MathOverflow is a question and answer site for professional mathematicians. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

Let $X$ be a $C^\infty$-manifold, $Y$ be its $C^\infty$-submanifold. Hörmander defines the set $I^m(X,Y)$ of conormal distributions as the set of all $u \in \mathscr D'(X)$ such that $$ L_1 \ldots L_N u \in {}^{\infty}H^{\mathrm{loc}}_{(-m-n/4)}(X), \quad n = \dim X, $$ for any smooth vector fields $L_1$, $\ldots$, $L_N$ on $X$ that are tangent to $Y$ and for any $N$, where ${}^\infty H^{\mathrm{loc}}_{(-m-n/4)}(X)$ is the Besov space.

Is there some practical way to check whether some distribution $u \in \mathscr D'(X)$ is in fact from $I^m(X,Y)$ or I should always check the above inclusions? What are the examples of the most simple nontrivial conormal distributions for general $X$ and $Y$? If I take, e.g. $Y = f^{-1}(0)$ for some smooth submersion $f \colon X \to \mathbb R$ and take some $C^\infty$-density $\mu$ on $Y$ and then define for any $v \in C^\infty_c(X)$ $\langle \delta_\mu, v \rangle = \int_Y v\mu$ is it possible to say that $\delta_\mu \in I^m(X,Y)$ for some $m$?

share|cite|improve this question
up vote 4 down vote accepted

Let us take a look at the case where $Y\equiv x_n=0$ in $\mathbb R^n$.

Let $a(\underbrace{\overbrace{x_1,\dots,x_{n-1}}^{x'},x_n}_x; \xi_n)$ be a symbol in $S^m(\mathbb R^n\times \mathbb R)$ and let us consider the following "oscillatory integral": $$ u(x)=\int_{\mathbb R} e^{ix_n\cdot \xi_n} a(x,\xi_n) d\xi_n. $$ This integral is not absolutely converging and we could define it as a distribution. Note that we may assume that $a$ is supported where $\vert \xi_n\vert\ge 1$ since the contribution of a compactly supported $a$ is a smooth function. Then we have for $\phi$ in the Schwartz space, $l\in 2\mathbb N$, formally integrating by parts, \begin{align*} \langle u,\phi\rangle_{\mathscr S',\mathscr S}&=\int_{\mathbb R^n\times\mathbb R} \vert \xi_n\vert^{-l} \vert D_{x_n}\vert^l (e^{ix_n\cdot \xi_n}) a(x,\xi_n)\phi(x) dx d\xi_n\\ &=\sum_{\alpha+\beta=l}c_{\alpha\beta}\int_{\mathbb R^n\times\mathbb R} e^{ix_n\cdot \xi_n} \vert \xi_n\vert^{-l} (D_{x_n}^\alpha a)(x,\xi_n)(D_{x_n}^\beta\phi)(x) dx d\xi_n, \end{align*} which makes sense for $l>1+m$. We can take the last integral as the definition of the duality product. We note also that $u$ can easily be differentiated with respect to $x'$ (tangential derivatives) since $\partial_{x'}a\in S^m$. However the distribution $u$ is more difficult to differentiate wrt $x_n$ since $\xi_na\in S^{1+m}$; however calculating $$ x_n\partial_{x_n} u=\int_{\mathbb R} x_n\partial_{x_n}(e^{ix_n\cdot \xi_n}) a(x,\xi_n) d\xi_n +\int_{\mathbb R} e^{ix_n\cdot \xi_n} x_n\partial_{x_n}(a)(x,\xi_n) d\xi_n. $$ $$ =\int_{\mathbb R} x_n i\xi_n e^{ix_n\cdot \xi_n} a(x,\xi_n) d\xi_n +\int_{\mathbb R} e^{ix_n\cdot \xi_n} x_n\partial_{x_n}(a)(x,\xi_n) d\xi_n. $$ $$ =\int_{\mathbb R} \partial_{\xi_n}(e^{ix_n\cdot \xi_n} )\xi_n a(x,\xi_n) d\xi_n +\int_{\mathbb R} e^{ix_n\cdot \xi_n} x_n\partial_{x_n}(a)(x,\xi_n) d\xi_n. $$ $$ =-\int_{\mathbb R} e^{ix_n\cdot \xi_n} \underbrace{\partial_{\xi_n}(\xi_n a(x,\xi_n))}_{\in S^m} d\xi_n +\int_{\mathbb R} e^{ix_n\cdot \xi_n} \underbrace{x_n\partial_{x_n}(a)(x,\xi_n)}_{\in S^m} d\xi_n. $$ We can prove in particular that the wave-front-set of $u$ is included in the conormal bundle of $Y$, i.e. $$ WF u\subset\{(x',0;0,\xi_n)\}_{x'\in \mathbb R^{n-1}, \xi_n \in \mathbb R^*}$$ Your simple layer $u(x)= \omega(x')\otimes \delta(x_n)$ with a smooth $\omega$ is indeed the prototypical conormal distribution since $$ u(x)= \int_\mathbb R\omega(x') e^{2iπ x_n\xi_n}d\xi_n. $$ It is not difficult to generalize the previous discussion to the case involving $k$ codimensions.

share|cite|improve this answer

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

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