Consider the operator $S^+ = \sum_{i=1}^L S^+_i$ acting on a spin-chain of spin-1/2 particles. Denote the half-chain Von Neumann entanglement entropy of a state $|\psi\rangle$ by $\mathbb{S}[|\psi\rangle]$. (For simplicity in notation in the following, take $\mathbb{S}[0] = 0$.)

Consider the following example. Define the state $|\Omega\rangle$ as having all spins down: $|\Omega\rangle = \otimes_{i=1}^L|\downarrow\rangle$. Then $\mathbb{S}[|\Omega\rangle] =0$ because $|\Omega\rangle $ is a product state, while $\mathbb{S}[S^+|\Omega\rangle] = \ln(2)$, as can be seen by a quick Schmidt decomposition by hand. Notice in particular that

$$\mathbb{S}[S^+|\Omega\rangle] - \mathbb{S}[|\Omega\rangle] = \ln(2)$$

After playing around a little with numerics, I have the following conjecture: $$\max_{|\psi\rangle \in \mathscr{H}} \left( \mathbb{S}[S^+|\psi\rangle] - \mathbb{S}[|\psi\rangle] \right)= \ln(2)$$

That is, $S^+$ can only increase the entropy of a state by $\ln(2)$ and no more. Similarly, I conjecture that $$\max_{|\psi\rangle \in \mathscr{H}} \left( \mathbb{S}[(S^+)^n|\psi\rangle] - \mathbb{S}[|\psi\rangle] \right)=\mathbb{S}[(S^+)^n|\Omega\rangle] - \mathbb{S}[|\Omega\rangle]$$

Are these conjectures correct? How can I prove these conjectures?

A small piece of supporting evidence (not near a proof) for the first conjecture is that it is easy to check that all product states $|p\rangle$ in the $S^z$-basis obey $\mathbb{S}[S^+|p\rangle] \leq \ln(2)$, as the resulting state's Schmidt decomposition has at most two states. Another small piece of supporting evidence, when I feed in random states for $|\psi\rangle$, the entanglement entropy decreases relative to the entropy of the random state. However, this is just supporting evidence that is far from a statement about all possible states in the Hilbert space.

  • $\begingroup$ Only saw this now: The point is that the operator S+ has Schmidt rank 2 (as an operator), and thus, it can increase the Schmidt rank of any state by at most a factor of 2. However, what happens to the Schmidt coefficients can be rather different, as you correctly note in your answer. $\endgroup$ May 2 at 22:03
  • $\begingroup$ @NorbertSchuch Thanks, that adds clarity to what's happening. $\endgroup$
    – user196574
    May 2 at 22:34

My conjectures in my question were incorrect. Since $S^+$ can destroy certain states, it's possible to begin with a low entanglement entropy state that gains a high entropy after being acted upon by $S^+$.

For example, consider (for some large $L$) a state $$|\psi\rangle = \sqrt{.999999999} \otimes_{i=1}^L|\uparrow\rangle_i + \sqrt{.000000001} (\text{scrambled, normalized superposition of a massive number of states}).$$ The entropy of this initial state arising from the second term is suppressed by the tiny coefficient. However, the action of $S^+$ will destroy the first term, removing the suppression of the second term after normalizing. The second term after being acted upon by $S^+$ will still have some large amount of entanglement, perhaps smaller than the second term had before, but still a much larger amount than the full initial state. Thus, the entanglement of the final state will be much larger than the initial state, easily exceeding $\ln(2)$.

I will see if I can formulate a similar question in the context of unitary operators, which cannot annihilate states.


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.