4
$\begingroup$

Given a list of $m$ quantum states $$|\phi_0\rangle, |\phi_1\rangle, ... |\phi_{m-1}\rangle$$

each on $n$ qubits, with unitaries to prepare these ($U_0, U_1, ...$), I'd like to prepare a superposition of these states weighted by their overlap with another quantum state ${|\psi\rangle}$, such that the resultant state is:

$$ \frac{1}{\sum_i |\langle \psi | \phi_i \rangle|^2} \sum_i |\langle \psi | \phi_i \rangle| | \phi_i \rangle $$

Is there a simple method to prepare such a state where the postselection cost is not exponential in $n$ or $m$?

I've looked at arXiv:1401.2142, but this requires amplitude estimation which I'd like to avoid.

$\endgroup$
1
  • $\begingroup$ How are the input states given to you? Are they a list of quantum states in qubits, or are they classical descriptions of the quantum states, or just the descriptions of the unitaries? $\endgroup$
    – Sam Jaques
    Aug 14 at 17:39

1 Answer 1

0
$\begingroup$

Hm. If you removed the absolute value from the coefficients, you'd have $\sum_i \langle \phi_i|\psi\rangle|\phi_i\rangle$, which equals $|\psi\rangle$ if it is in the span of the $|\phi_i\rangle$ or else the projection of $|\psi\rangle$ onto that subspace. So maybe you want an algorithm that changes the phases of the coefficients. Now that is easy to do, if the states $|\phi_i\rangle$ are orthonormal: apply $$\exp(-i\arg \langle \phi_0|\psi\rangle)\exp(-i\arg \langle \phi_1|\psi\rangle)\cdots \exp(-i\arg \langle \phi_{m-1}|\psi\rangle)$$ to your state and you're done.

If all of these conditions are met (i.e. the states are orthonormal and $|\psi\rangle$ lies within their span) then all you need is to know/measure/learn the phase arguments of the overlaps $ \langle \phi_0|\psi\rangle$ and then they can all be fixed with unitaries.

$\endgroup$

Your Answer

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

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