A question from Richard Hamilton's paper "A matrix Harnack estimate for the heat equation"

78 Views Asked by At

Richard Hamilton "A matrix Harnack estimate for the heat equation. Communications in Analysis and Geometry. 1(1993), 113-126." On page 125, at the end of the proof of Theorem 4.3, I abstract the question and can be described as (Einstein's summation convention):

Let $P$ be a $(0,4)$-tensor. $P_{ikjl}$ is skew-symmetric in $i$ and $k$, skew-symmetric in $j$ and $l$, symmetric in $\{ik\}$ and $\{jl\}$. For any vectors $v$ and $w$, $P_{ikjl}v_iw_kv_jw_l\ge 0$. Let $(N_{ij})$ be a real symmetric matrix. Prove that if $v$ is a vector satisfying $N_{ij}v_j=0$, then $P_{ikjl}N_{kl}v_iv_j\ge 0$.

How to prove this proposition?

Professor Hamilton says that "it is clear if we diagonalize $N_{kl}$ with respect to a basis". However, I have not thought clearly about this, and the following example seems to negate this proposition.

Suppose that the basic vector space is $2$-dimensional. $N=\begin{bmatrix} -1 & 0\\ 0 & 0 \end{bmatrix}$, $v=\begin{bmatrix} 0\\ 1 \end{bmatrix}$, then $P_{ikjl}N_{kl}v_iv_j=-P_{2121}\le 0$.

What is wrong with this example?