I am trying to build intuition on Legendre transforms. Arnold's Mathematical Methods of Classical Mechanics has some nice geometric interpretations, but he does not provide a proof that the Legendre transform of a convex function is convex. I know we can prove it by applying some inequalities, but is there a nice geometric argument (with a picture) that allows us to see that it is true?


The proof you seek, and the geometric insights you desire, and the links to physical dynamics, all are given in an outstanding survey article:

@article{Author = {Zia, R. K. P. and Redish, Edward F. and McKay, Susan R.}, 
Journal = {American Journal of Physics}, 
Number = {7}, Pages = {614-622}, 
Title = {Making sense of the {L}egendre transform}, 
Volume = {77}, Year = {2009}}

For the convexity proof in particular, see Equation 47 and the discussion that follows.


If you draw a collection of straight lines on a paper and you inspect the "upper contour" of these lines, you notice that this contour is convex. Similarly, the Legendre transform of a function is basically a supremum (=upper contour) over a family of affine functions (=straight lines).


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.