PDA

View Full Version : Re: Brownian motion on S^n, H^n


tessel@um.bot
Nov4-06, 03:38 PM
[Note: I am trying to find a technological solution to the word wrap
problem mentioned by a moderator, but for the nonce, if you quote, please
try to reformat any long lines in quoted text yourself, in order to save
work for the moderators! I happen to use a newsreader which makes this
task very easy, but apparently many other posters here have been unable or
unwilling to fix long lines in quoted text.]

On Tue, 5 Sep 2006, Alan wrote:

> I am looking for (refs. to) explicit formulas for the heat kernel
> p(t,x,y) on the n-sphere S^n?
>
> Also, while I have found the formulas for the analogous known kernels on
> the hyperbolic spaces H^n, I am still looking for an "elementary
> derivation" of those.

Consider the following familiar statement about the heat equation on R:

Fact: The initial value problem

u_t = u_(xx)

u(0,x) = f(x) on -infty < x < infty

has the solution

u(x,t) = 1/sqrt(2 Pi)

int_-infty^infty f(xi) exp(-(x-xi)^2/(4 t))/sqrt(2 t) dxi

Here the "kernel" in the convolution, i.e. the function of t,x,xi which we
integrate against, is the function

exp(-(x-xi)^2/(4 t))/sqrt(2 t)

which is also called the "fundamental solution", or the "Green function"
for this problem. This is the solution which describes how an initial
"delta mass" concentration of some scalar field on R spreads out as time
increases, if its evolution is governed by the heat equation.

If I understand your question correctly, you are asking for the "kernels"
appearing in analogous statements for the heat equation

u_t = Lap(u)

where Lap is the Laplace-Beltrami operator on H^n or S^n, as expressed in
some convenient coordinate chart.

E.g. for S^2 we can use a stereographic chart

ds^2 = 1/(1+x^2+y^2)^2/4 (dx^2+dy^2),

-infty < x, y < infty

and then

Lap(u) = (1+x^2+y^2)^2/4 (u_(xx) + u_(yy))

so that the "S^2 heat equation" (written in the stereographic chart) is

u_t = k (1+x^2+y^2)^2/4 (u_(xx) + u_(yy))

And on H^2 we can use the upper half plane chart

ds^2 = (dx^2 + dy^2)/y^2,

-infty < x < infty, 0 < y < infty

and then the "H^2 heat equation" (written in the UHP chart) is

u_t = y^2 ( h_(xx) + h_(yy) )

And so on.

If so, IIRC, I asked and answered these very questions (for n= 2, 3) as an
exercise here when I was yakking about the symmetries of PDEs a few years
ago. I think I also discussed the wave equation

u_(tt) = Lap u

the vibration equation

u_(tt) = Lap Lap u

and the Calabi flow

u_t = -Lap Lap u

IIRC, I compared the symmetry method with the transform method (taking
Laplace transform wrt the time variable and then multivariable Fourier
transform wrt space variables, solving the resulting algebraic equation,
and inverting the transforms to obtain the desired convolution formula). I
would consider these elementary methods, although the first may be less
familiar to many students.

Here, needless to add, a "convenient" chart is one in which these methods
work out with a minimum of fuss!

Or if I do not recall correctly, this is similar to questions studied
detail in the book by Peter J. Olver, Applications of Lie groups to
Differential Equations, 2nd Ed., Springer, 1993. Note that this book
discusses the relationship of a fundamental solution, as per symmetry
method with the Green function, as per transform method.

HTH,

"T. Essel"

Alan
Nov4-06, 03:38 PM
<tessel@um.bot> wrote in message news:edq0b5$2h0e$1@godfrey.mcc.ac.uk...

<snip>

> Or if I do not recall correctly, this is similar to questions studied
> detail in the book by Peter J. Olver, Applications of Lie groups to
> Differential Equations, 2nd Ed., Springer, 1993. Note that this book
> discusses the relationship of a fundamental solution, as per symmetry
> method with the Green function, as per transform method.
>
> HTH,

Thank you, it does -- I was able to locate some of your previous SPR posts
and they also helped. Since my original post, I have located an explicit
(understandable to me!) derivation of the fundamental heat eqn solution
for H^2 in "Eigenvalues in Riemannian Geometry" by Isaac Chavel.
So that part of my orig. question is now answered.

I would still much appreciate any reference to a work deriving,
(ideally, at a level similar to Chavel), the analogous formulas for S^n --
well,
I can do S^1 myself :-) [If it's in Olver, please ignore this, as I'll find
it there.]

Thanks!
alan

Alan
Nov4-06, 03:38 PM
"Alan" <info@optioncity.REMOVETHIS.net> wrote in message
news:PNOdnWM3mdxdJZ3YnZ2dnUVZ_r2dnZ2d@adelphia.com ...
> <tessel@um.bot> wrote in message news:edq0b5$2h0e$1@godfrey.mcc.ac.uk...

> I would still much appreciate any reference to a work deriving,
> (ideally, at a level similar to Chavel), the analogous formulas for
> ^n --
> well,
> I can do S^1 myself :-) [If it's in Olver, please ignore this, as I'll
> find
> it there.]

Nevermind -- I was able to do this calculation myself for S^n.

alan