The question should be pretty simple, but it turns out that there's more to it with respect to what I initially expected.

Starting from the definition of the gate $Y = \begin{bmatrix} 0 & -i \\ i & 0 \end{bmatrix}$, Wikipedia states that the eigenvectors are $$\lambda_{+1} = \frac{1}{\sqrt{2}}\begin{bmatrix}1\\i\end{bmatrix}=: |i\rangle, \lambda_{-1} = \frac{1}{\sqrt{2}}\begin{bmatrix}1\\-i\end{bmatrix} =: |-i\rangle$$

So, I should be able to derive the matrix Pauli-Y as

$$Y = (+1) |i\rangle \langle i| + (-1) |-i\rangle \langle -i| = \begin{bmatrix} 0 & i \\ i & 0 \end{bmatrix}$$

which is clearly different from the first matrix.

numpy, instead, gives as eigenvectors $$\lambda_{+1} = \frac{1}{\sqrt{2}}\begin{bmatrix}-i\\1\end{bmatrix}=: |i\rangle, \lambda_{-1} = \frac{1}{\sqrt{2}}\begin{bmatrix}1\\-i\end{bmatrix} =: |-i\rangle$$

which, by using the previous formula, returns a value for the Pauli Y gate equal to

$$Y = \begin{bmatrix} -1 & 0 \\ 0 & -1 \end{bmatrix}$$

which is even stranger.

So, what am I doing wrong? Is it possible to uniquely define the eigenvectors?

  • 1
    $\begingroup$ Can you write out what $\langle i|$ and $\langle -i|$ explicitly? Check it, you will know. $\endgroup$
    – narip
    Aug 10, 2022 at 13:39

2 Answers 2


Your first equation is wrong.

If $|i\rangle=\frac1{\sqrt2} (|0\rangle + i|1\rangle)$,

then its dual is the tranposed-conjugated version

$\langle i|=\frac1{\sqrt2} (\langle 0| - i\langle 1|)$,

which yields

$|i\rangle\langle i|=\frac12(|0\rangle\langle 0|-i|0\rangle\langle 1|+i|1\rangle\langle 0|+|1\rangle\langle 1|)$.

Similarly $|-i\rangle\langle -i|=\frac12(|0\rangle\langle 0|+i|0\rangle\langle 1|-i|1\rangle\langle 0|+|1\rangle\langle 1|)$.

So we get $|i\rangle\langle i|-|-i\rangle\langle -i|=-i|0\rangle\langle 1|+i|1\rangle\langle 0|=Y$ as you know it.

Note that I can modify one of the the eigenvectors as such $|i'\rangle=e^{i\alpha}|i\rangle$, with any real $\alpha$, and the relation still holds. So no, the eigenvector is not unique, it is defined up to a phase factor. numpy is providing valid eigenvectors, but be sure you are using the right complex functions to build $Y$.

  • $\begingroup$ Of course, you'and @Ohad are both right, shame on me for my stupid mistake. I'll leave my question here as a reminder :) $\endgroup$
    – tigerjack
    Aug 10, 2022 at 14:17
  • $\begingroup$ @tigerjack you can still accept the answer that seems more adequate to you ;-) $\endgroup$
    – Mauricio
    Aug 10, 2022 at 16:28

The eigenvectors that you have written above (both from Wikipedia and those plotted with Numpy) are valid eigenstates of the $pauli-Y$ operator.

I have done the same calculations that you have presented and the result in both cases is the $Y$ operator matrix.

I think that your mistake is in the computation of each outer product - it seems like that you forgot to complex-conjugate the entries of the bras $⟨i|$ and $⟨-i|$.

If $|i⟩ = \frac{1}{\sqrt2} \begin{bmatrix} 1 \\ i \end{bmatrix}$ then $⟨i| = \frac{1}{\sqrt2} \begin{bmatrix} 1 & -i \end{bmatrix}$.

Fix this for both $⟨i|$ and $⟨-i|$, perform the calculations again and you should get the right answers.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.