A speaker from a talk/lecture I attended recently quoted the following inequality without proof, and for no good reason I have taken it upon myself to prove it:

Let $A\in SL(2,\mathbb{C})$, so $\det\left(A\right) = 1$. Also, let $\not x \equiv x^\mu \sigma_\mu$ where $x^\mu=(x^0,x^1,x^2,x^3)$ is an arbitrary real-valued four-vector (tuple of 4 real numbers) and $\sigma_\mu=(\sigma_0,\sigma_1,\sigma_2,\sigma_3)$ are the usual Pauli matrices with:

$$\begin{align} \sigma_0&=-\begin{pmatrix}1 & 0 \\ 0 & 1 \end{pmatrix}\\ &\\ \sigma_1=\sigma_x\hspace{0.7cm}\,\,\,\sigma_2&=\sigma_y\hspace{1cm}\,\,\,\sigma_3=\sigma_z \end{align} $$

I would like to prove the following harmless looking inequality:

$$\text{Tr} [A\not x A ^{\dagger}\not x]\geq0$$

My progress:

I know that the trace is definitely real, and so the inequality is well-defined:

$$\begin{align} \text{Tr}[A\not x A ^{\dagger}\not x]^*&=\text{Tr}[(A\not x A ^{\dagger} \not x )^\dagger]\\ &=\text{Tr}[(\not x )^{\dagger}(A^{\dagger})^{\dagger}(\not x )^{\dagger}(A)^{\dagger}]\\ &=\text{Tr}[\not x A \not x A ^{\dagger} ]\\ &=\text{Tr}[A\not x A ^{\dagger} \not x] \end{align}$$

Honestly, apart from this I don't know what I could do. Hints and/or suggestions would be much appreciated.


Let $x=\sigma_3$, $A=\mathrm{i}\sigma_2$. Then $\det A = 1$, so $A\in\mathrm{SL}_2\mathbb{C}$. Nevertheless $$\begin{split}AxA^{\dagger}x&=(\mathrm{i}\sigma_2)\sigma_3(\mathrm{i}\sigma_2)^{\dagger}\sigma_3\\ &=\sigma_2\sigma_3\sigma_2\sigma_3\\ &=-\sigma_2^2\sigma_3^2\\ &=\sigma_0\end{split}$$ whence $$\mathrm{Tr}\,AxA^{\dagger}x=-2\text{.}$$

  • $\begingroup$ Thanks, clearly the inequality is false. I'm thinking now, should the inequality sign be reversed, or is any such inequality invalid? $\endgroup$ – Arturo don Juan Oct 19 '18 at 20:24
  • 1
    $\begingroup$ It's true when $x$ is a positive-semidefinite (psd) matrix, which is the case whenever $x^{\mu}$ is a spacelike or lightlike vector. For then $x$ has a psd square root $y$ and the trace is of the form $\mathrm{Tr}\,B B^{\dagger}$ where $B= y A y$. $\endgroup$ – K B Dave Oct 19 '18 at 21:25
  • 1
    $\begingroup$ "..which is the case whenever $x^\mu$ is a spacelike or lightlike vector." how did you arrive at this? And what convention are you using when you say "spacelike"? I think the 4-vector $x^\mu=(0,0,0,1)$ is definitely spacelike, but you have shown that the trace is negative with that choice. $\endgroup$ – Arturo don Juan Oct 19 '18 at 23:29
  • $\begingroup$ Sorry, I was hasty and mixed up spacelike and timelike :/. By direct diagonalization, the eigenvalues of $x$ are $-x^0 \pm \sqrt{(x^1)^2+(x^2)^2+(x^3)^2}$. For $x$ to be psd, we need the eigenvalues to both be nonnegative, whence $x^0\leq 0$ and $\lvert x^0\rvert\geq \sqrt{(x^1)^2+(x^2)^2+(x^3)^2}$—a timelike vector. $\endgroup$ – K B Dave Oct 19 '18 at 23:45

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.