next up previous
Next: About this document ...

Fourth Problem set for 18.155
Due November 7 in class or 2-174.

Especially for undergraduates I will accept just one of these problems as enough work for the week. Notice that the first problem is considered the deepest result in `elementary' distribution theory.

Problem 1 (Schwartz kernel theorem)  

  1. Show (mostly recalling things from class) that the topology of $ \mathcal{S}(\mathbb{R}^n)$ can be expressed as the projective limit of the Hilbert topologies on the weighted Sobolev spaces $ \langle
x\rangle^{-k}H^k(\mathbb{R}^n)$ as $ k\to\infty$
  2. Conclude that the dual space $ \mathcal{S}'(\mathbb{R}^n)$ can be topologized as the inductive limit of the weighted Sobolev spaces $ \langle
x\rangle^{k}H^{-k}(\mathbb{R}^n)$ as $ k\to\infty.$
  3. Show that the Fourier transform, multiplication by $ \langle x\rangle
^m$ for any $ m$ and the map $ \langle D\rangle ^m$ (defined as the preceeding map conjugated by the Fourier transform) are all isomorphisms of both $ \mathcal{S}(\mathbb{R}^n)$ and $ \mathcal{S}(\mathbb{R}^n).$
  4. Show that for a linear map (an operator)

    $\displaystyle A:\mathcal{S}(\mathbb{R}^n)\longrightarrow \mathcal{S}'(\mathbb{R}^n)$ (1)

    continuity with respect to these topologies is equivalent to the existence of some $ k$ such that $ A$ extends by continuity (in the Hilbert norms) to $ A:\langle x\rangle^{-k}H^k(\mathbb{R}^n)\longrightarrow
\langle x\rangle^{k}H^{-k}(\mathbb{R}^n).$
  5. Show how to compose $ A$ on the left and right with maps from (3) and arrive at a continuous linear map $ A'$ from $ H^{-n}(\mathbb{R}^n)$ to $ H^n(\mathbb{R}^n).$
  6. Recall the Sobolev embedding theorem and that the delta `function' at any point is in $ H^{-n}(\mathbb{R}^n)$ and use this to conclude that the formula $ a(x,y)=(A'(\delta _y))(x)$ defines a continuous bounded function on $ \mathbb{R}^{2n}).$
  7. Show that the map

    $\displaystyle \mathcal{S}(\mathbb{R}^n)\times \mathcal{S}(\mathbb{R}^n)\longrig...
...bb{R}^{2n}), (\phi,\psi )\longmapsto \phi \boxtimes\psi (x,y)=\phi (x)\psi (y)$    

    is jointly continuous (i.e. is continuous in the product metric topology).
  8. Show that if $ \beta \in\mathcal{S}'(\mathbb{R}^{2n})$ then the formula

    $\displaystyle (B\psi)\phi) =\beta(\phi \boxtimes\psi ) \forall \phi,\psi \in\mathcal{S}(\mathbb{R}^n)$ (2)

    defines a continuous linear map $ B:\mathcal{S}(\mathbb{R}^n)\longrightarrow\mathcal{S}'(\mathbb{R}^n).$
  9. Going back to (6) show that $ a$ thought of as a distribution defines $ A'$ in this way.
  10. Conclude that every continuous linear operator (0.1) arises from the construction (0.2).
  11. If you get this far and still have the energy, show that conversely $ \beta$ is determined by $ B$ through (0.2) and finally arrive at Schwartz kernel theorem:
  12. Theorem 0.1   There is a 1-1 correspondence between continuous linear maps

    $\displaystyle A:\mathcal{S}(\mathbb{R}^n)\longrightarrow \mathcal{S}(\mathbb{R}^{n'})$ (3)

    and elements of $ \mathcal{S}'(\mathbb{R}^{n+n'}).$

Problem 2   Work out the elementary behavior of the heat equation.
i)
Show that the function on $ \mathbb{R}\times\mathbb{R}^n,$ for $ n\ge1,$

$\displaystyle F(t,x)=\begin{cases}t^{-\frac n2}\exp\left(-\frac{\vert x\vert^2}{4t}\right) &t>0\  0& t\le0 \end{cases}$    

is measurable, bounded on the any set $ \{\vert(t,x)\vert\ge R\}$ and is integrable on $ \{\vert(t,x)\vert\le R\}$ for any $ R>0.$
ii)
Conclude that $ F$ defines a tempered distibution on $ \mathbb{R}^{n+1}.$
iii)
Show that $ F$ is $ \mathcal{C}^{\infty}$ outside the origin.
iv)
Show that $ F$ satisfies the heat equation

$\displaystyle (\partial_t-\sum\limits_{j=1}^n\partial^2_{x_j})F(t,x)=0$ in $\displaystyle (t,x)\not=0.$    

v)
Show that $ F$ satisfies

$\displaystyle F(s^2t,sx)=s^{-n}F(t,x)$ in $\displaystyle \mathcal{S}'(\mathbb{R}^{n+1})$ (4)

where the left hand side is defined by duality `` $ F(s^2t,sx)=F_s$'' where

$\displaystyle F_s(\phi)=s^{-n-2}F(\phi_{1/s}), \phi_{1/s}(t,x)=\phi(\frac t{s^2},\frac xs).$    

vi)
Conclude that

$\displaystyle (\partial_t-\sum\limits_{j=1}^n\partial^2_{x_j})F(t,x)=G(t,x)$    

where $ G(t,x)$ satisfies

$\displaystyle G(s^2t,sx)=s^{-n-2}G(t,x)$ in $\displaystyle \mathcal{S}'(\mathbb{R}^{n+1})$ (5)

in the same sense as above and has support at most $ \{0\}.$
vii)
Hence deduce that

$\displaystyle (\partial_t-\sum\limits_{j=1}^n\partial^2_{x_j})F(t,x)=c\delta (t)\delta (x)$ (6)

for some real constant $ c.$

Hint: Check which distributions with support at $ (0,0)$ satisfy (0.5).

viii)
If $ \psi \in\mathcal{C}^{\infty}_c(\mathbb{R}^{n+1})$ show that $ u=F\star\psi$ satisfies

\begin{multline}
u\in\mathcal{C}^{\infty}(\mathbb{R}^{n+1})\text{ and }\\
\sup\...
...u(t,x)\vert<\infty\
\forall S>0,\alpha \in\mathbb{N}^{n+1}, N.
\end{multline}

ix)
Supposing that $ u$ satisfies (0.7) and is a real-valued solution of

$\displaystyle (\partial_t-\sum\limits_{j=1}^n\partial^2_{x_j})u(t,x)=0$    

in $ \mathbb{R}^{n+1},$ show that

$\displaystyle v(t)=\int_{\mathbb{R}^n} u^2(t,x)$    

is a non-increasing function of $ t.$

Hint: Multiply the equation by $ u$ and integrate over a slab $ [t_1,t_2]\times\mathbb{R}^n.$

x)
Show that $ c$ in (0.6) is non-zero by arriving at a contradiction from the assumption that it is zero. Namely, show that if $ c=0$ then $ u$ in viii) satisfies the conditions of ix) and also vanishes in $ t<T$ for some $ T$ (depending on $ \psi ).$ Conclude that $ u=0$ for all $ \psi.$ Using properties of convolution show that this in turn implies that $ F=0$ which is a contradiction.
xi)
So, finally, we know that $ E=\frac1cF$ is a fundamental solution of the heat operator which vanishes in $ t<0.$ Explain why this allows us to show that for any $ \psi\in\mathcal{C}^{\infty}_c(\mathbb{R}\times\mathbb{R}^n)$ there is a solution of

$\displaystyle (\partial_t-\sum\limits_{j=1}^n\partial^2_{x_j})u=\psi, u=0$ in $\displaystyle t<T$ for some $\displaystyle T.$ (7)

What is the largest value of $ T$ for which this holds?
xii)
Can you give a heuristic, or indeed a rigorous, explanation of why

$\displaystyle c=\int_{\mathbb{R}^n}\exp(-\frac{\vert x\vert^2}4)dx?$    




next up previous
Next: About this document ...
Richard B. Melrose 2002-11-01