Calculus of variations changing variables

10 replies · 2K views · last reply
bobred
Messages
170
Reaction score
0

Homework Statement


Hi
I am given the functional

Integral S[y] = integral from a to b of x^3 y'^2 / y^4 dx.


I am asked to show that if
Equation x = u^3 written as plain text.
and with an appropriate value for
png.png
that

Integral S[y] equals the integral from B to A of y'(u) squared divided by y to the fourth power, du.


Homework Equations


[/B]

The Attempt at a Solution


So I get

Chain-rule equation: dy/dx = (dy/du)(du/dx) = y'(u)/(dx/du) = y'(u)/(βu^(3-1)).


Equation S[y] = integral of (u^(3/3) / y^4) times (y'(u) / (beta u^(beta-1)))^2 times beta u^(beta-1) du, with the square applied to the middle fraction.

Equation S[y] = integral of (1/y^4) times (u^(2β+1)/β) times y^2 d u, written as plain text: S[y] = ∫ (1/y^4)(u^(2β+1)/β) y^2 du.


If I set
Equation: beta = -1/2
then I get

Equation S[y] = 2 times the integral from B to A of y'(u)^2 / y^4 du, written as a functional with integrand y prime squared over y to the fourth.

I think that it is correct but what about the factor of 2?
 
Your work looks correct. Unless you do another variable transformation, you will not get rid of the 2.
 
I imagine that since you are trying to find stationary values of this integral, they are not affected by a constant multiple, so you can drop the factor. I think it should be -2, but that's a minor point.
 
PeroK said:
I think it should be -2, but that's a minor point.

This would depend on the order of the integration limits (note that ##a## was the lower integration limit in ##x## but ##A## is the upper integration limit in ##u## - of course I am just making the arbitrary inference that ##a## corresponds to ##A## here ...).
 
Orodruin said:
This would depend on the order of the integration limits (note that ##a## was the lower integration limit in ##x## but ##A## is the upper integration limit in ##u## - of course I am just making the arbitrary inference that ##a## corresponds to ##A## here ...).

Yes, didn't notice that.
 
Thanks guys.
I may have missed this in my notes PeroK but if we are trying to find stationary points of a functional constant multiples can be ignored?
James
 
bobred said:
but if we are trying to find stationary points of a functional constant multiples can be ignored?

Yes, but there is nothing strange about this. It works this way for functions as well, if ##f(x)## has the stationary point ##x=2##, then so does ##2f(x)##.
 
  • Like
Likes   Reactions: bobred
bobred said:
Thanks guys.
I may have missed this in my notes PeroK but if we are trying to find stationary points of a functional constant multiples can be ignored?
James

Positive multiples can be ignored, but omitting negative multiples changes the direction of optimization.
 
  • Like
Likes   Reactions: bobred
Ray Vickson said:
Positive multiples can be ignored, but omitting negative multiples changes the direction of optimization.

But it does not change the fact that the point is stationary. Just exchanges minima for maxima.
 
  • Like
Likes   Reactions: bobred
Orodruin said:
But it does not change the fact that the point is stationary. Just exchanges minima for maxima.

Of course, but that is another issue. It certainly changes the second-order tests, so that instead of looking to see if the Hessian is positive-definite, we look instead to see if it is negative-definite. However, when I taught this stuff I recommended that students just "memorize" the conditions for a minimum, then switch the sign of the objective if the problem was a maximization; that eliminates the need for a whole raft of special cases, etc.