04-08-2024

Problem: Let $n > 1$ be a fixed natural number. Which of the following is an inner product on the vector space of $n \times n$ real symmetric matrices?
  • $\left\langle A, B \right\rangle = \operatorname{trace}\left( A \right) \operatorname{trace}\left( B \right) $
  • $\left\langle A, B \right\rangle = \operatorname{trace}\left( AB \right) $
  • $\left\langle A,B \right\rangle = \det (AB)$
  • $\left\langle A, B \right\rangle = \operatorname{trace}\left( A \right) + \operatorname{trace}\left( B \right) $
Solution: We recall that $\left\langle \cdot, \cdot \right\rangle $ on a vector space $V$ is a real inner product if
  • $\left\langle v,v \right\rangle > 0$ for any $v \neq 0$.
  • $\left\langle v,v \right\rangle = 0$ if and only if $v = 0$.
  • For any $v,w \in V$, $\left\langle v,w \right\rangle = \left\langle w,v \right\rangle $.
  • For any $\lambda \in \mathbb{R} $ and $v_1,v_2,v_3 \in V$ $\left\langle \lambda v_1 + v_2, v_3 \right\rangle = \lambda \left\langle v_1, v_3 \right\rangle + \left\langle v_2, v_3 \right\rangle $.
We will verify each property for each options.
  • $\left\langle A,B \right\rangle = \operatorname{trace}\left( A \right) \operatorname{trace}\left( B \right) $.

    This is NOT an inner product. Take \[ A = \begin{bmatrix} 1 & 0 \\ 0 & -1 \\ \end{bmatrix}. \] Then, \[ \left\langle A,A \right\rangle = \operatorname{trace}\left( A \right) \operatorname{trace}\left( A \right) = 0, \] but $A \neq 0$. Thus, this is not an inner product.

  • $\left\langle A,B \right\rangle = \operatorname{trace}\left( AB \right) $.
    This is TRUE. We recall that for any real symmetric matrix all the eigenvalues are real and the matrix is diagonalizable.
    • Let $A$ be any real symmetric matrix. Then, \begin{align*} \left\langle A, A \right\rangle & = \operatorname{trace}\left( A^2 \right) \\ & = \sum_{\lambda }\lambda ^2 > 0, \end{align*} where $\lambda$ is an eigenvalue of $A$.
    • If $\left\langle A,A \right\rangle = 0$, then $\operatorname{trace}\left( A^2 \right) = 0$, which means all the eigenvalues are zero. Since $A$ is diagonalizable and all the eigenvalues are zero, it is a zero matrix. Thus, $A = 0$.
    • For any real symmetric matrices $A,B$, \[ \left\langle A,B \right\rangle = \operatorname{trace}\left( AB \right) = \operatorname{trace}\left( BA \right) = \left\langle B,A \right\rangle . \]
    • Also, the linearity follows from the linearity property of $\operatorname{trace}$. That is, for any $\lambda \in \mathbb{R} $ and for any symmetric matrices $A,B,C$ \begin{align*} \left\langle \lambda A + B, C \right\rangle & = \operatorname{trace}\left( (\lambda A + B) C \right) \\ & = \operatorname{trace}\left( \lambda AC + BC \right) \\ & = \operatorname{trace}\left( \lambda AC \right) + \operatorname{trace}\left( BC \right) \\ & = \lambda \operatorname{trace}\left( AC \right) + \operatorname{trace}\left( (BC) \right) \\ & = \lambda \left\langle A,C \right\rangle + \left\langle B,C \right\rangle . \end{align*} Therefore, it is an inner product.
  • $\left\langle A, B \right\rangle = \det (AB)$.

    This is NOT an inner product. Take \[ A = \begin{bmatrix} 1 & 0 \\ 0 & 0 \\ \end{bmatrix}, \] then \[ \left\langle A, A \right\rangle = \det (A^2) = 0, \] but $A \neq 0$.

  • $\left\langle A,B \right\rangle = \operatorname{trace}\left( A \right) + \operatorname{trace}\left( B \right) $.

    This is NOT an inner product. Take \[ A = \begin{bmatrix} -1 & 0 \\ 0 & 0 \\ \end{bmatrix}. \] Then, \[ \left\langle A, A \right\rangle = \operatorname{trace}\left( A \right) + \operatorname{trace}\left( A \right) = - 2 < 0. \]