5
$\begingroup$

When I read the Wiki about Legendre transformation, there is a statement

The Legendre transformation is an application of the duality relationship between points and lines.

What's the meaning of this statement?

$\endgroup$
1
  • $\begingroup$ Legendre transformations connect point geometry and Pluecker geometry. Check this link as well. $\endgroup$
    – user35952
    Apr 4, 2014 at 8:31

1 Answer 1

2
$\begingroup$

For a convex function you can do the following:

For each point on the graph of the function, draw the line tangent to the function at that point. That point can now be identified by its original $x$ and $y=f(x)$ coordinates, or by specifying the slope of that tangent line and its corresponding y-intercept. Each point maps to one and only one line, and vice versa. For convex functions, the mapping is one-to-one. There is no ambiguity. Draw a sketch and you will soon be convinced.

The Legendre transformation gives you the value of the y-intercept if you give it the slope. So the Legendre transform is a plot of $b(m)$ vs $m$ (y-intercept as a function of slope) rather than $f(x)$ vs $x$. Either function represents the same data or concept. In a sense they contain the same information.

update Thanks to @EmilioPisanty for improved wording. See comments.

$\endgroup$
4
  • $\begingroup$ This is not quite correct (in language; the content is OK). You can't specify both the slope and the y-intercept of a line, in the same way that you don't have freedom over $x$ and $y$ independently. $\endgroup$ Apr 4, 2014 at 13:56
  • $\begingroup$ If I understand you correctly, I agree completely. Is this it? --- given $x$, then $y=f(x)$ is specified and "fixed". Similarly, given $\mathrm{d}f/\mathrm{d}x$, the L.T. specifies (and "fixes") the y-intercept (of the line having that slope that is also tangent to $f(x)$ somewhere). $\endgroup$
    – garyp
    Apr 4, 2014 at 14:08
  • $\begingroup$ I would phrase your third sentence as something like "That point can now be identified by its original $x$ adn $y=f(x)$ coordinates, or by specifying the slope of that tangent line and its corresponding $y$-intercept." (But of course that's my phrasing.) $\endgroup$ Apr 4, 2014 at 14:11
  • $\begingroup$ Yes, it's your phrasing, but I still like it better than mine. Thanks. Editing answer. $\endgroup$
    – garyp
    Apr 4, 2014 at 14:13

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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