Chapter 6 Calculus of Variations Elementary Calculus Calculus of Variations

❑ We are given function ❑ We are given functional

Introduction: 𝑓(𝑥). 𝑏
𝐼(𝑦(𝑥)) = ∫ 𝑓(𝑦, 𝑦 ′ , 𝑥)𝑑𝑥 .
❑ The prior chapters have focussed on Newtonian 𝑎
❑ To find ❑ To find
approach to classical mechanics, which is based on
stationary points stationary functions
vector quantities like force, momentum, and
of the of the
acceleration. Newtonian mechanics leads to second-
function functional
order differential equations of motion. Newton relied we usually solve
we solve
heavily on geometrical arguments. 𝑑𝑓
❑ From here on we are starting a new approach to =0 differential equations
𝑑𝑥
classical mechanics called analytical mechanics (Euler-Lagrange equation)
(Lagrangian and Hamiltonian mechanics). The for 𝑥. for 𝑦(𝑥).
mathematics required to study this new version of ❑ These values 𝑥 ❑ This 𝑦(𝑥) minimizes or
mechanics is called calculus of variations. minimize or maximize maximizes the given
❑ Although variational calculus was developed originally the given function. functional.
for classical mechanics, now it has grown to be an ❑ (Further testing is ❑ (Further testing is
important branch of mathematics that can be used to required to determine required to determine
whether 𝑥 is a minimum whether 𝑦(𝑥) minimizes
formulate not only almost every branch of physics like
or maximum.) or maximizes 𝐼.)
relativity, electrodynamics, quantum mechanics, optics,
statistical mechanics, quantum field theory etc. (so
variational methods have given unity to physics), but 6.1 Two Examples:
can handle many other fields outside of physics like
economics, medicine, biology and so on! ❑ Let’s look at a couple of examples:
❑ The calculus of variations involves finding an extremum
◼ The shortest path between two points
(maximum or minimum) of a quantity that is expressible
as an integral. ◼ Fermat’s principle (light follows a path that is an
❑ The brachistochrone problem, formulated in 1696 by extremum)
Johann Bernoulli (a student of Leibniz), marks the
beginning of the calculus of variations. He solved the
brachistochrone problem which involves finding the
path for which the transit time between two points is The Shortest Path between Two Points
the shortest. ❑ What is the shortest path between two points in a
❑ The rigorous (thorough) mathematics of calculus of
plane? You certainly know the answer—a straight
variations was developed by Euler (student of
line—but you probably have not seen a proof of this—
Bernoulli).
the calculus of variations provides such a proof.

Functional ❑ Proof: (Our task is to find path 𝑦(𝑥) that has the
shortest length between points 1 and 2 and to show
❑ Function: A function is a mapping from one number (or that it is in fact a straight line.)
set of numbers) to another value e.g.
𝑓(𝑥) = 𝑥 2
is a function. It maps e.g. 𝑥 = 2 to 𝑓(2) = 22 = 4
❑ Functional: A functional is a function of a function and
depends on the entire path of one or more functions
rather than a number of discrete variables.
Functional is a mapping from a function (or a set of
functions) to a value. For example, If 𝐽 depends on
𝑓(𝑥) then 𝐽 is a functional and it can be written as
𝐽(𝑓(𝑥)).
𝜋
𝐽(𝑦(𝑥)) = ∫0 [𝑦(𝑥)]2 𝑑𝑥 is a functional.
𝜋
If 𝑦(𝑥) = 𝑥: then 𝐽(𝑦(𝑥)) = ∫0 𝑥 2 𝑑𝑥 = 𝜋 3 /3
𝜋
If 𝑦(𝑥) = sin 𝑥: then 𝐽(𝑦(𝑥)) = ∫0 𝑠𝑖𝑛2 𝑥 𝑑𝑥 = 𝜋/2
So we can see a functional maps a function to a ❑ Consider two points in the 𝑥𝑦-plane, as shown in the
number. Because an integral maps a function to a figure.
number, a functional usually involves an integral.
❑ An arbitrary path joining the points follows the general
Calculus of Variations curve 𝑦 = 𝑦(𝑥), and an element of length along the
path is 𝑑𝑠 = √𝑑𝑥 2 + 𝑑𝑦 2 .
It is a branch of mathematics whose aim to is to find path,
curve, surface, etc. for which given function has a ❑ We can write:
stationary value (which, in physical problems is usually a 𝑑𝑦
𝑑𝑦 = 𝑑𝑥 = 𝑦 ′ (𝑥)𝑑𝑥
minimum or maximum). Mathematically, this involves 𝑑𝑥
finding stationary values of integral of the form
❑ Using it we can rewrite 𝑑𝑠 as
𝑏
𝐼 = ∫ 𝑓(𝑦, 𝑦 ′ , 𝑥)𝑑𝑥 𝑑𝑠 = √𝑑𝑥 2 + (𝑦 ′ (𝑥)𝑑𝑥)2
𝑎

𝑑𝑠 = √1 + 𝑦 ′ (𝑥)2 𝑑𝑥
❑ Thus, the total length between points 1 and 2 is
2 depend on functions of one variable.
𝐿 = ∫ 𝑑𝑠.
1 ❑ We are now going to discuss a variational method due
𝑥2
= ∫ √1 + 𝑦 ′ (𝑥)2 𝑑𝑥
𝑥1
❑ Note that we have converted the problem from an
integral along a path, to an integral over x:
❑ We have thus succeeded in writing the problem down,
but we need some additional mathematical machinery
to find the path for which 𝐿 is an extremum (a minimum
in this case).
Fermat’s Principle:
❑ Fermat's principle states that “light travels between
two points along the path that requires the least time,
as compared to other nearby paths.”
❑ Recall that light travels more slowly through a medium,
and we define the index of refraction as 𝑛 = 𝑐/𝑣.
where c is the speed of light in vacuum, and 𝑣 is the
speed of light in the medium.
❑ If light enters from one medium to another between
point 1 and 2, the correct path is the one for which the
time taken between these points is minimum. The total
travel time is then
2 2
𝑑𝑠 1 2
𝜏 = ∫ 𝑑𝑡 = ∫ = ∫ 𝑛𝑑𝑠
1 1 𝑣 𝑐 1
1 𝑥2
= ∫ 𝑛(𝑥, 𝑦)√1 + 𝑦 ′ (𝑥)2 𝑑𝑥 .
𝑐 𝑥1
❑ Here we are allowing the index of refraction to vary
arbitrarily vs. 𝑥 and 𝑦.
❑ If 𝑛 is uniform, then it can be taken outside the integral
and problems reduces to finding shortest path 𝑦(𝑥)
between 1 and 2 and the answer is a ‘straight line’, of
course!
Variational Principles:
❑ Obviously, both problems are similar, and such cases
arise in many other situations.
❑ In our usual minimizing or maximizing of a function
𝑓(𝑥), we would take the derivative and find its zeroes
(i.e. the values of 𝑥 for which the slope of the function
is zero). These points of zero slope may be minima,
maxima, or points of inflection, but in each case, we can
say that the function is stationary at those points,
meaning for values of 𝑥 near such a point, the value of
the function does not change (due to the zero slope).
❑ In analogy with this familiar approach, we want to be
able to find solutions to the integrals that are stationary
for infinitesimal variations in the path. This is called
calculus of variations.
❑ The methods we will develop are called variational
methods, and a principle like Fermat’s Principle are
called variational principles.
6.2 Euler-Lagrange Equation:
❑ Here we use calculus of variations and derive the
general Euler-Lagrange equations for functionals that
to Euler and Lagrange, which seeks to find an extremum
(to be definite, let’s consider this a minimum) for an as
yet unknown curve joining two points 𝑥1 and 𝑥2 ,
satisfying the integral relation
𝑥2
𝑆 = ∫ 𝑓[𝑦(𝑥), 𝑦′(𝑥), 𝑥]𝑑𝑥 − − − (1)
𝑥1
❑ The function 𝑓 is a function of three variables, but
because the path of integration is 𝑦 = 𝑦(𝑥), the
integrand can be reduced to a function of just one
variable, 𝑥.
❑ To start, let’s consider two curves joining points 1 and
2, the “right” curve 𝑦(𝑥), and a “wrong” curve 𝑌(𝑥) that
is a small displacement from the “right” curve, as shown
in the figure.
❑ We will write the difference between these curves as
some function 𝜂(𝑥).
𝑌(𝑥) = 𝑦(𝑥) + 𝜂(𝑥)
❑ Since 𝑌(𝑥) must pass through the endpoints 1 and 2,
𝜂(𝑥) must satisfy
𝜂(𝑥1 ) = 𝜂(𝑥2 ) = 0.
❑ There are infinitely many functions 𝑌(x), that can be
“wrong,” but we require that they each be longer that
the “right” path. To quantify how close the “wrong”
path can be to the “right” one, let’s write
𝑌(𝑥) = 𝑦(𝑥) + 𝛼 𝜂(𝑥) − − − (2)
so that
𝑥2
𝑆(𝛼) = ∫ 𝑓[𝑌(𝑥), 𝑌′(𝑥), 𝑥]𝑑𝑥 − − − (3)
𝑥1
❑ We want to find particular 𝑌(𝑥) that make 𝑆(𝛼)
stationary. Note that 𝑆(𝛼) now only depends on 𝛼. ( By
introducing parameter 𝛼, the problem now is the same
as in elementary calculus, i.e. function is stationary at
points where its derivative is zero.)
❑ 𝑆(𝛼) corresponds to extreme path when
𝑑𝑆
| =0
𝑑𝛼 𝛼=0
𝑥2
𝑑
| ∫ 𝑓[𝑌(𝑥), 𝑌′(𝑥), 𝑥]𝑑𝑥 = 0
𝑑𝛼 𝛼=0 𝑥1
 𝑥2
𝜕 Example 6.1: Shortest Path Between Two Points:
∫ 𝑓[𝑌(𝑥), 𝑌′(𝑥), 𝑥]| 𝑑𝑥 = 0
𝑥1 𝜕𝛼
𝛼=0 ❑ We earlier showed that the problem of the shortest
𝑥2
𝜕𝑓 𝜕𝑌 𝜕𝑓 𝜕𝑌′ path between two points can be expressed as
∫ [ + ]| 𝑑𝑥 = 0
𝑥1 𝜕𝑌 𝜕𝛼 𝜕𝑌′ 𝜕𝛼 2 𝑥2
𝛼=0
𝐿 = ∫ 𝑑𝑠 = ∫ √1 + 𝑦 ′ (𝑥)2 𝑑𝑥 .
1 𝑥1
Since
𝜕𝑌 ❑ The integrand contains our function
𝑌(𝑥) = 𝑦(𝑥) + 𝛼 𝜂(𝑥) ⇒ = 𝜂(𝑥)
𝜕𝛼
𝑓(𝑦, 𝑦 ′ , 𝑥) = √1 + 𝑦 ′ (𝑥)2 .
𝜕𝑌 ′
𝑌′(𝑥) = 𝑦′(𝑥) + 𝛼 𝜂′(𝑥) ⇒ = 𝜂′(𝑥)
𝜕𝛼 ❑ The two partial derivatives in the Euler-Lagrange
So above integral becomes equation are:
𝑥2
𝜕𝑓 𝜕𝑓 Since 𝑓 does not depend on 𝑦 so if we differentiate 𝑓
∫ [ 𝜂(𝑥) + 𝜂′(𝑥) ]| 𝑑𝑥 = 0 − − − (4) w.r.t. 𝑦
𝑥1 𝜕𝑌 𝜕𝑌′
𝛼=0
𝜕𝑓
❑ We can handle the second term in the previous =0
𝜕𝑦
equation by integration by parts:
𝑥2 Now differentiate 𝑓 w.r.t. 𝑦′
𝜕𝑓 ′ 𝜕𝑓 𝑥2 ′ 𝑥2
𝑑 𝜕𝑓
∫ 𝜂 𝑑𝑥 = ∫ 𝜂 𝑑𝑥 − ∫ (∫ 𝜂 ′ 𝑑𝑥) { ′ } 𝑑𝑥
𝑥1 𝜕𝑌
′ ′
𝜕𝑌 𝑥1 𝑥1 𝑑𝑥 𝜕𝑌 𝜕𝑓 𝑦′
= .
𝑥2 𝑥2 𝜕𝑦 ′ √1 + 𝑦 ′2
𝜕𝑓 ′ 𝜕𝑓 𝑥2 𝑑 𝜕𝑓
∫ ′
𝜂 𝑑𝑥 = ( ′
𝜂|𝑥1 ) − ∫ 𝜂 { ′ } 𝑑𝑥
𝑥1 𝜕𝑌 𝜕𝑌 𝑥1 𝑑𝑥 𝜕𝑌 ❑ Thus, the Euler-Lagrange equation gives us

Here first term is zero as 𝜂(𝑥1 ) = 𝜂(𝑥2 ) = 0, so that 𝑑 𝜕𝑓 𝑑 𝑦′

= = 0.
𝑥2
𝜕𝑓 ′ 𝑥2
𝑑 𝜕𝑓 𝑑𝑥 𝜕𝑦 ′ 𝑑𝑥 √1 + 𝑦 ′2
∫ ′
𝜂 𝑑𝑥 = ∫ 𝜂 { } 𝑑𝑥
𝑥1 𝜕𝑌 𝑥1 𝑑𝑥 𝜕𝑌 ′ ❑ This says that
So that equation (4) becomes 𝑦′
=𝐶
𝑥2 √1 + 𝑦 ′2
𝜕𝑓 𝑑 𝜕𝑓
∫ [ 𝜂(𝑥) + 𝜂(𝑥) { ′ } ]| 𝑑𝑥 = 0
𝑥1 𝜕𝑌 𝑑𝑥 𝜕𝑌
𝛼=0 Squaring both sides
𝑥2
𝜕𝑓 𝑑 𝜕𝑓 or 𝑦 ′2 = 𝐶 2 (1 + 𝑦 ′2 )
∫ 𝜂(𝑥) [ + { } ]| 𝑑𝑥 = 0
𝑥1 𝜕𝑌 𝑑𝑥 𝜕𝑌 ′
𝛼=0 𝑦 ′2 − 𝐶 2 𝑦 ′2 = 𝐶 2
❑ At 𝛼 = 0, equation (2) says: 𝑌(𝑥) = 𝑦(𝑥), so
𝐶2
𝑥2 𝑦 ′2 =
𝜕𝑓 𝑑 𝜕𝑓 1 − 𝐶2
∫ 𝜂(𝑥) [ + { } ] 𝑑𝑥 = 0
𝑥1 𝜕𝑦 𝑑𝑥 𝜕𝑦
𝐶2
❑ Since 𝜂(𝑥) is an arbitrary function, we must have 𝑦′ = √
1 − 𝐶2
𝜕𝑓 𝑑 𝜕𝑓
− = 0 − − − (5) 𝐶2
𝜕𝑦 𝑑𝑥 𝜕𝑦 ′ √ = constant = C1
1 − 𝐶2
This is the Euler-Lagrange equation.

❑ Conclusion: 𝑦 ′ = 𝐶1
𝑑𝑦
If 𝑦(𝑥) is an extremal of 𝑆, then it can be obtained using = 𝐶1
𝑑𝑥
Euler-Lagrange equation.
Integrate both sides

𝑦 = 𝐶1 𝑥 + 𝐶2
Procedure for Solving Problems
We can find constants 𝐶1 and 𝐶2 using boundary
(1) Set up the problem so that the quantity whose conditions:
stationary path you seek is expressed as
𝑥2
𝑦(𝑥1 ) = 𝑦1 and 𝑦(𝑥2 ) = 𝑦2
𝑆 = ∫ 𝑓[𝑦(𝑥), 𝑦 ′ (𝑥), 𝑥]𝑑𝑥 , So we can write
𝑥1

where 𝑓[𝑦(𝑥), 𝑦 ′ (𝑥), 𝑥] 𝑦1 = 𝐶1 𝑥1 + 𝐶2

is the function appropriate to your problem. 𝑦2 = 𝐶1 𝑥2 + 𝐶2

Solve these for 𝐶1 and 𝐶2, and doing some algebra, we
(2) Write down the Euler-Lagrange equation,
get
(3) Solve for the function 𝑦(𝑥) that defines the required
stationary path.
 𝑦2 − 𝑦1
𝑦 = 𝑦2 + ( ) (𝑥 − 𝑥2 ) 𝑦
𝑥2 − 𝑥1 𝑥 = ∫√  𝑑𝑦.
2𝑎 − 𝑦
Which is equation of a straight line, so a straight line is
the shortest path between two points. ◼ It is not obvious, but this can be solved by
the substitution
Example 6.2: The Brachistochrone:
𝑦 = 𝑎(1 − 𝑐𝑜𝑠 𝜃) − − − (1)
❑ This is a famous problem in the history of the calculus of
variations. Statement of the problem: ⇒ 𝑑𝑦 = 𝑎 sin 𝜃 𝑑𝜃 = 𝑎√1 − cos 2 𝜃 𝑑𝜃
◼ Given two points 1 and 2, with 1 higher above the which gives
ground, in what shape could we build a track for a
frictionless roller-coaster so that a car released from 𝑎(1 − 𝑐𝑜𝑠 𝜃)
point 1 would reach point 2 in the shortest possible 𝑥 = ∫√   𝑎√1 − cos2 𝜃 𝑑𝜃
2𝑎 − [𝑎(1 − 𝑐𝑜𝑠 𝜃)]
time? See the figure, which takes point 1 as the origin,
with y positive downward.
𝑎(1 − 𝑐𝑜𝑠 𝜃)(1 + 𝑐𝑜𝑠 𝜃)(1 − 𝑐𝑜𝑠 𝜃)
𝑥 = 𝑎∫√   𝑑𝜃
❑ Solution: 𝑎 + 𝑎 𝑐𝑜𝑠 𝜃

◼ The time to travel from point 1 to 2 is

𝑎(1 − 𝑐𝑜𝑠 𝜃)(1 + 𝑐𝑜𝑠 𝜃)(1 − 𝑐𝑜𝑠 𝜃)
2 𝑥 = 𝑎∫√   𝑑𝜃
𝑑𝑠 𝑎(1 + 𝑐𝑜𝑠 𝜃)
𝜏=∫
1 𝑣
𝑥 = 𝑎 ∫(1 − 𝑐𝑜𝑠 𝜃)  𝑑𝜃
where 𝑣 = √2𝑔𝑦, from kinetic energy considerations.

◼ Since this depends on y, we will take y as 𝑥 = 𝑎(𝜃 − 𝑠𝑖𝑛 𝜃) + const − − − (2)

the independent variable, hence
◼ Using initial conditions i.e. 𝑥 = 𝑦 = 0
𝑑𝑠 = √𝑑𝑥 2 + 𝑑𝑦 2 = √𝑥 ′ (𝑦)2 + 1 𝑑𝑦. Equation (1): 0 = 𝑎(1 − cos 𝜃)
◼ Our integral now becomes: 1 = cos 𝜃 ⇒ 𝜃 = 0°
𝑦2 √ ′2 Putting in equation (2)
1 𝑥 +1
𝜏= ∫ 𝑑𝑦
√2𝑔 0 √𝑦 0 = 𝑎(0 − 𝑠𝑖𝑛 0) + const
◼ From the Euler-Lagrange equation: const. = 0
𝜕𝑓 𝑑 𝜕𝑓 ◼ The two equations that give the path are
=
𝜕𝑥 𝑑𝑦 𝜕𝑥 ′ then:
◼ Since 𝑥 = 𝑎(𝜃 − 𝑠𝑖𝑛 𝜃)
𝑦 = 𝑎(1 − 𝑐𝑜𝑠 𝜃)
√𝑥 ′2 + 1
𝑓=
√𝑦 in terms of 𝜃.

clearly ◼ This curve is called a cycloid, and is a very

special curve indeed.
𝜕𝑓
= 0, ◼ Conclusion: Brachistochrone is a segment
𝜕𝑥
on a cycloid.
and so
◼ As you will show in the homework, it is the
𝜕𝑓
= constant curve traced out by a wheel rolling (upside
𝜕𝑥 ′ down) along the x axis traces out a cycloid.
◼ Evaluating this derivative and squaring it
◼ Another remarkable thing is that the time
for convenience, we have
it takes for a cart to travel this path from
𝑥 ′2 1 2→3 is the same, no matter where 2 is
′2
= constant =
𝑦(𝑥 + 1) 2𝑎 placed, from 1 to 3! Thus, oscillations of
the cart along that path are exactly
where the constant is renamed 1/2𝑎 for future
isochronous (period perfectly independent
convenience.
of amplitude).
◼ Solving for 𝑥′ we have:

𝑦
𝑥′ = √
2𝑎 − 𝑦

Finally, to get x we integrate:

 ❑ This may seem more difficult than the previous case,but
in fact we just proceed in the same manner with the
“wrong” paths

𝑥 = 𝑥(𝑢) + 𝛼𝜉(𝑢); 𝜉(𝑢1 ) = 𝜉(𝑢2 ) = 0.

𝑦 = 𝑦(𝑢) + 𝛽𝜂(𝑢); 𝜂(𝑢1 ) = 𝜂(𝑢2 ) = 0.
❑ The stationary path is the one for which
𝜕𝑆 𝜕𝑆
=0 and =0.
𝜕𝛼 𝜕𝛽
❑ Following the same procedure as before, we end up
with two Euler-Lagrange equations that the function f
More than Two Variables:
must satisfy:
❑ When we discussed the shortest path between two
𝜕𝑓 𝑑 𝜕𝑓 𝜕𝑓 𝑑 𝜕𝑓
points, we examined the situation in the figure below. − =0 and − = 0.
𝜕𝑥 𝑑𝑢 𝜕𝑥 ′ 𝜕𝑦 𝑑𝑢 𝜕𝑦 ′
H.W. Derive these equations.

Follow example 6.3

Preparation for Chapter 7:

Euler-Lagrange equation in Lagrangian Formulism

❑ In Chapter 7, we will meet problems with several

variables, with time 𝑡 as the independent variable. In
❑ However, this is not the most general path. Here is one these cases, there will be two or more Euler-Lagrange
that cannot be expressed as a function 𝑦(𝑥). equations to satisfy (for example, equations for 𝑥, 𝑦 and
𝑧).

❑ One of the great strengths of Lagrangian mechanics is

its ability to deal with cartesian, cylindrical, spherical,
and any other coordinate systems with ease.

❑ In order to generalize our discussion, we write (𝑥, 𝑦, 𝑧),

(𝜌, 𝜙, 𝑧) or (𝑟, 𝜃, 𝜙) as generalized coordinates
(𝑞1 , 𝑞2 , … , 𝑞𝑁 ).

❑ In the next chapter, we will use slightly different

notation, and refer to our integrand function

𝑓[𝑞1 , 𝑞2 , ⋯ , 𝑞𝑁 , 𝑞̇ 1 , 𝑞̇ 2 , ⋯ , 𝑞̇ 𝑁 , 𝑡],
as the Lagrangian
❑ To handle this case, we need to consider a path
𝐿 = 𝐿[𝑞1 , 𝑞2 , ⋯ , 𝑞𝑁 , 𝑞̇ 1 , 𝑞̇ 2 , ⋯ , 𝑞̇ 𝑁 , 𝑡].
specified by an independent variable (parameter) 𝑢
along the path, i.e. 𝑥 = 𝑥(𝑢), 𝑦 = 𝑦(𝑢) . Note that because the independent variable is t, we can
use 𝑞̇ 𝑖 rather than 𝑞𝑖′ .
❑ The length of a small segment of the path is
❑ We will then make the action integral stationary
2 2
𝑑𝑥 𝑑𝑦
𝑑𝑠 = √𝑑𝑥 2 + 𝑑𝑦 2 = √( 𝑑𝑢) + ( 𝑑𝑢) 𝑆 = ∫ 𝐿[𝑞1 , 𝑞2 , ⋯ , 𝑞𝑁 , 𝑞̇ 1 , 𝑞̇ 2 , ⋯ , 𝑞̇ 𝑁 , 𝑡] 𝑑𝑡,
𝑑𝑢 𝑑𝑢

𝑑𝑠 = √𝑥 ′ (𝑢)2 + 𝑦 ′ (𝑢)2 𝑑𝑢 which requires that L satisfy

❑ Now the length of the path is then 𝜕𝐿 𝑑 𝜕𝐿 𝜕𝐿 𝑑 𝜕𝐿 𝜕𝐿 𝑑 𝜕𝐿

= ;  = ; ⋯ ;  = .
𝜕𝑞1 𝑑𝑡 𝜕𝑞̇ 1 𝜕𝑞2 𝑑𝑡 𝜕𝑞̇ 2 𝜕𝑞𝑁 𝑑𝑡 𝜕𝑞̇ 𝑁
𝑢2
𝐿=∫ √𝑥 ′ (𝑢)2 + 𝑦 ′ (𝑢)2 𝑑𝑢 We will develop this more next time.
𝑢1

And our goal is to find 𝑥(𝑢) and 𝑦(𝑢) for which the
integral is minimum.

❑ In general, the problem is: Given an integral of the form

𝑢
𝑆 = ∫𝑢 2 𝑓[𝑥(𝑢), 𝑦(𝑢), 𝑥′(𝑢), 𝑦′(𝑢), 𝑢]𝑑𝑢
1

between two [𝑥1 (𝑢), 𝑦1 (𝑢)] and [𝑥2 (𝑢), 𝑦2 (𝑢)], find
the path [𝑥(𝑢), 𝑦(𝑢)] for which the integral 𝑆 is
stationary.