1
$\begingroup$

Is there a way to write the negative Laplacian on the 2-sphere as a decomposition of an operator $A$ and its adjoint $A^*$? I am interested in finding such a decomposition, but I could not get one by simple algebraic manipulations. Does anybody know if something like that exists?

So again, I want to get: $- \Delta = A^* A$ on the $\mathbb{S}^2$. In case that you know a way to get this, I would be also interested in possible domains(if this is non-trivial), such that $A$ is closed.

I think, in principle, the chances that something like that exists are fairly good, as the Laplacian is positive and self-adjoint.

Edit: I know that there is a (i.e. unique positive) square root for these kinds of operators, but I am rather looking for an analogue to the decomposition $-\frac{d^2}{dx^2}= (- \frac{d}{dx}) ( \frac{d}{dx}).$ So, I am rather looking for a casual root( in the sense that the root is comfortable for calculations). I hope I could give you an idea, what I am looking for.

$\endgroup$

3 Answers 3

2
$\begingroup$

Let $B$ be the non-negative square root of the usual self-adjoint realization of $-\Delta$. Then I claim that $A^*A=-\Delta$ precisely if $A=UB$ for a $U$ satisfying $U^*U=1$. In other words, $U$ varies over the isometries $U: L^2\to R(U)$.

Clearly, such an $A$ satisfies $A^*A=-\Delta$. Conversely, if $A^*A=-\Delta$, then $A1=0$ also (I write $1$ for the constant function $f=1$). This follows because $N(A^*)=R(A)^{\perp}$ and $-\Delta 1=0$. On the orthogonal complement $H$ of $1$, I can invert $B$. Let $C$ be this inverse on $H$, and $C1=0$, say. Let $V=AC$. Then $V^*V = CB^2C=P$, the projection onto $H$. So $V$ is a partial isometry with initial space $H$, and $VB=ACB=AP=A$, since $A$ annihilates $H^{\perp}=L(1)$. Since $B1=0$ also, I can replace $V$ by an isometry $U$ here, without destroying the identity $UB=A$.

$\endgroup$
2
$\begingroup$

Let me follow your notations with $\Delta=\sum_{1\le j\le 3}\partial_{x_j}^2$. You have with $r=\Vert x\Vert$ (the Euclidean norm) $$ r^2\Delta=(r\partial_r)^2+r\partial_r+\Delta_{\mathbb S^2},\quad\text{where $\Delta_{\mathbb S^2}$ is the Laplace operator on the sphere.} $$ Considering an homogeneous harmonic polynomial $p_k$ of degree $k$ on $\mathbb R^3$, we have $$ 0=r^2\Delta p_k=(k^2+k)p_k+\Delta_{\mathbb S^2}p_k, $$ and from that identity, it is easy to get that the spectrum of $-\Delta_{\mathbb S^2}$ is $$\{k^2+k\}_{k\in \mathbb N}=\{(k+\frac{1}{2})^2\}_{k\in \mathbb N}-\frac{1}{4}.$$ We have thus $$ -\Delta_{\mathbb S^2}+\frac{1}{4}=\sum_{k\in \mathbb N}(k+\frac{1}{2})^2\mathbb P_k, $$ where $\mathbb P_k$ is the $L^2(\mathbb S^2)$ orthogonal projection on the space $\Sigma_k$ of spherical harmonics with degree $k$ ($\Sigma_k$=restriction to the sphere of harmonic polynomials of degree $k$). As a result, we get $$ -\Delta_{\mathbb S^2}+\frac{1}{4}=\Lambda ^2,\quad \Lambda=\sum_{k\in \mathbb N}(k+\frac{1}{2})\mathbb P_k. $$ Note that the same procedure could be done in any dimension, that $\Lambda$ is an elliptic pseudodifferential operator of order $1$ on the sphere, which is differential only in the case $\mathbb S^1$ with $\Delta_{\mathbb S^1}=d^2/(d\theta)^2$.

$\endgroup$
2
$\begingroup$

Laplace operator on functions (as a positive operator convention) is defined by $\Delta=d^*d$ where $d:C^\infty(M)\to C^\infty(T^*M)$ is the exterior derivative.

$\endgroup$

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.