAA 9

=  j= v= j
258

at

at at at

9 9

A  A ) A ) A 6

(26.14) (26.15) (26.16) (26.17)

The solution to the problem is given by

v =A
26.2.7 Temperature prole

K K = K  j ) ) j L X K ) L

9 9

(26.18) (26.19)

From the above equation, it can be seen that the maximum temperature is generated at the center of the wire and the temperature prole in the wire is parabolic. Equation 26.13 is very similar to the equation of motion for laminar ow in a pipe under applied pressure gradient. Refer to eq.14.1: it represents the balance between pressure forces and the viscous forces. The latter are due to diffusion of momentum. However, it was pointed out in discussing momentum balance that forces can also be thought of as generators of momentum. Thus, eq.26.13 which is a balance between heat generation and diffusion, is similar to eq.14.1. Therefore a parabolic temperature prole can be expected in this problem! It is always worth while to think in terms of analogy between heat transfer and uid mechanics because experience in one area can help in the other. Figure 26.3 shows the temperature prole. Notice the continuity of temperature

w i

Figure 26.3: Temperature prole in a heated insulated wire at the interface between wire and the insulation. Notice also that since the conductivity of the insulation is less than that of the wire, the temperature gradient at the interface in the insulation has to be more than that in the wire in order that heat ux remains constant.

26.2.8 Looking back

The maximum temperature is expected to increase with (current density), and decrease as the ratio of the thermal conductivity of wire to that of the insulation increases. As the radius of the wire increases, the heat generated increases as square of radius while the area available through which it can be lost increases only linearly as the radius. Further, the gradients tend to decrease with increasing radius. Thus, the maximum in temperature can be expected to increase with the radius. The maximum temperature is calculated from the solution to be

 B A

P ; K H j= ) K ) K j= j X L )V

It can be seen that all these expectations are borne out. 259

26.3 Heat transfer transfer coefcient


A detour is needed before shell balances in forced convection can be discussed. Two concepts, that of bulk temperature and heat transfer coefcient have to be introduced.

26.3.1 Bulk temperature


The rate at which enthalpy carried by a owing stream of uid across an area is given by

, - P

m 

The present notes deals with mostly uids, which can be treated as incompressible. Thus, the above is equal to

, - P %$ r  m 

where r is some reference temperature. It is often convenient to dene a mean or bulk temperature,  , of a stream such that, when multiplied by the mass ow rate and specic heat, it gives the total enthalpy being carried by the stream per unit time. Hence

 r  P % m 
P %$ r  m 

,-

,-

If properties do not change over the cross section, the above equation can be written as

, - m 
 A , - m 
' , P m 

(26.20)

The above equation is easily linked to measurements. If liquid coming out of cross section is 4 collected in a cup for some duration of time, , the amount of mass collected in it will be

If specic heat is assumed to be not a function of temperature, the heat capacity of the mass collected is

' , -

P % m 

If the mass collected is well stirred or mixed perfectly, it will attain a uniform temperature, % D  . Hence, the total enthalpy of the stream collected is

 D % r  P % m 

' , 260

Figuratively speaking of course, since we may need a big cup.

However this must also equal the enthalpy of the liquid collected:

! , -

P % r  m 

Thus, it is easily seen that the bulk temperature is identical to the uniform temperature reached by the stream if it were well stirred. Hence it is also referred to as cup mixing temperature by BSL.

26.3.2 Heat transfer coefcient


In case of uid mechanics, where possible, the velocity proles were obtained rst. They were used to calculate the drag forces etc. The results of those calculations were summarized as plots of drag coefcient ( ) or friction factor ( Z ). These plots contain all the information needed to calculate forces, rate of work done etc. They are sufcient if one is interested only in those quantities, which is usually the case in applications. Similar is the case in heat transfer, where usually the rate of heat transfer between two phases is the main quantity of interest. Thus, it appears that if a quantity similar to drag coefcient or friction factor is dened for heat transfer problems, it would be of great practical value. The corresponding factor is heat transfer coefcient. It is dened such that use of heat transfer coefcient, along with the driving forces, allows calculation of the heat ux into a stream of owing uid from an interface between two phases. In this sense, it is similar to friction factor or drag coefcient. The heat transfer coefcient is dened as follows. Let a normal be constructed from an interface pointing into the uid. Notice that the normal is located at an interface: uid-uid or solid-uid, usually the latter is the most commonly encountered. The heat transfer coefcient, h is dened such that the heat ux into the uid from the interface is given by

where  is the temperature 5 of the interface. This is of course the denition that is so widely used in Unit Operations. Notice however that we have precisely dened the temperature of the uid to be used: it is the bulk temperature. This denition makes sense since we are often interested in calculating changes in the total enthalpy of owing uid streams, which is related to the bulk temperature of the stream. The above denition of heat transfer coefcient allows us to link the rate of heat transfer to the bulk temperature. Notice that as soon as the heat transfer coefcient to a uid stream is known, the heat ux into that stream can be calculated. No more information is required. The main objective of solving convective heat transfer problems is to calculate the heat transfer coefcient. This in turn allows calculation of rate of heat transfer into owing uid streams. This is similar to the situation in uid mechanics. The idea of solving uid mechanics problems from a fundamental view point is to calculate the friction factors, and drag coefcients. If the friction factor or the drag coefcient is known, the drag forces, the power requirements etc can be calculated. In studying uid mechanics, it was found that the objective of calculation of friction factors is only partly fullled, and success has eluded when the ow becomes turbulent. Similar is the situation in heat transfer. When ow becomes turbulent, heat transfer coefcients also can not be calculated, but have to be measured. However, a lot of insight can be gained by learning how to calculate heat transfer coefcients where it can be done.
5

Cm

  

(26.21)

The subscript is being used since the most common interface is a solid wall.

261

26.4 Heat transfer in laminar forced convection


26.4.1 Problem statement
Consider a liquid owing through a tube due to a pressure gradient. Let the conditions be such that laminar ow prevails. The uid enters the tube at a uniform temperature of  = . The beginning portion of the tube is at the same temperature  = . But after some length, the tube is being heated and the arrangement is such that the heat ux to the wall, - , is constant. The overall arrangement is shown in Figure 26.4 The process has been in progress for long and

T i T i Upto z = 0, the tube wall is at the same temperature as the entering fluid

w Heating section begins at z = 0

Figure 26.4: Overall ow arrangement. The control volume shown in dotted line will be referred to later. hence steady state can be assumed to prevail. The objective is to calculate the temperature in the uid as a function of radius and axial position in the tube, and of course, the heat transfer coefcient.

26.4.2 Heat balance over a control volume


We start in this problem with a heat balance, instead of the usual statements of hypotheses, because it is the best way to show the kind of simplications that are made in convection problems. A part of the tube, as shown in Figure 26.5, is chosen to be the control volume. The heat balance over the control volume can be written as Rate of genRate of rate Rate of Net eration of Rate of ingeneration of input of accumuput of heat + heat due + of heat due lation of = enthalpy into + to chemical through to viscous enthalpy in the CV by reactions and surfaces dissipation convection the CV other means As this is steady state problem, the accumulation term can be set to zero. No chemical reactions are occurring in the system, and the generation of heat due to this need not be considered. The procedure that will be useful in estimating the magnitude of heat generated due to viscous dissipation is considered in the next section. But for the moment, it will be assumed that it is negligible. We introduce one hypothesis. In view of the uniformity of boundary conditions on ^ v = 0. the circumference of the tube, axisymmetry is assumed, i.e., p , shown as ABC, and the The control volume has two cylindrical surfaces. One is at p M other at p . It also has two annular circular faces. One is at T and is shown shaded in Figure 26.5. The other is directly below it, at TM T . The rate of input of enthalpy across any interface

4 4

262

r z

A B C

Figure 26.5: Control volume in pipe for heat balance of area


and in the direction of the normal n to it is given by

P m 

Thus the net rate of input of enthalpy into the control volume through all the control surfaces is equal to

L k~p I r P  r L k~p I r P  I  I  T T ryr M L k~p p P L k~p p P y

" %

" %

The heat ux through the surfaces can be by conduction and radiation. However, radiation is not being considering in the present notes. Thus, heat ux is given by , and the net rate of input of heat by conduction through the control surfaces is given by

equation will then be obtained:

A L k p T -r  r L ~ k p T -r  ryr M L ~ k p p-o  L ~ k p p-o  ly p These results can be substituted into the heat balance, the resulting equation divided by T, p and the limit of it equation as T tends to zero can be evaluated. The following differential p I r  M p p 6

, -Cm 

" %

" %

26.4.3 Coupling to uid mechanics

I *  A 6 p-,r  p p T

4 4

o T
-

The rst thing to be noticed is that the above equation can not be solved till the velocity proles are known. Thus, the equations of convective heat transfer are coupled to the equation of continuity and motion. The natural question to ask is whether, if at all, the equation of motion can be solved independent of the heat balance equation. The Navier Stokes equations were derived by assuming that physical properties, specically viscosity and density, remain constant. But these are affected by temperature. Thus, unless some simplications are made, the equation of motion and the thermal energy balance equation can not be decoupled. This is true even for forced convection problems as well. 263

In general, a simplication is made in case of liquids. Mass conservation implies that velocities should change as density varies. Thus, for ideal gases, the density varies signicantly with temperature, and these effects can be signicant. However, the variation in density with temperature is not so important for liquids. Thus, variations in velocity caused by the small changes in density are also small and hence are not expected to have important effect on the momentum balance. Hence, unless it is essential, density is assumed to be constant. One instance where such an approximation can not be made is natural convection. Here, if the density variation is neglected, the very force that drives motion, namely buoyancy, also gets thrown out! But density is assumed to remain constant in forced convection problems. One more simplication, though difcult to justify, is also made for the sake of obtaining simple solutions. Thus, it is assumed that viscosity also remains constant! Usually, this is in signicant error and needs to be corrected for practical applications. This is the origin of the wall correction applied in calculating heat transfer coefcients for designing heat exchangers. Thus, if both density and viscosity are assumed to be constant, the equations of continuity and motion are decoupled from the heat balance equation and can be solved independent of the latter. In the present notes also, it will be assumed that both density and viscosity remain constant in dealing with forced convection problems.

26.4.4 Simplications and assumptions


As stated just now, it will be assumed that the physical properties remain constant. This permits solution of the equation of motion independent of heat transfer equation. As uid ows through the tube, the velocity prole develops as a result of the imbalance between the pressure and viscous forces. If the length prior to where heating begins is long enough, then the ow can be expected to be fully developed before the uid reaches the heated section. Further, it is given that the ow is laminar. Hence, the velocity prole is given by

I A   r

I A L H 6 pK a KV

where a is the average velocity of the uid. It was briey mentioned in chapter 24 that heat transfer problems are classied using several criteria. One of them is about the state of ow. Thus the present problem comes under the category of fully developed ows. Link to thermodynamics In chapter 23 the issue of relating enthalpy to temperature and pressure was briey discussed. Thus, if it is assumed, as it will be here, that the state of the uid is not far from equilibrium, thermodynamic relationships can be applied. In the present problem, as mentioned earlier, incompressibility or constancy of density is assumed. Thus,

In the spirit of the spate of constancies being assumed, let also join the list. Substituting all these results and assumptions into the enthalpy balance gives

% A 

K H LP % a 6 p K V

It may be noted that the equation of continuity was used to effect algebraic simplication. 264

 A 6 p-r  p p T

, T
-

26.4.5 Combining rate law with balance


Fouriers law of heat conduction is then substituted into the balance equation to obtain,

H K LP % a 6 p K V

where, in the same spirit as the other ones, thermal conductivity was also assumed to remain constant.

K H  A j 6  K p j V M p p p T T

4 4

26.4.6 Scaling
The appropriate length scale in the dimensional radius:

direction is the radius of the pipe. Let

A p

be the non-

However, there are no obvious characteristic quantities for temperature and the length along the axial direction. As mentioned in the discussion of the previous problem, selection of  = as a temperature is unphysical, and a characteristic temperature difference is needed to scale  scale  =  . In the present problem, the heat ux is specied. Thus, in a manner similar to what was done in the problem dealing with heated wire, the scale for the temperature difference should be derived from the heat ux. The heat ux is supplied at the wall, where velocity is zero by no slip condition. Hence, right at the wall, the heat must be removed into the interior of the uid by conduction. By using Fouriers law of conduction as the basis for scaling, a good ^j. scale for temperature difference is obtained to be -

 =  j A v

A length scale in the axial direction is not there in the problem, but a point will be illustrated ^ by assuming that one is available. Let is be W . Let the non-dimensional length be T A T W . Substituting all these, the following non-dimensional equation is obtained:

as the velocity prole develops into parabolic shape, momentum diffuses from bulk toward the wall. There, in the beginning portions of entry zone, a boundary layer develops near the wall, and the velocity gradients are conned to the wall. However, when the thickness of the boundary layer increases along the length of the pipe, velocity gradients percolate all the way to the center. After the boundary layer extends up to the center of the tube, the velocity prole further changes till it achieves the fully developed parabolic shape. Similarly, in the case of heat transfer also, a thermal boundary layer forms on the wall, and in the rst parts of the entry zone, temperature gradients are conned to the wall region. Further downstream, the thermal boundary layer thickens, and the temperature gradients reach the center. The temperature prole further changes along the length. Now consider K^ a K location in the tube where the gradients W is expected to be small. Thus the second have already percolated into the center. There, term in the right hand side of the above equation can be neglected. It appears that by the same argument, the entire left hand side of the equation also should be put to zero! That is unrealistic 265

9 4 9 4 9 9 4 9 4 4 transfer.4 During ow in a pipe, We consider the analogy between momentum4 transfer4 and heat

%a W H K K v A 6 L P 6 j W V T

K K v V M K vK W T

since the term being neglected is the convection term, which is the heart of the present problem. It appears therefore that it should be kept, and for this to be logical, the following condition 6 must be valid: P % More insight can be gained by recognizing that a length scale for axial distance has to be built, much the same way a characteristic temperature difference was derived, when a length scale is not available. This is also reminiscent of the experience with similarity solutions. The central issue here is to nd when convective terms in the axial direction are comparable to conduction terms in the radial direction since only then they can be retained. One way is to compare the heat transferred by conduction and convection while following the uid element. Thus, the time scale for convection can be compared with the time scale of conduction. K P At% any ^ 7 T , the time scale of convection is T a . The conduction (or diffusion ) time scale is ^ j . Thus

a j

A P %T j K a

is a possible way to remove dimensions of axial length. Thus, as long as is not small, the convective heat transfer remains important. Yet another way to think about this is to construct a length scale for axial distance, the direction of convection, by combining velocity with (since we want the convective terms to be comparable to the diffusion terms) the radial diffusion time scale. With these variables the non-dimensional equation is given by

L 6 K  v A 6
The group

9 4 4

vV 9 9 49 4 9 4 %4 P a j

M P %j a

K K

vK

is called the Peclet number, . From our earlier efforts to nd the scale for axial length, the number is easily interpreted as the ratio of the conduction (or thermal diffusion) time scale to is large, the axial conduction can be the convection time scale for a length of . Thus, if neglected compared to the radial conduction term. Thus we nally have the equation we are looking for:

26.4.7

9 4 4 Boundary conditions 99 Y

L 6 K  v A 6

vV 9 9 49 4 9 4 4

(26.22)

order with The above is a partial differential equation, rst order with respect to and second respect to . Solution of the equation therefore needs one boundary condition in coordinate  A and two in coordinate. The boundary conditions at T and at the center of the tube are  A easily specied. These correspond to v and the gradient of temperature equal to zero,
6 7

If the ratio is very large, the convective terms dominate radial conduction as well, and our scaling is wrong. Remember that is the thermal diffusivity.

266

respectively: At At

9
A

c c

v A  v A 

49 4

(26.23) (26.24)

Now consider the boundary at the wall. Denote wall as phase I and the uid as phase II. Application of the temperature boundary condition suggests that the wall temperature and the uid temperature are equal. While this is correct, it does not help us since the wall temperature is not known. The other boundary condition concerns continuity of heat ux. Refer to eq.25.2. The wall-uid boundary is xed, and hence its velocity is zero. Take the normal to be qr . There is no mass ux across the interface, and and all other effects like radiation impinging, heat released due to surface reactions etc are absent. Thus, the boundary condition reduces to

However

mqyr

C C

 mqUr A 

is the heat supplied to the wall, - . The boundary condition then reduces to At or At

p A A

c 6c

4 94 4

j p A v A 6

(26.25)

26.4.8 Solution of the equation

9 3 After substituting the assumed solution and separating variables, we can obtain: b 6 6   A K  L 6 9  9  9 g9  3 9 6  6 6 6    A L 6 9 K  9  9 \9  9 As the left hand side of both equations is a function of only while the right hand side is a function of only 9 , both must be equal to a constant. Let the constant in the rst equation be , while that in the second is . Then
v A #  M b  M

The above partial differential equation can be solved by using separation of variables. The whole solution is not of interest to us at the present, and so we give a quick outline to justify the special solution which is the focus here. Let us assume

H K A L 6nV M g M K X From the second equation, However we do not expect the solution to blow up exponentially. Thus, ) has to be negative, K ) . Thus, the total solution can be written as a sum of which are conveniently written as  solutions:  A v M M )c
and

9 9

bA

 9

: 9 >9

267

H K L A v 6nV M g M K M X The boundary conditions have to be used to determine the various constants. The boundary condition at the center of the tube implies that g is zero. The boundary condition at the  L A A wall gives . A problem occurs in implementing the boundary condition at . It

Recall that the formulation of the problem focused at large lengths into the pipe. Under those conditions, the last term decays to zero leaving only the rst two terms.

9 9

 9

is easily found that the above solution can not satisfy the boundary condition at = 0. The above solution represents the form in the limit of large lengths, and it does make sense that the boundary condition is not satised by it. Thus in order to incorporate the boundary condition, an integral form of heat balance is used. The total heat supplied into the tube from the beginning up to some length, should be carried outward by the uid at that location. This is same as heat balance over the control volume shown by dotted line Figure 26.4. After noting that we have neglected axial conduction, the heat balance gives

Lk T

A La

, ,


K H P % 6 p K V   =  p  p  g v 6 pT  

to In non-dimensional form, this reduces A L

99
K. X

(26.26)

can be used to determine in view of the axisymmetry. This condition The nal solution is found to be

m ! This is exact in the limit . Surprisingly, as BSL note, it is accurate to 2% for Usually, limiting solutions are good, even when the limit is far away! Yet another reason to attempt to nd limiting solutions.
26.4.9 Looking back

K L A v M

9 9 X

L

(26.27)

 6

The above solution does make intuitive sense. The temperature of the uid is expected to rise linearly along the length of the tube as heat is being supplied at a constant rate. However, none could not have guessed that this is valid at all radial positions. This is a gain from the solution. The solution also implies that the wall temperature increases along the length of the tube. This also expected. As the uid gets heated, its temperature rises. Thus to supply heat at a constant rate, as dictated by the boundary condition, the wall temperature has also to rise.

26.4.10 Bulk temperature


The bulk temperature is easily calculated by using the denition:

  =A

La

, K , P % H 6 p K K V   =  p  p  ( K K H L a , , P % 6 p K V p  p  (
 

268

where  = has been used as the reference temperature. The denominator is equal to the mass ow rate multiplied by the specic heat. The numerator, according to the integral boundary condiL tion just derived, is equal to k T - . After substituting all these into the equation dening the bulk temperature, it is found that

As expected, it shows that the bulk temperature increases linearly with length of the tube.
26.4.11 Heat transfer coefcient

  =A L - j

It has been mentioned earlier that the results of the temperature proles are summarized in the form of heat transfer coefcient. The heat transfer coefcient is dened as
-

A  c T   T 
-

But from the temperature proles,

 c T   =A

v 9 j
A X 686 j ] j

Substituting this equation and the equation for the bulk temperature into the dening equation, it is found that ] where

686 A 6 A L M L X

is the diameter of the tube. The group

is the well known Nusselt number. This is interesting since the Nusselt number and hence the heat transfer coefcient is a constant and not a function of Reynolds number. This idea is discussed in more detail later on. It turns out that this is the property of fully developed temperature proles.

26.4.12 Conclusion
This example has to be studied carefully. Ideas regarding heat balances in the context of convection, the coupling between uid mechanics and heat balances, the approximations made to effect decoupling, and evaluation of heat transfer coefcient have been illustrated in this example. Heat transfer coefcient plays the same role as friction factor does in uid mechanics. It is sufcient to know only the heat transfer coefcient to tackle problems in the manner of Unit Operations. The example here illustrates that, in some cases at least, heat transfer coefcient itself can be calculated. Often temperature proles are also required. Unit Operations approach can not provide such an information. The example considered shows that with the fundamental approach, such an information can also be obtained.

269

26.5 Convective heat transfer to slabs


We have mentioned that if heat transfer coefcient is given, problems involving convective heat transfer can be calculated without needing any more information. Let us illustrate this by returning to the problem of heat transfer to slabs in series. Refer to Figure 26.1. Let a uid at a bulk temperature of  B ow at a fast rate parallel to the face
] . Similarly let another uid at a bulk temperature of  ow at a fast rate parallel to the face ` . The uids will gain or lose heat and hence their temperature should change as they ow past the surface. But if the ow rate is sufciently large, the temperature of the uid will not change. The earlier solution to this problem indicates that the heat ux through the slabs can be calculated one the surface temperatures are known. The rst thought that comes to our mind is that a shell balance in the uid has to be made and coupled to the solids to determine the required surface temperatures. However, the heat transfer coefcient neatly contains all the information needed for calculating the heat uxes, and that permits us to bypass the shell balance in the uid. Let us illustrate this, which of course is already well known to you!  are the temperature continuity and ux The the possible boundary conditions at A continuity. The former can not be implemented since only the bulk temperature of the uid is known. Thus, we can try the ux continuity and it reads

 A  B A   j  c

A  A
%

Similar boundary condition is written at the other face:

A   A j B  A B  B    A  B 

  c

g  % g %  % K

The problem can now be solved. The solution is well known, and is given by

j M j M

6 j % g% g 6 j 6 % g M

The expression for  remains unaltered. coefcient reects the additional resistance to heat transfer present in the two uids. The main point we wanted to illustrate is that all the vital information about uid and convection in it is contained in the heat transfer coefcient. Once it is known, the equations for the uid phase can be ignored as far as heat ux calculations are concerned.

j% M K g

% K j j Q % % % % K K M The ratios of thermal conductivity to heat transfer

26.6 Temperature proles in a bearing


Bearings oat on a uid lm and their carrying capacity depends upon viscosity, and the gap between the rotor and stator, refer to section 15.2. Usually gaps are small, the uids used are viscous, and the work done to over come the friction to keep the rotor moving is large. The work done is dissipated as heat in the uid, and as it is large, there could be a large variation in the temperature and hence viscosity. Thus the carrying capacity of the bearing could be effected by the temperature distribution. In the example to be considered, the density and viscosity are assumed to be constant and make an estimate of the maximum temperature in the bearing. It 270

gives an idea of the kind of effects one can expect, and whether it is necessary to account for viscous heat generation in various cases. A more exact solution of course is required where it is important to account for viscous dissipation as well as for property variation. A bearing is slightly eccentric as shown in section 15.2. But for the purposes of the present analysis, it is assumed that the bearing is concentric.

26.6.1 Problem statement


The gap in a concentric set of cylinders is lled with a uid. The gap is small compared to the radius of the cylinders. The outer cylinder is stationary while the inner one rotates at a constant speed. The ow can be assumed to be laminar. The outer cylinder is maintained at a constant temperature,  B . Let us also assume that the inner cylinder is being maintained at a constant temperature  g . The objective is to calculate the maximum temperature produced in the uid at steady state.

26.6.2 Simplications and approximations


Consider a bearing that is very long compared to its radius. This permits to neglect the small variation in temperature and velocities in the axial direction. Axisymmetry can be expected to prevail so that all I quantities are independent of the angle as well. The ow is given to be laminar and hence t is the only non-zero component of the velocity. Further,

I A I  t t p

and

 A  p 

In the spirit of obtaining the rst approximation, all physical properties are assumed to be independent of temperature. This decouples the uid mechanics problem from the heat transfer problem, and the velocity prole can be determined immediately. Velocity prole The velocity prole for this geometry can be obtained by solving the NSE. If the gap is small, the velocity prole is very nearly linear:

where is the rotational velocity of the inner cylinder. As the gap is small, the Cartesian coordinates can be used. Thus, the annular geometry is replaced by a parallel plate geometry. The radial direction can coincide with one of the coordinates, say . The angular direction coincides with another coordinate, say . The geometry and the control volume are shown in Figure 26.6. Thus,

I A p t )y 6 + )

the inner cylinder is now written as:

A p ) The angular rotational velocity becomes the component of the velocity and the velocity at I A )
at

A 

271

x+ x x y y x R

Figure 26.6: Coordinate system and geometry for the bearing and control volume.

This problem is now identical to Couette ow, and hence the velocity prole is given by

I0 A 6 ) 6  )

In cylindrical coordinates, axial symmetry implied that all variables like temperature, and velocity are independent of v , the angular coordinate. An important thing here is to translate this condition also into Cartesian coordinates. It is easy to see that this condition is equivalent to

26.6.3 Heat balance

A 

The control volume has four surfaces: two each perpendicular to the and axes. Convection is only in the direction. The net rate of input of enthalpy by convection through the surfaces perpendicular to the surfaces is given by

I I P P y

This can be written in terms of temperature as discussed in the previous section:

% I0 P %  I0 y P 

The heat ux through the surfaces can also be by conduction while that through the surface is only by conduction: Thus, The net rate of input of heat by conduction through the control surfaces is given by

-  -  y -  -  y M

As stated earlier, effects due to radiation are absent. There are no chemical reactions and so heat liberated due to that cause is zero. However, the viscous dissipation is not zero. The rate of viscous dissipation per unit volume is given by

R b

For this example, it is equal to

c   I
272

The uid is Newtonian. Thus, the above is equal to

K H  I K H ) _  V A 6  ) V

Thus the rate of viscous heat dissipation in the control volume is obtained by multiplying the above by . All these results are substituted into the heat balance, and the limit of the resulting equation, after dividing by is evaluated as tends to zero. The result is

K  A  - _ H ) 6 M ) V

In deriving the above, the condition equivalent to independence of angular coordinate has been used.

26.6.4 combining rate law with heat balance


The heat ux can be linked to Fouriers law of heat conduction and the following equation is obtained in terms of temperature

K  A j  K  K _ H ) 6 ) V M

where the thermal conductivity has also been assumed to be independent of temperature. The above equation is easily interpreted. The rst term represents the net heat ux out from a small volume and the second is the heat generated by viscous dissipation, both per unit volume basis. The two balance each other. This suggests that temperature prole will remain unaltered because generation is balancing the out ux. Thus, even in a rectangular channel, it is possible that the temperature prole will remain unaltered in the ow direction, after some distance downstream, if the heat generated can be conducted away through the walls.

26.6.5 Boundary conditions


The temperature is being maintained constant at the boundaries of both the cylinders. Thus,


26.6.6 Scaling

 A

B

 g

at at

A  A 6  )  

The coordinate can obviously be scaled  with ) . A characteristic temperature difference is available in this problem:  g  B . Thus the following non-dimensional variables can be dened B

A 6  hc ) +

 v A  g B

The non-dimensional differential equation is then given by

Kv  K M p A 

(26.28)

273

where p is known as the Brinkmann number and is given by

The factor

K _ K 6 )  K K H ) _ A )g B  A p j  g  B  6 + V ) j   K K H 6 + )  _ 6 ) V j )

can be thought of as the temperature difference that is required to lose the heat generated through the walls. Hence Brinkmann number is the ratio of the temperature difference required to remove the dissipation to that available. If Brinkmann number is small, viscous dissipation is small and the available temperature is enough to remove the heat generated. Thus, there will be little temperature rise in the bearing. Viewed differently, the heat ux created by the imposed temperature difference is very much larger than the heat generated by viscous dissipation. Under these conditions, the problem reduces to that of heat conduction in a slab and the temperature prole is linear. If the rate of viscous dissipation is large, the available driving force or temperature difference is not enough to remove the heat generated, and the uid temperature will have to increase. Thus, large viscous dissipation will increase the temperature beyond  g and  B . Thus, Brinkmann number is an estimate of the rise in uid temperature one can expect due to viscous heat dissipation.

26.6.7 Temperature proles


The solution of the above equation with the boundary conditions is given by

If p , the source term is zero and the quadratic part of the solution is insignicant and the linear part is the solution. In other words, the only heat ux is due to conduction. The maximum temperature is given by

Lp 6  M 6  v A

9 9

(26.29)

It occurs at

H 6 6 A ! p v pKV L X 6 6 A L p

26.6.8 Looking back


The effects due to viscous dissipation should increase with increasing viscosity and rotational speed, and decrease with increasing thermal conductivity and gap between the cylinders. The Brinkmann number has all these characteristics. The maximum in the temperature increases with increasing Brinkmann number conrming all our expectations. If viscous dissipation is 6 very large, p , the conduction term is insignicant. Then equal heat is lost through both the walls, and the maximum in the temperature occurs in the middle. This is conrmed by the above equation.

274

26.7 Fully developed temperature prole


In the previous section, we found that temperature itself does not change in the ow direction. Examination of the problem on convection reveals a more complex pattern. We found that

L = A   j

and that Combining the two, we nd that

K L A v M

 j 

9 9 X
K

9 9 X

L X

In other words, a temperature prole, does not change in the ow direction. This is a useful concept, reminiscent of the fully developed velocity prole. Consider heating of a uid owing in a duct. A non-dimensional temperature prole can be dened as follows:

where  is the temperature of the wall of the duct, which need not be a constant. This denition is like the non-dimensional velocity prole in a tube.  and  are similar to the average velocity, and velocity at the wall, respectively. Thus, we can dene the temperature to be fully
developed for a ow in T direction if We leave it to you to show that you get the same results for Nusselt number if you follow this approach in the convection problem. EXERCISE Solve the problem of viscous dissipation but in cylindrical coordinates without assuming that the gap is small. Further, implement the following change in boundary condition. Consider that the inner cylinder is insulated. Though not a normal practice, use the following nondimensional temperature:

 v A   

4T

v A 

B v A   B

This gives the temperature rise as a fraction of the outer temperature.

275

Problems for Chapter 26.


26.1 A liquid at its melting point  is brought into contact with a large mass of a highly conducting solid at  B  . a) Plot the expected temperature prole as a function of distance and time? b) What is theB jump boundary condition at the liquid-solid boundary? c) If   is small, the rate of freezing will be low. Under those conditions the temperature proles in the solid can be assumed to be linear. Obtain a simple expression for the rate of freezing.
Liquid far away at Solid at T = To x T=T m

Figure for problem 26.1.

materials A and B are in 26.2 Innitely long bars of rectangular cross section of two solid perfect contact with each other as shown in the gure. The thermal conductivities of the  two materials are constant and are denoted by j and j . The surfaces at A and A W are maintained at  g and  K . The surfaces at A and A are insulated. Under steady conditions: a) Is the temperature a function of only or of both and ? Does this depend upon whether the thickness of the solids A and B are the same or not? b) Simplify the heat conduction equation. c) Specify the boundary conditions.

W A

Figure for problem 26.2. 26.3 An innitely long bar shown in the gure has the following properties: = 3000 %   ^ ^ ^ < < j R , = 1 j j R , j = 300 . It absorbs all the radiative ux

276

falling on its surface at = 0.5 < (shown by curly arrows) at the rate of 500 < . Further, the surface at = 0.5 < is exposed to a owing liquid bulk temperature ^ < K at a of 50 C and with a heat transfer coefcient of = 50 . The face at = 0 is exposed a owing uid at 100 with a heat transfer coefcient of = 100  m 6L and ^ < K  to . The surfaces at 9 < are insulated. Under steady conditions, in which direction is the heat owing in the solid, M or ? Calculate the heat ux at the = 0 face.

^ K

0.5 m

0.5 m

Figure for problem 26.3.

26.4 e are interested in nding the steady temperature proles in a body of trapezoidal cross section shown in the gure. The body extends to innity in the T direction.  A The surfaces at and A W are maintained at constant temperatures  B and g  respectively. The other surfaces are insulated. i) Assume that temperature is not a function of . Choose an appropriate control volume and make a shell therml energy balance to derive the relevant differential equation for temperature. Specify the boundary conditions neede. Solve the equation and nd the temperature prole.

ii) Is it correct to assume that temperature is not a function of ? Specify the exact boundary condition on the insulated surface.
y Insulated 2Wo x Insulated L 2W 1

Figure for problem 26.4. 26.5 A very long nuclear fuel element is shaped in the form of a very thick walled tube. A uid at constant temperature  = is owing through the inner tubular space, and the heat transfer coefcient is = . Another uid at constant temperature  B is owing over the outer side of the fuel element, and the heat transfer coefcient is B . Heat A
^p is generated inside the fuel and it is given that - i) Make a shell balance to derive the thermal energy balance equation for nding the steady state temperature prole in the fuel element. The thermal conductivity of the fuel can be denoted by j . 277

ii) Specify the boundary conditions. iii) Find the temperature prole.
Fluid flowing at To Heat transfer coefficient = h o Ro Ri Fuel Fluid flowing at T Heat transfer coefficient = h i i

Figure for problem 26.5.

26.6 Heat Recovery from a Solar Pond: A solar pond is a large lake of water. The bottom of the lake is made of a black surface so as to absorb all the radiation that falls on it. Suns radiation falls on the surface of water, and passes through the body of water and is absorbed at the bottom. Very little of the Suns radiation is absorbed in the body of water itself. At the bottom of the lake and beneath the black surface, provision is made to remove heat. The sides of the pond are insulated and can be assumed to lose little heat. However, the pond loses heat from the top surface of water by natural convection. Suppose that the ^ K radiant heat ux from Sun reaching the surface of the pond is given by 800 < . Consider a pond 3 < in depth. a) The heat transfer coefcient K ^ for natural convection through the top surface of wa ^ < ter is equal to 10 c ) (a reasonable value for natural convection to gases). Assume that the water in the pond is stagnant. The thermal conductivity of water is  7 ^ ^ < equal to 0.6 7 c . If the bottom is being maintained at 90 C and the ambient temperature is 27 C how much of the radiant energy is being recovered? b) Is it reasonable to assume that the water in the pond will be stationary, and that there will be no convective heat transfer from the bottom surface and water? c) Suppose that the pond becomes unstable and the heat transfer coefcient from  ^ ^ < c (a reasonable value for turbulent natthe bottom surface to water is 60 ural convection heat transfer coefcient in liquids). For this condition, recalculate the temperature attained by the bottom surface and the heat recovered. d) What are the ways in which the pond can be stabilized? 26.7 a) Consider a circular rod being used as a n. The radius of the rod is while its length is W . The heat transfer coefcient between the n and the ambient is . Suppose the thermal conductivity of the rod is innitely large. Make suitable shell balance and derive an expression for the rate of loss of heat from the entire n. A n with innite thermal conductivity is referred to as an ideal n. b) Even when the n has nite thermal conductivity, the resistance to heat transfer in the radial direction is usually dominated by the lm. Thus, the radial gradients in the n are negligible. This is a one dimensional approximation. Make suitable shell balance and derive an expression for the rate of loss of heat from the entire n. Fin efciency is dened as the ratio of the rate of heat loss in the n to the rate of heat loss from an ideal n. Derive an expression for the n efciency. c) Suppose we use a cone of length W with the base radius being equal to in place of the rod. Derive an expression for the n efciency. This leads to Bessel equation, 278

Heat loss through convection Ta

Heat gain by radiation

Water Tb

Heat Removed from the bottom

Figure for problem 26.6.

and you might refer to Appl. Math. in Chem. Engg. by Mickley, Sherwood and Reid or any other book. d) Which amongst the two shapes is better per unit weight of the n?
T b T ,h a z T b T ,h a z

R L

L Conical fin

Cylindrical fin

Figure for problem 26.7.

26.8 A cylindrical n of diameter is made of two materials. The two pieces are of the same length. The thermal conductivities of the two materials are j g and j K . The n is attached to a hot surface maintained at a temperature of  B . The n loses heat to ambient at a temperature of  and the heat transfer coefcient is given by . Heat loss at circular surface at the end of the second piece is negligible. In the following use one dimensional approximation for the n. i) What are the simplied energy equations for the n? What are the boundary conditions? Solve these to derive an expression for heat loss through the n. ii) To obtain greater n efciency, which material should be put near the hot surface: one with greater conductivity or smaller conductivity? Give your reasons. 26.9 Sun is very hot because of thermonuclear reactions taking place inside. We wish to calculate the temperature at Suns surface by treating him as a sphere with heat generation inside. We will assume that Sun is at steady state. (This is not correct since he is going to die. But dont worry since he will live for another 5 billion years.) Assume that heat transfer occurs inside Sun is only by conduction. (Of course it is a bad assumption since there are other mechanisms of heat transfer.) The thermal 279

Temperature = T o

Ambient temperature = T a Heat transfer coefficient = h z L k k L

Negligible heat loss from this surface

Figure for problem 26.8. conductivity ( j ) however is a very strong function of temperature. This dependence can be approximated K K by treating j as a function of radial position (p ) inside the Sun: j A 6 6  m9 6 n ^ p  c ^ < 7  . Here is the radius of the Sun and is equal to < . 

The heat generation is also a function of temperature and K  approx m 6 can ; it also K ^ p be c ^< imated by treating it as a function of radial position: A (This is a cooked up value!). Sun loses heat from its surface by radiation to the outer space which can be assumed to be at 0 K.

a) Calculate the surface temperature of the Sun. b) Simplify the energy equation inside the Sun. What are the boundary conditions? c) Solvethe and nd the temperature at the center. The observed value L 6 K. equation is Compare the calculated value with this and give your comments. 26.10 A water vapor bubble is growing in an innitely large pool of super-heated water. and that of the super-heated liquid far away from Denote the boiling point by  the bubble by  . a) Show that I if the liquid can be assumed to be of constant density, the velocity in the liquid ( r ) created due to the growth of the bubble is given by

to be small and the processes can be analyzed by assuming pseudo steady state to prevail. For these conditions, simplify the relevant equations to nd the temperature prole in the liquid. What will be the temperature prole in the bubble? Specify the boundary conditions. c) Solve the equations to nd; the heat transfer coefcient. d) Derive an expression for in terms of  .
R r

KI K ; p r A ; is the rate of growth of the bubble. where is the radius of the bubble and b) For small values of super-heat  A  f  , the growth rate can be expected

Figure for 26.10.

280

26.11 Show that heat transfer coefcient is a constant for when temperature prole is fully developed. 26.12 Formulate the differential equation that describes fully developed temperature prole for constant wall temperature boundary condition. 26.13 The velocity prole for fully developed laminar ow of a power-law uid in a tube of circular cross section is given by

I  A 6 M  a p 6 M 

6 M 6 6 p

Derive the expression for Nusselt number for thermally fully developed conditions for constant wall heat ux boundary conditions. 26.14 Consider fully developed laminar ow of a Newtonian incompressible liquid in a tube of circular cross section. The wall temperature is maintained as a linearly increasing function of the axial distance. Consider the portion where thermally fully developed prole exists.

    a) Derive a relationship between and  T T b) Derive an expression for Nusselt number

26.15 A uid is owing between two parallel plates in the direction. The two plates are separated by a distance  in the direction. The velocity prole is fully developed I and is given by = a  .

i) The plate at A is receiving a constant heat ux - while the plate at A  is insulated. Calculate the Nusselt number if the temperature is fully developed.

ii) Determine the Nusselt number(s) if both walls are being supplied different but constant heat uxes? 26.16 A uid is owing in insulated cylindrical tube of radius . The ow is fully developed and; laminar. Heat is being generated in the tube at a constant rate per unit volume, . Neglect heat generation by viscous dissipation. i) Do you think that the temperature will be independent of p ? iii) Is it possible for the temperature prole to become fully developed? 26.17 Fermi is one of the great physicists of our times. He went to hospital where his friend was recovering from a heart attack. (Story taken from Physics Today, page 43, June 2002.) His friend complained that he was being given too little to eat, 1500 calories per day. Fermi asked his friend You are a great lover of detective stories. How long does it take for a corpse to cool to, say, a degree above room temperature? His friend replied About four hours. After some thought, Fermi concluded that his friend will not survive if he continued to get only that much food. Is he right? 281

Chapter 27 CONSERVATION EQUATIONS FOR NON-ISOTHERMAL SYSTEMS


So far we have dealt with nding temperature proles in non-isothermal single component systems by making shell balances. Based on the experience with uid mechanics where Navier Stokes equations were formulated, one will naturally expect that procedures of shell balances can be systematized, and partial differential equations applicable in general can be derived. That is the topic of this chapter. The Navier Stokes equations were derived by taking a parallelopiped and applying the transport theorem for a stationary control volume. To break the monotony, and to show a different way, a slightly different procedure is followed here. Needless to say, the same procedure can be followed to derive the Navier Stokes equations also. In the new procedure, balances are made on the system itself.

27.1 Balances on a system


Recall the transport theorem derived for a system:

The system moves through a uid, and as it moves its properties change. The rate of change of any property is given by the transport theorem. Properties change due to some external actions. Its momentum would change since forces are acting on it. Similarly, the sum of its kinetic and internal energies change because of work done on it and heat supplied to it.

5) c l  , 5 ) l  *) c l
( , A ,  6- ( c a  m 7 ( M .- " F&% 4 - " F&% " F&% 4 

(5.2)

27.2 Thermal energy balance on a system


The basic steps in deriving the thermal energy balance are the same as followed in chapter 23. There, the rst law of thermodyanmics, modied to take into account the kinetic energy, was stated as: The rate of change in the internal energy plus the kinetic energy of a closed system is equal to the rate at which heat is supplied to it minus the rate at which work done is by it on the surroundings.

282

,. - P ;  a , - mJC 
M , - m O " F&% " F&% " F&% , P ;  a

This was applied to a system to obtain:


2

 6 I K l  m 
 P P A m a    M L a M  F F

., - " &% t

 ., - " &%

(23.1)

After applying the transport theorem, eq. 23.2 was obtained:

, - m C 
M , - m O R  m 
" F&% " F&% " F&% P 6 I K  , K I   P P P , t A , M 6" F&% m a 6- " F&% 4  M L a M - " F&% m  M L
4 As the focus is on the thermal energy, the contributions from mechanical work and kinetic

,6 - P t m  a , - O m 
M , - m R3m 
M ., - O bm  a F&% F&% " F&% " F&% " " 6 P I K l 6 PIK ,  , A ,  - R b a  - L a M - m L
" F&% " F&%k4 " F&% 4 When the mechanical energy balance is subtracted from the equation representing the rst law
of thermodynamics, the equation for the thermal energy balance was obtained:

energy have to be removed from the above equation. This was done by deriving the mechanical energy balance equation, which is given by eq. 23.3:

, -

" F&%

P ;  a

, - mJC 
 , - O b m  a M  , - R b " F&% " F&% " F&%

a A  F

The internal energy is eliminated from the above equation in favour of enthalpy anticipating convenience it affords for incompressible uids. The equation in terms of enthalpy is given by:

, - P  l a M , - m P  
" &% 4 " F&% 4 (23.4)

,  - P ;  a , - mJC 
,  - ]] O  a M ,  - R b

a A 

27.2.1 Divergence theorem

, - P 4 4

 a P   m+ M

,-

(23.5)

, 6- P  a 4 4 Collecting all terms into one side, it can be rewritten as ,  - P ;  a ,  - bm C 


M ,  - ]] O  a M ,  - R b  a ,  - P l a 4 4
a A 

Now a new step is taken by converting the above equation into an integral over volume of the system. The integrals on the area of the system are converted into volume integrals by using the divergence theorem:

, .- P ;  a , - m C 
, .- ] ] O  a M , .- R b

 M  m P a

, .- $b

,  - m$b

P   a A

283

27.3 Application to arbitrary system


The above equation suggests that the integral of sum of certain terms, collectively referred to as the integrand, should equal zero. The system that was selected was arbitrary. Suppose that initially system
occupying a volume a was selected. The integral of the integrand over a must be equal to zero. Suppose we selected initially a system which was a part of system
. Call it system , and let its volume be a . System will be contained in system
. But the above equation says that the integral of the integrand over a should also be zero. In other words, the integral is zero over a part of the volume a . The argument can be generalized and it can be argued that the integral must be zero over any arbitrary part of a . In fact, a itself is arbitrary. The only way the integral over any arbitrary part of any system is that the integrand itself is equal to zero, excepting for a nite number of discontinuities in the integrand. However, discontinuities are not expected in a single phase system. Thus, the integrand must be equal to zero every where. In other words

/ /

/ / $b

P ; m M ] ] O M

b C

R b

The equation can be rearranged to put it in the form of a conservation equation like the Navier Stokes equations:

Left hand side of the above equation can be expressed as

 J b M mJb P V 4 4 But the last term is equal to zero by application of equation of continuity. The thermal energy 4 4 balance can be then written as ] ; P M P mJb A b mJC M ] O M R b M P 4 substantial derivative of enthalpy. Thus, the equation can be written The right hand side is the 4 in compact form as ] P ] ] A (27.1) b m C M ] O M R b M P ; P M P m M

4 P

]  P A m M ] O M M m +

Jb

b C

4 P

 m P  A  ; M P

R b

H P

The above equation is known as differential thermal energy balance.

27.4 Temperature form


The differential thermal energy balance equation is more useful in terms of temperature. This was discussed in chapter 23. There it was shown that

6 6 H %  A  M P M

; 4  M P 4 to the Cauchys equation of motion. At this stage, the above equation is similar
Substituting this relationship into the differential thermal balance

H P % ] ]  M ]] O

P A  V % m M
284

4 b JC R b

4 

 V %

 O

(23.6)

27.4.1 Constitutive equation


The equation is still not entirely in terms of temperature. The heat ux vector has to be replaced by a suitable constitutive equation, in this case, Fouriers law of heat conduction. Substituting this, the nal form is obtained:

properties are constant, excepting where it is important, e.g., natural convection. Under such restrictive assumptions, the following widely used form is obtained:

4    V % A bmjeb  M R b M P ; (27.2) 4 Navier Stokes equation. It is generally assumed that The above equation is similar to the
H P ] ]  M ]] O

P % ]]  A j K  M

R b

; M P

(27.3)

This equation is the analog of the Navier Stokes equation and is the one which will be used in this notes. It is easy to interpret the equation. The left hand side is the rate of increase in the enthalpy content per unit volume as observed when moving with the uid. Enthalpy increases due to net heat conducted into the volume, which is represented by the rst term on the right hand side. The last term is that due to any heat being generated per unit volume per unit time.

27.5 Boundary conditions


The boundary conditions have already been discussed earlier in chapter 25. Boundaries are at the surfaces between homogeneous phases. The balance equations are applicable in each phase and the boundary conditions are appicable at the boundary. These are the continuity of temperature and heat uxes. The should in general include any rate of ;  heat ux ;  continuity  r and r . Consider the boundary between two phases heat generation per unit area, and  and let be the normal pointing from pahse to phase  . As discussed earlier, the boundary conditions will be given by

 A  A < ;  m ;  r ;  r E M M

C C 8

e e / /

 

(25.1) (25.2)

27.6 Coordinate systems


The thermal energy balance equation and the boundary conditions are vector equations. Thus, they need to be decomposed into components for use. This has already been discussed in connection with the Navier Stokes equations. The components of the gradient and divergence operators have appeared in those equations, and so the same tables can be used.

27.7 Solution methodology


The thermal energy balance equation is a scalar equation. Thus it is simpler to solve than the equations of motion. However, it is coupled to the equations of motion since velocity 285

appears in the equation. Thus, the equations of motion and thermal energy balance have to be solved simultaneously. This aspect has already been discussed in the chapter26, and will be mentioned only briey here. Natural convection and cases where the physical properties are a sensitive function of temperature are obviously where simultaneous solution is a must. In other instances, it is customary to assume that the physical properties, in particular density and viscosity, remain constant. Then, the equations of motion are decoupled from the equation of thermal energy balance. This is the most common procedure 1 that is used. The solution method is same as adopted with the Navier Stokes equations. A suitable coordinate system is selected once the problem is specied. Depending upon the problem and the level of complexity of solution required, assumptions are made. The assumptions are translated into mathematical terms, in particular about the various derivatives that vanish. The resulting ordinary or partial differential equations are solved along with the boundary conditions.

It must be emphasized that all these assumptions are not needed when numerical solutions are sought. Such is the case with using CFD approaches to problem solving.

286

Appendix 27.A Equations of energy balance in different coordinate systems


As with the equation of motion, it may be necessary to solve the energy equation in different coordinate systems. Therefore it is necessary to have the energy balance equation in different coordinate systems. Here the equation is given in cylindrical and spherical coordiate systems.

Equation of thermal energy balance for constant thermal conductivity represents rate of generation of heat per unit volume. It could be due to chemical reactions, or other sources of energy. The viscous dissipation  term is written as g . It is given by

Rb

Rectangular coordinates

4 4 4 4 Fouriers law 4 4 4 4 Fouriers law 4 4 4 4 4 Fouriers law 4


I H P %  M A j

H K H K M P %  M I0  M I  M I  V A j T

A j  c - A j  c-o A j  T

4 4

K M

4 4

 K VSM ; M T

(A27.4)

Cylindrical coordinates

(A27.5)

I H H 6 K  P %  M I r  M t  M I  V A j 6 p p v p p p p VSM p K v K M T

r A j  p c -ot A
-

Spherical coordinates

4 4 4 4 6  j p v c o 4 4
-

A j  T

4 4

K M ; M T

(A27.6) (A27.7)

I I  t  r p M p v M p v  V H K H 6 6 6   K M ; M K K K p v K V M S V M p p p p v v v p v

 

4( 4  4 4 4 4
287

,r A j  p c -ot A
-

 4( 4 6  6  A j p v c j p  v ( 4 4 4 4
-

(A27.8)

(A27.9)

Chapter 28 Solution of thermal energy balance equation


Some examples of solution of the thermal energy balance equation are given in this chapter.

28.1 Old wine in new bottles


Formulation of problems solved by shell balances by the generalized procedure can be reconsidered.

28.1.1 Heat transfer through slabs in series


The slabs are rectangular in cross section. Hence, the Cartesian coordinate system is chosen. Refer to gure 26.1. The length and height of the slabs is very large compared to their thicknesses. In view of it, it is assumed that the conduction occurs in the direction of thickness or in the direction or that

 A  T ^ The heat transfer occurs at steady state, and hence 

= 0. Heat transfer occurs through 4 4 ; solids and so convection is absent or v is zero. Furhter heat generation is absent and hence 4 4 is equal to zero. After these assumption are incorporated, the thermal energy balance equation
simplies to which is identical to what was derived earlier. The boundary conditions are to be applied at the boundary of solid and surroundings, the interfaces between the solids, and nally between boundary of solid @ and surroundings. Temperature of the surface of solid at = 0 is being maintained at  B . Thus it forms a boundary condition. The normal vector to the interfaces between the solids is i. At the inner interfaces between the solids, the continuity of temperatures and component of the heat uxes applies. No heat is being generated at the ;   boundaries, and so r is equal to zero. Mass is not being exchanged across the boundaries ; < either, and hence is equal to zero. Thus, the boundary conditions are given by

 A

 K =   K A c A c c

 A c

and -

 A  c
288

at

A
%

These are identical to the ones derived earlier. The other boundaries are treated similarly to arrive at the same equations as before.

28.1.2 Temperature distribution in a heated wire


The problem is one of determining the temperature distribution in a electrically heated wire at steady state. Cylindrical coordinates are the correct choice. The heating is uniform and hence axisymmetry is expected. As the length of the wire is very large compared to the radius, the gradients in the axial direction are expected to be small. Thus,

4 4 volume, 4 and 4 is given by The heat generation occurs in the entire 4


 A v

 A T

 A 

P ; A

where is the current density, . Heat transfer occurs in the solid and hence convection is absent. The energy balance equation simplies to

K K   A 6 p -,r  M A j 6   p    M p p j p p p j

Heat generation is absent in the insulation and hence, in that phase, the energy balance equation simplies to 6  H 

 A j =  p  = p p p V

The boundary condition at the center is one of symmetry or that temperature must be nite. The normal vector to the interface between the insulation and the wire is qr . The temperature and the radial component of the heat uxes must be continuous at the interface between the wire and the insulation. Finally the surface of the insulation is kept at temperature  B . These give the same boundary conditions as before:

 p   j p

A = A Y 

A A

 = = j =  p B

c

or

is nite at

p A  at p A j at p A ) at p A

28.1.3 Heat transfer in laminar forced convection


In this problem, steady state heat transfer to a uid owing through a tube under laminar ow conditions is examined. coordinates are the correct choice. The ow is fully I Cylindrical  developed and hence only p is non-zero and the prole is parabolic. As the wall of the tube is being heated at a constant rate, the temperature proles are expected to be axisymmetric. Thus,

4 4 289 4
 A v

 A 

The following equation is obtained from the energy balance equation when the above simplications are used: K H K 6 H

boundary condition is then given by

4 4 4 4 which is identical to the one obtained earlier. 4 One 4 of the 4 boundary 4 conditions is heat ux continuity at the interface between the wall and the uid. The normal to the wall is qr . The
At

P % 6 p K V

 A j  V M j K p p p p T T j p A

4 Another boundary condition is the symmetry at the 4 center of the tube: c j  A  At p A 4p or niteness of temperature can also be used equivalently. These are identical to those used 4 earlier.
-

p A c

28.2 New wine in old bottles


Let us consider a few more complex problems. As you will notice some of them have great similarity to momentum transfer problems and the same signicance.

28.2.1 Viscous heating in Couette ow


Consider ow between a wide but narrow channel. We can approximate the geometry by ow between two parallel plates with the gap separating them, , being very small. Let the top plate move parallel to itself with a constant velocity while the bottom plate remains stationary. Let the two plates be maintained at a constant temperature of  B . We are interested in calculating the steady and fully developed temperature prole after assuming that physical properties remain constant. See gure 28.1.
V y

Figure 28.1: Viscous heating in Couette ow Velocity prole Temperature proles can not be fully developed unless the velocity proles are fully developed. Thus, we have I and linear:

I A I A  . Navier Stokes equations can be solved easily and the velocity prole is I A ^   a .

A 

290

Energy balance There is a characteristic temperature and hence we dene nondimensional temperature as

 B v A B

Fully develped temperature prole implies

 A  T Substituting these assumptions and results into the energy balance equation  A j   K  K M  . For the Couette ow, the rate of viscous dissipation is given by _ I8 ^  K . But, A Thus, the nal form of energy balance equation is given by K  A j  K  K _ H a V M

The temperature does not change in the ow direction. As the channel is very wide, we can assume that

4 4

 A 

Rb

4 4

Boundary conditions The above is a second order differential equation and its solution requires two boundary conditions. They are given by

 A B

at

A 

and at

Scaling

Characteristic length scale is . As discussed in the previous chapter,temperature differences are of interest and be nondimensionalized with  B . Thus,

  B A c  v A B

Nondimensional energy balance The energy balance and boundary conditions can now be written in terms of the nondimensional variables: K

 A  v K p M _ a K A p j B

where p is Brinkmann number and is given by

(28.1)

The boundary conditions are given by

v A c

at

A 
291

and at

A 6

28.2.2 Solution
The solution is given by

Lp v A

9 9 K

Similarity to momentum transfer It is good to always ask which problem in uid mechanics is similar to this? And vice versa! This kind of exercise helps build intuition and also creates an ability to translate experience in one transfer process into the other. Equation 28.1 is suggesting that the heat generated is being conducted away. This kind of balance, translated into momentum balance, should read that momentum generated is being diffused away. Forces generate momentum while viscous action diffuses the momentum away. Therefore, the problem of falling lm is identical to the present heat transfer problem, though laminar ow in a pipe is also similar except that it is in a different coordinate system. Now if we refer to the falliing lm in chapter 12, we see that the A equations are identical but that boundary conditions are different. The surface at was at zero shear stress or that velocity gradient was equal to zero. This would correspond to the top surface being insulated. However, as both plates are at the same temperature, the middle plane must be a plane of symmetry or the temperature gradient must be equal to zero on that plane. Thus, we can reduce the problem, in all respects, to the falling lm problem by using the middle plane as the top surface of the falling lm problem. This is accomplished by letting

Then,

which is identical to eq. 12.5.

K v L 

9 A L 9

6 6 9  K>

28.2.3 Heat trasnfer in a radial ow catalytic reactor


In a radial catalytic ow reactor, the catalyst is shaped like a ring. Reactants ow in the radial direction entering the inner side of the ring. Reactants and products ow out at the far side. As the ow is outward and radial, velocity decreases in the ow direction. Therefore, the contact time increases as the reactants ow through the reactor. This can compensate to some extent for the decreasing concentration of reactants and attain higher reaction rates. See gure 28.2.
Vo
Ro Re

Figure 28.2: Schematic of a radial ow reactor. Typically, plug ow can be assumed to occur in porous catalyst beds. The velocity prole is therefore easy to compute from equation of continuity and pressure drop can be computed by substituting this into the equation of motion. Equation of continuity can be used to show I that p r is a constant. Thus we have B

I A r a p
292

Let us assume that the walls of the reactor are insulated. This implies that the temperature is only a function of radius at steady state. Assuming that physical properties are constant, the energy balance equation simplies to

P % a B p

 6  H  ; A p j p p p p V M

where we have substituted the equation derived for the radial velocity prole in the energy balance equation. The rate at which heat is liberated depends upon the rate of reaction and the enthalpy change of the reaction. Thus, the energy balance has to be solved simultaneously with a species balance ; equation, which we have not yet derived. So, to make mattes simple, let us assume that remains constant. Boundary conditions Let us assume that we are maintaining the inlet to the reactor at a constant temperature of while the outlet is being maintained at a constant temperature of  . Scaling The radial coordinate can be scaled with the radius of the inlet. Let the temperature difference be scaled with the temperature difference between the outlet and the inlet. Thus,

 =

9
. Dimensionless equations

B A p B c  v A   B  

The energy balance and the boundary conditions are rewritten in terms of the nondimensional variables as follows.  6  H  where is the Peclet number. The nondimensional parameters appearing in the above equation are given by ; K

!  v A

vV  9 9 9 9 9 

M A

(28.2)

A a B A P j % c  c

B j   B 

It is easy to interpret these parameters. is the ratio of heat liberated to that which can be is the ratio of the conducted away given the temperature difference being maintained, and amounts of heat removed by convection to that removed by conduction. OThe nondimensional boundary conditions are given by

v A  0

A 6c  v A 6 

A B

293

Solution The solution to the energy balance equation is given by

v A
@

K  H 6 6 M L @ L 6 

66
@

9 KK

6uV

(28.3)

Figure 28.3 shows temperature proles for = 7.5 and for Pelcet numbers of 0.1 and 3. At low Peclet nubmer convection is ineffective and heat is lost by conduction to both ends. As heat is liberated is large for the value of selected, a maximum in temperature is created. However, as Peclet number is increased, heat is swept down and less heat is lost from the inner surface, and hence the temperature gradient is less at the inner surface. As now more heat is carried by convection, the maximum in the temperature is reduced.
1.6 1.4 1.2 1 0.8 0.6 0.4 0.2 0 Pe = 0.1
the1(x) the3(x)

Pe = 3

1.2

1.4

1.6

1.8

Figure 28.3: Temperature proles in radial ow reactor

Limiting cases There are a few limiting cases of the above solution that are of interest. 1. If is zero, the problem is referred to as transpiration cooling. Suppose that   B . The situation is similar to that of a refrigerator. Then heat would leak into the inside cylinder. Convection opposes the heat leak and so reduces its magnitude. This helps in keeping the temperature low and hence refrigeration. In an earthern pot, convection is created by evaportion of water through the pores of the pot, and that is how water becomes and maintained cool in hot summer. The ratio of heat leak with and without transpiration is an indicator of the effectiveness of transpirtion. The temperature proles ! are plotted for a few values of . Figure 28.4 shows temperature proles in the absence of heat generateiion. Notice how the temperature gradients become less steep as increases and hence heat leak is reduced. = 0, even though can still be large due to 2. If thermal conductivity is very large, large heat liberation. The problem reduces to that of a pure conduction problem and is identical to the problem of conduction in a heated annulus.

28.2.4 Leveque solution: convective heat transfer for short contact times
Let us consider heating of a cold uid when it enters a hot tube, and the ow is laminar. Let the process be at steady state. Here, three possibilities exist. One of them is that the ow as 294

1 0.8 0.6 0.4 0.2 0 1 1.2 1.4 1.6 1.8 2 Pe = 0.1 Pe 3

Figure 28.4: Temperature proles in transpiration cooling well as the temperature proles are fully developed. A similar problem, but with constant wall ux boundary condition already been discussed in chapter 26. The next level of complexity arises if only the velocity prole is fully developed. Physically this occurs if uid rst enters a long inlet section which is at the same temperature as that of uid before encountering the  A hot portion. See gure 28.5. Thus, ow is fully developed when it reaches T and it is at a temperature of  B . It remains at the temperature it entered with since the wall is also at the  B temperature og  .  However, the pipe wall is at a temperature  for T . The uid will then get heated for T assuming that   B . If one observes at locations where T is very large, the temperature proles will be found to be fully developed. In this section we are interested in the temperature proles for small values of T , i.e., in the inlet section. If we were to move with

1 0.8 0.6 0.4 0.2 0 1 1.2 1.4 1.6 1.8 2 Pe = 0.1 Pe 3

Figure 28.5: Schematic of fully developed ow entering a heated pipe the uid, we would observe that, in the inlet section, the uid has own past the hot wall only for a short time. In chemical reaction engineering, we would have stated that the residence time is small. Hence, the title contains the words short contact times. Velocity prole We assume that physical properties remain constant. Hence, the equation of motion is decoupled from the energy balance. Since the ow is fully developed, only axial component of the velocity vector is not zero and it is a function of only the radial coordinate, p . The velocity prole is parabolic: H K where

I A L a

6 p K V

is the radius of the tube. 295

Thermal energy balance Due to the cylindrical symmetry of the tube, temperature is not a function of the angle, v . The temperature has to be a function of radius, p , since the uid receives heat from the wall. It has to be a function of the axial distance, T , since it gets heated as it travels downstream. After substituting these assumed dependences into the energy equation, the following partial differential equation is obtained:

H K LP % a 6 p K V

4 4

K H  A j 6  K p j V M p p p T T

Boundary conditions

o 4  p A   p A  As second derivative of the temperature with respect to T 4 appears, we should require one more boundary condition. We will discuss this in greater detail a little later.
 pc 8 A  B c c T  A  c 
Scaling

Fluid is assumed to enter at a uniform temperature,  B . Further, the temperature prole must be symmetric with respect to p . Thus, the following boundary conditions can be written:

4 4

A characteristic temperature difference is available which can be used to nondiemnsionalize temperature: As we are interested in short distances from the entry point of the uid, we expect changes in temperature to be conned to small distances from the wall. Thus, the effects of curvature of the wall can be expected to be not important, i.e., we expect to approximate the important portion as a thin at layer of uid. We implement this idea after nondimensionalizing p with , and rewriting the energy equation:

B v A  B

9 4 T 9 4 9 9 4 9 4 4 wall 4 H 4 K Let us now dene a nondimensional distance 4 from the A 6 s 9 c 4 9 A 4 ce9 4 9 4 c 4 9 9 4 9 V A 4 K O 4 is given 4 as if4 we are in 4 Cartesian 4 4 coordinates. We see that the conduction term It makes sense
L `p P % jya K 6 K I A   I A Xa Substituting these results into the energy balance equation, K 6 Kv K % v v P A K K K X `p jya T M T

v A 6

K v V M 6 K K T

since the basic idea is that curvature effects are not important in the since the temperature gradients are conned to is a very thin zone. Using the same argument, the velocity prole can be approximated by

4 4

296

4 4

K K % v P ` p y j a T A vK X Still a length scale is not apparent in the T direction, and we can suspect that a similarity transformation may work. But to make it convenient, let us nondimensionalize the T coordinate

It is easy to judge the relative importance of the two conduction terms on the right hand side. As long as we are not interested in the same lengths in the and the T direction, i.e. T , the conduction term in the T direction is negligible compared to that in the direction. This  approximation breaks down near the edge around T , but will be valid farther down. The equation can be recast after accepting this approximation to obtain

4 4

by a length scale obtained in terms of conduction time scale and velocity:

A T K a X v A

The energy balance equation reduces to

Similarity variable

vK

Discovering the form of a similarity variable is by trial and error. Let us try the following form

where c < and are constants. As discussed in chapter 14, a similarity transformation works if the partial differential equation is reduced to an ordinary differential equation in terms of only the similarity variable, and the boundary conditions can also be stated in terms of only the similarity variable. Substituting the assumed form of the similarity variable, and a bit of manipulation, the energy balance is converted to
v  K Kv  v 6 N A < < <    M K ^ 6^ It is easy to see that, for the similarity transformation to work, we need < A 6 equation will also simplify, if we choose <A . Hence g

A`

A`

` A`

. The

" %

The energy balance now simplies to

` K A `v A A `K v K ` A`

The equation will simplify to the following convenient form

Kv Kv A   K M 

A`

297

if we choose

where is the thermal diffusivity. This is equivalent to dening the similarity variable as

H 6

"g %

As required, the partial differential equation has reduced to an ordinary differential equation in terms of only the similarity variable. Notice that the differential equation is a second order and hence would need only two boundary conditions. Hence we still need to examine the boundary conditions. The boundary conditions are given by

with small penetration distances or that the range of last boundary condition is modied as

o 4  A   A 4 Keeping in mind that we are dealing It is easy to see that the three boundary would not reduce.
  A  B c   c  A  c   c[ T
of interest is much smaller than , the

cT A  B

In terms of the nondimensional temperature, the boundary conditions are given by


 [     A v c cv c  A 6 c  v

c  A   A 

These can be rewritten in terms of the similarity variable as

*`

 A  cv 8 A 6 c  v

o *`
  A 

The last and the rst boundary conditions have now collapsed into one and thus we can conclude that the similarity transformation will work. Now the energy balance and boundary conditions can be written in terms of the similarity variable as

Kv  K M [ A 6 cfv v

A`

` K A `v A *`

(28.4) (28.5)

Solution The solution to the above equation is found to be

v A

, ,

 
298

l  H XV

(28.6)

Heat ux The major quantity of interest is the heat ux into the uid. As velocity is zero at the wall, the heat ux is only by conduction, and it is in the p direction. Hence,

g  H 6 6 6 -r p A  - A j  A j   B  V  v and the heat ux can be evaluated to be The derivative can be evaluated from eq. 28.6 6 H 6 g 6  A B j   V H [ X V

" %

A`

" %

leading edge. It is because, the temperAs can be seen, the heat ux is innitely large at the ature gradient in the direction suffers a jump at A F . The uid gets heated as it ows downstream, and hence the temperature gradient should decrease. Consequently, the heat ux also decreases as increases.

Heat transfer coefcient As discussed in chapter 26, the heat transfer coefcient is sufcient to calculate changes in the average enthalpy of a stream due to heating or cooling. Thus, it alone is sufcient for the practically useful engineering calculations. One therefore summarizes the detailed calculations on temperature proles by calculating the heat transfer coefcient. The heat transfer coefcient is dened as 

K/ where n is the unit normal and the temperature difference is taken in its direction. Further, 
m A  
is the bulk temperature. For the present problem, then
-

JC

 A B T ~   T  where B is the heat transfer coefcient at a given value of T . As the uid extends to ininity, the value of  will be the same as the entry temperature or  B . Thus, g 6 B A 6 H 6 H j V [ XV or g H H 6 L _ L g H ] g L B a L B A (28.7) p T V _ T V A V j ] and p are diameter of the tube, Nusselt, Reynolds, and Prandtl numbers where, , ,

" %

" %

respectively. Some times, results are also reported as the average Nusselt number which is dened as 6

 " %

" %

j T  A T

 B l

which can be easily calculated from the above. 299

Graetz-Nusselt-Leveque solution It is possible to solve the energy balance equation after neglecting the axial conduction term by separating variables. Such a solution is referred to as Graetz-Nusselt-Leveque solution. The similarity solution and the fully developed temperature prole for large T form the asymptotes of the more complete solution. The latter gives that Nusselt number attains a constant value of 3.66. A sketch of the entire solution is shown in gure28.6.

log

hD k

Slope =

1/3

Aymptote, Nu = 3.66

log

1 z Re Pr D

Figure 28.6: Schematic of Nusselt number for fully developed laminar ow entering a heated pipe

Consider a semi-innite solid slab initially at  = . Suppose that its surface temperature is raised to  B and kept constant at that value. The temperature in the slab will now change with time and the change will be different at different locations. We want to calculate the spatial and temporal variation of the temperature. Energy balance Let the slab be semi-innite in the direction and as it is innite in extent in the other two   , , and T . directions, i.e., the slab occupies the domain, In view of the innite extent of the body in the and T directions, we do not expect temperature l to depend upon those coordinates. Thus, we expect  c . As the body is a solid, convection is absent. Let us assume that physical properties remain constant. The energy equation simplies to K

28.2.5 Heat trasnfer to a semi-innite slab

Intial and boundary conditions

4 4

 A K

The initial and boundary conditionsare given by

  A  = c   c l A  B c  c  c8

The temperature can be nondimensionalized with the temperature difference:

c l

A =

= v A  B =  The initial and boundary are now given by v c [ A  cv  c l A 6 c  c v

c l

A 

300

Similarity to momentum transfer This problem is identical to the problem of a plate dragged suddenly with a constant velocity in a Newtonian uid. The energy equation is identical to eq. 14.3. The initial and boundary conditions are identical to eqs. 14.4, 14.5 and 14.6. Thus the solution must also be identical 1 . The problem admits a similarity solution and is given by

v A 6
where erf is the error function dened as

` A K X

s

All one had to do was to replace by ! Thus, while momentum diffusion is characterized by , heat condution (or diffusion of heat) is governed by . The temperature prole is given by gure 14.1.

s A L k

, :  T

Penetration times and depths One can therefore two quantities: penetration time and penetration distance 2 . Pene% , is dene the time required for tration time, % the effects of heat conduction to be felt at a specied distance. Similarly penetration depth, is the depth up to which the effects of heat conduction are felt by a specied time. We of course have to specify what we mean by the term: the effects are felt. It is arbitrary and we can say that the effects are not felt when the nondimensional temperature at location and time specied is less than 5% of that imposed by the  m  the 9 . The absolute number we calculate will therefore be arbitrary but the boundary, i.e., v> numbers will allow us to compare different situations. Thus, effects are deemed to be felt when   vi m 9 . In our problem, the value of xed once the value of v is xed and we xed v = 0.95. Hence K

l%

Thermal diffusivity is the diffusion coefcient for heat. Thus, penetration depth should increase, and penetration time decrease with an increase in thermal diffusivity. These are conrmed by the above. Suppose we had a slab of thickness W instead of the semi-ininite slab. We can then guess that the effect K ^ of changed boundary temperature will not be felt up to the thickness W for a time and for times much smaller than this, the solution we derived can be used, even less than W though the slab is not semi-innite in extent. Thus we see that ideas of penetration time and depth have enormous value in thinking about limiting solutions.
1 2

and

c%

To quote Feynman, the same equations have the same solutions! In the modern paralance, these are referred to as diffusion time scale and diffusion length scale

301

Heat ux The heat ux is by conduction and the heat to be supplied to keep the surface at temperature is of interest. Thus,

A 

constant

  c l A j 

Heat ux is seen to be similar to the drag force. It is easily calculated to be

 B =   l  j  A c k

As expected heat ux increases with thermal conductivity. As time progresses, temperature of the solid increases and hence the temperature gradients decrease. Consequently, the heat ux decreases with time.

and Consider another problem in heat conduction. Suppose a slab bound by the planes A  = . Let the surface A% is initially at  = . Now let the surface at A be maintained at  A be exposed to a uid at a bulk temperature  owing past it. The conditions are at such that the heat transfer coefcient is given by . See gure 28.7. Once again, as with the

28.2.6 Heat conduction in an innite slab

Infinite slab Surface maintained at T o

Flowing fluid at Tb

Figure 28.7: Innite slab being heated by convection. previous problem, the temperature will change with time and position and we are interested in calculating this dependence. Energy balance

As the slab extends to innity in the and T ldirections, temperature is not expected to depend upon these coordinates. Thus,  A  c . Convection is absent as the slab is solid and therefore the energy equation simplies to

4 4

K  A K

302

A  are given by  c [ A  = c   c l A  = We need one more boundary condition at A W . There are two possible boundary conditions at A W . One is temperature continuity. However, only the bulk temperature of the uid, and not its temperature at A W , is known. Thus, we have to apply the other possible boundary
The initial and boundary condition at

Intial and boundary conditions

condition, namely the energy balance at the interface:

C C  m 8
-

;  ;  E M r M r A < ;

/ / :

The heat generation terms are zero and there is no mass ux across the interface. taken to be the unit normal in the direction. Then

 

(25.2) can be

A W  A

A  W

But heat is lost by convection into the uid, and hence


-

A  A l  W  W3c 

Note that we have used the temperature continuity in writing the above expression. Substituting Fouriers law for heat ux in the solid, we get
 j A  W3c l  

Scaling

W can be used as the length scale. A time scale can be dened from the thermal diffusivity and the length scale and can be used to nondimensionalize time. The characteristic temperature   = can be used to nondimensionalize temperature difference,

A c W

= A W K cv A   =  

The energy balance, the initial condition and the boundary conditions can be written in terms of the nondimensional variables as K

v A vK (28.8) c 4 49 The boundary conditions can be rewritten 4 as 4 v 9yc 8 A  cv  c c  A  c! 9v g A v 6 c c  6  c A j W (28.9) 4 4 In the above is the Biot number which is the ratio of resistance to conduction in the slab to
g Y

the resistance due to convection.

303

Solution The solution to the above equation is obtained by separation of variables. Let us assume that v A   . Substituting this into eq. 28.8, and separating variables

9+ c

where can be a constant or constants3 . The rst equation can be solved to get

 +c

K A  K A

The temperature does not diverge with time, and hence we know . The solution corre A sponding to is the steady state solution. Let it be denoted by v . The entire solution can then be written as4 (28.10) v A vD M v

I/ satises the steady state heat conduction K equation,  9 vIK / A  The boundary conditions for vH/ are the same at steady state and hence remain the same as specied in 28.9. Solution is easily found and is given by vI/ A !M 6 9
v vD A v KD

I/

 

Now let us turn our attention to v D . By substituting eq. 28.10 into eq. 28.8, it can be shown that v D satises K

c 49 4 The boundary and initial conditions are given 4 by 4 D v D 9yc 8 A hvI/ 9  cv D  c c  A  c v 9 g A !v D 6 c c  4 c  94  Now we separate variables and write v D A . Repeating the procedure adopted earlier, we get  K K A c 9K A
# *)

where we have used the idea that the separation constants must be negative. We have the following solutions K    In view of the boundary condtion at 6 gives at A
)
3 4

v DA

*)

c  g  9 A
,

,8

A i ) c c  )

M K ,8 A

must be equal to zero. The boundary condition

9
)

Does the word eigenvalues ring a bell? The method is reminiscent of the way we solved the unsteady ow of a Newtonian uid in a pipe. It is being done here in a slightly different way to illustrate the idea that zero eigne value corresponds to steady state.

304

There are several values of ) that can satisfy the above equation. Now, the initial condition that has to be satised is Clearly the above condition can not be satisied with any single value of ) . Looking at the above expression, we see that we need to express the function on the right hand side in terms of the eigen functions that have arisen out of solution by separation of variables. SturmLiouville theory guarantees that this is possible, and tells that the eigen functions are complete and any arbitrary function can be expressed in terms of the eigen functions. Thus, we write

g )  A iv  A M 6

 9

I/ 9

 A

? ) ?  ? g

 9

The eigen values are obtained by solving the transcendental equation

?  ) ?A )

The roots are shown graphically in gure 28.8 There are innite number of roots and can be

Figure 28.8: Roots of the equation

)

?A ) ? . The curves are plots of tan of angle.

found. The constants can be found using the orthogonal property of sines from the initial condition

which gives

M 6 A ? g g ?A M 6 g

? ) ? 
)

 9
)

,  ,  K

 ?l  ?l *) ?K  ) ? 

The entire solution can now be written as

=   = A 6 M ? ? g   M

c  9

Initially, the second term on the right hand side is equal to the rst term, since that is what we used to determine the constants, and hence  A  = . The solution shows that as time increases, the second term on the right hand side decreases and K goes to zero. The solution ^ then corresponds to that of steady state. is proportional to W . Thus, for a given thickness

305

, the exponential terms decrease more rapidly for materials with greater thermal diffusivity. This makes sense since greater thermal conductivity and or smaller heat capacities enable the material to respond faster to the changes in the environment. Similarly, for a given , steady state is reached faster as W decreases. This also makes sense since resistance to conduction has reduced and the slab should adjust more quickly to environmental changes. These trends are as expected. Figure 28.9 shows qualitatively the temperature proles as a function of time.
1 Bi Bi+1 Increasing time

Figure 28.9: Qualitative trends in the temperature proles in a slab exposed to hot uid. constant. Limiting cases

is

If , resistance due to convection is negligible. Here, the surface temperature should equal  . This is inagreement with the gure. Steady temperature proles in a rectangular solid

306

Problems for Chapter 28.


28.1 A semi-innite slab initially at  = is suddenly exposed to a uid at transfer coefcient is while the thermal conductivity is j .

1 . The heat

i) Sketch how temperature proles will change with time and distance. Show the effect of small and large in a qualitative but comparative manner. ii) Formulate the unsteady state conduction equation and specify the initial and boundary conditions. iii) We dene nondimensional temperature

 v A   = 
and a new dependent variable

1 1 4

(28.11)

We now look for a similarity solution for

iv) Derive an expression for v .

a) Show that satises the same partial differential equation as  , i.e., the unsteady state heat conduction equation. b) Show that partial differential equation for does give rise to a similarity solution with is a similarity variable. What are the initial and boundary conditions for ? c) Solve the equation to nd .

( 4 in terms of the similarity variable ` A K (28.13) X (

A v j v

(28.12)

Liquid far away at T=T i T f h x

Figure for problem 28.1. 28.2 Consider a semi-innite body occupying the positive half of the axis. It is initially at a temperature of  B . For 0, the face at = 0 is supplied with a periodically varying heat ux given by - , :~ . Determine the steady periodic temperature prole.

307

Heat Flux

Figure for problem 28.2. and A . The solid is initially 28.3 Consider a solid bounded by two planes at A at a uniform temperature of  B .  At , the surface at A is maintained at  A cos while the surface at is maintained at  B itself. (a) Find the steady periodic temperature reached in the body. (b) Which problem of uid mechanics is this analogous to?

Tscos t x y T x=1

x=0

Figure for problem 28.3. 28.4 Consider a solid bounded by two planes at A and A . The solid is initially  at a uniform temperature  B . At , heat is generated uniformly inside the ; of  A A A
cos . Both the surfaces at body at a rate given by and at are maintained at  B itself. (a) Find the steady periodic temperature reached in the body. (b) Which problem of uid mechanics is this analogous to?

28.5 It is known that near the surface of the earth. the temperature (averaged over long periods of time, e.g., a year) increases with depth. It is referred to as geothermal 7 ^ gradient and is measured (in the recent times) to be 0.04 < . We wish to estimate the age of earth from this measurement. Suppose that, at its time of birth, earth was a solid mass at its melting point: 3500 . Since its birth, it has been losing heat to the outer space. This cooling process can be approximated by assuming that the surface of the earth is at a constant temperature 7 of 0 . It is also known that, even now, the depth of penetration is very small compared to the diameter of the earth. Estimate the age of earth given that the 308

T x
Q

x=1 y
v = A cos t

x=0 T

Figure for problem 28.4.

thermal diffusivity of earth is 10 Fourier, yes, the same fellow.)



< K ^ : . (If you dont like this problem, curse Mr.

28.6 Consider an innitely long slab of solid 1 and thickness W in perfect contact with a semi-innite solid 2. Both solids are initially at  = . At A , the temperature of B the free surface of solid 1 is raised to  . Simplify the thermal energy balance for this situation. Specify the initial and boundary conditions. solve the problem using Laplace transformation.

2 x

L
Figure for problem 28.6. 28.7 Two semi-innite bodies are initially at  and r . Their thermal diffusivities are and r respectively. Their thermal conductivities are j and j8r respectively. They are brought into contact with each other at = 0. The temperature proles in both bodies are expected to obey similarity solution i.e.,

 A
M

where
and

are constants.

K l O A X

` A`

i) Find the temperature proles in both the solids as a function of time. Some may need the Leibnitz formula given at the bottom of page 68 of BSL. 309

ii) Sketch the temperature proles for = 0,

A g and K where g K .

iii) What is the value of the interface temperature? iv) When you touch a metal or wood, both at the same temperature but lower than the body temperature, your hand instantaneously feels colder with metal rather than wood. Can you explain why?
Semininfinite body initially at T l x<0 x Semiinfinite body initially at T r x>0

Figure for problem 28.7.

28.8 Find the steady state two dimensional temperature distribution in an innitely long are exposed to a owing uid at rectangular solid. The surfaces at A a  A temperature of  and the heat transfer coefcient is . The surface at is being maintained at  B while the surface at A W is being maintained at  g .
y B x -B L

Figure for problem 28.8. 28.9 Consider an innite slab of thickness W . It is initially at temperature  B . From 0, the face at = 0 is being supplied with a constant heat ux - B while the other face is being maintained at  B . We are interested in nding the unsteady temperature l  prole  c

i) Find the temperature prole at steady state,  . ii) What is the energy equation for unsteady state? What are the boundary conditions? iii) Are the boundary conditions in form required by Sturm-Liouville theory? iv) Dene v A   . Substitute this form in the energy equation and derive the equation for v . What are the boundary conditions for v ? Are the boundary conditions in form required by Sturm-Liouville theory? v) Find the unsteady temperature prole. 310

K/

Constant heat flux being supplied to this face for t > 0

Initial Temperature is T o x L
Figure for problem 28.9.

This face is being maitained at To for all time

28.10 Find the steady state temperature prole in the solids shown in the gures. a) An innitely long rectangular block of dimensions 2 by 2 W . The faces whose lengths are 2 W are insulated. One of the faces measuring 2 is maintained at non-dimensional temperature of 0 while the other has a temperature prole varying linearly from 0 to 1 with length along the side. b) An innitely long cylindrical arc. The non-dimensional temperature on both the edges of the arc are 0.5, and one of the circular faces is being maintained at 0 while the other is being maintained at 1.

D y 2L x

C
0.5

1.0

1.0

2M

0.5

Figure for problem 28.10.

28.11 Find the steady state two dimensional temperature distribution in an innitely long rectangular solid. The surfaces at A a are exposed to a owing uid at  A temperature of  and the heat transfer coefcient is . The surface at is B A g being maintained at  while the surface at W is being maintained at  .
y B x -B L

Figure for problem 28.11.

311

28.12 The one dimensional approximation for n has been considered in many books. It can be compared with the exact solution to get a feel for how the approximation works. a) Write down the two dimensional heat conduction equation for the n shown in the diagram. b) Specify the boundary conditions. c) Solve the equation by using the separation of variables technique. d) Does the two dimensional K solution reduce to the one dimensional approximation when j and dimensional solution.

uW j

6 ?
h, T a

The latter is usually used in obtaining one

y x

2 W

This face is insulated

Tb

L a Figure for problem 28.12. h, T

28.13 Two parallel faces of an innitley long rectangular bar are maintained at  B while the other two perpendicular faces are mainatined at  g and  K . See the attached gure. i) Derive an expression for the steady temperature proles. ii) Draw a few heat ux lines. Draw on a separate gure isotherms corresponding to  B ,  g , and  K given  g  B  K . 28.14 Consider fully developed laminar ow of a Newtonian incompressible liquid in an annulus. Derive expressions for Nusselt numbers for the inner wall and outer wall for constant wall temperature boundary condition in the entry zone, i.e., the temperature prole is developing. 28.15 For fully developed temperature prole conditions, derive expressions for Nusselt number for constant wall ux and wall temperature boundary conditions for laminar ow in ducts of isoscles triangular cross section of side . For a Cartesian coordinate system set up with origin at the centroid of the triangle, the T component velocity prole is given by

6  I A L_  T

6 K K 6 K LK  L  L M L S

Only to be set up. it can not be solved in closed form. 312

These faces matained at T

This face is maintained at T 1 2L

y x 2W

This face is maintained at T2


Figure for problem 28.13.

28.16 An ideal gas bubble is placed in an innite extent of a Newtonian liquid. The initial radius of the bubble is solubility of the gas in the liquid is negligible. The B , and is in equilibrium with the liquid surrounding it. Initially, the pressure in the liquid is and the temperature of the liquid is  . The surface tension of the O liquid is . a) Determine the initial pressure and temperature  in the bubble.  O b) For , heat is generated ; in the bubble at a uniform rate. The rate of genera tion in the entire bubble is . Will the bubble expand or contract? c) Assume that the pressure and the temperature of the bubble to be uniform. Simplify the necessary balance equations and boundary conditions to determine the radius of the bubble as a function of time. State any assumptions that you make.

28.17 Nylon bers are made by extruding a molten lament of radius B at a velocity B The lament is wound at a higher speed, and hence the radius of the lament, a ,. reduces. The T component velocity as a function of the distance from the point where nylon comes out, T , can be found by solving equation of motion and is given by I where and W are constants related to operating conditions. The radius at the corresponding point can therefore be found by mass conservation

I K  A B BK T a  be  B . As it is being wound in ambient Let the temperature of the lament at T A conditions, the lament cools. It is desired to calculate the temperature as a function of the distance from the point it comes out into the room. The following data are _ ] = 100, = 10_ m, L = 3 m, j of air = 0.025 w/(m C), j given: B = 100 m, K of molten nylon = 0.3 w/(m C), thermal diffusivity of nylon = 10 m /s, h between K
air and nylon = 4 w/(m C).

A a B]

" %

313

i) What are the nonzero velocities? Simplify the thermal energy balance equation neglecting viscous dissipation as well as axial conduction, and specify the boundary conditions needed to solve it. ii) We wish to calculate the way the temperature of the bre falls as a function of T in a simple way. Assume that the variation of temperature in the radial direction in the ber is negligible. Simplify the above equation to derive a differential equation for the temperature of the ber as a function of T . DO NOT SOLVE IT. iii) Are we justied in assuming that there is no variation of temperature in the radial direction in the ber?
L
Ro r R This portion is shown expanded on the left

z Air

Nylon fibre

Figure for problem 28.17 28.18 Two large plates (extending to innity in the and directions) are separated by distance W , and the gap is lled with a uid (with properties j c P and % . p 6 ). The two plates are being maintained at constant temperatures, the left one at  and the right one at  . Assume that natural convection is not present in answering the following. Steady state is allowed to reach in the system.

1 1

a. What is the temperature prole at steady state? b.

P and % ) at a uniform temA solid sphere of radius (with properties j c  ) in the middle of the perature of  B is placed and held stationary (at T A gap between the plates. It is given that W . Use spherical coordinates to do the following: (i) simplify the relevant energy equations needed to determine the temperature proles in the uid and the solid, (ii) specify the boundary conditions needed to solve the resulting equations. 5 marks.

/ /

2 marks.

c. Suppose that the solid sphere (instead of being held stationary) is moved with a constant velocity a in the T direction (i.e., moves with a velocity a in ^ h 6 . As pQ 6 , velocities in the negative T direction) such that a the uid can be described by the steady Stokes proles in a frame moving with the sphere. They are given by

} h

I A r a

H 6 L M L V , 8 v p p

314

and

I A t a

6 H 6 M p M V v X X p

Use spherical coordinates moving with the sphere to do the following:(i) simplify the relevant energy equations needed to determine the temperature proles in the uid and the solid, (ii) specify the boundary conditions needed to solve the resulting equations. 5 marks.
Plate at T L r z L 2 L 2 Plate at T R

Sphere

Figure for problem 28.18 approaches a long cylinder with a uniform velocity a . The surface 28.19 Fluid at  of the cylinder is being maintained at a constant temperature,  . Reynolds number of the ow is small and the velocity proles near the cylinder are given by

 I A a o r L v I A a  t L

k 

H 6  p L 6 M pKV H v 6 L p 6  M L K V p

(28.14) (28.15)

28.20 A uid enters an annulus at temperature  B . The radius of the inner wall is = while that of the outer wall is B . The ow is laminar, fully developed and steady. A constant pressure gradient is applied in the axial direction. Only the inner cylinder is rotating with a constant speed 0 . The uid is being heated since the inner wall is   B maintained at constant temperature  for T and  for T . The outer wall is insulated. (i) Specify a convenient coordinate system that you will use to solve the problem, (ii) specify the nonzero velocities, (iii) specify the coordinates on which temperature of uid depends. (iv) simplify the thermal energy balance equation, (v) specify the boundary conditions needed for solving the energy balance equation.

We are interested in short contact time solution for heat transfer to the cylinder. Simplify the convection equations for this case and specify the boundary conditions.

28.21 a. A steam bubble is growing in a large mass of superheated water at constant pressure. Specify all the boundary conditions that may be useful. b. Suppose the bubble is growing in an environment where forces due to gravity are zero. Simplify the convection equation.

315

28.22 An ideal gas bubble is placed in an innite extent of a Newtonian liquid. The initial radius of the bubble is solubility of the gas in the liquid is negligible. The B , and is in equilibrium with the liquid surrounding it. Initially, the pressure in the liquid is and the temperature of the liquid is  . The surface tension of the liquid is O . and temperature  in the bubble. a) Determine the initial pressure  O b) For , heat is generated ; in the bubble at a uniform rate. The rate of genera tion in the entire bubble is . Will the bubble expand or contract? c) Assume that the pressure and the temperature of the bubble to be uniform. Simplify the necessary balance equations and boundary conditions to determine the radius of the bubble as a function of time. State any assumptions that you make. DO NOT SOLVE the equations.

28.23 The annular space in an innitely long concentric cylindrical annulus is lled with a liquid. The outer cylinder is rotating at a constant velocity, 0 while the inner cylinder is stationary. The outer wall is being maintained at a constant temperature of  B while the inner wall is stationary. Heat ; is being generated in the annular space at a constant rate per unit volume, . Neglect heat generation by viscous dissipation. The velocity prole is given by:

Simplify the thermal energy balance for this situation. Specify the boundary conditions. Determine the temperature prole.
Inner cylinder is stationary Inner cylinder is insulated

h B I p B B h Bc A  0 h B where B is the radius of the outer wall while is the radius of the inner wall.

I A  I A r c t

Outer cylinder rotates at constant speed Outer cylinder is maintained at constant temperature.

Figure for problem 28.23

316

Chapter 29 THERMAL BOOUNDARY LAYER

317

Problems for Chapter 29.

29.1 in an innite extent of uid owing past the plate with a uniform velocity a parallel to the plate. The plate is supplied with constant wall heat ux - . Find Nusselt number using integral methods and assuming Pr 1. 29.2 Consider heat transfer to a boundary layer on a horizontal at plate. Here the wall is equal temperature is maintained constant at  for B . The wall temperature   B to that of the incoming uid away from the plate,  , between . Derive an B expression for local Nusselt number for by using the integral balance method. For simplicity assume the velocity proles and temperature proles are I0 linear in the boundary layer:

I A  A "   " where and are the thicknesses of the momentum and thermal boundary layers 6. respectively.Assume that p

8 ,

V 8 x T 8

o T w

Figure for 29.2. 29.3 Consider fully developed laminar couette ow between two parallel plates separated by distance  in the direction, and the top plate moving with a velocity in  .a For the  direction. The uid and both the plates are at temperature  B for , the bottom plate is maintained at a constant temperature  while the top plate is maintained at  B . (See gure). We want to calculate the heat transfer coef ) where cient (between the bottom plate and the uid) for short distances ( temperature prole is developing by using integral methods for steady state. (i) Assume a suitable linear prole for temperature in the uid. (ii) Obtain the integral energy balance equation. (iii) Determine the local heat transfer coefcient. (iv) This problem could have been solved using the short contact time approximation as was done in the class. This method gives the following solution:

this compare with that predicted by the integral method? 318

  B A g T l T B   O u g  and g and K are constants and is the thermal diffusivity where A K dependence of heat transfer coefcient on predicted by of the uid. How does the

o V

To

y x

Tb

w Figure for problem 29.3.

29.4 Are integral method of solution valid for steady turbulent ows?

319

Chapter 30 NATURAL CONVECTION

The Boussinesq approximation, where density variations are accounted for in dealing with forces but not in continuity equation, is discuss A fully developed temperature and velocity profile problem, where the buoyancy forces exactly match viscous forces is examined. Boundary layer problems, where both temperature and velocity profiles are then considered. Heat transfer by natural convection is more complex to analyze than forced convection heat transfer. The reason is that a velocity scale is not available a priori. The motion is generated due to the buoyancy forces and hence the temperature distribution plays an important role in determining the motion. But we know motion determines the temperature prole! Thus the equation of motion and thermal energy balance are coupled and have to be solved simultaneously.

30.1 Boussinesq approximation


Buoyancy forces are generated due to density variations produced by temperature variation. There is one more effect caused by density variation: the velocities must change to conserve mass ow rates. The Boussinesq approximation suggests that the changes in velocities due to density variation are negligible. Consider ow in a pipe as an example. Suppose the average axial velocity is a . Let the temperature change be  when the uid ows over some length W . Let the reference temperature be  B , and the density at that temperature be P B . The corresponding in density while the uid has owed over the length W would be proportional  ^change  P B B  . Hence the velocity would have to change by a factor of 6  ^  B to keep to mass ow rate the same. The change in velocity is reected in an additional acceleration of uid as it ows through the length W , and hence it affects the motion causing inertial forces. K by  P ^ ^ B B An estimate of these additional inertial forces is given by a K   W . This term will P ^ W as be negligible compared to the magnitude of inertial forces B a long as the temperature rise is not large. Similar is the argument about the additional viscous forces generated due to changes in velocities due to expansion of uid in response to temperature variations. The essence of Boussinesq approximation is then to ignore the effect of density variations in the equation of continuity but keep it in the equation of motion as far as buoyancy forces are

320

concerned. Thus we solve the following equations:

where is recognized to be a function of temperature. It is obvious that if it is not done so in the equation of motion, there is no buoyancy force and then occurrence of motion can not be predicted!

b m A  P B ] ] A b M _ B b O ] P B % ]  A j B b

K K 

M P

30.2 Fully developed temperature and velocity prole


As before we can classify convection problems according to the status of developedness of proles. First we note that since temperature and velocity proles are linked, there is no possibility of velocity prole being fully developed while the temperature prole is developing. The present section thus examines a case where both velocity and temperature proles are fully developed. Consider a tall and wide channel with the walls being separated by a gap  . See Figure 30.1. The left wall is hotter and is being maintained at  g while the colder right wall is kept at

T1

y x b

T2

gravity

Figure 30.1: Natural convection in a vertical channel.

channel is sufciently tall, then the buoyancy forces will exactly match the viscous forces and the ow will be fully developed. This of course will not be true near the end and exit of the channel.

 K . Due to buoyancy, the uid will rise at the left wall and descend near the right wall. If the

30.2.1 Equation of continuity


If the ow is fully developed we expect the component of the velocity to not change in the direction. As the width of the channel i.e., in a direction perpendicular to the plane of the paper is very large, we do not expect either a velocity to exist in the T direction or the other

321

dynamical quantities to depend upon the T coordinate. The equation of continuity then gives, P after using the idea that is treated as constant in this equation,

Since the walls are impervious, hence = 0. The component equation of motion simply gives us the result that pressure can not depend upon .

I

I0  A

30.2.2 Equation of motion


Now we turn to the component of the equation of motion. At steady state, it simplies to

where R is the acceleration due to gravity, and we have used the fact that R A R . Here of course we will have to treat the density as a function of temperature. Let us take the mean of the wall temperatures as the reference temperature,  B . Then the density can be written as

KI0  A  _ PR K O M

 P A P B 6 M P 6  P  B  B 

as long as the temperature difference between the walls is not too large.

30.2.3 Coefcient of expansion


The change of density with temperature is a physical property. It is the tendency of the uid to expand when heated. This property is described by the coefcient of expansion, :

Hence where

is evaluated at  B . Substituting this in the equation of motion we get

  B  PB A 6

A 6  P P 

 KI  A   _ O M  K P B RM P B R   B

(30.1)

30.2.4 Thermal energy balance


Now we turn to the equation of thermal energy balance. We can simplify it to get

I0  A H K  K M

4 4

K V

(30.2)

322

30.2.5 Fully developed temperature prole


Let us now examine the consequences of temperature prole being fully developed. For this purpose, a non-dimensional temperature can be dened using one of the wall temperatures and bulk temperature:

Since the temperature prole is fully developed,

g   d g A v  g

vg   A A

A non-dimensional temperature could have been dened using one of the wall temperatures and bulk temperature:

K v K A   K

   S g v 

The fully developed condition would have given the following condition:

vK   A A

This implies that

v gdA v K c

Since the right hand side of the second equality is only a function of , this implies that    ^ g K    is not a function of x. The only possibility is that the temperature differences, and hence, temperature is linear in . If temperature is a linear function of , the heat supplied at one wall is equal to the heat removed at the other wall. Hence the temperature itself can not change in the direction. Thus eq.30.2 simplies to

g A  g    or  K   K

   S K v 

Hence the temperature prole is linear in :

 K  K A

  A  g  g  K  LM  B   B A L  

We can rewrite the above in terms of the mean of the temperatures of the walls, 

as

where

 A  g  K .

30.2.6 Equation of motion


Let us examine the equation of motion in light of these ndings. We notice that all terms   ^ only be a function involving velocity and temperature are only functions of . Hence O ofcan of or be a constant. But we deduced from the component equation motion that pressure   ^ B can only be a constant. Now since  is the mean of the is not a function of . Hence

323

 B must be two wall temperatures and temperature prole is linear in , the term B R positive for negative values of and negative for positive values of . Thus the terms involving velocity must change sign accordingly. Thus we infer that the viscous forces are balanced by the buoyancy forces, and the average weight of the uid is balanced by pressure gradient. Hence 
and

 KI _  K M P B R   B  A 

 O A P B R

30.2.7 Scaling and velocity prole


Let us try to get a scale for velocity. The buoyancy forces per unit must balance the P B R volume  . Hence an estimate of viscous forces per unit volume. An estimate of the former is K ^_ P B   . We can nondimensionalize the length with  . the characteristic velocity, a is R The temperature prole obtained earlier can be substituted into the resulting equation. The equation of motion becomes:  KI 6 Integrating this with the no slip boundary conditions, we get

K L A 

I A 6 P B R  K  6L _

From the procedure we followed, the ratio B R   a represents the ratio of buoyancy to viscous forces. If we multiply this ration with a Reynolds number formulated with the characteristic velocity, we get the following ratio:

30.2.8 Grashoff number

^ _

 p R  K 

It is referred to as referred to as Grashoff number. Thus Grashoff number represents the product of the ratio of buoyancy to viscous forces with the ratio of inertial to viscous forces. Thus the velocity prole can be written as

 I P B A  p _

30.2.9 Looking back


We were looking for a fully developed velocity prole, and by denition, it can not change in the ow direction. The uid must rise near the hot plate and descend near the cold one. Thus, invariance with ow direction implies that these two ows must match each other. In view of only one length scale and independence of properties on temperature, the upward and downward ow have to be mirror images. Thus, the velocity prole has a minimum and a maximum. The buoyancy forces increase with density as well as with the sensitivity of density variation with temperature. Thus, the convection velocity must increase with both 324

these quantities. Larger imposed temperature differences increase the buoyancy forces and hence convection. Increased viscosities must have exactly the opposite effect. All these are as conrmed by the expression obtained.

30.2.10 Transitions
We have assumed that the ow is laminar. One can expect a laminar to turbulent ow transition to occur if inertial forces dominate viscous forces or if Reynolds number is large. As we discussed Grashoff number is the Reynolds number for natural convection. Thus, ow transition  p exceeds a critical value. As we shall see in a later section, can be expected to occur when a more insightful number is Rayleigh number, . It is equal to the product of Grashoff and Prandtl numbers. We need to pay attention to the heat transfer from the wall to the uid element to understand this aspect. We have taken the total imposed temperature difference to evaluate the characteristic velocity. If ow becomes turbulent, it is likely that gradients of velocity and temperature are conned to layers near the walls. In such an instance, the velocity and temperature gradients can be spread over different distances from the wall depending upon Prandtl number. Thus, the selection of characteristic velocity can be expected to depend upon 8 for ow L  c 8 between two Prandtl number as well. It turns out transition occurs when A  vertical plates with large aspect ratio, i.e., is very small compared to the length of the plates.

30.3 Boundary layer formed due to natural convection


Now we go to the next level of complexity where both the temperature and velocity proles are developing1. A prototype of such a situation is heat transfer through boundary layers. Here we consider a hot at plate placed vertically in an innitely large medium. See Figure 30.2.
Expected velocity profile

Expected temperature profile x Boundary layer y

Ambient fluid is stationary. Temperature here is T 8

Figure 30.2: Boundary layer developing on a vertical hot plate The uid adjacent to the wall gets heated and hence its temperature begins to rise. As the uid next to the wall gets hot, it will rise due to buoyancy. As it moves up, it gets heated more and hence will continue to rise. Far away from the plate the temperature will remain at the undisturbed value of  . Thus we see a thermal boundary layer developing on the wall. As it moves up, it continues to get hotter, and the buoyancy increases. Hence uid elements will

This material is paraphrased from the book Convective heat transfer by Bejan

325

also accelerate. Far away from the wall, the uid will be stationary as buoyancy forces are absent there. Thus we also see a momentum boundary layer developing. Thus we expect that the temperature as well as velocity will change in both the vertical and horizontal direction. We can qualitatively sketch the velocity and temperature proles. The temperature is expected to monotonically change from the wall temperature to that of the ambient. The uid far away from the plate will be stationary as buoyancy force is absent there. However, the uid velocity must be zero at the plate due to the no slip condition. Thus we expect the velocity prole to have a maximum. This is shown in Figure 30.2.

30.3.1 Inuence of Prandtl number


As with heat transfer to a boundary layer formed under forced convection, we expect the relative rates of diffusion of heat and momentum to inuence the characteristics of heat transfer here as well. Consider the simple case of Prandtl number being unity. Here momentum and heat diffuse at the same rate. Hence we expect the velocity to be nonzero over the same thickness to which temperature gradients are conned to. The thermal and momentum boundary layers are of the same thickness. Suppose Prandtl number is less than unity. Then heat diffuses faster than momentum. We get an interesting situation here. It is clear that buoyancy forces exist in the entire region where temperature gradients exist and hence motion must exist through out the region where temperature gradients exist. Hence the thermal boundary layer must have two regions as far as momentum diffusion is concerned. We recall that viscous forces are denitely important in the region where momentum diffusion is important. Thus, near the wall where viscous forces dominate due to the no slip boundary conditions, the viscous forces must balance buoyancy forces. This region then constitutes the momentum boundary layer. Out side the momentum boundary layer, the buoyancy forces are balanced by inertial forces. The situation where Prandtl number is greater than unity also gives rise to an interesting situation. Here heat is diffusing slower than momentum. The buoyancy forces are created in a region to which temperature gradients are conned to. As motion is created in this region, the momentum diffuses rapidly and causes motion outside the thermal boundary layer even though temperature gradients are absent there. Thus, inside the thermal boundary layer, the buoyancy forces are balanced by viscous forces. The momentum boundary layer thickness is determined by a balance of inertial and viscous forces. The expected velocity and temperature proles are shown in Fig.30.3.

30.4 Scaling in natural convection boundary layers


The heat transfer rates in natural convection boundary layers can be determined if the thermal boundary layer thickness is found. The numerical values can be found only by integrating equations of motion and thermal energy balance. Scaling arguments can be used to nd the relationships between the various dynamical quantities. Let us illustrate this type of arguments here.

326

Momentum boundary layer edge Momentum boundary layer edge Thermal boundary layer edge Buoyancy ~ inertial Inertial ~ viscous Thermal boundary layer edge

x Buoyancy ~ Viscous Buoyancy ~ Viscous

Prandtl Number << 1

Prandtl Number >> 1

Figure 30.3: Inuence of Prandtl number on force balances in boundary layer

30.4.1 Estimate of buoyancy force

The characteristic temperature difference is known  to us. Let it be denoted by  . Then an esP B timate of buoyancy forces per unit volume is R . Here we take the reference temperature to be the ambient temperature,  . We use this force to estimate the characteristic velocity. The balance depends upon Prandtl number as discussed above.

30.4.2 Prandtl number is unity


Here we expect the thermal and momentum boundary layers to be of equal thickness. Let the boundary layer thickness be at some length W along the plate. We need characteristic velocities in both directions. In the direction we can choose it to be the maximum velocity. ^ Let it be represented by a . From the equation of continuity, a W is then an estimate of the characteristic velocity in the direction. The buoyancy force is of the same magnitude as either the inertial or viscous forces since both these are of same magnitude in the thermal boundary layer. The balance between the buoyancy and inertial forces gives

Now we can either turn to equation of energy and balance magnitude of convection terms with diffusion terms or go to equation of motion and balance the viscous and inertial forces to relate the boundary layer thickness to the position along the plate. Both will give the same result since both heat and momentum diffuse at the same rate. The balance of viscous and inertial forces is given by

K  c P a P R W

or

Wau

W  R

like it does with momentum boundary layer in the absence of heat transfer. It is as expected since this dependence arises from a balance of viscous and inertial forces. The heat transfer coefcient can be evaluated from an estimate of the heat ux by conduction at the wall:  6 c or uW W j c or

W  V R We notice that the momentum boundary layer thickness scales with Reynolds number exactly
or

K a K P a _ u W

Wauh

327

Hence we get the following scaling result

h The above result is also presented in terms of Rayleigh number dened as


Thus

A W j

6^  H W RK  V X  p K 

A W R  s H K p V

30.4.3 Prandtl number

'

Here we expect the thermal boundary layer to be thicker than the momentum boundary layer. The thermal boundary layer will have a zone where the viscous forces are balanced by buoy " be ancy forces and a zone where the inertial forces balance the buoyancy forces. Let and the thicknesses of momentum and thermal boundary layers respectively. Our estimate of the characteristic velocity will still be correct since the maximum in the velocity is being created by the buoyancy forces. Thus we have

Now we have to look at the equation of thermal energy to estimate the thickness of the thermal boundary layer. Thus by balancing the convection terms with diffusion terms we get,

Wa

W  R

au  " K  c W

or

As before, the heat transfer coefcient can be evaluated from an estimate of the heat ux by conduction at the wall:

K H K G" H W R  V  W Wa V  6c
or

 G" j "  c or j Thus we have the scaling result H W A uW j number It can be rewritten in terms of Rayleigh as

uW W " j

K R K  V  K p 

The momentum boundary layer thickness can be obtained by balancing the viscous forces and inertial forces. Thus we get

Wah Q

 p K W  V R  
328

h K

30.4.4 Prandtl number

'

Here the thermal boundary layer will be thinner than the momentum boundary layer. The maximum in velocity can be expected to occur at the edge of the thermal boundary layer. As thermal boundary layer is thinner than the momentum boundary layer, we expect that the maximum in velocity to occur close to the wall. Hence we expect the buoyancy forces to be balanced by viscous forces. Thus

a _ u " K

P R c a "K c W
or

or

au

 "K R

From the balance of heat transfer by convection with diffusion, we get

 "K "K R W

This can be rearranged to get

Thus we get

Zh The thickness of the momentum boundary layer can be found out by balancing the inertial and viscous forces K " H h K : h h V   p K
W auW W W R  

G" H W  W V R H R W V A uW j

'h

 K   K

30.4.5 Heat ux in natural convection


Recall that one of the questions we raised about unit operations approach in the beginning of the course was about the nonlinear dependence of heat ux on temperature difference. Now we are ready to answer this. Once the heat transfer coefcient is known, the heat ux can be calculated:   g K
-

Thus the simple analysis does show that heat ux depends nonlinearly on temperature difference!

30.5 Similarity solution of natural convection on a vertical at plate


We have seen that the results of scaling analysis give powerful insight into the relationships between various quantities. However, the numerical coefcients that appear in the relationships can only be obtained by solving the exact equations. Considering that the plate is innitely long and that the medium is unbounded, we can expect a similarity solution. We consider this solution briey. 329

As with isothermal boundary layers, it is expected that the velocity will be a function of only the ratio of the perpendicular distance the wall to the boundary layer thickness. Thus, I0 ^ U from referring to Figure 30.3, A a b only. Scaling analysis carried out earlier gives


Thus, it may be expected that

R  V

h K

I0 A au b

g H RK  V

In other words, the argument of the function is expected to be the similarity variable

g H RK  V X

where the number four is introduced in the denominator to make the nal differential equations look pretty. The ow here is two dimensional and hence it is more convenient to use stream function. Thus it is postulated  y The function of is introduced into the above since we are requiring that only velocities be a function of the similarity variable and not necessarily the stream function. However,

A Z

*`

 I A  A  Z

Thus, similarity hypothesis requires that

or

Introducing a numerical factor for reasons already mentioned, the stream function is then dened as K H

au  K K  R   

4 4

4 A` 4

A X

Turning our attention to temperature, it is postulated that a non-dimensional temperature is only a function of the same similarity variable :

v A

` K   h4 4

R K  V  s Z X

A v y

*`

The equation of motion, after using the boundary layer approximations, is given by

I I I0 I A M

K I0 K MSR   

330

The thermal energy balance equation, after using the boundary layer approximations, simplies to K

I0  I0  A K M

Introducing the non-dimensional temperature and the stream function, the above equations simplify to K  and

L Z M Z Z Z M v A v M p Z v A 

where prime refers to differentiation with respect to . Boundary conditions needed to solve the above are given below. The no-slip velocity condition at the plate and that velocity approaches zero far away from the plate requires

 Z A 0

A   

o `
A 

As stream function can be set zero at any convenient point, the following is used

 Z A 

The non-dimensional temperature must be unity at the plate and zero far away from the plate. Thus v A  at A  and v A 6

The thermal energy balance equation can be integrated in terms of Z :

v y A
where

*`

, 9 , 9

l l

p  c  p

>9  c@9 

  c

@9 4 4

, Z 

The equation of motion is nonlinear and hence solutions have to be obtained numerically. Once the solutions are obtained, the heat transfer coefcient can be calculated. Thus, A [   B  -   A A B A H R j X

or

The derivative of the non-dimensional temperature has to be evaluated numerically. It turns out that the following is a very good t over a wide range of Prandtl numbers:

g H RK  V j  A X H K  K 8 A A K V  v p V  v

j  K `

*` A v *` A [
L

[

K L p A  K 6 L  L  p M p X 9 M
331

As expected Nusselt number is a function of and p . It will be interesting to see how scale analysis our We need to only examine the K compares with the exact solution. K factor mul 6 tiplying purpose. If pS the term is proportional to p  for and hence  K this 6 , the term is,independent K . p . If p> of p , and hence    Both these are in agreement with the results of scale analysis.

30.5.1 Looking back

30.5.2 Transitions
The ow was assumed to be laminar and this assumption breaks down at

p 6 T .

332

Problems for Chapter 30.


30.1 Find Nusselt number by integral method for heat transfer from a hot vertical plate maintained at a constant temperature  and place in a n innite extent of uid at  . Heat transfer is occurring only by natural convection. It is also given that Pr = 1. You can assume the following temperature and velocity proles:

where

 A 

 . (Please refer to Bejan, p 155).

K H 6 A    G" V K I0 A G " H 6 G " a V

333

Chapter 31 PHYSICAL LAWS FOR MULTICOMPONENT SYSTEMS


We first discuss the complications arising in systems having several components and define a multicomponent body. We state the law of conservation of mass for species.

The concept of macroscopic flux due to diffusion is introduced and discussed. The law of conservation of species for a single component has been discussed in the context of uid mechanics. Often however, chemical engineers deal with multicomponent systems. Hence it is of interest to know how to generalize the approach used to derive balance equations to multicomponent systems. Briey recall the approach used for single component systems. Physical laws are dened for a closed system. Transport theorem connects the laws applied to a closed system to a control volume. By denition, there is no mass exchange across the boundary of a closed system. Consider a closed system containing many components and no mass is exchanged across its boundary. The composition of such a system then remains unchanged in the absence of chemical reactions. Thus, the dynamics of such a multicomponent closed system, i.e., whose composition is unaltered, can be described by the equations of change derived in the earlier chapters for a single component system. All that has to be done is to use the physical properties of the mixture of the prescribed constant composition. While this idea is useful, it is too restrictive since it would not allow generalization to instances where composition in a control volume changes due to either chemical reactions or exchange of species with the surroundings caused by concentration difference between the two. The main problem then is to be able to meaningfully dene a closed system when many components are present and when its composition does not remain constant. This situation is often encountered in science. Usually, under such circumstances, a hypothesis consistent1 with the existing structure has to be made. The longevity of the hypothesis depends entirely on the success with which predictions based on it match with experimental
A hypothesis in contradiction with the existing structures can also be proposed. Such would be a revolutionary and not an evolutionary development. Such a revolution face stiff resistance, and when it succeeds, it forms one of the fundamental laws of science!
1

334

observations. If it is consistently successful, it becomes part of the edice of scientic structure.

31.1 Multicomponent body


Now, it is proposed that a system containing many components is treated as a closed system in a restricted sense. In this sense, such a system will be referred to as a multicomponent body2 . A multicomponent body is dened as a collection of particles across whose boundary no net exchange of mass occurs as it moves in a continuum. In other words, in a multicomponent body, mass of some species can decrease and mass of some others can increase, but the total mass of the system remains constant. The surface of a multicomponent body can be imagined to be like a porous sieve which however manages to make sure that the mass of the body remains constant. Examine how such a system can be followed as it moves in a continuum. At some instant, identify some volume in the continuum, and that is the multicomponent body to be followed. Species will be crossing the boundary of the volume at that instant. Now move every part of the boundary of the identied volume by a differential amount, equal to the product of some innitesimal time interval and a velocity, and mark the new boundary. The velocities chosen need not be constant and can vary from point to point on the boundary. This would be an assumed position of the multicomponent body after a small time interval since it is not known that the velocities chosen were that of the multicomponent body. Suppose the velocity distribution chosen was such that the mass contained in the newly marked volume is same as that in the originally marked one. Then, the newly marked space occupied by the initially identied volume is such that, on the net, no mass was exchanged across its boundary. That of course is the denition of a multicomponent body. Thus, the volume being followed is a multicomponent body. If this process of marking the position of the boundaries of the originally identied volume is continued in time, the boundary of the multicomponent closed system would have been traced. The next task then is how to determine such a velocity distribution on the boundary of an arbitrarily chosen volume.

31.1.1 Mass average velocity


Individual species can have different velocities in the multicomponent body. The denitional restriction of no net mass exchange gives the handle needed to relate the velocities of individual species to the velocity with which every part of the boundary of a multicomponent body must P move. Let v= be the velocity of species . Let the mass concentration3 of species be = . The rate at which mass crosses a differential area at any point on the boundary, which moves with a velocity v, is given by

where n is the normal and


is the differential area. This must be zero for any arbitrary multicomponent body whose boundary passes through the chosen point. Surfaces of bodies arbitrarily chosen will have normal vectors pointing in arbitrary directions. Then, the only way in which the net mass ux across all arbitrarily chosen boundaries can be zero is to equate the
2 3


P = m = l =

To avoid confusion, it is not referred to it as a closed system since it is, in a strict sense, not a closed system This is the mass concentration and should not be confused with the density of species p .

335

velocity of the boundary of the multicomponent body to the mass average velocity:

Or

 A  P = = = A =

P = = P P P ^P where is the density of the solution. But the ratio = species . Hence A = = =

is equal to the weight fraction,

of

Since the velocity of each species is being weighted with the mass fraction, v is referred to as the mass average velocity 4 . Hence, a multicomponent body can be dened as one whose boundary moves with the local mass average velocity. As its velocity is dened, its contours can be traced and followed as it moves in space. Now it is proposed that all physical laws, namely the law of conservation of mass, Newtons laws of motion and the rst law of thermodynamics are applicable to a multicomponent body. This hypothesis forms the basis for deriving the balance equations for systems containing many components.

31.2 Physical laws applied to a multicomponent body


Now, a brief review of the procedure of deriving the balance laws for a control volume will be useful. Consider application of law of conservation of mass. The law of conservation of mass states that the mass of a multicomponent body remains constant. Now following the approach used earlier, we can write Total mass of the system

,  - = P = 5 ) c l a A ,  - P *) c l a

Thus application of law of conservation of mass to a multicomponent body gives

where a is the volume of the multicomponent body. By applying transport theorem and specializing it to bodies occupying the control volume of interest, i.e, using eq.6.2, the following result can be obtained:

0/

 A 

A , .- P *) c ll a ,

 A    P c l a M

A , *)

m P c ll

*)

(31.1)

where a and
are volume and surface area of the stationary control volume currently being occupied by the multicomponent body. This is identical to what was derived earlier, and is interpreted as
Averages based on other weights, e.g., mole fraction can be dened and are useful as will be seen in later chapters.
4

336

Rate of accumulation of mass in the CV

Rate of input of mass to the CV

Rate of output of mass from the CV

In a similar manner, balance equations can be derived by applying Newtons second law and rst law of thermodynamics. Referring back to the earlier chapters, it is seen that body forces and surface forces have to be dened to apply Newtons law. Similarly, the heat ux across the surface of multicomponent body has to be dened. These tasks will be considered in later chapters. However, the above discussion and equation does not throw any light on the law that governs the balance of mass of individual species in a multicomponent body, which is the focus of the present chapter.

31.3 Law of conservation of mass of species


The mass of a multicomponent body remains constant by denition. However, as species move across its boundary, the composition of the body changes. Thus the mass of individual species present in a multicomponent body will not remain constant. Similarly, chemical reactions could be occurring and for this reason also, the mass of individual species in a multicomponent body can change. Thus it is seen that we need a law governing changes mass of individual species to write an equation for balance of mass of individual species. The following hypothesis is stated for this purpose. The rate of accumulation of mass of any species in a multicomponent body is equal to the net rate of input of mass through its boundaries plus the rate of generation of mass of species due to reactions. The boundary of a multicomponent body moves with the mass average velocity. However since all species do not have the same velocity, some will move out of the multicomponent body while some will move in. The ux of species with respect to the mass average velocity 5 is dened as the diffusive mass ux. Thus, the diffusive mass ux relative to the mass average velocity is given by 

=A P = = =A =

As the diffusive ux is relative to the average ux, it is clear that sum of the diffusive uxes must be zero:  Thus, in a mixture containing components, only (n-1) uxes are independent. Before leaving this topic, it is advantageous to give another interpretation of the above P expression. The mass ux of species as measured by a stationary observer would be = v= . Suppose all species had the same velocity, v. The mass ux of any species as measured by the P observer would have been = v. Thus the diffusive ux is the ux over and above what is caused by the average motion. Thus it is advantageous to write Total mass ux of species

A = M P =

The last term is easily recognized as the usual convective ux. P In % this  sense, it is similar to the expression describing the total heat ux at any point: q +  v. It can therefore be seen
Many kinds of average velocities can be dened, and there would be corresponding diffusive uxes with respect to such average velocities. These ideas will be discussed later in greater detail.
5

337

=
=A  = = is the stoichiometric coefcient of F species in the F reaction. where F The stoichiometric coefcient is positive for products by convention. The intrinsic rate of reaction is denoted ; by . It is the rate of F reaction per unit volume, and is dened such a way that the ; rate of F production of moles of species F per unit volume due to reaction is given by = . The rate of production of moles of species per unit volume due to all reactions is equal to 2 = ; g where r is the total number of independent reactions occurring in the F F system. Let = be the molecular weight of species. The rate of production of mass of species per unit volume

that the diffusive mass ux is analogous to the diffusive momentum and heat uxes. Hence the same kind molecular mechanisms can be expected to dictate the diffusive migration of species, and that a constitutive relationship would be needed to predict the magnitude of the diffusive ux. can be applied, an expression is needed Before the law of conservation of mass of species for the rate of generation of mass of species by chemical reactions. The notation used by Aris6 is best suited for this purpose. Let there be r independent reactions occurring. Let F
= represent the species involved in the reactions. The chemical reaction can then be represented by the equation

due to all reactions is then given by

A rate expression is needed to compute ship is available from the domain of chemical reaction engineering. The balance equation for individual species in a multicomponent body can now be derived by applying the law of conservation. rate of accumulation of mass of species A 

= = ; g ; and it is assumed that such a constitutive relationF

 3

 ,  - " F&% P = *) c l a




By the law of conservation of mass of species, the above is equal to the net rate of input of mass through the boundaries of the multicomponent body plus the rate of generation due to reactions. The former of course is the diffusive ux. Hence the rate of accumulation of mass F of species is equal to

Thus, we have

, - " &%

m = 
M F

, 6- " F&% g = 


= ; a

, - " F&% m = 

, .- " F&% = 


338

 = ;  a A   F P = c ll a

 , .- " &% 5)

Introduction to chemical reactor analysis, Prentice Hall.

The balance equation for individual species in a stationary control volume is derived by using the same method followed earlier. The above balance equation is now specialized to a multicomponent body whose boundaries coincide with a stationary control volume. Thus, the following equation is obtained after using eq.6.2

 * ) l  * ) l l   P P , = A , = ,  6  c a M  " F&% c 

m P = c ll

*)

Substituting this result into the mass balance for the multicomponent body and after a little rearrangement, the following equation is obtained:

m P = c ll

*)

; A   = = = m
M g   a 

where a and
are the volume and surface areaF of the control volume. This equation is easily P interpreted. = n.v is the outward mass ux of component due to the movement caused by average F velocity. This is the usual convective term. Similarly, n.j = is the outward diffusive mass ux of component. Hence the above equation can be written as Net rate of input of mass of any component into the CV by average velocity plus by diffusion Similarity to heat balance The above balance equation is obviously similar to the heat balance equation. The terms of accumulation, and convection are identical. The conduction term in heat transfer is similar to the diffusion term here. The heat generation was due to chemical reactions or viscous dissipation. Here, it is due to chemical reactions. Hence experience gained in solving heat transfer problems should be of great value in understanding and solving mass transfer problems. Rate of generation Rate of accumulation of mass of that of mass of that + component due to = component in the CV chemical reactions

 3

 ,  P = 5) c l a

31.3.1 Consistency with balance of total mass


As no net mass crosses the boundaries of the multicomponent body, the balance equation for the total mass of the system has been derived in the previous section. It is given by

 A    P c l a M

A , *)

m P c ll

*)

(31.2)

It would be necessary to check if the balance equation derived for individual species is consistent with the above. The rate of change in the total mass of the system is obviously equal to sum of rates of change of of mass of individual species. Thus, if the balance equations for all the individual species are summed, the above equation must be recovered. The stoichiometric coefcients obey the following restriction.

= =A  =
339

for all . Further, sum of all the diffusive uxes is zero. Thus it is easily seen that the sum of balance equations does yield the balance equation for the total mass. It is important to point out a consequence of this. Mathematically speaking, obtaining the solution of mass balance equations for all the species is equivalent to solving the overall mass balance equation and the mass balance equations for all but any one of the species. As will be discussed in greater detail, it is the latter procedure that is followed often. From past experience, it can be anticipated that if the above balance equation for species P is applied to a small volume, it would give a differential equation for = . If that equation is solved, the mass concentration proles as a function of position and time in the control volume can be determined. Once again from past experience, it can be expected that a constitutive relationship for the diffusive mass ux is needed before the differential equation can be solved. These will be the topics for discussion in the next chapters. However, the momentum balance for a multicomponent body need to be discussed rst.

31.4 Momentum balance for a multicomponent body


According to the hypothesis made, Newtons second law is also applicable to a multicomponent body. The body forces acting on the multicomponent body are easily computed. Their magnitude is given by where g= is the body force per unit volume on species . A mass averaged body force g can be dened now:

,  - = P =5tu=  a tA

where = is the weight fraction of species in the mixture. In terms of the mass averaged body force, the body force acting on the multicomponent body is given by

= = =

t

, .- P t  a

It is seen that this expression is identical to that for a single component as long as the density corresponds to the density of the mixture, and the body force is the mass averaged quantity. The surface forces are assumed to not act separately on different components. It is assumed that there is a single stress tensor that represents the effects of all the components. Hence, the surface forces acting on the multicomponent body are given by

, - mJD 

, - O 
M , - m R 
A , -

The above is identical to what was used in a single component system. The multicomponent nature of the substance is reected in the constitutive relationship. Thus, if Newton Stokes law is valid for the mixture, the viscosity will be a function of the composition. The rate of change of momentum of the multicomponent body is given by

 ,  - =

 P = = a A  P  a 
340

where the denition of mass average velocity has been used to convert the rst term into the second. This expression is also identical to that derived for a single component. Without going through all the other steps, it is easy to see that the momentum balance will be identical to that for a single component provided the velocity and body force are the mass averaged quantities, and density is the density of the mixture. Thus, in summary, the momentum balance for a multicomponent body can be written as:

 O
M m 
M  P  a

, R

, t

A A

 , 6 , 5) 5)
= = = = = = = P=

  P a   P l l l  l l   P   c c a M m c c

5) 5)

where

t
31.5 Birds eye view

P A A

t

It is useful now to get an overall picture of what has been done. Consider a multicomponent system. Mass averaged quantities of velocity, and body force were dened for such a system. The balance equation for the total mass turns out to be identical to that derived for a single component, namely the equation of continuity. The linear momentum balance for a multicomponent system also is found to be identical to that for a single component. In other words, the mass average velocity satises the equation of continuity and the Navier Stokes equations 7 of motion. Consider the system to be isothermal, i.e., heat effects are not involved. Under those circumstances, the equation of continuity and the three components of NSE can be thought of the equations for determining the four variables and v. O fractions. Only i 6  of them can be The other variables in the system are the weight considered as variables since weight fractions must add up to unity. The overall mass balance has been used up when the equation of continuity was solved. As was pointed out in the 6  independent discussion above, for an component mixture, only equations of species 6 balances are available. That of course, is exactly the number needed to determine the  6 weight fraction variables. Thus, if the species balance equations can be solved along with the equation of continuity and NSE, the concentration distributions can be determined.   6 6 The species balance equations involve independent diffusive uxes, = . Constitutive equations are needed to predict these. The rst step then is to examine the mechanism of diffusion, which is the topic of the next chapter.

It is being assumed that the multicomponent mixture obeys Newton Stokes law of viscosity.

341

Chapter 32 MECHANISM OF DIFFUSION


Discuss the mechanisms of diffusion The driving forces for diffusion Constitutive relationships for binary systems It was shown in the previous chapter that the denition of a multicomponent body, when used in the species mass balance, gave rise to ux of species with respect to the mass average velocity of the boundary of the body. The ux is referred to as diffusive ux. In the earlier chapters, we have discussed the molecular mechanism of momentum as well as heat transfer. The mechanism of diffusion of mass of species is similar to that of heat and momentum. It is the random molecular motion that forms the basis of diffusion of mass as well. Under the inuence of a concentration gradient, species acquire a macroscopic velocity as they move by diffusion. The diffusive ux enters the species mass balance equation, and in order to solve it, we need a constitutive relationship for the mass ux vector. We will develop the relationship in a manner similar to that presented in the case of heat ux vector.

32.1 Diffusive motion and diffusive ux


In the appendix on turbulent dispersion, we showed that a particle steadily moves away from its starting point due to random motion. Molecules execute random motions and hence drift away from any location. This motion will be referred to as diffusive motion. However diffusive motion alone can not cause a ux that can be observed at a macroscopic level. Consider a homogeneous mixture of gases in a box. If a molecule is marked and observed, it will be found to move randomly from place to place. The drift velocity vector however is randomly oriented. In other words if we take a plane and observe drift motion, we will nd that, on the average, equal number of molecules move away from the plane in opposite directions. Thus if we consider a thin slab shown by dotted lines in Figure 32.1, the ux to both sides of the slab will be the same if the there is no gradient in the composition of the mixture. Hence net ux in any direction will be zero. Thus diffusive motion alone can not cause macroscopically observable ux. A concentration gradient is needed for a net diffusive ux to exist. The situation when there is a concentration gradient is also shown in the gure indicating the way a macroscopic ux arises. It is the macroscopic diffusive ux that appears in the species mass balance that we derived in the previous chapter, and a constitutive relationship is needed to link it to the concentration gradient.

342

No concentration gradient
Equal numbers move to left or right

Low concentration

High concentration

Equal numbers move to left or right

Less flux approaches the plane from this side Same flux approachesthe plane from both sides. No net flux. Plane

More fluxapproaches the plane from this side

Figure 32.1: Flux and concentration gradient

32.2 Diffusive ux vector


Chemical potential is a well dened concept in thermodynamics. For any species, it determines the activity of a species in a mixture. Thermodynamic equilibrium prevails when the chemical potential of the species is equal in all phases1 present in a system provided the temperature and pressure in the entire system are constant. If the chemical potential of any species in different locations in a single phase is different or is different in different phases, the system will tend to attain equilibrium by migration or diffusion of species. Just as temperature differences caused heat transfer to drive the system toward equilibrium, inequalities in chemical potential cause mass transfer so that the system can tend toward a state of equilibrium. As the species move, they obviously must have a direction and hence, the diffusive ux is also a vector just as heat ux is a vector.

32.2.1 Direction of mass transfer


The direction of movement of species must be such that, at constant temperature and pressure, free energy of the system is reduced due to diffusion. Consider a thin slab F in a single phase of thickness located at Figure 32.2. Let the chemical potential of species at the two . See   _ _ edges be = and = M . Let = moles of species be transported across the slab. Thus,

dn i

dn i

x + x

=
1

Figure 32.2: Direction of ux and gradient in chemical potential moles enter slab from with a chemical potential of

_ =  , and leave from M

with a

The species under consideration must be soluble in the phase, and the phase must be accessible to the species, i.e., the phases can not be separated by impermeable barriers.

343

chemical potential of = this transfer is given by

mass transfer occurs in the direction of decreasing chemical potential. This is similar to the heat ux occurring from regions at higher temperature to those at lower temperatures. The chemical potential is proportional to the concentration. Hence, we can say that the direction of mass transfer2 is from high concentration to low concentration. The same can be said of transfer of species between phases too.

_ M  . The the change in the free energy change of  = moles due _ = M  _ =   =  The change must be negative. = is positive as shown in Figure 32.2, since mass transfer is   _ _ = = M for the free energy to decrease. Thus in the positive direction. Hence,

32.3 Mass transfer by diffusion


The mechanism underlying the mass transfer is diffusive motion caused by random molecular motions. This is similar to heat conduction. Thus, we can expect a constitutive relationship similar to the Fouriers law of heat conduction. To get the basic ideas clearly in a simple setting, we will restrict the discussion to isothermal and isobaric binary systems in this chapter. In later chapters, these restrictions will be relaxed. The two components in a binary are denoted by 1 and 2. The constitutive relationship that links diffusive ux to the concentrations is Ficks law of diffusion. It is written as where g K is the binary diffusion coefcient. In a binary, g M K A and by denition, j g + jK = 0. Hence, only one diffusion coefcient is required to characterize diffusion in isothermal isobaric binary systems. The similarity between this and the Fouriers law of heat conduction is apparent. Thus, we can conclude that the diffusive mass ux vector is normal to the curves on which weight fraction remains constant, i.e., mass ux is directed along the steepest gradient of mass fraction. The above constitutive relationship is often given a physical interpretation which offers some insights into the constitutive relationships. Let us reconsider the free energy F decrease   _ _ = = for ux of species in argument we just discussed. We stipulated F = that M the direction to be positive. Suppose is the component of the diffusive velocity of F species. Then of moles of species that move F in the direction is = = , the number  =A in time = , where is the molar concentration of species. Hence, for the given by change in free energy to be negative,

P gK g gA

! !

' 4 and write it as We can generalize this to three dimensions 4 = mJb _ =  = ' 
We can rewrite this as

_ =

 _=

_ =  = = 

= = 

This direction is dictated by the tendency to equilibrate. If the system is deliberately maintained at a nonequilibrium state, then it is possible to have mass transfer from regions of low concentration to those at low concentration in systems containing three or more components. In binary systems, the direction of mass transfer is always from high to low concentrations.

344

Since both

and

! are positive, the above can be simplied to = m$b _ =  

At constant temperature and pressure,

_ =A
where

b 4 4 or a potential. Thus it can be regarded as The above is in the form of gradient of a free energy a force: b _ =A =
_ =A
m

is the Gibbs free energy, and

is the number of moles if

4=

  =

species. Thus,

Thus we have the relationship

= m = 

This is easily satised if the velocity of species is proportional to the force:

=A

where

is the mobility. Since mass ux is proportional to velocity, it is easy to see that

= = _ =

==A = _ =

P=

which is the essence of Ficks law. The argument presented is general and is valid even if diffusive motion were to occur under the inuence of gradient of some other potential. We will have occasion to refer to this kind of argument again.

32.3.1 Alternative denitions of diffusive uxes


We have dened diffusive ux as the macroscopic ux of species over and above that caused by the mass average velocity. This denition has arisen since we have postulated that the physical laws are applicable to a multicomponent body. However, it is possible to dene different average velocities and hence different diffusive uxes. There is no evidence to suggest whether any of these averages have an intrinsic advantage over the other or have a more fundamental signicance. Different average velocities and diffusive uxes are used based on convenience and we will dene one more here. The differences and advantages will be considered as and when examples are taken for study.

32.3.2 Molar average velocity


If the velocities of species are weighted by mole fractions, the molar average velocity, v is obtained.

b A = = =
345

The molar ux of

species as seen by a stationary observer is given by

where = is the molar concentration. If the species velocity was also equal to the molar average velocity, the molar ux would have been = . The diffusive ux then is the ux over and above that due to average motion:

=A = =

= A = = b 

It turns out that Ficks law for binary systems is equivalent to

g A

gK g

where is the molar density of the solution, and the diffusion coefcient is the same as that in the expression that relates j g to the gradient of mass fraction. It turns out that use of molar diffusive uxes is advantageous when dealing with gases and when chemical reactions occur. In case of gases, especially at low pressures and high temperatures, the molar density is related to the pressure and temperature by the ideal gas law and hence it is advantageous to use molar units. The rates of chemical reactions are usually related to molar concentrations, and hence once again it is better to use molar units. Volume average velocity is also used some times. It is obtained obtained by weighting the velocities of species with respect to volume fractions.

32.4 convection
We have seen that heat can be transported from location to location by bulk movement of uid since the temperature can vary from place to place. As uid moves from one place to another, it can encounter environment at different concentrations. Then it will exchange mass of species with the environment and mass transfer occurs. However, for mixing to occur at molecularly homogeneous level, diffusion has to occur, just like conduction was the ultimate mechanism of spatial homogenization of temperature in heat transfer. The convection or bulk movement of uid caused by an external agency is forced convection. If density gradients caused by differences in the density of mixtures of different compositions coupled with gravity cause motion, it is referred to as natural convection. This nomenclature is similar to what was stated in the case of heat transfer as well.

32.5 Multicomponent systems and other mechanisms


Emphasis has been placed on binary systems. Diffusion however can involve many components. Diffusion in multicomponent systems will be considered later so that ideas can be xed in simpler situations rst. However, it is worthwhile to note that in multicomponent systems, diffusion can occur against the concentration gradient, and diffusive ux can be zero even when concentration gradient is present! One more complication also exists. There are other mechanisms by which diffusion can occur. These are 1. Thermal diffusion: Temperature gradients cause diffusion and this is referred to as thermal diffusion. Isotopes are separated using this principle. 346

2. Forced diffusion: If charged species are present, they can diffuse under the inuence of electrical elds. Electrolytic cells, fuel cells, batteries are important examples of this. 3. Pressure diffusion: Pressure gradients can also cause diffusion. Centrifugal separation is an example. These also will be considered later on.

347

Chapter 33 PHYSICS OF MASS TRANSFER


The physics of mass transfer process is We discuss the boundary conditions. discussed.

In this chapter, we would like to develop a physical picture of how the effects of diffusion develop in systems. For mass transfer to occur, there must be a concentration gradient. Typically, such a gradient is created by bringing two phases into contact. Consider an example. Suppose we add sugar to water. Then, as water does not contain sugar, the chemical potential of sugar in water is less than the chemical potential of the solid sugar. Thus a chemical potential difference is created and sugar begins to dissolve into water. The boundary conditions will be discussed in detail later but we will state one boundary condition that is needed to examine the development of mass transfer process in detail.

33.1 Equilibrium at the interface


A basic hypothesis of diffusion is that thermodynamic equilibrium prevails at the interface between two phases. Thus, chemical potential of any species which can be present in both the phases will be equal on either side of the phase boundary, i.e., chemical potential is continuous across the phase boundary. This is similar to the condition of continuity of temperature across phase boundary. Consider once again the example of addition of solid sugar to water. The boundary condition then states that the concentration of sugar in water, at the interface between sugar solid and liquid phases, is equal to the solubility of sugar in water. Water is not soluble in solid sugar, and hence we do not apply any equilibrium boundary condition for water. However if water could be present in sugar, then chemical potential of water in the liquid phase and in solid phase at the interface must be equal according to the boundary condition.

33.2 Development of concentration proles


Now let us consider dissolution of sugar crystals into water in detail. To simplify the issues, let us imagine that sugar is present as an innitely wide slab and water is in contact with one side of it. Refer to Figure 33.1. As soon as water comes into contact with the solid sugar phase, the concentration in the water phase at the interface rises to the value of the solubility of sugar in water. However, the sugar concentration in water at all other locations is zero. Hence, a very large concentration 348

Solubility in water Increasing time

Concentration
0

Sugar
Direction of diffusion

Water

Figure 33.1: Development of concentration proles

gradient is set up. This causes a diffusive ux. The magnitude as well as direction of it is given by the Ficks law. The direction of diffusive ux is in the direction in which the concentration decreases. As the concentration decreases in the direction away from the solid-uid interface, sugar diffuses into the bulk of the liquid phase away from the solid phase. Part of the diffusive ux increases the concentration while a part of it diffuses further according to the species mass balance. With increasing time, the diffusion causes sugar to penetrate deeper and deeper into the bulk of the liquid phase. The concentration proles thus become less and less steep with increasing time. It might be noted that this is similar to the way the velocity proles develop in drag driven ows and temperature proles driven by conduction. However there are some dissimilarities as well. Unlike with temperature prole development, as sugar molecules move into the liquid phase, they create a velocity in the liquid phase1 . This is a complication which will be discussed in detail in the next chapter. One more difference, though in detail, also exists. In the example of drag driven ow, see Figure 10.1, the stationary plate plays the role of a momentum sink. Similarly in case of heat transfer, if at some distance from a hot slab, we placed a slab being maintained at a lower temperature, the latter will play the role of a heat sink. In both these instances, steady state can be reached when the supply from the source matches the loss at the sink. In the example we just considered, solid sugar can be considered as a source of sugar, and steady state will not be reached since a boundary where sugar molecules can be consumed is not available2 . However the example being considered is almost identical to the drag ow set up when a plate is suddenly set into motion in an innite extent of an initially stationary uid, and the corresponding problem of conductive heat transfer from a semi-innite slab. Another difference between mass transfer and momentum as well as heat transfer can also arise. In case of both heat and momentum transfer, problems involving specied heat or momentum ux can arise. Thus, a plate can be supplied with a constant heat ux or a plate can be dragged with a constant force. In mass transfer, such is hard to imagine, though not impossible to arrange e.g., in controlled drug delivery.

33.2.1 Mass transfer with chemical reactions


If in the above problem, a chemical reaction consumes the diffusing species, it is equivalent to providing a sink in the bulk of the uid. In such an instance, a steady state can be reached when
Further, the solid-liquid interface will have to recede in a direction opposite to that of diffusive ux. A velocity will be created in liquid phase due to this. This will be briey considered a little later. 2 An situation analogous to a sink can arise if sugar is consumed at a surface for example by a catalytic reaction.
1

349

the rate at which the component diffuses into the liquid matches the rate at which it is consumed in the entire volume. Pressure drop and body forces can be thought of as bulk sources or sinks of momentum, while ohmic heating is an example of bulk source in heat transfer.

33.3 Boundary conditions


Boundary conditions dictate the development of concentration proles. Let us consider the general boundary conditions for isothermal and isobaric systems. Generalization to nonisothermal and varying pressure conditions will be discussed later on.

33.3.1 Concentration conditions


As mentioned earlier, one of the general conditions concerns the chemical potential. The chemical potential of all species that can be present in both the phases must be continuous across the phase boundary. Refer to Figure 33.2. The condition is stated as:
I II v

Figure 33.2: Boundary conditions

_= A _=

However, as chemical potential can not be measured directly, it is more convenient to specify boundary conditions in terms of some measurable quantities. Usually, concentrations and partial pressures are therefore used. It must be remembered that when conditions are specied in terms of these alternative measures, it is equivalent to continuity chemical potential. These conditions, though they are very different in form, are therefore equivalent to the continuity of velocity vector and temperature across a phase boundary. Thermodynamics provides the relationship between the chemical potential and the concentration variables, e.g., concentration or mole fraction etc. Let us consider a few examples which are commonly used. Gas-liquid interfaces Suppose we have a gas-liquid interface. Typically, the equivalence of chemical potential is stated in the form of Henrys law. For example,

desired. Henrys law is commonly used in dealing with gases dissolving in nonvolatile liquids.

=A = = (33.1) O F where = is the partial pressure of the component in the gas phase, = is the Henrys law F O and = is the molar concentration of the component constant, F in the liquid phase. This can be converted to a relationship between mole fractions of the component in both the phases if

350

Vapor-liquid interfaces Suppose we have a vapor-liquid interface. Raoults law is commonly used. It is valid for ideal systems only. It states that = A = = (33.2)

where is the total pressure, = is the mole fraction of component, superscripts and refer F = to the vapor and liquid phases, and is the saturation vapor pressure of component. For non-ideal systems, corrections through activity coefcients can be introduced.

Liquid-liquid interfaces The equivalence of chemical potential is stated in terms of the slope of a plot of the locus of mole fractions in the two phases at equilibrium, <= . The slope is referred to as the partition coefcient. = A<>= = (33.3) where superscripts and  refer to the two liquid phases in equilibrium. Solid-liquid interfaces The condition of equality of chemical potential is equivalent to the concentration of the solute in the liquid phase being equal to its solubility.

=A =
33.3.2 Flux conditions

Another set of boundary conditions pertain to ux of species. Conditions on ux are equivalent to continuity of stresses and heat uxes across phase boundaries. As net mass can be transferred across a phase boundary, the boundary is not made up of the same particles at all times. As a result the boundary may move with a velocity different from the mass average velocity of both phases in contact. Consider the example of a volatile liquid placed in a beaker. The level of the gas-liquid interface will fall while the vapor will move up. Thus velocity of the gas-liquid interface is not equal to the mass average velocity in either the gas phase or the liquid phase at the interface. In general we need boundary conditions at such an interface. Let V be the velocity of the interface. The interface has no volume. Thus, the rate of accumulation of mass of any species has to be zero. The interface can be a catalytic surface in which case, chemical reactions could occur there. Thus, a massF balance of species can be made on the interface. This would state that the net F rate of ux of species to the interface from both the phases plus the rate of generation of species per unit area must be equal to zero. To implement this, we need to specify the rates of surface reactions. As with bulk chemical reactions, the overall rate of production can be specied in terms of rates of independent surface ; reactions. Let us followF the same methodology as with the bulk chemical reactions. Let be the intrinsic rate of surface F reaction. It is F dened as rate per unit area. Let = be the stoichiometric coefcient of species in the surface reaction. F F Then the rate of generation of species is given by summing the rate of production of species due to all reactions.

/ 3 /

351

Now we are in a position to write the boundary condition involving ux. Refer to Figure 33.2. The mass balance states that

P = = \  m P = = \  m M

? ? 8

8 

= ; A  g

/ / 3

(33.4)

The rate of surface reactions has to be specied and it belongs in the domain of chemical reaction engineering. Let us take a simple example to make application of this boundary condition clear. Consider a stationary liquid-liquid interface. Since it is stationary, V = 0. If there are no reactions occurring at the interface, the boundary condition gives

P= =m A P= = m

  8

This is similar to the conditions of continuity of heat ux or of stress continuity in uid mechanics. Let us take a slightly more complex example. Consider a stationary impervious catalyst surface. Denote it as phase . Due to impervious nature of the catalyst, v = = 0, and since the catalyst is stationary, V is also equal to zero. Let a surface reaction that is occurring be denoted by

M L

Let the normal point into the adjoining uid phase containing the reactants and products. Ap plying the ux boundary condition we get

surface in appropriate stoichioThe above simply states that species must reach the catalytic ] metric ratio. Such equalities play an important role in solving mass transfer problems. L A The stoichiometry of the reaction also suggests that . Thus by adding all M the three conditions we get ] ] P   M P M P m A P m A

8  / 8  3 / 8  3/ 3 where the superscript corresponding to phase was dropped for simplicity. The above condi tions are rearranged some times to give the following equalities: 8 P m8 P m8 P A L A m   
; A  P m ; P ] ] m L ] A  ; A  P m M
 ] ] ]

Thus boundary conditions for the mass average velocity can thus be derived from those for individual species.

352

33.3.3 Mass transfer coefcient


Heat transfer coefcient was introduced earlier. If it is known, the heat ux can be calculated from the overall driving force. Mass transfer coefcient is similar to the heat transfer coefcient. We jump ahead and illustrate the use of mass transfer coefcient in boundary conditions. The diffusive mass ux can be calculated using the mass transfer coefcient, and is written as the driving force. The total mass ux of species
the product of mass transfer coefcient and in any uid phase is given by

mass transfer coefcient,

P A M P calculated using the mass transfer coefcient: can be The diffusive ux at the interface m A j P = P  j P  P Here, n is the normal to the interface of the uid phase and points into the uid phase, = is P bulk concentration the concentration at the interface, and is the in the uid phase. j is the

The above expression for the ux is used as a boundary condition if j is known. As explained in the chapters on convective heat transfer, the heat transfer coefcient can, in principle, be obtained by solving the equations of thermal energy balance. Similarly, the mass transfer coefcient can also be obtained by solving the equations of species mass balance. However, in turbulent ows, this is not possible and hence a correlation is used to obtain the ux in the uid phases. Under such conditions the above representation is useful. Once again, as with heat transfer, if mass transfer coefcient is given for a phase, the Navier Stokes equations need not be solved as far as calculation of uxes is concerned. It must be mentioned that mass transfer coefcients are dened on the basis of other driving forces as well. The most common one with chemical engineers if of course based on molar concentration difference. EXERCISES Derive boundary conditions for a liquid insoluble in liquid.

P  P m P m A j  M

vaporizing into a gas

. Assume that gas is

353

Problems for Chapter 33.


33.1 A pool of stagnant water contained in a circular tube of diameter is evaporating from its surface into the vapor phase above. Assume that vapor phase is also com l  posed of water only. Let W be the height of the liquid level measured from the bottom of the beaker. a) What  is the velocity of water in the liquid phase? 

b) Let  be the rate of evaporation of water at any time. Relate it to c) What is the velocity in the vapor phase at the interface?

 .
W

L(t)

Water

Figure for problem 33.1.

33.2 A candle is known to get extinguished slowly after lighting it when gravitational forces are absent. Why?

A gas stream containing A at temperature  ows past a 33.3 (iii) , and catalytic wall with a velocity a . Species A reacts  at the catalyst surface at a rate of  j$ (moles per unit area per unit time) where is the mole fraction of A in the gas phase adjacent to the catalyst surface. The heat of reaction is per mole of A consumed. The catalyst surface is insulated on the side not exposed to the gas stream. We are interested in obtaining temperature and mole fraction proles at steady state. (iv) An insulated long tube is partly lled with a volatile liquid A initially at  . The rest of the tube is lled with a stagnant gas B which is insoluble in A. Initially gas B is also at a temperature of  , does not contain any A, and is separated from  A A by a membrane. At , the membrane is removed and the liquid and gas come into contact with each other. Due to vaporization, the liquid interface moves to the left. We are interested in calculating the position of the liquid interface as a function of time, Z(t). Denote the saturation vapor pressure of A by and it is known that it O is very small compared to where is the total pressure in the system. The total molar concentration can be considered to be constant.

6

354

A V T Catalytic surface
X

X A z Z(t) Insulated Tube 8

=0

8 8

X A

8 8 Insulated Part iii

Part iv

Figures for problem 33.3.

355

Chapter 34 SPECIAL FEATURES OF MASS TRANSFER PROBLEMS


If the sequence used in momentum and heat transfer chapters were to be followed here as well, shell balances should come now. Mass transfer creates some special difculties and, unaware of these, if one proceeds to make shell balances, confusion will be the result. This chapter is therefore devoted to discuss these difculties. In chapter 31, it was pointed out that heat balance and mass balance equations are very similar. Heat transfer problems were classied into two categories: conduction and convection. The rst attempt would then be to classify the problems in mass transfer also in a similar manner. Thus, one might expect that, similar to where heat transfer occurred only due to conduction, mass transfer may also occur purely by diffusion. These class of problems are referred to as pure diffusion problems. Then, there can also be problems where convection is present, and they can be referred to as convective mass transfer problems. An important difference between pure diffusion problems and heat conduction problems exists and it has to be clearly understood. In heat conduction, there is no macroscopic 1 motion at all. Thus, all convective terms in heat balance could be put to zero. However, when mass transfer occurs, it is necessarily accompanied by motion. This motion has to be distinguished from forced convection where an external agency creates motion and from natural convection where motion occurs due to buoyancy forces created through density gradients caused by mass transfer.

34.1 Bulk velocity due to diffusion


A brief review of the concepts of velocities of species and diffusive uxes will be of help here. Species must have a velocity if they move from one place to another. Let the velocity of species to this motion. of species per unit be denoted by v= . A ux of species exists due F If the mass P = = volume of solution, or mass concentration, is , the ux of species n is given by

=A P = =

In fact, in general, only the left hand side is what can be measured in a laboratory frame and hence the above equation can be thought of as the dening relationship for the measurement of
The word macroscopic is being used to exclude motion at molecular level, which always exists. Macroscopic motion refers to that which can be detected by the usual methods.
1

356

velocity. If all species in the mixture possess the same velocity, relative motion of one species with respect to others is absent. A diffusive ux exists only if there is relative motion. The diffusive ux is therefore dened relative to an average velocity of all the species. Average velocity can be dened by weighting the individual velocities with some property. If averaging is carried out with respect to mass fraction, = , the resulting quantity is the mass average velocity2 :

A = = =

If the species velocities are weighted with mole fraction, obtained:

= , the molar average velocity, v

is

A = = =

It is obvious that if all species move with the same velocity, the average velocity is same as the velocity of the species. Thus there is only ux due to bulk motion and diffusive ux is absent. The ux relative to the mass average velocity is dened as

 =A P = = A P = = P =

The ux relative to the molar average velocity is dened as

by bulk motion are the diffusive uxes. If the velocities of species are different, since the molecular weights of species are different, the mass average velocity and the molar average velocity will not be equal. In particular, if one is zero, the other need not be zero. Suppose that motion occurs only due to diffusion. In other words, there is no external agency to create motion and that motion is not due to buoyancy forces caused by density gradients. Even then it is possible that v is not zero. Hence macroscopic motion can be created by diffusion alone. There is no analogue of this in heat transfer.

= A = =  A = = = where = is the molar concentration of species . Thus the uxes over and above that caused

34.2 Bulk velocity and equation of motion


There are several tricky questions associated with the velocity created by diffusion. Consider the following experiment being done in the absence of gravity. A very long tube is partitioned into two halves separated by a membrane. The two halves are lled with ideal gases
and . They are at the same temperature and pressure. At some instant, the membrane is suddenly ruptured and the two gases are exposed to each other. There is no pressure gradient initially since both compartments are at the same pressure. The gases will diffuse into each other due to the concentration gradient. One would expect that the pressures in both compartments will remain the same. As a result (since the temperature also is constant), the total molar concentration will remain constant. Hence, a balance of the total number of moles will yield the result that the total molar ux must be independent of the position in the tube, and hence a constant. Consequently, the molar average velocity is also a constant, and it must be zero since the tube ends are impermeable. This situation is what is normally referred to as equimolar
It is called mass average velocity also since the product of density of the solution with the mass average velocity gives the total mass ux, n.
2

357

counter diffusion: for each mole of


diffusing to one side, a mole of diffuses to the other side. Thus the molar density remains constant and the total molar ux is zero. However, as the molecular weights of
and are not the same, mass average velocity will not be equal to zero. If the no slip condition is satised at the tube walls, and that is expected to be valid, then there must exist a pressure drop. This contradicts our assumption that pressure must remain constant. However, apparently, no mechanism seems to exist by which pressure gradients can develop. Several such intricate questions can be found in diffusion theory 3 . Thus, simple looking problems seem to get very complex if it is insisted that the velocity caused by diffusion alone should also satisfy the equation of motion. The bulk velocities created by diffusion alone are very small and hence the inaccuracies caused by the assumptions, which have to be made if we do not insist that the equation of motion be satised, are likely to be not important. Thus, it is customary to neglect the equation of motion in solving problems involving only diffusion. It must be once again emphasized that this pertains only to a situation where only diffusion occurs and there are no other sources of motion, i.e., forced convection or buoyancy driven natural convection are absent. If we do not use the equation of motion for determining the bulk velocity, the number of equations available to solve falls short of the variables to be determined. This is understood as follows. Take a binary incompressible system as an example. Since incompressibility is P assumed, the density of the mixture is treated as a known. There are two diffusive uxes: j and j . However, only one of them is independent since j + j = 0. Ficks law relates the ux to the gradients of weight fraction:

A P

The mass conservation equation in the absence of reactions for both species
can be written using Ficks law as

m P 
M P 

  P A m


As M A , only two unknowns are there: and v. Only the above mass balance is available. Hence another equation is needed to determine v. Thus, if the equation of motion is not used, the number of equations fall short of the number of unknowns. Ignoring the equation of motion implies that other conditions4are needed to solve for the bulk velocity. Some times, the problem posed provides scope for some approximation. The well known equimolar counter-diffusion approximation arises from energy considerations in distillation. Insoluble nature of a component provides the approximation that velocity of that component can be set to zero. These points will be illustrated in the next chapter. In the next section, a widely used condition, called dilute solution approximation is discussed. It is easily understood and also helps clarify the ideas discussed.
It must be emphasized that there may be no contradiction here. In principle there are enough number of equations to nd a solution for all the variables. In the particular issue being discussed, perhaps the assumptions of constancy of pressure and one dimensional diffusion may have to be abandoned. It is counter intuitive, but a solution can be found! 4 The discussion presented here is similar to that presented by S. Whittaker, Transport in porous media, 2, 269-299 (1987).
3

A , 

P a

358

34.3 Dilute solutions


Consider a mixture containing -components. Then,

the others. That component is referred to as the solvent. The other components are referred to as solutes, and this mixture is called a dilute solution. Thus in a dilute solution,

= A = M =P F Suppose one of the components, say the component, is present in large excess compared to = 6 c A 6 c L cnmnmom,c C 6  h c ? 6 =A = c 6

In such a situation, the diffusive ux of the solutes are of interest and solvent ux is of no 6 , the total ux of solutes can be approximated by the diffusive ux interest. Since = itself:  As the weight fractions are small, their gradients can also be expected to be small. Thus, the uxes would also be small. Under these circumstances, the bulk velocity caused by diffusion alone is negligible and one need not be concerned about the issues of satisfying the equation of motion etc. The mass balance equation for solutes simplies to

  P A m


 , 

P a

and this equation can be solved for . As concentrated solutions may involve high uxes, diffusion in such solutions is referred to as high ux mass transfer case. In contrast, problems involving dilute solutions are also referred to as low ux cases. Hence, it is customary to ignore the equation of motion in solving problems involving diffusion alone in dilute solutions. As mentioned, problems of diffusion alone but in concentrated solutions will have to be dealt case by case and some examples of it are discussed in the next chapter.

34.4 Molar form of mass balance


Some times it is found that use of molar units of species is more convenient than the mass units. For example, when dealing with mixtures of vapors at low constant pressure and temperature, it is advantageous to use the ideal gas law and the inference that total molar density of vapor phase is constant. Note that as the composition of mixture will in general vary from point to point, the mass density will not be constant. Thus, it is some times useful, especially when dealing with diffusion problems, to formulate balance in terms of moles rather than in terms of mass. Consider the species mass balance over a stationary control volume: Net rate of input of + mass of any component into the CV by average velocity plus by diffusion Rate of generation = Rate of accumulation of mass of that of mass of that component due to component in the CV chemical reactions

359

The above can be converted into molar units by dividing by the molecular weight of species . The last term of the balance is given by

 ., - " &% 5)


  P = c l F

If the above is divided by the molecular weight of the species, be in molar units: 

  = c l F where = is the molar concentration of species . The rst term of the balance is given by l l   P m = c
m = 

 ., - " &% *) 5)

= , the resulting expression will

From the above, it can be seen that for stationary control volumes, the mass ux of species at P any point on the boundary is given by = M = . Expressing this in terms of velocities,

P = M =A P = M P = =  A P = = P = = = A = =A

If the above is divided by the molecular weight,

Thus the rst term of the balance, after dividing by molecular weight of species is given by

m = c ll

*) 3

The second term of the equation is given by

, = g  

= ; a

After dividing by the molecular weight of species , it is simply equal to

, = g 
2 2

Thus, in molar units, the balance for species is given by

;

; A  l   = = m c
M g  a 

5)

 ,  = *) c l a

This equation is also easily interpreted: Net rate of input of + moles of any component into the CV by average velocity plus by diffusion Rate of generation = Rate of accumulation of moles of that of moles of that component due to component in the CV chemical reactions

360

34.5 Classication of mass transfer problems


Now the discussion can be summarized and mass transfer problems can be classied. I Pure Diffusion problems a) Dilute solutions: Only solute uxes are of interest and for these components, n = = j= and v = 0.

b) Concentrated solutions. Fluxes of all species are of interest, However, v is small and, its value is not determined from equation of motion. Some condition in the problem of diffusion has to give rise to some approximation that allows determination of the bulk velocity. II Convection problems a) Forced convection. An external agency creates motion. Equation of motion is solved to nd the bulk velocity. b) Natural convection. Diffusion causes density gradients which in turn cause motion. Equation of motion is solved to nd the bulk velocity. Convection problems can further be classied into those involving dilute and concentrated solutions. The concentrated solutions involve high mass uxes. In this notes, concentrated solutions are not considered. The high ux mass transfer problems are normally treated by using models and arriving at high ux corrections which can be applied to the corresponding low ux or dilute solution problems. Interested reader is referred to the text by Bird, Stewart and Lightfoot.

361

Chapter 35 SHELL BALANCES IN PURE DIFFUSION PROBLEMS


Analysis of mass transfer problems, involving only diffusion, through shell balances are taken up in this chapter. As mentioned in the earlier chapter, diffusion problems are solved by making an assumption to avoid the solution of equation of motion. Three examples are considered in this chapter, and in several cases, the differences between the low ux (dilute solutions) and high ux (concentrated solutions) cases will be examined to illustrate the issues discussed earlier. Finally, the exact analogy between the low ux pure diffusive mass transfer and heat transfer by conduction is demonstrated.

35.1 Equimolar counter diffusion


35.1.1 Problem statement
A liquid mixture of
and is placed in a tube. Let
be the more volatile component. Let the mole fraction of
in the liquid be B . The mixture is ideal and follows Raoults law. It is also given that their molar latent heats are nearly the same. A vapor mixture of
and is owing on top of the tube. Let the mole fraction of
in the vapor be . The composition of the vapor is different from that of the vapor which would be in equilibrium 1 with the liquid. Let the owing vapor mixture be richer in the more volatile component. Therefore, the more volatile component will diffuse from the vapor into the liquid and less volatile component will diffuse in the opposite direction. The objective of the problem is to calculate the mass ux at steady state2 . The physical arrangement is shown in Figure 35.1.

35.1.2 Simplications and approximations


The problem is very reminiscent of distillation. As diffusive ow rates are small, it will be assumed that the pressure drop required for ow is small, and the pressure in the two phases is equal and constant. The vapor and the liquid are assumed to be at the same temperature 3 . Let
There will be no mass transfer if the two compositions correspond to an equilibrium state. It is possible to reach steady state if a liquid stream is fed to the tube to maintain liquid level in the tube constant, and if its composition and the rate are equal to that of the vapor diffusing out of the tube. 3 As the owing vapor is richer in the more volatile component than the vapor that would be in equilibrium with the liquid, the temperature can be uniform.
2 1

362

N A,z N B,z

Flowing mxiture of vapors of A and B. Mole fraction of A is y AL

L N A,z N B,z

Well mixed liquid. Mole fraction is o x Mixture of A and B being suppplied to keep level constant

Figure 35.1: Schematic for equimolar counter diffusion.

the system be insulated so that heat is not exchanged between it and the surroundings. Even if the pressure drop is small, there would be a boundary layer developing. Further, even if the ow becomes fully developed, radial velocity gradients would exist. This can lead to a composition gradient in the radial direction. However, if the radius of the tube is very small compared to the axial length, it is possible that concentration proles in the radial direction would be evened out since mass cannot diffuse through the tube wall. In view of this, it will be assumed that radial concentration gradients do not exist, and this is tantamount to assuming that diffusive uxes exist only in the axial or T direction. As the pressure and temperature are constant through out the system, energy balance should yield a relationship between the ux of
and . If the latent heats of vaporization of the two components are equal, for each mole of the more volatile component condensing, a mole of the less volatile component can be expected to evaporate. An energy balance boundary condition can be used to derive this. As boundary conditions for simultaneous heat and mass transfer have not yet been spelt out, this result can not be shown as of now. However, an intuitive argument can be given here. As temperature is the same through out the system, conduction heat transfer is absent. There is no heat exchange between the system and the surroundings. ! Hence, energy balance at the interface suggests that
)

M
)

A 

if the sensible heat effects are neglected. But the latent heats are equal. This implies that the net molar ux is zero or the situation corresponds to equimolar counter diffusion. Notice that by using this argument, a relationship between uxes was arrived at, and it is this kind of ! of motion. The condition of equimolar counter equation that allows bypassing the equation 4 diffusion can be written as 

35.1.3 Balance of mass of species


In view of the above relationship, it is clear that the present example is a t case for using molar units. As before a thin shell of thickness T is chosen as the control volume for making
4

Notice that

d 4

and hence it follows that mass average velocity is not zero.

363

a balance of moles of
. Refer to Figure 35.1. Net rate of input of + moles of any component into the CV by average velocity plus by diffusion Rate of generation = Rate of accumulation of moles of that of moles of that component due to component in the CV chemical reactions

As the process is in steady state, the last term is zero. As no chemical reactions are occurring in the !above in the control volume, the second term equation is zero. Thus the above reduces to    Dividing the above equation by

and taking the limit of the equation as

 y A

 A  T

goes to zero

35.1.4 Combining balance with constitutive equation


In order to solve for the concentration proles, we need to substitute the constitutive relationship into the mass balance. It should be noticed that this procedure is similar to what was used in momentum and heat transfer problems, where we made a shell balance (for momentum or heat) and substituted the constitutive relationship to determine (the velocity or temperature) ! any species ! ! the diffusive ux: is the sum of the convective proles. The total ux of and

I  M M A M In this problem, the second term is !zero. The diffusive ux is governed by Ficks law. Hence  A A  T Substituting this in the balance equation K K  A  K c or  K A  (35.1) T T A

Here it was assumed that vapors are assumed to behave like ideal gases, and as temperature and pressure are constant, the total molar density remains constant. It was also assumed that the diffusion coefcient also remains constant. In general it can depend upon composition, though not for ideal gases.

35.1.5 Boundary conditions


The order of presentation is changed slightly from the usual one since scaling can not be done in this problem till boundary conditions are considered. As with heat transfer problems, the ux and or the concentration boundary conditions can be used at the interface. At the top of the tube, the boundary condition is determined by the composition of the owing vapors. If the ow rate of the vapors is large compared to the diffusive ux of vapors from the tube, and if the diffusive stream is assumed to mix quickly 364

The ux of
is what is to be determined and so it is not known, and consideration of the ux continuity boundary condition is of not much use. The same is true at the vapor-liquid boundary as well. The boundary at the vapor liquid boundary is very much akin to phase boundary in heat transfer problems. As ux is unknown, ux continuity condition is of not much help. Thus, one has to look for the condition equivalent to continuity of temperature, and that is the condition of thermodynamic equilibrium. Typical vapor-liquid equilibrium boundary conditions have been discussed in chapter 33. The vapor and the liquid phases are in equilibrium and both phases show ideal behavior. Hence Raoults law can be applied. The mole fraction of
in the well mixed liquid is specied to be B . Hence

with the owing stream, it is reasonable to assume that the composition of the vapors at the top owing stream. Thus, the boundary condition of the tube will not be different from that of the at the top is given by A c T A W (35.2)

where

F

and

F B B

(35.3)

are the saturation vapor pressure of


and the total pressure of the system.

35.1.6 Scaling
The length of the tube forms the scale to nondimensionalize U T . The mole fraction difference can be used to scale the mole fraction: s

A T h A B c v W Kv   K A

The non-dimensional equations become

and the boundary conditions become

v A c

at

A 6

and

v A 6c

at

A 

The solution is

v A 6 from this. The ! ux can be easily calculated U s H  H   B  v A A A B A   V V W W T The ux is in the negative direction since, as stated in the problem, mole fraction of the more volatile in the owing vapor stream is greater than B .

In a binary mixture only one mole fraction is independent. Therefore an independent equation for the other mole fraction, in this case that of , can not be obtained. Thus ux of has to be determined without solving for mole fraction of . But of course, the ux of is equal and opposite in sign to the ux of
. Thus ux of is immediately known. 365

35.1.7 Looking back


The results show that as the diffusion coefcient increases, and the path length decreases, the ux increases. The ux also increases as the difference in the mole fractions at the top and at the vapor-liquid interface increases. These are expected. If the mole fraction at the top and at the interface are equal, the system is in equilibrium and there will be no mass transfer.

The similarity between this problem and the heat conduction in a slab is obvious. The idea of the resistance to diffusion is relevant here and similar to that in the heat transfer. The diffusion ^ . As expected the resistance decreases with increased diffusivity resistance is given by W and increases with an increase in the path length. These conclusions are also entirely analogous to what was learnt in conduction heat transfer. The above problem is a model of the resistance to mass transfer in the vapor phase if W is considered to be the thickness of the gas diffusion lm. This is commonly referred to as the lm model of mass transfer and is familiar to all chemical engineers. The assumption that the liquid is well mixed is equivalent to assuming that there is no resistance to diffusion in the liquid phase. There could be resistance to mass transfer in the liquid phase as well. In a distillation tray, typically both do exist. A model can be developed when diffusional lm resistance exists in both phases, and this is the two lm theory. This would require knowledge of the diffusion lm thickness in the liquid phase. This situation is also encountered when mass transfer takes place through any two immiscible phases placed in series, a situation encountered in liquidliquid systems. The ideas of conduction heat transfer though series of slabs can be extended to calculate ux in such problems. The only difference in such a treatment arises in treating the condition of equilibrium at the boundary of phases. In case of heat transfer, this condition is equivalent to the continuity of temperature. Additionally, the ux continuity was the other condition needed. The ux continuity condition is valid in case of mass transfer as well. The condition of thermodynamic equilibrium is equivalent to equality of chemical potential and not the concentration (or mole fraction) of the same species in the two phases as seen from eq.35.3. The condition of thermodynamic equilibrium dictates a relationship between the concentration phases. In sthis (or mole fraction) of the same ! species in the two example, the ux could be rewritten as H

35.1.8 Similarity to heat transfer

A W

and this form demonstrates the way the mole fraction of the liquid phase can be incorporated. EXERCISE Suppose that the latent heats of vaporization are not equal. Let the ratio of the latent heats be . How does the analysis change?

35.2 Vaporization of solute into an insoluble gas


35.2.1 Problem statement
Consider a liquid
which is placed in a beaker and vaporising into a gas . Let us assume that steady state has been reached. To make matters simple let us assume that the liquid and 366

gas are at the same temperature5 . See Figure 35.2. Liquid A is being supplied to the bottom of the beaker so that the liquid level remains constant. Its rate must be equal to the rate of vaporization, which of course in not known, and the objective of the problem is to determine it.
Pure gas B flowing
N A,z

L N A,z

Liquid A Liquid A being suppplied at a velocity of v to keep level constant

Figure 35.2: Vaporization into a stagnant gas As the liquid exerts its own saturation vapor pressure, the concentration of
will be more at the vapor-liquid interface. In contrast, as pure is owing at the top of the tube, the concentration of
will be zero there. Thus a concentration gradient is created and
will diffuse from the liquid to the top of the tube. We want to calculate the diffusion ux of the vaporizing solute.

35.2.2 Simplications and assumptions


Suppose the whole experiment is being carried out at atmospheric pressure. As the liquid vaporises, it will move through the tube and leave at the top. Clearly mass average velocity is not zero and hence we have to solve the equation of motion to determine it. If we wish to avoid solving the equation of motion for the reasons mentioned earlier, we have to look for some condition to get the average velocity. The vaporization problem is usually solved in the context of an insoluble gas. If the gas is insoluble in the liquid, then the ux of at the gas liquid interface must be zero. In other words, we know some thing about one ux. As we will see later, this condition then allows us to avoid the need to solve for the equation of motion. We expect that the ow created by vaporization is so small6 that there will hardly be any pressure drop due to it. Hence we assume that the pressure in the tube is atmospheric. In problems of vaporization, e.g., humidication, some heat effects are involved. The central theme here is to analyze diffusional phenomena and so, as mentioned in the problem statement, it will be assumed that temperature remains constant. Generally, the vaporization problems are encountered in low pressure environment. In view of this, it will be assumed that the gas phase is ideal. For reasons enumerated in the previous example of equimolar counter diffusion, it will be assumed that diffusion is one dimensional.
This assumption can be relaxed if energy balance is incorporated. This problem is solved in many undergraduate texts and is referred to as vaporization into a stagnant lm. The word stagnant can be interpreted to mean the velocity is small. In this particular problem, as we shall see later, the gas remains stagnant. Reference to is of course exact.
6 5

367

35.2.3 Balance of species

control volume must be zero:

Once again, constancy of pressure and temperature coupled with the assumption of ideal gases  makes use of molar units appropriate in this problem also. Thus we use molar uxes,  and mole fraction . Consider any thin shell of thickness T , as shown in Figure and 35.2. As there are no reactions and we have steady state, the net molar ux of
and into the !  
or

Thus we have that both molar uxes are constant.

 A   8  8 y A  8 y c       A  c and  T A T

35.2.4 Combining balance with constitutive equation


The constitutive relationship is the Ficks law:

T This can used! to for the write an expression total ux !   I A  A   M  M   M T T ! as The above equation can be rewritten   A   6 T M 6 Substituting this into the mass
, we get balance for  H H     6  V A  6 V T T T

 A



(35.4)

where we have assumed the diffusion coefcient to be constant. equation. Thus it requires two boundary The above equation is a second order differential conditions to solve the differential equation, and we shall show later that two are available. The  . We above equation has two more unknowns: and have a thermodynamic equation of state, and since the gas mixture was assumed to be ideal, can be determined if the pressure pressure drop is negligible. Hence the pressure in the tube is known. We have assumed that the is same as in the ambient, . The system was assumed to be isothermal. Hence is known.  Supercially it would appear that can be calculated from an equation for , which in by combining the balance for with the constitutive relationship, turn can be obtained which will not give an equation for independent will be in terms of . However this procedure of the one for , since we know that the sum of mole fractions is equal to unity. Thus the . Thus we are falling short by one equation balance for can not be used to determine or condition. This is because we are not using the equation of motion. Thus we need some condition which allows us to determine the ux of one component or relate the uxes of the two components to avoid solving the equation of motion. 368

7 Another way of understanding this is that a total of three conditions are needed since we  have a second order differential equation, and the unknown . Hence one more condition is required apart from the two boundary conditions.

Two boundary conditions at T A and T A W are obvious. They constitute chemical potential equality and imposed concentration respectively. If we assume that vapor liquid equilibrium is given by Raoults law, 

35.2.5 Boundary conditions

 T A F  and are the saturation vapor pressure of


and total pressure in the system where
respectively. Since pure gas is owing at the top of the tube,

/ F

Bc
at

at

A c

T A W

We need one more condition and we examine the ux continuity conditions at the vapor liquid interface which are yet to be utilized. is insoluble in liquid
, and hence is not present in the liquid phase. Thus ux of coming to the interface from the liquid phase is zero. Hence ux continuity at the interface requires that ux of in the gas phase is zero:

 A  c

at

 T A

This makes sense: ux of in gas phase at the interface must be zero since it is not soluble in liquid and hence can not enter the liquid phase. As discussed earlier, the three boundary conditions are sufcient to nd a solution, and we have them. But, we have not yet utilized the ux balance of
. Does it over determine Continuity of the ux of
at the the system? interface gives PI

P where is the velocity in the liquid phase, is the density of liquid I


, and is the molecular weight of
. As can be seen this equation is useful to determine , or the rate at which the
liquid must be supplied for the level to remain constant, and it does not constitute an additional equation for the gas phase. Notice how insoluble nature of gas has given us one more condition, and thereby allowing us to bypass NSE. This is equivalent to the equimolar ux condition of the last problem.

 A

35.2.6 Mass balance equation


The balance equation for is given by

This implies that ux of is a constant. However, ux of is zero at the liquid-gas phase boundary is zero according to the boundary condition. Hence

    A T

 A  c

for all values of

Yet another way to see this is to note that eq.35.4 is a rst order differential equation involving two uxes, that of and , as unknowns. Hence we require three conditions to solve for the unknowns.

369

The balance equation for


is given by

! 

T ! in equation 35.4, Thus ux of


is a constant. Setting ux of to zero   A   Constant A M T Applying this in particular at T = 0,  A 6 H   6 B V T ! Substituting this result into the right hand side of the constitutive H  relationship, we get  6  A  V  6 M B T T are specied in terms of It is convenient to differentiate it conditions once since the boundary the mole fractions. Thus, the above equation is equivalent to K  6 H   K M 6 B  V  A  T T T used to solve the above second order Note that we have two boundary conditions which can be

A 

differential equation. Thus we did not need to solve for the equation of motion. This problem is similar to heat transfer in a slab whose opposite faces are maintained at two different temperatures. Heat balance in that problem gave only a term like the rst one in the above equation. The above equation is like

Therefore, the second term represents convection caused by diffusion alone. We will discuss this aspect a little later.

 K I    K  A T T

35.2.7 Scaling
The length can obviously scaled with W . The mole fraction can be scaled with v represent the dimensionless variables. Further let

B . Let

larger is . Thus, it is clear that the second term in the mass balance is a correction due to convection. 370

v   A ! related to the ux. Thus It may be noted that is H   A 6 6 B  T V A 3 W velocity. The larger the ux, the is seen basically to be a non-dimensional ux or an average
The mass balance then simplies to

H  A 6 B B vV

9 and

Kv  K

9 ( 9

35.2.8 Solution
Solving the mass balance after implementing the two boundary conditions we get

However

6 Sv A H  B A 6 B 

and substituting our solution into this we get

vV

H A 6 6 c or A 6 6 V B B to be ! is obtained Thus the ux in dimensional terms  A H 6 6 BV W As expected, the ux is in the positive T direction since the mole fraction decreases in the T direction. We also can determine the concentration proles: H 6 H 6 T A 6 6 BV W BV

( u 

It may be noticed that the concentration proles are logarithmic, and not linear. Let us once again return to the analogy we drew with the heat transfer in a slab. If the analogy was perfect, the mole fraction prole should have been linear. The logarithmic terms are therefore an effect of convection caused by diffusion alone.

As there is a mole fraction gradient, also should diffuse, and the natural question to ask is the constitutive about the cause for the ux of to be zero. From relationship,

35.2.9 How is  stagnant?

The bulk  ux term, i.e., the second term on the right hand side, is in the positive T direction since is positive. However, the mole fraction gradient is negative. Thus the diffusive ux is being opposed and exactly being balanced by the bulk ux so that the net ux is zero. EXERCISES We want to illustrate issues regarding the assumptions that allow specication of the ux by the following problem. Consider vaporization of water from an aqueous solution of Na(OH) into a blanket of CO K . COK dissolves into aqueous Na(OH), and the assumption of insolubility can not be used. Suppose we assume that the reaction between dissolved CO K and Na(OH) is instantaneous and that the concentration of the alkali is large so that we need not consider its depletion. Discuss how this problem can be set up. 371

 A

 M 6  T

the problem, we have not made any assumptions regarding the nature of the comIn solving position of the gas mixture. Suppose we have a dilute solution. In this example, this means that B is small. This is satised by relatively non-volatile solutes. In this case we are not interested in what happens to and we are simply interested in calculating the ux of
. The mass balances are unaffected by the nature of the mixture and they still give the same result:   T  A c
and

35.2.10 Dilute solutions

    A T  A

or that both molar uxes are constant. We are not interested in 6 For dilute solutions, since ! , Ficks law is given by

and so we focus on only


.

 A

 M I A

 T

This is the form of Ficks law that is used for dilute solutions. Implicitly we are assuming that bulk velocity is zero. This of course implies that we need not solve the equation of motion. The pressure and molar concentration are constant since there is no bulk velocity. It is easily seen that the mass balance equation reduces to

due to diffusion alone is negligible, the differential mass balance gives an identical equation to that obtained from heat conduction.

( from the more exact analysis. But ( is proportional to the convective velocity. Thus, in dilute solutions, convective velocity is negligible. It then corresponds to low ux situations. Once the convection

Kv   K A We can then integrate the above equation to nd the concentration proles, 6 A T B W ! to be and they are linear. The ux is easily found  A B W  The above differential equation is equivalent to that obtained in the

The ux of
is seen to increase with increased diffusion coefcient, and with increased B or the vapor pressure of the solute. Similarly, the ux decreases as the length of the column increases. These are expected since an increase in vapor pressure reects an increased driving force, while an increase in W increases the resistance to diffusion. Let us examine the effect of the convection induced by diffusion on the ux. The ratio of the ux obtained by exact analysis to that obtained by neglecting convection would reect the effect. The latter is that obtained by assuming the solutions to be dilute or that corresponding to low ux case. The ratio is referred to as correction due to high ux or simply the high 372

35.2.11 Looking back

ux correction factor, . Thus, the actual ux is given by multiplying the ux obtained by ! factor. Hence, assuming the solution to be dilute with the high correction

 A WB

It is easily worked out that

 A

 6

B  A 6 M L B M mnmom B

The ux is more8 than that obtained by dilute solution approximation. Though surprising at rst instance, it is understandable.
can diffuse at the low ux rate corresponding to a linear mole fraction gradient consistent with the imposed boundary conditions. The same gradient will make to diffuse in the opposite direction. Hence,
must diffuse more in order to keep stagnant.

35.2.12 Similarity to conduction heat transfer problems


Once again, consider the problem of heat transfer by conduction across a slab with thickness W and with a temperature difference maintained across it. The heat ux is constant and constancy of mass ux derived earlier is identical to the constancy of heat ux. Ficks law is similar to the Fouriers law of heat conduction, but is not identical to it because the former relates the diffusion ux and not the total ux to the concentration gradients. However for dilute solutions diffusion ux is equal to the mass ux. The problem we considered for dilute solutions then is identical to that of heat transfer by conduction across a slab with thickness W and with a temperature difference maintained across it. It will be seen later that this identity between heat transfer by conduction and mass transfer by diffusion in dilute solutions holds even in convective transport.

35.3 Catalytic reaction


35.3.1 Problem statement
Consider a catalyst on which the following reaction occurs:

can be imagined as a tube whose bottom is a catalytic surface. At the top of the The apparatus tube a mixture of gases
and with a known composition is owing. Let its composition be  . See given by Figure 35.3. The reaction is also a rate process and the rate expression is to surface, in moles per unit area per unit be specied. Let the intrinsic rate of the reaction at the time, be given by
As
reacts at the surface, its concentration there will deplete. Hence it will diffuse toward the catalyst surface. In contrast as is produced at the catalyst surface, its concentration will
This must not be confused with the effect of non-condensibles on condensation rate and other similar phenomena, where the introduction of non-condensibles creates a resistance which would not have been there otherwise. Here we are discussing the effect of diffusion induced convection.
8

L


A j

373

N A,z

A mixture of A and B flowing.

L N A,z

Catalytic surface
Figure 35.3: Reaction occurring on a catalytic surface

increase there and it will diffuse away from the surface. A steady state will be established when the rate of consumption of
matches the rate of diffusion from the bulk. We are interested in calculating the rate of reaction or rate of consumption of
or equivalently, rate of production of . Earlier, we have referred to the lm model. The similarity between the setting of the present problem and the lm model is obvious and intended!

35.3.2 Simplications and assumptions


Based on the discussion in the previous sections, we expect that the ow caused by the reaction at the surface to be small and hence we assume the pressure drop to be negligible. Thus, the pressure can be treated as constant. It will be assumed that temperature remains constant. To keep things simple, it will be assumed that the gas phase is ideal. For reasons enumerated in the previous examples, it will be assumed that diffusion is one dimensional.

35.3.3 Balance of species


There are two components and the reaction is occurring at the catalyst surface. There is no ! bulk reaction. Hence mass balance over a differential element of thickness T is given by:

 

or

Thus, both molar uxes are constant.

 8 y A  c  8  8 A  y       A A  c and  T T

35.3.4 Combining balance with constitutive equation


In order to solve for the concentration proles, we need to substitute the constitutive relationship into the mass balance. The constitutive relationship is the Ficks law,

 A



374

! for the total ux of


is given by and the expressions

where we have assumed the diffusion coefcient to be constant. The balance for after substitution of the constitutive relationship will not give an independent equation since we know that the sum of mole fractions is equal to unity. The above is a second order differential equation and it needs two boundary conditions. We  will guess that the rate of reaction at the catalyst surface and the imposed concentration at T A provide the two boundary conditions. Apart from that, the equation contains two unknowns: and sum of the uxes. Thus two more conditions are required to solve for the concentration proles. As discussed before, we assumed pressure and temperature to be constant and that the gas phase is ideal. Thus, once the pressure and temperature of the system are known, is known. Based on the previous experience, we now look toward boundary conditions to get some information about the sum of uxes.

!   I A  A   M  M   M T T balance for Substituting this into the mass


, we get ! H     V A    M   T T T

(35.5)

35.3.5 Boundary conditions

The composition of the gas at the top is known, and it forms one boundary condition:

T A W  A

= = \  S = ? = 8 \  m 8 M /  = / A  where superscripts and refer to the phases, is the normal  pointing from phase to phase , V is the velocity of the interface between the phases, / = are the stoichiometric coefcient of species in the surface reaction and / is the intrinsic rate of surface reaction. Since the catalytic surface is stationary, V is zero. Further, neither reactants nor products are present in the catalyst phase, and 8 hence all velocities will be zero in that phase. Let the uid phase be vector in T direction. the boundary conditions denoted by . Hence is same as unit Then
 

We do not have a concentration boundary condition at the catalyst surface. However we do have mass ux balances. The general form of the mass ux balance at an interface in presence of a single surface reaction is

simplify to

 L j A  c

and

 M j A 

 A L  This makes physical sense since for each mole of


diffusing toward the catalyst surface, two Now we use the relationship between moles of have to diffuse away to satisfy ! stoichiometry. the uxes to nd their sum    A L M
Thus, once again, by using the relationship between uxes, we managed to bypass the NSE. 375

Thus we have a total of three needed conditions. We can treat the rst one as one boundary condition. It can be seen that the two uxes to each other through the boundary are related conditions by virtue of the stoichiometry:

35.3.6 Mass balance equation


Since the ux of
is not known, it is more convenient to eliminate it from the mass balance. We rewrite eq.35.5, using the relationship between ! uxes as

 A

The second term of the left hand side obviously corresponds to correction due to the velocity created by diffusion. The above is a second order ordinary differential equation and needs two boundary conditions. They are

L A L 6 B T V L  where B is mole fraction of


at T A . Using this, the ux of
can be rewritten as ! H   6  A  B  V L M 6 T T L Substituting this into the mass balance equation and recognizing that is constant since pressure and temperature are constant, we get H    6   M L 6 B  V A  T T T L H  A 6 L B  T V s A c 6  A L j B c  T A c

equation is valid at all values of Since the ux is constant, the Thus, H   6

 M L T

and in particular at

 T A

and

at

at

T A W s
. Let

Thus we have successfully eliminated the need for solving the equation of motion.

35.3.7 Scaling
The length can be scaled using W while the mole s fraction can be scaled with be the corresponding non-dimensional variables. Further, let

9 and v
(35.6)

L H  v 6 LB  V

The mass balance equation then becomes

Kv  K

9 ( 9
376

v   A

where v

The solution will be implicit since the mole fraction at the surface is not known. Let us formally write the solution however. Let us use the following boundary conditions in nondimensional form v A 6 c at A 6 c and v A v B at A 

is the scaled mole fraction of


at the catalyst surface which is to be determined.

35.3.8 Solution
Integrating the equation once, we get

6 s L B H 
A vV A L Hence 6 s L B H  vV A L Integrating once more and using the boundary condition that v found to be 6 s L B 6 v A L

But from the denition of

v  A

denition, i.e., eq.35.6, it follows that

= 1 at

A 6 , the solution is

Thus the concentration proles are nonlinear. Only an implicit expression for the ux can be found as follows. Let us evaluate the solution at = 0: s B

This equation still has , which is however related to from eq.35.6

  B  V A L W A W j T s s or B A j W U for : Substituting this into eq.35.7, we have an implicit equation U H H 6 L 6 M 6 A M L j W V j W V

6 6 B A L L

6
!

(35.7)

WL H A 6 LB

through the rate of reaction. Thus,

377

This equation can be solved iteratively! to nd . The ux is related as seen from above:

We expect to be small for low uxes. Low uxes will occur when is small, i.e., for  dilute solutions. The ux under those circumstances can be found by taking the limit as U of the above implicit equation for . Thus we have

 A L

Thus we nd the ux to be

A L 6 M L j W U ! 6 A W 6 M L j W s

It is however instructive to resolve the problem from the beginning again to see the principles.

35.3.9 Dilute solutions

Suppose we have a dilute solution. In this example, this means that is small. In dilute solutions, we are interested in solving only for the ux of the dilute component, and not interested in calculating the ux of . The mass balances are unaffected by the nature of the mixture and they still give the same result:

or that both molar uxes are constant. We are not interested in 6 For dilute solutions, since ! , Ficks law is given by

 T

A c

and

    A T  A

and so we focus on only


.

 A

 M I A

 T

Notice that the unknown v has been eliminated which is equivalent to knowing one ux. Hence we need not solve the equation of motion. We can assume that pressure is constant and relationship into the mass balance we get hence is constant. Substituting the constitutive

The solution is of course linear. Integrating once

  A Constant T  But from the boundary condition that matches the ux with reaction rate at T A  L j$ B  A T
378

K  K A  T

, we have

Integrating once more and using the boundary s condition at T

or

The ux is given by

A W , we get L j$ B W B A s 6 B A 6 L j$ W M U ! 6 A L j BA W 6 M L j W A 6 W L j M

The answer of course coincides with that derived earlier. s is as follows: An interesting way of recasting the ! above equation

Thus we see that the driving force is the mole fraction at T A W , and there are two resistances to be accounted for. The rst is the resistance due to diffusion. The other is due to the surface reaction. The ux is given by the ratio of driving force to the sum of the resistances.

35.3.10 Looking back


The ux or the reaction rate increases with increasing diffusion coefcient, mole fraction in the feed ( T A W ), reaction rate constant and with decreasing W . All these are as expected.

35.4 Summary and nal comments


Diffusion is similar but not identical to heat conduction. Diffusion alone can generate convection. Such convective velocities however are small. Their effects are negligible for dilute solutions and pure diffusion problems in dilute solutions are identical to heat conduction problems. In general, the effects of the convective velocity can be accounted for by solving the species balance equations along with the equation of motion. However, the small values of the convective velocity allows us to assume in general that pressure drops due to them are negligible. This alone is not enough to bypass the necessity to solve the NSE. Some condition that allows specication of a ux or relationship between uxes is needed to avoid simultaneous solution of the equation of motion. A few examples of this have been illustrated. The ratio of the ux calculated by accounting for the effects of convective ux and that obtained by neglecting them is referred to as high ux correction. The setting of the problems is deliberately made similar to the classical lm model. Thus high ux corrections can be calculated for lm model. Though not dealt with in this notes, lm model is used in this way in practice.

379

Problems for Chapter 35.


35.1 A very thin liquid lm is suddenly exposed to equal volumes of pure gases A and B at pressure B and temperature  . The mass of the lm is negligible, and hence it can not support any pressure difference across it. In other words, it will move instantaneously in order to equalize pressures across it. The Henrys law constants of A and B relating their partial pressures in the gas phase to their concentrations in respectively. Diffusion will cause movement of A and B and the liquid are lm. Assume that the diffusion in the liquid lm can be assumed across the liquid to be the controlling resistance. (What would be the criteria for testing this assump] and ] be the diffusion coefcients of A and B in the liquid lm. tion?) Let a) Will the lm move or not due to the diffusion of A and B? b) What will be the nal position of the lm when equilibrium has been reached? c) What will be the nal position of the liquid lm if gas A is insoluble in the liquid lm? Hint: Make a jump balance at the liquid lm and write an equation for the movement of the lm, and couple this to the mass balance of one of the components.

Gas A

Gas B

Liquid film separating gases

Figure for problem 35.1.

in a large volume of stagnant nitrogen, is evaporat35.2 A drop of octane, suspended ing at constant temperature and pressure. Denote the mole fraction of octane at the vapor-liquid interface by B , the total vapor phase density by and the density of P liquid octane by . a) Assuming steady state, simplify the species conservation equation for octane. What are the relevant boundary conditions? b) From the above, derive an expression for the rate of evaporation of octane. c) Assuming pseudo-steady state, derive an expression for the time required to evap orate a drop whose initial radius is B .
35.3 A tube contains highly volatile liquid C. The bottom of the tube is made of a catalyst where A reacts irreversibly with C. The tube is thermally insulated. The following additional information is available. a) The height of the liquid is W and C is being supplied at such a rate to maintain the liquid level constant. The vapor space above the liquid level is . b) At the top of the tube, pure gas A is owing. A is sparingly soluble in C. The equilibrium relationship between the concentration of A in the liquid phase, 380


, and the mole fraction of A in the vapor phase,

A
, is given by:

c) Since C is in excess, the rate of reaction per unit area of the catalyst is given by:

where

is the Henrys law constant.

The reaction is exothermic and the heat liberated is reacted. d) The latent heat of vaporization (per mole) of C is] )

binary diffusivity by e) Denote the vapor phase ] the liquid phase by .

per mole of A

, and the diffusivity of A in

f) Assume that the product formed does not inuence the diffusion processes. We are interested in calculating the rate of consumption of A at steady state. i) Simplify the diffusion equation for species A in the liquid phase and derive an [   A expression for the ux of A as a function of . T ii) Simplify the diffusion equation for A in the vapor phase. iii) The thermal conductivity of vapors is known to be very small. As it is insulated, the heat losses through the tube are negligible. What is the mechanism through which heat of reaction is being removed? iv) Derive the jump energy balance at the vapor-liquid interface. v) Set up the differential equations and boundary conditions required for solving for the rate of consumption of A. vi) Discuss the effect of heat of reaction and latent heat of vaporization on conversion. 35.4 a) Derive an expression for the effectiveness factor for a spherical porous catalyst pellet for the following reversible reaction:


Assume that the reaction rate expression is rst order with respect to both A and B. Assume that the solutions are dilute. b) Is the catalyst more effective when the reaction mixture is far away from equilibrium or nearer to equilibrium? 35.5 Ultraltration of aqueous solutions of a protein is being carried out using a semipermeable membrane, which is permeable only to water. The protein solutions exhibit osmotic pressure k . Hence ltration is possible only if, the applied pressure gradient across the membrane, , exceeds I k . Under such conditions, the supercial O velocity of water through the membrane, , is given by:

I A % 
381

 k O

Pure A flowing here

H z

Liquid C

Catalyst surface

            

Figure for problem 35.3. where is the permeability of the membrane. In above, k is the value of the osmotic pressure at the weight fraction of protein in the solution adjacent to the membrane. The osmotic pressure is related to the weight fraction of protein in the solution by the relationship:

k A !

In answering the following questions, use lm model, and assume that steady state prevails. ii) The weight fraction of protein in the bulk solution is . Assume that the P densities of protein solutions and water are the same and given by . Simplify the diffusion equation for protein to nd as a function of position in the lm. Specify the boundary conditions. iii) Indicate how i) Sketch the weight fraction proles of protein in the solution side.

can be found.

35.6 A polymer membrane which is permeable to oxygen but impermeable to nitrogen is available. It is desired to use this membrane to make pure oxygen from air. Normally thin membranes are used and hence only ux in the direction perpendicular to the membrane is important. The mass ux through this membrane is given by

B is the difference in the partial pressure of oxygen across the membrane. where is supplied to one side of the membrane. Assume that O Suppose air at pressure air consists of only oxygen and nitrogen with oxygen mole fraction equal to 0.21. Suppose we maintain the pressure on the other side of the membrane at . Use the lm model to derive an equation which can be used to predict the ux of oxygen through the membrane.

A B O

382

Solution side z Membrane Figure for problem 35.5. Water side

Film Pressure is Pd z Air at P f

Membrane

Figure for problem 35.6.

383

35.7 Consider a porous solid with pores of very small size. The pores are coated with a thin lm of solid A which sublimes. A slab of this porous solid is attached on one side to a non-porous, thin and highly conductive slab, which is being maintained at a constant temperature  B . The other side of the porous solid is exposed to a owing stream of gas B. The gas is at a bulk temperature of  and the mole fraction of A in the gas is zero. The solid sublimes into the gas since  B  . We wish to calculate the rate of sublimation of A into the gas stream under steady state conditions.

The saturation vapor pressure of A at temperature  (  B   ) is given by A  . Further where is the total pressure in the system. As the Orange of temperaturesO is small, the total molar concentration can be considered to be constant.

(i) Will there be temperature and mole fraction gradients in the porous space of the solid in directions perpendicular to T ? (See gure). Give reasons for your answer. (ii) Denote the latent heat of sublimation per mole of A by , the heat and mass transfer coefcients between the gas and the porous solid by and j respectively, ] the effective diffusion coefcient of A in porous solid by , and the effective thermal conductivity of the porous solid by j . Derive the differential equations that govern energy and mass balance. Specify the boundary conditions. Derive an expression for the mass ux of A into the gas B.

                                                                    Porous solid z            L                                                               Thin nonporous
plate

Figure for problem 35.7.

35.8 Two immiscible liquid phases I and II are in contact with each other. Phase I is a dilute solution of A while II is a dilute solution of B. A is insoluble in II while B is insoluble in I. However A and B can react with each other and do so at the interface. A+B Products

The rate of this reaction per unit area is given by j where and are concentrations of A and B at the interface. a) Simplify the relevant conservation equations for steady state using lm model. b) Specify the boundary conditions. c) Determine the concentration proles and derive an expression for the rate of consumption of A. 35.9 A stream of inerts containing
as a dilute component is available. A spherical porous catalyst pellet (radius = , area per unit volume = ) is being used to convert reactant
in this stream to product . However,
also gets converted to a side product by a parallel reaction. Both reactions are exothermic and the heat of 384

"!

II

Figure for problem 35.8.


] ] reactions can be taken to be given by rate of formation of and . The and per unit area of catalyst are respectively given by jy and j$ , where depend upon is the concentration of A in the pore space. The rate constants the temperature. The rates of external heat and mass transfer processes are very ] high. Amongst all the components, diffusion coefcient of
( ) is the smallest. Denote the effective thermal conductivity of the pellet by j . Outside the pellet, the temperature is  B and the concentration of
is B . Derive the differential equations to nd the concentration of
in the catalyst pellet at steady state. Specify the boundary conditions required. How can the fraction of
converted to B can be

found?

35.10 The following gas phase isomerization reaction occurs catalytically:

The equilibrium constant, is a function of temperature. The reaction is endother mic and the enthalpy change upon reaction , denoted by , is positive. Hence decreases with decreasing temperature. A mixture of
and is placed between two large parallel plates which are made of catalyst. The distance between the plates is . The isomerization reaction occurs instantaneously at the catalyst surface. One plate is maintained at  B while the other plate is maintained at  g where  B N g . The thermal conductivity, j of the gases can be taken to be equal,
and constant. The diffusivity can also be taken to be constant. Assume that neither forced nor natural convection is present. At steady state, calculate the energy ux at the plate which is at  g . Is the energy ux towards the other plate or away from it?



35.11 Photo-acoustic effect: Consider a closed cell and well insulated cell. At one end, a very thin lm of a volatile liquid absorbs radiation falling on it. However, its temperature is low so that it hardly emits any radiation. If the intensity of radiation falling on is varied in a periodic manner, sound is emitted by the cell. The frequency of the sound is the same as that of the intensity. a) Explain why sound is emitted? b) Derive the jump mass and energy balances at the liquid-vapor interface. 35.12 The following reaction occurs at a catalytic surface: A 385 2B

CA

CB

Thin liquid film

Figure for problem 35.11.

The reaction is endothermic, and the heat of reaction is nearly independent of tem perature is given by . Gas A at  is owing past a catalytic surface at a fast rate. The catalytic surface is at temperature  . We wish to calculate the rate of conversion using lm theory. The rate of the reaction per unit area of the catalyst surface is given by

the rest are constants. a) Apply the lm theory to derive the relevant differential equations for determining the temperature and mole fraction proles. Assume that the molar concentrations are constant even though the temperature is varying. This would be valid if  B  is not large. b) Calculate the heat ux to the catalyst surface. Is it larger than if the reaction is absent? Is it meaningful to speak of either heat or mass transfer controlling the reaction rate? c) Suppose we nd that temperature at which the heat ux is zero. What would be the signicance of this temperature. d) What will be the effect of having an inert gas in the system? Is it meaningful to speak of either heat transfer or mass transfer controlling the reaction rate? e) What would be the changes needed if the wall is adiabatic and   ?

H j ! O  V where is the mole fraction of A,  is the temperature of the catalyst surface, and

Wall Film

Bulk

Figure for problem 35.12. and K are mutually insoluble. A is insoluble in The following reaction occurs at the interface

35.13

 g

K

while B is insoluble in

 g .


M  L
386

and its intrinsic rate is given by

where = stand for the concentrations of   the respective species in the  bulk but  at the interface. is soluble in both g and K . The solubility of in g and K is g and K respectively. Specify all the boundary conditions that may be useful.

I/ /

p ; A j j

1/
V

Channel wall

S1 containing A

Steam

S2 containing B
Channel wall

Problem 1

Problem 2

Problem 3

Figure for problem 35.13.

35.14 Consider a cylindrical catalyst pore of radius tion occurs at the catalyst walls:

and length W . The following reac-

and Stefan Maxwell equations have to be used in solving the problem. They are given by = = As

L K The intrinsic rate of reaction in mole/(sec m ) is given by j  B The mole fraction of


at the pore mouth T A is . The mixtures are not dilute,

=A

, gradients in the radial directions are negligible.

a. Consider steady state has been reached and that only


and are present in the system. Do the following: (i) simplify the relevant species conservation Figure 35.4: Figure for problem 3

b. Suppose an inert is also present in the system and the mole fraction of  A at T is B while that of
is B . What are the equations and boundary conditions needed to calculate the rate of production of in a single pore? 7 marks.
]

equations to derive a differential equation for , (ii)specify the boundary conditions needed to solve the differential equation, (DO NOT SOLVE THE  EQUATION) (iii) relate T to the rate of production of in a single pore. 6 marks

387

Catalyst surface z Catalyst surface


Figure for problem 35.14.

388

Chapter 36 EQUATIONS OF SPECIES MASS BALANCE

389

Problems for Chapter 36.


36.1 A porous spherical particle of radius has been placed in a stream of gas B at a pressure and temperature  . B is present in the pore space but does not adsorb on without changing the surface. Suddenly the composition of the stream is changed, the pressure and the temperature by introducing gas A in to the stream. The composition is now maintained constant at a mole fraction of . Gas A can adsorb total surface area of the particle per unit volume on the surface of the particle. The is . The adsorption equilibrium in the range of interest is linear. The mole fraction in equilibrium with the surface is related to the surface concentration is given by

where is a constant and is the surface concentration in moles per unit area. The rate of adsorption can be assumed to be also linear:

A ! 

amount adsorbed per unit time per unit area

where j is a constant. a) Will the particle reach a state where nothing changes with time? What will be the mole fraction of A in the pore space of the particle? change with b) Will the concentration of A on the surface of the sphere, i.e., p A location in general? Under what conditions can we neglect this change? c) If the mass transfer coefcient between the gas stream and the spherical particle is large, and if the variations of composition on the surface of the sphere are negligible, what will be the boundary conditions needed for determining the concentration proles inside the sphere? d) Write the species balance for component A inside the sphere. Use a pseudo homogeneous model. e) Write the mass balance equation for species A on the surface to determine the  rate of change of at any given position inside the sphere. f) Specify the initial and boundary conditions needed to solve the problem. Neglect the gas phase mass transfer resistance to the surface of the sphere. 6? g) How will the equations simplify if j i

#!

A j !

36.2 Consider a porous spherical catalyst particle of radius suspended in a large volume of a well stirred dilute solution of species A. Solute A diffuses into the catalyst particle and reacts irreversibly to form products. The reaction can be modelled as psuedo-homogeneous and to be rst order with respect to concentration of A. The initial concentration of A in the solution is B . The mass transfer coefcient between the solution and the sphere is iven by j . Find the rate of consumption of A. See 19. I of BSL.

36.3 Solute A present in a solvent decomposes irreversibly by a homogeneous reaction to yield product B:

L

390

and the rate of the above reaction follows rst order kinetics. Compound A also reacts instantaneously on a catalyst surface to yield product C:

A solution containing A at a concentration of B is placed at A between two planar catalytic surfaces separated by distance W . Find the ratio of B to C formed for short contact times. (Refer to 19. E of BSL). 36.4 Consider a membrane containing catalytic sites where the following irreversible reaction occurs:

The reaction can be modelled as psuedo-homogeneous and to be rst order with respect to concentration of A. A solvent stream containing A at a concentration of  ows past one face of the membrane while pure solvent ows past the other face. The mass transfer coefcients for both faces is the same and is denoted by j . If the membrane does not contain any A initially, nd the unsteady concentration proles. (See 19. I of BSL.)

36.5 Hydrogen and oxygen are being fed to a tube in stoichiometric ratio to form water. The average velocity of the stream entering the tube is a . The tube wall is made of a very good catalyst and oxidation of hydrogen is known to proceed irreversibly ] and instantaneously on the catalyst surface. The tube diameter is while its length ^ ] is W . To avoid danger of explosions, a very, very short tube is used, i.e., W 1. Correspondingly, the conversions are small. The ow is laminar. Derive an expression for the bulk concentration of water in the exit stream? 36.6 An insulated and innitely long tube is partly lled with liquid A. The other part of the tube is lled with gas B, insoluble in A. Initially the gas and liquid are not in contact, and are initially at a temperature  . They are brought into contact  A with each other suddenly at . The vapor pressure of A, is dependent on temperature and is given by the Clausius Clapeyron equation: O

where is a constant, is the heat of vaporization and is the universal gas constant. a) Write appropriate balance equations needed to obtain the rate of evaporation of A. Assume that the effect of the fall in the liquid level can be neglected. b) Specify the necessary boundary and initial conditions. c) Sketch the proles of the mole fraction of A in the gas phase and the temperature proles in both the phases. 36.7 A solid sphere of radius is placed in a uniform ow. Let us say that the uniform velocity is in the T direction and is equal to a . The solubility of the solid in the uid is very small. Let it be . The uid far away from the sphere does not contain any solute and hence the sphere slowly dissolves into the uid. Assuming that the radius of the sphere is constant and that its radius remains constant at , (i) specify a convenient coordinate system that you will use to solve the problem, (ii) specify the nonzero velocities, (iii) specify the coordinates on which concentration depends. (iv) simplify the mass balance equation, (v) specify the boundary conditions needed for solving the mass balance equation.

  A H O O  V

391

Gas B

Gas A

Figure for problem 36.6. 36.8 A spherical drop of hexane containing acetic acid (
) is rising at its terminal velocity in an aqueous solution of NaOH ( ).
is soluble in both hexane and water. The following irreversible reaction can occur:

are insoluble in hexane, and it can be assumed that water and hexane are mutually insoluble. We are interested in calculating the rate of loss of acetic acid from the drop. It is convenient to solve this problem from a frame located at the centre of the drop and moving with the terminal velocity of the drop. If spherical coordinates are used in such a frame, it is clear that only p and v components of the velocity are not zero. Note that solutions are dilute. i) Simplify the relevant equations of conservation of mass of species ii) Specify the boundary conditions for all the species. The concentrations of acetic acid in water and hexane at equilibrium are linearly proportional to each other. iii) We want to investigate an important limiting case: alkali is in excess. How do the boundary conditions change if j for this case?

M M K F and the rate of reaction is given by js . and

392

Chapter 37 DIFFUSION BOUNDARY LAYER

393

Problems for Chapter 37.


37.1 A plate of benzoic acid is embedded ush in a metal plate and a dilute solution of NaOH ows over it as shown in the gure. The reaction between benzoic acid and NaOH can be considered to be instantaneous. Develop the relevant equations and boundary conditions needed to nd the local value of the ux of benzoic acid.
y

V 8

Soluble solid

Figure for problem 37.1.

37.2 A large disc is rotating in an innitely large extent of liquid. The disc is made of a sparingly soluble solute which can dissolve into the liquid. The solubility is given by . The ow patterns in this ow have been discussed in the class, and it is known that the following differential equation describes the concentration proles in the uid: K

is the T velocity of the uid. A very good approximation of the T velocity in K   ^ B the momentum boundary layer is given by a T where is the momentum boundary layer thickness. We want to derive an expression for the mass transfer coefcient for the dissolution of the solute by integral methods. i) Derive an integral balance equation to determine the concentration boundary layer thickness.  Assume that the solute concentration far away from the disc 6 is zero, and that . ii) We have to assume a prole for concentration to solve the above equation. K  6 ^ A Which of the following is better? T or A 6 T ^  where is the concentration boundary layer thickness. iii) Assume the linear prole and derive an expression for ] B coefcient in terms of a , , and .

T 4 4 where T is the distance along the axis perpendicular to the disc measured from the 4 4 ] disc surface, is the diffusion coefcient, is the concentration of the solute, and I

T A ]

/g

, and the mass transfer

/g

394

Chapter 38 ADVANCED TOPICS IN MASS TRANSPORT

395

Problems for Chapter 38.


38.1 Liquid (component 1) is vaporizing into a mixture of gases (components 2 and 3). The gases are insoluble in the liquid. An equimolar mixture of the gases is owing past the top of the tube at T A W . Derive an expression for the ux of component 1 using Stefan-Maxwell equations.

Gases 2 and 3 flowing

L
z

Liquid 1
Figure for problem 38.1.

38.2 Consider a cylindrical pore on whose walls the following irreversible catalytic reaction occurs:

A , where is the At the pore mouth, T A , pure A ows at a fast rate, i.e., mole fraction of A in the mixture of gases. The intrinsic surface reaction rate per unit area of the surface is given by j . Derive the differential equation and the boundary conditions needed to compute the steady state rate at which A is being ] consumed. Assume that the pore diameter is very small compared to its length W . Calculate the rate at which A is being consumed.

3A

Figure for problem 38.2.

38.3 A mixture of benzene and toluene is evaporating from a lm 7 into a stagnant nitrogen atmosphere. The temperature of the liquid lm is 90 . The total pressure is 1 atm. The vapors7 condense at a cold wall and form a lm. The cold wall is being maintained at 0 . The partial pressures exerted by benzene and toluene in the liquid lm on the hot wall are 0.6 and 0.35 atm respectively. The vapor pressures 396

of both benzene and toluene at 0 are nearly equal to zero. Assume steady state prevails. a) Specify the species conservation equations. How many boundary conditions are needed? Specify them. Using Ficks law with effective binary diffusivity, = A ] = = M = , derive expressions for the mass uxes of benzene and toluene. b) What is the composition of the liquid condensing on the cold wall?

T = 90 C z

T=0C

Vaporizing Liquid

Condensate

Figure for problem 38.3.

38.4 A mixture of volatile liquids A and B is placed in a beaker. They form ideal solutions. The mole fraction of A in the liquid is . Liquid is a fed to the beaker and both its level and composition are being maintained constant. Above the liquid, the beaker was lled with an inert gas C which is insoluble in the liquid mixture. The distance between the liquid level and the top of the beaker is . Pure C is being owed past the top of the beaker. As a result, A and B vaporize into the gas stream. This stream is passed through a condenser to remove A and B. a) Using Stefan Maxwell equations, derive an expression for the composition of the condensate. b) Is there a possibility that the liquid composition and the condensate composition are the same exhibiting an azeotrope like behavior even though the solutions are ideal? 38.5 a) Consider the following reaction occurring at a catalytic wall:

M
Assume steady state prevails. How many independent uxes are there? Simplify the Stefan Maxwell equations for this case assuming that diffusion occurs in only z direction. b) It is desired that lm model be used to solve the problem. Suppose that both the mole fractions of
and at the edge of the lm are equal to 0.5. Suppose that the reaction occurs instantaneously at the catalytic wall. How many boundary conditions are needed to determine the rate of production of per unit area of the catalyst? Specify them. 397

Catalyst wall z
Figure for problem 38.5.

bulk

38.6 Silicon dioxide ( iF K ) is deposited on quartz in a process of making optical bres.  The deposition is achieved by owing an equimolar mixture of gases and K F past the hot quartz surface. The following reaction occurs at the hot quartz surface:   where the iF K formed is in solid phase while the rest are in gaseous form. We want to calculate the rate of formation of the layer of iF K under steady state =A conditions using lm model and Ficks law with effective binary diffusivity: ] = = M = . is the total molar  concentration and can be assumed to be constant. The rate of formation of F K at the quartz surface is given by j$ =  Z  where =  , Z  represent mole fraction of  E and F K in the gas phase adjacent to the hot quartz surface. (i) Simplify the relevant mass conservation equations and substitute the Ficks law into them. (ii) What are the relevant boundary conditions?  (iii) Solve the equations to nd the rate of increase in the thickness of F K layer if P its molar density in the solid phase is = .

iF K M L K

Si Cl 4

O2

= 0.5

SiO

layer

Quartz

Figure for problem 38.6.

38.7 The following irreversible reaction occurring at a catalyst surface is being used to produce C from A and B. 398

A+B

4C

However, the mixture of A and B also contains an impurity D, which also reacts irreversibly at the catalyst surface to produce an undesirable product E:

The rate of these reactions per unit area of the catalyst surface in the order listed fractions above are given by j g and j K where , and are mole of A, B and D at the interface, respectively. A mixture of A, B and D is owing past the catalyst at a large ow rate, and its composition is given by = 0.45, = 0.45 and = 0.1. We wish to use lm model to nd the steady state rates of production of C and E. a) There are ve uxes, one for each of the ve species. Of them, how many are independent? b) Use the effective diffusivity model

A+D

$!

#!

c) Specify the boundary conditions. d) Solve the equations to derive the necessary expressions to obtain the rate of production of D and E.
x = 0.45 A Catalyst Surface x = 0.45 B x = 0.1 C

i ] = = M = % to derive equations for determining = as a function of position in the lm.

Figure for problem 38.7. 38.8 Consider a dilute solution of A ( ) in water (B). We are trying freeze B from  a solution of concentration by bringing the solution in contact with plane cool of B,  , is lowered due to the wall being maintained at  . The melting point presence of A. The following equation gives the relation between the depression in the freezing point and the concentration of A, :

be the latent heat of melting. ? a) Will pure solid B formed be at a temperature higher or lower than j3 b) Use lm model and pseudo-steady state assumption to determine the concentration and temperature proles, and calculate the rate of formation of ice per unit area of the solid from the solution containing A at a bulk concentration of and at a temperature  . c) Is this process limited by diffusion of water, salt or heat removal from solution or heat removal through ice?

Let

A ij

399

Solid B

Solution, C = 10 A

Figure for problem 38.8.

38.9 Saturated steam at 100 C is condensing at a constant wall maintained at 90 C. The K heat transfer transfer coefcient corresponding to the liquid lm is 1000 k cal/(m hr C). a) What is the condensation rate per unit area if the latent heat of condensation is 10 kcal/kmol? b) Look at example 18.5-1 of Bird, Stewart and Lightfoot. Choose liquid at  B as the reference, instead of vapor at  B chosen in the book. How does eq. (18.5-7) change? c) Verify if eq. (18.5-8) is still valid. d) Consider condensation in the presence of a small amount of air. The temperature of the vapor phase is still 100 C while the mole fraction of air in the vapor is 0.04. What should be the wall temperature to air in the vapor is 0.04. What should be the wall temperature to maintain the same rate of condensation as in part a)? e) Can you give an approximate estimate of the decrease in the condensation rate compared to part a) if the wall temperature was maintained at 90 C? The following data will be of use: K Diffusion coefcient = 0.018 m /hr, lm thickness = 10 m, concentration = 3 X K 10 kmol/m , k = 0.025 kcal/(m hr C), c = 10 kcal/(kmol C). Vapor pressure of by about 35 mm Hg per degree water decreases C near 100 C.

38.10 Data obtained on melting of small ice spheres of radius in a stirred vessel of pure L ^ j is equal to 2. Similar ice spheres water indicate that the Nusselt number are now being melted in a stirred vessel in the same way but using 5% by weight 7 aqueous solution of  NaCl at -1.0 C. The following data on the solution are available. K j = 1350 X 10 cal/(sec cm C), viscosity is 1.8 cp, ] = 0.67 X 10 cm /sec, melting density of ice is 0.91 g/cc, and enthalpy of fusion of ice is 80 cal/g. The 7 point in C data for aqeuous solution of NaCl can be approximated by -0.6 X weight percent.  L a) Estimate the temperature at the surface of ice spheres of size m < assuming that the concentration and temperature of brine does not change? b) Calculate the time required for melting of ice spheres of initial size 0.2 cm.

38.11 Saturated steam at 100 C is condensing at a constant wall maintained at 90 C. The K heat transfer transfer coefcient corresponding to the liquid lm is 1000 k cal/(m hr C). a) What is the condensation rate per unit area if the latent heat of condensation is 10 kcal/kmol? b) Look at example 18.5-1 of Bird, Stewart and Lightfoot. Choose liquid at  B as

400

the reference, instead of vapor at  B chosen in the book. How does eq. (18.5-7) change? c) Verify if eq. (18.5-8) is still valid. d) Consider condensation in the presence of a small amount of air. The temperature of the vapor phase is still 100 C while the mole fraction of air in the vapor is 0.04. What should be the wall temperature to maintain the same rate of condensation as in part a)? e) Can you give an approximate estimate of the decrease in the condensation rate compared to part a) if the wall temperature was maintained at 90 C? The following data will be of use: K Diffusion coefcient = 0.018 m /hr, lm thickness = 10 m, concentration = 3 X K 10 kmol/m , k = 0.025 kcal/(m hr C), c = 10 kcal/(kmol C). Vapor pressure of by about 35 mm Hg per degree water decreases C near 100 C.

38.12 A thin layer of coke has been deposited on the surface of a spherical catalyst par] ticle of diameter . The coke is being burnt by passing hot gaseous oxygen. While it is desirable to have a very high burning rate by employing high temperatures and pure oxygen, if the temperature exceeds a certain value, the catalyst sinters and such conditions have to be avoided. Thus, we wish to calculate the burning rate of coke as a function of temperature and composition of the gas using the lm model. The bulk temperature of the gas is  , and mole fraction of O K in the bulk is Z . The reaction occurs only at the surface and the rate of burning reaction per unit area is given by: H

gas mixture. a) Simplify the energy equation to determine the unsteady temperature proles in the particle. Specify the boundary conditions. b) Determine the temperature proles in the solid under pseudo-steady conditions. c) Simplify the species conservation equation under pseudo-steady conditions in the gas phase assuming that the physical properties are constant and that the total molar concentration is constant. Derive expressions for the temperature and concentration proles in the gas phase. d) Derive an expression for the burning rate. e) When the coke on the catalyst particle is burnt in pure oxygen, the reaction rates are high and hence the temperature of the catalyst particle can reach high values resulting in sintering of the catalyst. To avoid this, a mixture of nitrogen and oxygen are used in place of pure oxygen. Rework the problem using Stefan-Maxwell equations. Denote the physical properties by subscript f and of the solid by subscript s, e.g, P c P cGj cGj etc. The heat transfer coefcient between the uid and the solid is

 V Z where
c and are constants, and Z is the mole fraction of O K at the surface. Denote the heat of combustion of coke by . The particle is initially at a uniform temperature of  B before being exposed to the

1 / 1 A/ 1 j given by = .

401

Vous aimerez peut-être aussi