I know for a separable state $\rho_{AB}$ we have

$$P_{AB}= P_{A}=P_{B}=1$$

Where P is purity. What if the relation between the purities when the bipartite state is entangled? Is there any inequality relation?

  • $\begingroup$ I don't understand how you came up with the first equation. If you have a pure product state, say $|\psi\rangle_A |\phi\rangle_B$, Then $$P_{AB} = P_A = P_B= 1$$ $\endgroup$
    – FDGod
    Commented Mar 5 at 20:26
  • $\begingroup$ Thanks you i fixed that. $\endgroup$
    – reza
    Commented Mar 5 at 20:52

2 Answers 2


This question seems to presume a pure bipartite state. For any such state, there exists a Schmidt decomposition that effectively finds a local basis in which to expand each of the subsystems such that $$|\Psi\rangle=\sum_{i=1}^n \lambda_i |\psi_i\rangle\otimes |\phi_i\rangle$$ for some integer $n$, some coefficients satisfying $\sum_{i=1}^n |\lambda_i|^2=1$, some set of orthonormal states $\{|\psi_i\rangle\}$ for system A, and another set of orthonormal states $\{|\phi_i\rangle\}$ for system B. Any bipartite pure state can be expressed in this form; it is separable if and only if $n=1$, equivalent to a single coefficient $|\lambda_1|=1$ and the rest of the coefficients vanishing.

Because any state can be written like this, we know that the reduced state for system A is $$\rho_A=\sum_{i=1}^n|\lambda_i|^2 |\psi_i\rangle\langle\psi_i|$$ with purity $$P_A=\mathrm{Tr}(\rho_A^2)=\sum_{i=1}^n|\lambda_i|^4$$ and that the reduced state for system B is $$\rho_B=\sum_{i=1}^n|\lambda_i|^2 |\phi_i\rangle\langle\phi_i|$$ with purity $$P_B=\mathrm{Tr}(\rho_B^2)=\sum_{i=1}^n|\lambda_i|^4=P_A.$$

This directly tells us all equalities and inequalities that we would like. First, if the bipartite state is pure, then $P_A=P_B$. Second, if it is separable, then $|\lambda_1|=1$, so $P_A=P_B=1$, and if it is not separable, then there is more than one nonzero coefficient $\lambda_i$, and we always have that $$\sum_i |\lambda_i|^4\leq \sum_i |\lambda_i|^2=1,$$ so the reduced states are always mixed. The more coefficients $\lambda_i$ there are, the more mixed the reduced state will be, in general (maximally mixed comes from maximal entanglement which comes from $n$ as large as possible and equal-magnitude $|\lambda_i|=\mathrm{constant}$.

What happens if the initial state is not pure; then what can we say? Take a generic state expressed in some orthonormal basis: $$\rho=\sum_{ij;kl}\rho_{ij;kl}|i\rangle\langle j|\otimes|k\rangle\langle l|$$ subject to positivity and normalization ($\sum_{ik}\rho_{ii;kk}=1$) constraints. The purity is $$P_{AB}=\mathrm{Tr}(\rho^2)=\sum_{ij;kl}\rho_{ij;kl}\rho_{ji;lk}=\sum_{ij;kl}|\rho_{ij;kl}|^2.$$ Then, tracing out one system yields something like $$\rho_A=\sum_{ij}(\sum_k \rho_{ij;kk})|i\rangle\langle j|$$ with purity $$P_A=\sum_{ij}|\sum_k \rho_{ij;kk}|^2;$$ similarly, $$P_B=\sum_{kl}|\sum_i \rho_{ii;kl}|^2.$$ The first thing to note is that, in general $P_A\neq P_B$ for mixed states. Next, we can note some inequalities with complex numbers like $$|\sum_k \rho_{ij;kk}|^2\leq \sum_k |\rho_{ij;kk}|^2 \leq \sum_{kl} |\rho_{ij;kl}|^2,$$ which directly show that $$P_A\leq \sum_{ij}\sum_{kl} |\rho_{ij;kl}|^2=P_{AB}$$ and the same inequality can be done to show that $P_B\leq P_{AB}$.

  • $\begingroup$ Thank you for answer. I was looking for a relation between purities for a state ( mixed or pure), i guess $P_{AB}<=P_A.P_B$ holds. But im not sure about it because could not prove it. $\endgroup$
    – reza
    Commented Mar 6 at 7:51
  • $\begingroup$ Seems for pure states, where you described in your answer, the relation i mentioned in the above comment, holds. $\endgroup$
    – reza
    Commented Mar 6 at 7:58
  • 1
    $\begingroup$ @reza I updated with some generic results for mixed states to show that $P_{AB}\geq P_A$ and $P_{AB}\geq P_B$ (so, if this is what you meant, $P_{AB}\geq P_A\times P_B$. $\endgroup$ Commented Mar 6 at 22:44
  • 1
    $\begingroup$ @Quantum Mechanic Thanks, trying to summarize things in a final conclusion always has this danger of oversimplifying! I glanced through it too quickly, of course bipartite doesn't imply qubits (and the $1/2$ should be $1/n$). $\endgroup$ Commented Mar 7 at 7:30
  • 1
    $\begingroup$ @reza the right side includes all of the terms on the left side when $k=l$, then adds the additional terms when $k\neq l$, and each term is positive because of the absolute value $\endgroup$ Commented Mar 22 at 5:00

Special case I: Product states

A bipartite quantum state$^1$ $\rho_{AB}$ is said to be a product state if it can be written as $\rho_{AB}=\rho_A\otimes\rho_B$ for some quantum states $\rho_A$ and $\rho_B$ of the subsystems $A$ and $B$. If $\rho_{AB}$ is a product state, then $$ P_{AB}=\mathrm{tr}(\rho_{AB}^2)=\mathrm{tr}(\rho_A^2\otimes\rho_B^2)=P_A\cdot P_B.\tag1 $$

Special case II: Pure product states

If a product state $\rho_{AB}$ is pure, then $\rho_A$ and $\rho_B$ are pure, too$^2$. Therefore, if $\rho_{AB}$ is a pure product state, then $$ P_{AB}=P_A=P_B=1.\tag2 $$

General case: Separable states

A bipartite product state $\rho_{AB}$ is said to be separable if it is a convex combination of product states, i.e. if $\rho_{AB}=\sum_k c_k\rho_{A,k}\otimes\rho_{B,k}$ for some real numbers $c_k\in(0,1]$ that add up to one and quantum states $\rho_{A,k}$ and $\rho_{B,k}$ of the subsystems. Consequently, if $\rho_{AB}$ is separable, then $$ \begin{align} P_A&=\sum_{ij}c_ic_j\cdot\mathrm{tr}(\rho_{A,i}\rho_{A,j})\tag3\\ P_B&=\sum_{ij}c_ic_j\cdot\mathrm{tr}(\rho_{B,i}\rho_{B,j})\tag4\\ P_{AB}&=\sum_{ij}c_ic_j\cdot\mathrm{tr}(\rho_{A,i}\rho_{A,j})\cdot\mathrm{tr}(\rho_{B,i}\rho_{B,j})\tag5 \end{align} $$ but $0\leqslant \mathrm{tr}(\rho_{A,i}\rho_{A,j})\leqslant 1$ by Cauchy-Schwarz inequality and we obtain $$ P_{AB}\leqslant\min(P_A,P_B).\tag6 $$

$^1$ Here, a quantum state is a positive semidefinite operator with unit trace.
$^2$ This follows by considering the rank of the tensor product of two matrices. However, it also follows from $(1)$.

  • $\begingroup$ Thank you adam for the answer. Im wondering about a general case ( a mixed entangled state) $\endgroup$
    – reza
    Commented Mar 6 at 9:47

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.