From: Maury Barbato on
Hello,
let C be a compact convex subset of the usual euclidean
space R^n, and suppose that C has at least two points. Does there exist two distinct points x, y in C such that
the linear function f(z) = z*(x-y) (where * denotes the
dot product in R^n) attains its maximum on C in x and
its minimum on C in y?


Thank you very much for your attention.
My Best Regards,
Maury Barbato

PS Note that my question has a precise geometric meaning,
and my geometric intuition suggests me it has a positive
answer.
From: Robert Israel on
Maury Barbato <mauriziobarbato(a)aruba.it> writes:

> Hello,
> let C be a compact convex subset of the usual euclidean
> space R^n, and suppose that C has at least two points. Does there exist
> two distinct points x, y in C such that
> the linear function f(z) = z*(x-y) (where * denotes the
> dot product in R^n) attains its maximum on C in x and
> its minimum on C in y?

Yes. Consider the continuous function g defined on C x C by g(s,t) = |s-t|
(where |.| is the Euclidean norm).
By compactness, g attains its supremum, i.e. there are x,y in C such that
|x-y| >= |s-t| for all s,t in C. Then for any s, t in C
f(s) - f(t) = (s - t)*(x - y) <= |s - t| |x - y| <= |x - y|^2 = f(x) - f(y).
In particular (taking t = y) f(s) <= f(x) and (taking s = x) f(t) >= f(y),
so f attains its maximum at x and its minimum at y.
--
Robert Israel israel(a)math.MyUniversitysInitials.ca
Department of Mathematics http://www.math.ubc.ca/~israel
University of British Columbia Vancouver, BC, Canada
From: Maury Barbato on
Robert Israel wrote:

> Maury Barbato <mauriziobarbato(a)aruba.it> writes:
>
> > Hello,
> > let C be a compact convex subset of the usual
> euclidean
> > space R^n, and suppose that C has at least two
> points. Does there exist
> > two distinct points x, y in C such that
> > the linear function f(z) = z*(x-y) (where * denotes
> the
> > dot product in R^n) attains its maximum on C in x
> and
> > its minimum on C in y?
>
> Yes. Consider the continuous function g defined on C
> x C by g(s,t) = |s-t|
> (where |.| is the Euclidean norm).
> By compactness, g attains its supremum, i.e. there
> are x,y in C such that
> |x-y| >= |s-t| for all s,t in C. Then for any s, t
> in C
> f(s) - f(t) = (s - t)*(x - y) <= |s - t| |x - y| <=
> |x - y|^2 = f(x) - f(y).
> In particular (taking t = y) f(s) <= f(x) and (taking
> s = x) f(t) >= f(y),
> so f attains its maximum at x and its minimum at y.
> --
> Robert Israel
> israel(a)math.MyUniversitysInitials.ca
> Department of Mathematics
> http://www.math.ubc.ca/~israel
> University of British Columbia Vancouver,
> BC, Canada

Chapeau, monsieur le professeur! Your solutions are
always clever and clear!
Thank you very very ... much, prof. Israel!
Friendly Regards,
Maury Barbato