Given a system whose density operator is $\rho$, the (Von Neumann) entanglement entropy is defined as $$S[\rho]=-\text{Tr}(\rho\log\rho)$$ My question: Is $\log\rho$ simply the matrix whose entries are the logarithm of the entries of $\rho$? Or does it denote something more exotic (e.g. related with the matrix exponential)?

My thoughts on it: I know that $S$ must coincide with Shannon's Entropy $-\underset{j}{\sum}p_{j}\log p_{j}$ (if one does the calculations on an eigenbasis $\{|n\rangle\}$ of $\rho$ with eigenvalues $p_{j}$, which are the probabilities of the system being in the state $|\psi_{j}\rangle$). In fact, if one assumes that $\log\rho$ is the matrix whose entries are the logarithm of the entries of $\rho$, this happens (the calculations are not difficult). This indicates (but of course does not prove) that the $\text{log}\rho$ has this simple interpretation. Does it?

  • $\begingroup$ $\log\rho\equiv A\Leftrightarrow \rho=\mathrm e^A$. In other words: yes, the definition is something more exotic, and it is indeed related to the matrix exponential. You can only evaluate the logarithm entry-wise if the operator is diagonal. In any case, this seems to be a math.SE question ("how is the logarithm of an operator defined?") $\endgroup$ – AccidentalFourierTransform Jan 21 '17 at 20:24
  • $\begingroup$ I really don't understand the downvotes and I would like to ask whomever downvoted to reconsider their decision: the answer might be off-topic, but it is not a bad question at all. $\endgroup$ – AccidentalFourierTransform Jan 21 '17 at 21:44
  • $\begingroup$ @AccidentalFourierTransform "this seems to be a math.SE question"- not really, since the question was merely about notation: I was asking if it was the logarithm related with the matrix exponential. Anyway, thanks for the answer. $\endgroup$ – Soap Jan 21 '17 at 21:45
  • $\begingroup$ @AccidentalFourierTransform That is, it was a question about notation that physicists use when talking about entanglement entropy. $\endgroup$ – Soap Jan 21 '17 at 21:51
  • $\begingroup$ The answer can be instantly found on the Wikipedia entry for "von Neumann entropy": ln denotes the (natural) matrix logarithm. $\endgroup$ – Norbert Schuch Jan 22 '17 at 2:14

Firstly, we can define the exponential of an $n\times n$ matrix $A$ through the Taylor series, namely,

$$e^A = \sum_{k = 1 }^\infty \frac{A^k}{k!}.$$

If the matrix $A$ is diagonalizable then it can be expressed as $A = M \Lambda M^{-1}$ where $M$ is the matrix of column eigenvectors and $\Lambda = \mathrm{diag}(\lambda_1, \lambda_2, \dots, \lambda_n)$ where $\lambda_n$ are the eigenvalues. It then follows,

$$e^A = \sum_{k= 1}^\infty \frac{1}{k!}M\Lambda^k M^{-1} = M \left( \sum_{k=1}^\infty \frac{\Lambda^k}{k!}\right) M^{-1} = Me^{\Lambda} M^{-1}.$$

In the special case that $A$ is a diagonal matrix, you can show that the definition of the matrix $\exp$ reduces to the exponential of each diagonal element. Therefore, if $B=\ln A$, then we define the logarithm $B$ as being the matrix such that $e^B = A$, in the matrix exponential sense above.

  • $\begingroup$ The question was merely: what does $\text{log}$ denote in this case. However, I appreciate the fact that you elaborated on it. $\endgroup$ – Soap Jan 21 '17 at 21:50
  • $\begingroup$ @XicoSim If you didn't want an answer which elaborates, why didn't you just check the Wikipedia entry? $\endgroup$ – Norbert Schuch Jan 22 '17 at 2:17
  • $\begingroup$ @NorbertSchuch Of course I did! I did not find it though: en.wikipedia.org/wiki/Entropy_of_entanglement $\endgroup$ – Soap Jan 22 '17 at 11:54
  • 1
    $\begingroup$ @XicoSim en.wikipedia.org/wiki/Von_Neumann_entropy. I'm not sure why you keep calling it entanglement entropy. $\endgroup$ – Norbert Schuch Jan 22 '17 at 17:38

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.