Vous êtes sur la page 1sur 7

5.

Constraints, generalised coordinates and configuration space


You might have the impression that all problems in mechanics can be reduced to
solving a set of differential equations:
X
mi r̈i = Fei + Fji (5.1)
j

i.e. substitute all the forces acting on a system and churn out the motions. This
is oversimplified. All mechanical systems have constraint forces acting on them to
some degree, i.e. particles in a box.
Holonomic constraints are those in which the equations connecting the coor-
dinates of the particles (and possibly the time) have the form:

f (r1 , r2 , r3 , ....t) = 0, (5.2)

a particle constrained to move along a curve, or a car along a road, are examples
of holonomic constraints. Gas particles in a balloon suffer non-holonomic con-
straints i.e. r2 − a2 ≤ 0, where a is the radius of the spherical balloon.
If the constraints contain the time as an explicit variable they are rheonomous
(e.g. a bead sliding on a moving wire) or non-explicitly they are scleronomous (e.g.
a bead sliding on a rigid curved wire).
Constraints lead to two difficulties: Firstly the coordinates ri are no longer all
independent since they are connected by the equations of the constraint, hence the
equations of motion are no longer independent. Secondly, the forces of the constraint
are not given a priori - they are among the unknowns of the problem and must be
obtained from the solution we seek. This means that there are forces present in the
system that cannot be specified directly but are known in terms of their effect on
the motion of the system.
The first difficulty is solved by the introduction of generalised coordinates.
So far, we have thought in terms of Cartesian coordinates. A system of N particles
free from constraints has 3N degrees of freedom. If there exists holonomic constraints
expressed in k equations of the form given above, we can use these to eliminate k of
the 3N coordinates, leaving a system with n = 3N − k independent coordinates and
n degrees of freedom.
We can introduce new independent variables qa = q1 , q2 , ...., q3N −k in terms of
which the old coordinates r1 , r2 , ...., rN are expressed by equations of the form:

ri = ri (q1 , q2 , ...., q3N −k , t) (5.3)

which implicitly contain the constraints. These are transformation equations from
the set of (r) variables to the qa set. You can also think about these equations as

– 22 –
parametric representations of the (r) variables. The generalised coordinates q1 , ..., qn
where n = 3N − k completely define the system. The derivatives q˙a are known as
the generalised velocities.
For example, the generalised coordinates of the double pendulum may be the
angles θ1 , θ2 that each pendulum makes to the vertical. This reduces the number of
coordinates in the standard cartesian description from four x1 , y1 , x2 , y2 to just two.
Generalised coordinates do not have to be linked to spatial surfaces but could have
the dimensions of energy, angular momentum for example. As another example,
consider a rigid body like a chair composed of 1024 atoms. This system can be
specified by the three coordinates of the centre of mass and three angles which define
the rotation of the body with respect to the centre of mass. These 6 parameters
determine the position of the chair completely - the coordinates of all the component
atoms can be expressed in terms of these 6 parameters.
For non-holonomic constraints, the dependent coordinates cannot be eliminated.
e.g. a football rolling on a surface can be described by angular coordinates (to
set the orientation of the football) and a set of coordinates to describe the point
of contact. The constraint of rolling connects these two sets of coordinates - they
are not independent. Problems involving holonomic constraints are amenable to a
formal solution, but there is no general way to tackle non-holonomic problems - we
will come back to this point later. However, most real-life problems can be solved
without worrying about constraints.

5.0.4 Example - nonholonomic constraints on a rolling penny (on the


board only)
5.1 Configuration space
If we associate with the three numbers x, y, z with a point in 3-d Cartesian space,
there is no reason that we can’t do the same with the n values of qa giving the
coordinates of a point in configuration space. The above example of a chair, with
its near infinity of mass points is symbolized as a single point of a six dimensional
space. There is a mapping between the component particles of the chair and the 6-D
space. This is often called the C-point and the motion of the object in this space
defines a C-curve.
The methods of Riemannian geometry are employed to study the motion of
bodies in this new ”configuration space”. Geometry and mechanics are connected
through the concept of inertial mass. We have seen that the kinetic energy of a
system can be replaced by one single particle which can be considered as moving in
an n-dimensional space.

6. D’Alembert’s principle and Lagrange’s equations


Lagrange’s equations can be derived in several ways. It is such an important topic

– 23 –
Figure 4: The path of particles in real space (on the left) and in configuration space (on
the right).

that we will look at two of these. Here we consider tiny (virtual) displacements of a
system from its instantaneous state - a differential principle. Later we discuss a more
general and powerful integral principle that allows us to obtain Lagrange’s equations
by considering the entire motion of the system between times t1 and t2 . The latter
approach is perhaps more general since it can be used in GR, quantum mechanics,
electrodynamics etc. So before delving into the wonderful concept of least action,
lets look at virtual displacements.

6.1 Virtual work


Consider a virtual (arbitrarily small, infinitesimal) displacement to a system δr.
Note that this is not the same as the actual displacement that might occur in time
dt during which the forces and constraints may change. Suppose also that the system
is in equilibrium such that total force on each particle vanishes Fi = 0, then the dot
product Fi · δri also vanishes. This latter term is the virtual work of the force in the
direction of the displacement. The sum of these vanishing products over all particles
must also be zero.
X
Fi · δri = 0. (6.1)
i

Nothing too profound yet. Introduce some dynamics via Newton:

Fi = ṗi , (6.2)

which D’Alembert cleverly wrote as:

Fi − ṗi = 0. (6.3)

which states that the particles in the system will be in equilibrium under a force
equal to the actual force plus a ”‘reversed effective force”’ −pi , sometimes called the
inertial force.
Part of the force Fi must be due to the constraints such that

Fi = Fnc
i +f
constraint
(6.4)

– 24 –
where Fnc i are the non-constraint forces. In general the constraint force does no
work since the motion is perpendicular to the force f constraint · δri = 0. (One obvious
exception being friction.) Thus we can write i (Fnc
P
i − ṗi ) · δri = 0.
Soon we will transform this principle into an expression involving virtual dis-
placements of the generalized coordinates which are then independent of each other
for holonomic constraints. This allows us to set each of the coefficients of δqa to zero.
This is more than just a reformulation of Newton’s equation. It is the expression
of a principle. We know that the vanishing of a force in mechanics means equilibrium.
Equation (6.3) says that the addition of the force of inertia to the other acting forces
produced equilibrium. But this means that if we have any criteria for the equilibrium
of a system, we can extend that criterion to a system which is in motion. Thus
dynamics can be reduced to statics! And of course, motion is relative and we can
introduce a coordinate system that moves with the system. If the reference system
is accelerated motion, this is like having a system with external forces which have to
be added to the existing forces. For example, if the system is in uniform acceleration
then the apparent force is that of a uniform gravitational field.
Lanczos states the d’Alembert’s principle as: ”The virtual work of the forces of
reaction is always zero for any virtual displacement which is in harmony with the
given kinematical constraints”. i.e. δW = F1 · δr1 + F2 · δr2 + ... = 0. He argues
this is the only principle of analytical mechanics and that this can’t be derived from
Newton’s laws. We will see later that Hamilton’s principle is derived from the above.
R
Work is defined to be the line integral W ≡ F · dr, the integral of the force
along the direction of displacement. Virtual work is defined in the same way except
the displacement is virtual such that

δW = F · δr. (6.5)

or

δW − ṗ · δr = 0. (6.6)

Consider an infinitesimal displacement (δqa ) to one of the degrees of freedom


(qa0 s).
(Recall that each coordinate can be varied independently of the other gener-
alised coordinates.) Consider the effects on a part of the system, the ith part, whose
position ri will change in space by an amount
∂ri
δri = δqa . (6.7)
∂qa
The virtual work is the sum of the virtual work done by each generalised coordinate
variation, so we have
!
X X X
δW ≡ Fi · δri ≡ (6.8)
i i i=1,N

– 25 –
from which we can obtain partial derivatives connecting the virtual displacements of
the different parts with the virtual displacement of qa :
!
X ∂ri
δWa ≡ Fi · δqa (6.9)
i
∂q a

where
!
X X X
δW = δWa ≡ (6.10)
a a a=1,n

Since the constraint forces always act to maintain the constraint, they point
perpendicular to the movement of the parts of the system. i.e. they do not contribute
anything to the virtual work. Equations (6.9) and (6.10) can be rewritten in terms
of the non constraint forces as a sum over all parts of the system followed by a sum
over the n generalised coordinates:
!
X X ∂r i
δW = Fnc
i · δqa . (6.11)
a i
∂qa

The generalised force FG associated with the ath degree of freedom is defined
X ∂ri δW
FG ≡ Fi · = (6.12)
i
∂qa δqa

where Fnci are the non constraint forces. In this last step I have suppressed the
superscript nc on Fi so from now on we only refer to forces that do virtual work.
The dimensions of FG depend on the dimensions of qa which may not be the same
as ordinary force.

6.2 Generalised velocities


Generalised velocities are the first time derivatives of the generalised coordinates qa .
qa (t) and q̇a (t) are independent variables. The positions ri will change as qa varies.
If one particular qa is varied and the other (n − 1) qa0 s are held constant, the partial
derivative can be written:
∂ri
. (6.13)
∂qa
Relate the real velocity of the ith particle, ṙi , to the generalised velocities q̇a by
∂y
applying the chain rule ∂u
∂s
= ∂u ∂x
∂x ∂s
+ ∂u
∂y ∂s
+ ∂u ∂z
∂z ∂s
to our generalised coordinates,
equation (5.3):
n
X ∂ri ∂ri
vi = ṙi = q̇a + (6.14)
a=1
∂qa ∂t

– 26 –
The last term is absent if we have scleronomic constraints. Taking the partial deriva-
tive with respect to q̇a gives us:
∂ ṙi ∂ri
= (6.15)
∂ q̇a ∂qa
This useful relation is as if the ”dots” can be canceled out.

6.3 Kinetic energy as a function of the generalised coordinates and veloc-


ities
The kinetic energy is additive:
N
1X
T = mi ṙi · ṙi = T (q1 , ..., qn , q̇1 , ..., q̇n , t) (6.16)
2 i=1

where the summation is over all parts of the system. The kinetic energy T is only a
function of the 2n variables representing the generalised coordinates and their first
time derivatives, the generalised velocities.
Using the chain rule we can write the partial derivatives of T :
∂T X ∂ ṙi X ∂ ṙi
= mi ṙi · = pi · (6.17)
∂qa i
∂qa i
∂qa

∂T X ∂ ṙi X ∂ri
= mi ṙi · = pi · (6.18)
∂ q̇a i
∂ q̇a i
∂qa

The second part of this equation is only true for holonomic constraints that allow
”dot cancellation”.
The virtual work and thus the generalised force can be related to the partial
derivatives of the kinetic energy. This a crucial step in the development of analytical
mechanics. Write down the total time derivative of the above equation using the
chain rule:
  X
d ∂T ∂ri X ∂ ṙi
= ṗi · + pi · (6.19)
dt ∂ q̇a i
∂qa i
∂qa

Now the last step is to replace the momentum terms by the force on the index i. The
first term is simply the generalised force, FG (c.f. equation 6.12) and substituting
from equation 6.17 we have:
 
G d ∂T ∂T
F = − (6.20)
dt ∂ q̇a ∂qa
which gives n equations of motion for a = 1, ....., n. This is the basic formula of
analytical mechanics which gives a set of differential equations for the generalised

– 27 –
equations of motion for any holonomic system. Only a single scalar function of the
generalised coordinates and velocities, the kinetic energy and generalised force are
needed to use these n equations.
It is unnecessary to find the constraint forces. Newton’s laws were not used
explicitly but they are contained within equation (6.20). The advantage is that the
number of unknowns equals the number of degrees of freedom, unlike for Newtonian
vector mechanics in which the constraint forces are additional unknowns and more
equations must be solved.
If the forces are conservative forces we can eliminate the need to find FG explic-
itly.

– 28 –

Vous aimerez peut-être aussi