This paper was converted on www.awesomepapers.org from LaTeX by an anonymous user.
Want to know more? Visit the Converter page.

Quantum Discords of Tripartite Quantum Systems

Jianming Zhou Zhou: Department of Mathematics, Shanghai University, Shanghai 200444, China 272410225@qq.com Xiaoli Hu Hu: School of Artificial Intelligence, Jianghan University, Wuhan, Hubei 430056, China xiaolihumath@jhun.edu.cn (Corresponding author)  and  Naihuan Jing Jing: Department of Mathematics, North Carolina State University, Raleigh, NC 27695, USA jing@ncsu.edu
Abstract.

The quantum discord of bipartite systems is one of the best-known measures of non-classical correlations and an important quantum resource. In the recent work appeared in [Phys. Rev. Lett 2020, 124:110401], the quantum discord has been generalized to multipartite systems. In this paper, we give analytic solutions of the quantum discord for tripartite states with fourteen parameters.

Key words and phrases:
Quantum discord, quantum correlations, tripartite quantum states, optimization on manifolds
Key words and phrases:
Quantum discord, quantum correlations, tripartite quantum states, optimization on manifolds
2010 Mathematics Subject Classification:
Primary: 81P40; Secondary: 81Qxx
*Corresponding author: Xiaoli Hu (xiaolihumath@jhun.edu.cn)

1. Introduction

The quantum discord usually involves with quantum entanglement and umentangled quantum correlations in quantum systems. It measures the total non-classical correlation in a quantum system, and has attracted widespread attention since its appearance. Applications of the non-entanglement quantum correlations in quantum information processings have been extensively studied, including the quantum computing scheme of DQC1 [1] and Grover search algorithm [2] etc. This partly explains why quantum schemes surpass classical schemes. Meanwhile, the quantum discord as a non-classical correlation is one of the important quantum resources and is ubiquitous in many areas of modern physics ranging from condensed matter physics, quantum optics, high-energy physics to quantum chemistry, thus can be regarded as one of the fundamental non-classical correlations besides entanglement and EPR-steerable states [3, 4].

The quantum discord is defined as the maximal difference between the quantum mutual information without and with a von Neumann projective measurement applying to one part of the bipartite system. For tripartite and lager systems, some generalizations of the discord have been proposed [5, 6, 7, 8, 9, 10], and have been used in quantum information processings. It is well-known that quantum discord is extremely difficult to evaluate and most exact solutions are only for the X-type quantum states (cf. [11, 12, 13, 14]). This paper is devoted to quantification of the quantum correlation in tripartite and larger systems to derive some exact solutions for non-X-type states, and we hope it can contribute to better understanding and more effective use of quantum states in realizing quantum information processing schemes.

The paper is organized as follows. We first introduce the generalized discord for tripartite systems [10] based on that of bipartite systems [3]. We derive analytic solutions for tripartite states with fourteen parameters. Furthermore, the quantum discord of some well-known states (such as GHZ states) are computed.

2. Generalization of quantum discord to tripartite states

For a bipartite state ϕbc\phi^{bc} on system HBHCH_{B}\otimes H_{C}, the quantum mutual information is I(ϕbc):=SB(ϕb)+SC(ϕc)SBC(ϕbc)I(\phi^{bc}):=S_{B}(\phi^{b})+S_{C}(\phi^{c})-S_{BC}(\phi^{bc}), where S(ϕX)=TrϕXlog2(ϕX)S(\phi^{X})=\mathrm{Tr}\phi^{X}\log_{2}(\phi^{X}) is the von Neumann entropy of the quantum state on system X. Set {ΠkB}\{\Pi^{B}_{k}\} to be an one-dimensional von Neumann projection operator on subsystem BB which satisfies kΠkB=I,(ΠkB)2=ΠkB,ΠkBΠkB=δkk\sum_{k}\Pi_{k}^{B}=I,(\Pi_{k}^{B})^{2}=\Pi_{k}^{B},\Pi_{k}^{B}\Pi_{k^{\prime}}^{B}=\delta_{kk^{{}^{\prime}}}. Then the state ϕbc\phi^{bc} under the measurement {ΠkB}\{\Pi^{B}_{k}\} is changed into

ϕkc=1pkTrB(IΠkB)ϕbc(IΠkB)\phi^{c}_{k}=\frac{1}{p_{k}}\mathrm{Tr}_{B}(I\otimes\Pi^{B}_{k})\phi^{bc}(I\otimes\Pi^{B}_{k})

with the probability pk=Tr(IΠkB)ϕbc(IΠkB)p_{k}=\mathrm{Tr}(I\otimes\Pi^{B}_{k})\phi^{bc}(I\otimes\Pi^{B}_{k}). For simplicity, we denote by ΠX\Pi^{X} the measurement {ΠkX}\{\Pi_{k}^{X}\} on system XX. The quantum conditional entropy is simply given by SC|ΠB(ϕbc)=kpkS(ϕkc)S_{C|\Pi^{B}}(\phi^{bc})=\sum_{k}p_{k}S(\phi^{c}_{k}). Then the measurement-induced quantum mutual information is given by

C(ϕbc)=SC(ϕc)minSC|ΠB(ϕbc).C(\phi^{bc})=S_{C}(\phi^{c})-\min S_{C|\Pi^{B}}(\phi^{bc}).

By Olliver and Zurek [3], the original definition of the quantum discord Q(ρ)Q(\rho) is the difference of the quantum mutual information I(ϕbc)I(\phi^{bc}) and the measurement-induced quantum mutual information C(ϕbc)C(\phi^{bc}), i.e.

(2.1) Q(ϕbc)=I(ϕbc)C(ϕbc)=minΠB{SC|ΠB(ϕbc)SC|B(ϕbc)},Q(\phi^{bc})=I(\phi^{bc})-C(\phi^{bc})=\min_{\Pi^{B}}\{S_{C|\Pi^{B}}(\phi^{bc})-S_{C|B}(\phi^{bc})\},

where SC|B(ϕbc)=SBC(ϕbc)SB(ϕb)S_{C|B}(\phi^{bc})=S_{BC}(\phi^{bc})-S_{B}(\phi^{b}) is the unmeasured conditional state on subsystem CC.

For the tripartite system HAHBHCH_{A}\otimes H_{B}\otimes H_{C}, we consider the BCBC composite system as the first subsystem and AA-system as the second subsystem. The state ρabc\rho^{abc} of system HAHBHCH_{A}\otimes H_{B}\otimes H_{C} gives arise to a state on BCBC-subsystem after the von Neumann measurement {ΠjA}\{\Pi_{j}^{A}\} on AA subsystem. Namely, it takes the following form:

(2.2) ρjbc=1pjbcTrA(ΠjAI)ρabc(ΠjAI)\begin{split}\rho^{bc}_{j}=\frac{1}{p^{bc}_{j}}\mathrm{Tr}_{A}(\Pi_{j}^{A}\otimes I)\rho^{abc}(\Pi_{j}^{A}\otimes I)\end{split}

with probability pjbc=Tr(ΠjAI)ρabc(ΠjAI)p^{bc}_{j}=\mathrm{Tr}(\Pi_{j}^{A}\otimes I)\rho^{abc}(\Pi_{j}^{A}\otimes I).The measured quantum mutual information of ρabc\rho^{abc} is naturally given by

(2.3) 𝒥(ρabc|ΠA)=SBC(ρbc)SBC|ΠA(ρabc).\begin{split}\mathcal{J}(\rho^{abc}|\Pi^{A})=S_{BC}(\rho^{bc})-S_{BC|\Pi^{A}}(\rho^{abc}).\end{split}

The quantity of classical correlation of the tripartite state ρabc\rho^{abc} is

(2.4) 𝒞(ρabc)=maxΠA𝒥(ρabc|ΠA)=SBC(ρbc)minΠASBC|ΠA(ρabc).\begin{split}\mathcal{C}(\rho^{abc})=\max_{\Pi^{A}}\mathcal{J}(\rho^{abc}|\Pi^{A})=S_{BC}(\rho^{bc})-\min_{\Pi^{A}}S_{BC|\Pi^{A}}(\rho^{abc}).\end{split}

We know that the quantum mutual information I(ρabc)=SA(ρa)+SBC(ρbc)SABC(ρabc)I(\rho^{abc})=S_{A}(\rho^{a})+S_{BC}(\rho^{bc})-S_{ABC}(\rho^{abc}). Similar to Eq.(2.1), the generalized quantum discord of the tripartite state ρabc\rho^{abc} can be defined as

(2.5) 𝒬(ρabc)=I(ρabc)C(ρabc)=minΠA{SBC|ΠA(ρabc)SBC|A(ρabc)},\begin{split}\mathcal{Q}(\rho^{abc})=I(\rho^{abc})-C(\rho^{abc})=\min_{\Pi^{A}}\{S_{BC|\Pi^{A}}(\rho^{abc})-S_{BC|A}(\rho^{abc})\},\end{split}

where SBC|A(ρabc)=SABC(ρabc)SA(ρa)S_{BC|A}(\rho^{abc})=S_{ABC}(\rho^{abc})-S_{A}(\rho^{a}) is the unmeasured conditional entropy on BCBC-bipartite subsystem.

In order to evaluate the quantity minΠASBC|ΠA(ρabc)\min_{\Pi^{A}}S_{BC|\Pi^{A}}(\rho^{abc}), the multipartite measurement based on conditional operators can be constructed as follows: [15]

(2.6) ΠjkAB=ΠjAΠk|jB\begin{split}\Pi_{jk}^{AB}=\Pi_{j}^{A}\otimes\Pi_{k|j}^{B}\end{split}

with the measurement ordering from AA to BB. The projector Πk|jB\Pi_{k|j}^{B} on subsystem BB is conditional measurement outcome of AA. These projectors satisfy kΠk|jB=IB,jΠjA=IA\sum_{k}\Pi^{B}_{k|j}=I^{B},\sum_{j}\Pi_{j}^{A}=I^{A}. Then after the measurement ΠjkAB\Pi_{jk}^{AB}, the state ρabc\rho^{abc} is collapsed to a state on subsystem CC, i.e.

(2.7) ρjkc=1pjkcTrAB(ΠjkABI)ρabc(ΠjkABI)\begin{split}\rho^{c}_{jk}=\frac{1}{p^{c}_{jk}}\mathrm{Tr}_{AB}(\Pi_{jk}^{AB}\otimes I)\rho^{abc}(\Pi_{jk}^{AB}\otimes I)\end{split}

with the probability pjkc=Tr(ΠjkABI)ρabc(ΠjkABI)p^{c}_{jk}=\mathrm{Tr}(\Pi_{jk}^{AB}\otimes I)\rho^{abc}(\Pi_{jk}^{AB}\otimes I). The conditional entropy after the ABAB-bipartite measurement is

SC|ΠAB(ρabc)=jklpjkcλl(jk)log2λl(jk),S_{C|\Pi^{AB}}(\rho^{abc})=\sum_{jkl}p_{jk}^{c}\lambda_{l}^{(jk)}\log_{2}\lambda_{l}^{(jk)},

where λl(jk)\lambda_{l}^{(jk)} are eigenvalues of state ρjkc\rho^{c}_{jk}.

Let ρΠX=ΠXΠXρΠX\rho_{\Pi^{X}}=\sum_{\Pi^{X}}\Pi^{X}\rho\Pi^{X} be the state after measurement ΠX\Pi^{X}. Then for a bipartite state ρab\rho^{ab}, the conditional entropy on subsystem BB after the measurement on subsystem AA is

(2.8) SB|ΠA(ρab)=jpjSB(ρjb).\begin{split}S_{B|\Pi^{A}}(\rho^{ab})=\sum_{j}p_{j}S_{B}(\rho^{b}_{j}).\end{split}

By [10, Eq.(6)], the entropy of the measured system can always be decomposed as

(2.9) SAB(ρΠAab)=SA(ρΠAab)+SB|ΠA(ρab).S_{AB}(\rho^{ab}_{\Pi^{A}})=S_{A}(\rho^{ab}_{\Pi^{A}})+S_{B|\Pi^{A}}(\rho^{ab}).

For the tripartite system, using the measurement ΠAB\Pi^{AB}, we have

(2.10) SABC(ρΠABabc)SAB(ρΠABabc)=SC|ΠAB(ρabc),S_{ABC}(\rho^{abc}_{\Pi^{AB}})-S_{AB}(\rho^{abc}_{\Pi^{AB}})=S_{C|\Pi^{AB}}(\rho^{abc}),

when the measurement on AA system is ΠA\Pi^{A}, then we have

(2.11) SABC(ρΠAabc)SA(ρΠAabc)=SBC|ΠA(ρabc).S_{ABC}(\rho^{abc}_{\Pi^{A}})-S_{A}(\rho^{abc}_{\Pi^{A}})=S_{BC|\Pi^{A}}(\rho^{abc}).

By Eq.(2.9), Eq.(2.10), Eq.(2.11), we have that

SBC|ΠA(ρabc)=SB|ΠA(ρab)+SC|ΠAB(ρabc).S_{BC|\Pi^{A}}(\rho^{abc})=S_{B|\Pi^{A}}(\rho^{ab})+S_{C|\Pi^{AB}}(\rho^{abc}).

Meanwhile, SA(ρΠABabc)=SA(ρΠAabc)S_{A}(\rho^{abc}_{\Pi^{AB}})=S_{A}(\rho^{abc}_{\Pi^{A}}), so the generalization discord of a tripartite state can be written as [10]

(2.12) 𝒬(ρ):=minΠAB[SBC|A(ρ)+SB|ΠA(ρ)+SC|ΠAB(ρ)].\begin{split}\mathcal{Q}(\rho):=\min_{\Pi^{AB}}[-S_{BC|A}(\rho)+S_{B|\Pi^{A}}(\rho)+S_{C|\Pi^{AB}}(\rho)].\end{split}

3. Quantum Discord of non-X Qubit-Qutrit state

For the product states in the tripartite system, the discord has the special property that it reduces to the standard bipartite discord when only bipartite quantum correlations are present. This means 𝒬ABC(ρxρy)=𝒬X(ρx)\mathcal{Q}_{ABC}(\rho^{x}\otimes\rho^{y})=\mathcal{Q}_{X}(\rho^{x}) for X=AB,BCX=AB,BC and ACAC subsystem. We consider the following tripartite states

(3.1) ρabc=18(I8+a3σ3I4+I2b3σ3I2+I4i3ciσi+i3riσiσiI2+i3siσiI2σi+i3Tiσiσiσi),\begin{split}\rho^{abc}=&\frac{1}{8}(I_{8}+a_{3}\sigma_{3}\otimes I_{4}+I_{2}\otimes b_{3}\sigma_{3}\otimes I_{2}+I_{4}\otimes\sum_{i}^{3}c_{i}\sigma_{i}\\ +&\sum_{i}^{3}r_{i}\sigma_{i}\otimes\sigma_{i}\otimes I_{2}+\sum_{i}^{3}s_{i}\sigma_{i}\otimes I_{2}\otimes\sigma_{i}+\sum_{i}^{3}T_{i}\sigma_{i}\otimes\sigma_{i}\otimes\sigma_{i}),\end{split}

where IdI_{d} represents the unit matrix of order dd, and σi(i=1,2,3)\sigma_{i}(i=1,2,3) are Pauli matrices. The parameters a3,b3,ci,ri,si,Tia_{3},b_{3},c_{i},r_{i},s_{i},T_{i}\in\mathbb{R} and they are confined within the internal [1,1][-1,1]. Its matrix has the following form:

(3.2) ρ=(000000000000000000000000).\begin{split}\rho=\left(\begin{array}[]{cccccccc}*&*&0&0&0&*&*&*\\ &*&0&0&*&0&*&*\\ 0&0&*&*&*&*&0&*\\ 0&0&*&*&*&*&*&0\\ 0&*&*&*&*&*&0&0\\ &0&*&*&*&*&0&0\\ &*&0&*&0&0&*&*\\ &*&*&0&0&0&*&*\\ \end{array}\right).\end{split}

Let {|jj|,j=0,1}\{|j\rangle\langle j|,j=0,1\} be the computational base, then any von Neumann measurement on system XX can be written as {ΠjX=V|jj|V|,j=0,1}\{\Pi_{j}^{X}=V|j\rangle\langle j|V^{\dagger}|,j=0,1\} for some unitary matrix VSU(2)V\in\mathrm{SU}(2). Any unitary matrix can be written as V=tI+1kykσkV=tI+\sqrt{-1}\sum_{k}y_{k}\sigma_{k} with t,ykt,y_{k}\in\mathbb{R}. When the measurement {ΠjX}\{\Pi^{X}_{j}\} is performed locally on one part of the composite system YXY\otimes X, the ensemble {ρjY,pjY}\{\rho_{j}^{Y},p_{j}^{Y}\} is given by ρjY=1pjYTrX(IΠjX)ρYX(IΠjX)\rho_{j}^{Y}=\frac{1}{p_{j}^{Y}}\mathrm{Tr}_{X}(I\otimes\Pi^{X}_{j})\rho^{YX}(I\otimes\Pi^{X}_{j}) with the probability pjY=Tr[ρYX(IΠjX)]p_{j}^{Y}=\mathrm{Tr}[\rho^{YX}(I\otimes\Pi_{j}^{X})].

It follows from symmetry that

(3.3) Vσ1V=(t2+y12y22y32)σ1+2(ty3+y1y2)σ2+2(ty2+y1y3)σ3,Vσ2V=(t2+y22y32y12)σ1+2(ty1+y2y3)σ3+2(ty3+y1y2)σ1,Vσ3V=(t2+y32y12y22)σ1+2(ty2+y1y3)σ3+2(ty1+y2y3)σ2.\begin{split}V^{\dagger}\sigma_{1}V&=(t^{2}+y_{1}^{2}-y_{2}^{2}-y_{3}^{2})\sigma_{1}+2(ty_{3}+y_{1}y_{2})\sigma_{2}+2(-ty_{2}+y_{1}y_{3})\sigma_{3},\\ V^{\dagger}\sigma_{2}V&=(t^{2}+y_{2}^{2}-y_{3}^{2}-y_{1}^{2})\sigma_{1}+2(ty_{1}+y_{2}y_{3})\sigma_{3}+2(-ty_{3}+y_{1}y_{2})\sigma_{1},\\ V^{\dagger}\sigma_{3}V&=(t^{2}+y_{3}^{2}-y_{1}^{2}-y_{2}^{2})\sigma_{1}+2(ty_{2}+y_{1}y_{3})\sigma_{3}+2(-ty_{1}+y_{2}y_{3})\sigma_{2}.\\ \end{split}

Introduce new variables z1X=2(ty2+y1y3),z2X=2(ty1+y2y3),z3X=(t2+y32y12y22)z_{1}^{X}=2(-ty_{2}+y_{1}y_{3}),z_{2}^{X}=2(ty_{1}+y_{2}y_{3}),z_{3}^{X}=(t^{2}+y_{3}^{2}-y_{1}^{2}-y_{2}^{2}), then (z1X)2+(z2X)2+(z3X)2=1(z_{1}^{X})^{2}+(z_{2}^{X})^{2}+(z_{3}^{X})^{2}=1. Therefore ΠjXσkΠjX=(1)jzkXΠjX\Pi_{j}^{X}\sigma_{k}\Pi_{j}^{X}=(-1)^{j}z_{k}^{X}\Pi_{j}^{X} for j=0,1j=0,1 and k=1,2,3k=1,2,3.

For the tripartite state ρabc\rho^{abc}, the conditional state on BCBC subsystem after measurement {ΠjA(j=0,1)}\{\Pi_{j}^{A}(j=0,1)\} on subsystem AA is

(3.4) ρjbc=1pjbc(Π1AI2I2)ρabc(Π1AI2I2)=1pjbc[(1+(1)ja3z3A)I2I2+b3σ3I2+(1)ji3riziAσiI2+i3(ci+(1)jsiziA)I2σi+(1)ji3TiziAσiσi],\begin{split}\rho_{j}^{bc}&=\frac{1}{p_{j}^{bc}}(\Pi_{1}^{A}\otimes I_{2}\otimes I_{2})\rho^{abc}(\Pi_{1}^{A}\otimes I_{2}\otimes I_{2})\\ &=\frac{1}{p_{j}^{bc}}[(1+(-1)^{j}a_{3}z_{3}^{A})I_{2}\otimes I_{2}+b_{3}\sigma_{3}\otimes I_{2}+(-1)^{j}\sum_{i}^{3}r_{i}z_{i}^{A}\sigma_{i}\otimes I_{2}\\ &+\sum_{i}^{3}(c_{i}+(-1)^{j}s_{i}z_{i}^{A})I_{2}\otimes\sigma_{i}+(-1)^{j}\sum_{i}^{3}T_{i}z_{i}^{A}\sigma_{i}\otimes\sigma_{i}],\end{split}

where the probabilities are

pjbc=Tr((ΠjAI2I2)ρabc(ΠjAI2I2))=12[1+(1)ja3z3A],p_{j}^{bc}=\mathrm{Tr}((\Pi_{j}^{A}\otimes I_{2}\otimes I_{2})\rho^{abc}(\Pi_{j}^{A}\otimes I_{2}\otimes I_{2}))=\frac{1}{2}[1+(-1)^{j}a_{3}z_{3}^{A}],

and i3(ziA)2=1\sum_{i}^{3}(z^{A}_{i})^{2}=1. Therefore the reduced state of ρjbc\rho_{j}^{bc} is

ρjb=TrCρjbc=12(1+(1)ja3z3A)[(1+(1)ja3z3A)I2+b3σ3+(1)ji3riziAσi]\rho_{j}^{b}=\mathrm{Tr}_{C}\rho_{j}^{bc}=\frac{1}{2(1+(-1)^{j}a_{3}z_{3}^{A})}[(1+(-1)^{j}a_{3}z_{3}^{A})I_{2}+b_{3}\sigma_{3}+(-1)^{j}\sum_{i}^{3}r_{i}z_{i}^{A}\sigma_{i}]

with the probability pjb=pjbc=12[1+(1)ja3z3A]p_{j}^{b}=p_{j}^{bc}=\frac{1}{2}[1+(-1)^{j}a_{3}z_{3}^{A}]. The eigenvalues of ρjb\rho_{j}^{b} are

λj±=12(1+(1)ja3z3A)[1+(1)ja3z3A±(b3+(1)jr3z3A)2+i2(riziA)2].\lambda^{\pm}_{j}=\frac{1}{2(1+(-1)^{j}a_{3}z_{3}^{A})}[1+(-1)^{j}a_{3}z_{3}^{A}\pm\sqrt{(b_{3}+(-1)^{j}r_{3}z_{3}^{A})^{2}+\sum_{i}^{2}(r_{i}z_{i}^{A})^{2}}].

We define the following entropy function

(3.5) Hε(x)=12[(1+ε+x)log2(1+ε+x)+(1+εx)log2(1+εx)].\begin{split}H_{\varepsilon}(x)=\frac{1}{2}[(1+\varepsilon+x)\log_{2}(1+\varepsilon+x)+(1+\varepsilon-x)\log_{2}(1+\varepsilon-x)].\end{split}

Then measured conditional entropy of BB subsystem can be obtained as [3, 4, 14, 16, 17, 18]

(3.6) SB|ΠA(ρ)=jpjb(λj+log2λj++λjlog2λj)=12[Ha3z3A(A+)+Ha3z3A(A)2H(a3z3A)2],\begin{split}S_{B|\Pi^{A}}(\rho)&=-\sum_{j}p_{j}^{b}({\lambda^{+}_{j}\log_{2}\lambda^{+}_{j}}+\lambda^{-}_{j}\log_{2}\lambda^{-}_{j})\\ &=-\frac{1}{2}[H_{a_{3}z_{3}^{A}}(A_{+})+H_{-a_{3}z_{3}^{A}}(A_{-})-2H(a_{3}z_{3}^{A})-2],\end{split}

where A±=(b3±r3z3A)2+i2(riziA)2A_{\pm}=\sqrt{(b_{3}\pm r_{3}z_{3}^{A})^{2}+\sum_{i}^{2}(r_{i}z_{i}^{A})^{2}}.

After measurement Πk|jB\Pi_{k|j}^{B} on BCBC system, the state ρjbc\rho_{j}^{bc} is changed to

(3.7) ρjkc=1pjkc[(1+(1)ja3z3A+(1)kb3z3B+(1)j+ki3riziAziB)I2+i3(ci+(1)jsiziA+(1)k+jTiziAziB)σi],(j,k=0,1)\begin{split}\rho_{jk}^{c}=&\frac{1}{p_{jk}^{c}}[(1+(-1)^{j}a_{3}z_{3}^{A}+(-1)^{k}b_{3}z_{3}^{B}+(-1)^{j+k}\sum_{i}^{3}r_{i}z_{i}^{A}z_{i}^{B})I_{2}\\ +&\sum_{i}^{3}(c_{i}+(-1)^{j}s_{i}z_{i}^{A}+(-1)^{k+j}T_{i}z_{i}^{A}z_{i}^{B})\sigma_{i}],(j,k=0,1)\end{split}

with the probability (k=0,1k=0,1)

(3.8) p0kc=12(1+a3z3A)(1+αk),p1kc=12(1a3z3A)(1+βk),p_{0k}^{c}=\frac{1}{2(1+a_{3}z_{3}^{A})}(1+\alpha_{k}),\ \ p_{1k}^{c}=\frac{1}{2(1-a_{3}z_{3}^{A})}(1+\beta_{k}),

where αk=a3z3A+(1)k(b3z3B+i3riziAziB),βk=a3z3A+(1)k(b3z3Bi3riziAziB)\alpha_{k}=a_{3}z_{3}^{A}+(-1)^{k}(b_{3}z_{3}^{B}+\sum_{i}^{3}r_{i}z_{i}^{A}z_{i}^{B}),\beta_{k}=-a_{3}z_{3}^{A}+(-1)^{k}(b_{3}z_{3}^{B}-\sum_{i}^{3}r_{i}z_{i}^{A}z_{i}^{B}). The non-zero eigenvalues of ρjkc\rho^{c}_{jk} are given by

(3.9) λ0k±=12(1+αk)(1+αk±γk),λ1k±=12(1+βk)(1+βk±δk),k=0,1,\lambda_{0k}^{\pm}=\frac{1}{2(1+\alpha_{k})}(1+\alpha_{k}\pm\gamma_{k}),\ \ \lambda_{1k}^{\pm}=\frac{1}{2(1+\beta_{k})}(1+\beta_{k}\pm\delta_{k}),k=0,1,

where

γk=[i3(ci+siziA+(1)kTiziAziB)2]12,δk=[i3(ci+siziA+(1)kTiziAziB)2]12.\begin{split}\gamma_{k}&=[\sum_{i}^{3}(c_{i}+s_{i}z_{i}^{A}+(-1)^{k}T_{i}z_{i}^{A}z^{B}_{i})^{2}]^{\frac{1}{2}},\\ \delta_{k}&=[\sum_{i}^{3}(-c_{i}+s_{i}z_{i}^{A}+(-1)^{k}T_{i}z_{i}^{A}z^{B}_{i})^{2}]^{\frac{1}{2}}.\end{split}

According to the fact that the eigenvalues in Eq.(3.9) are nonnegative, we have i3ai2+i3bi2+i3ri21\sqrt{\sum_{i}^{3}a_{i}^{2}}+\sqrt{\sum_{i}^{3}b_{i}^{2}}+\sqrt{\sum_{i}^{3}r_{i}^{2}}\leq 1.

The entropy of ρabc\rho^{abc} under the measurement ΠAB\Pi^{AB} is given by

(3.10) SC|ΠAB(ρ)=j,kpjkc(λjk+log2λjk++λjklog2λjk)=12(1+a3z3A)[Hα0(γ0)+Hα1(γ1)2Ha3z3A(α0α12)]12(1a3z3A)[Hβ0(δ0)+Hβ1(δ1)2Ha3z3A(β0β12)]+2.\begin{split}S_{C|\Pi^{AB}}(\rho)=&-\sum_{j,k}p^{c}_{jk}(\lambda_{jk}^{+}\log_{2}\lambda_{jk}^{+}+\lambda_{jk}^{-}\log_{2}\lambda_{jk}^{-})\\ =&-\frac{1}{2(1+a_{3}z_{3}^{A})}[H_{\alpha_{0}}(\gamma_{0})+H_{\alpha_{1}}(\gamma_{1})-2H_{a_{3}z_{3}^{A}}(\frac{\alpha_{0}-\alpha_{1}}{2})]\\ &-\frac{1}{2(1-a_{3}z_{3}^{A})}[H_{\beta_{0}}(\delta_{0})+H_{\beta_{1}}(\delta_{1})-2H_{-a_{3}z_{3}^{A}}(\frac{\beta_{0}-\beta_{1}}{2})]+2.\end{split}

In particularly, a3z3A=α0+α12a_{3}z_{3}^{A}=\frac{\alpha_{0}+\alpha_{1}}{2}.

Let G(z1A,z2A,z3A)=1SB|ΠA(ρ)G(z_{1}^{A},z_{2}^{A},z_{3}^{A})=1-S_{B|\Pi^{A}}(\rho) and F(z1A,z2A,z3A,z1B,z2B,z3B)=2SC|ΠAB(ρ)F(z_{1}^{A},z_{2}^{A},z_{3}^{A},z^{B}_{1},z^{B}_{2},z^{B}_{3})=2-S_{C|\Pi^{AB}}(\rho), then we have the following result.

Theorem 3.1.

For the non-X-states ρ\rho in Eq(3.1) with 14 parameters, the quantum discord is given by

(3.11) 𝒬(ρ)=SABC(ρ)+SA(ρ)+min{SB|ΠA(ρ)+SC|ΠAB(ρ)}=3+i=18λilog2λik=12λkalog2λkamaxziX[0,1],i(ziX)2=1{G+F},\begin{split}\mathcal{Q}(\rho)=&-S_{ABC}(\rho)+S_{A}(\rho)+\min\{S_{B|\Pi^{A}}(\rho)+S_{C|\Pi^{AB}}(\rho)\}\\ =&3+\sum_{i=1}^{8}{\lambda_{i}\log_{2}\lambda_{i}}-\sum_{k=1}^{2}\lambda_{k}^{a}\log_{2}\lambda_{k}^{a}-\max_{z^{X}_{i}\in[0,1],\sum_{i}(z_{i}^{X})^{2}=1}\{G+F\},\end{split}

where λi(i=1,,8)\lambda_{i}(i=1,\cdots,8) are the eigenvalues of ρabc\rho^{abc}, λka=12[1+(1)ka3],(k=0,1)\lambda_{k}^{a}=\frac{1}{2}[1+(-1)^{k}a_{3}],(k=0,1) are eigenvalues of ρabc\rho^{abc} on subsystem AA and XX represents subsystem A,BA,B.

Theorem 3.2.

Let r=max{|r1|,|r2|}r=\max{\{|r_{1}|,|r_{2}|\}}, then maxziX[0,1],i(ziX)2=1{G+F}\max_{z^{X}_{i}\in[0,1],\sum_{i}(z_{i}^{X})^{2}=1}\{G+F\} can be explicitly computed as follows.

Case1: when a3b3r30,r32r2a3b3r3a_{3}b_{3}r_{3}\leq 0,r_{3}^{2}-r^{2}\geq a_{3}b_{3}r_{3}, and (b3+r3)(c3+s3)0(b_{3}+r_{3})(c_{3}+s_{3})\leq 0, we have

(3.12) maxziX[0,1],i(ziX)2=1{G+F}=G(0,0,1)+F(0,0,1,0,0,1),\max_{z^{X}_{i}\in[0,1],\sum_{i}(z_{i}^{X})^{2}=1}\{G+F\}=G(0,0,1)+F(0,0,1,0,0,1),

where

(3.13) G(0,0,1)=12[Ha3(|b3+r3|)+Ha3(|b3r3|)2H(a3)]\begin{split}G(0,0,1)=\frac{1}{2}[H_{a_{3}}(|b_{3}+r_{3}|)+H_{-a_{3}}(|b_{3}-r_{3}|)-2H(a_{3})]\end{split}

and

(3.14) F(0,0,1,0,0,1)=12(1+a3)[Hα0(γ0)+Hα1(γ1)2Ha3(b3+r3)]+12(1a3)[Hβ0(δ0)+Hβ1(δ1)2Ha3(b3r3)].\begin{split}F(0,0,1,0,0,1)=&\frac{1}{2(1+a_{3})}[H_{\alpha_{0}}(\gamma_{0})+H_{\alpha_{1}}(\gamma_{1})-2H_{a_{3}}(b_{3}+r_{3})]\\ +&\frac{1}{2(1-a_{3})}[H_{\beta_{0}}(\delta_{0})+H_{\beta_{1}}(\delta_{1})-2H_{-a_{3}}(b_{3}-r_{3})].\end{split}

In this case, the parameters are degenerated into (k=0,1)(k=0,1)

αk=a3+(1)k(b3+r3),γk=[i3ci2+s32+T32+2(c3s3+(1)k(c3T3+s3T3))]12,\alpha_{k}=a_{3}+(-1)^{k}(b_{3}+r_{3}),\gamma_{k}=[\sum_{i}^{3}c_{i}^{2}+s_{3}^{2}+T_{3}^{2}+2(c_{3}s_{3}+(-1)^{k}(c_{3}T_{3}+s_{3}T_{3}))]^{\frac{1}{2}},
βk=a3+(1)k(b3r3),δk=[i3ci2+s32+T32+2(c3s3+(1)k(s3T3c3T3))]12.\beta_{k}=-a_{3}+(-1)^{k}(b_{3}-r_{3}),\delta_{k}=[\sum_{i}^{3}c_{i}^{2}+s_{3}^{2}+T_{3}^{2}+2(-c_{3}s_{3}+(-1)^{k}(s_{3}T_{3}-c_{3}T_{3}))]^{\frac{1}{2}}.

Case 2: (1) When b3=0,c1s10,s1|c1|b_{3}=0,c_{1}s_{1}\leq 0,s_{1}\leq|c_{1}| and max{|r1|,|r2|,|r3|}=|r1|\max\{|r_{1}|,|r_{2}|,|r_{3}|\}=|r_{1}|, we have

(3.15) maxziX[0,1],i(ziX)2=1{G+F}=G(1,0,0)+F(1,0,0,1,0,0),\max_{z^{X}_{i}\in[0,1],\sum_{i}(z_{i}^{X})^{2}=1}\{G+F\}=G(1,0,0)+F(1,0,0,1,0,0),

where

(3.16) G(1,0,0)=12[Ha3(r1)+Ha3(r1)2H(a3)]G(1,0,0)=\frac{1}{2}[H_{a_{3}}(r_{1})+H_{-a_{3}}(r_{1})-2H(a_{3})]

and

(3.17) F(1,0,0,1,0,0)=12[Hr1(γ0)+Hr1(γ1)+Hr1(δ0)+Hr1(δ1)4H(r1)].\begin{split}F(1,0,0,1,0,0)=\frac{1}{2}[H_{r_{1}}(\gamma_{0})+H_{-r_{1}}(\gamma_{1})+H_{r_{1}}(\delta_{0})+H_{-r_{1}}(\delta_{1})-4H(r_{1})].\end{split}

In this case, the parameters are degenerated into (k=0,1)(k=0,1)

γk=[i3ci2+s12+T12+2(c1s1+(1)k(c1T1+s1T1))]12,δk=[i3ci2+s12+T12+2(c1s1+(1)k(c1T1s1T1))]12.\begin{split}\gamma_{k}&=[\sum_{i}^{3}c_{i}^{2}+s_{1}^{2}+T_{1}^{2}+2(c_{1}s_{1}+(-1)^{k}(c_{1}T_{1}+s_{1}T_{1}))]^{\frac{1}{2}},\\ \delta_{k}&=[\sum_{i}^{3}c_{i}^{2}+s_{1}^{2}+T_{1}^{2}+2(-c_{1}s_{1}+(-1)^{k}(c_{1}T_{1}-s_{1}T_{1}))]^{\frac{1}{2}}.\end{split}

(2) When b3=0,c1s10,s1|c1|b_{3}=0,c_{1}s_{1}\leq 0,s_{1}\leq|c_{1}| and max{|r1|,|r2|,|r3|}=|r2|\max\{|r_{1}|,|r_{2}|,|r_{3}|\}=|r_{2}|, we have

(3.18) maxziX[0,1],i(ziX)2=1{G+F}=G(0,1,0)+F(0,1,0,0,1,0),\max_{z^{X}_{i}\in[0,1],\sum_{i}(z_{i}^{X})^{2}=1}\{G+F\}=G(0,1,0)+F(0,1,0,0,1,0),

where

(3.19) G(0,1,0)=12[Ha3(r2)+Ha3(r2)2H(a3)]G(0,1,0)=\frac{1}{2}[H_{a_{3}}(r_{2})+H_{-a_{3}}(r_{2})-2H(a_{3})]

and

(3.20) F(0,1,0,0,1,0)=12[Hr2(γ0)+Hr2(γ1)+Hr2(δ0)+Hr2(δ1)4H(r2)].\begin{split}F(0,1,0,0,1,0)=&\frac{1}{2}[H_{r_{2}}(\gamma_{0})+H_{-r_{2}}(\gamma_{1})+H_{r_{2}}(\delta_{0})+H_{-r_{2}}(\delta_{1})-4H(r_{2})].\end{split}

In this case, the parameters are degenerated into (k=0,1)(k=0,1)

γk=[i3ci2+s22+T22+2(c2s2+(1)kc2T2+s2T2)]12;δk=[i3ci2+s22+T22+2(c2s2+(1)kc2T2s2T2)]12.\begin{split}\gamma_{k}&=[\sum_{i}^{3}c_{i}^{2}+s_{2}^{2}+T_{2}^{2}+2(c_{2}s_{2}+(-1)^{k}c_{2}T_{2}+s_{2}T_{2})]^{\frac{1}{2}};\\ \delta_{k}&=[\sum_{i}^{3}c_{i}^{2}+s_{2}^{2}+T_{2}^{2}+2(-c_{2}s_{2}+(-1)^{k}c_{2}T_{2}-s_{2}T_{2})]^{\frac{1}{2}}.\end{split}
Proof.

By definition, we have

(3.21) G+F=Ha3z3A(B+)+Ha3z3A(B)2H(a3z3A)+12(1+a3z3A)[Hα0(γ0)+Hα1(γ1)2Ha3z3A(α0α12)]+12(1a3z3A)[Hβ0(δ0)+Hβ1(δ1)2Ha3z3A(β0β12)].\begin{split}G+F&=H_{a_{3}z_{3}^{A}}(B_{+})+H_{-a_{3}z_{3}^{A}}(B_{-})-2H(a_{3}z_{3}^{A})\\ &+\frac{1}{2(1+a_{3}z_{3}^{A})}[H_{\alpha_{0}}(\gamma_{0})+H_{\alpha_{1}}(\gamma_{1})-2H_{a_{3}z_{3}^{A}}(\frac{\alpha_{0}-\alpha_{1}}{2})]\\ &+\frac{1}{2(1-a_{3}z_{3}^{A})}[H_{\beta_{0}}(\delta_{0})+H_{\beta_{1}}(\delta_{1})-2H_{-a_{3}z_{3}^{A}}(\frac{\beta_{0}-\beta_{1}}{2})].\end{split}

Note that FF is a function of six variables and the first three are exactly the variables of GG. Our strategy of locating the extremal points of G+FG+F is first finding the critical points z1A,z2A,z3Az_{1}^{A},z_{2}^{A},z_{3}^{A} of GG and verify that at those points the critical points of GG are attainable, then we can find the maximal points of F+GF+G.

For case 1: a3b3r30a_{3}b_{3}r_{3}\leq 0 and r32r2a3b3r3r_{3}^{2}-r^{2}\geq a_{3}b_{3}r_{3}, by [14] we know that maxG(z1A,z2A,z3A)=G(0,0,1)\max{G(z_{1}^{A},z_{2}^{A},z_{3}^{A})}=G(0,0,1), then the parameters in function FF are degenerated into (k=0,1k=0,1)

αk=a3+(1)k(b3z3B+r3z3B),βk=a3+(1)k(b3z3Br3z3B),γk={i3ci2+s32+T32(z3B)2+2[c3s3+(1)k(s3T3z3B+c3T3z3B)]}12,δk={i3ci2+s32+T32(z3B)2+2[c3s3+(1)k(s3T3z3Bc3T3z3B)]}12.\begin{split}\alpha_{k}&=a_{3}+(-1)^{k}(b_{3}z_{3}^{B}+r_{3}z_{3}^{B}),\\ \beta_{k}&=-a_{3}+(-1)^{k}(b_{3}z_{3}^{B}-r_{3}z_{3}^{B}),\\ \gamma_{k}&=\{\sum_{i}^{3}c_{i}^{2}+s_{3}^{2}+T_{3}^{2}(z^{B}_{3})^{2}+2[c_{3}s_{3}+(-1)^{k}(s_{3}T_{3}z^{B}_{3}+c_{3}T_{3}z^{B}_{3})]\}^{\frac{1}{2}},\\ \delta_{k}&=\{\sum_{i}^{3}c_{i}^{2}+s_{3}^{2}+T_{3}^{2}(z^{B}_{3})^{2}+2[-c_{3}s_{3}+(-1)^{k}(s_{3}T_{3}z^{B}_{3}-c_{3}T_{3}z^{B}_{3})]\}^{\frac{1}{2}}.\end{split}

Therefore, we have

(3.22) F(z1A,z2A,z3A,z1B,z2B,z3B)=F(0,0,1,z3B)=12(1+a3)[Hα0(γ0)+Hα1(γ1)]+12(1a3)[Hβ0(γ0)+Hβ1(γ1)].\begin{split}&F(z^{A}_{1},z^{A}_{2},z^{A}_{3},z^{B}_{1},z^{B}_{2},z^{B}_{3})=F(0,0,1,z^{B}_{3})\\ &=\frac{1}{2(1+a_{3})}[H_{\alpha_{0}}(\gamma_{0})+H_{\alpha_{1}}(\gamma_{1})]+\frac{1}{2(1-a_{3})}[H_{\beta_{0}}(\gamma_{0})+H_{\beta_{1}}(\gamma_{1})].\end{split}

When (b3+r3)(c3+s3)0(b_{3}+r_{3})(c_{3}+s_{3})\leq 0, it can be observed that FF is an even function for z3B[1,1]z_{3}^{B}\in[-1,1], so we just need to consider z3B[0,1]z_{3}^{B}\in[0,1]. The derivative of FF on z3Bz_{3}^{B} is given by

(3.23) Fz3B=14(1+a3){(b3+r3)log2(1+α1)2[(1+α0)2γ02](1+α0)2[(1+α1)2γ12]+c3T3s3T3+T32z3Bγ1log21+α1+γ11+α1γ1+c3T3+s3T3+T32z3Bγ0log21+α0+γ01+α0γ0}+14(1a3){(b3r3)log2(1+β1)2[(1+β0)2δ02](1+β0)2[(1+β1)2δ12]+c3T3s3T3+T32z3Bδ1log21+β1+δ11+β1δ1+c3T3+s3T3+T32z3Bδ0log21+β0+δ01+β0δ0};\begin{split}&\frac{\partial{F}}{\partial{z_{3}^{B}}}=\frac{1}{4(1+a_{3})}\{(b_{3}+r_{3})\log_{2}\frac{(1+\alpha_{1})^{2}[(1+\alpha_{0})^{2}-\gamma_{0}^{2}]}{(1+\alpha_{0})^{2}[(1+\alpha_{1})^{2}-\gamma_{1}^{2}]}\\ &+\frac{-c_{3}T_{3}-s_{3}T_{3}+T_{3}^{2}z_{3}^{B}}{\gamma_{1}}\log_{2}\frac{1+\alpha_{1}+\gamma_{1}}{1+\alpha_{1}-\gamma_{1}}+\frac{c_{3}T_{3}+s_{3}T_{3}+T_{3}^{2}z_{3}^{B}}{\gamma_{0}}\log_{2}\frac{1+\alpha_{0}+\gamma_{0}}{1+\alpha_{0}-\gamma_{0}}\}\\ &+\frac{1}{4(1-a_{3})}\{(b_{3}-r_{3})\log_{2}\frac{(1+\beta_{1})^{2}[(1+\beta_{0})^{2}-\delta_{0}^{2}]}{(1+\beta_{0})^{2}[(1+\beta_{1})^{2}-\delta_{1}^{2}]}\\ &+\frac{c_{3}T_{3}-s_{3}T_{3}+T_{3}^{2}z_{3}^{B}}{\delta_{1}}\log_{2}\frac{1+\beta_{1}+\delta_{1}}{1+\beta_{1}-\delta_{1}}\ +\frac{-c_{3}T_{3}+s_{3}T_{3}+T_{3}^{2}z_{3}^{B}}{\delta_{0}}\log_{2}\frac{1+\beta_{0}+\delta_{0}}{1+\beta_{0}-\delta_{0}}\};\end{split}

If b3+r30b_{3}+r_{3}\leq 0 and c3+s30c_{3}+s_{3}\geq 0, we have γ0γ1\gamma_{0}\geq\gamma_{1}, α1α0\alpha_{1}\geq\alpha_{0}, δ0δ1\delta_{0}\geq\delta_{1} and β1β0\beta_{1}\geq\beta_{0}, then

(3.24) (b3+r3)log2(1+α1)2[(1+α0)2γ02](1+α0)2[(1+α1)2γ12]0;(b3r3)log2(1+β1)2[(1+β0)2δ02](1+β0)2[(1+β1)2δ12]0;(b_{3}+r_{3})\log_{2}\frac{(1+\alpha_{1})^{2}[(1+\alpha_{0})^{2}-\gamma_{0}^{2}]}{(1+\alpha_{0})^{2}[(1+\alpha_{1})^{2}-\gamma_{1}^{2}]}\geq 0;(b_{3}-r_{3})\log_{2}\frac{(1+\beta_{1})^{2}[(1+\beta_{0})^{2}-\delta_{0}^{2}]}{(1+\beta_{0})^{2}[(1+\beta_{1})^{2}-\delta_{1}^{2}]}\geq 0;
(3.25) c3T3s3T3+T32z3Bγ1log21+α1+γ11+α1γ1+c3T3+s3T3+T32z3Bγ0log21+α0+γ01+α0γ0c3T3s3T3+T32z3Bγ0log21+α1+γ11+α1γ1+c3T3+s3T3+T32z3Bγ0log21+α1+γ11+α1γ1=2T32z3Bγ0log21+α1+γ11+α1γ10;\begin{split}&\frac{-c_{3}T_{3}-s_{3}T_{3}+T_{3}^{2}z_{3}^{B}}{\gamma_{1}}\log_{2}\frac{1+\alpha_{1}+\gamma_{1}}{1+\alpha_{1}-\gamma_{1}}+\frac{c_{3}T_{3}+s_{3}T_{3}+T_{3}^{2}z_{3}^{B}}{\gamma_{0}}\log_{2}\frac{1+\alpha_{0}+\gamma_{0}}{1+\alpha_{0}-\gamma_{0}}\\ \geq&\frac{-c_{3}T_{3}-s_{3}T_{3}+T_{3}^{2}z_{3}^{B}}{\gamma_{0}}\log_{2}\frac{1+\alpha_{1}+\gamma_{1}}{1+\alpha_{1}-\gamma_{1}}+\frac{c_{3}T_{3}+s_{3}T_{3}+T_{3}^{2}z_{3}^{B}}{\gamma_{0}}\log_{2}\frac{1+\alpha_{1}+\gamma_{1}}{1+\alpha_{1}-\gamma_{1}}\\ =&\frac{2T_{3}^{2}z^{B}_{3}}{\gamma_{0}}\log_{2}\frac{1+\alpha_{1}+\gamma_{1}}{1+\alpha_{1}-\gamma_{1}}\geq 0;\end{split}
(3.26) c3T3s3T3+T32z3Bδ1log21+β1+δ11+β1δ1+c3T3+s3T3+T32z3Bδ0log21+β0+δ01+β0δ0c3T3s3T3+T32z3Bδ0log21+β1+δ11+β1δ1+c3T3+s3T3+T32z3Bδ0log21+β1+δ11+β1δ1=2T32z3Bδ0log21+β1+δ11+β1δ10.\begin{split}&\frac{c_{3}T_{3}-s_{3}T_{3}+T_{3}^{2}z_{3}^{B}}{\delta_{1}}\log_{2}\frac{1+\beta_{1}+\delta_{1}}{1+\beta_{1}-\delta_{1}}+\frac{-c_{3}T_{3}+s_{3}T_{3}+T_{3}^{2}z_{3}^{B}}{\delta_{0}}\log_{2}\frac{1+\beta_{0}+\delta_{0}}{1+\beta_{0}-\delta_{0}}\\ \geq&\frac{c_{3}T_{3}-s_{3}T_{3}+T_{3}^{2}z_{3}^{B}}{\delta_{0}}\log_{2}\frac{1+\beta_{1}+\delta_{1}}{1+\beta_{1}-\delta_{1}}+\frac{-c_{3}T_{3}+s_{3}T_{3}+T_{3}^{2}z_{3}^{B}}{\delta_{0}}\log_{2}\frac{1+\beta_{1}+\delta_{1}}{1+\beta_{1}-\delta_{1}}\\ =&\frac{2T_{3}^{2}z^{B}_{3}}{\delta_{0}}\log_{2}\frac{1+\beta_{1}+\delta_{1}}{1+\beta_{1}-\delta_{1}}\geq 0.\end{split}

Hence in this case we get Fz3B0\frac{\partial{F}}{\partial{z^{B}_{3}}}\geq 0 when z3B[0,1]z_{3}^{B}\in[0,1].

If b3+r30b_{3}+r_{3}\geq 0 and c3+s30c_{3}+s_{3}\leq 0, we also can show that Fz3B0\frac{\partial F}{\partial z_{3}^{B}}\geq 0 similarly. So FF is a strictly monotonically increasing function with z3B[0,1]z^{B}_{3}\in[0,1]. Similarly we can check that FF is a strictly monotonically increasing function with respect to z1B[0,1]z^{B}_{1}\in[0,1] or z2B[0,1]z^{B}_{2}\in[0,1] in case 2. ∎

Theorem 3.3.

For the Werner-GHZ state ρw=c|ψψ|+(1c)I8\rho_{w}=c|\psi\rangle\langle\psi|+(1-c)\frac{I}{8}, where |ψ=|000+|1112|\psi\rangle=\frac{|000\rangle+|111\rangle}{2}, the quantum discord is

(3.27) 𝒬=18(1c)log2(1c)+1+7c8log2(1+7c)14(1+3c)log2(1+3c).\begin{split}\mathcal{Q}=\frac{1}{8}(1-c)log_{2}(1-c)+\frac{1+7c}{8}log_{2}(1+7c)-\frac{1}{4}(1+3c)log_{2}(1+3c).\end{split}
Proof.

Obviously, max{G(z1A,z2A,z3A)}=H(c).\max{\{G(z_{1}^{A},z_{2}^{A},z_{3}^{A})\}}=H(c). Let θ=cz3B\theta=cz_{3}^{B}, then

(3.28) F(z1A,z2A,z3A,z1B,z2B,z3B)=F(θ)=12[Hθ(|c+θ|)+Hθ(|cθ|)+Hθ(|c+θ|)+Hθ(|cθ|)]2H(θ).\begin{split}&F(z_{1}^{A},z_{2}^{A},z_{3}^{A},z_{1}^{B},z_{2}^{B},z_{3}^{B})=F(\theta)\\ =&\frac{1}{2}[H_{\theta}(|c+\theta|)+H_{\theta}(|c-\theta|)+H_{-\theta}(|c+\theta|)+H_{-\theta}(|c-\theta|)]-2H(\theta).\end{split}

It is easy to see that F(θ)F(\theta) is monotonically increasing with respect to θ[0,1]\theta\in[0,1]. So max{F(θ)}=F(max{θ})=F(c)\max{\{F(\theta)\}}=F(\max{\{\theta\}})=F(c). Fig.1 shows the behavior of the function 𝒬\mathcal{Q}.

Refer to caption
Figure 1. The behavior of the quantum discord 𝒬\mathcal{Q} for the Werner-GHZ state in Theorem 3.3.

Next, we consider the following general tripartite state

(3.29) ρ=18(I8+i3aiσiI4+I2i3biσiI2+I4i3ciσi+i3riσiσiI2+i3siσiI2σi+i3viI2σiσi+i3Tiσiσiσi).\begin{split}\rho&=\frac{1}{8}(I_{8}+\sum_{i}^{3}a_{i}\sigma_{i}\otimes I_{4}+I_{2}\otimes\sum_{i}^{3}b_{i}\sigma_{i}\otimes I_{2}+I_{4}\otimes\sum_{i}^{3}c_{i}\sigma_{i}\\ &+\sum_{i}^{3}r_{i}\sigma_{i}\otimes\sigma_{i}\otimes I_{2}+\sum_{i}^{3}s_{i}\sigma_{i}\otimes I_{2}\otimes\sigma_{i}\\ &+\sum_{i}^{3}v_{i}I_{2}\otimes\sigma_{i}\otimes\sigma_{i}+\sum_{i}^{3}T_{i}\sigma_{i}\otimes\sigma_{i}\otimes\sigma_{i}).\end{split}

Let a=i3ai2a=\sqrt{\sum_{i}^{3}a_{i}^{2}} and b=i3bi2b=\sqrt{\sum_{i}^{3}b_{i}^{2}}, then we can get the quantum discord for some special cases.

Theorem 3.4.

For the general tripartite state ρ\rho in Eq.(3.29), we have the following results:

Case 1: when ai=vi=Ti=0,r1=r2=r3=ra_{i}=v_{i}=T_{i}=0,r_{1}=r_{2}=r_{3}=r, we have that

(3.30) 𝒬(ρ)=i8λilog2λi+4+Hb(r)+Hb(r)H(|b+r|)12[Hb+B(r)+HbB(r)+Hb+B(r)+HbB(r)],\begin{split}\mathcal{Q}(\rho)=&\sum_{i}^{8}{\lambda_{i}\log_{2}\lambda_{i}}+4+H_{b}(r)+H_{-b}(r)-H(|b+r|)\\ -&\frac{1}{2}[H_{b+\mathrm{B}}(r)+H_{b-\mathrm{B}}(r)+H_{-b+\mathrm{B}}(r)+H_{-b-\mathrm{B}}(r)],\end{split}

where B=[i3(sibib+ci)2]12\mathrm{B}=[\sum_{i}^{3}(s_{i}\frac{b_{i}}{b}+c_{i})^{2}]^{\frac{1}{2}}.

Case 2: when bi=vi=Ti=0,r1=r2=r3=rb_{i}=v_{i}=T_{i}=0,r_{1}=r_{2}=r_{3}=r, we have that

(3.31) 𝒬(ρ)=i8λilog2λi+3H(a2)12[Ha(r)+Ha(r)2H(a)]12(1+a)[Ha+A(r)+HaA(r)2Ha(r)]12(1a)[Ha+A(r)+HaA(r)2Ha(r)],\begin{split}\mathcal{Q}(\rho)&=\sum_{i}^{8}{\lambda_{i}\log_{2}\lambda_{i}}+3-H(a^{2})-\frac{1}{2}[H_{a}(r)+H_{-a}(r)-2H(a)]\\ &-\frac{1}{2(1+a)}[H_{a+\mathrm{A}}(r)+H_{a-\mathrm{A}}(r)-2H_{a}(r)]\\ &-\frac{1}{2(1-a)}[H_{-a+\mathrm{A}}(r)+H_{-a-\mathrm{A}}(r)-2H_{-a}(r)],\end{split}

where A=[i3(siaia+ci)2]12\mathrm{A}=[\sum_{i}^{3}(s_{i}\frac{a_{i}}{a}+c_{i})^{2}]^{\frac{1}{2}}.

Case 3: when ri=Ti=vi=0r_{i}=T_{i}=v_{i}=0, we have that

(3.32) 𝒬(ρ)=i8λilog2λi+3H(a2)12[Hb(a)+Hb(a)2H(a)]12(1+a)[Ha+A(b)+HaA(b)2Ha(b)]12(1a)[Hb+A(b)+HaA(b)2Ha(b)],\begin{split}\mathcal{Q}(\rho)&=\sum_{i}^{8}{\lambda_{i}\log_{2}\lambda_{i}}+3-H(a^{2})-\frac{1}{2}[H_{b}(a)+H_{-b}(a)-2H(a)]\\ &-\frac{1}{2(1+a)}[H_{a+\mathrm{A}}(b)+H_{a-\mathrm{A}}(b)-2H_{a}(b)]\\ &-\frac{1}{2(1-a)}[H_{b+\mathrm{A}}(b)+H_{-a-\mathrm{A}}(b)-2H_{-a}(b)],\end{split}

where A=[i3(siaia+ci)2]12\mathrm{A}=[\sum_{i}^{3}(s_{i}\frac{a_{i}}{a}+c_{i})^{2}]^{\frac{1}{2}}.

Case 4: when ai=ci=si=Ti=0,r1=r2=r3=r,v1=v2=v3=va_{i}=c_{i}=s_{i}=T_{i}=0,r_{1}=r_{2}=r_{3}=r,v_{1}=v_{2}=v_{3}=v, we have that

(3.33) 𝒬(ρ)=i8λilog2λi+Hb(r)+Hb(r)+4H(|b+r|)12[Hb+v(r)+Hbv(r)+Hb+v(r)Hbv(r)].\begin{split}\mathcal{Q}(\rho)=&\sum_{i}^{8}{\lambda_{i}\log_{2}\lambda_{i}}+H_{b}(r)+H_{-b}(r)+4-H(|b+r|)\\ -&\frac{1}{2}[H_{b+v}(r)+H_{b-v}(r)+H_{-b+v}(r)H_{-b-v}(r)].\end{split}

Case 5: when ri=Ti=si=ci=0,v1=v2=v3=vr_{i}=T_{i}=s_{i}=c_{i}=0,v_{1}=v_{2}=v_{3}=v, we have that

(3.34) 𝒬(ρ)=i8λilog2λi+3H(a2)12[Hb(a)+Hb(a)2H(a)]12(1+a)[Ha+v(b)+Hav(b)2Ha(b)]12(1a)[Ha+v(b)+Hav(b)2Ha(b)].\begin{split}\mathcal{Q}(\rho)&=\sum_{i}^{8}{\lambda_{i}\log_{2}\lambda_{i}}+3-H(a^{2})-\frac{1}{2}[H_{b}(a)+H_{-b}(a)-2H(a)]\\ &-\frac{1}{2(1+a)}[H_{a+v}(b)+H_{a-v}(b)-2H_{a}(b)]\\ &-\frac{1}{2(1-a)}[H_{-a+v}(b)+H_{-a-v}(b)-2H_{-a}(b)].\end{split}

Case 6: when bi=si=ci=Ti=0,r1=r2=r3=r,v1=v2=v3=vb_{i}=s_{i}=c_{i}=T_{i}=0,r_{1}=r_{2}=r_{3}=r,v_{1}=v_{2}=v_{3}=v, we have that

(3.35) 𝒬(ρ)=i8λilog2λi+3H(a2)12[Ha(r)+Ha(r)2H(a)]12(1+a)[Ha+v(r)+Hav(r)2Ha(r)]12(1a)[Ha+v(r)+Hav(r)2Ha(r)].\begin{split}\mathcal{Q}(\rho)&=\sum_{i}^{8}{\lambda_{i}\log_{2}\lambda_{i}}+3-H(a^{2})-\frac{1}{2}[H_{a}(r)+H_{-a}(r)-2H(a)]\\ &-\frac{1}{2(1+a)}[H_{a+v}(r)+H_{a-v}(r)-2H_{a}(r)]\\ &-\frac{1}{2(1-a)}[H_{-a+v}(r)+H_{-a-v}(r)-2H_{-a}(r)].\end{split}
Proof.

All cases can be shown similarly. Let’s consider case 1: ai=vi=Ti=0,r1=r2=r3=ra_{i}=v_{i}=T_{i}=0,r_{1}=r_{2}=r_{3}=r, max{G(z1A,z2A,z3A)}=G(b1b,b2b,b3b)\max{\{G(z_{1}^{A},z_{2}^{A},z_{3}^{A})\}}=G(\frac{b_{1}}{b},\frac{b_{2}}{b},\frac{b_{3}}{b}). Let θ=i3rziBbib\theta=\sum_{i}^{3}rz_{i}^{B}\frac{b_{i}}{b}, then

(3.36) F(z1A,z2A,z3A,z1B,z2B,z3B)=F(b1b,b2b,b3b,θ)=12[Hb+B(θ)+HbB(θ)+Hb+B(θ)+HbB(θ)]Hb(θ)Hb(θ)2,\begin{split}&F(z_{1}^{A},z_{2}^{A},z_{3}^{A},z_{1}^{B},z_{2}^{B},z_{3}^{B})=F(\frac{b_{1}}{b},\frac{b_{2}}{b},\frac{b_{3}}{b},\theta)\\ =&\frac{1}{2}[H_{b+\mathrm{B}}(\theta)+H_{b-\mathrm{B}}(\theta)+H_{-b+\mathrm{B}}(\theta)+H_{-b-\mathrm{B}}(\theta)]-H_{b}(\theta)-H_{-b}(\theta)-2,\end{split}

where B=[i3(sibib+ci)2]12\mathrm{B}=[\sum_{i}^{3}(s_{i}\frac{b_{i}}{b}+c_{i})^{2}]^{\frac{1}{2}}.

The derivative of FF over θ\theta is equal to

(3.37) Fθ=14[log2(1+b+B+θ)(1+bB+θ)(1+bθ)2(1+b+Bθ)(1+bBθ)(1+b+θ)2+log2(1bB+θ)(1b+B+θ)(1bθ)2(1bBθ)(1b+Bθ)(1b+θ)2].\begin{split}\frac{\partial F}{\partial\theta}=&\frac{1}{4}[\log_{2}\frac{(1+b+\mathrm{B}+\theta)(1+b-\mathrm{B}+\theta)(1+b-\theta)^{2}}{(1+b+\mathrm{B}-\theta)(1+b-\mathrm{B}-\theta)(1+b+\theta)^{2}}\\ +&\log_{2}\frac{(1-b-\mathrm{B}+\theta)(1-b+\mathrm{B}+\theta)(1-b-\theta)^{2}}{(1-b-\mathrm{B}-\theta)(1-b+\mathrm{B}-\theta)(1-b+\theta)^{2}}].\end{split}

Obviously, Fθ0\frac{\partial F}{\partial\theta}\geq 0 when θ[0,1]\theta\in[0,1]. Then F(θ)F(\theta) is a strictly increasing function and maxF(θ)=F(max{θ})\max{F(\theta)}=F(\max\{\theta\}).

Let Y=θ+μ[1(z1B)2(z2B)2(z3B)2]Y=\theta+\mu[1-(z_{1}^{B})^{2}-(z_{2}^{B})^{2}-(z_{3}^{B})^{2}], Yz1B=rb1b2μz1B\frac{\partial Y}{\partial z_{1}^{B}}=r\frac{b_{1}}{b}-2\mu z_{1}^{B}, Yz2B=rb2b2μz2B\frac{\partial Y}{\partial z_{2}^{B}}=r\frac{b_{2}}{b}-2\mu z_{2}^{B}, Yz3B=rb3b2μz3B\frac{\partial Y}{\partial z_{3}^{B}}=r\frac{b_{3}}{b}-2\mu z_{3}^{B}, Yμ=1(z1B)2(z2B)2(z3B)2\frac{\partial Y}{\partial\mu}=1-(z_{1}^{B})^{2}-(z_{2}^{B})^{2}-(z_{3}^{B})^{2}. Imposing Yz1B=0,Yz2B=0,Yz3B=0,Yμ=0\frac{\partial Y}{\partial z_{1}^{B}}=0,\frac{\partial Y}{\partial z_{2}^{B}}=0,\frac{\partial Y}{\partial z_{3}^{B}}=0,\frac{\partial Y}{\partial\mu}=0, we have ziB=bibz_{i}^{B}=\frac{b_{i}}{b}. So max{θ}=r\max\{\theta\}=r and maxF(θ)=F(r)\max{F(\theta)}=F(r), then case 1 is shown. ∎

Example 1. For a state in Eq.(3.1), when a1=0,a2=0,a3=0.03,b1=0,b2=0,b3=0.25,c1=0.12,c2=0.12,c3=0.01,r1=0.1,r2=0.1,r3=0.3,s1=0.13,s2=0.13,s3=0.26,v1=0,v2=0,v3=0,T1=0.02,T2=0.02,T3=0.36a_{1}=0,a_{2}=0,a_{3}=0.03,b_{1}=0,b_{2}=0,b_{3}=0.25,c_{1}=0.12,c_{2}=0.12,c_{3}=0.01,r_{1}=0.1,r_{2}=0.1,r_{3}=-0.3,s_{1}=0.13,s_{2}=0.13,s_{3}=-0.26,v_{1}=0,v_{2}=0,v_{3}=0,T_{1}=-0.02,T_{2}=-0.02,T_{3}=-0.36. According to the case 1 of Theorem 3.2, we have 𝒬=0.8889\mathcal{Q}=0.8889. Fig. 2 shows the behavior of the quantum discord 𝒬\mathcal{Q}.

Refer to caption
Figure 2. The behavior of the quantum discord 𝒬\mathcal{Q} with respect to the parameters in Example 1. In this case, z1A=z2A=z1B=z2B=0z_{1}^{A}=z_{2}^{A}=z_{1}^{B}=z_{2}^{B}=0, the quantum discord 𝒬\mathcal{Q} is only related to variables z3A,z3Bz_{3}^{A},z_{3}^{B}. Then 𝒬=0.8889\mathcal{Q}=0.8889.

Example 2. For a state of the case 1 in Theorem 3.4, when a1=a2=a3=0,b1=0.2,b2=0.05,b3=0.1,c1=0.04,c2=0.06,c3=0.11,r1=r2=r3=0.17,s1=0.08,s2=0.15,s3=0.25,v1=v2=v3=T1=T2=T3=0a_{1}=a_{2}=a_{3}=0,b_{1}=0.2,b_{2}=0.05,b_{3}=0.1,c_{1}=0.04,c_{2}=0.06,c_{3}=0.11,r_{1}=r_{2}=r_{3}=0.17,s_{1}=0.08,s_{2}=0.15,s_{3}=0.25,v_{1}=v_{2}=v_{3}=T_{1}=T_{2}=T_{3}=0. Then the quantum discord is 𝒬=0.9970\mathcal{Q}=0.9970. Fig. 3 and Fig. 4 show the behavior of the function GG and FF respectively.

Refer to caption
Figure 3. The behavior of G(z1A,z2A,z3A)G(z_{1}^{A},z_{2}^{A},z_{3}^{A}) with the variables z1A,z2A,z3Az_{1}^{A},z_{2}^{A},z_{3}^{A} in Example 2, where zA=(z1A,z2A,z3A)z^{A}=(z_{1}^{A},z_{2}^{A},z_{3}^{A}) is on a unit sphere. This is a four-dimensional figure. Among them, the intensity of light is used to indicate the magnitude of the GG value. The brighter the point, the greater the value of GG and maxG=0.1182\max G=0.1182.
Refer to caption
Figure 4. The behavior of F(0.8729,0.2182,0.4364,z1B,z2B,z3B)F(0.8729,0.2182,0.4364,z_{1}^{B},z_{2}^{B},z_{3}^{B}) with the variables z1B,z2B,z3Bz_{1}^{B},z_{2}^{B},z_{3}^{B} in Example 2, where zB=(z1B,z2B,z3B)z^{B}=(z_{1}^{B},z_{2}^{B},z_{3}^{B}) is on a unit sphere. The brighter the point, the greater the value of FF and maxF=0.1107\max F=0.1107.

4. Conclusions

Quantum discord is one of the important correlations in studying quantum systems. It is well-known that the quantum discord is hard to compute explicitly, and only sporadic formulas are known, for instance, the Bell state and the X-state etc. Recently important progresses are made to generalize the notion to multipartite quantum systems [10], and their explicit formulas are expectedly not easy to find. In this work, we have found explicit formulas of the quantum discord for tripartite non X-states with 14 parameters, including some famous states such as the Werner-GHZ state.

Acknowledgments

The research is supported in part by the NSFC grants 11871325 and 12126351, and Natural Science Foundation of Hubei Province grant no. 2020CFB538 as well as Simons Foundation grant no. 523868.

References

  • [1] Datta, A., Shaji, A., Caves, C. M.: Quantum discord and the power of one qubit, Phys. Rev. Lett., 2008, 100:050502.
  • [2] Cui, J., Fan, H.: Correlations in the Grover search, J. Phys. A:Math. Theor., 2010, 43(4): 045305.
  • [3] Ollivier, H., Zurek, W. H.: Quantum discord: a measure of the quantumness of correlations, Phys. Rev. Lett., 2001, 88(1): 017901.
  • [4] Luo, S.: Quantum discord for two-qubit systems, Phys. Rev. A, 2008, 77(4): 042303.
  • [5] Rulli, C. C., Sarandy, M. S.: Global quantum discord in multipartite systems, Phys. Rev. A, 2011, 84: 042109.
  • [6] Okrasa, M., Walczak, Z.: Quantum discord and multipartite correlations, Euro. Phys. Lett., 2011, 96:60003.
  • [7] Giorgi, G. L., Bellomo, B., Galve, F., et al.: Genuine quantum and classical correlations in multipartite systems, Phys. Rev. Lett., 2011, 107:190501.
  • [8] Chakrabarty, I., Agrawal, P., Pati, A. K.: Quantum dissension: Generalizing quantum discord for three-qubit states, Euro. Phys. J. D, 2011, 65(3):605-612.
  • [9] Modi, K., Paterek, T., Son, W., et al.: Unified view of quantum and classical correlations, Phys. Rev. Lett., 2010, 104:080501.
  • [10] Radhakrishnan, C., Laurière, M., Byrnes, T.: Multipartite generalization of quantum discord, Phys. Rev. Lett., 2020, 124: 110401.
  • [11] Jing, N., Zhang, X., Wang, Y.-K.: Comment on “One-way deficit of two qubit X states”, Quant. Inf. Process, 2015, 14:4511–4521
  • [12] Streltsov, A. Quantum discord and its role in quantum information theory, In: Quantum Correlations beyond Entanglement, Springer Briefs in Phys. (New York: Springer), 2015, pp 22-43.
  • [13] Ye, B., Fei, S.-M.: A note on one-way quantum deficit and quantum discord, Quant. Inf. Process, 2016, 15:279.
  • [14] Jing, N., Yu, B.: Quantum discord of X-states as optimization of a one variable function, J. Phys. A:Math. Theor., 2016, 49(38): 385302.
  • [15] Wilde, M. M.: Quantum information theory, Cambridge University Press, Cambridge, 2017, 2nd. ed.
  • [16] Vedral, V. The role of relative entropy in quantum information theory, Rev. Mod. Phys., 2002, 74(1): 197.
  • [17] Groisman, B., Popescu, S., Winter, A.: Quantum, classical, and total amount of correlations in a quantum state, Phys. Rev. A, 2005, 72(3): 032317.
  • [18] Schumacher, B., Westmoreland, M. D.: Quantum mutual information and the one-time pad, Phys. Rev. A, 2006, 74(4): 042305.