3
$\begingroup$

Let $(M^2,g)$ be a noncompact orientable Riemannian surface without boundary. Let $A \in \Gamma(\operatorname{Sym}(TM))$ be a section of the bundle of symmetric endomorphisms of $TM$, that is, for each $p \in M$, the linear map $A_p : T_p M \to T_p M$ is symmetric with respect to the inner product $g_p$. Assume that the eigenvalues $\lambda_1(p), \lambda_2(p) \in \mathbb{R}$ of $A_p$ are distinct for all $p \in M$.

Do there exist two smooth globally defined unit vector fields $E_1, E_2$ on $M$ such that $E_i(p)$ is an eigenvector of $A_p$ associated with $\lambda_i(p)$ for $i=1,2$ and all $p \in M$?

Does the topology of $M$ impose any restrictions on the existence of such vector fields?

$\endgroup$
1
  • $\begingroup$ I think you want to look at the moonstar and similar examples in McCleary, Geometry from a differentiable viewpoint. $\endgroup$
    – Ben McKay
    Commented Feb 15, 2022 at 10:33

1 Answer 1

7
$\begingroup$

Not necessarily. To construct a counter-example, start from the other direction. Suppose that the tangent bundle of $M$ can be split as the direct sum $TM = L_1\oplus L_2$ where $L_1$ and $L_2$ are non-trivial smooth line bundles. Then you can easily construct a metric $g$ on $M$ such that $L_1$ and $L_2$ are $g$-orthogonal. Let $A:TM\to TM$ be the linear transformation that is the identity on $L_1$ and minus the identity on $L_2$. Then $A$ is symmetric with respect to $g$, but the two eigenbundles of $A$ have no non-trivial sections.

A simple example of such a splitting is to let the (orientable) cylinder $M$ be the quotient of the plane $\mathbb{R}^2$ by the translation $(x,y)\mapsto (x{+}\pi,y)$, and let $L_1$ be the bundle spanned by the vector field $Z = \cos x\,\partial_x + \sin x\,\partial_y$ (which is only defined up to a sign on $M$), while $L_2$ is spanned by the vector field $W = -\sin x\,\partial_x + \cos x\,\partial_y$ (again defined only up to a sign on $M$).

$\endgroup$
4
  • $\begingroup$ Can such a splitting exist if $M$ is orientable (I forgot to add this hypothesis..)? $\endgroup$ Commented Feb 15, 2022 at 10:40
  • $\begingroup$ Yes. See above. $\endgroup$ Commented Feb 15, 2022 at 10:42
  • $\begingroup$ So, "every line bundle is trivial on an orientable surface" is true only when the surface is compact? $\endgroup$ Commented Feb 15, 2022 at 10:49
  • 2
    $\begingroup$ @EduardoLonga: Actually, the only connected surfaces such that every (real) line bundle over them is trivial are the sphere and the plane. Meanwhile, the example I gave of the cylinder also drops to the (compact) torus if you also reduce by the translations $(x,y)\mapsto(x,y{+}\pi)$, so even a compact orientable surface can have a splitting of its tangent bundle into non-orientable lines. $\endgroup$ Commented Feb 15, 2022 at 12:42

You must log in to answer this question.

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