REVIEW
On the distribution of entanglement changes produced by unitary operations
Quantum Physics · Submitted 7 Mar 2006 · 36 references
Pith's verdict: Honest but modest Monte Carlo survey of two-qubit gate entanglement changes under the product measure; the main conclusion about entangled mixed states is not backed by the unconditional histograms.
T0 · AI review
Why Pith says this
Highlight any text to comment on it.
Referee report
The paper
The paper claims
For two-qubit states sampled from a natural product measure, the change in entanglement under CNOT peaks sharply at zero and is biased toward decreasing entanglement, especially for mixed states.
The reading
What carries the argument
The object that carries the argument is the entanglement of formation $E[\rho]$, evaluated through the concurrence $C(\rho)$ obtained from the square roots of the eigenvalues of $\rho\tilde\rho$, with $\tilde\rho=(\sigma_y\otimes\sigma_y)\rho^*(\sigma_y\otimes\sigma_y)$, via $E=h((1+\sqrt{1-C^2})/2)$ with binary entropy $h$. For each randomly generated two-qubit state $\rho$, the concurrence is computed before and after the unitary transformation $\hat U$, giving $\Delta E = E(\hat U\rho\hat U^\dagger)-E(\rho)$. The space of states is sampled with the product measure $\mu=\nu\times L_{N-1}$ for $N=4$, where $\nu$ is the uniform measure on complete sets of orthonormal projectors induced by the Haar measure and $L_{N-1}$ is Lebesgue measure on the simplex of eigenvalues. The gate family consists of the block-diagonal CNOT and $\hat U_\theta$, whose lower block is a rotation by $\theta$; applying these gates to many sampled states builds the histograms of $\Delta E$, its mean, dispersion, and conditional averages.
Extended reading notes
Core claim
The central numerical discovery is that, for two-qubit states distributed according to the product measure on state space, the probability distribution $P(\Delta E)$ of the change $\Delta E$ in entanglement of formation under the CNOT gate has a sharp peak at $\Delta E=0$; the same is true for the family $\hat U_\theta$ with rotation angle $\theta$ in the lower block. For the full space of pure and mixed states, the gate is more likely to reduce the entanglement of an entangled state than to increase it, and initially separable states have only a tiny probability of becoming entangled: the mean entanglement generated by CNOT from $E_0=0$ is $\langle E_F\rangle=0.0052$. For pure states the behavior differs qualitatively: CNOT raises the mean final entanglement to about $0.5$ for initial entanglements in $[0,0.5]$, and only reduces it for states with large initial entanglement. Because the sampling measure is invariant under unitary transformations, the marginal distribution of final entanglement equals that of initial entanglement, so the entire gate effect is contained in the joint distribution that $P(\Delta E)$ describes.
Load-bearing premise
All reported probabilities are averages over the specific product measure $\mu=\nu\times L_{N-1}$, and if that measure does not represent the physically relevant ensemble of two-qubit states, the numerical conclusions would change.
Editorial extensions
If this is right
- For mixed two-qubit states, generic unitary gates such as CNOT act typically as disentanglers: entangled mixed states are more likely to see their entanglement decrease than increase.
- Separable mixed states are almost untouched by the gates in terms of entanglement: CNOT creates a mean final entanglement of only $0.0052$ from initially separable states, so entangling mixed states requires more than a single standard gate.
- For pure states the same gates are effective entanglers at low initial entanglement: CNOT pushes the mean final entanglement to about $0.5$ for $E_0\in[0,0.5]$, while for large $E_0$ it tends to reduce it.
- Because the measure is unitarily invariant, the gate's effect cannot be seen in the marginal distribution of final entanglement; it appears only in the joint distribution, so $P(\Delta E)$ is the right and sufficient object for comparing a gate against a random replacement of the state.
- The conditional data $\langle E_0\rangle$ versus $\Delta E$ show that large entanglement increases come from initially highly entangled states, and for mixed states the dispersion in initial entanglement is smallest exactly at $\Delta E=0$, confirming that zero-change events are dominated by near-separable states.
Reading between the lines
- A testable extension is to compute $P(\Delta E)$ for the q-entropy family mentioned in the paper; if the sharp peak at zero is robust across entropic measures, the 'typical disentangling' conclusion is a property of the state-space geometry rather than of the particular measure of entanglement used.
- The unitarity-invariance argument suggests that the shape of $P(\Delta E)$ is fully determined by the correlation between a state and its image under the gate; one could quantify this by the mutual information between initial and final entanglement, a quantity the paper does not compute.
- If the physically relevant ensemble of two-qubit states is biased toward separable or low-entanglement states, as is common in noisy preparations, the paper's numbers imply that standard CNOT operations will rarely lift such states into the entangled region; protocols that need entanglement from mixed sources would require either gate sequences, measurement postselection, or non-unitary operations
- Since the distribution changes with the parameter $\theta$ and approaches identity as $\theta\to0$, the $\hat U_\theta$ family offers a controlled way to interpolate between no entanglement change and the full CNOT effect; this could serve as a benchmark for the entangling power of a gate on mixed states under the same measure.
Original abstract
We consider the change of entanglement of formation $\Delta E$ produced by a unitary transformation acting on a general (pure or mixed) state $\rho$ describing a system of two qubits. We study numerically the probabilities of obtaining different values of $\Delta E$, assuming that the initial state is randomly distributed in the space of all states according to the product measure introduced by Zyczkowski {\it et al.} [Phys. Rev. A {\bf 58} (1998) 883].
The compiled record
The paper's own sentences, arranged, kept beside the review and separate from it. Each label is drawn from a closed vocabulary; the number beside it is the probability the classifier gave that label. Numbers in the tables are the authors', checked by code. No sentence in this record was written by a model.
The abstract, sentence by sentence
We consider the change of entanglement of formation $\Delta E$ produced by a unitary transformation acting on a general (pure or mixed) state $\rho$ describing a system of two qubits. We study numerically the probabilities of obtaining different values of $\Delta E$, assuming that the initial state is randomly distributed in the space of all states according to the product measure introduced by Zyczkowski {\it et al.} [Phys. Rev. A {\bf 58} (1998) 883].
background contribution method result interpretation main claim
| # | sentence | role | the authors' own work | states the main conclusion | strength |
|---|---|---|---|---|---|
| s0 | We consider the change of entanglement of formation $\Delta E$ produced by a unitary transformation acting on a general (pure or mixed) state $\rho$ describing a system of two qubits. | contribution 0.65 | 0.65 | 0.04 | asserted 0.58 |
| s1 | We study numerically the probabilities of obtaining different values of $\Delta E$, assuming that the initial state is randomly distributed in the space of all states according to the product measure introduced by Zyczkowski {\it et al.} [Phys. Rev. A {\bf 58} (1998) 883]. | method 0.68 | 0.93 | 0.08 | asserted 0.92 |
What was examined, and what was not
- arXiv listing abstract v1
- full text
- figures and tables
- references and citing sentences
- relations to other papers
- P abstract
- S sections
- T body sentences
- C citations
- R cross-paper relations
Figures
Showing 1 of the paper's 5 figures. See all figures in the PDF ↗ Figures reproduced from arXiv:quant-ph/0603057 by the authors.
References
Discussion
Disagree with the review, or know something it missed? Leave a comment, highlight any passage above to comment on it, or challenge this review formally.