IM Gelfand and SV Fomin, Calculus of Variations
Sec. 4.2, Case 2: Walkthrough

Mark Vuletic

Note: This document uses MathJax to typeset LaTeX. Please enable JavaScript or you will just see raw LaTeX.

Source: IM Gelfand, SV Fomin. 1963. Calculus of Variations. Mineola, NY: Dover. Tr. RA Silverman. p. 18-19.

Problem: In Sec. 4.2, Case 2, Gelfand and Fomin claim without further comment that we can move from

\( F_y - \frac{d}{dx} F_{y'} = F_{y} - F_{y'y}y' - F_{y'y'}y'' \)


\( F_y y' - F_{y'y}y'^{2} - F_{y'y'}y'y'' = \frac{d}{dx}(F - y'F_{y'}) \)

by multiplying the first equation by \(y'\). Verifying this turned out to be non-trivial for me, though (as is the way in math) transparent in retrospect, so I walk through it here.


We start with

\( F_y - \frac{d}{dx} F_{y'} = F_{y} - F_{y'y}y' - F_{y'y'}y'' \tag{1} \)

Multiply by \( y' \) as instructed:

\( F_yy' - y' \frac{d}{dx} F_{y'} = F_{y}y' - F_{y'y}y'^2 - F_{y'y'}y'y'' \tag{2} \)

Reverse the order (you wouldn't write this kind of thing out in a normal proof, of course, but this is a walkthrough):

\( F_{y}y' - F_{y'y}y'^2 - F_{y'y'}y'y'' = F_y y' - y' \frac{d}{dx} F_{y'} \tag{3} \)

You can see above that we already have the left-hand side of the target, so we start to work on the right-hand side. The key thing to do now is to insert \( F_{y'}y'' - F_{y'}y'' \) into the middle of the right-hand term.1 How on earth do we know to do this? I can't speak for the Russians, whose mathematical powers are so far beyond mine as to seem magical to me, but I eventually figured it out by working backward. Before that, I went down innumerable blind alleys. Anyway, doing so gives us

\( F_{y}y' - F_{y'y}y'^2 - F_{y'y'}y'y'' = F_y y' + \left( F_{y'}y'' - F_{y'}y'' \right) - y' \frac{d}{dx}F_{y'} \tag{4} \)

I regroup the terms above, simply as a visual aid:

\( F_{y}y' - F_{y'y}y'^2 - F_{y'y'}y'y'' = \left( F_y y' + F_{y'}y'' \right) - \left( F_{y'}y'' - y' \frac{d}{dx}F_{y'} \right) \tag{5} \)

Rewrite the first group in 5 as \( \frac{dF}{dx} \). To see the equivalence, remember that in Case 2—the case Gelfand and Fomin are dealing with here—\( F \) is a function only of \(y\) and \(y'\). Hence, \( \frac{dF}{dx} = \frac{dF}{dy}\frac{dy}{dx} + \frac{dF}{dy'}\frac{dy'}{dx} \), which is just a different way of writing \( F_y y' + F_{y'}y'' \). So we have

\( F_{y}y' - F_{y'y}y'^2 - F_{y'y'}y'y'' = \frac{dF}{dx} - \left( F_{y'}y'' - y' \frac{d}{dx}F_{y'} \right) \tag{6} \)

Rewrite the second group as \(\frac{d}{dx}\left(y'F_{y'}\right)\). To see the equivalence, just apply the product rule to this new term, then do a tiny bit of rearranging, and you will get the second group in 5. We now have:

\( F_{y}y' - F_{y'y}y'^2 - F_{y'y'}y'y'' = \frac{dF}{dx} - \frac{d}{dx}\left(y'F_{y'}\right) \tag{7} \)

Finally, pull the \( \frac{d}{dx} \) out from the right-hand side of 7:

\( F_{y}y' - F_{y'y}y'^2 - F_{y'y'}y'y'' = \frac{d}{dx} \left( F - y'F_{y'} \right) \tag{8} \)

And there you go.


1 We can do this because \( F_{y'}y'' - F_{y'}y'' = 0 \).

Last updated: 25 Nov 2016

Pleased? Angered? Confused? Have something else you would like
me to write about? Please send in your questions and comments!