Lagrangian dynamics

Newton examined forces. From this, he derived accelerations and such. Lagrange did something dierent.
He examined energy, by using generalized coordinates. And thats what well do in this chapter as well.
1 System congurations and coordinates
1.1 Degrees of freedom
Lets consider a particle in a three-dimensional space. We need three numbers to fully describe this
particle. We thus say that the particle has three degrees of freedom. On the other hand, a rigid body
has six degrees of freedom. (Three due to its position, and three due to its orientation.) For a general
system, the number of degrees of freedom is denoted by NDOF. We usually thus need NDOF numbers
(called coordinates) to describe the system.
1.2 System congurations
Describing the conguration of a system can be done in many ways. (We could use many kinds of
coordinate systems.) However, we want to be able to work with any description of the system. To
accomplish this, we dene generalized coordinates q
i
. The collection of all possible sets of coordinates
(q
1
, q
2
, . . . , q
n
) is called the conguration space. By the way, the formulation of dynamics problems in
terms of generalized coordinates is known as Lagrangian dynamics.
1.3 Transforming coordinates
Once a problem is described in certain generalized coordiantes, it can also be described in other coordinate
systems. For this, we use coordinate transformations, like
q
i
= q
i
(x
1
, x
2
, . . . , x
n
, t) and similarly x
i
= x
i
(q
1
, q
2
, . . . , q
n
, t). (1.1)
The latter part of the above equation is known as the inverse transformation.
2 Generalizing energy, momentum and forces
We have generalized coordinates. It would be nice if we could generalize other parameters as well. Thats
what well do in this part.
2.1 Kinetic energy
Lets examine a system with generalized coordinates q = (q
1
, q
2
, . . . , q
n
). The generalized velocities
q
i
of the system are the time derivatives of the coordinates. In other words,
q
i
=
dq
i
dt
. (2.1)
From the generalized velocities, the kinetic energy, in terms of the generalized coordinates, can be
derived. It can be shown that
T =
1
2
n

i=1
n

j=1

ij
q
i
q
j
+
n

i=1

i
q
i
+ = T
2
+T
1
+ T
0
. (2.2)
1
In this equation, T
2
is the collection of terms with the
ij
coecients, T
1
with the
i
coecients and
T
0
= . If the transformations from the generalized coordinates to the actual coordinates do not depend
on time, then T
1
= T
0
= 0. (This is the case if x
i
= x
i
(q
1
, q
2
, . . . , q
n
), or, equivalently, x
i
/t = 0 for all
i.) In this case, we call the system a natural system.
Sadly, the above relation isnt a very easy one. The coecients
ij
,
j
and depend on the system and
the way in which the generalized coordinates are dened. Finding them requires some skill and work.
2.2 The generalized momentum
The generalized momentum in q
i
direction p
i
can be derived from the kinetic energy. It is dened as
p
i
=
T
q
i
. (2.3)
When calculating this partial derivative, all other variables (including the time t) remain xed.
2.3 Generalized forces
Now lets examine generalized forces Q. To do this, we examine work. We can remember that work
is force times displacement in the corresponding direction. We thus have
W =
n

i=1
F
i
x
i
. (2.4)
By using the transformations to generalized coordinates, we can rewrite this to
W =
n

i=1

j=1
F
j
x
j
q
i

q
i
=
n

i=1
Q
i
q
i
. (2.5)
In this equation, Q
i
is the generalized force in the direction of the coordinate q
i
. It is given by
Q
i
=
n

j=1
F
j
x
j
q
i
. (2.6)
It may be important to note that the amount of work performed does not depend on changes in time.
Lets examine a force that is conservative in the real physical word. The corresponding potential function
is V . (We thus have F = V .) In this case, it can be shown that the corresponding generalized force
is also conservative. In fact, we have
Q
i
=
V
q
i
or, equivalently, Q = V. (2.7)
Note that, to use this function, we rst have to transform V . First, V was a function of the physical
coordinates. (So V = V (x
1
, x
2
, . . . , x
n
).) But, to use the above equation, V has to be a function of the
generalized coordinates. (Thus V = V (q
1
, q
2
, . . . , q
n
).)
3 The generalized equations of motion
3.1 Finding the generalized equations of motion
Newton once stated that F = mr or, equivalently, F = p. This relation doesnt always work with gener-
alized forces. This is because, in generalized coordinates, force is not the time derivative of momentum.
2
(So p = Q.) Instead, we have
p = Q+T, or, in components, p
i
= Q
i
+
T
q
i
. (3.1)
We can also recall that p
i
=
dpi
dt
=
d
dt

T
qi

. Putting things together gives

T
q
i

T
q
i
= Q
i
. (3.2)
This very important relation is known as Lagranges Equations of Motions. (It holds for every
coordinate q
i
.)
3.2 The Lagrangian function
In some cases, we can simplify Lagranges equations of motion. Lets suppose that the force Q
i
acting
on the system is conservative. We thus have a conservative system. So there is a function V (q) such
that Q
i
= V/q
k
. If this is the case, then
d
dt

L
q
i

L
q
i
= 0, (3.3)
where L(q, q, t) = T(q, q, t) V (q) is known as the Lagrangian function.
Sometimes, however, we cant nd a potential function V (q) such that Q
i
= V/q
i
. But, we may be
able to nd a function V (q, q, t) such that
Q
i
=
d
dt

V
d q
i

V
q
i
. (3.4)
The function V (q, q, t) is then known as the generalized potential. If there is such a generalized
potential, then the system is called a Lagrangian system. And, if we again dene L(q, q, t) =
T(q, q, t) V (q, q, t), then equation (3.3) still holds.
3.3 Lagrangian systems
Energy is not always conserved in a Lagrangian system. However, there may be another quantity that is
conserved. We dene the Jacobi energy function h as
h(q, q, t) = L +
n

i=1
q
i
L
q
i
. (3.5)
It can now be shown that
dh
dt
=
L
t
. (3.6)
So, if L does not explicitly depend on the time t, then h is constant. It is thus conserved. If this is indeed
the case, then we have
h = T
2
T
0
+ V. (3.7)
For natural systems, we have T
1
= T
0
= 0 and thus h = T +V . In this case, the mechanical energy T +V
is thus conserved as well.
3
3.4 Ignorable coordinates
Lets consider a Lagrangian system with n degrees of freedom. We suppose that there are m generalized
coordinates q
nm+1
, . . . , q
n
that do not appear in the Lagrangian L. These coordinates are called inactive
or ignorable coordinates. (We will soon see why.) For these coordinates, we have L/q
i
= 0, where
n m < i n. This implies that
d
dt

L
q
i

= 0, or, equivalently,
L
q
i
= C
i
, (3.8)
where again n m < i n. The above equation is known as an integral of motion (a quantity
that stays constant during the motion). The constant C
i
is related to the momentum corresponding to
coordinate q
i
.
We use the above relation, when dening the Routhian function. The Routhian function R is dened
as
R = L +
n

i=nm+1
C
i
q
i
. (3.9)
The ignorable coordinates do not appear in the Routhian function. Instead, the Routhian function
contains the constants C
i
. By using the Routhian function, we can rewrite the equations of motion for
the nonignorable coordinates to
d
dt

R
q
i

R
q
i
= 0, (3.10)
where this time 1 i n m. We now only have n m equations of motion left. We therefore ignore
the equations of motion corresponding to the ignorable coordinates. Once the equations of motion have
been solved, the ignorable coordinates can be found using
q
i
=
R
C
i
or, equivalently, q
i
(t) =

t
t0
R
C
i
d, (3.11)
where n m < i n. By the way, we can also express the Jacobi energy function h as a function of R.
We then nd that
h = R
nm

i=1
q
i
R
q
k
. (3.12)
3.5 Steady motion
A special type of motion is steady motion. In a steady motion, all the nonignorable coordinates are
constant. We thus have q
i
= q
i
= 0 for 1 i nm. Because of this, the equations of motion reduce to
R
q
i
= 0, (3.13)
for 1 i n m. On the other hand, the ignorable coordinates have a constant velocity. We thus have
q
i
= v
i
= constant for n m < i n.
3.6 Disippative systems
Lets examine the forces Q
i
acting on a non-Lagrangian system. Part of these forces Q
lagr
i
can be derived
from a generalized potential function V (q, q, t), according to equation (3.4). However, another part can
not be derived from such a potential function.
4
Examples of forces that cant be derived from a potential are frictional forces. These forces can not be
connected to a potential, since energy is dissipated. In real (physical) coordinates, such forces are usually
described by functions like F
fr
i
= c
i
x
i
. (This relation holds for all i.) From this, we can derive that
the frictional generalized forces Q
fr
i
are given by
Q
fr
i
=
1
2

q
i
n

i=1
c
i
x
2
i
=
D
q
i
. (3.14)
The parameter D is known as Rayleighs Dissipation Function. It is dened as
D =
1
2
n

i=1
c
i
x
2
i
. (3.15)
By using this function, we can rewrite the equations of motion to
d
dt

L
q
i

L
q
i
+
D
q
i
= Q

i
. (3.16)
In this equation, Q

i
denotes the part of the force Q
i
that is not derivable from a potential function or a
dissipation function.
4 Constraints
4.1 Holonomic constraints
Systems can have certain constraints. Constraints reduce the number of degrees of freedom. Lets
examine a system normally having n degrees of freedom. If this system is given c constraints, then the
remaining number of degrees of freedom is NDOF = n c.
There are many types of constraints. So-called holonomic constraints can be written as
f(q
1
, q
2
, . . . , q
n
) = constant or f(q
1
, q
2
, . . . , q
n
, t) = constant. (4.1)
If the constraint does not depend on time (as in the rst relation), then the constraint is scleronomic.
Otherwise, it is rheonomic. We can put holonomic constraints in the so-called dierential form. To
do this, we have to use the chain rule. We then nd that
df =
f
q
1
dq
1
+
f
q
2
dq
2
+ . . . +
f
q
n
dq
n
+
f
t
dt = 0. (4.2)
4.2 Nonholonomic constraints
Some constraints can not be written in the form shown in equation (4.1). These constraints are known
as nonholonomic constraints. There are two important types of nonhomolic constraints. These are
inequalities and nonintegrable dierential expressions. Their forms are, respectively,
g(q
1
, q
2
, . . . , q
n
, t) 0 and A
1
dq
1
+A
2
dq
2
+. . . + A
n
dq
n
+A
0
dt = 0, (4.3)
where the dierential expression (the one on the right) is nonintegrable. By the way, the coecients A
i
dont have to be constant. They can depend on time. Nonholonomic constraints are usually a bit more
dicult to deal with than holonomic constraints. Luckily, they appear less frequently too.
5
4.3 Forces caused by constraints
Lets examine a system. This system has a set of J constraints that can be written as
A
j
q = A
j1
q
1
+ A
j2
q
2
+. . . + A
jn
q
n
= 0, (4.4)
where the index j is between 1 and J. Lets consider one of these constraints. This constraint demands
that the position of the system (in the conguration space) moves along a certain n1-dimensional plane.
To keep the position of the system in this plane, a reacting constraint force R
j
acts on the system.
The only eect/goal of this force is to keep the conguration of the system in the plane. It thus acts
perpendicular to the plane.
Now lets examine a movement q of the system. Due to the constraint, this movement q must be
performed in the n1-dimensional plane. Since A
j
q = 0, the vector A
j
must be perpendicular to the
plane. This implies that R
j
and A
j
have the same direction. We can thus write R
j
=
j
A
j
, where the
Lagrange multiplier
j
is (at the moment) an unknown number.
In the equations of motion, we need to take the reaction forces R
j
into account. We thus rewrite these
equations to
d
dt

T
q
i

T
q
i
= Q
i
+
J

j=1
R
ji
= Q
i
+
J

j=1

j
A
ji
. (4.5)
We now have n + J equations, being n equations of motion, and J constraint equations. We also have
n + J unknowns, being the n coordinates q
i
and the J Lagrange multipliers
j
. To nd the unknowns,
all the equations have to be solved simultaneously.
6