Let us consider Minkowski inner product on $\mathbb R^{1+n}$, defined by $$ \langle v,w \rangle = -v_0w_0+\sum_{j=1}^n v_j w_j\quad \,\forall\, v,w \in \mathbb R^{1+n}.$$ We say that a vector $v$ is null if $\langle v,v\rangle =0$. Moreover, given a symmetric matrix $A$, we say that $A$ is null positive definite if $\langle Av,v\rangle >0$ for all non-zero null vectors $v$, and null semi-positive definite if $\langle Av,v\rangle \geq 0$ for all null vectors $v$.

Suppose $A(t)$ is a family of symmetric matrices smoothly depending on $t \in [0,1]$ and suppose that $A(t)$ is null positive definite for all $t \in (0,t_0)$. Suppose also that $A(t_0)$ is null semi-positive definite but it is not null positive definite. Does it follow that there exists a null vector $v$ such that $$ A(t_0) v =0?$$


Consider the case $n = 1$.

Let $A = \begin{pmatrix} -1 \\ & -2t\end{pmatrix}$

If $v = (1,1)$, then $Av = (-1,-2t) \neq 0$. If $w = (1,-1)$, then $Aw = (-1, 2t) \neq 0$.

But $\langle Av,v\rangle = \langle Aw,w\rangle = 1-2t$, and so $A$ is null positive definition for $t \in (0,\frac12)$, null semi-positive definition at $t = 1/2$, but there are no non-trivial null vectors in the kernel of $A$ at $t = 1/2$.


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.