Vous êtes sur la page 1sur 136

Cosserat Continuum Mechanics, I.

Vardoulakis 2009

3rd National Meeting on Generalized Continuum Theories and Applications Thessaloniki February 13 and 14, 2009

LECTURE NOTES ON COSSERAT CONTINUUM MECHANICS WITH APPLICATION TO THE MECHANICS OF GRANULAR MEDIA
Ioannis Vardoulakis, N.T.U.A.

MEDIGRA (EU Programme Ideas, ERC-2008-AdG 228051)

January 2009

Cosserat Continuum Mechanics, I. Vardoulakis 2009

COSSERAT CONTINUUM MECHANICS, 2009. Lecture Notes by Ioannis Vardoulakis, Dr-Ing., Professor of Mechanics at N.T.U. Athens, Greece, P.O. box 144, Paiania Gr-19002, http://geolab.mechan.ntua.gr/, I.Vardoulakis@mechan.ntua.gr

Cosserat Continuum Mechanics, I. Vardoulakis 2009

Table of Contents
LECTURE NOTES ON COSSERAT CONTINUUM MECHANICS WITH APPLICATION TO THE MECHANICS OF GRANULAR MEDIA ............................................................................. 1 Preface ............................................................................................................................................. 5 1 Introduction ............................................................................................................................ 6 2 Mathematical preliminaries .................................................................................................... 7 2.1 Line integrals ................................................................................................................. 7 2.2 The transport law of a von Mises motor................................................................... 12 2.3 Rigid body motion ....................................................................................................... 14 2.4 General rigid-body motion........................................................................................... 16 3 Cosserat continuum kinematics ............................................................................................ 20 3.1 Description of Cosserat kinematics in curvilinear coordinates.................................... 20 3.2 Integrability conditions and compatibility equations ................................................... 22 3.3 Compatibility equations in Cartesian coordinates........................................................ 24 3.4 Strain, spin, curvature and torsion in Cartesian coordinates ........................................ 28 3.5 2D Cosserat kinematics in Cartesian coordinates ........................................................ 30 3.6 Exercise: Deformation of Cosserat continuum in special curvilinear coordinates ...... 36 3.6.1 Polar cylindrical coordinates ................................................................................... 36 3.6.2 Polar spherical coordinates ...................................................................................... 38 4 Cosserat continuum statics ................................................................................................... 40 4.1 The virtual work equation ............................................................................................ 40 4.2 Equilibrium equations in curvilinear coordinates ........................................................ 45 4.3 Equilibrium equations in Cartesian coordinates .......................................................... 47 4.4 The Mohr circle of non-symmetric stress in 2D .......................................................... 48 4.5 Exercise: Differential equilibrium equations in special curvilinear coordinates ......... 55 4.5.1 Polar cylindrical coordinates ................................................................................... 55 4.5.2 Polar spherical coordinates ...................................................................................... 55 5 Cosserat continuum dynamics .............................................................................................. 56 5.1 Balance of mass ........................................................................................................... 56 5.2 Balance of linear momentum ....................................................................................... 57 5.3 Balance of angular momentum .................................................................................... 58 5.4 The micro-morhic continuum interpretation ................................................................ 60 5.5 Exercise: Balance of angular momentum in curvilinear coordinates........................... 64 5.6 Exercise: Dynamic equations in plane polar coordinates............................................. 65 5.7 Stress power in micro-morhic media ........................................................................... 66 6 Cosserat continuum energetics ............................................................................................. 67 6.1 Energy balance equation .............................................................................................. 67 6.2 Entropy balance............................................................................................................ 72 6.3 Linear, isotropic Cosserat elasticity theory.................................................................. 72 6.4 A 2D linear, isotropic Cosserat- elasticity theory ........................................................ 75 6.5 Examples of simple Cosserat elasticity b.v. problems ................................................. 76 6.5.1. Pure bending of a Cosserat-elastic beam............................................................. 76 6.5.2. Annular shear of a cylindrical hole in Cosserat-elastic solid .............................. 79 6.5.3. Sphere under uniform radial torsion.................................................................... 87 6.6 Cosserat thermo-elasto-plasticity ................................................................................. 90 7 Micromechanics of solid granular materials......................................................................... 94 7.1 Stress and couple stress in granular media................................................................... 94 7.1.1 Definitions ............................................................................................................... 95 7.1.2 The virtual work equation for a discrete assembly of particles in contact............... 97 7.1.3 Compatibility in the discrete setting ........................................................................ 98

Cosserat Continuum Mechanics, I. Vardoulakis 2009

7.1.4 Remark on incompatible deformations in granular media..................................... 102 7.1.5 Example: Buckling of rigid-plastic, frictional hinged mechanism ........................ 104 7.1.6 Equilibrium conditions for compatible virtual kinematics .................................... 105 7.1.7 A micromechanical definition of average stress and couple stress........................ 107 7.1.8 Example: Computation of the Love stress in a regular hinged lattice ................... 110 7.2 Mass and moment of inertia considerations............................................................... 113 7.3 Grain scale energy dissipation considerations ........................................................... 114 7.4 The 2-grain circuit of homothetically rotating grains ................................................ 116 7.5 Statistical averaging ................................................................................................... 122 8 References .......................................................................................................................... 129 9 Appendix: The meaning of the Lode angle in Boltzmann Continuum Mechanics............. 134

Cosserat Continuum Mechanics, I. Vardoulakis 2009

Preface
These Lectures were prepared for and presented at the 3rd National Meeting on Generalized Continuum Theories and Applications in Thessaloniki, February 13 and 14, 2009. For the presentation of the material at hand we use the vectorial and the indicial tensorial notation, and fixed in space Cartesian- or curvilinear coordinate description. For easy reading of this material a basic course in Continuum Mechanics is considered as a prerequisite1. For easy reference, some concepts and definitions that derive from basic Analysis and Mechanics are summarized in sect. 2. In the first part of these Lecture Notes (sects. 13 to 6) we compile the basic results that pertain to the mechanics of infinitesimal deformations of Cosserat Continua. In the second part (sect. 7) and in the light of micromechanical considerations the general Cosserat Continuum Theory is applied to the Mechanics of Solid Granular Media. Some of the material presented is published here for the first time and the author would appreciate any comments and critique. The support of the EU project MEDIGRA (ERC-2008-AdG 228051) and the help of my co-worker Dr. Stefanos-Aldo Papanicolopulos are acknowledged.

http://geolab.mechan.ntua.gr/teaching/lectnotes.html#postcm

Cosserat Continuum Mechanics, I. Vardoulakis 2009

Introduction

There is an ongoing discussion concerning the origins of generalized continuum theories such as the Cosserat Theory. On p. 2 of these Notes the reader can find the learned opinion on the subject put on paper in 1966 by late Maurice Antoine Biot. As precursors of the Cosserat theory are mentioned in the literature the theory of Lord Kelvin, concerning the light-aether and the works of W. Voight on the physics of crystalic matter. A nice historical note on the subject can be found in the introduction of the CISM Lecture on Polar Continua by Rastko Stojanovi2. Classical continuum mechanics is based on the axiom that the stress tensor is symmetric. According to Schaefer [43] it is Hamel [26] who has named this statement the Boltzmann axiom, since it is Boltzmann who has pointed first already in the year 1899 to the fact that the assumption about the symmetry of the stress tensor has an axiomatic character. Thus the continuum mechanics of media with non-symmetric stress tensor may be termed also as non-Boltzmann continuum mechanics. Such a theory is the theory of the Cosserat Continuum, that derives from the seminal work of the brothers Eugne and Franois Cosserat [14]. A three dimensional Boltzmann continuum is a continuous manifold of material points that possess 3 degrees of freedom (dofs) of displacement. As already pointed the Boltzmann continuum is juxtaposed to the Cosserat continuum that is in turn a manifold of oriented particles, called tridres rigides or rigid crosses, with 6 dofs, namely 3 dofs of displacement and 3 dofs of rotation. This property is the reason why Cosserat continua are also called polar continua. For example the Timoshenko beam is a typical example of an one-dimensional Cosserat continuum [23],[47],[54][54]. The Bernoulli-Euler beam is seen as a special case of a one-dimensional Cosserat continuum for which the rotation of the material point (i.e. of its cross-section) is related to its displacement through the well known orthogonality condition of J. Bernoulli. Such a continuum is called a restricted Cosserat continuum [34].

Stojanovi, R., Recent developments in the theory of polar continua, CISM, No. 27, Springer, 1970.

Cosserat Continuum Mechanics, I. Vardoulakis 2009

Mathematical preliminaries

2.1 Line integrals The following summary on the basic properties of line integration is taken from the reference book on Tensor Analysis by Duschek & Hochrainer [18]. Let A( xi ) , i = 1, 2,3 (2.1)

be a continuous function of the Cartesian coordinates of point P ( xi ) . Let also () be a piecewise smooth Jordan curve. The expression,

Ii =

()

A ( x ) dx
k

(2.2)

is called a line integral. If the curve is closed then we write,


Ii =

()

A ( x ) dx
k

(2.3)

The above introduced line integral can be transformed to an ordinary (Riemann) Integral if we select a parametric description of the considered curve, say () : xi = i (u ) u [u1 , u2 ] and evaluate the definition eq. (2.2)
Ii =
u2

(2.4)

u1

d A ( ( u ) ) du
k

du

(2.5)

The selection of the parameter u is irrelevant, since the transformation


u = (t ) du = d dt dt

(2.6)

leads to () : xi = i (t ) t [t1 , t2 ] and


I i = A k ( ( t ) )
t1 t2
2 d i d d dt = A ( k ( t ) ) i dt du dt dt t1

(2.7)

(2.8)

where the integration limits are derived from the equations

( t1 ) u1 = 0 ( t 2 ) u2 = 0

(2.9)

Cosserat Continuum Mechanics, I. Vardoulakis 2009

The representation of the line integral given above by eq. (2.5) includes the case of closed-curve line integral, if we assume that

i (u2 ) = i (u1 )
consists of a number of consecutive smooth segments,
() = (1 ) ( 2 ) ( )
( ) : xi = i (u ) u [u1 , u2 ] , = 1

(2.10)

It is obvious that this definition of the line integral includes the case where the line ()

(2.11)

i (u2 ) = i (u1 ) , = 1 1
+1 +1

since Ii =
()

A ( x ) dx
k

= Ii =

( 1 )

A ( x ) dx + A ( x ) dx
k i k ( 2 )

( )

A ( x ) dx
k

(2.12)

Note that a special case of such a curve is a polygonal line, consisting of straight line segments (Figure 2-1).

Figure 2-1: A polygonal curve consisting of consecutive straight-line segments From the definition of the line integral follows also that if A( xk ) is a scalar, then I i is a vector. To prove this let us consider the coordinate transformation

xr = arm xm

dxr = arm dxm

(2.13)

and assume that

A( xr ) = A( arm xm ) = A( xm ) then

(2.14)

Cosserat Continuum Mechanics, I. Vardoulakis 2009

Ii =

()

A ( x ) dx = A ( x ) a
k i k
()

im

dxm =aim

()

A ( x ) dx
k

=aim I m

(2.15)

i.e. it transforms like a vector. Similarly we prove that the line integral of a vector I ij =
()

A ( x ) dx
j k

(2.16)

is a 2nd order tensor, and so on. A special case of the line integral, eq. (2.1) arises, if we select A = 1 I i = dxi = i (u2 ) i (u1 )
u1 u2

(2.17)

In that case the vector I i is the vector that connects the starting point (1) with the endpoint (2) on the considered curve and coincides thus with the oriented secant of that curve through these points (Figure 2-2).

Figure 2-2: Oriented secant on a curve, passing through points (1) and (2) The integral representation of the secant (12) , eq. (2.17), is not to be confused with the integral, s=
u2

u1

xixidu , xi =

d i du

(2.18)

that computes the arc length between points (1) and (2). Let us consider the tensor I ij , eq. (2.16). With indices contraction we may define the scalar,

10

Cosserat Continuum Mechanics, I. Vardoulakis 2009

I = I ii =

()

A dx
i

(2.19)

In continuum mechanics applications this is the most commonly appearing line integral. Thus some authors will call eq. (2.19) a line integral of the vector Ai . For example, if a force fi applies on a particle that moves along a curve () , then the work done by this force during the passage of the particle from point (1) to point (2) along this path is given by the line integral, W=

()

fi dxi

(2.20)

The sign of a line integral depends on the orientation of the curve () , since it changes sign if we change the orientation,
u2

u1

Adxi = Adxi
u2

u1

(2.21)

We consider now two curves


(1 ) : xi = i(1) (u ) u [u1 , u2 ] ( 2 ) : xi = i(2) (v) u [v1 , v2 ]

(2.22)

that are having the same start- and end-points (Figure 2-3),

i(1) (u1 ) = i(2) (v1 ) i(1) (u2 ) = i(2) (v2 )

(2.23)

Figure 2-3: Two curve segments with common end-points In general the two line integrals I i(1) =
( 1 )

Adxi

, I i(2) =

( 2 )

Adxi

(2.24)

will have different values.

Cosserat Continuum Mechanics, I. Vardoulakis 2009

11

If we change the orientation of one of the two curves, say of ( 2 ) we may compute the line integral along the closed loop (1, 2,1) :

( 1 )

Adxi +

( 2 )

Adxi =

( 1 )

Adxi

( 2 )

Adxi =

()

Adx

, = 1 ( 2 )

(2.25)

If the value of the line integral does not depend on the integration path, then for the two considered curves that possess the same end-points we have that,

( 1 )

Adxi =

( 2 )

Adxi

(2.26)

and that the closed path integral, eq. (2.25), vanishes,


()

Adx

=0

(2.27)

The converse is also true, since any partition of the closed loop ( ) , would lead from eq. (2.27) to (2.26). We return now to the line integral, eq. (2.19), and ask the question when the integral

I=

()

A dx
i

(2.28)

is path independent. It is evident that this true, if the vector field Ai derives as the gradient of a scalar (called the potential), i.e. if

Ai = iU
Indeed for
U ( xi ) = U ( i (u ) ) = (u )

(2.29)

(2.30)

we have
2 2 U d i d I = iUdxi = du = du = dU = U 2 U1 xi du du u1 u1 U1 ( )

u2

(2.31)

This means that the value of the line integral I depends only on the values of the underlying potential function at the end points of the connecting curve. It is thus path independent.

12

Cosserat Continuum Mechanics, I. Vardoulakis 2009

In textbooks of Tensor Analysis we find finally the following fundamental,


Theorem

Necessary and sufficient condition for a vector field to be the gradient of potential scalar function is that its rotor vanishes,

Ai = iU

ijk j Ak = 0

(2.32)

This means that sufficient and necessary for the path independence of the line integral (2.28) is the condition

rot A = 0 ijk j Ak = 0 2.2 The transport law of a von Mises motor3

(2.33)

Figure 2-4: Equipollent reductions of a system of forces


Let us consider a system of forces F1 , that are acting on a rigid body. These forces can be reduced by their resultant force F that is acting at a point P . We denote this setting by F ( P) . This force F ( P) can be replaced by a force, that is passing through another point O1 , denoted as F (O1 ) , and a moment, denoted as M (O1 ) The force F (O1 ) arises through parallel translation of the force F and the moment of the pair of forces

{F ( P), F (O )} . The moment M (O ) is a free vector, computed as


1 1
3

Schaeffer, H. (1968). The basic affine connection in a Cosserat Continuum, In: Mechanics of Generalized Continua, Springer, p. 57-62.

Cosserat Continuum Mechanics, I. Vardoulakis 2009

13

M (O1 ) = O1 P F ( P )

(2.34)

The point O1 is called a reduction point. We notice that for all reduction points P ( ) along the straight line ( ) , that is parallel to F and passes through point P , the moment

M ( P) = 0 We may select another point O2 with,


M (O2 ) = O2 P F ( P) = O2O1 + O1 P F ( P ) = O2O1 F (O1 ) + M (O1 )

(2.35)

= M (O1 ) F (O1 ) O2O1 = M (O1 ) + F (O1 ) O1O2 = M (O1 ) + F (O1 ) OO2 OO1 (2.36)

or M (O2 ) = M (O1 ) + F O2 O1

(2.37)

In summary, we have above two reductions that are called equivalent or equipollent, when the following transport law is holding,
F (O2 ) = F (O1 )

M (O2 ) = M (O1 ) + F O2 O1

(2.38)

The compound of the two vectors


F = M

(2.39)

is called a v. Mises motor4, if these vectors obey the transport law, eq. (2.38). The motor eq. (2.39) in particular is a dynamic motor5. The same is the case with the velocity field of a rigid body. Let w be the angular velocity and v the displacement velocity. From two points A and B we have the same transport law,

Mises, R.v. (1924). Motorrechnung, ein neues Hilfsmittel der Mechanik, and Anwendungen der Motorrechnung, ZAMM, 4, 155-181, 193-213. 5 , Greek for dynamic action.

14

Cosserat Continuum Mechanics, I. Vardoulakis 2009

w( B) = w( A) v ( B) = v ( A) + w B A

(2.40)

The corresponding motor is the vector pair, w = v The motor eq. (2.41) in particular is a kinematic motor6. We remark that in the motor space V 6 we can define the power,
F w A = = := F v + M w = F i vi + M i wi M v

(2.41)

(2.42)

It can be shown that power A = is independent of the choice of the reduction point!
2.3 Rigid body motion

We consider Cartesian coordinates. The affine mapping 7 of the points of a body xi = ci (t ) + Rij (t ) j body is called a rigid body motion. Let two such points A(i ) B( i ) , and let
xi = ci (t ) + Rij (t ) j yi = ci (t ) + Rij (t ) j

(2.43)

that has the property to keep the distance constant between two arbitrary points of that

(2.44)

From the above definition we get that, ( xi yi )( xi yi ) = (i i )(i i ) where xi yi = Rij ( j j ) thus (2.46) (2.45)

6 7

, Greek for spiral motion. Lat. affinitas=neighboring

Cosserat Continuum Mechanics, I. Vardoulakis 2009

15

Rij Rik ( j j )( k k ) = jk ( j j )( k k ) Rij Rik = jk

(2.47)

We recall that a square matrix [ R] is called orthogonal, if


det [ R ] = 1

(2.48)

For an orthogonal matrix we have the following relations

[ R ]T [ R ] = [ R ][ R ]T = [ I ]
where [ I ] is the unit matrix, and

(2.49)

[ R ]T = [ R ]1
Or in components Rik Ril = kl , Rki Rli = kl

(2.50)

(2.51)

From the above we conclude that if the affine mapping eq. (2.43) is describing a rigid body motion, then Rij must be orthogonal, since from eq. (2.47) we get the condition, eq. (2.48). The case where det[ Rij ] = +1 corresponds to a real motion, whereas the case
det[ Rij ] = 1 is not, since the corresponding mapping is a reflection! In case when

Rij = ij From eq. (2.43) we get that xi = ci + i vector of all material points of the rigid body is a unique function of time only, ui = xi i = ci (t ) Since for t = 0 : ui = 0 xi i ci (0) = 0

(2.52)

(2.53)

In that case the motion is a translation of the rigid body (Figure 2-5) and the displacement

(2.54)

(2.55)

Thus from eq. (2.55) follows that the vector ci (t ) describes the motion of that material point, which at t = 0 was posiotined at the origin of the selected coordinate system.

16

Cosserat Continuum Mechanics, I. Vardoulakis 2009

Figure 2-5: Translation of a rigid body


2.4 General rigid-body motion

A pont of a body is called a fixed point if after the application of the motion, eq. (2.43), is mapped onto. Based on the definition of the affine mapping and the definition of the fixed point of a motion, we can prove the follwing theorems8:
Theorem 1:

If a motion, eq. (2.43), has four fixed points, that are not on the same plane, then the motion is an identy mapping of all points onto themselves ci (t ) = 0 , Rij (t ) = ij
Theorem 2:

(2.56)

If a motion, eq. (2.43), has a fixed plane and the motion is not the identy mapping, eq. (2.56), then this motion is not a real motion; it is a pseudomotion that corresponds to a reflection of all points of the considered body with respect to the given fixed plane.
Therome 3:

If a motion, eq. (2.43), possesses a fixed straight line, then this motion is a rotation with respect to that line.

cf. Grottemeyer, K.P., Analytische Geometrie, Gschen, Bd. 65-65a, 1964.

Cosserat Continuum Mechanics, I. Vardoulakis 2009

17

For the determination of the fixed points of a motion we consider the following equations,

i = ci + Rij j Tij j + ci = 0
where Tij = Rij ij To that we have te follwing matix

(2.57)

(2.58)

[T ] = [ R] [ I ]
rank of matrix [T ] as follows: 1) rng[Tij ] = 3 : In that det[Tij ] 0 , and there exists only one fixed-point,

(2.59)

The classification of the solutions of the problem, eqs. (2.57), is done on the basis of the

k = Tki1ci

(2.60)

If we assume that the mapping Rij corresponds to a real motion, i.e. if det[ Rij ] = +1 , then
T Rij T jk = R ji ( R jk jk ) = ik Rki = ( ik Rik ) T

(2.61)

thus
det[ R ji ( R jk jk )] = (+1) det[ R jk jk ] = det[ Rik ik ] det[ Rij ij ] = 0 rng[Tij ] < 3

(2.62)

This is in contradiction we the initial assumption. This case is impossible. 2) rng[Tij ] = 2 : Inthat case the homogeneous system of equations, Tij j = 0 has a solution of the form (2.63)

i = i
Thus
Tij j = 0 jT ji = j ( R ji ji ) = 0

(2.64)

j ( R ji ji ) Rik = 0 j ( R jk jk ) = 0
or

(2.65)

18

Cosserat Continuum Mechanics, I. Vardoulakis 2009

{1 2

R12 R13 R11 1 R 3 } 21 R22 1 R23 = {0 0 0} R31 R32 R33 1

(2.66)

This means that the vector

{ i }

is normal to the column-vectors of the matrix

[T ] = [ R I ] , and two such vectors are be hypothesis linearly independent. From eq. (2.57) we get,

i ( Rij ij ) j + i ci = 0 i ci = 0

(2.67)

Thus vector { i } must be normal to vector {ci } . In the considered case eq. (2.67) is a necessary condition for eq. (2.57) two have a solution. Thus a sufficient condition for a solution to exist in the considered case is that the vector {ci } is a linear combination of two vector-columns of the matrix [T ] = [ R I ] . In that case the solution we seek has the form,

i = i + i
where i is any particular solution of eq. (2.57). We consider now the vector
mi = k k ci k ck i m = (c )

(2.68)

(2.69)

Thus,

i mi = i k k ci i k ck i = 0
and ci = mi + i

(2.70)

i i

, =

i ci i i

(2.71)

With the help of this analysis we can replace the motion, eq. (2.43),

xi = Rij j + ci
by the following consequtive motions:

(2.72)

xi = Rij j + mi

(2.73)

Cosserat Continuum Mechanics, I. Vardoulakis 2009

19

xi = xi + i

(2.74)

Based on Theorem 2, above, the first component of the motion that is given by eq. (2.73), is a rotation with respect an axis that has the orientation of vector

{ i } .

This is so

because eq.(2.69) holds by construction. The second component of the motion, eq. (2.74), is a translation in the direction of { i } . A motion like that is called helicoidal. Figure 2-6 shows the geometric characteristics of a helix.

Figure 2-6: Right, circular helix


3) rng[Tij ] = 1 : This case is impossible when det[ Rij ] = +1 , since in this case the existence of a fixed plane is only compatible with a reflection. 4) rng[Tij ] = 1 : In this case Rij = ij and the motion is a translation. Theorem of Chasles9: The general motion of a rigid body is a helicoidal motion, that combines a rotation with respect to an axis and a parallel translation of screws 13.
10

. This theorem is originally attributed by

some authors 11 to Giulio Mozzi12, and is considered as the basis of the mechanical theory

Chasles, M. (1830). Note sur les proprits gnrales du systme de deux corps semblables entre eux, Bulletin de Sciences Mathmatiques, Astronomiques Physiques et Chimiques, Baron de Ferussac, Paris, 321-326. 10 P. Chadwick, Continuum Mechanics, Chapt.2, Dover, 1976. 11 Ceccarelli, M. (2000). Screw axis defined by Giulio Mozzi in 1763 and early studies on helicoidal motion, Mechanism and Machine Theory, 35, 761-770. 12 Mozzi, G. Discorso matematico sopra il rotamento momentaneo dei corpi, Stamperia di Donato Campo, Napoli, 1763, 13 R.S. Ball, Treatise on the Theory of Screws, Hodges Dublin 1876, Cambridge University Press 1900, Cambridge Mathematical Library 1998.

20

Cosserat Continuum Mechanics, I. Vardoulakis 2009

Cosserat continuum kinematics

3.1 Description of Cosserat kinematics in curvilinear coordinates

Figure 3-1: Cartesian and curvilinear coordinates of point in the plane


We restrict our analysis here to infinitesimal deformations. The following demonstrations follow closely the lines of the seminal paper of late Professor Wilhelm Gnther [23]. The position vector of the material point is (Figure 3-1)
OP = R = x i ei

(3.1)

where xi are the underlying Cartesian coordinates of the position vector with,
xi = i ( k ) ; i 0 k

(3.2)

This transformation allows us to write the position vector as a function of the curvilinear coordinates i (i = 1, 2,3) of the point P

R = ( i )
We introduce the local affine basis

(3.3)

gi =

= ,i i

, (),i

(3.4)

and we assume that the vectors, g1 , g 2 , g3 , in the given order build a right handed system. The infinitesimal displacement vector at point P is denoted as (Figure 3-2),

Cosserat Continuum Mechanics, I. Vardoulakis 2009

21

u = u i ( k ) g i
by another vector,

(3.5)

and the infinitesimal rotation of the rigid cross attached to the material point P , is given

= i ( k ) g i

(3.6)

Figure 3-2: Dofs of a material polar point of a 2D Cosserat continuum projected on to the local affine basis ( g1 , g 2 , g3 ).
We introduce the following vector deformation measures,

i = u ,i + gi i = ,i
Note that the i-th component of the vector product of two vectors is computed ( x y )i = eikl x k y l

(3.7) (3.8)

(3.9)

where eklm is the corresponding Levi-Civita 3rd-order fully antisymmetric tensor, g if : (k , l , m) = cycl (1, 2,3) = g if : (k , l , m) = cycl (2,1,3) , g = gij 0 else

eklm

(3.10)

where gij is the co-variant metric tensor. In sects. 3.4 and 3.5 . we will demonstrate that the above introduced measures, eqs. (3.7) and (3.8), really describe deformation in a Cosserat continuum. We observe that the gradient of a vector is expressed by means of its covariant derivative,

22

Cosserat Continuum Mechanics, I. Vardoulakis 2009

u, j = u ij gi

(3.11)

The 2nd term on r.h.s. of eq. (3.7) becomes,


gi = k gi g k = k eikl g l = eilk k g l

(3.12)

or,
gi = il g l

(3.13)

where

il = eilk k
With

(3.14)

i = ik g k

(3.15)

eqs. (3.7), (3.11) and (3.13) yield the following expression for the deformation tensor,

ik = u ik + ik
where

(3.16)

ik = il g lk
Similarly from

(3.17)

i = ik g k
and

(3.18)

,i = ik g k
we get the following expression for the Cosserat rotation gradient

(3.19)

ik = ik

(3.20)

Thus from the 6 placements u i and i (i = 1, 2,3) we have generated 18 deformations

ik and ik .
3.2 Integrability conditions and compatibility equations Let now () be a curve in space that is passing through points P0 and P . Starting from a point P0 we can compute the value of one of the kinematic fields, say the particle

Cosserat Continuum Mechanics, I. Vardoulakis 2009

23

rotation, by means of a line integral that is evaluated along the considered curve () . Thus from

( P) = ( P0 ) + ,k d k
P0

(3.21)

and eq. (3.8) we get

( P) = 0 + k dk
P0

(3.22)

For uniqueness purposes we require that the value of the Cosserat rotation at point P , as computed from eq. (3.22), is independent of the particular choice of the curve () that joins the points P0 and P ; assuming that at point P0 the value of is known. According to the fundamental theorem of Tensor Analysis [18] the sufficient and necessary condition for this integrability requirement is that rot k = 0 or
i = eijk k , j = 0 The first-order system k is called the 1st incompatibility form. With
(1) (1)

(3.23)

(3.24)

i = ik g k = ik g kl g l = il g l ; il = ik g kl
eq. (3.24) yields 0 = eijk k , j = eijk ( kl gl ) = eijk ( kl , j gl + kq g q , j ) = eijk ( kl , j + kq lqj ) gl
,j

(3.25)

(3.26)

or
I kl = e kpq ql p = 0
(1)

(3.27)

Similarly from,
u ( P ) = u ( P0 ) + u, k d = u0 + ( k g k ) d k
k P0 P0 P P

(3.28)

and due to eq. (3.4) through partial integration we get

24

Cosserat Continuum Mechanics, I. Vardoulakis 2009

u ( P ) = u0 + 0 P 0 + k + P k d k
P0

) (
P

(3.29)

The integrability of eq. (3.29) results to the following condition 0 = eijk k + P k

( (

= eijk k , j + , j k + P k , j

,j

(3.30)

Due to eqs. (3.24) and (3.4), eq. (3.30) yields the following condition,
= eijk ( k , j + g j k ) = 0
(2) k

(3.31)

The first-order system k is called the 2nd incompatibility form. In order to derive the analytic form of the 2nd compatibility condition we start from eq. (3.28) and set it as, u ( P ) = u0 + (
P0 P m k

(2)

+ eklp g
p

lm

)g

d = u0 + Ek d k
k P0

(3.32)

Ek = Ekm g m

, Ekm = km + eklp g lm p

After some extended algebraic manipulations eq. (3.32) yields [23],


I = eijk kr j ri + ri kk
(2) i r

(3.33)

where

rk j = g pr k pj
3.3 Compatibility equations in Cartesian coordinates

(3.34)

As an application of the above derivations we consider a Cartesian description of the motion of the Cosserat continuum. In that setting the infinitesimal displacement vector and Cosserat rotation of the polar material point P are given as, u = ui ei ; = i ei (3.35)

where ei is the orthonormal Cartesian basis.

Cosserat Continuum Mechanics, I. Vardoulakis 2009

25

In Figure 3-3 we see in a 2D setting that the polar material point is symbolized as a material cross that is moved from position P( xi ) to position P ( xi + ui ( xk ) ) and at the

same time is rotated in positive sense by a small angle 3 = ( xi ) .

Figure 3-3: Displacement and rotation of the polar material point in a 2D setting.

In Cartesian coordinates the relative deformation tensor, as defined above in eq. (3.16), becomes

ik = i uk + ik ; i =
with

xi

(3.36)

ij = ijk k
where ijk is the Cartesian permutation tensor,
1 if : (i, j , k ) = cycl (1, 2,3) = 1 if : (i, j , k ) = cycl (2,1,3) 0 else

(3.37)

ijk

(3.38)

Note that an antisymmetric system of 2nd order is always determined by a system of 1st order,
0 [ ij ] = 3 2 3
0

2 1 ij = ijk k 0

(3.39)

and with that ,

l = mnl mn

1 2

(3.40)

26

Cosserat Continuum Mechanics, I. Vardoulakis 2009

In Cartesian coordinates eq. (3.20) yields that the gradient of the Cosserat rotation is,

ik = i k k = k ; k k = 1

(3.41)

where the angle is infinitesimal and i is the axial unit vector. In particular we call the components (i )(i ) torsions and the rest components curvatures. Similarly from eqs.(3.36) and (3.37) we get,

ik = i uk ikl l

(3.42)

For completeness we write down first in Cartesian form the line integrals for the computation of the relative displacement and the relative rotation between two distant points P2 and P , eqs. (3.21) and (3.28), 1
( P2 , P ) 1 i

= i ( P2 ) i ( P ) = k i dxk = ki dxk 1
P 1 P 1

P2

P2

(3.43)
P2

( P2 , P ) 1 i

= ui ( P2 ) ui ( P ) = k ui dxk = ki dxk ikl l dxk 1


P 1 P 1 P 1

P2

P2

(3.44)

Eq. (3.43) can be written formally as I (i ) =

( )

(i ) k

dxk

A( i ) k = ki

(3.45)

The path independence if this line integral is guaranteed, if

pjk j A( i ) k = 0
or if
I pi = pjk j ki = 0
(1)

(3.46)

(3.47)

Similarly, eq. (3.44) formally reads, J ( i ) = B(i ) k dxk


P 1 P2

; B( i ) k = ki ikl l

(3.48)

leading to

pjk j B( i ) k = 0 pjk j ( ki ikl l ) = 0


or

(3.49)

Cosserat Continuum Mechanics, I. Vardoulakis 2009

27

pjk j ki + pjk ilk j l = 0 pjk j ki + ( pi jl pl ji ) jl = 0 pjk j ki + pi ll ip = 0


Thus, in Cartesian coordinates the compatibility conditions, eqs. (3.27) and (3.33), become
I kl = kpq p ql = 0 I pi pjk j ki + pi ll ip = 0
(2) (1)

(3.50)

(3.51) (3.52)

Explicitly these compatibility equations read as follows, I11 = 0 = 1 pq p q1 = 123 2 31 + 132 3 21 = 2 31 3 21 I12 = 0 = 1 pq p q 2 = 123 2 32 + 132 3 22 = 2 32 3 22
(1) (1)

(3.53)

and I11 = 0 = 1 jk j k1 11 + 11 kk = 2 31 3 21 + 22 + 33 I12 = 0 = 1 jk j k 2 21 + 21 kk = 2 32 132 3 22 21


(2) (2)

(3.54)

If ij are the components of the gradient of a vector field k , eq. (3.41), then the compatibility eqs. (3.53) reduce to the differentiability conditions for the named vector field, I11 = 0 = 2 3 1 3 2 1 I12 = 0 = 2 3 2 3 2 2
(1) (1)

(3.55)

Similarly, if the ij are given by eqs. (3.42), then the compatibility eqs. (3.54) reduce to the differentiability conditions for the vector field ui ,
I11 = 0 = 2 ( 3u1 31l l ) 3 ( 2u1 21l l ) + 2 2 + 3 3 = 2 3u1 3 2u1 2 2 3 3 + 2 2 + 3 3 = 2 3u1 3 2u1
(2)

(3.56)

28

Cosserat Continuum Mechanics, I. Vardoulakis 2009

3.4 Strain, spin, curvature and torsion in Cartesian coordinates


Let xi be the coordinates of a point of a rigid body before the motion and xi the coordinates of the same point after the motion. It can be shown that the motion [5], xi = xi + sin ikj nk + (1 cos ) ( ni n j ij ) x j where ni ni = 1 , 0 (3.58)

(3.57)

is a rigid-body rotation at a finite angle around a fixed axis with unit director ni . We consider two neighboring points P ( xi ) and Q( yi ) in the undeformed configuration of a Cosserat continuum, such that yi = xi + dxi . The material line element that connects these two points is given by the vector,
PQ = dxi ei

(3.59)

Using eq. (3.57) the positions of points P and Q in the deformed configuration are computed as,
xi = xi + ui + sin ikj k + (1 cos ) ( i j ij ) x j xi + ui + ikj k x j yi = xi + dxi + ui + j ui dx j + sin ikj k + (1 cos ) ( i j ij ) ( x j + dx j ) xi + dxi + ui + j ui dx j + ikj k ( x j + dx j )

(3.60)

Thus = xi + dxi + ui + j ui dx j + ikj k ( x j + dx j ) ( xi + ui + ikj k x j ) = ( ij + j ui ) dx j + ikj k dx j or dxi = dxi dxi = ( j ui + ikj k ) dx j From eqs. (3.62), (3.37) and (3.36) we get (3.62) dxi = yi xi

(3.61)

Cosserat Continuum Mechanics, I. Vardoulakis 2009

29

dxi = ( j ui + ikj k ) dx j = ( j ui ijk k ) dx j = ( j ui ij ) dx j = ( j ui + ji ) dx j = ji dx j

(3.63)

In matrix notation eq. (3.63) reads,


dx1 11 21 31 dx1 dx2 = 12 22 32 dx2 dx3 13 23 33 dx3

(3.64)

or

{dx} = [ ] {dx}
T

(3.65)

The length of the line element before and after the deformation is

ds 2 = dxi dxi ds2 = dxidxi = ds 2 + 2 ( ij ) dxi dx j where ( ij ) is the symmetric part of the relative deformation,

(3.66)

(ij ) =
=

1 ( ij + ji ) 2

1 ( iu j + ij + j ui + ji ) 2 1 = ( i u j + j ui ) 2

(3.67)

This means that the infinitesimal strain in the Cosserat continuum is given as the symmetric part of the relative deformation tensor and coincides with common infinitesimal strain tensor in the Boltzmann continuum

(ij ) = ij
where 1 ( iu j + j ui ) 2

(3.68)

ij =

(3.69)

Let us now consider the antisymmetric part of the relative deformation tensor,

30

Cosserat Continuum Mechanics, I. Vardoulakis 2009

[ij ] =

1 ( ij ji ) = 1 ( iu j + ij jui ji ) 2 2 1 1 1 = ( i u j j ui ) + ( ij ji ) = ( i u j j ui ) + [ij ] 2 2 2 1 1 = ( i u j j ui ) + ij = ( i u j j ui ) ijk k 2 2

(3.70)

The antisymmetric part of the transposed displacement gradient is denoted as

ij =
thus

1 ( jui iu j ) 2

(3.71)

[ij ] = ij ij
We may define the axial vector that corresponds to ij ,

(3.72)

ij = ijk k
with

(3.73)

k = k ; k k = 1
With that, eq. (3.73) becomes,

(3.74)

[ij ] = ij ij = ijk k + ijk k


= ijk ( k k ) Summarizing the above results, from eqs. (3.67) to (3.75) we get,

(3.75)

ij = ij + ( ij ij )
= ij ijk ( k k )
3.5 2D Cosserat kinematics in Cartesian coordinates

(3.76)

As an application we assume a 2D setting. In this case we have the following placements [44], u = u1e1 + u2 e2 (3.77)

= 3e3
The relative deformation tensor becomes

Cosserat Continuum Mechanics, I. Vardoulakis 2009

31

11 = 1u1 12 = 1u2 + 12 = 1u2 123 3 = 1u2 21 = 2u1 + 21 = 2u1 213 3 = 2u1 + 22 = 2u2
The curvature tensor becomes

(3.78)

11 = 12 = 0 ; 13 = 1 3 = 1 21 = 22 = 0 ; 23 = 2 3 = 2 11 = 22 = 33 = 0
The compatibility conditions, eq. (3.27) and eq. (3.33), yield
I 33 = 3 pq p q 3 = 321 213 + 312 1 23 = 213 + 1 23 = 0
(2) (1)

(3.79)

(3.80)

I 31 = 3 jk j k1 13 = 312 1 21 + 321 2 11 13 = 1 21 2 11 13 = 0 I 32 = 3 jk j k 2 23 = 312 1 22 + 321 2 12 23 = 1 22 2 12 23 = 0 Introducing the infinitesimal strain tensor and the infinitesimal background spin tensor,
(2)

(3.81)

11 = 1u1 12 = 21 = 22 = 2u2
1 ( 1u2 + 2u1 ) 2

(3.82)

11 = 0
1 ( 2u1 1u2 ) = 1233 = 2 1 21 = ( 1u2 2u1 ) = 2133 = + 2 22 = 0

12 =

(3.83)

we get,

32

Cosserat Continuum Mechanics, I. Vardoulakis 2009

11 = 11
1 2 1 21 = 2u1 1u2 + = 21 ( ) 2 22 = 2u2

12 = 1u2 2u1 = 12 + ( )

(3.84)

Thus

(ij ) = ij
and
1 ( 12 21 ) 2

(3.85)

[12] =
=

1 ( 1u2 ( 2 u1 + ) ) = 1 ( 1u2 2 u1 ) 2 2 =

(3.86)

In order to visualize this decomposition, we consider a line element PQ that is originally parallel to the x1 -axis, eq. (3.59), with
dx1 1 = dx dx2 0

(3.87)

With,

[ ]

21 = 11 12 22

(3.88)

we get from eq. (3.63),


dx1 11 21 1 11dx = 0 dx = dx dx2 12 22 12

(3.89)

We consider a line element PR that is originally parallel to the x2 -axis,


dx1 0 = dy dx2 1

(3.90)

Then
dx1 11 21 1 21dy = 0 dy = dy dx2 12 22 22

(3.91)

Cosserat Continuum Mechanics, I. Vardoulakis 2009

33

In Figure 3-4 we show the geometrical visualization of the deformation of the solid orthogonal element PQ, PR , that is computed from eqs. (3.89) and (3.91).

Figure 3-4: The deformation of a solid orthogonal element From this figure it becomes clear that the diagonal terms of the relative deformation matrix describe normal strains,

( PQ) =

( (1 + ) dx ) + (
2 11

12

dx ) dx 2 (1 + 2 11 ) dx (1 + 11 )
2

( PQ) ( PQ) (1 + 11 ) dx dx = 11 = 1u1 = 11 ( PQ) dx Similarly we get that ( PR) ( PR) 22 = 2u2 = 22 ( PR) Angular strains are given by,

(3.92)

(3.93)

(QPR)

12 dx 21dy + + = 212 = 2 21 = (1 + 11 ) dx (1 + 22 ) dy 12 21

(3.94)

We return now to eq. (3.64) and we consider a deformation that is generated by [ij ]
dx1 [11] [21] dx1 0 [21] dx1 = = 0 dx2 dx2 [12] [22] dx2 [12]

(3.95)

34

Cosserat Continuum Mechanics, I. Vardoulakis 2009

or 0 213 ( 3 3 ) dx1 dx1 = 0 dx2 123 ( 3 3 ) dx2 0 ( 3 3 ) dx1 = 0 ( 3 3 ) dx2 0 ( ) dx1 = 0 ( ) dx2 3-5): For the line element PQ we get, ( ) 1 0 dx1 0 = = 0 dx2 ( ) 0 and for the line element PR we get, ( ) 0 ( ) dx1 0 = = 0 0 dx2 ( ) 1 (3.98) (3.97) (3.96)

We visualize this motion using again the concept of the rigid orthogonal element (Figure

Figure 3-5: The relative rotation of a solid orthogonal element As is shown in Figure 3-6, this motion is the relative rotation of the polar material points in the neighborhood of point P that is due to the displacement field with respect to the rotation that is due to their spin. In Figure 3-7, for the visualization of the curvature of the

Cosserat Continuum Mechanics, I. Vardoulakis 2009

35

Cosserat deformation we consider the relative rotation of the rigid crosses attached at points Q and R , with respect to the rigid cross attached at point P ,

( Q ) ( P ) + 13 x1 ( R ) ( P ) + 23x2

(3.99)

The curvature tensor is seen as a measure of the bend of the neighbourhood of point P .

Figure 3-6: Visualization of the relative spin

Figure 3-7: Visualization of the curvature of the deformation

36

Cosserat Continuum Mechanics, I. Vardoulakis 2009

3.6 Exercise: Deformation of Cosserat continuum in special curvilinear coordinates 3.6.1 Polar cylindrical coordinates

Figure 3-8: Cartesian and polar cylindrical coordinates The polar cylindrical coordinates of a point P (r , , z ) , are related to its Cartesian coordinates by the following set of equations (Figure 3-8),

x = x1 = 1 cos 2 = r cos y = x 2 = 1 sin 2 = r sin z = x 3 = 3 for r (0, ) and [0, 2 ) .

(0 2 ) (3.100)

Prove that the deformation tensors in cylindrical polar cylindrical coordinates are as follows, ur r ur u 1 [ u ] = r r ur z u r u ur 1 + r r u z u z r 1 u z r u z z

(3.101)

Cosserat Continuum Mechanics, I. Vardoulakis 2009

37

rr r rz sym [ ] = r z = zr z zz ur 1 1 ur u u + r r r 2 r 1 1 ur u u 1 u ur = + + r r r r 2 r 1 ur u z 1 u 1 u z + z + r r 2 2 z
asym [ ] = 0 1 u 1 ur u = + + z r 2 r r u 1 u r z 2 z r (3.103)

1 ur u z + r 2 z 1 u 1 u z + 2 z r u z z

(3.102)

1 u 1 ur u + z 2 r r r 0 1 1 u z u + r 2 r z

1 u z ur r z + 2 1 1 u z u r z r 2 0

r r 1 r [ ] = r r r z

r 1 r + r r z

z r 1 z r z z

(3.104)

38

Cosserat Continuum Mechanics, I. Vardoulakis 2009

3.6.2 Polar spherical coordinates

Figure 3-9: Cartesian and polar spherical coordinates The polar spherical coordinates of a point be P(r , , ) are related ti its Cartesian coordinates by the following set of equations (Figure 3-9)
x = x1 = 1 sin 2 cos 3 = r sin cos y = x 2 = 1 sin 2 sin 3 = r sin sin z = x3 = 1 cos 2 = r cos for r (0, ) , [0, ) and [0, 2 ) . Prove that the deformation tensors in polar spherical coordinates are as follows,
sym[ ] = ur r 1 1 ur u u 2 r + r r u u 1 + 1 ur r r sin r 2 1 1 ur u u + r 2 r r 1 u ur + r r 1 1 u 1 u cot + u r sin r r 2 1 u 1 ur u + r r sin 2 r 1 u 1 u cot 1 + u r 2 r sin r 1 u ur u + + cot r sin r r

(3.105)

(3.106)

Cosserat Continuum Mechanics, I. Vardoulakis 2009


asym [ ] = 0 1 u 1 ur + u r 2 r r 1 ur u 1 u 2 r r sin + r + 1 u 1 ur u + + r 2 r r 0 1 1 u 1 u cot + u r r 2 r r sin

39

1 u 1 ur u + r 2 r r sin 1 1 u 1 u cot + u + r r 2 r r sin 0

(3.107)
r r 1 r [ ] = r r 1 r r r sin r 1 r + r r 1 1 r sin r tan r 1 r 1 r 1 + + r sin r r tan

(3.108)

40

Cosserat Continuum Mechanics, I. Vardoulakis 2009

Cosserat continuum statics

4.1 The virtual work equation


We consider a Cosserat continuum , that occupies a domain with volume V that has the boundary V . Body is assumed to be in a state of stress in static equilibrium. In order to formulate the equilibrium conditions we consider fields u i (k ) and i (k ) , that are defined uniquely at all points of the given body. These fields will be called virtual particle displacement and virtual particle rotation fields, respectively and it will be assumed that they are sufficiently differentiable. We define the virtual relative deformation tensor,

ik = uk i + ik
where

(4.1)

ik = eikl l
and the virtual curvature tensor

(4.2)

ik = k i

(4.3)

We define the tensor fields ij (m ) and ij (m ) , through the so called virtual work of the internal forces, that is in turn defined per unit volume of the considered continuum,

w(int) = ik ik + ik ik

(4.4)

We assume that w(int) is an invariant scalar quantity. In this case the quantities and are called the stress- and couple-stress tensors, respectively. From the point of view of continuum thermodynamics the stress- and couple stress tensors are intensive quantities, that are dual in energy to the deformation gradient an spin, that are in turn the corresponding mechanical extensive quantities of the considered continuum14. We remark at this point that the above definition of the virtual work of internal forces is consistent with the Cosserat continuum energetics, that are discussed below in sect. 6.

14

An intensive property (also called a bulk property) is a physical property of a system that does not depend on the system size or the amount of material in the system. By contrast, an extensive property of a system does depend on the system size or the amount of material in the system

Cosserat Continuum Mechanics, I. Vardoulakis 2009

41

In view of eq. (3.76) we decompose additively the virtual relative deformation into symmetric and antisymmetric part

ij = (ij ) + [ij ]
where 1 k l ( i j + il jk ) kl = 1 ( ij + ji ) = ij 2 2 1 1 = eijp e klp kl = ( ij ji ) = ij ij 2 2

(4.5)

(ij ) = [ij ]
or

(4.6)

ij = ij + ( ij ij )
= ij eijk ( k k ) = ij + eijk ( k k )

(4.7)

where

ij =

1 u j i + ui j 2

(4.8)

ij = eijk k =
and

1 ui j u j i 2

(4.9)

ij = eijk k
Thus

(4.10)

[ij ] = eijk ( k k )
part

(4.11)

Similarly we decompose additively the stress tensor into symmetric and antisymmetric

ij = (ij ) + [ij ]
where 1 i j (kl + ilkj ) kl = 1 ( ij + ji ) 2 2 1 ijp 1 i j 1 = e eklp kl = ( k l il kj ) kl = ( ij ji ) 2 2 2

(4.12)

(ij ) =
[ ij ]

(4.13)

With this decomposition the virtual work of the internal forces, eq. (4.4), becomes,

42

Cosserat Continuum Mechanics, I. Vardoulakis 2009

w(int) = (ij ) ij + [ ij ] [ij ] + ij ij


= (ij ) ij + [ij ] ( ij ij ) + ij ij = (ij ) ij + [ij ]eijk ( k k ) + ij ij Eq. (4.14) is summarized in the following, (4.14)

Theorem
In a Cosserat continuum the virtual work of the internal forces is such that: a) the symmetric part of the stress tensor (ij ) is dual in energy to the strain ij = u( i j ) (i.e. to the symmetric part of the displacement gradient), b) the antisymmetric part of the stress tensor [ i , j ] is dual in energy to the relative spin15 eijk ( k k ) , and c) the couple stress tensor ij is dual in energy to the distortion tensor ij = j i . We remark that the term in eq. (4.14) that pertains to the virtual work of the antisymmetric part of the stress tensor can be written as follows,

[ij ] [ ij ] = [ij ] ( ij ij )
= [ij ] eijm ( m m )
* = 2tm ( m m )

(4.15)

where ti is the axial vector, that corresponds to the non-symmetric part of the stress tensor. 1 1 ti = eijk jk = eijk [ jk ] 2 2 [ jk ] = eijk ti (4.16)

With this remark eq. (4.14) becomes

w(int) = (ij ) ij + 2ti* ( i i ) + ij ij w(int) over the volume V ,


W (int) =

(4.17)

The total work of the internal forces is defined as the integral of the density function

(V )

(int)

dV

(4.18)

15

For this reason we call [ij ] the relative stress tensor (see sect. 5.7 ).

Cosserat Continuum Mechanics, I. Vardoulakis 2009

43

For the formulation of the principle of virtual work (p.v.w.), we assume that on the considered Cosserat continuum three types of external forces are acting16: a) Volume forces f i dV , b) surface tractions t i dS and c) surface couples m k dS . In these expressions
dV is the volume element and dS is the surface element. These external actions are

related to the internal forces and it is the virtual work equation that defines this relation. As we will see in sect.4.2, the equations that couple locally the internal and the external forces are the equilibrium equations.

Figure 4-1: Local coordinates in a point at the bounding surface The bounding material surface V of a material volume V is seen as a two dimensional smooth point manifold17 with each point of that manifold possessing two vectorial degrees of freedom, the one of particle displacement and that of particle rotation. At any point P V we define a local coordinate system, say

( ,
1

, 3 ) , where the

coordinates 1 and 2 , describe the position of the considered point in the bounding surface and 3 is the coordinate positive along the outward normal to it At the arbitrary point P ( 1 , 2 , 0 ) on the surface we can define the corresponding covariant basis,

16 17

In general one may assume the existence of body couples as well; cf. sect. 5.4 . Bounding surfaces with sharp corners and edges are not considered here.

44

Cosserat Continuum Mechanics, I. Vardoulakis 2009

( 1 , 2 , 3 )

as shown in Figure 4-1. From that basis we may construct the corresponding

contavariant basis ( 1 , 2 , 3 ) 18.. A set of admissible boundary conditions at point P ( 1 , 2 , 0 ) are of purely Diriclet type, with data projected on the local contravariant basis u1 u2 u3 P V : S D : [ p ]P = S N = : [ q ]P = 1 2 3 (4.19)

Accordingly Neumann type boundary conditions are expressed on the covariant basis, t1 P V : S D = : [ p ]P = S N : [ q ]P = 1 m t2 m2 t 3 m3 (4.20)

Mixed-type boundary conditions are also allowed, however if some information, say pij is given no information concerning q ij can be given et vice versa; e.g. t1 u3 S N : [ q ]P = 1 P V : S D : [ p ]P = 3 m t 2 2 m (4.21)

In the example given above by eq. (4.21) in the neighbourhood of point P ( 1 , 2 , 0 ) and along the -surface lines ( = 1, 2) (i.e. in the tangential plane) tractions and couples are prescribed, whereas in normal to the surface direction the displacement and the spin are restricted. On the basis of the above definitions we define a functional that is called the virtual work of external forces,

W (ext ) =

(V )

f i ui dV +

(SN )

(ti ui + mi i )dS

(4.22)

We assume that on S D , the virtual kinematics vanish,

ui = 0 i = 0
18

(4.23)

For a concise presentation of the geometry of a surface in the Euclidean space we refer to: McConnell, A.J., Applications of Tensor Analysis, Ch. XIV, Dover, 1957, and to: Klingbeil, E., Tensorrechnung fr Ingenieure, Kap. 4, BI, 1966.

Cosserat Continuum Mechanics, I. Vardoulakis 2009

45

We assume that these data are continuously extended into V and on the disjoint parts of the boundary. Thus the functional can be extended over the whole boundary,

W (ext ) =

(V )

f i ui dV +

(V )

(ti ui + mi i )dS

(4.24)

on S D : i = 0 ui = 0

On the basis of the above definitions the p.v.w. in a Cosserat continuum is defined as follows: The system { ij , ij ; f i , t i , mi } is called an equilibrium set, if for any choice of the virtual fields of displacement and rotation that satisfy eq. (4.23), the virtual work equation holds,

W (ext ) = W (int)

(4.25)

From eq. (4.25) and the definitions for the virtual work of internal- and external forces, eqs. (4.18), (4.4) and (4.24), we obtain the following integral equation, that is the expression of the virtual work equation in a Cosserat continuum:

(V )

f i ui dV +

(V )

t i ui dS +

(V )

mi i dS =

(V )

ik

ik + ik ik dV

(4.26)

4.2 Equilibrium equations in curvilinear coordinates


We remark first that the density of the virtual work of the internal forces can be written as follows,

w(int) = ik uk i + ik + ik k i
= ik uk + ik k

) i ( ik i uk + ik i k ) eikl ik l

(4.27)

With the notation qi = ik uk + ik k we observe that


divq = q|ii

(4.28)

(4.29)

and with the use of Gauss theorem we get,

46

Cosserat Continuum Mechanics, I. Vardoulakis 2009

(
V

ik

uk + ik k
ik ik

) i dV = divq dV = q n dS )
V v

uk + k ni dS

(4.30)

With eq. (4.30) the virtual work equation (4.26) becomes


(V )

f k uk dV +
ik

(
(V )

(V )

t k uk dS +

uk + ik k ni dS
i uk dV

(V )

m k k dS = (4.31)

or

ik

(V )

ik

+ eiml g lk im k dV

(V )

ik

+ f k uk dV +

(V )

ik ni t k uk dS +

(V )

ik

+ eiml g lk im k dV =

(V )

ik ni mk k dS

(4.32)

The test functions uk (k ) and k ( k ) can be chosen arbitrarily. In particular they may be chosen in such a way that from eq. (4.32) the following equations follow,

(V )

( (

ik

+ f k uk dV = 0 V ' V + eiml g
lk im

ik

(V )

) dV = 0
k

V ' V

(4.33)

(V )

( ik ni t k ) uk dS = 0 (
ik

V ' V V ' V (4.34)

ni m

(V )

) k dS = 0

These equations result finally to the following set of stress equilibrium equations,

ik i + f k = 0 P(i ) V

(4.35) (4.36)

ik ni = t k

P(i ) V

and moment stress equilibrium equations,

Cosserat Continuum Mechanics, I. Vardoulakis 2009

47

ik i + eiml g lk im = 0 P(i ) V

(4.37) (4.38)

ik ni = mk

P(i ) V

We remark here that in [23] we find the following equivalent form of eq. (4.37), g kp ik i + eiml g kp g lk im = 0 ip i + eimp im = 0 (4.39)

4.3 Equilibrium equations in Cartesian coordinates


We apply eqs. (4.35) to (4.38) for a Cartesian description, thus yielding

ik ni = ti
i ik + f k = 0 and

(4.40) (4.41)

ik ni = mk
i ik + imk im = 0

(4.42) (4.43)

We observe that the equilibrium equations (4.40) and (4.41) are identical to the ones holding for the Boltzmann continuum and that the equilibrium equations (4.40) and (4.42) introduce the stress- and couple stress tensors as lineal densities for the internal forces in the sense of Cauchy. Due to the moment equilibrium equation (4.43), however, the stress tensor in a Cosserat continuum is in general non-symmetric.

Figure 4-2:Stress and couple stress in the sense of Cauchy in 2D. As an example we apply the above equilibrium equations for a 2D setting, thus yielding [44] (Figure 4-2),

48

Cosserat Continuum Mechanics, I. Vardoulakis 2009

t1 = 11n1 + 21n2 t2 = 12 n1 + 22 n2 m3 = 13 n1 + 23 n2 and (Figure 4-3) 11 21 + + f1 = 0 x1 x2 12 22 + + f2 = 0 x1 x2 13 23 + + 12 21 = 0 x1 x2 (4.45) (4.44)

Figure 4-3: Stress and moment stress equilibrium in 2D

4.4 The Mohr circle of non-symmetric stress in 2D19


We consider a 2D state of stress in the O( x1 , x2 ) -plane and we identify x1 = x , x2 = y . As shown in Figure 4-4 we may introduce another coordinate system O( , ) that results from the original coordinate system O ( x, y ) by a rotation of the coordinate axes by an angle and we want to compute the components of the stress vector in planes parallel to the rotted system.

19

. For a 3D non-symmetric state of stress the concept of the Mohr circle is meaningfull only for very special cases [41]. However, as we will see in this section, it is always meaningful for a 2D setting.

Cosserat Continuum Mechanics, I. Vardoulakis 2009

49

Figure 4-4: Plane state of stress As can be seen from Figure 4-4 the normal and tangential vectors on the plane that is normal to the axis are, n1 = cos ; n2 = sin m1 = sin ; m2 = cos and with that (4.46)

1 ( xx + yy ) + 1 ( xx yy ) cos 2 + 1 ( yx + xy ) sin 2 2 2 2 1 1 1 = ( xx yy ) sin 2 + ( xy yx ) + ( xy + yx ) cos 2 2 2 2 1 1 1 = ( xx yy ) sin 2 ( xy yx ) + ( xy + yx ) cos 2 2 2 2 1 1 1 = ( xx + yy ) ( xx yy ) cos 2 ( yx + xy ) sin 2 2 2 2

(4.47)

We observe that, if the stress tensor is symmetric ( xy = yx = ), then the transformation rule, eq. (4.47), collapses to the one known from Boltzmann Continuum Mechanics.

50

Cosserat Continuum Mechanics, I. Vardoulakis 2009

Figure 4-5: Normal and shear stresses on an arbitrary plane In order to check the applicability of concept of the Mohr circle of stresses for nonsymmetric states of stress, we consider here the geometric locus of the normal- and shear stress vector, acting on a plane with unit outward normal ni with the angle as curve parameter (Figure 4-5), 1 ( xx + yy ) + 1 ( xx yy ) cos 2 + 1 ( yx + xy ) sin 2 2 2 2 1 1 1 n = ( xx yy ) sin 2 + ( xy yx ) + ( xy + yx ) cos 2 2 2 2

n =

(4.48)

Let the mean normal stress is an in-plane invariant and is denoted as

M =

1 ( xx + yy ) = 1 ( + ) 2 2

(4.49)

The shear stress difference,


a = t3 =

1 ( xy yx ) = 1 ( ) 2 2

(4.50)

is also an invariant and measure of the stress-tensor asymmetry20. With this notation we get from eq. (4.48),

20

cf. eq. (4.17).

Cosserat Continuum Mechanics, I. Vardoulakis 2009

51

1 ( xx yy ) cos 2 + ( xy ) sin 2 2 1 tn = n a = ( xx yy ) sin 2 + ( xy ) cos 2 2 sn = n M = Thus

(4.51)

xx yy 2 = s +t = + ( xy ) 2
2 n 2 n

(4.52) (4.53)

tn = tan ( 2 ( ) ) sn where

tan 2 =

2 ( xy )

xx yy

(4.54)

Figure 4-6: Mohr circle for a non-symmetric stress tensor in 2D

As can be seen from Figure 4-6, M is the radius of the Mohr circle. The center ( M , a ) of the Mohr circle is shifted normal to the O n -axis by the amount of the asymmetry of the stress tensor, given by a .This observation allows us to transfer some geometrical results from Mohr-circle analysis, holding for symmetric stress, to the 2D

52

Cosserat Continuum Mechanics, I. Vardoulakis 2009

Mohr-circle holding for non-symmetric stress. For example the pole of normals is defined as in the Boltzmann continuum. From Figure 4-6 we can also compute the angular displacement of the center of the Mohr circle, expressed by the stress obliquity of the antisymmetric part of the stress tensor, tan a = tan ( ( ) ) =

yx a = xy M xx + yy

(4.55)

The opening angle of the tangent lines, drwn from the origin to the shifted Mohr circle is 2s , where
tan s = tan ( (C ) ) = tan ( (C ) ) = tan s =

( C ) = ( C ) ( C ) ( )2 ( C )2

M
2 2 2 M + a M

(4.56)

With,

tan M = M = M
eq. (4.56) becomes,
tan s =

xx

( xx + yy )

yy ) + 4 (2xy )
2 2

(4.57)

tan M 1 + tan 2 a tan 2 M

(4.58)

We observe that for symmetric states of stress we retrieve a well known result,

a = 0 tan s =
or
sin s = tan M =

tan M 1 tan 2 M
yy ) + 4 (2xy )
2 2

(4.59)

xx

( xx + yy )
2

(4.60)

With eqs. (4.55) and (4.57) eq. (4.58)becomes,


1 tan s = 2

xx

yy ) + ( xy + yx )

xx yy xy yx

(4.61)

Cosserat Continuum Mechanics, I. Vardoulakis 2009

53

The slopes of these tangents with respect to to the n - axis are,


2 M a 2 2 2 M + a M M 2 a M 1 2 2 2 M + a M M

tan (s a ) =

tan s tan a = 1 tan s tan a

(4.62)

With the notation,


p = M ; q =M 0 ; r = a

(4.63)

and tan a = ; tan s = (4.64)

from eqs. (4.55) and (4.56) we get, tan a = tan s =

a r= p M M
2 2 2 M + a M

q= p2 + r 2 ) 2 ( 1+

(4.65)

With that from eq. (4.62) we get,

tan (s a ) =

tan s tan a = 1 tan s tan a 1

(4.66)

If a 0 ( a > 0 , > 0 ), then the critical Mohr-Coulomb envelope has the slope angle,

s + a ; for a < 0 ( a < 0 , < 0 ), the critical value is s a . Thus we set


m=

+ 1

(4.67)

If we assume that m is a constitutive friction parameter, we get a constraint equation between the friction parametrs

m 1 + m m p 1 + m

0< <m

(4.68)

This means that


r =

r +m p 0 r m+ p

r m p

(4.69)

From eqs. (4.65) and (4.69) we get

54

Cosserat Continuum Mechanics, I. Vardoulakis 2009

q=

1 ( p + r m )( p m r ) ; r m p 1 + m2

(4.70)

In ( p, q, r ) -space of stresses this is a compound conical surface, as shown in Figure 4-7., that degenerates into straight lines on the varios coordinate planes,
for r = 0: q = 1 p = sin s p 1 + m2

(4.71)

and (Figure 4-8),


for q = 0 : r = m p sin 2 s p

(4.72)

Figure 4-7: Limit condition in the ( p, q, r ) -stress space of the Mohr-Colomb type for a 2D frictional Cosseart material ( m = 0.5 ).

Figure 4-8: Relation between friction coefficients m and sin s

Cosserat Continuum Mechanics, I. Vardoulakis 2009

55

4.5 Exercise: Differential equilibrium equations in special curvilinear coordinates 4.5.1 Polar cylindrical coordinates Prove that the equilibrium equations for a Cosserat continuum in terms of physical

components in polar, cylindrical coordinates are the following:


rr 1 r 1 + + ( rr ) + zr + f r = 0 r r r z r 1 1 + + ( r + r ) + z + f = 0 r r r z rz 1 z 1 zz + + rz + + fz = 0 r r r z

(4.73)

and
rr 1 r zr 1 + + + ( rr ) + z z + r = 0 r z r r r 1 z 1 + + + ( r + r ) + zr rz + = 0 r z r r rz 1 z zz 1 + + + rz + r r + z = 0 r z r r

(4.74)

4.5.2 Polar spherical coordinates Prove that the equilibrium equations for a Cosserat continuum in terms of physical

components in polar, cylindrical coordinates are the following:


rr 1 r 1 r cot 1 + + + r + 2 rr + f r = 0 r r sin r r r r 1 1 1 cot + + + ( 2 r + r ) + + f = 0 r r sin r r r r 1 1 1 cot + + + r + r + + + f = 0 r r r sin r r

(4.75)

rr 1 r 1 r cot 1 + + + r + 2rr + + r = 0 r r r sin r r r 1 1 1 cot + + + ( 2 r + r ) + + r r + = 0 (4.76) r r r sin r r r 1 1 1 cot + + + 2 r + r + + + r r + = 0 r r r sin r r

56

Cosserat Continuum Mechanics, I. Vardoulakis 2009

Cosserat continuum dynamics

The equations that describe mass balance and balance of linear momentum in a Cosserat continuum are the same as the ones holding for a Boltzmann continuum. The difference between the two types of continua arises while considering the action of the extra dofs of the Cosserat continuum, i.e. in the formulation of the momentum balance- and energy balance equations. For completeness we derive here also the equations that describe balance of mass and balance of linear momentum.
5.1 Balance of mass

The material point of a Cosserat continuum is equipped with a linear particle velocity
Du i v = Dt
i

(5.1)

We remark that within a small deformation theory we have that


i Du u i k u = +v t x k Dt i

i u i u i v k k = x t
k

(5.2)

or
i v k Fk = t u i i v k Fk F 1ji = t u i F 1ji

v j = t u i F 1ji t u i + O k u l

(5.3)

The mass of the particle is,


dm = dV

(5.4)

where density ( i , t ) is the mass density at the considered point. The total mass of a body B at a given time t is,
M (t ) =

(V )

dV

(5.5)

Mass balance is expressed by the requirement,


dM =0 dt

(5.6)

We recall Reynolds transport theorem,

Cosserat Continuum Mechanics, I. Vardoulakis 2009

57

S (t ) =

(V )

s (k , t )dV

dS = dt

i t s + sv dV i (V )

( )

(5.7)

Thus, from eqs. (5.5) and (5.6) follows that mass balance is expressed as
(V )

( + ( v )
i t

= 0 dV = 0 V V

(5.8)

If we assume that mass balance holds for any subdivision of the considered body, then from eq. (5.8) we get the following local form for the mass balance equation, + ( vi ) = 0 P(i ) V i t quantity
S (t ) =

(5.9)

Note that if eq. (5.9) holds, then Reynolds transport theorem applied for the global

(B)

sdm = sdV
(V )

(5.10)

yields,
dS = sdV dt (V )

(5.11)

where s is the material time derivative of the specific quantity s (k , t ) :


s= Ds s k = + v sk Dt t

(5.12)

5.2 Balance of linear momentum

The total force that is acting on a body B at a given time t is,


F i (t ) =

(V )

f i dV +

( V )

t i dS

(5.13)

The total linear momentum of the consider body is, i (t ) =

(V )

v dV
i

(5.14)

Balance of linear momentum is expressed as,


d i = Fi dt

(5.15)

From Reynolds transport theorem, eq. (5.12), we get that,

58

Cosserat Continuum Mechanics, I. Vardoulakis 2009

d i d i i k i = ) v dV = (V ) t v + v v k dV dt dt (V

(5.16)

Thus from eqs. (5.13) to (5.16) we get,

(V )

( v
t

+ v k vik dV =

(V )

f i dV +

( V )

t i dS

(5.17)

We assume that linear momentum balance holds for any subdivision of the considered body. If we apply eq. (5.17) in particular for the elementary tetrahedron under suitable mathematical restrictions [24] the volume integrals tend to zero and the remaining surface integral yields Cauchys theorem21,

ki nk = t i P(m ) V
From eqs. (5.17) and (5.18) and Gauss theorem we get
(V )

(5.18)

t vi + vik v k dV =

(V )

f i dV +

( V )

ki nk dS
(5.19)

(V )

f dV +
i

(V )

ki

dV k

V V

We observe that the material time derivative of the velocity coincides with the particle acceleration,
ai = Dv i = t v i + vik v k Dt Dvi Dt

(5.20)

From eq. (5.19) and (5.20) we get the dynamic equations,

ki k + f i =

(5.21)

We observe that if we assume that the particle acceleration is vanishing, then eqs. (5.21) reduce to the static equilibrium eqs. (4.35).
5.3 Balance of angular momentum

The total moment of the forces and couples acting on a body B at a given time t is,
M=

( V )

tdS +

(V )

fdV +

mdS

(5.22)

( V )

21

cf. eq. (4.36).

Cosserat Continuum Mechanics, I. Vardoulakis 2009

59

where is the position vector. On the other hand the total angular momentum is
L=

(V )

( v ) dV +

dV

(5.23)

(V )

where is the angular momentum of the spinning polar material point. Balance of angular momentum is expressed as,
dL =M dt

(5.24)

Assuming that mass balance is holding we have that,


dL d = v + dV dt dt (V ) Dv D = dV dV + Dt Dt (V ) (V )

(5.25)

We consider the 1st term on the r.h.s. of eq. (5.22), and evaluate it for convenience in a Cartesian description,

( V )

( t ) dS =
i
(V )

ijk x j tk dS =

( V )

( V )

ijk x j mk nm dS mk + x j m mk ) dV
(5.26)

= = Let,

ijk

m ( x j mk ) dV =

(V )

(
ijk

mj

(V )

ijk

jk + ijk x j m mk ) dV

t = tiei

1 , ti = ijk jk 2
i

(5.27)

and with that eq.(5.22) becomes


Mi =
(V )

2t dV +
(V ) i

ijk

x j m mk dV +

(V )

ijk

x j f k dV +

( V )

ki nk dS
(5.28)

(V )

( 2t

+ k ki ) dV +

(V )

ijk

x j ( m mk + f k ) ) dV

By combining eqs. (5.24), (5.25) and (5.28) we obtain

(V )

Di Dv dV = ( 2ti + k ki ) dV + ijk x j m mk + f k k dV Dt Dt (V ) (V )

(5.29)

60

Cosserat Continuum Mechanics, I. Vardoulakis 2009

If we assume that balance of linear momentum holds, then the last term on the r.h.s. of eq.(5.29) is vanishing, thus yielding

(V )

Di dV = ( 2ti + k ki ) dV Dt (V )

(5.30)

The local form of eq. (5.30) is k ki + ikl kl =


Di Dt

(5.31)

5.4 The micro-morhic continuum interpretation

Figure 5-1: The microstructure of an (REV) with sub-particles sharing a rigid-body motion

For the determination of the angular momentum of the spinning polar material point we resort to the micro-morphic continuum interpretation22. In this case we assign to the material (polar) particle (or macro-particle) of the continuum the average properties of a Representative Elementary Volume (REV) of an assembly of sub-particles, as shown in Figure 5-1. The (REV) may consist of sub-particles (or micro-particles). The spatial position of the polar macro-particle is identified with the position of the center of mass

22

The terminus micro-morphic is introduced by Eringen: A volume element of a micro-morphic medium consists of micro-elements which undergo micro-motions and micro-deformations. Micro-polar media are a subclass, in which the micro-elements behave like rigid bodies; Eringen, A.C. (1965). Theory of micropolar continua. Proc. 9th Midwestern Mechanics Conference, Madison.

Cosserat Continuum Mechanics, I. Vardoulakis 2009

61

S ( xi ) of the sub-particles in the (REV). The velocity vi of the center of mass S ( xi ) is

defined as the velocity of the particle itself,


vi = xi

(5.32)

The sub-particle, at position P ( xi + yi ) , has the mass m = pV , where p is the subparticle mass density, V , and a velocity
vi ( P) = vi ( S ) + vi

(5.33)

The total mass of the macro-particle is the sum of the masses of its constituents,
m = m

(5.34)

The linear- and angular momentum of the macro-particle are computed as follows,
ii = mvi d k = m ijk xi v j + ( m ijk yi v j )

(5.35) (5.36)

The volume of the material (REV) is V and the total volume of the sub-particles inside the (REV) is
V p = V

(5.37)

The volume fraction

V Vp V

= 1

V
V

(5.38)

is the porosity of the (REV). The density of the macro-particle is

m m V = = V V V
si = mvi = vi V

= (1 ) p

(5.39)23

Similarly we introduce the linear momentum of the macro-particle, (5.40)

and its angular momentum


23

If the particles consist of different substances then we should replace in eq. (5.39) instead of p with an

average particle density < p > .

62

Cosserat Continuum Mechanics, I. Vardoulakis 2009

Dk =

dk 1 = ijk xi v j + ( m ijk yi v j ) V V

(5.41)

The relative velocity vi of the sub-particle at point P ( xi + yi ) with respect to the center of mass S ( xi ) is assumed to be function of its position inside the (REV) and of time. We expand this function in a Taylor series in the vicinity of the center of mass S ( xi ) of the REV,
vi vij (t ) y j + vijk (t ) y j yk +

(5.42)

We can develop a special theory, if we consider only the linear term in the series expansion, eq. (5.42),
vi vij (t ) y j

(5.43)

This assumption is interpreted as a statement for local homogeneity of the microdeformation; i.e. of the deformation inside the (REV), and our demonstration will follow the steps of [6]. In this case from eqs. (5.41), and (5.43) we get
Dk = ijk xi v j (t ) + ijk J il v jl (t )

(5.44)

where J jl is the inertia tensor of the (REV) with respect to its center of mass,
J il =

1 1 ( myi yl ) = p V yi ylV V

(5.45)

For simplicity we assume that on the (REV) only volume external forces are acting. In this case the moment per unit volume of external forces acting on the (REV) is,

k = ijk xi f j + ijk f ji
where,
1 fiV V 1 fij = f iVy j V fi =

(5.46)

( ) (

(5.47)

If we integrate eq. (5.46) over the volume of the continuum body B we get the expression for the total moment of body forces acting on B,

Cosserat Continuum Mechanics, I. Vardoulakis 2009

63

M k (b. f .) =

(V )

dV =
k

ijk i

x f j dV +

(V )

(V )

ijk

f ji dV

(5.48)

In view of eq. (5.22) we recognize the 1st term on the r.h.s. of eq. (5.48), as the moment of body forces. The 2nd term is the contribution of body-couples, that were systematically ignored in the previous derivations since there was no real motivation to introduce such body-couples until this point in the demonstration. Thus we introduce here body couples, k = ijk f ji and eq. (5.48) becomes
M k (b. f .) =

(5.49)

(V )

ijk i

x f j dV +

(V )

dV
k

(5.50)

With this background we may re-write the linear- and angular momentum equations for the considered special micro-morphic continuum; these are,
d vi dV = (V ) fi dV + dt (V ) d Dk dV = (V ) ijk xi f j dV + (V ) k dV + dt (V )

(5.51) (5.52)

where the dots stand for the actions of surface tractions and surface couples. Eq. (5.52) with (5.44) becomes,
d ( ijk xi v j + ijk J il v jl ) dV = (V ) ijk xi f j dV + (V ) k dV + dt (V )

(5.53)

The balance equations for the Cosserat continuum can be derived from eq. (5.53) if we set,
vij = ij = ijk wk

(5.54)

where
wk = D k t k Dt

(5.55)

This assumption means that the considered macro-element is a swarm of sub-particles that they all share a rigid body motion: The center of mass of these sub-particles is

64

Cosserat Continuum Mechanics, I. Vardoulakis 2009

translated by the velocity vi and at the same time all the sub-particles rotate around an instantaneous axis with director nk and have an angular velocity w , such that,
wk = nk w motor24 that characterizes the rigid-body motion of the sub-particles,
w( P ) = w( S ) v ( P) = v ( S ) + w P s

(5.56)

Thus, the spin and the velocity of a sub-particle inside the (REV) is given by the v. Mises

wi = vi ijk wk y j

(5.57)

In this case we have that

ijk J il v jl = ijk J il ( jlm wm ) = J km wm = k


follows,

(5.58)

With these results we return to the momentum balance eq. (5.30), that is written now as

(V )

im

Dwm dV = ( ti + k ki + i ) dV Dt (V )

(5.59)

Its local form reads, p pk + kpq pq + k = J km


Dwm Dt

(5.60)

In general curvilinear coordinates the dynamic eq. (5.60) takes the following form,
p k p + ekpq pq + k =

Dwm J km Dt

(5.61)

5.5 Exercise: Balance of angular momentum in curvilinear coordinates

Prove that in curvilinear coordinates eq. (5.31) becomes25,

jij + ei jk
Proof:

jk

Dk Dt

(5.62)

From,

24 25

see sect. 2.2 cf. eq. (5.58)

Cosserat Continuum Mechanics, I. Vardoulakis 2009

65

Mi = = =

( V )

ei x j t k dS + jk

(V )

i jk

x j f k dV +

( V )

mi dS +

(V )

dV
i

( V )

ei x j lk nl dS + jk
ji j

(V )

(
(V )

(V )

i jk

x j f k dV +

( V )

ji n j dS +

+ ei jk + i + ei x j lkl + f k dV jk jk

))

(V )

dV
i

(5.63)26

Li =

i jk

x j v k dV +

(V )

dV
i

(5.64)

and eq. (5.24) we get, ji i jk Di Dv k + ei x j lkl + f k + j + e jk + i dV = 0 jk ) Dt Dt (V (5.65)


q.e.d.
5.6 Exercise: Dynamic equations in plane polar coordinates

In the absence of body forces and body couples the dynamic equations in physical components and in plane polar coordinates are (Figure 5-2),
Dv rr 1 r 1 + + ( rr ) = r r r Dt r r 1 1 Dv + + ( r + r ) = r r Dt r D z rz 1 z 1 + + rz + r r = r r Dt r

(5.66)

(5.67)

Figure 5-2: Dynamic equilibrium in a Cosserat medium

26

For completeness in eq. (5.63) we considered the action of body couples i dV as well.

66

Cosserat Continuum Mechanics, I. Vardoulakis 2009

5.7 Stress power in micro-morhic media

At this point we would like to make a reference to the more general formulation of the stress power that applies for a micro-morphic medium [32]. In this context we define the following kinematic variables: The macro-strain rate

ij =

1 ( i v j + j vi ) 2

(5.68)

the micro-deformation vij , the relative strain-rate ij = i vk vij and the micro-strain rate gradient, ijk = i v jk Based on these definitions we define the stress power as
w(int) = ij ij + ij ij + ijk ijk

(5.69)

(5.70)

(5.71)

The tensor ij is called the macro-stress, the tensor ij is the relative stress and the tensor ijk is the double stress [54]. We specialize now the micro-deformation so as to correspond to a rigid-body rotation (cf. eq. (5.54))
vij = ij = ijk wk

(5.72)

and with that, ijk = i v jk = jkl i vl = jkl il In that case from eq. (5.71) we get,
w(int) = ij ij + ij ij + ijk ijk = ij ij + ij ( i v j ij ) jkl ijk il

(5.73)

= ij ij + ij ( ij + ij ij ) jkl ijk il = ( ij + (ij ) ) ij + [ ij ] (ij ij ) jkl ijk il

(5.74)

If we compare eqs. (4.14) and (5.74) we obtain the following identification among the stress fields defined in the micro-morphic and the Cosserat continuum, respectively,

(ij ) = ij + (ij ) [ij ] = [ij ] il = jkl ijk


(5.75)

Cosserat Continuum Mechanics, I. Vardoulakis 2009

67

6
6.1

Cosserat continuum energetics


Energy balance equation

The total energy E (t ) of a continuum body B is split into two parts: One part that depends on the position of the observer, that is called the kinetic energy K (t ) of the considered body and another part that does not depend on the observer, called the internal energy U (t ) ,

E (t ) = K (t ) + U (t )
translationary motion of the particles, 1 dmv k vk 2 and of the contribution that is due to their spin, 1 dmk wk 2 Thus the total kinetic energy is computed as,
K (t ) =

(6.1)

The kinetic energy of a Cosserat continuum consists of the contribution that is due to the

(6.2)

(6.3)

(V )

2 v v dV +
k k

1 k wk dV 2 (V )

1 1 = v k vk + k wk dV 2 2 (V )

(6.4)

The internal energy is assumed to be given by means of a specific internal energy density function, e ( i , t ) ,
U (t ) =

(B)

edm = edV
(V )

(6.5)

The 1st Law of Thermodynamics requires that the change of the total energy of a material body B is due to two factors: a) the power W ( ext ) of all external forces acting on B in the current configuration, and b) the non-mechanical energy Q , which is supplied per unit time to B from the exterior domain; i.e.:
dE = W (ext ) + Q dt

(6.6)

68

Cosserat Continuum Mechanics, I. Vardoulakis 2009

By eliminating dE / dt from eqs. (6.1) and (6.6) we arrive to the fundamental energy
balance equation27 dU dK + = W ( ext ) + Q dt dt

(6.7)

The work of external forces is computed as follows,


W ( ext ) =
( V )

(t v
k

+ m k wk ) dS +

(V )

(f

vk + k wk ) dV

(6.8)

The influx of energy in the form of heat flow is defined through a heat-flow vector
q i ( k , t ) , that is taken positive if heat flows into the considered body,
Q=

(V )

q k nk dS

(6.9)

We introduce into eq. (6.8) the stress- and couple-stress tensors, according to eqs. (4.36) and (4.38),
W ( ext ) =

( V )

ik

vk + ik wk ) ni dS +

(V )

(f

vk + k wk ) dV

(6.10)

and we apply Gauss theorem,


W ( ext ) =
(V )

(
k

ik i

+ f k vk dV + ik i + k vk dV +
(V )

(V )

ik

vk i + ik wk i dV

(6.11)

Similarly from eq. (6.9) we get


Q=

(V )

q k dV

(6.12)

The l.h.s. of eq. (6.6) is computed by means of eq. (6.1) to (6.5) and Reynolds transport theorem
dE 1 D k D = ( v vk ) dV + 1 Dt ( k wk ) dV + de dV 2 dt (V ) 2 Dt dt (V ) (V )

(6.13)

We remark that the following relations hold,

27

According to Truesdell & Toupin [50][50] the first formulation of the Energy balance Law is due to

Duhem (1892).

Cosserat Continuum Mechanics, I. Vardoulakis 2009


k k 1 D k D ( v vk ) = 1 Dt ( v k gkl vl ) = gkl vl Dv = vk Dv 2 Dt 2 Dt Dt

69

(6.14) (6.15)

Dwk Dk wk = k Dt Dt

The latter will be demonstrated below. With that we get,


Dwk Dk 1 D k 1 Dk wk ) = = wk wk + k ( Dt Dt 2 Dt 2 Dt

(6.16)

By combining eqs. (6.6), (6.11) to (6.16) we get

(V )

(V )

Dv k Dk De vk dV + wk dV + dV = Dt Dt Dt (V ) (V )
ik i

+ f k vk dV + ik i + k wk dV
ik

+ vk i + wk i dV
ik (V )

(V )

(6.17)

(V )

k k

dV

or

(V )

Dt dV
ik Dv k Dk i + fk vk dV + ik i + k Dt Dt (V ) (V )
(V )

De

+ ik vk i + ik wk i dV

wk dV

(6.18)

(V )

k k

dV

If we assume that linear- and angular momentum balance equations, eqs. (5.21) and (5.62), are holding, then we get from eq. (6.18)

(V )

Dt dV = ( e
De
(V )

iml

g lk im wk + ik vk i + ik wk i qkk dV

(6.19)

With
wim = eiml g lk wk = eiml wl

(6.20)

we get

(V )

Dt dV = (
De
(V )

im

(v

mi

+ wim + im wm i qkk dV

(6.21)

In accordance to eq. (4.1) we define the rate of deformation tensor,

70

Cosserat Continuum Mechanics, I. Vardoulakis 2009

ik = vk i + wik

(6.22)

and the rate of curvature tensor,


K ik = wi k

(6.23)

With this notation in accordance to eq. (4.4) we define the stress power in a Cosserat continuum as
P = im im + im K mi

(6.24)

With this definition, from eq. (6.21) we obtain the following local form of the energy balance equation,

De = P qkk Dt

(6.25)

Remark

In order to have the above derivation complete we must prove the validity of eq. (6.15); cf. [6]. We use Cartesian coordinates and eqs. (5.45) and (5.58) as starting points,
* k = J km wm
where J ij is the (symmetric) inertia tensor referred the unit of mass

(6.26)

* J il =

J il =

1 ( myi yl ) m

* ; J ij = J il

(6.27)

In view of the l.h.s. of eq. (6.15) ,eq. (6.26) and the symmetry of the moment of inertia tensor we get,
DJ Dk DJ km Dwm Dwm wk = km wm + J km wk = wk wm + J km wk Dt Dt Dt Dt Dt DJ km Dwm wm + J mk wk = wk Dt Dt

(6.28)

In a Cosserat continuum the material points move like rigid bodies. This means that for a fixed coordinate system, the inertia tensor is,
J ij (t ) = Qin (t )Q jm (t ) J nm (0)

(6.29)

Cosserat Continuum Mechanics, I. Vardoulakis 2009

71

The proper orthogonal tensor Qij (t ) describes the (finite) rotation of the of the material point between its configuration at time t = 0 and time t > 0 . We recall that tensor Qij (t ) fulfills the orthogonality conditions,
Qik Qil = kl

, Qki Qli = kl

(6.30)

Thus,
D * J ij = QinQ jm J nm (0) + Qin (t )Q jm (t ) J nm (0) Dt

(6.31)

If we take the current configuration as reference configuration, then eq. (6.31) yields
D * J ij = Qin jm J nm + in Q jm J nm = Qin J nj + Q jm J im Dt

(6.32)

or
D * J ij = in J nj J im mj Dt

(6.33)

where the tensor


kl = Qkl

(6.34)

is antisymmetric and has the angular velocity vector wk as an axial vector,


0 [] = w3 w2 w3
0 w1

w2 w1 0

(6.35)

or
ij = ijk wk

(6.36)

Thus,
D * J ij = ink wk J nj + mjk wk J im Dt

(6.37)

and
w j = wi ( imk wk J mj + mjk wk J im ) w j Dt = mik wk wi J mj w j + mjk wk w j J mi wi = 2rm ( mik wi wk ) = 0

wi

DJ ij

(6.38)

72

Cosserat Continuum Mechanics, I. Vardoulakis 2009

6.2

Entropy balance

Let be the total entropy of the considered material body B in the current configuration
(t ) = sdV
V

(6.39)

In eq. (6.39) s = s ( xk , t ) , is the specific entropy. Let also T = T ( xk , t ) > 0 be the absolute temperature. We define further the following quantities: a) The Helmholtz free energy is defined as that portion of the internal energy, which is available for doing mechanical work at constant temperature
f = e sT

(6.40)

b) The local dissipation is


DT Df +s Dloc = P Dt Dt

(6.41)

With the above definitions the energy balance eq. (6.25) becomes,

Ds k = q k + Dloc Dt

(6.42)

This equation is called the balance law for local entropy production. The entropy balance law, eq.(6.42), is further worked out by introducing appropriate constitutive assumptions.
6.3 Linear, isotropic Cosserat elasticity theory

For an elastic Cosserat material that is stressed under isothermal conditions, the energy balance equation (6.25) provides the means to compute the rate of the internal energy density function. In Cartesian notation we have,

De = ij ij + ij K ij Dt

(6.43)

Within the frame of a small-deformation theory we assume that the density remains practically constant,

dV (0) = = (0) (1 kk ) (0) 1 + kk (0) dV (0)

(6.44)

Cosserat Continuum Mechanics, I. Vardoulakis 2009

73

where (0) is the density of the material in the initial, unstrained configuration. Thus for small deformations we get,
ij =

D ij Dt D ij Dt

t ij

(6.45)
t ij

K ij = and

De De (0) ij ij + ij ij Dt Dt

(6.46)

We assume that the elastic energy density, w( el ) = (0) e is a function of the 18 kinematic variables, ij and ij , w( el ) = F ( ij , ij ) Then from eqs. (6.48) and (6.46) we get that (6.48) (6.47)

ij =

w( el ) ij

; ij =

w( el ) ij

(6.49)

Within the frame of linear elasticity we assume that w( el ) = F ( ij , ij ) is homogeneous of degree 2 with respect to its arguments. From

ij ij = (ij ) ( ij ) + [ij ] [ij ]


we get that the elastic strain energy is split into three terms, as
( ( w( el ) = w1( el ) ( ( ij ) ) + w2el ) ( [ij ] ) + w3el ) ( ij )

(6.50)

(6.51)

A simple elasticity model arises if we assume that the 1st term on the r.h.s. of eq. (6.46) reflects Hookes law for linear isotropic elastic materials [43], 1 w1( el ) = (ij ) ij 2

(6.52) ; = 2 G 1 2 (6.53)

1 w1( el ) = mm nn + G mn mn 2

74

Cosserat Continuum Mechanics, I. Vardoulakis 2009

thus yielding,

(ij )

w( el ) w1( el ) kk ij = = = 2G ij + 1 2 ij ij

(6.54)

We notice also that both, [ ij ] and [ij ] , are antisymmetric tensors of 2nd order. Thus both are possessing axial vectors, say
* [ ij ] = ijk tk ; [ij ] = ijk k

(6.55)

where according to eq. (4.11)

k = k k
the antisymmetric part of the deformation, thus yielding
ti = 21G i
(1 > 0) [ij ] = 21G [ij ]

(6.56)

Isotropy calls for proportionality between the axial vector of the antisymmetric stress and

(6.57)

and with that 1 ( w2el ) = [ij ] [ij ] = 1G [ij ] [ij ] = 61G k k 2 (6.58)

The couple stress tensor and the gradient of the Cosserat rotation k are decomposed additively also into symmetric and antisymmetric parts,

ij = (ij ) + [ij ] ; ij = ( ij ) + [ij ]


Then the isotropic linear-elastic law for the couple stress reads,

(6.59)

(ij ) = [ ij ]
where

( w3el ) =G ( ij )

( ij )

+ 2 ij kk ) , 2 > 0

w( el ) = 3 = G 23[ij ] , 3 > 0 [ij ]

(6.60)

is a material constant with the dimension of length, called also material or

internal length. Thus

1 ( w3el ) = G 2 [28].

( mn )

( mn ) + 2 ( mm ) ( nn ) + 3[ij ][ ij ] )

(6.61)

Note that the general anisotropic Cosserat elasticity was formulated in a paper by Kessel

Cosserat Continuum Mechanics, I. Vardoulakis 2009

75

6.4

A 2D linear, isotropic Cosserat- elasticity theory

Here we summarize some results from the paper of M. Schfer [44], that pertain to a proposition for a simple two-dimensional, linear elasticity theory for isotropic materials of the Cosserat type. This is a theory of plane stress states. First we assume that the symmetric part of the stress tensor is related to the symmetric part of the deformation tensor (i.e. to the symmetric part of the displacement gradient, that is identified with the infinitesimal strain tensor), trough the usual equations of planestress, isotropic Hooke elasticity,
1 ( 11 22 ) E 1 22 = ( 22 11 ) E 1 12 = (12) 2G

11 =

(6.62)

where

ij =

1 ( ui, j + u j ,i ) 2

(6.63)

The antisymmetric parts of the relative deformation and of the stress are also linked by a linear relation,

[12] =

1 [12] , Gc = 1G > 0 2Gc

(6.64)

where according to eqs. (3.86) and (4.13)

[12] = [12]

1 ( 12 21 ) = 2 1 = ( 12 21 ) 2

(6.65)

Due to the isotropy requirement we assume that the couple stress are linked to the curvatures by means of only one additional material constant,

13 = D13 ; 23 = D 23 , D > 0
where

(6.66)

13 = 1 3 = 1 23 = 2 3 = 2

(6.67)

76

Cosserat Continuum Mechanics, I. Vardoulakis 2009

6.5

Examples of simple Cosserat elasticity b.v. problems

6.5.1. Pure bending of a Cosserat-elastic beam We consider a beam, with rectangular cross section, as shown in Figure 6-1. The only

stresses that are considered are the axial stress xx and the couple stress xy . Motivated by the classical beam theory we assume that [44],

xx = cz ; xy = c
where c and c are positive constants.

(6.68)

Figure 6-1: Pure bending of a Cosserat beam

In the considered case the only significant equilibrium equations are,


xy xx =0 ; =0 x x

(6.69)

The stress fields, eqs. (6.68), are equilibrium fields. The elasticity equations and the ansatz (6.68) yield,

xx = xy =

1 c xx = z E E 1 c xy = D D

(6.70) (6.71)

The only surviving compatibility condition is


I 21 =
(2)

xx xy = 0 z

(6.72)

which in turn yields a restriction for the introduced constants,

Cosserat Continuum Mechanics, I. Vardoulakis 2009

77

c c D =0 c= c E D E

(6.73)28

We normalize the material constant D by the Youngs modulus, by setting.


D=

1 E 12

(6.74)

where

is a micro-mechanical length that is in most cases considered to be small, if

compared with the typical geometric dimension of a structure. As we will see below, the factor 1/12 is put for convenience in the computation. With this remark from eqs. (6.71) and (6.74) we get

xy =
and

1 E 2 xy 12

(6.75)

xy =

c E

(6.76)

As in classical beam bending theory, from Figure 6-2 we read


x u x dx dx = z R c 1 = E R

xx
z

1 R

(6.77)

where R is the radius of curvature of the beam. On other hand from eqs. (6.77) and (6.76) we get
y x =

1 R

dx = Rd y

(6.78)

28

In the paper by M. Schfer [44] we find the derivation of stress functions that satisfy equilibrium and compatibility conditions.

78

Cosserat Continuum Mechanics, I. Vardoulakis 2009

Figure 6-2: Pure bending

The total bending moment that is taken by the rectangular cross-section of the beam is computed as
h/2

M=

h / 2 3

xx zbdz +

h/2

h / 2 3 2

xy bdz = cb

h/2

h / 2

h/2

z 2 dz + cb
2

h / 2

dz

bh bh bh =c + cbh = c + 12 12 12

= cI 1 + h

(6.79)

where I is the surface moment of inertia of the rectangular cross-section of the beam,
bh3 I= 12

(6.80)

Then from eq. (6.79) we get


M= EI R

(6.81)

where
2 I = I 1 + h

(6.82)

This is a typical result of Cosserat elasticity theory, meaning that a structure made of Cosserat elastic material is stiffer then the corresponding classical elastic structure.

Cosserat Continuum Mechanics, I. Vardoulakis 2009

79

6.5.2. Annular shear of a cylindrical hole in Cosserat-elastic solid

Figure 6-3: Cylindrical hole in plane strain annular shear

We consider a cylindrical hole of radius R under the action of internal shear as shown in Figure 5-2. Axial symmetry yields to the following equilibrium equations,
d r 1 + ( r + r ) = 0 dr r d rz 1 + rz + r r = 0 dr r

(6.83) (6.84)

Remark

Note that for Boltzmann continua the considered problem is isostatic; i.e. for the determination of the stress field one does need to specify the constitutive equation. Indeed in that case the only valid equilibrium equation is ,
d r 2 + r = 0 dr r

(6.85)

The boundary conditions for the classical problem are,


r = R : r =
r : r <

(6.86) (6.87)

These b.c. admit the solution

80
2

Cosserat Continuum Mechanics, I. Vardoulakis 2009

r r = R

(6.88)

Figure 6-4: Stress state at the the cavity lips in case of a Boltzamann continuum

Figure 6-5: Principal stress trajrectories in case of Boltzamann continuum, indicating the isoststaticity of the considered problem.

As mentioned, the probem of a cylindrical cavity under annular shear is isostatic (Figure 6-4, Figure 6-5). In this case principal stresses exist and their trajectories are logarithmic spirals,
( 1 ) : r = Re xp cot , x = r cos ( + 0 ) , 4 3 ( ) : r = Re xp cot 4
2

y = r sin ( + 0 )

, x = r cos ( + 0 ) ,

(6.89)
y = r sin ( + 0 )

Cosserat Continuum Mechanics, I. Vardoulakis 2009

81

In the case, however, of a Cosseart continuum, none of the above holds. The problem is not isostatic, the solution depends on the constitutive assumptions met fror the stresses and the couple stresses and a boundary layer is forming close to the cavity wall, where Cosserat effects are dominant. Indeed in case of a Cosserat continuum, from sect. 3.6 we get that the following expressions for the deformation measures,
0 rz 1 u u z = 2 r r zz 0

rr sym[ ] = r zr

r z

1 u u 0 2 r r 0 0 0 0
z 0 0 0

(6.90)

0 1 u u asym [ ] = + + z r 2 r 0 z 0 0 r [ ] = 0 0 0 0 0 0

1 u u + r 2 r 0 0

(6.91)

(6.92)

These expressions are combined here with the constitutive equations of linear isotropic Cosserat elasticity, that are derived in sect. 6.3and provide the following set of generalized stress-strain relationships,

( r ) = 2G r = G

du u r dr du u + r dr 2 z

(6.93) (6.94)

[ r ] = 21G [ r ] = 1G

And couple-stress-curvarure relationhips

82

Cosserat Continuum Mechanics, I. Vardoulakis 2009

( rz ) = G 2 ( rz ) = G [ rz ] = G 3[ rz ]
2

d z dr 1 d z = G 23 2 dr

1 2

(6.95)

Thus

r = ( r ) + [ r ] = G (1 + 1 )

du u (1 1 ) 21 z dr r d z dr

(6.96)

and

rz = ( rz ) + [ rz ] = G

1 2

(1 + 3 )

(6.97)

Introducing the above set of constitutive equations into the equilibrium eqs. (6.83) and (6.84) we get a set of two coupled differential equations for the particle displacement in tangential direction u and for the particle rotation, z ,
d 2 u dr
2

d z 1 du u = 2a 2 r dr dr r

(6.98)

and
2

d 2 z

u 1 d z du + 2 2b z = b + r dr r dr dr 2

(6.99)

where
a=

1 ; b=4 1 1 + 1 1 + 3

(6.100)

For 1 = 0 (a = b = 0) the stress tensor is symmetric and eqs. (6.98) and (6.99) become decoupled,
d 2u dr dr
2

+ +

1 du u =0 r dr r 2 1 d z =0 r dr

(6.101) (6.102)

d 2 z
2

The solution of eq. (6.101) and eq. (6.102) is


u = C3 r + C4

1 r

(6.103)

Cosserat Continuum Mechanics, I. Vardoulakis 2009

83

z = C1 + C2 ln r

(6.104)

The boundeness condition at infinity for the particle rotation angle z and for the circumferential displacement u is fulfilled, if C2 = 0 and C3 = 0 . The solution for
C1 0 is physically meaningless, thus we accept the solution u = C4 r

; z = 0

(6.105)

The integration constant C4 is determined from the boundary condition for the shear stress,
r = R : r =

(6.106)

Thus
u = u R ; u= R 2G r

(6.107)

In the uncoupled case (1 = 0 ), the valid solution for the displacement, eq.(6.107), together with the constitutive equation for the symmetric part of the stress, eq. (6.93), yield the classical solution, eq. (6.88). In the general case (1 > 0 ), eqs. (6.99) and eq. (6.98) yield
2

d 2 z + 1 d z 2 dr r dr
> 0 and

2 z = 0

(6.108)

where

2 = 8
Let

1 1 > 0 (1 > 0) 1 + 3 1 + 1

(6.109)

(6.110)

the general solution of eq. (6.108) is given in terms of 0th-order modified Bessel functions

z = C1 I 0 ( ) + C2 K 0 ( )
and from that

(6.111)

84

Cosserat Continuum Mechanics, I. Vardoulakis 2009

d z = C1 I1 ( ) C2 K1 ( ) dz

(6.112)

The extra boundary conditions are given in terms of the particle rotation and/or of the couple stress. In order to introduce these extra boundary conditions within the Cosserat continuum description we resort to the concept of ortho-fiber [21] An ortho-fiber is a rigid bar of length

aligned normally to the surface of the considered Cosserat

continuum body and it is pointing outwards. On the end of this fiber we assume either displacements or tractions are applied thus giving to the surface actions the meaning of v. Mises motors (Figure 6-6). Accordingly we assume here that at the cavity wall the shear stress and the couple stress are prescribed and at infinity the particle displacement and rotation must vanish,
r = R : r = rz r = R :

(6.113)

z 0 = u 0

(6.114)

The sign of the surface couple in eq. (6.113) as follows: As shown in Figure 6-6 the shear traction of magnitude is assumed to be applied on an ortho-fiber of length , thus yielding and equivalent set of surface actions, a surface traction and a surface couple. If the surface traction is positive then the surface couple must negative,

Figure 6-6: Shear traction applied on an ortho-fiber at distance , resulting into a surface traction and a surface couple when transported to the surface.

For large argument we have the following asymptotic expression for the solution, eq. (6.111)

Cosserat Continuum Mechanics, I. Vardoulakis 2009

85

z = C1

exp ( ) 2

+ C2

exp ( ) 2

(6.115)

From eqs. (6.114), (6.113) and (6.115) we get e c1 + e c2 = 0 where


ci = 2 (1 + 3 ) G 1 Ci (i = 1, 2) ; R = R

( * >> 1)

I1 ( R ) c1 + K1 ( R ) c2 = 1

(6.116)

(6.117)

Using Kramers rule, the solution of the system of linear eqs. (6.116) takes the following form,
c1 = c2 = e 2

K1 ( R ) I1 ( R )e

0 ( )

1
K1 ( R ) I1 ( R )e
2

1 K1 ( R )

(6.118) ( )

This means that the valid solution here is the logarithmic one (C1 = 0)

z = C2 K 0 ( )
and with that

(6.119)

rz = C2 G

(1 + 3 ) K1 ( )

(6.120)

The b.c. at the cavity wall ,eq. (6.113) for the couple stress, yields,
C2 =

1 4u 1 (1 + 3 ) R K1 ( R ) 1 4u K 0 ( ) >0 (1 + 3 ) R K1 ( R )

(6.121)

and with that (Figure 6-7)

z =

(6.122)

86

Cosserat Continuum Mechanics, I. Vardoulakis 2009

Figure 6-7: Boundary particle displacement and rotation

For small values of the internal length 0 < << R we have the following asymptotic solution for the particle rotation.

z 4 R

u R

e ( R )

(6.123)

This means that the particle rotations are confined in boundary layer and they decay faster then exponentially with the radial distance from the cavity wall. On the other hand we observe that the particle rotation depends linearly on the ratio of the roughness length scale to the material length scale hoop displacement to the radius of the cavity Eq. (6.98) becomes
d 2u dr
2

, and on the ratio of first-order imposed

1 du u d C K 0 ( ) = 0 ; C = 2aC2 2 r dr r dr

(6.124)

or 1 u = C3 r + C4 2aC2 K1 ( ) r (6.125)

As already explained above the only meaningful solution is the one for C3 = 0 , thus
K1 ( ) 1 C u = C4 2 1 + 1 K1 ( R ) r

(6.126)

where

Cosserat Continuum Mechanics, I. Vardoulakis 2009

87

C = C21 =

41 1 u (1 + 3 ) K1 ( R ) R

(6.127)

The b.c. at the cavity wall ,eq. (6.113) for the shear stress, yields,
1 1 K0 ( R ) C4 = u + CR 2 2 + K0 ( R ) 1+ K1 ( R ) 1 R

(6.128)

and with that


u u(0) u

(6.129)

In this expression with u(0) we denote the classical solution eq. (6.107)
u(0) = u R ; u= R 2G r

(6.130)

and u is the perturbation that stems from the Cosserat terms,


u
2 R ( R ) C R 2 R e + 1 + 1 1 2 C = 2 1 (1 + 3 ) K1 ( R ) G

(6.131)

The perturbation for the displacement contains both hyperbolically and exponentially decaying terms, it is proportional to the classical solution, eq. (6.107), and it scales linearly with the interfacial length scale , ,
R (0) u = u u u + C f ( R , ) R r

(6.132)

6.5.3. Sphere under uniform radial torsion We Consider a spherical body of radius R made of linear-elastic, isotropic Cosserat-type

material that is subjected on its surface to uniform radial-torsional laoding of intensity

rr ( R) = m (6Figure 6-8). We want to analyze its state of stress and the deformation that
this object suffers.

88

Cosserat Continuum Mechanics, I. Vardoulakis 2009

6Figure 6-8: Sphere under uniform surface torsion

In the considered setting the deformations, distortions and torsions are given in pola spherical coordinates as follows:
0 0 0 0 0 asym [ ] = r 0 r 0

(6.133)

d r dr [ ] = 0 0 rr [ ] = 0 0

r
r

0 0

0 0 r r

(6.134)

0 0

(6.135)

The only significant stress and couple-stress components are (Figure 6-9):

= 1G r ; = 1G r
d rr = G 2 (1 + 2 ) r + 22 r dr r d = = G 2 2 r + (1 + 22 ) r dr r

(6.136)

(6.137)

Cosserat Continuum Mechanics, I. Vardoulakis 2009

89

Figure 6-9: Stress state in the element of a sphere under uniform torsion The governing equilibrium equation is: d rr 1 + 2 rr + = 0 dr r

(6.138)

The solution that is acceptable (as being regular at the origin) is (Figure 6-10):

r = C( ) ; =

cosh sinh 2

, =

(6.139)

Figure 6-10: Torsional boundary layer solution In this case the torsion is also confined in a boundary layer. For example the solution for the isotropic part of the torsion reads,

90

Cosserat Continuum Mechanics, I. Vardoulakis 2009

1 ( rr + + ) = 1 G 3 3 sinh = = +O ( 3 )

T =

(1 + 32 )

d r +2 r r dr

= C ( )

(6.140)

At this point we should make a remark concerning the so called the restricted Cosserat Continuum theory. In this theory we assume that the relative particle spin is zero, meaning that the particle rotation coincides with the antisymmetric part of the displacement gradient [34]. This theory, although widely used, it has led to difficulties, since the isotropic part of the couple-stress remains indeterminate [33]. This observation has lead to some controversies that we believe that it has been resolved recently by Froiio et al. [21], who have shown, by using the concept of ortho-fiber, that such a restricted Cosserat continuum is incapable to absorb boundary conditions that refer to the torsional dof. In other words the above example of the sphere under uniform torsion illustrates vlearly the ability of the Cosserat continuum theory to a provide a unique solution for the torsion and the mean torsion!

6.6

Cosserat thermo-elasto-plasticity

As for a Boltzmann Continuum so for the Cosserat Continuum within the realm of smalldeformations elastoplasticity thery the total rate-of-deformation measures are decomposed additively into elastic and plastic part; say
e p ij = ij + ij e p ; ij = ij + ij

(6.141)

For isotropic, linear thermo-elastoplatic materials the small strain-theory is based on the following constitutive equations for the elastic rates-of-deformation,
e ij =

1 1 1 DT ij kl ( kl ) + [ij ] + ( e ) ij ( ik jl + il jk ) 1 2G 2 1 + Dt
1 1 ( ki lj + kj li ij kl ) ( kl ) + 1 [ij ] 2 G 2 3

(6.142)

e K ij =

(6.143)

where ( e ) is the coefficient of elastic thermal expansion and T ( xk , t ) is the absolute temperature field. With a superimposed we denote the Zaremba-Jaumann derivative of a tensor,

Cosserat Continuum Mechanics, I. Vardoulakis 2009

91

DSij d + Sik kj ik Skj S ij = J Sij = dt Dt

; ij =

1 j vi i v j 2

(6.144)

The elastic rate-of-deformation tensors are set dual in free energy to the stress and couple stress respectively29,

DT Df +s Dt Dt

e e = ij ij + ij ij

(6.145)

The local dissipation is DT Df Dloc = P +s Dt Dt and with P = ij ij + ij ij (6.147) (6.146)

we conclude that the local dissipation coincides here with the power of stress and of the couple stress on the plastic part of the rate of the corresponding rate of deformation tensors,
p p Dloc = P p = ij ij + ij ij

(6.148)

We recall here the balance law for local entropy production, eq. (6.42),

Ds p = k qk + P Dt

(6.149)

We assume that the rate of entropy production consists of two terms as follows, Ds = s ( th / e ) + s (i ) Dt effects, such as grain breakage and grain attrition. For the thermo-elastic part we assume the following form: s ( th / e ) = jc DT + Akl (ekl ) Dt (6.151) (6.150)

The first contribution is due to thermo-elastic effects and the second is due to in-elastic

29

see esct. 6.2

92

Cosserat Continuum Mechanics, I. Vardoulakis 2009

In eq. (6.151) jc is the specific heat of the material and Aij is the stress temperature tensor, which gives the stress resulting from a given temperature distribution, when the strain vanishes30. In the isotropic case we have Aij = A ij , A= 2(1 + ) G ( e ) 1 2 (6.152)

where ( e ) is the thermal expansion coefficient of skeleton. From the above constitutive assumptions we get

Ds DT + A ekk ) + s ( i ) = k qk + P p ( = jc Dt Dt DT jc = k qk + P p A ekk ) s (i ) ( Dt

(6.153)

In this expression of the energy balance law, we observe that only a part of the local dissipation is converted to heat. One part of the entropy production goes to thermo-elastic expansion and another is dissipated in other forms than heat, such attrition and breakage of grains. If we neglect these effects, we end up with the simplest form of the energy balance law,

jc

DT = k qk + P p Dt

(6.154)

We consider further Fourier's constitutive law of isotropic heat conduction qk = k F k T where k F is the thermal conductivity of the material. From classical thermodynamics we retain as an expression of the 2nd Law the axiom that heat must flow opposite to the temperature gradient (i.e. from hot regions to cold regions). This means that the entropy production by heat conduction must be nondecreasing (6.155)

con =

1 q T 0 2 k k

(6.156)

30

As an example think of an elastic bar heated and clamped at both ends.

Cosserat Continuum Mechanics, I. Vardoulakis 2009

93

With Fouriers law the entropy production by heat conduction becomes a quadratic form in the temperature gradient, and Ineq. (6.156) reduces simply to a constitutive inequality for the thermal conductivity,

con =

kF T T 0 2 k k

kF > 0

(6.157)

With these assumptions and restrictions eq. (6.154) results in the following heatconduction/generation equation, DT 1 = 2T + ( p + ij ijp ) jc ij ij Dt where (6.158)

kF >0 c

(6.159)

is Kelvins coefficient of thermal diffusion, with dimensions [ L2T 1 ]. Thus for the evaluation of the heat eq. (6.158) we need to specify the plastic part of the rate-of-deformation measures.

94

Cosserat Continuum Mechanics, I. Vardoulakis 2009

Micromechanics of solid granular materials

Granular materials are large assemblies of particles interacting with each other through pair-wise frictional contacts. In this section the emphasis will be laid on solid granular matter, where particle contacts are semi-permanent31, i.e. intergranular contacts exist for finite time intervals. We assume that deterministic solid granular assembly configurations exist for substantial time intervals, making thus time averaging unnecessary to describe the dynamics of the assembly. However, spatial averaging is applied here as a tool for bridging the gap between discrete and continuum realizations.

7.1 Stress and couple stress in granular media


The application of Cosserat continuum mechanics to the description of the mechanical behavior of granular media is traced by many authors to the pioneering work of Oshima [39] and later by Kanatani [27] and Mhlhaus & Vardoulakis [35]. In recent years a discussion could be followed where the pros and cons of the applicability of the Cosserat continuum model to granular materials were vividly debated [2], [1], [29], [30], [3], [4], [19], [10], [15]. The central issue in these papers was the discussion concerning the validity of stress asymmetry hypothesis. Froiio et al. [20] followed closely this debate and tried to provide a platform where the various viewpoints could find their position. This deductive mathematical approach is encouraging us to try to learn more about the basic physics that may govern the mechanics of granular media and try to describe them within the frame of Cosserat Continuum Mechanics. It is obvious that the application of the Cosserat continuum concept will be meaningful if the phenomena that we want to describe run at grain scale. Indeed the analysis presented here is meant for the mathematical description of localization phenomena such as shearbands occurring in the interior of a granular body [35] or interfacial bands, occurring at shear interfaces between hard material and granular material [7], [57].

As opposed to granular gases and granular fluids, granular solids possess granular contacts that survive finite time intervals (not impacting contacts).

31

Cosserat Continuum Mechanics, I. Vardoulakis 2009

95

7.1.1 Definitions

Figure 7-1: (REV) containing set B of particles in contact among each other and other exterior particles Following Bardet & Vardoulakis [2] we consider an (REV) that consists of N subparticles (grains), some of which are subjected to external forces or couples, applied from the exterior of the considered (REV), Figure 7-1. These particles inside the (REV) are grouped in the set B = { p = 1, , N } . The forces and couples acting on the particles of B are reduced at M points that form the set of contact points, C = {cc c = 1, , M } . The subset I C contains the contact points between two particles of B , whereas the subset E C contains the points where external actions are applied, I = c1 , , cM I , E = cM I +1 , , cM C = I M , = I M

(7.1)

Sets I , E and C denote the contact points on particle p corresponding internal actions, external actions and all actions. Sets C , I , E , I , E and C are related as follows: C = C
p B

, C = I E (7.2)

I = I
p B

E = E
p B

96

Cosserat Continuum Mechanics, I. Vardoulakis 2009

The intersections of I and E are either empty or reduced to a single (contact) point, E E = ( ) I I = {ci I } .or. ( ) (7.3)

The particle assembly inside the (REV) is in equilibrium, when each sub-particle or grain is in equilibrium. Let us assume that the action of an internal or an external grain onto the considered grain p is reduced to the dynamic v. Mises motor of a force acting on contact point c , fi (c) f i c and the moment mi c :
f c ( c ) = i c (i = 1, 2,3) i mi

(7.4)

The resultant of forces acting on particle p is the force


fi =

cC

f i c

(i = 1, 2,3)

(7.5)

We transport all forces and moments acting on grain p to a center point 32 of the considered particle, thus obtaining the equivalent total dynamic v. Mises motor,
f ( ) = i (i = 1, 2,3) i mi

(7.6)

where fi ( ) = fi (c) = mi ( ) =

cC

fi c
c j

cC

(m

ac i

+ ijk ( x x j ) f k

(7.7)33

Note that force and moment equilibrium for grain p is expressed by34 0 ( ) = i 0 (7.8)

At this point of the derivation this point should be seen as an arbitrarily chosen point inside the grain Note that in eq. (7.7) summation of repeated lower indices is meant! 34 As an exercise the reader should replace this equilibrium condition by Newtons dynamic equation between the dynamic and the kinematic motor of the considered grain.
32 33

Cosserat Continuum Mechanics, I. Vardoulakis 2009

97

7.1.2 The virtual work equation for a discrete assembly of particles in contact
The virtual kinematics of a grain p is given by the virtual rotation i of it along an axis passing through a point and the virtual displacement ui of that point35. If we select the center , then from eqs. (7.7) and (7.8) we get fi ui( ) + miac + ijk ( x cj x ) f k c i = 0 j cC

(7.9)

or
cC

( f u + (m
c i

( ) i

ac i

+ ijk ( x cj x ) f k c i = 0 j

(7.10)

This equation holds for all grains in B , thus,

( f u + (m
c B cC i ( ) i

ac i

+ ijk ( x c x ) f k c i = 0 j j

(7.11)

We note that the double sum over C and B can be split into two separate sums over I and E , respectively. In addition we observe that for any two grains p and p in contact at point c we have from Newtons 3rd law that,
f i c = f i c = f i ( , ) c mi c = mi c = mi( , ) c

(7.12)

Thus from eq. (7.11) we get

W (int) = W ( ext )
where

(7.13)

W (int) = f i ( , )c ( ui c ui c ) + mic (i i )
cI

(7.14) (7.15)

W ( ext ) = ( fi e uie + mieie )


eE

A silent assumption made here is that in granular media particle displacement and particle rotation are independent degrees of freedom.

35

98

Cosserat Continuum Mechanics, I. Vardoulakis 2009

Figure 7-2: Relative displacement of two grains in contact


The relative displacement and relative rotation of two homothetically rotating grains and in contact is (Figure 7-2)
c c uic = ui c ui c = ui ui + ijk ( xk xk ) ( xk xk ) j j

ic = i i

(7.16)

or
c uic = ( ui ui ) + ijk ( xk xk ) + ijk ic ( xk xk ) j

i = i + ic

(7.17)

7.1.3 Compatibility in the discrete setting Let us consider an open line of grains36, as is shown in Figure 7-3. If we apply eq.(7.16)
consecutively, then we get the following expressions for the difference in rotation and displacement between two grains in remote contact,

i3 i1 = ic 2 + ic 3
and

(7.18)

ui3 ui1 = uic (2) + uic (3)

c c c 3 c + ijk 1 ( xk (2) x1 ) + 2 ( xk (3) xk (2) ) + 3 ( xk xk (3) ) j k j j

(7.19)

36

usually termed also a granular column.

Cosserat Continuum Mechanics, I. Vardoulakis 2009

99

Figure 7-3: Open line of homothetically rotating grains


In general for a column of N -grains we shave,

i( N ,1) = iN i1 = ic ( +1)
=1

N 1

(7.20)

and
c c ui( N ,1) = uiN ui1 = uic ( +1) ikj ( xk ( +1) xk ( 1) ) j N 1

=1

(7.21)

where
c c xk (1) x1 , xk ( N +1) xkN k

(7.22)

Following a remark by Satake [41], eqs. (7.20) and (7.21), should be seen as the discrete manifestation of the integrability, holding for a Cosserat continuum, eqs. (3.43) and (3.44):
i( N ,1) = iN i1 = ic ( +1)
=1
N 1 ( P2 , P ) 1 i

= i ( P2 ) i ( P ) = ki dxk 1
P 1

P2

(7.23)

and

100

Cosserat Continuum Mechanics, I. Vardoulakis 2009


P2 P2

( P2 , P ) 1 i

= ui ( P2 ) ui ( P ) = ki dxk ikj j dxk 1


P 1 P 1

c c ui( N ,1) = uiN ui1 = uic ( +1) ikj ( xk ( +1) xk ( 1) ) j N 1 N 1

(7.24)
=1 =1

Satakes analogy, displayed above, allows us to identify: a) the Cosserat continuum rotation as that kinematical property of the continuum that is meant to reproduce the particle rotation, b) the relative deformation of the Cosserat continuum as measure for relative or inter-particle displacement37, and c) the curvature of the Cosserat continuum as the measure for the relative inter-particle rotation. The discrete and the continuous realization of the relative displacement and relative rotation between two points are given by eqs. (7.23) and (7.24). If these relative motions are path independent then we are dealing with a compatible deformation. In particular the relative motions in a compatible deformation should vanish, if evaluated in a closed loop. This is not always the case, in granular media. To demonstrate this we consider the paradigm of the planar, 3-rain circuit of Figure 7-4.

Figure 7-4: Three-particle assembly of two homothetically and one antithetically rotating particle, forming two rolling contacts and one sliding contact.

37

cf. eq. (7.17) and sect. 7.3

Cosserat Continuum Mechanics, I. Vardoulakis 2009

101

For simplicity we assume that the grains are circular rods of equal radius Rg and that grains (1) and (2) are spinning homothetically, grain (3) is spinning antithetically, all with the same strength . We see immediately that in this constellation provides two pure rolling contacts (rc) at c1 and c3 , and a pure sliding contact ( sc) at c2 . We note that the relative displacement between to neighboring grains is null across pure rolling contacts. For this circuit we compute,
(2,1) 1 3 = 32 3 = = 0 (3,2) 3 3 = 3 32 = = 2 (1,3) 1 3 3 = 3 3 = ( ) = 2

(7.25)

3( , ) = 0
cycl

and
(2,1) c 1 c 1 1 u2 = u2 (2) 2 k 33 ( xk (2) x1 ) = 2 Rg 2133 ( x1c (2) x1 ) = 2 Rg + Rg = Rg k (3,2) c c c u2 = u2 (3) 2 k 332 ( xk (3) xk (2) ) = 0 21332 ( x1c (3) x1c (2) ) = + (+ )( Rg 2 Rg ) = Rg (1,3) c 3 c 3 1 u2 = u2 (1) 2 k 33 ( x1 xk (1) ) = 0 2133 ( x1 x1c (1) ) = + ( )( Rg ) = Rg k ( u2 , ) = Rg 0 cycl

(7.26)
1 c 1 c u1(2,1) = u1c (2) 1k 33 ( xk (2) x1 ) = 0 1233 ( x2 (2) x1 ) = 0 k 2

3 Rg 4 3 3 3 c 3 c Rg ) = u1(1,3) = u1c (1) 1k 33 ( x1 xk (1) ) = 0 1233 ( x1 x2 (1) ) = ( )( k 2 4 4 3 ( u2 , ) = Rg 0 2 cycl (7.27)


c c c c u1(3,2) = u1c (3) 1k 332 ( xk (3) xk (2) ) = 0 12332 ( x2 (3) x2 (2) ) =

This means that incorporation of antithetically rotating particles into our consideration, would mean to extend the Cosserat model to incompatible deformations as is the case for example in Gnthers interpretation of Krners theory of dislocations [23]. In the following we will restrict our analysis to compatible deformations.

102

Cosserat Continuum Mechanics, I. Vardoulakis 2009

7.1.4 Remark on incompatible deformations in granular media The question arises if incompatible deformations are important in the study of the
mechanics of dense granular matter. The answer to this question seems today to be affirmative. To this end we recall an early statement by Oda & Kazama [38] who remarked that: ..that a shear band grows through buckling of columns together with rolling at contacts; it can be said that the thickness of a shear band is determined by the number of particles involved in a single column.... Indeed from the micro-mechanical point of view an important structure that appears to dominate localized deformation in 2D simulations is the formation and collapse (buckling) of grain columns, as this was demonstrated experimentally by Oda and was given a simple theoretical description by Satake [42]. These load-carrying columns belong to the so-called competent grain

fraction introduced by Dietrich [16] and Vardoulakis [52] (cf. also [40], [45]) and their
current length reflects more or less the current shear band thickness. Recently Tordessilas [48] picked up on this matter and pointed that One such unjamming mechanism is the buckling of force chains and associate growth of surrounding voids.This mechanism is characteristically non-affine . The term non-affine in connection to an open line of grains is in our understanding not outside Gnthers original idea of incompatible deformations and Satakes integrability and dislocation concepts. This can be seen in Figure 7-5 taken from [48], where we clearly observe that the line of grains that caries non-affine deformation information includes antithetically rotating grains. Indeed, as we have sketched in Figure 7-6, these columns could contain cross-links between antithetically rotating lines, such as the grain pairs (3,3) and (5, 6) in that figure. These cross-links contain strong rolling contacts, of reduced resistance, if compared to sliding contacts and serve as buckling hinges in case they belong to the strong-force network. To our understanding, the essential feature here is the incompatibility of grain rotation across a grain contact, that is leading to the possibility of an internal instability in the form of an internal (frictional) plastic hinge. Thus shear-banding in the sense of Oda should include the formation of plastic hinges among grains along strong force chains as an internal instability.

Cosserat Continuum Mechanics, I. Vardoulakis 2009

103

Figure 7-5: Picture taken from Tordesillas [48]

Figure 7-6: Two lines of compatible rotations (1, 2,3, 4,5) and (3, 4,5, 6) . If we integrate along the lines (1, 2,3, 4,5) and (1, 2,3,3, 4,5, 6,5) we may end up with different results, reflecting the dislocation that is trapped inside the closed loop (3,3, 4,5, 6,5, 4,3) .Antithetically rotating grains-pairs, such as (3,3) and (5, 6) , are connected with plastic hinges (colored in red) and are leading to buckling of strong force columns (colored here in yellow and light blue).

104

Cosserat Continuum Mechanics, I. Vardoulakis 2009

7.1.5 Example: Buckling of rigid-plastic, frictional hinged mechanism38

Figure 7-7: Buckling of frictional hinged mechanism consisting of two rigid rods
In order to explain grain-column buckling it is not necessary to make any assumption concerning the elasticity of grains. It suffices only to consider rolling friction. In order to demonstrate this statement let us consider the mechanism of Figure 7-7, consisting of two rigid bars of length L connected with a frictional hinge. We may assume that the shear forces acting as reactions at the end-points of the mechanism are limited by Coulombs law,

fs fn
The limit moment that the plastic frictional hinge can sustain is

(7.28)

ml = f n e
is stable as soon as

(7.29)

where e is the rolling friction coefficient, with dimensions of length. The mechanism

f s L < ml
Thus, stability is achieved as soon as

(7.30)

f n L < f n e e > L
The shear-band thickness is then estimated as

(7.31)

d B 2 Lcr = 2

(7.32)

This model problem is a very crude first attempt to expalain the effect of plastic hinges forming in grain columns, that should lead to instabilities like shear banding and micro-structural softening.

38

Cosserat Continuum Mechanics, I. Vardoulakis 2009

105

7.1.6 Equilibrium conditions for compatible virtual kinematics The virtual displacement and rotation of the grains are assumed to be continuous
functions of position of the center-points of the grains and expand them in series of position vector of the center of the particle. We truncate these (test) functions to the second order for the displacement,
ui = ai + bij x + cijk x xk j j

(7.33)

and to the first order for the rotations

i = i + ij x j
where ai , bij , cijk , i , ij are arbitrary coefficients.

(7.34)

The above (continuity) assumption, that allows us to write down eqs. (7.33) and (7.34), restricts the variations of the kinematic variables to compatible sets39. From eqs.(7.16) to (7.34) we get
uic = bij ( x x ) + cijk ( x xk x xk ) ijk j ( xk xk ) j j j j

+ ijk jl

(( x

c k

c xk ) xl ( xk xk ) xl

(7.35) (7.36) (7.37)

ic = ij ( x x ) j j
uie = ui + ijk ( xke xk e ) j e e = ai + bij x e + cijk x e xk e + ijk j ( xk xk e ) + ijk jl xl e ( xk xk e ) j j

where xi e is the position of the center of particle p , where contact e takes place. From eqs. (7.35) to (7.37) we get the following expressions for the virtual work of internal and external actions in the considered (REV),

W (int) = bij fi ( , ) c ( x x ) j j
cI

+ cijk f i
cI cI

( , ) c

(x

xk x xk ) j

j fi ( , ) c ijk ( xk xk )
c c + ji ijk fi ( , ) c xl ( xk xk ) xl ( xk xk ) + m(j , ) c ( xl xl ) cI

(7.38)

39

cf. sect. 3.3 .

106

Cosserat Continuum Mechanics, I. Vardoulakis 2009

W ( ext ) = ai fi e + bij f i e x e j
+ cijk f x x
eE eE eE eE e e e i j k

e + j mej + ijk f i e ( xk xk e )

+ ji

( (m
cI

e j

e + ijk fi e ( xk xk e

) ) )x

(7.39)
e

The virtual work equation (7.13), with eqs. (7.38) and (7.39), applies for arbitrary choice of the coefficients ai , bij , cijk , i , ij . Thus by independent variation of these coefficients we get the following set of algebraic equations:

f
eE cI

e i

=0
( , )c j i

(7.40)
= x e f i e j
eE j k i e e e j k

( x x ) f
j ( , )c cI i j k

(7.41)
x x

f ( x x x x ) = f
eE

(7.42) (7.43)

( x x ) f
cI ijk j j

( , )c k

= mi( e, )
eE

(
cI

ikl

c c fl ( , ) c x ( xk xk ) x ( xk xk ) + mi( , ) c ( x x ) = mi( e, ) x e (7.44) j j j j j eE

where
mi( e, ) = mie + ijk ( x e x e ) f ke j j

(7.45)

is the moment acting that results by transporting the external contact force and couple from point e on particle p to its center . Eq. (7.40) is expressing the equilibrium of external forces that are applied to the whole assembly of particles in the considered (REV). On the other hand from eq. (7.41) we get

( x x ) f
cI ijk j j

( , )c k

= ijk x e f ke j
eE

(7.46)

and with that eq. (7.43) transforms into

(m
eE

( e , ) i

+ ijk x e f ke ) = 0 j

(7.47)

or the moment equilibrium equation for all external actions on the considered (REV),

Cosserat Continuum Mechanics, I. Vardoulakis 2009

107

(m (m
eE eE

e i e i

+ ijk ( x e x e ) f ke + ijk x e f ke = 0 j j j + ijk x f


e e j k

)=0
j e i e e c + inm xn f m )x e ikl f l ( , ) c ( x x )xk j j j cI

(7.48)

If we combine eqs.(7.42) and (7.44) we obtain,

m ( x x ) = ( m
cI ( , )c i j eE

(7.49)

We summarize below the set of equations that we derived by applying the virtual work equation on an (REV) of particles that are in a state of static equilibrium under the action of external forces and couples,

f
eE eE

e i

=0
e i

(7.50) (7.51) (7.52) (7.53) (7.54)

(m
cI

+ ijk x ej f ke ) = 0
( , )c j i

( x x ) f
j cI j k j k

= x e f i e j
eE ( , )c

( x x x x ) f
i cI ( , )c i j j

= fi e x e xk e j
eE e i e e c + inm xn f m )x e ikl f l ( , ) c ( x x )xk j j j cI

m ( x x ) = ( m
eE

7.1.7 A micromechanical definition of average stress and couple stress We consider now a strategy for a transition from the discrete medium to the continuum.

This is by far not a unique procedure, thus having always the character of a working hypothesis. The mathematical limitations of such strategies are discussed in detail by Froiio et al. [20]. The analysis starts from the stress equilibrium equations that apply for the continuum, e.g. eqs. (4.41). We assume the small volume V that is occupied by the (REV) and observe that in that case (as was done already in the discrete medium analysis) the effect of volume forces can be neglected. Thus we assume the existence of a stress field that satisfies the following equations
ij xi = 0 xk VREV

(7.55) (7.56)

ij ni = t j xk VREV

108

Cosserat Continuum Mechanics, I. Vardoulakis 2009

For the computation of a mean value of the stress within the (REV) we follow a standard procedure40: We multiply eq. (7.55) with xk , integrate over V and apply Gauss theorem, 0=

VREV

( ) x dV = ( x )dV
i ij k i ij k VREV VREV

ij i xk dV
(7.57)

VREV

VREV

ij xk ni dS

ij ki dV = 0

or
VREV

kj dV =

VREV

t j xk dS

(7.58)

First we observe that the quantity

ij =

1
VREV

VREV

ij dV

(7.59)

is describing the volume-averaged stress. Secondly we juxtapose eqs. (7.58) and (7.52)

VREV

xk t j dS =

VREV

kj dV

x
eE k

f je = ( xk xk ) f j( , ) c
cI

(7.60)

This identification suggest a formula for the computation of the mean stress inside the (REV) using micromechanical information41,

ij

1
VREV

( x x ) f
cI i i

( , )c j

(7.61)

Eq. (7.61)is a celebrated formula that is attributed to Love42 and has been used by many authors since. Similarly we assume the existence of a couple stress field that satisfies the following equations

ik ni = mk
i ij + imj im = 0

(7.62) (7.63)

From these equations we derive,


40 41

L.D. Landau and E.M. Lifshitz, Theory of Elasticity, Vol.7, sect. 2, p.7, Pergamon Press, 1959. e.g. information stemming from a DEM simulation of a granular medium. 42 A.E.H. Love, A Treatise of the Mathematical Theory of Elasticity, Cambridge University Press, 1927.

Cosserat Continuum Mechanics, I. Vardoulakis 2009

109

VREV

(
i

ij

+ imj im ) xk dV =

VREV

( x )dV
i ij k

ij i xk dV +

VREV

VR EV

imj im xk dV = 0 (7.64)

or
VREV

kj dV = kj dV =

VREV

ij ni xk dV +
m j xk dS +

VR EV

imj im xk dV = 0 imj im xk dV
(7.65)

VREV

VREV

VR EV

We remark that if surface couples and couple stresses are zero then eq. (7.65) reduces to a condition that implies symmetry of stress tensor. In general however, with 0=

VREV

VREV

( ) x x dV = (
i ij k l i kj xl + lj xk )dV = VREV

ij

xk xl ) dV

VREV

ij i ( xk xl )dV
(7.66)

VREV

t j xk xl dS

the last integral on the r.h.s. of eq. (7.65) becomes,

imj im xk dV =

VR EV

VR EV

imj [im ] xk dV =

VR EV

imj [ mi ] xk dV imj ti xk xm dS
(7.67)

VR EV

imj mi xk dV =

VR EV

imj ki xm dV

VREV

and with that


VREV

kj dV =

VREV

m j xk dS +

VR EV

imj ki xm dV

VREV

imj ti xk xm dS

(7.68)

or
VREV

kj

+ jmi xm ki )dV =

VREV

(m

+ jmi xmti ) xk dS

(7.69)

First we observe that the quantity

ij =

1
VREV

VREV

ij dV

(7.70)

is the mean couple stress computed over the considered volume. Secondly we juxtapose eqs. (7.69) and (7.54), that is written as follows,

( x x )( m
cI k k

( , )c j

c e + jmi xm f i ( , ) c ) = ( mej + mij xm fi e )xk e eE

(7.71)

This comparison suggests the introduction of the following definitions fort the computation of the mean couple stress inside the (REV):

110

Cosserat Continuum Mechanics, I. Vardoulakis 2009

1) The mean couple stress referred to particle centroid, introduced by Oda [37]:

kj =

1
VREV

( x x ) m
cI k k

( , )c j

(7.72)

2) The transported couple stress at interparticle contact points, introduced by Bardet & Vardoulakis [2] and Tordessilas & Walsh [49]:
( c kjc ) = ( xk xk )( m(j , ) c + jmi xm f i ( , ) c ) cI

(7.73)

The existence of these two definitions explains in part also the controversy that exists in relation to the statement that couple stresses in granular media are: a) only due to contact couples [37], an assumption that would support definition (7.72), or b) that they are also generated in part also by the contact forces, as in definition (7.73). Both definitions are meaningful and valid. We will demonstrate, however in sect. 7.3 , that indeed the
( transported to the contact point couple stress ijc ) is the one that does work on the rolling

contact.
7.1.8 Example: Computation of the Love stress in a regular hinged lattice For the illustration of Loves formula for the computation of the stress in a continuum

that is supposedly carrying the average stress of an underlying discrete medium, we consider here the example of a regular truss under the action of a force F as shown in Figure 7-8.

Figure 7-8: Regular lattice under shear

Cosserat Continuum Mechanics, I. Vardoulakis 2009

111

Table 7-1: Evaluation of the regular lattice of Figure 7-8

We focus on the central node a and we define an (REV) that includes all connecting rods

(1) to (6) . Loves formula applied in this example will yield a symmetric stress tensor,
because the hinged rods cannot transmit an moments, thus

ij =

VREV

(
c =1

c i

Sc + j
c i

c j

Sic )

(7.74)

In this expression

are the projections of the lengths of the rods on the coordinate axes

and Si are the corresponding components of the forces of the respective rods. The Volume of the (REV) is computed as follows: The central region considered is span by a hexagon whose total surface is
S=

3 3 2

(7.75)

Each rod has a volume


Vrod = A

(7.76)

where
A = D2 / 4

(7.77)

is the cross-sectional area of each rod (assuming that they are equal). Thus the volume occupied by this structure is
VREV = SD =

3 3 2

(7.78)

Note that the solid fraction has a volume

112

Cosserat Continuum Mechanics, I. Vardoulakis 2009

Vs = 6Vrod =

3 2 D 2

(7.79)

and the porosity of the structure is

VREV Vs 1 D = 1 ( D << ) VREV 3

(7.80)

In Table 7-1 we summarize the results concerning the forces carried by the central rods, that are computed using an elastic analysis of the considered truss. The components of the resulting stress tensor, evaluated from eq. (7.74) in the considered Cartesian coordinate system are
0 / 4 0 [ ] = / 4 0 0 0 0 0

(7.81)

where

F 2 D = VREV 3 3

; =

F A

(7.82)

Exercises:

3) Verify the accuracy of Loves formula and the result given by eq. (7.82), by solving numerically the problem of a triangular disc of thickness D , loaded by a concentrated force43 as in Figure 7-8. The disc should be made of poro-elastic material with porosity as given above by eq. (7.80). 4) Solve the same problem as above in sect. , by assuming that 7.1.8 by assuming that all rods are clamped and not hinged. Develop the formulae for the stress and couple stress at collocation point (a ) and verify the result numerical as in Exercise 3).

The problem of wedge loaded with a force at is tip is an ill-posed problem within linear Boltzmann elasticity, thus the linear Cosserat elastic solution is needed here; cf. Bogy, D.B. and Sternberg. (1968). The effect of couple-stress on the corner singularity due to an asymmetric shear loading. Int. J. Solids Structures, 4, 159-174.

43

Cosserat Continuum Mechanics, I. Vardoulakis 2009

113

7.2 Mass and moment of inertia considerations

Within the realm of Cosserat continuum mechanics the smallest elementary unit in granular medium is the grain, which moves in space in good approximation as a rigid body. For example, a spherical grain with radius Rg has a volume 4 3 Vs = Rg 3 The grain is made of mineral with bulk density g and has the mass,
mg = gVs

(7.83)

(7.84)

The statistical (REV) has a volume V and includes grains that do not fill all the available space, but leave some void space, Vv = V free of solid mass,
V = Vs + Vv = Vs + V V= Vs 1

(7.85)

where is the porosity of the granular medium. Thus the mass of (distributed) solid material per unit volume is

mg V

= (1 ) g

(7.86)

On the other hand the moment of inertia of a homogeneous sphere is,


Js =

8 2 5 2 g Rg = mg Rg 15 5

(7.87)

The moment of inertia distributed over the volume V is44


J= Js 2 2 = J ; J = Rg 5 V

(7.88)

This means that even on that nave level of simplification a physical property with dimension of length enters the mathematical description of the material behavior of granular media through the micro-polar inertia of that medium.

44

cf. eq. (6.27).

114

Cosserat Continuum Mechanics, I. Vardoulakis 2009

Let N ( Dg < D) be the cumulative distribution curve (or sieve curve) of the grain diameters per unit mass that appear in a statistical (REV) that is given as function of the grain diameter Dg = 2 Rg . Then the representative material moment of inertia is
< J >= J * ; J * =

1 2 Dg 10

(7.89)

where
D =
2 g Dmax

Dmin

N D 2 dD , N =

dN dD

(7.90)

7.3 Grain scale energy dissipation considerations

Figure 7-9: Picture taken from Cole & Peters [12]

As stated by Cole & Peters [12] (Figure 7-9)the relationship between the contact motions and resisting forces define the micro-scale properties of the medium. The constitutive response of the material at the macro-scale is an emergent property that is the result of the collective response of the aggregate, and depends on the micro-scale properties, the stochastic nature of the particle arrangement and boundary conditions . A similar response is modeled for rolling resistance which is assumed to be proportional to the magnitude of the normal force. Not shown is the torsional mode in which the

Cosserat Continuum Mechanics, I. Vardoulakis 2009

115

relative rotation vector is aligned with the normal axis. The torsional mode is of greatest interest in bonded materials.....

Figure 7-10: Two grain circuit with sliding contact and rolling contact respectively.

As shown in Figure 7-10, the contact of two homothetically rotating grains will involve strong contact sliding and weak contact rolling, whereas the contact of two antithetically rotating grains will involve strong contact rolling and weak contact sliding. In this context we like to quote directly from Tordesillas & Walsh [49]: Johnson45 classified the various sources of rolling resistance to be: (a) those arising from microslip and friction at the contact interface, (b) those due to the inelastic properties of the contacting bodies, (c) those due to the roughness of the rolling surfaces. The latter two may be safely ignored in quasistatic loading of densely packed assemblies: inelastic deformation may be ignored for small particle deformations while surface irregularities only influence rolling resistance in the following two ways. Firstly, they intensify the contact pressure at certain points in the contact area, causing local plastic deformation, but this may be ignored for the same reasons for neglecting inelastic deformation. Secondly, as particles move past each other, energy is lost from impacts between irregularities on the opposing contacting surfaces, but this may be considered negligible in systems undergoing quasistatic deformation. A basic hidden assumption met in earlier studies was that almost all energy dissipation is localized at sliding contacts [31]. In general, however, energy dissipation due to rolling

45

K.L. Johnson, Contact Mechanics, Cambridge Univ. Press, Cambridge, 1985.

116

Cosserat Continuum Mechanics, I. Vardoulakis 2009

cannot be excluded due to micro slip and friction at the contact interface. Thus, in recent numerical simulations, energy dissipation is admitted to rolling contacts as well. This assumption is in line with the Cosserat plasticity theory of granular materials [11], [49].
7.4 The 2-grain circuit of homothetically rotating grains

Figure 7-11: Two-grain circuit: kinematic embedment

Let two homothetically rotating grains of equal radius Rg with a strong sliding contact, as seen in Figure 7-11. The branch vector that connects the centers of the two grains is
( K1 K 2 )i = 2li ;
i

= Rg ni , nk nk = 1

(7.91)

The velocities of the centers of the grains (1) and (2) are denoted by vi(1) and vi(2) , and the grains are rotating homothetically with angular velocities k(1) and k(2) , respectively. At the midpoint c of the center line ( K1 K 2 ) the velocities of the grains are
vi(1c ) = vi(1) + ( (1) Rg n ) = vi(1) + ilk l(1)
i i k k

vi(2 c ) = vi(2) + ( (2) Rg ( n ) ) = vi(2) ilk l(2)

(7.92)

Thus the relative velocity and relative rotation of grain (2) with respect to grain (1) at the contact point c are [11], [49]46:

46

cf eq. (7.16)

Cosserat Continuum Mechanics, I. Vardoulakis 2009

117

vi(2,1) = vi(2 c ) vi(1c ) = vi(2) vi(1) ijk (j2) + (1) j

(7.93) (7.94)

i(2,1) = i(2) i(1)

We assume that the particle velocity is embedded now into a continuous field, such that
vi(1) = vi , vi(2) vi + j vi 2
j

(7.95)

Similarly for the particle spin we assume that

l(1) = wl

l(2) wl + 2 m m wl

(7.96)

With this notation eq. (7.93) yields


vi(2,1) = vi(2) vi(1) ijk (j2) + (1) j vi( 2,1) = 2 j vi
j

k k

+ ikj ( 2w j + 2 m m w j )
m m w j ) k

= 2 ( k vi + ikj w j + ikj = 2 ( k vi + wki + ikj With ki = k vi + wki we get


vi(2,1) = 2 ( ki + ikl K lm
m

(7.97)

mwj )

; K ik = k wi

(7.98)

(7.99)

Similarly we get that

i(2,1) = 2 m m wi

(7.100)

Since the two grains are in contact, we assume that they interact with contact forces and contact couples. The force and the couple acted upon grain (1) by grain (2) are denoted as fi (2,1) and mi(2,1) , respectively; their reactions are the contact force fi (1,2) and the contact couple mi(1,2) that are acted by grain (1) on grain (2) . These force- and couple pairs satisfy Newtons 3rd law, fi (1,2) = fi (2,1) ; mi(1,2) = mi(2,1) (7.101)

The interface at the contact of the two grains is identified as an intergranular surface. This is a continuum material band of vanishing thickness, whose boundaries share the

118

Cosserat Continuum Mechanics, I. Vardoulakis 2009


47

motion of the two adjacent faces of the contact

c (1,2) and c (2,1) . On the faces of this

infinitesimal slip the reactions of the intergranular forces are acting. On the face of the intergranular surface that contains point c (1,2) , with the outer unit normal ni the force fi (1,2) is acting and on the face that contains point c (2,1) with the outer unit normal + ni the force fi (2,1) is acting. The rate of work per unit volume, done by these forces at the considered contact due to sliding is (Figure 7-12) P ( ns ) = 1 (1,2) (1c ) 1 1 ( fi vi + fi (2,1)vi(2c ) ) = V fi (1,2) ( vi(1c ) vi(2c ) ) = V fi (1,2)vi(2,1) V (7.102)

Figure 7-12: Two-grain circuit: strong sliding contact Similarly the rate of work of contact couples at the considered contact due to (weak) rolling is (Figure 7-13), P ( nr ) = 1 (1,2) (1) 1 ( mi i + mi(2,1)i(2) ) = V mi(1,2) (i(1) i(2) ) V (7.103)

In the terminology of Tribology this interface is called the third body. As stated by Godet [22], Interfaces, or third bodies can be defined in a material sense, as a zone which exhibits a marked change in composition from that of the rubbing specimens or in a kinematic sense, as the thickness across which the difference in velocity between solids is accommodated.

47

Cosserat Continuum Mechanics, I. Vardoulakis 2009

119

Figure 7-13: Two-grain circuit: weak rolling sliding contact We assume that the force fi (1,2) and the couple mi(1,2) are generated by a stress field and a couple stress field, which are defined in turn at the center of the considered grain. Thus f i (1,2) = ki ( nk ) S , fi (2,1) = ki ( + nk ) S (7.104)

where S = a 2 is the area of a the face of a cube with unit volume , V = a 3 , and with that S P ( ns ) = ( ki ( nk )vi(1c ) + ki nk vi(2 c ) ) V 1 1 V = ki nk ( vi(2 c ) vi(1c ) ) = ki nk vi(2,1) ; a = a a S or, due to eq. (7.99) 1 P ( ns ) = ki nk 2 ( ni + inl K lm m ) n a 2R = g ( ki nk ) ( ( nn ni ) + inl ( K lm nm ) nn ) a Let 2 Rg a

(7.105)

(7.106)

(7.107)

With the notation ti( n ) = ki nk (7.108)

120

Cosserat Continuum Mechanics, I. Vardoulakis 2009

i( n ) = ki nk i( n ) = K im nm = nm m wi = Dwi
; D = nm m

(7.109)

the expression eq. (7.106) for the work per unit volume done by the grain contact forces becomes, P ( ns ) = ti( n ) ( i( n ) + ikl k l( n ) ) We observe that the compound
(1) (n) v v = (2) = ( n ) + (n) v

(7.110)

(7.111)

is a v. Mises kinematic motor. Moreover the component


(2)

v = (n) + (n)

(7.112)

is dual in energy to the stress vector

= t (n)
Since according to eqs. (7.110), (7.112) and (7.113),
P ( ns ) = v
(1) (2)

(1)

(7.113)

(7.114)

( We postulate the couple-stress field ijc ) , that we call hereafter the contact couple stress,

such that,
( ijc ) = ij ijk l lk

(7.115)

and we define the moment vectors,


( i( cn ) = ijc ) nk

; i( n ) = ki nk

(7.116)

Thus
( (jcn ) = ijc ) ni = ( ij ijk l lk ) ni

= ij ni ijk lk l ni = ij ni ijk i lk nl = (j n ) + jik i tk( n )

(7.117)

We observe that the compound

Cosserat Continuum Mechanics, I. Vardoulakis 2009

121

(1) t (n) t = t = (2) ( n ) (n) + t t

(7.118)

is a v. Misses dynamic motor. Indeed the selection of

(jcn ) = (j n ) + jik i tk( n )

(7.119)

is meaningful, because this corresponds the transport of the traction and the couple defined at the center of the grain to the contact point, according to the transport law of statics,
t ( n ) (c) = t ( n ) ( K1 ) ; t ( n ) ( K1 ) = ki ( K1 )nk

( n ) (c) = ( n ) ( K1 ) + t ( n ) ; ( n ) ( K1 ) = ki ( K1 )nk
.With these remarks we introduce the inner product
P(n) = t v = t v + t v
(1) (2) (2) (1)

(7.120)

( ( = (t
(n)

= t (n) (n) + (n) + (n) + t (n) (n)


(n)

) (

)
)

(7.121)

+ t (n)

(n) + (n) (n) +

t (n) (n)

)
(7.122)

or
P(n) = t (n) (n) + (n) (n)

We showed already that the first on the r.h.s. of eq. (7.122) reflects the work done by the forces due to (strong) sliding, eq. (7.114). The 2nd term corresponds on the r.h.s. of eq. (7.122) corresponds to the work done by the couples at the considered contact due to (weak) rolling,
P ( nr ) =
1 (1,2) (1) ( mi i + mi(2,1)i(2) ) V

S ( ( = ( kic ) (nk )i(1) + kic ) nk i(2) ) V =


2R ( 1 (c) 1 ( ki nk (i(2) i(1) ) = kic ) nk i(2,1) = g kic ) nk a a a
m

(7.123)
m wi

or due to eqs. (7.100), (7.116) and (7.109)


P ( nr ) = ( cn ) ( n )

(7.124)

122

Cosserat Continuum Mechanics, I. Vardoulakis 2009

where the componenets of the couple stress vector at the contact ( cn ) are given by eq. (7.117). Thus
P ( nr ) = ( n ) + t ( n ) ( n ) = t v

(2) (1)

(7.125)

and with that


P ( n ) = P ( ns ) + P ( nr )

(7.126)

7.5 Statistical averaging


As a starting point we consider Cauchys fundamental theorem, that relates the components of stress vector on a plane with unit outward normal ni as linear functions of the stress tensor,
t j = ij ni

(7.127)

Let now ni be the unit normal that characterizes an intergranular contact plane, as this was discussed in the previous section. We select all such contact-plane normal vectors and transfer them parallelly to the center of the unit sphere in R 3 (Figure 7-14). Let E be the corresponding point on the surface of the unit sphere. At this point we attach the stress vector, that derives from eq. (7.127).

Figure 7-14: Mapping of the unit contact plane vectors on the unit sphere

Following this reasoning one could ask for example the question as of what is the mean value of the normal component of the stress vector if one considers all probable normalcontact directions in the considered statistical (REV). Thus we define first the scalarquantity

Cosserat Continuum Mechanics, I. Vardoulakis 2009

123

t ( n ) = ti ni = ki nk ni

(7.128)

and try to compute its mean value


p =< t ( n ) >

(7.129)

The simplest possible model derives from the assumption that the probability distribution of the unit contact-plane normal vectors is uniform. This assumption is rather crude as gar as granular media is concerned, and for realistic modeling considerations should be replaced by suitable anisotropic probability distributions [49]. In case of isotropy, averaging over all contact normals is done on the unit sphere as follows [27], [35], [11]:
< t (n) > =

1 4

0 0

(n) t sin d d =

1 4

t
0 0

(n)

sin d d

(7.130)

where r = 1 , and are the polar, spherical coordinates of point E . We observe that the corresponding Cartesian coordinates of the position vector OE = n on the unit sphere are (Figure 7-15):
n1 = sin cos ; n2 = sin sin ; n3 = cos

(7.131)

Figure 7-15: Position vector on the unit sphere: spherical and Cartesian coordinates.

It can be shown that following identities hold [27]:

124

Cosserat Continuum Mechanics, I. Vardoulakis 2009

< ni >= 0 < ni n j > = 1 4


2

n n
i 0 0

1 sin d d = ij 3

< ni n j nk > = 0 1 1 ( ij kl + ik jl + il jk ) = 15 ijkl 35 < ni n j nk nl nm > = 0 < ni n j nk nl >= < ni n j nk nl nm nn > = 1 1 ( in jklm + jn klmi + kn lmij + ln mijk + mn ijkl ) = 105 ijklmn 357

(7.132)

Thus 1 1 < t ( n ) > =< ji n j ni > = ji < n j ni > = ij ij = kk 3 3 measure for the mean normal traction, 1 p =< t ( n ) >= kk 3 (7.134) (7.133)

We recover here the well known statement that the trace of the stress tensor48 is a

Before we proceed with the statistical interpretation of further stress- and couple stress invariants we return to the expression for the work of contact forces and contact couples done at intergranular contact, eq. (7.122) [49],
P ( n ) = t ( n ) ( n ) + ( cn ) ( n ) = ki nk li nl + ( ij ijk i lk

= ( ki li nk nl + ij K jm ni nm Rg ijk lk K jm nl ni nm )

) )n K
i

jm m

(7.135)

Thus
< P ( n ) >= ( ki li < nk nl > + ij K jm < ni nm > Rg ijk lk K jm < nl ni nm > ) = =

( (

ki

li kl + ij K jm im ) ki + ij K ji )

(7.136)

ki

By comparing this result with eq. (6.24) we conclude that for

48

i.e. the mean normal stress

Cosserat Continuum Mechanics, I. Vardoulakis 2009

125

2 Rg
a

=3 a=

V 2 = Rg S 3

(7.137)

and with that


< P ( n ) >= ki ki + ij K ji

(7.138)

the particular choice of micromechanical variables at the level of intergranular contact has allowed us to recover the stress power of the Cosserat continuum as the average value of the work done by contact forces and contact couples at the third body of strong sliding contact. We may now return to the stress analysis. From eqs. (7.127) and (7.128) we get the expression for the shear stress vector, that is tangential to unit sphere,
ti(t ) = ti t ( n ) ni = ki nk kl nk nl ni

(7.139)

We remark that by introducing the decomposition of the stress tensor in spherical and deviatoric part,

ij = kk ij + sij , skk = 0
we get 1 t ( n ) = sij ni n j + kk 3
t (jt ) = sik ni ( jk n j nk )

1 3

(7.140)

(7.141) (7.142)

Thus from eq. (7.141) we retrieve eq. (7.133), 1 1 1 1 1 < t ( n ) >= sij ni n j + kk = sij ij + kk = sii + kk = kk 3 3 3 3 3 From eq. (7.142) we get
t (jt )t (jt ) = sik ni ( jk n j nk ) snm nn ( jm n j nm ) = sik ni snm nn ( mk nk nm ) = sik snk ni nn sik snm ni nn nk nm

(7.143)

(7.144)

and its average value


< ti(t )ti(t ) >=

4 1 ski ski skp s pk 35 35

(7.145)

Based on this computation we define here the shearing stress intensity as,

126

Cosserat Continuum Mechanics, I. Vardoulakis 2009

5 (t ) (t ) 2 1 < ti ti > = ski ski skp s pk 2 3 6

(7.146)

Remark
We notice that if the stress tensor is symmetric, then the average of the square of the shear stress magnitude is related to the shearing stress intensity [56], 1 21 2 < ti(t )ti(t ) >= ski ski = ski ski = 2 5 52 5 5 (t ) (t ) = < ti ti > 2 or ; = 1 ski ski 2

(7.147)

mean =

2 5

(7.148)

We note that the so-called shearing stress intensity in case of a Boltzamnn continuum, differs but little from the maximum shear stress49,
min ( max ) = 0.87 max ( max )

(7.149)

( We may now repeat the above procedure for the contact couple-stress tensor, ijc ) ,

defined above through eq. (7.115)


( ijc ) = ij Rg ijk lk nl

(7.150)

The corresponding couple stress vector is defined through Cauchys fundamental theorem
( (jc ) = ijc ) ni = ij ni Rg ijk lk nl ni

(7.151)

The normal component


( ( cn ) = (jc ) n j = ijc ) ni n j = ij ni n j Rg ijk lk nl ni n j

(7.152)

( The 1st statistical moment of the contact couple-stress tensor ijc ) is a measure for the

mean torsion at the contact,

49

see sect. 9

Cosserat Continuum Mechanics, I. Vardoulakis 2009

127

1 < ( cn ) >= ij < ni n j > Rg ijk lk < nl ni n j >= kk 3 or 1 mT =< ( cn ) >= kk 3 With
( ( i( ct ) = i( c ) ( cn ) ni = kic ) nk klc ) nk nl ni

(7.153)

(7.154)

(7.155)

( we may also compute the 2nd statistical moment of the contact couple-stress tensor, ijc ) ,

< i( ct ) i( ct ) >=

4 1 1 mki mki mqi miq + Rg2 ( 4smr smr slm sml ) 35 35 35

(7.156)

where mij is the couple-stress deviator,

ij = mij + kk ij
Similarly we define the intensity of deviator couples, as
=

1 3

(7.157)

5 2 1 1 < i( ct ) i( ct ) > = mki mki mqi miq + Rg2 2 2 3 6 6

(7.158)

The above introduced stress- and couple-stress invariants can be used in the formulation of constitutive equations for granular media. In case that someone wishes for example to generalize plasticity models that incorporate in their formulation the effect of the 3rd invariant, then one could consider the computation of 3rd order moments [56] of the
( deviators of the stress tensor ij and the contact couple stress tensor ijc ) . For example

we may introduce the standard deviations


( ( 2 = ti(t ) t (jt ) t (pt ) t (pt ) ij ti( t ) t (jt ) tqt ) tqt ) ij 3 3

1 1 ( ( = i( ct ) (jct ) (pct ) (pct ) ij i( ct ) (jct ) qct ) qct ) ij 3 3


2

(7.159)

Excersise: Compute the above expressions for the statistically meaningful 3rd invariants and see if in case of Boltzmann continuum give simple relations that can be expressed in terms of the Lode angle (see sect. 9).

128

Cosserat Continuum Mechanics, I. Vardoulakis 2009

Cosserat Continuum Mechanics, I. Vardoulakis 2009

129

References
Solids Struct. 40, 13291331

[1] Bagi, K. (2003). Discussion on The asymmetry of stress in granular media. Int. J. [2] Bardet, J.P., Vardoulakis, I. (2001). The asymmetry of stress in granular media. Int. J. Solids Struct. 38, 353367. [3] Bardet, J.P., Vardoulakis, I. (2003a). Reply to discussion by Dr. Katalin Bagi. Int. J. Solids Struct. 40, 1035. [4] Bardet, J.P., Vardoulakis, I. (2003b). Reply to Dr. Kuhns discussion. Int. J. Solids Struct. 40, 1809. [5] Beatty, M. F. (1963). Vector representation of rigid body rotation. American Journal of Physics, 31, (2), 134-135. [6] Becker, E. und Brger, W., Kontinuumsmechanik, Teubner Stuttgart, 1975. [7] Bogdanova-Bontcheva, N. and Lippmann, H. (1975). Rotationssymmetrisches ebenes Flieen eines granularen Modellmaterials, Acta Mech., 21, 93113. [8] Buggisch, H. (173). Herleitung der mechanischen Bilanzgleichungen des CossertKontinuums aus der Energiegleichung und ihrem Verhalten beim bergang zum rotierenden System. ZAMM, 53, T68-T69. [9] Chambon, R., Caillerie, D. and El Hassan N. (1996). Etude de la localisation unidimensionelle laide dun modle de second gradient. C.R. Acad. Sci. Paris, Srie II b, 231-238. [10] Chang, C.S., Kuhn, M.R. (2005). On virtual work and stress in granular media. Int. J. Solids Struct. 42, 37733793. [11] Chang, C.S. and Ma, L. (1991). A micromechanically-based micropolar theory for deformation of granular solids. Int. J. Solids Struct., 28, 67-86. [12] Cole, D.M. and Peters, J.F. (2007). A physically based approach to granular media mechanics: Grain-scale experiments, initial results and implications to numerical modeling., Granular Matter 9, 309-321, DOI: 10.1007/s10035-007-0046-2. [13] Corfdir, A., Lerat, P. and Vardoulakis, I. (2004). A cylinder shear apparatus. ASTM GTJ, 27(5),1-9. [14] Cosserat; E. & F., Thorie de Corps dformables, Paris, 1909.

130

Cosserat Continuum Mechanics, I. Vardoulakis 2009

[15] D Addetta, G.A., Ramm, E., Diebels, S. And Ehlers W. (2004). A particle center based homogenization strategy for granular assemblies. Engineering Computations, 21, 360-383. [16] Dietrich, Th.. Der Psammische Stoff als mechanisches Modell des Sandes. Dissertation Universitt Karlsruhe 1976. [17] Dietsche, A. and Willam, K. (1997). Boundary effects in elastoplastic Cosserat Continua. Int. J. Solids Structures, 34, 877-893. [18] Duschek, A. Und Hochrainer, A., Tensorrechnung in Analytischer Darstellung, Bd. II. Tensoranalysis, Springer, 1970. [19] Ehlers, W., Ramm, E., Diebels, S., DAddetta, G.A. (2003). From particle ensembles to Cosserat continua: homogenization of contact forces. [20] Froiio, F., Tomassetti, G. and Vardoulakis, I. (2006). Mechanics of granular materials: the discrete and the continuum descriptions juxtaposed, Int. J. Solids Structures, 43, 76847720. [21] Froiio, F., Zervos, A. and Vardoulakis, I. (2008). Linear 2nd-grade elasticity revisited, J. of Elasticity, submitted. [22] Godet, M. (1990). Third-bodies in tribology. Wear, 136, 29-45. [23] Gnther, W. (1958). Zur Sataik und Kinematik des Cosseratschen Kontinuums. Abh. Braunschweigische Wiss. Ges., 10, 195-213. [24] Gurtin, M.E., Martins, L.C. (1976). Cauchys theorem in classical physics. Arch. Rat. Mech. An. 60, 305-324. [25] Gurtin, M.E., Williams, W.O. (1967). An axiomatic foundation of continuum thermodynamics. Arch. Rat. Mech. Anal. 26, 83117, Gurtin, M.E., Martins, L.C., 1976. Cauchys theorem in classical physics. Arch. Rat. Mech. Anal. 60, 305324. [26] Hamel, G., Elementare Mechanik, Leipzig u. Berlin, 1912. [27] Kanatani, K. (1979). A micropolar continuum theory for flow of granular materials. International Journal of Engineering Science, 17, 419-432. [28] Kessel, S. (1964). Lineare Elastizittstheorie des anisotropen Cosserat-Kontinuums. Abh. Braunschweig. Wiss. Ges., 16, 1-22. [29] Kruyt, N.P. (2003). Static and kinematics of discrete Cosserat-type granular materials. Int. J. Solids Struct. 40, 511534.

Cosserat Continuum Mechanics, I. Vardoulakis 2009

131

[30] Kuhn, M. (2003). Discussion on The asymmetry of stress in granular media. Int. J. Solids Struct. 40, 18051807. [31] Mehrabadi, M.M. and Cowin, S.C. (1978). Initial planar deformation of dilatant granular materials. J. Mech. Phys. Solids, 269-284. [32] Mindlin, R. D. (1964). Microstructure in linear elasticity. Arch. Rational Mech. Anal. 16, 51-78. [33] Mindlin, R.D. and Eshel, N.N. (1968). On first strain-gradient theory in linear elasticity. Int. J. Solids Struct., 4, 109-124 [34] Mindlin, R. D. and Tiersten, H. F. (1962) Effects of Couple stresses in linear elasticity. Arch. Rational Mech. Anal., 11, 415-448. [35] Mhlhaus, H.-B. and Vardoulakis, I. (1987). The thickness of shear bands in granular materials. Gotechnique, 37, 271-283 [36] Noll, W. (1959). The foundation of classical mechanics in the light of recent advances in continuum mechanics. In: The axiomatic method, with special reference to Geometry and Physics, North-Holland Publishing Co., Amsterdam pp. 266281. [37] Oda, M. and Iwashita, K. (2000). Study on couple stress and shear band development in granular media based on numerical simulation analyses. Int. J. Engineering Sci., 38, 1713-1740. [38] Oda, M. and Kazama, H. (1998). Microstructure of shear bands and its relation to the mechanisms of dilatancy and failure of dense granular soils. Gotechnique, 48, 465-481. [39] Oshima, N. (1955). Dynamics of granular media. Memoirs of the Unifying Study of the Basic Problems in Engineering Sciencs by means of Geometry, Vol. 1, Division D, 563-572. [40] Radjai, F., Wolf, D.E., Jean, M. and Moreau J.-J. (1998). Bimodal character of stress transmission in granular packings. Physics Review Letters, 80, 61-64. [41] Satake, M. (1968). Some considerations on the mechanics of granular materials. In: Mechanics of Generalized Continua (E. Krner, Ed.), Springer, 156-159. [42] Satake, M. (1998). Finite Difference Approach to the Shear Band Formation from Viewpoint of Particle Column Buckling. 13th South Asian Geotechnical Conference, 16-20 November, 1998,Taipei, Taiwan, ROC.

132

Cosserat Continuum Mechanics, I. Vardoulakis 2009

[43] Schaefer, H. (1967). Das Cosserat-Kontinuum, ZAMM, 47 (8), 485-498. [44] Schfer, M. (1962). Versuch einer Elastizittstheorie des zweidimensionalen Cosserat-Kontinuums. Miszellaneneen der Angewandten Mechanik, Akademie Verlag, Berlin, 277-292. [45] Staron, L., Vilotte and Radjai, F. (2001). Friction and mobili-zation of contacts in granular numerical avalanches. Powders and Grains 2001, Kishino (ed.), c 2001 Swets & Zeitlinger. [46] Tejhman , J. and Wei Wu. (1995). Experimental and numerical study of sand-stell inter-faces. Int. J. Num. Anal. Meth. Geomech., 19, 513-536. [47] Timoshenko, S. P. (1921) On the correction for shear of the differential equation for transverse vibrations of prismatic beams. Philosophical Magazine, Sec. 6., 41, 744-746. [48] Tordesillas, A. (2007). Force chain buckling, unjamming transitions and shear banding in dense granular assemblies. Phil. Magazine, 87, 4987-5016. [49] Tordesillas, A. and Walsh, D.C.S. (2002). Incorporating rolling resistance and contact anisotropy in micromechanical models of granular media. Powder Technology, 124, 106-111. [50] Truesdell, C and Toupin, R.. The Classical Field Theories. Vol. III/1, Sect. 240ff, Springer, 1960. [51] Unterreiner, F. and Vardoulakis, I. (1994). Interfacial localisation in granular media. Computer Methods in Geomechanics (Ed. Siriwardane & Zaamn), Balkema, 1711-1715. [52] Vardoulakis I. (1989). Shear banding and liquefaction in granu-lar materials on the basis of a Cosserat continuum theory. In-genieur Archiv, 59, 106-113. [53] Vardoulakis, . and Georgopoulos, .-O. (2005) .The Stress - dilatancy hypothesis revisited: shear - banding related instabilities. Soils & Foundations, 45 (2), 61-76. [54] Vardoulakis, I. and Giannakopoulos, A. (2006). An Example of double forces taken from Structural Analysis, Int. J. Solids Structures, 43, 4047-4062. [55] Vardoulakis, I., Shah, K.R. and Papanastasiou, P. (1992). Modelling of tool-rock interfaces using gradient dependent flow-theory of plasticity. Int. J. Rock Mech. & Min. Sci. and Geomechanics. Abstr., 29, 573-582.

Cosserat Continuum Mechanics, I. Vardoulakis 2009

133

[56] Vardoulakis, I. and Sulem, J., Bifurcation Analysis in Geomechanics, Blackie Academic & Professional, 1995. [57] Zervos, A., Vardoulakis, I., Jean, M. and Lerat, P. (2000). Numerical investigation of granular kinematics. Mechanics of Cohesive-Frictional Materials , 5, 305324.

134

Cosserat Continuum Mechanics, I. Vardoulakis 2009

Appendix: The meaning of the Lode angle in Boltzmann Continuum Mechanics

In Boltzmann Continuum Mechanics we introduce the so-called stress invariant angle of similarity or Lode angle cos 3 s 0 = where
J 2s =

3 3 J 3s , 0 s0 / 3 3/ 2 J 2s 2

(8.1)

1 sij s ji 2

(8.2)

and 1 J 3 s = sij s jk ski 3 are the second and third deviatoric stress invariants . The Lode angle is the angle in the deviatoric plane that defines the position of the stress deviator (Figure 9-1). (8.3)

Figure 9-1: Lode angle in the deviatoric plane

The Lode angle in turn defines in turn the deviation of maximum shear stress from its mean value (Figure 3-1),

Cosserat Continuum Mechanics, I. Vardoulakis 2009

135

1,max 2 3 / 2 5 sin( s ) = = mean mean 2 2,max 3 1 / 2 5 sin( / 3 s ) = = 2 mean mean 3,max 1 2 / 2 5 sin( / 3 + s ) = = 2 mean mean
where (8.4)

mean =

2 2 J 2s = T 5 5

(8.5)

Figure 9-2: Mean and max deviator in the deviatoric plane

We observe that the least deviation between maximum and mean deviatoric stress holds for the cases of triaxial extension and compression

max mean

=
min

5 5 3 15 = 1.37 sin( / 3) = 2 2 2 8

(8.6)

The maximum deviation takes place at s 0 = / 6

max mean

=
max

5 5 1.58 sin( / 3 + / 6) = 2 2

(8.7)

In this case we have also that the intermediate principal stress is equal to the mean of the other two,

136

Cosserat Continuum Mechanics, I. Vardoulakis 2009

s 0 = / 6 3 = ( 1 + 2 ) ( 2 3 1 )

1 2

(8.8)

The case with s 0 = / 6 characterizes approximately the so-called plane-strain states.

Vous aimerez peut-être aussi