I have a conjecture about quantum channels. On which examples should I test it before I try to prove it, ask it on StackExchange, or write a paper about it?

(Note: This is meant to be a reference question. But whenever I have a conjecture, I do test it on the channels listed, and it has saved me a lot of time trying to prove wrong statements.)

  • $\begingroup$ Is the conjecture to be found somewhere now? $\endgroup$ Oct 1 '20 at 11:08
  • $\begingroup$ @DescheleSchilder What do you mean? $\endgroup$ Oct 1 '20 at 11:35
  • $\begingroup$ I mean the conjecture you concocted. Has it been published? $\endgroup$ Oct 1 '20 at 12:05
  • $\begingroup$ @DescheleSchilder Have you read the post fully, including the part in the brackets? $\endgroup$ Oct 1 '20 at 15:35
  • 1
    $\begingroup$ @DescheleSchilder Precisely. There were a few questions here which made me realize that such a catalogue would be useful (since I could shoot down the conjectures with one of those examples). In fact, Emilio Pisanty lured me into writing the reference question by promising a bounty which he never awarded ;) As you can see from the linked question, I have also used those examples since. $\endgroup$ Oct 1 '20 at 17:40

The following is a list of channels you can use to test your conjecture. (Some are special cases of subsequent ones -- it makes more sense to first test the special cases.) Here, $d$ is the dimension of the space.

  1. The identity channel: $$ \mathcal E(\rho)=\rho\ . $$

  2. The fully depolarizing channel: $$ \mathcal E(\rho) = \mathrm{tr}(\rho)\,\tfrac1d\mathbb I\ . $$

  3. The depolarizing channel: $$ \mathcal E(\rho) = \gamma\rho + (1-\gamma)\mathrm{tr}(\rho)\,\tfrac1d\mathbb I $$ for $-\tfrac1d\le\gamma\le1$.

  4. The dephasing channel $$ \mathcal E(\rho) = \gamma\rho + (1-\gamma)Z\rho Z $$ for qubits (with $Z$ the Pauli $Z$ matrix), and possibly some suitable generalizations for $d>2$. If your conjecture is not rotationally symmetric, test rotated versions as well.

  5. The "Wirf weg und mach neu™" ("throw away and make new") channel: $$ \mathcal E(\rho) = \mathrm{tr}(\rho)\,\sigma $$ with $\sigma$ a density matrix. Obviously a generalization of 2, but test e.g. pure states $\sigma$.

  6. The Holevo-Werner channel: $$ \mathcal E(\rho) = \tfrac{1}{d-1}(\mathrm{tr}(\rho)\,\mathbb I -\rho^T)\ , $$ where $\rho^T$ is the transposition.

  7. Entanglement breaking channels: $$ \mathcal E(\rho) = \sum_i \mathrm{tr}(\rho F_i) \sigma_i $$ with $\sigma_i$ density matrices and the $F_i$ a POVM (i.e., $F_i\ge0$ and $\sum F_i=\mathbb I$). Obviously a generalization of 5: These are all channels which can be realized by first measuring the input and then preparing a new output conditioned on the measurement outcome. (You might want to test specific instances of these, like $F_i$ projectors onto subspaces, and $\sigma_i$ supported in the same subspace, etc.)

If your conjecture has passed all examples: Congratulations! It is probably true, and you can start proving it!

  • $\begingroup$ Nice list. Is it useful to consider only depolarization along a specific subspace? $\endgroup$ Nov 9 '16 at 19:12
  • $\begingroup$ @EmilioPisanty Probably. Guess I usually encounter qubit channels ... To be honest, in my experience 1, 2, 5 with pure $\sigma$, and 6 are enough to shoot down almost all wrong conjectures. $\endgroup$ Nov 9 '16 at 19:35

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.