2
$\begingroup$

In my studies I've come across the hessian in the context of Riemannian geometry. I use the following definition of the hessian

$$ H^f(X,Y)=XYf-(\nabla_XY)f=\langle \nabla_X(\operatorname{grad} f),T\rangle. $$

I here want to show that at a critical point $p$ of $f$ we get the following

$$ H^f(X_p,Y_p)=X_p(Yf)=Y_p(Xf),\quad \text{for all } X,Y. $$

I'm not quite sure how to start.

$\endgroup$
  • 1
    $\begingroup$ You will know the answer once you write down the definition of a critical point. $\endgroup$ – user99914 Sep 29 '15 at 13:16
  • $\begingroup$ (You'll also need the fact that $[X,Y]$ is a vector field to get the second equality.) $\endgroup$ – Anthony Carapetis Sep 30 '15 at 5:27
1
$\begingroup$

I'll record a quick answer here: if $M$ is a smooth manifold and $\nabla$ is a linear connection in $M$, then the covariant hessians according to $\nabla$ are symmetric tensors if and only if $\nabla$ is torsion free. This happens, for instance, if $(M,\langle\cdot,\cdot\rangle)$ is a pseudo-Riemannian manifold and $\nabla$ is its Levi-Civita connection.

If given a tensor field $A \in \mathscr{T}^r_{\;\;s}(M)$, we define $\nabla A \in \mathscr{T}^r_{\;\;s+1}(M)$ by $$\begin{align}\nabla A(&\omega^1,\ldots, \omega^r, X_1,\ldots,X_s;X) = \nabla_XA(\omega^1,\ldots,\omega^r,X_1,\ldots,X_s)\\ &= X(A(\omega^1,\ldots, \omega^r,X_1,\ldots, X_s)) - \sum_{i=1}^r A(\omega^1,\ldots, \nabla_X\omega^i, \ldots, \omega^r,X_1,\ldots,X_r) \\ & \qquad\qquad - \sum_{i=1}^r A(\omega^1,\ldots, \omega^r, X_1,\ldots \nabla_XX_i,\ldots, X_s),\end{align}$$then by definition $${\rm Hess}(f)(X,Y) \doteq \nabla(\nabla f)(X;Y) = Y(X(f)) - {\rm d}f(\nabla_YX).$$If $\tau(X,Y) = \nabla_XY - \nabla_YX - [X,Y]$ is the torsion of the connection, we have $${\rm Hess}(f)(X,Y) = {\rm Hess}(f)(Y,X) + {\rm d}f(\tau(X,Y)).$$If we have no torsion, and $p$ is a critical point of $f$, then ${\rm d}f_p = 0$, and so $${\rm Hess}(f)_p(X_p,Y_p) = {\rm Hess}(f)_p(Y_p,X_p) = X_p(Y(f)) = Y_p(X(f)).$$

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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