3
$\begingroup$

In page 79 of Davies's book on Heat Kernels and spectral theory, the author proves that $$\lVert e^{-Ht}f \rVert_2 \leq c_1t^{-\mu/ 4}\lVert f \rVert_1$$ where the norms are $L^p$ norms. He states

by duality, it follows that $$\lVert e^{-Ht}f \rVert_\infty \leq c_1t^{-\mu/ 4}\lVert f \rVert_2$$

Can someone explain what exactly this "duality" argument is? Above, $e^{-Ht}f$ we can take to be the solution of the heat equation with initial data $f$ (where $H$ is the Laplacian).

$\endgroup$
1
  • 1
    $\begingroup$ I do not think this is research level. However, I do think it would be useful to record a discussion of certain duality arguments, maybe over at StackExchange. See Q1 Q2 Q3 $\endgroup$
    – user70229
    Feb 9, 2016 at 14:10

1 Answer 1

5
$\begingroup$

You have shown that $e^{-Ht}\colon L^1\to L^2$ is bounded with the given operator norm. It follows that the dual operator $$\left(e^{-Ht}\right)^*\colon (L^2)^* \to (L^1)^*$$ is bounded with the same operator norm, because of the characterisation $$\|T\|_{L^1\rightarrow L^2}=\|T^*\|_{L^\infty\rightarrow L^2}=\sup_{f,g}\frac{\langle Tf,g\rangle}{\|f\|_1\|g\|_2}\,.$$
Since $(L^1)^* = L^\infty$ and $(L^2)^* = L^2$ (these are isomorphisms, actually), it remains to check that $\left(e^{-Ht}\right)^* = e^{-Ht}$.

For this note that $$\int f(x)(e^{-Ht}g)(x)dx = \iint f(x)e(x,y;t)g(y)dxdy$$ where $e(x,y;t)$ is the heat kernel, which is real and symmetric in the variables $x,y$. It follows that $\left(e^{-Ht}\right)^*$ is again integration against the heat kernel as claimed.

$\endgroup$
7
  • 1
    $\begingroup$ Thanks. your domain and range are mixed in the first line btw. $\endgroup$ Feb 9, 2016 at 13:32
  • $\begingroup$ In the second line too. Fixed now. $\endgroup$ Feb 9, 2016 at 13:48
  • $\begingroup$ In the last step, all we really need is that the semigroup $e^{-Ht}$ is a bounded self-adjoint operator, which follows from the fact that $H$ is self-adjoint. We don't need to use the fact that a heat kernel exists, which is quite a bit more difficult to prove. $\endgroup$ Feb 9, 2016 at 15:09
  • $\begingroup$ This is not obvious to me. I agree that $T=\exp(-Ht)\colon L^2\to L^2$ has $T^*=T$, but I don't see why that automatically shows something about the dual of $S=\exp(-Ht)\colon L^1\to L^2$. $\endgroup$ Feb 9, 2016 at 15:32
  • 1
    $\begingroup$ Basically you differentiate both sides wrt $t$ and use the self-adjointness of the Laplace operator, but this brushes asides various minor technical issues. $\endgroup$ Feb 9, 2016 at 17:29

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.