8
$\begingroup$

Let $S\subset \mathbb{R}^3$ be a smooth surface with the Gauss normal map $N:S\to S^2$.

Then for every $x\in S$, the differential $(dN)_x:T_xS\to T_{N(x)}S^2$ can be considered as an endomorphism of the tangent space $T_xS$ since $T_xS$ is parallel to $T_{N(x)}S^2$. So from now on, without any ambiguity and in a unique way, we count $dN$ as an endomorphism of the tangent bundle $TS$ of the surface $S$. So $dN$ defines a linear operator $$dN:\chi^{\infty}(S)\to \chi^{\infty}(S)$$ where $\chi^{\infty}(S)$ is the space of all smooth vector fields on $S$.

Under which geometric conditions on $S$ this operator preserve the Lie bracket of $\chi^{\infty}(S)$? Under which conditions on $S$, the range $dN(\chi^{\infty}(S))$ of this operator is a Lie algebra?

Of course we can ask the same question for every codimension $1$ submanifold $S$ of $\mathbb{R}^n$.

$\endgroup$
5
  • 2
    $\begingroup$ $dN$ is known as the shape operator and is a $(1,1)$-tensor, i.e. an endomorphism of $T_*S$. Based on a quick back-of-the-envelope calculation, I believe that $dN$ has to be either the identity map or the zero map. In the first case, it is a sphere. In the second, it is a flat surface. $\endgroup$
    – Deane Yang
    Commented Aug 10, 2018 at 21:38
  • $\begingroup$ @DeaneYang I think the eigenvalues of $dN$ are principal curvatures, then for cylinder it is neither identically zero nor "identity". Right? $\endgroup$ Commented Aug 11, 2018 at 3:06
  • 1
    $\begingroup$ that is correct. The eigenvalues of the shape operator are the principal curvatures. And the shape operator of the cylinder is neither the identity map nor zero. $\endgroup$
    – Deane Yang
    Commented Aug 11, 2018 at 3:36
  • $\begingroup$ @DeaneYang so my question can be read, when the shape operator generate a lie algebra morphism on the space of vector fields on our surface?When its range is a Lie algebra? $\endgroup$ Commented Aug 11, 2018 at 9:29
  • $\begingroup$ @DeaneYang Dear prof. Yang I am sorry if my question is elementary. To be honnest I did not understand a relation between my question and the part of your statement that "......this familly consist of either whole morphisms or 0 or identity.." I would appreciate if you more explain. $\endgroup$ Commented Aug 14, 2018 at 7:53

1 Answer 1

8
+50
$\begingroup$

ADDED: I checked my calculation of the first displayed equation, and it appears to be correct. If so, it looks to me that it already implies that $A$ is either the identity or zero and therefore the second paragraph isn't even needed.

The shape operator is an example of a bundle map $A: S\rightarrow S$, where $S$ is a $d$-dimensional smooth manifold. If it preserves the Lie algebra structure of vector fields, then $[AX,AY] = A[X,Y]$ for any smooth vector fields $X$ and $Y$. For any $p \in S$ choose local coordinates $x = (x^1, \dots, x^d)$ such that $p$ is at the origin. For any $1 \le i,j,k \le d$, let $X = \partial_i$ and $Y= x^j\partial_k$. A straightforward calculation shows that at the origin, $$ [AX,AY] =A^i_jA_k^p\partial_p\text{ and }A[X,Y] = \delta_i^jA_k^p\partial_p. $$ Therefore, at each point in $S$, $A^2 = A$ and therefore $A$ is diagonalizable with eigenvalues $0$ and $1$.

Now suppose $A$ is neither the identity nor $0$ at a point $p \in S$. Let $X$ and $Y$ be smooth vector fields that are nonzero at $p$ and satisfy $AX = 0$ and $AY = Y$ at $p$. Then $A[X,Y] = [AX,AY] = 0$ at $p$. Let $f$ be a smooth function such that $Xf(p) \ne 0$. Then on one hand, $$ [AX,A(fY)] = 0. $$ On the other hand, at $p$, $$ [AX,A(fY)] = A[X,fY] = A((Xf)Y + f[X,Y]) = (Xf)AY + fA[X,Y] = (Xf)Y \ne 0. $$ This is a contradiction and therefore $A$ is either the identity or $0$.

$\endgroup$
1
  • $\begingroup$ Thank you very much for your very interesting answer and your attention to my question. $\endgroup$ Commented Aug 20, 2018 at 9:50

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .