Abstract
The quantum entanglement E of a bipartite quantum Ising chain is compared with the mutual information I between the two parts after a local measurement of the classical spin configuration. As the model is conformally invariant, the entanglement measured in its ground state at the critical point is known to obey a certain scaling form. Surprisingly, the mutual information of classical spin configurations is found to obey the same scaling form, although with a different prefactor. Moreover, we find that mutual information and the entanglement obey the inequality I ≤ E in the ground state as well as in a dynamically evolving situation. This inequality holds for general bipartite systems in a pure state and can be proved using similar techniques as for Holevo's bound.
Export citation and abstract BibTeX RIS
1. Introduction
Recently the study of quantum spin chains from the perspective of quantum information theory has attracted considerable attention. This applies in particular to entanglement studies of quantum spin chains in their ground state ρ = |ψ0〉〈ψ0|. In these studies a quantum chain is fictitiously divided into two parts A and B (see figure 1). The quantum entanglement between the two parts is then given by
where S[ρA,B] =− tr[ρA,BlnρA,B] denotes the von Neumann entropy of the reduced density matrices ρA = trB[ρ] and ρB = trA[ρ]. The entanglement EA:B is particularly interesting to study in the context of critical quantum chains with an underlying conformal symmetry [1], which are characterized by long-range correlations in the ground state. Using methods of conformal field theory it was shown in [2]–[6] that the entanglement in such systems obeys the scaling form
where L is the total length of the chain, l is the length of section A, and f is a scaling function. In this expression the constant a is non-universal, meaning that it depends on the specific microscopic realization of the respective model, while the constant c turns out to be universal. Remarkably, c is equal to the so-called central charge of the underlying conformal field theory which labels the universality class. For example, the Ising universality class is characterized by the central charge c = 1/2.
To determine the entanglement experimentally one has to perform a variety of repeated measurements acting simultaneously on all spins in one of the sectors. Such a task is usually difficult to perform. In fact, it would be much simpler to measure individual spin orientations locally and to study the resulting classical spin configurations. Such a measurement is expected to destroy the existing entanglement and to convert it to some extent into classical correlations between the two parts of the system. Such classical correlations are usually quantified by the mutual information
where the terms on the right-hand side denote the Shannon entropy of the classical spin configurations after the measurement in the sections A, B and in the whole chain.
In this paper we address the question as to what extent the quantum entanglement and the classical mutual information are related to each other. The main results are:
- In the quantum Ising model the mutual information obeys the same scaling law (2) as the entanglement at the critical point, although with a different prefactor.
- The mutual information obeys the inequality IA:B ≤ EA:B for general bipartite systems in a pure state.
The paper is organized as follows. In section 2 we first discuss the example of the quantum Ising chain, determining the entanglement and the mutual information in finite-size systems by different methods. In section 3 we give a general proof of the inequality IA:B ≤ EA:B. Finally, in section 4 we present a summary and discuss the relationship between our inequality evoked by local measurements and the monotonicity of the quantum relative entropy.
2. Entanglement and mutual information in the quantum Ising chain
The quantum Ising chain with the length L is defined by the Hamiltonian
where σx,y,z are Pauli matrices and g is the strength of the transverse field. We use periodic boundary conditions by setting L + 1 ≡ 1. The ground state of this model exhibits an order–disorder phase transition at the critical point gc = 1: At g = 0 all spins are aligned in one direction of the z axis, and as g is increased the magnetization is weakened and vanishes at g = 1, which means that the system goes to the paramagnetic phase in the z direction at g > 1.
2.1. Scaling behavior EA:B and IA:B in the ground state
To obtain the mutual information of classical spin configurations in the ground state, we have to compute the probability for each classical configuration after a measurement of local spin orientations. If we choose the z axis for the spin orientations, measuring , these probabilities can be computed from the coefficients of the ground state wavefunction |ψ0〉 represented in the product basis of the eigenvectors of . For example, if we consider an Ising chain with two spins in the ground state
where |↑↓〉 ≡ |↑〉1⊗|↓〉2 is the product basis of the eigenvectors |↑〉i and |↓〉i of , the probability P({σz}) to obtain the classical configuration {σz} is given by , and . To determine the ground state of the Hamiltonian (4) at finite L, we use the Lanczos method [7].
Each classical configuration obtained from the measurement is now divided into two segments, . Determining the Shannon entropy of the probability distribution in the segments and in the whole chain,
where and , one can obtain the classical mutual information IA:B(L,l) = HA + HB − HAB for a given length l of section A.
On the other hand, to compute the entanglement EA:B(L,l) between sections A and B, we perform the partial trace on the density matrix ρ = |ψ0〉〈ψ0| obtained from the Lanczos method, leading to the reduced density matrices ρA and ρB. Since the entanglement of a pure state is given by the von Neumann entropy of the reduced density matrices (see equation (1)), we can determine the entanglement EA:B(L,l) by numerically diagonalizing ρA or ρB.
Using this method, we calculate EA:B(L,l) and IA:B(L,l) for various values of g, up to L = 24. As expected, at the critical point g = 1 the entanglement EA:B(L,l) is found to obey the scaling form (2), as shown in figure 2 with f(ξ) = 1/πsin(πξ). Surprisingly, the mutual information IA:B(L,l) also obeys the same type of scaling form. To illustrate this finding, we plot two scaling forms
as a solid and a dashed line in figure 2, together with the numerical data obtained by the Lanczos method. Here, we have used a ≈ 0.478, a' ≈ 0.329, and c' ≈ 0.715. This means that the initial entanglement and the mutual information after the measurement differ only by a factor at the critical point.
Download figure:
Standard imageNote that the classical mutual information after the measurement is always smaller than or equal to the initial entanglement in the ground state. The same observation holds in the off-critical case g ≠ 1 (not shown here).
2.2. Behavior in a non-stationary pure state
The inequality IA:B ≤ EA:B is valid not only in the ground state but also in time-dependent pure states of the quantum Ising chain. The time evolution of the density matrix is given by
where H is the quantum Ising Hamiltonian (4) and |ψ(0)〉 is the initial state. Since ρ(t) is also a pure state, one can obtain EA:B(L,l) and IA:B(L,l) using the same methods as in the ground state. In figure 3 we show results for g = 0.1 and 1.0 in a chain with L = 10 and l = 5, using an initial state that is fully magnetized in the z-direction. As can be seen, the quantities oscillate irregularly but satisfy the inequality IA:B ≤ EA:B at any time.
Download figure:
Standard imageIt turns out that this inequality holds generally for arbitrary local measurements on entangled pure states in bipartite system, as will be proved in section 3.
3. Proof of the inequality IA:B ≤ EA:B
To prove the inequality between mutual information and entanglement we use similar techniques as for the proof of Holevo's inequality [9]. This suggests that both inequalities may be closely related or even equivalent.
3.1. Measurement
In the following let us consider an arbitrary bipartite system on AB = A⊗B which is in a pure state ρAB = |ψ〉〈ψ| so that the von Neumann entropy S[ρAB] =− tr[ρABlnρAB] vanishes. In such a pure state the entanglement between the subsystems A and B is given by
where ρA = trBρAB and ρB = trAρAB denote the reduced density matrices. On both sides let us now perform local projective measurements
with the projectors Πa = |ϕa〉〈ϕa| and Πb = |ϕb〉〈ϕb|, which may be thought of as spin configuration measurements, as in the example given above. This measurement converts the pure state ρAB into a mixed state
where Πab = Πa⊗Πb. Moreover, the measurement completely destroys the entanglement between the two subsystems. To see this, let us consider the entanglement of formation
defined as the infimum of the entropy over all possible statistical ensembles represented by . Since is non-negative and the entropy of the particular representative
vanishes, we can conclude that after the measurement. This means that the measurement destroys the original entanglement and converts it to some extent into classical correlations which can be quantified by the mutual information IA:B in equation (3). Note that the classical mutual information IA:B is equivalent to the quantum mutual information of the post-measurement state , i.e.,
where and .
3.2. Expressing the measurement as an isometry in an extended space
According to the Stinespring theorem [8], the measurement process (16) can be carried out by embedding ρAB in a higher-dimensional Hilbert space, performing a unitary transformation on it, and finally tracing out the additional degrees of freedom. To this end let us extend the Hilbert space AB = A⊗B by an auxiliary space whose task will be to store the measurement outcome encoded in the form of orthonormal basis vectors . Moreover, let us define the linear map
which obeys the condition VV† = 1 so that it maps the initial state isometrically onto a subset of the extended space like a unitary (entropy-preserving) transformation. With the corresponding extended density matrix
the measurement process (16) can now be written as
3.3. Schmidt decomposition
According to Schmidt's theorem, any quantum state |ψ〉∈AB can be decomposed into
with certain vectors |iA〉∈A, |iB〉∈B and probabilities λi∈[0,1], called Schmidt coefficients, which sum up to 1. This means that the initial state can be written as
Calculating the partial traces we obtain
and a similar expression for ρB, meaning that the Shannon entropy of the Schmidt coefficients is equal to the initial entanglement:
3.4. Encoding the Schmidt decomposition in another auxiliary space
Let us now introduce another auxiliary Hilbert space C with canonical basis vectors |ij〉 whose task will be to store pairs of Schmidt coefficients. On the combined Hilbert space ABC = AB⊗C we define the density operator
If we compare this operator with its own diagonal part
It is easy to check that integer powers of these operators have always the same trace, i.e. for all k∈N, meaning that they have the same entropy
We now use the map V defined in equation (18) to extend this operator even further to the space by defining
As V is an isometry, this extension does not change the entropy, hence
3.5. Tracing out the original Hilbert space AB and the auxiliary space C
Tracing out the original Hilbert we have trAB[|ϕab〉〈ϕa'b'|] = δaa'δbb', leading to
Since the density matrix in equation (30) has only diagonal elements, the von Neumann entropy of and its reduced density matrices can be obtained easily as follows:
where and . Combining equations (31)–(33), one obtains a useful relation
Tracing out the auxiliary space C we obtain the reduced density matrices , and in which the outcome of the measurement is stored. Therefore, the classical mutual information after the measurement can be expressed in terms of these density matrices by
3.6. Apply strong subadditivity
To prove the inequality IA:B ≤ EA:B, we use the strong subadditivity of the von Neumann entropy [10], which is known to hold in the two following equivalent forms [11]:
Using the first relation the mutual information can be written as
Replacing one obtains in the same way
Adding (38) and (39) gives an inequality, where we again apply the second version of strong subadditivity (37)
proving the initial assertion.
4. Conclusion
We have investigated the relation between the initial entanglement and the classical mutual information after local projective measurements in a bipartite quantum system which is initially in a pure state. For the quantum Ising chain with periodic boundary conditions, which is divided into two segments, it is found that both quantities obey the same scaling form at the critical point, differing only in the prefactor of the scaling function. For the entanglement we find the prefactor c/3 = 1/6 in agreement with predictions of conformal field theory, whereas the prefactor of the mutual information is found to be less than 1/6.
Furthermore, we have observed that classical mutual information cannot exceed the initial entanglement in arbitrary pure states. The inequality IA:B ≤ EA:B has been proved generally by successively expanding and reducing the Hilbert space and by applying strong subadditivity of the von Neumann entropy. For entangled pure states, we conclude that a local projective measurement destroys the original quantum correlation and converts it into classical mutual information bounded from above by the initial entanglement.
In a general case, it is more convenient using the quantum mutual information I(ρAB) = S[ρA] + S[ρB] − S[ρAB], where the density operators can be defined in a pure or a mixed state, to measure the correlation in a quantum system at arbitrary temperature. Note that if a system is given by a pure state, the quantum mutual information becomes twice the entanglement entropy, because S[ρAB] = 0. According to the monotonicity of the quantum relative entropy, the local projective measurement does not increase the quantum mutual information [12], that is,
where is the post-measurement density operator. If the initial system is in a pure state, equation (41) yields IA:B ≤ 2EA:B. Consequently, our inequality IA:B ≤ EA:B provides a more enhanced bound for the classical mutual information after the local projective measurement. As a future work, generalized inequalities induced by the measurement on a mixed state will be studied in the context of the quantum mutual information.
Acknowledgments
This work was supported by the Basic Science Research Program of MEST, NRF grant No. 2010-0009697.