3 Lagrangian formulation

In this section, we will explore a new formulation of classical mechanics that addresses the limitations of
Newtonian mechanics. To reiterate, the need for this new formulation arises from the following challenges
associated with Newton’s laws:

• In a system with constraints, we must simultaneously solve both Newton’s second law and the equa-
tions that describe those constraints.
• The forces of constraint are often unknown, making it difficult to apply Newton’s laws directly in
many situations.
• Newton’s second law is not invariant under coordinate transformations. For example, the equation of
motion for a free particle in a two-dimensional plane, expressed in Cartesian coordinates, is given by:

d2 x d2 y
mx̂ + mŷ = F~ .
dt2 dt2
However, in spherical coordinates, the same equation takes a more complex form:
✓ ◆2 ! ✓ ◆
d2 r d✓ dr d✓ d2 ✓ ˆ ~
m r r̂ + m 2 +r 2 ✓ =F .
dt2 dt dt dt dt

This shows that the form of the equation depends on the choice of coordinates, which complicates
analysis.

To overcome these limitations, we will derive Lagrange’s equation of motion in this section. This
formulation provides a more general and coordinate-independent way to describe the motion of a system
between times t1 and t2 .

3.1 Hamilton’s Principle

Configuration space: Consider a system with f degrees of freedom. We can represent the system in an
f -dimensional space, where the coordinate system consists of f orthogonal axes, labeled Q1 , Q2 , . . . , Qf .
This f -dimensional space is referred to as a hyperspace. In this hyperspace, a single point corresponds to a
specific set of f coordinates, {q1 , q2 , q3 , . . . , qf }, which collectively describe the instantaneous configuration
of the system. This set of coordinates is called the system’s configuration. The f -dimensional space in
which these coordinates reside is known as the configuration space. For example, the configuration space
of a particle moving in three-dimensional space is a 3-dimensional Cartesian coordinate system.
Motion of a system in configuration space: Since the configuration of the system at any instant t is repre-
sented by a single point in the configuration space, it follows that the motion of the system corresponds
to a curve in this space. This curve is a function of the generalized coordinates, with time acting as the
parameter. However, since the generalized coordinates do not necessarily correspond to any physical co-
ordinate system, the trajectory in the configuration space does not necessarily resemble the actual path
traced by the system in real space.
Lagrangian formulation is based on the Hamilton’s principle1 . The principle is valid for systems where
all forces can be derived from a generalized scalar potential that may be a function of coordinates and
velocities. If the potential is explicit function of coordinates only then the system is also conservative.
Consider such a system that evolves from time t1 at point A ⌘ {q1 (t1 ), q2 (t1 ).....qf (t1 )} to another time t2
point B ⌘ {q1 (t2 ), q2 (t2 ).....qf (t2 )}. According to the Hamilton’s principle, the motion will be such that
the line integral
Z t2 ✓ ◆ Z t2
S= T (qi , q̇i , t) V (qi , t) dt = L(qi , q̇i , t)dt , (18)
t1 t1

1 Please read Feynman Lectures on Physics, Vol-2, Chapter 19 for an excellent introduction to this topic.

9
has a stationary value for the actual path of the system. Here, the value of the integral S is called the
action and the functional (function of a function is called a functional)

L(qi , q̇i , t) = T (qi , q̇i , t) V (qi , t) . (19)

is called the Lagrangian of the system. The meaning of “stationary” is that for any path that is infinitesi-
mally close to the original path, the change in the action is
Z t2
S= L(qi , q̇i , t)dt = 0 . (20)
t1

For a holonomic system, the Lagrange equation of motion can be derived from the Hamilton’s principle.

3.2 Variational Calculas

The Hamilton’s principle (20) is equivalent to the mathematical problem in calculus of variation. In one
dimension, the problem of calculus of variation is as follows: suppose there is a path y = y(x) between x1
and x2 . There is a function f (y, ẏ, x) that is defined on the path where y 0 = dy/dx. In the calculus of
variation, one finds the path y = y(x) given the stationary integral
Z x2
J= f (y(x), y 0 (x), x)dx . (21)
x1

Here x acts as the parameter of the path. The meaning of stationary is that J = 0 for any path that is
infinitesimally close to the original path. We can denote a neighbourhood path as y(x) + ↵⌘(x) such that
↵ is a parameter and ⌘(x1 ) = 0, ⌘(x2 ) = 0. To be specific all paths can be written as

y(x, ↵) = y(x, 0) + ↵⌘(x) , (22)

such that ↵ = 0 correspond to the original path. The required properties of the function ⌘(x) is that it is
continuous and nonsingular and has continuous first and second derivative between x1 and x2 .
Our aim is to find the actual path for which the value of the integral (21) is extremum – i.e., for an
infinitesimal deviation of the path from the original path, the deviation J. Since an infinitesimally close
path to the original path is parametrized by a non-zero value of ↵ one can write Hamilton’s principle as
dJ
= 0.
d↵ ↵=0

Di↵erentiating (18) w.r.to ↵ we get (from now onwards we omit |↵=0 sign for simplicity)
Z x2 Z x2 ✓ ◆
dJ df @f dy @f dy 0 @f dx
0= = dx = + + dx .
d↵ ↵=0 x1 d↵ x1 @y d↵ @y 0 d↵ @x d↵
The last term vanishes and the the second term can be written as
Z x2 Z x2
@f dy 0 @f d2 y
0
dx = 0
dx .
x1 @y d↵ x1 @y dxd↵

which can be integrated by parts as

Z x2 x2 Z x2 ✓ ◆
@f d2 y @f dy d @f dy
0 dxd↵
dx = 0 d↵
dx .
x1 @y @y x1 x1 dx @y 0 d↵

Since all the paths have the common end-points y(x1 ) and y(x2 ), and the end points are always fixed, the
first terms is zero in the above line. Hence we get
Z x2 ✓ ◆ Z x2 ✓ ◆
dJ @f d @f @y @f d @f
0= = dx = ⌘(x)dx .
d↵ ↵=0 x1 @y dx @y 0 @↵ ↵=0 x1 @y dx @y 0

10
Since y ⌘ y(x, ↵) are all independent paths, their variations ⌘(x) = dy/d↵ are independent of each other.
From the ”fundamental lemma” of calculus it implies that the terms in the right-hand-side of the parenthesis
vanishes ✓ ◆
@f d @f
=0 . (23)
@y df @y 0

This is known the Euler-Lagrange equation. It is the equation of the path y = y(x) for which (21) is
stationary. In other words, the solution of the EL equation (29) gives the path y = y(x).

11