## The Spectrum of the Partial Transpose of a Density Matrix

It is a simple fact that, given any density matrix (i.e., quantum state) , the eigenvalues of are the same as the eigenvalues of (the transpose of ). However, strange things can happen if we instead only apply the transpose to one half of a quantum state. That is, if then its eigenvalues in general will be very different from the eigenvalues of , where is the identity map on and is the transpose map on (the map is called the *partial transpose*).

In fact, even though is positive semidefinite (since it is a density matrix), the matrix in general can have negative eigenvalues. To see this, define and let , where

is the standard maximally-entangled pure state. It then follows that

,

which has eigenvalues equal to , eigenvalues equal to , and eigenvalues equal to .

The fact that can have negative eigenvalues is another way of saying that the transpose map is positive but not completely positive, and thus plays a big role in entanglement theory. In this post we consider the question of how exactly the partial transpose map can transform the eigenvalues of :

**Question.** For which ordered lists does there exist a density matrix such that has eigenvalues ?

### The Answer for Pure States

In the case when is a pure state (i.e., has rank 1), we can completely characterize the eigenvalues of by making use of the Schmidt decomposition. In particular, we have the following:

**Theorem 1.** Let have Schmidt rank and Schmidt coefficients . Then the spectrum of is

,

together with the eigenvalue with multiplicity .

*Proof.* If has Schmidt decomposition

then

It is then straightforward to verify, for all , that:

- is an eigenvector with eigenvalue ;
- is an eigenvector with eigenvalue ; and
- , from which it follows that the remaining eigenvalues are .

Despite such a simple characterization in the case of rank-1 density matrices, there is no known characterization for general density matrices, since eigenvalues aren’t well-behaved under convex combinations.

### The Number of Negative Eigenvalues

Instead of asking for a complete characterization of the possible spectra of , for now we focus on the simpler question that asks how many of the eigenvalues of can be negative. Theorem 1 answers this question when is a pure state: the number of negative eigenvalues is , where r is the Schmidt rank of . Since , it follows that has at most negative eigenvalues when is a pure state.

It was conjectured in [1] that a similar fact holds for general (not necessarily pure) density matrices as well. In particular, they conjectured that if then has at most negative eigenvalues. However, this conjecture is easily shown to be false just by randomly-generating many density matrices and then counting the number of negative eigenvalues of ; density matrices whose partial transposes have more than negative eigenvalues are very common.

In [2,3], it was shown that if then can not have more than negative eigenvalues. In [4], this bound was shown to be tight when by explicitly constructing density matrices such that has negative eigenvalues. Similarly, this bound was shown to be tight via explicit construction when in [3]. Finally, it was shown in [5] that this bound is tight in general. That is, we have the following result:

**Theorem 2.** The maximum number of negative eigenvalues that can have when is .

It is worth pointing out that the method used in [5] to prove that this bound is tight is not completely analytic. Instead, a numerical method was presented that is proved to always generate a density matrix such that has negative eigenvalues. Code that implements this numerical procedure in MATLAB is available here, but no general analytic form for such density matrices is known.

### Other Bounds on the Spectrum

Unfortunately, not a whole lot more is known about the spectrum of . Here are some miscellaneous other results that impose certain restrictions on its maximal and minimal eigenvalues (which we denote by and , respectively):

**Theorem 3 [3].** .

**Theorem 4 [2].** .

**Theorem 5 [6].** If has negative eigenvalues then

and

.

However, these bounds in general are fairly weak and the question of what the possible spectra of are is still far beyond our grasp.

**Update [August 21, 2017]:** Everett Patterson and I have now written a paper about this topic.

**References**

- R. Xi-Jun, H. Yong-Jian, W. Yu-Chun, and G. Guang-Can. Partial transposition on bipartite system.
*Chinese Phys. Lett.*, 25:35, 2008. - N. Johnston and D. W. Kribs. A family of norms with applications in quantum information theory.
*J. Math. Phys.*, 51:082202, 2010. E-print: arXiv:0909.3907 [quant-ph] - S. Rana. Negative eigenvalues of partial transposition of arbitrary bipartite states.
*Phys. Rev. A*, 87:054301, 2013. E-print: arXiv:1304.6775 [quant-ph] - L. Chen, D. Z. Djokovic. Qubit-qudit states with positive partial transpose.
*Phys. Rev. A*, 86:062332, 2012. E-print: arXiv:1210.0111 [quant-ph] - N. Johnston. Non-positive-partial-transpose subspaces can be as large as any entangled subspace.
*Phys. Rev. A*, 87:064302, 2013. E-print: arXiv:1305.0257 [quant-ph] - N. Johnston.
*Norms and Cones in the Theory of Quantum Entanglement*. PhD thesis, University of Guelph, 2012.

Is the Theorem 3 also fulfilled for n – infinite (so is the partial transpose of density operator in this case bounded)?

Nice post, suppose we have the density state which is finite-dimensional, suppose we take partial transpose can take absolute of it. If we take partial transpose again which one should think the number of negative eigenvalues should be less than the (m-1)(n-1). This is called binegativity. Is there any comment?

@EnjUI KUO – You’re right that the number of negative eigenvalues in that case would also be <= (m-1)(n-1). However, I do not know if that bound is tight. My naive first guess is that it should be tight, and we "just" need to find a construction of density matrices rho that attains that bound... but I don't know how to do that.

@Nathaniel, I just found that people already proved the bound is tight in https://arxiv.org/abs/1305.0257 ( I found you also have a post about it)

But now, I am interested how to connect the eigenvalue of density matrix to the eigenvalue of partial transposition. For example, suppose I know my rho has a eigenvalue>=1/2, can I get something bound about the eigenvalue of its partial transposition? It seems like all the inequality you mentioned above only related the eigenvalue of its partial transposition.

Thanks for this great post! It was a great overview that helped us find the relevant results. 🙂 And very nice results, actually!