I am reading the following paper on mixed three-qubit states. It states that any three–qubit pure state can be written as (equation 1 of the paper)

$|\psi_{GHZ}\rangle=\lambda_0 |000\rangle+\lambda_1 e^{i\theta} |100\rangle+\lambda_2 |101\rangle+\lambda_3 |110\rangle+\lambda_4 |111\rangle$

and W class states have the form (equation 2 of the paper)

$|\psi_{W}\rangle=\lambda_0 |000\rangle+\lambda_1 |100\rangle+\lambda_2 |101\rangle+\lambda_3 |110\rangle$

However, later they use the following form for W states (page 2 of the paper)

$|\psi_{W}\rangle=\kappa_0 |000\rangle+\kappa_1\left(|100\rangle+ |010\rangle+ |001\rangle\right)$

I did not understand how equation 1 is obtained; also which one amongst equation 2 and the last equation is the "correct" representation of W-class states (or whether these two forms are equivalent to each other). Any help would be greatly appreciated.


1 Answer 1


The GHZ form of the state follows from a bit of a manipulation after define what you mean by a GHZ state, remembering that all these parametrisations are "up to local unitaries". For me, the GHZ class of states is one that takes the form $$ \alpha|u_1\rangle|u_2\rangle|u_3\rangle+\beta|v_1\rangle|v_2\rangle|v_3\rangle, $$ so the first challenge is to see how we can reduce a state of this form, under the action of local unitaries, to the claimed standard form. Pick a convention for the global phase such that the $|000\rangle$ term is real. Start by looking at the two basis states $|111\rangle$ and $|011\rangle$. There must exist a rotation on the first qubit that can completely get rid of the $|011\rangle$ term, leaving the $|111\rangle$ term real. Now, we can apply the same trick using a single qubit rotation on the second qubit to cancel the $|010\rangle$ term out of the pair $|010\rangle$ and $\|000\rangle$. Finally, use the same trick to cancel the $|001\rangle$ term out of $|001\rangle$ and $|000\rangle$. You'll find the only basis states we have left are those from your statement of $|\psi_\text{GHZ}\rangle$, and the only one we've not had the freedom to make real has been the $|100\rangle$ term.

As for the correct representation of the W class, it's the first one. If you read the bit on page 2 more carefully, you'll see that because they're trying to maximise a specific overlap, they know that the state they need to achieve the maximum has certain symmetries that they've built into the description. It's perhaps a bit misleading to give it the same notation as the general version, but what they say seems clear enough.


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.