My questions:

  1. What does discarding the second register mean for the standard approach of hidden subgroup algorithm?
  2. Why does discarding let the first register end up in a mixed state?

My understanding so far:

I am trying to understand the effect of discarding the second register in the standard approach of hidden subgroup algorithm using this paper as resource. I will be using $S_5$ as my group of interest.

It is said in the paper that,

enter image description here

I assume $f$ maps $g$ to it's irrep. So, if we take the case of $S_5$, assuming an automorphism group of a $5$-node graph is hidden inside $S_5$, $|G| = |S_5| = 120$. Both $|g\rangle$ and $|f(g)\rangle$ are $120$ element column vectors.

So, when it is said that the second register is discarded what it actually means? Let's take the case of trapped ion qubits. When we say we discard the second register, does it mean that we will not do any unitary transformation or measurement on them anymore but they will always be there? In the Wikipedia page of Simon's problem the second register is measured.

I also don't understand why discarding (whatever it means) will let the first register end up in a mixed state. Shouldn't it be in a pure quantum state?


1 Answer 1


Discarding in this context can be carried out by performing a Partial Trace. The system you're interested in is that of the first register. If you want to know only its state, as opposed to the combined state of the two registers, then you take the density matrix of the whole system and trace out the second register. The resultant state will, in this case, be mixed. What you're seeing at work is more or less purification in reverse. The resultant mixed state is the reduced state of some other pure system, namely the system comprised of the two registers.

  • $\begingroup$ I have started with a pure state $$\frac{1}{\sqrt{|G|}} \sum_{g \in G} |g, f(g)\rangle$$. Why or how would I take a partial trace on it? $\endgroup$ Dec 19, 2015 at 14:30
  • 1
    $\begingroup$ Partial traces are covered in all the standard texts on Quantum computation and information. I'm not even going to attempt to explain them better than they are there. $\endgroup$ Dec 19, 2015 at 16:00
  • $\begingroup$ that's correct. Moreover it is also true that we can treat a pure state as a mixed state with probability 1. Are you suggesting that I would consider $\frac{1}{\sqrt{|G|}} \sum_{g \in G} |g, f(g)\rangle$ as a mixed state i.e. $\frac{1}{|G|} \sum_{g \in G} |g, f(g)\rangle \langle, f(g)I $ and take the partial trace on the second register? $\endgroup$ Dec 19, 2015 at 16:07
  • 1
    $\begingroup$ yes. that's the idea. $\endgroup$ Dec 19, 2015 at 16:28
  • $\begingroup$ is that the only way? What could be other alternatives? $\endgroup$ Dec 19, 2015 at 16:47

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.