2
$\begingroup$

if $HH = I$ and $XX =I$, then is $H=X$?

$HH = I = XX$ or, $HH = XX$ then, taking under root, is $H = X$?

This is absurd but how to disprove it?

$\endgroup$
1
  • 5
    $\begingroup$ The argument is obviously flawed since it fails even in the good ol' integers: $1\cdot 1=1$ and $(-1)\cdot(-1)=1$ but $1\ne-1$. One way to disprove it for the Hadamard and Pauli $X$ gate is to carry out the calculations. $\endgroup$ Nov 26 '21 at 8:20
5
$\begingroup$

Obviously $H\neq X$, as they are simply different matrices/operators.

The reason for the apparent "contradiction" is that the square root of a matrix (or of any number, for that matter) is not unique. What you are seeing here is that both $H$ and $X$ are valid square roots of the identity matrix $I$.

In fact, you can derive infinitely many possible square roots of $I$. Start observing that you can write $I=\mathbb P(u)+\mathbb P(v)$ for any pair of orthonormal vectors, $\langle u,v\rangle=0$ and $\langle u,u\rangle=\langle v,v\rangle=1$, where $\mathbb P(u)\equiv uu^\dagger$ denotes the projector onto $u$. Taking the square root via the eigendecomposition in the usual way then gives you $$\sqrt{I} = s \mathbb P(u) + t\mathbb P(v),$$ where $s,t\in\mathbb C$ are square roots of $1$, and thus $s,t\in\{+1,-1\}$. You thus get four possible solutions for every choice of two orthonormal vectors.

You get $H$ and $X$ choosing $s=-t=1$ and $u=(1,1)^T/\sqrt2$ and $u=(1,0)$, respectively.

It is interesting to note that much of the freedom that allowed the above solutions is due to the degeneracy of the matrix. A non-degenerate matrix will instead have only four solutions, as there is no freedom in the choice of the eigenvectors. For more information see e.g. the relevant Wikipedia page.

$\endgroup$

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.