Perfect Harmony II

Calculus Level 5

Let F \mathcal F be the collection of all continuously differentiable increasing functions ψ \psi defined on [ 0 , π 2 ] \left[0,\frac \pi2\right] such that ψ ( 0 ) = 0 \psi(0)=0 and ψ ( π 2 ) = 1. \psi\left(\frac \pi 2\right)=1. For any ψ F , \psi\in\mathcal F, consider the integral I ( ψ ) = 0 π 2 [ ( d d x ψ ( x ) ) 2 + ( d d x 1 ψ ( x ) ) 2 ] d x . I(\psi)=\int_0^{\frac \pi 2}\left[\left(\frac d{dx}\sqrt{\psi (x)}\right)^2+\left(\frac d{dx}\sqrt{1-\psi(x)}\right)^2\right]dx. As ψ \psi ranges over F , \mathcal F, the smallest possible value of I ( ψ ) I(\psi) is equal to A π . A\pi. What is the value of A ? A?


Resource: AMSS Summer Camp 2018.


The answer is 0.5.

This section requires Javascript.
You are seeing this because something didn't load right. We suggest you, (a) try refreshing the page, (b) enabling javascript if it is disabled on your browser and, finally, (c) loading the non-javascript version of this page . We're sorry about the hassle.

3 solutions

Mark Hennings
Mar 19, 2019

The functional is I ( ψ ) = 0 1 2 π ψ ( x ) 2 4 ψ ( x ) ( 1 ψ ( x ) ) d x I(\psi) \; = \; \int_0^{\frac12\pi} \frac{\psi'(x)^2}{4\psi(x)(1-\psi(x))}\,dx By the Cauchy-Schwarz Inequality, ( 0 1 2 π ψ ( x ) ψ ( x ) ( 1 ψ ( x ) ) d x ) 2 ( 0 1 2 π d x ) ( 0 1 2 π ψ ( x ) 2 ψ ( x ) ( 1 ψ ( x ) ) d x ) = 2 π I ( ψ ) ( 0 1 d u u ( 1 u ) ) 2 2 π I ( ψ ) π 2 2 π I ( ψ ) \begin{aligned} \left(\int_0^{\frac12\pi} \frac{\psi'(x)}{\sqrt{\psi(x)(1-\psi(x))}}\,dx\right)^2 & \le \; \left(\int_0^{\frac12\pi}\,dx\right)\left(\int_0^{\frac12\pi} \frac{\psi'(x)^2}{\psi(x)(1-\psi(x))}\,dx\right) \; = \; 2\pi I(\psi) \\ \left(\int_0^1 \frac{du}{\sqrt{u(1-u)}}\right)^2 & \le \; 2\pi I(\psi) \\ \pi^2 & \le \; 2\pi I(\psi) \end{aligned} for all suitable ψ \psi . Thus I ( ψ ) 1 2 π I(\psi) \ge \tfrac12\pi for all suitable ψ \psi (by "suitable" I mean that ψ ( x ) \psi'(x) has to be sufficiently well-behaved at both x = 0 x=0 and x = 1 2 π x=\tfrac12\pi for the integral I ( ψ ) I(\psi) to exist), with equality obtained when ψ ( x ) = k ψ ( x ) ( 1 ψ ( x ) ) \psi'(x) = k\sqrt{\psi(x)(1-\psi(x))} for some constant k k . With the boundary conditions ψ ( 0 ) = 0 \psi(0) = 0 and ψ ( 1 2 π ) = 1 \psi(\tfrac12\pi) = 1 , we solve this differential equation to obtain ψ ( x ) = sin 2 x \psi(x) = \sin^2x , and so the minimum value of 1 2 π \boxed{\tfrac12}\pi for I ( ψ ) I(\psi) can be achieved.

Given the potentially singular nature of the integrand at x = 0 x=0 and x = 1 2 π x=\tfrac12\pi , the Calculus of Variations approach would have to be established carefully!

@Mark Hennings Sir, could you please post a Calculus of Variations approach too???

Aaghaz Mahajan - 2 years, 2 months ago

Log in to reply

I've just posted one, I hope it helps! :)

Wesley Low - 2 years, 2 months ago
Wesley Low
Mar 21, 2019

Letting the integrand be L L , we can simplify it to be L = ψ 2 4 ψ ( 1 ψ ) L =\frac { { \psi ' }^{ 2 } }{ 4\psi \left( 1-\psi \right) } By the Euler-Lagrange equation, for I ( ψ ) I\left( \psi \right) to be minimised, L L must satisfy the equation x L d d x ( L ψ ψ L ) = 0 \frac { \partial }{ \partial x } L-\frac { d }{ dx } \left( L-\psi '\frac { \partial }{ \partial \psi ' } L \right) =0 L ψ ψ L = c L-\psi '\frac { \partial }{ \partial \psi ' } L=c , where c c is an arbitrary constant. This simplifies to ψ = α ψ ( 1 ψ ) \psi '=\alpha \sqrt { \psi \left( 1-\psi \right) } , where α = 4 c \alpha =\sqrt{-4c} . Now, integrating using the variable seperable method yields 1 ψ ( 1 ψ ) α ψ d x d x = α d x \int { \frac { 1 }{ \sqrt { \psi \left( 1-\psi \right) } } } \frac { \alpha \psi }{ dx } dx=\int { \alpha } dx 2 arcsin ψ = α x + β 2\arcsin { \sqrt { \psi } } =\alpha x+\beta with β \beta as the constant of integration. Making ψ \psi the subject and substituting the limits ψ ( 0 ) = 0 \psi \left( 0 \right) =0 and ψ ( π 2 ) = 1 \psi \left( \frac { \pi }{ 2 } \right) =1 gives ψ = sin 2 ( x ) \psi =\sin ^{ 2 }{ \left( x \right) } and L = 1 L=1 Therefore I ( ψ ) = 0 π 2 1 d x = π 2 I\left( \psi \right) =\int _{ 0 }^{ \frac { \pi }{ 2 } }{ 1 } dx=\boxed{\frac { \pi }{ 2 }}

The problem with this is that the function L L has the potential to be singular at both x = 0 x=0 and x = 1 2 π x=\tfrac12\pi , since the denominator vanishes at both of these points. You need to be more specific about the class of functions ψ \psi to which you are applying the argument. to ensure that all the integrals involved exist. The class of functions over which you are conducting your analysis could not include 2 π x \tfrac{2}{\pi}x , for example.

Mark Hennings - 2 years, 2 months ago

Log in to reply

Wait..Sir, I don't get what you mean....I only know about Calculus of Variations approach from seeing some solutions on Brilliant...Could you suggest me some book to start learning this from??

Aaghaz Mahajan - 2 years, 2 months ago

Log in to reply

He's saying that if ψ = 0 \psi =0 or 1 1 , then L = ψ 4 ψ ( 1 ψ ) L=\frac { \psi ' }{ 4\psi \left( 1-\psi \right) } would not be defined since the denominator is zero. As for books, I don't know any but you can look up some videos by Faculty of Khan (no relation to Khanacademy, I think) on YouTube. He's a really fantastic content creator.

Wesley Low - 2 years, 2 months ago

Log in to reply

@Wesley Low No no....That part was obvious.....What I didn't get was how to be more specific about the class of functions......Is there a method for that?? Thanks for the youtube channel......will surely check it out!!

Aaghaz Mahajan - 2 years, 2 months ago

I learned the subject reading Gelfand & Fomin's book (quite cheap on Amazon, and there may be a dubious .pdf download available).

Here is a quick run-through of the basic idea, Suppose that F F is a twice continuously differentiable function of three variables, and let X α , β X_{\alpha,\beta} be the set of all twice continuously differentiable functions f f such that f ( a ) = α f(a) = \alpha and f ( b ) = β f(b)=\beta . Consider the functional I I , defined on X α , β X_{\alpha,\beta} , given by I [ g ] = a b F ( x , g ( x ) , g ( x ) ) d x g X α , β I[g] \; = \; \int_a^b F(x,g(x),g'(x))\,dx \hspace{2cm} g \in X_{\alpha,\beta} If f X α , β f \in X_{\alpha,\beta} extremizes I I over X α , β X_{\alpha,\beta} , then for any h X 0 , 0 h \in X_{0,0} the function Q ( t ) = I [ f + t h ] Q(t) \; = \; I[f + th] has a turning point at t = 0 t=0 . But 0 = Q ( 0 ) = a b ( F f h ( x ) + F f h ( x ) ) d x 0 \; = \; Q(0) \; = \; \int_a^b \left(\frac{\partial F}{\partial f}h(x) + \frac{\partial F}{\partial f'}h'(x)\right)\,dx Integrating by parts, we obtain 0 = Q ( 0 ) = a b h ( x ) [ F f d d x ( F f ) ] d x 0 \; = \; Q'(0) \; = \; \int_a^b h(x)\left[\frac{\partial F}{\partial f} - \frac{d}{dx}\left(\frac{\partial F}{\partial f'}\right)\right]\,dx Since this holds for all h X 0 , 0 h \in X_{0,0} we can deduce (how!) that F f d d x ( F f ) = 0 \frac{\partial F}{\partial f} - \frac{d}{dx}\left(\frac{\partial F}{\partial f'}\right) \; = \; 0 You can see that there is a lot of technical work that has to be done. Integrals have to exist, and integration by parts has to work. When you are faced with the possibility that integrals could fail to converge, it is difficult to perform this analysis!

Mark Hennings - 2 years, 2 months ago

Log in to reply

@Mark Hennings Thanks a looooot Sir!!

Aaghaz Mahajan - 2 years, 2 months ago

I see what you mean, but I don't have enough experience with functions to provide a way to circumvent this. Do you know how to avoid this problem, or prove that for functions of ψ \psi where L L doesn't exist that I ( ψ ) I\left( \psi \right) is not minimised, using a calculus of variations approach?

Wesley Low - 2 years, 2 months ago

Log in to reply

Well, if the integral does not exist, then it is infinite, so the functional is not being extremized. However, as I said in my proof, this question is best not solved using CofV.

Mark Hennings - 2 years, 2 months ago

@Wesley Low Thanks!!! I just wanted to confirm my approach...:)

Aaghaz Mahajan - 2 years, 2 months ago

The given integral attains an extrimum for the function (sinx)^2 (a standard variational problem). The value of the integral for this function is π/2

0 pending reports

×

Problem Loading...

Note Loading...

Set Loading...