Physics 430: Lecture 16

Lagranges Equations
with Constraints
Dale E. Gary
NJIT Physics Department
 7.3 Constrained Systems in

General
Lets number a system of N particles as = 1, , N. The positions of
these N particles are r. We say that the parameters q1, , qn are a set
of n generalized coordinates for the system if each position r can
be expressed as a function of q1, , qn and possibly time t,
r (q1 , , qn , t ) [ 1,, N ],
and conversely each qi can be expressed in terms of r and possibly t,
qi qi (r1 , , rN , t ) [i 1, , n].
In addition, we require that the number n of the generalized
coordinates is the smallest number that allows the system to be
parametrized in this way. In three dimensions, the number of
generalized coordinates for N particles is certainly no more than 3N,
and for constrained systems is usually less. For a rigid body of, say 1023
particles, for example, the number of generalized coordinates is n = 6,
three for the position of the center of mass, and three for the
orientation.
For the case of the pendulum we discussed last time, there is one body
(the pendulum bob), and two coordinates (x and y), but there is only
one generalized coordinate, , since r = (x, y) = (l sin , l cos ).
October 26, 2010
 Generalized Coordinates-2
Consider the double pendulum, with two bobs confined to motion in a
plane.
Now we have two particles, four coordinates (x1, y1, x2, y2), but
only two generalized coordinates 1 and 2.
In these two examples, the transformation between Cartesian
coordinates and generalized coordinates did not depend on
time, but here is an example that does.
Consider a pendulum hanging from a car that is undergoing a constant
acceleration a to the right.
Because Lagranges equation was derived assuming that the
coordinates are defined in an inertial frame, we are not allowed to use
x
coordinates defined in the frame of the accelerating car! xs=1/2 at2
However, we can express them relative to the ground. y l
a (given)
In this case, the conversion from Cartesian to
generalized coordinates is
r ( x, y ) (l sin 12 at , l cos ) r ( , t ).
2

Generalized coordinates that do not depend on t are called natural.

October 26, 2010

 Degrees of Freedom
The number of degrees of freedom of a system is the number of coordinates
that can be independently varied, i.e. the number of directions a system
can move in small displacements from any initial configuration.
A simple pendulum has one degree of freedom, while the double pendulum
has two. A free particle has three, while a system of N free particles has 3N
degrees of freedom, i.e. each particle has complete freedom.
When the number of degrees of freedom of a system of N particles is less
than 3N, we say that the system is constrained. A system of free particles
constrained to move in two dimensions has 2N degrees of freedom. Some
further examples: a rigid body has 6 degrees of freedom, a bead on a wire
has 1 degree of freedom, and a particle on a surface has 2 degrees of
freedom.
In each of these examples, the number of degrees of freedom equals the
number of generalized coordinates (and so the number of Lagrange
equations that apply).
A system with this natural-seeming property is said to be holonomic. This
course will only treat holonomic systems, which are easier to solve.

October 26, 2010

 A Non-Holonomic System
You might think that a system that does not have this natural property
must be rare and bizarrely complicated. However, there are some simple
examples of a nonholonomic system. Here is one.
Imaging a rubber ball free to roll (but not slide or spin) on a 2-d surface.
Starting at position (x, y) on the 2-d surface, it can only move in two
independent directions and you might think that only two coordinates are
necessary to completely describe its configuration, the coordinates x and y of
its center.
But consider the followingplace the ball at the origin O, and paint a dot on its
top. Then roll it a distance equal to its circumference c along the x axis, so that
the dot is again on top.
Now roll it a distance c in the y direction to a point P where its dot returns to the
top.
Finally, roll it along the hypotenuse back to the origin. Now the dot is not on
top, even though its position is again at O.
Evidentally, the two coordinates x and y are not enough to uniquely
specify the configuration. In fact, we need three more, the orientation of
the ball.
So 5 coordinates are needed, even though the ball has only two degrees
of freedom. Such a system is nonholonomic.
October 26, 2010
 7.4 Proof of Lagranges
Equations with Constraints
We are now ready to prove Lagranges equations for any holonomic
system. We will prove it for one particle, but it can easily be
extended to an arbitrary number (see Prob. 7.13).
Lets take a particle constrained to move on a surface, so that it has
two degrees of freedom and hence two independent generalized
coordinates q1 and q2.
There are two types of forces on the particleforces of constraint
(whatever forces are keeping the particle constrained), which well
denote Fcstr, and all other forces F. The key is that the forces Fcstr can
do no work on the particle. Note that the Fcstr forces may not be
conservative, but this does not matter, since the Lagrange
equations are not going to include them.
We shall assume that the non-constraint forces do satisfy the
second condition, at least, of conservative forces, i.e. that they can
be derived from the gradientFof
a potential
U (r, t ). energy, U(r, t):

If all forces F are really conservative, then they do not depend on t,

but we do not need to assume this. The total force on the particle is
Ftot = Fcstr + F.
October 26, 2010
 Action Integral Stationary on
Right Path
Consider a constrained particle that moves through two points r1 and r2
at times t1 and t2. We will denote r(t) as the position when the particle is
on the right path and R(t) as the position along any neighboring
wrong path.
For a small displacement (t)Rbetween
(t ) r(t ) the
(t ).right and wrong path, we have
Note that (t) = 0 at the end points r1 and r2, since both paths go through
these points. Note also that r(t) and R(t) are in the surface, so (t) is also.
t2
We denote the action integralS by L (R, R&, t )dt ,
t1

taken along any path R(t) lying in the surface, and by So the
corresponding integral taken along the right path r(t).
We wish to prove that the integral S is stationary when R(t) = r(t), i.e.
when (t) = 0. Another way to say this is that the difference in the S S So ,
integrals
t2
is zero to
first dt , in . L L (R, R&, t ) L (r, r&, t ).
S Lorder
t1
Now where
R (t ) r(t ) (t ), L (r, r&, t ) T U 12 mr& U (r, t ).
2

We can substitute and

October 26, 2010
 Action Integral Stationary on
Right Path-2
This gives & & r &2 Ur ( , t ) Ur ( , t )
L 12 m r
2

mr&& U O ( 2 ).
Putting this back into the integral, and integrating the first term by
parts, we get t2
S L dt mr U dt mr&& dt ,
t2 t2

t1 t1
&& t1
U
where the end-point term is zero, as usual.
m&
By Newtons second law r& Ftot Fcstr F, U F,
and so
t2
S Fcstr dt ,
t1

But recall that is in the surface, while Fcstr is perpendicular to the

Fcstr 0,
surface. Therefore and we have proved that S = 0. We
have thus proved Hamiltons Principle, that the action integral is
stationary at the path that the particle actually follows.
Notice that this is only true for paths in the surface, i.e. consistent
with the constraints. Thus, it is not true for any coordinates x, y, z,
say, but only for our generalized coordinates q1 and q2.
October 26, 2010
 The General Result
For any holonomic system, with n degrees of freedom and n
generalized coordinates, and with the nonconstraint forces derivable
from a potential energy U(q1, , qn, t), the path followed by the system
is determined by the n Lagrange equations
L d L
[i 1,L , n],
qi dt q&i
where L is the Lagrangian L = T U and U(q1, , qn, t) is the total
potential energy corresponding to all of the forces excluding the
forces of constraint.
You might ask what you should do in the case that some force on the
particle is not conservative, i.e. friction. In that case you must modify
the Lagrange equations (see Prob. 7.12), but the result is not elegant
and we will not consider such cases.
We are now going to take a look at a number of examples where the
above Lagrange equations hold. You should try as many examples as
you can, to get a feel for how to do these problems. You WILL see
these sorts of problems on the exam.

October 26, 2010