Scribd
Upload
37 views

Calcvar

1) The document discusses calculus of variations and how to minimize or maximize functionals. 2) It explains that to find extrema of functionals, we consider "admissible curves" that are small variations of the desired curve and examine how the functional changes with these variations. 3) The necessary condition to minimize a functional is for the first variation δI to equal zero, and the sufficient condition is for the second variation δ2I to be positive.

Uploaded by

Kumardeep Mukhopadhyay
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
37 views

Calcvar

1) The document discusses calculus of variations and how to minimize or maximize functionals. 2) It explains that to find extrema of functionals, we consider "admissible curves" that are small variations of the desired curve and examine how the functional changes with these variations. 3) The necessary condition to minimize a functional is for the first variation δI to equal zero, and the sufficient condition is for the second variation δ2I to be positive.

Uploaded by

Kumardeep Mukhopadhyay
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 5

Advanced Mechanics of Solids (ME60402) Mechanical Engineering

Spring Semester, 2021 IIT Kharagpur

Calculus of Variations ∗

1 Introduction
Let us first consider something very familiar to us, a simple minimiza-
tion problem from differential calculus: What is the value of 𝑥 that will
minimize 𝑦 = 3𝑥 2 , with 𝑥 ∈ [𝑥1 , 𝑥2 ]?
Next, consider the integration 𝐼 = ∫𝑥1 2 𝑓 (𝑥) d𝑥, which has a definite value
𝑥

for a particular choice of 𝑓 (𝑥). However, suppose we wish to know the


minimum value of 𝐼 for various choices of 𝑓 (𝑥). So, what we are asking
is this: What is the 𝑓 (𝑥) that gives the minimum value of 𝐼 ? Question is,
what method do we use to find such a minimum? And, the answer is:
the calculus of variations.
One may wonder in what kind of scenario would be faced with such a
question. Let’s take a couple of concrete examples:
• Find the curve between two points which minimizes the distance
between the two points. We have
√ 𝑥2 √
𝐿 = d𝑥 2 + d𝑦 2 = ∫ 1 + (𝑦 ′ )2 d𝑥
𝑥1

• Find the shape of a frictionless path between two points such that a
particle takes the least time to slide down the path under the effect
of gravity. We have
√ √
𝑥2 𝑥2
d𝑥 2 + d𝑦 2 √ 1 + 𝑦′
𝑇 =∫ , with 𝑉 = 2𝑔𝑦. Thus, 𝑇 = ∫ d𝑥
𝑥1 𝑉 𝑥1 2𝑔𝑦
In the first example, the integrand is of the form 𝐹 (𝑦 ′ ), and in the second,
it is of the form 𝐹 (𝑦, 𝑦 ′ ). More generally, beyond these examples, we
want to minimize 𝐼 = ∫𝑥1 2 𝐹 (𝑥, 𝑦, 𝑦 ′ , 𝑦 ′′ , … 𝑦 (𝑛) )d𝑥. Even more generally,
𝑥

we want to extremize 𝐼 .
∗ Notes prepared by Jeevanjyoti Chakraborty. Contact: [email protected]

1
Advanced Mechanics of Solids Mechanical Engineering
Calculus of Variations IIT Kharagpur

Let us see how we can actually minimize (or, equivalently, maximize)


the integration, 𝐼 (𝑥, 𝑦, 𝑦 ′ , … 𝑦 (𝑛) ). To keep things simple we will consider
𝐼 (𝑥, 𝑦, 𝑦 ′ ). Note that 𝐼 is a function of functions; such an entity is referred
to as a functional.

A The first temptation when trying to minimize (or, maximize) a func-


d𝐼
tional, 𝐼 , is to try to do something like . But this is problematic!
d𝑦
To see the problem, consider as an example the problem of finding the
path on a plane that covers the shortest distance between two points.
When we write the differential d𝑦, it is associated with a change in 𝑥,
i.e. d𝑥. However, when we are considering various possibilities of the
path what we are actually doing is considering different values of 𝑦 for
the same 𝑥. Such changes in 𝑦 are completely different in meaning from
d𝑦. The notation that we use to represent this special kind of change in
𝑦 is 𝛿𝑦.
d𝐼
So we certainly do not want . What do we do then?
d𝑦
We go back to the basics and look once again into how extremization is
done for functions. We will then export that idea to functionals.

2 Extremization of functions
Consider a simple function 𝑦(𝑥). Suppose this function takes on extreme
values (either minimum or maximum) at 𝑥 = 𝑎. To concretize our ideas
consider the case of the minimum. We expand 𝑦(𝑥) about 𝑥 = 𝑎 as:
𝑥 −𝑎 ′ (𝑥 − 𝑎)2 ′′
𝑦(𝑥) = 𝑦(𝑥 − 𝑎 + 𝑎) = 𝑦(𝑎) + 𝑦 (𝑎) + 𝑦 +⋯ (1)
1! 2!
Since 𝑦(𝑎) is a minimum, we must have 𝑦(𝑥) − 𝑦(𝑎) > 0. Now, both the
conditions 𝑥 − 𝑎 < 0 and 𝑥 − 𝑎 > 0 are possible depending on where 𝑥
is located. However, because we are considering a particular function,
𝑦(𝑥), only a particular value of 𝑦 ′ (𝑎) is possible. Therefore, to ensure
that the sign of (𝑥 − 𝑎) does not influence 𝑦(𝑥) − 𝑦(𝑎), we need to have
𝑓 ′ (𝑎) = 0. This condition is the necessary condition for extremum value
which we are all familiar with from high school calculus.

2
Advanced Mechanics of Solids Mechanical Engineering
Calculus of Variations IIT Kharagpur

Next, to ensure that 𝑦(𝑥) − 𝑦(𝑎) > 0, it is sufficient that 𝑦 ′′ > 0 because
(𝑥 − 𝑎)2 > 0, always.
Now, we will export this idea of finding the conditions for extremum of
functions to functionals.

3 Extremization of functionals
At the outset, note that in the previous section for the extremization of
functions, we started by exploring the neighbourhood of 𝑥 = 𝑎. Similarly
for functionals, we need to explore the “neighbourhood” of the desired
function or curve 𝑦 that will ultimately lead to the extreme value of the
functional. But since derivatives and integrations are involved we need
to be careful in exploring this neighbourhood. We need to select those
alternative curves which satisfy certain conditions. This set of curves are
referred to as admissible curves. We consider an admissible curve 𝑦 + 𝜀𝜂.
The value of the functional corresponding to 𝑦 + 𝜀𝜂 is
𝑥2
𝐼̄ = ∫ 𝐹 (𝑥, 𝑦 + 𝜀𝜂, 𝑦 ′ + 𝜀𝜂′ ) d𝑥. (2)
𝑥1

For minimum, we must have


𝐼̄ − 𝐼 = Δ𝐼 > 0. (3)
Now,
𝑥2
Δ𝐼 = ∫ ′ ′ ′
[𝐹 (𝑥, 𝑦 + 𝜖𝜂, 𝑦 + 𝜖𝜂 ) − 𝐹 (𝑥, 𝑦, 𝑦 )] d𝑥. (4)
𝑥1

Using Taylor expansion on the integrand, we have


𝑥2
𝜕𝐹 𝜕𝐹 𝜖 2 𝜂2 𝜕 2 𝐹 𝜖 2 𝜂′2 𝜕 2 𝐹
Δ𝐼 = ∫ 𝐹 (𝑥, 𝑦, 𝑦 ′ ) + 𝜖𝜂 + 𝜖𝜂′ ′ + +
𝑥1 [ 𝜕𝑦 𝜕𝑦 2! 𝜕𝑦 2 2! 𝜕𝑦 ′2
𝜖𝜂𝜂′ 𝜕 2 𝐹
+2 + 𝑂(𝜖 3 ) − 𝐹 (𝑥, 𝑦, 𝑦 ′ ) d𝑥
2! 𝜕𝑦𝜕𝑦 ′ ]
𝑥2 2
𝜕 𝜕 1 𝜕 𝜕
=∫ 𝜖 𝜂 + 𝜂′ ′ 𝐹 + 𝜖 2 𝜂 + 𝜂′ ′ 𝐹 d𝑥 + 𝑂(𝜖 3 ).
𝑥1 [ ( 𝜕𝑦 𝜕𝑦 ) 2 ( 𝜕𝑦 𝜕𝑦 ) ]
(5)

3
Advanced Mechanics of Solids Mechanical Engineering
Calculus of Variations IIT Kharagpur

This equation is conventionally represented in the form


1
Δ𝐼 = 𝛿𝐼 + 𝛿 2 𝐼 + 𝑂(𝜖 3 ) (6)
2
The terms 𝛿𝐼 and 𝛿 2 𝐼 are called the first and second variations of 𝐼 .
If 𝜀 → 0, the term 𝛿𝐼 (which involves 𝜀) dominates over the term 𝛿 2 𝐼
(which involves 𝜀 2 ). Now, since the sign of 𝛿𝐼 is reversed by a reversal of
the sign of 𝜀, the necessary condition for Δ𝐼 to be positive for all small
values of 𝜀 is that 𝛿𝐼 = 0. Further, when 𝛿𝐼 = 0, the sufficient condition
for minimum is that 𝛿 2 𝐼 > 0.
Let us look at the necessary condition in more detail. We have

𝛿𝐼 = 0,
𝑥2
𝜕 𝜕
or, ∫ 𝜀 𝜂 + 𝜂′ ′ d𝑥 = 0,
𝑥1 ( 𝜕𝑦 𝜕𝑦 )
𝑥2 2 𝑥 𝑥2
𝜕𝐹 𝜕𝐹 d 𝜕𝐹
or, ∫ 𝜀𝜂 d𝑥 + 𝜀 ′ 𝜂 − ∫ 𝜀 𝜂 d𝑥 = 0.
𝑥1 𝜕𝑦 [ 𝜕𝑦 ]𝑥 𝑥1 d𝑥 ( 𝜕𝑦 ′ )
1

(Using integration by parts) (7)

2 𝑥
𝜕𝐹
Now, we focus on the boundary term 𝜀 ′ 𝜂 , and consider the fol-
[ 𝜕𝑦 ]𝑥
1
lowing cases:
Case 1: If among the conditions which define the admissibility of curves
that can serve as input to our functional, we have the conditions that the
values of the curves are specified at the boundaries 𝑥 = 𝑥1 and 𝑥 = 𝑥2 ,
then we must have 𝜂 = 0 at 𝑥1 and 𝑥2 . Then the boundary term becomes
zero, and we have
𝑥2 𝑥2
𝜕𝐹 d 𝜕𝐹
∫ 𝜀𝜂 d𝑥 − ∫ 𝜀 𝜂 d𝑥 = 0,
𝑥1 𝜕𝑦 𝑥1 d𝑥 ( 𝜕𝑦 ′ )
𝑥2
𝜕𝐹 d 𝜕𝐹
or, ∫ 𝜀𝜂 [ 𝜕𝑦 − d𝑥 ( 𝜕𝑦 ′ )] d𝑥 = 0.
𝑥1

4
Advanced Mechanics of Solids Mechanical Engineering
Calculus of Variations IIT Kharagpur

Since 𝜂 is arbitrary, the integral can be zero only if we have the follow-
ing†

𝜕𝐹 d 𝜕𝐹
− = 0, (8)
𝜕𝑦 d𝑥 𝜕𝑦 ′ )
(

which is famously known as the Euler-Lagrange equation.

Case 2: If the values of the curves are not specified at the boundaries
𝑥 = 𝑥1 and 𝑥 = 𝑥2 , then 𝜂 is not equal to zero at the boundaries. Then the
boundary terms in Eq. (7) are not zero, and so in addition to the Euler-
Lagrange equations we end up with the requirement that at 𝑥1 and 𝑥2 ,
𝜕𝐹
= 0.
𝜕𝑦 ′

† Actually, there is a proof for this statement. But we state it without proof here!

You might also like