1: %\documentclass[twoside,twocolumn]{IEEEtran}
2:
3: \documentclass[journal]{IEEEtran}
4:
5: \usepackage[cmex10]{amsmath}
6: \interdisplaylinepenalty=2500
7: \usepackage{array}
8: \usepackage{url}
9: \usepackage{amsfonts}
10: \usepackage{amssymb}
11: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
12: %\usepackage{amssymb}
13: \usepackage{graphicx}
14:
15: %\setlength\topmargin{-0.3in} \setlength\oddsidemargin {5mm}
16: %\setlength\evensidemargin {5mm} \setlength\textheight {235mm}
17: %\setlength\textwidth {156.0mm} \setlength\columnsep {4.1mm}
18:
19:
20:
21: %\usepackage{amsmath}
22:
23:
24: \begin{document}
25:
26: \newtheorem{corollary}{Corollary}
27: \newtheorem{definition}{Definition}
28: \newtheorem{example}{Example}
29: \newtheorem{lemma}{Lemma}
30: \newtheorem{proposition}{Proposition}
31: \newtheorem{theorem}{Theorem}
32: \newtheorem{fact}{Fact}
33: \newtheorem{property}{Property}
34: \newcommand{\bra}[1]{\langle #1|}
35: \newcommand{\ket}[1]{|#1\rangle}
36: \newcommand{\braket}[3]{\langle #1|#2|#3\rangle}
37: %%inner product
38: \newcommand{\ip}[2]{\langle #1|#2\rangle}
39: %%outer product
40: \newcommand{\op}[2]{|#1\rangle \langle #2|}
41:
42: \newcommand{\tr}{{\rm tr}}
43: \newcommand{\supp}{{\it supp}}
44: \newcommand{\sch}{{\it Sch}}
45:
46: %\newcommand{\mod}{{\rm mod}}
47: \newcommand {\E } {{\mathcal{E}}}
48: \newcommand {\F } {{\mathcal{F}}}
49: \newcommand {\diag } {{\rm diag}}
50:
51: \title{Distinguishability of Quantum States by Separable Operations\thanks{This work was partly supported
52: by the Natural Science Foundation of China (Grant Nos. 60621062 and
53: 60503001) and the Hi-Tech Research and Development Program of China
54: (863 project) (Grant No. 2006AA01Z102).}}
55: \author{Runyao Duan,\ \ Yuan Feng,\ \,\ \ Yu Xin, \ \ and Mingsheng Ying\thanks{The material in this paper was presented in part as a long talk
56: at the 2007 Asia Conference of Quantum Information Science (AQIS07),
57: Kyoto, Japan. The authors are with the State Key Laboratory of
58: Intelligent Technology and Systems, Tsinghua National Laboratory for
59: Information Science and Technology, Department of Computer Science
60: and Technology, Tsinghua University, Beijing 100084, China. Yu Xin
61: is also with the Department of Physics, Tsinghua University, Beijing
62: 100084, China. E-mails: dry@tsinghua.edu.cn (Runyao Duan),
63: feng-y@tsinghua.edu.cn (Yuan Feng), xiny05@mails.tsinghua.edu.cn (Yu
64: Xin), and yingmsh@tsinghua.edu.cn (Mingsheng Ying)}}
65: \date{\today}
66: \maketitle
67:
68: \begin{abstract}
69: We study the distinguishability of multipartite quantum states by
70: separable operations. We first present a necessary and sufficient
71: condition for a finite set of orthogonal quantum states to be
72: distinguishable by separable operations. An analytical version of
73: this condition is derived for the case of $(D-1)$ pure states, where
74: $D$ is the total dimension of the state space under consideration. A
75: number of interesting consequences of this result are then carefully
76: investigated. Remarkably, we show there exists a large class of
77: $2\otimes 2$ separable operations not being realizable by local
78: operations and classical communication. Before our work only a class
79: of $3\otimes 3$ nonlocal separable operations was known [Bennett et
80: al, Phys. Rev. A \textbf{59}, 1070 (1999)]. We also show that any
81: basis of the orthogonal complement of a multipartite pure state is
82: indistinguishable by separable operations if and only if this state
83: cannot be a superposition of $1$ or $2$ orthogonal product states,
84: i.e., has an orthogonal Schmidt number not less than $3$, thus
85: generalize the recent work about indistinguishable bipartite
86: subspaces [Watrous, Phys. Rev. Lett. \textbf{95}, 080505 (2005)].
87: Notably, we obtain an explicit construction of indistinguishable
88: subspaces of dimension $7$ (or $6$) by considering a composite
89: quantum system consisting of two qutrits (resp. three qubits), which
90: is slightly better than the previously known indistinguishable
91: bipartite subspace with dimension $8$.
92:
93: \smallskip\
94: {\it Index Terms} --- Quantum Nonlocality, Local distinguishability,
95: Separable operations, Orthogonal Schmidt number, Unextendible
96: Product Bases.
97: \end{abstract}
98:
99: \section{Introduction}
100: One of the most profound features of quantum mechanics is that
101: composite quantum systems can exhibit nonlocality. Such an effect
102: can be interpreted as there exist some global quantum operations
103: performing on a composite system cannot be implemented by the owners
104: of the subsystems using local operations and classical communication
105: (LOCC) only. Actually, it is well known that any locally realizable
106: quantum operation is necessarily separable. The converse part,
107: however, is not always true, as a consequence of the weird
108: phenomenon of ``nonlocality without entanglement" discovered by
109: Bennet and coworkers \cite{BDF+99}. On the one hand, although many
110: partial progresses have been made, the structure of LOCC operations
111: is far from well understood. On the other hand, the class of
112: separable operations is rather restricted and is with rich
113: mathematical structure. It is relatively easier to determine whether
114: a given quantum operation is separable by employing the well
115: developed tools for the separability of quantum states. Thus, a deep
116: understanding of separable operations is of particular importance in
117: quantum information theory.
118:
119: A general strategy for studying quantum nonlocality is to consider
120: what kind of information processing tasks can be achieved by LOCC.
121: Roughly speaking, if a certain task is accomplished with different
122: optimal global and local efficiencies, then we can construct a class
123: of quantum operations that cannot be realized by LOCC. Among these
124: tasks, discrimination of orthogonal quantum states is a very
125: effective one and has been received considerable attentions in
126: recent years. Many interesting results have been reported, see Refs.
127: \cite{BDF+99,BDM+99,DMS+00,WSHV00,VSPM01, CY02,
128: HM03,JCY05,ABB+05,OGA06,WH02, TDL01, DLT02,
129: EW02,CHE04,BW06,NAT05,OH06, GKR+01,FAN04,WAT05,HK05,
130: HMM+06,DFJY07,HSSH03,CL03,COH07, FS07} and references therein for a
131: partial list. There exist sets of orthogonal product states that
132: cannot be discriminated perfectly by LOCC \cite{BDF+99,BDM+99,
133: DMS+00}; a perfect discrimination between two multipartite
134: orthogonal quantum states can always be achieved locally
135: \cite{WSHV00}, and this result can be generalized to the case when
136: distinguishing two nonorthogonal states or two infinite dimensional
137: orthogonal states \cite{VSPM01,CY02,HM03,JCY05,ABB+05,OGA06}; a
138: complete characterization for the local distinguishability of
139: $2\otimes 2$ states also has been obtained \cite{WH02}. Despite of
140: these significant advances, determining the local distinguishability
141: of a finite set of orthogonal states on a multipartite state space
142: still is a formidable task. Actually, even for an arbitrary set of
143: multipartite orthogonal product states we still don't know how to
144: decide the local distinguishability analytically except for the
145: special cases of $2\otimes n$ and $3\otimes 3$, which were solved in
146: Ref. \cite{BDM+99} and Ref. \cite{FS07}, respectively.
147:
148:
149: Very recently separable operations have been widely used in order to
150: obtain useful criteria for local distinguishability. In particular,
151: Chefles employed separable operations as a tool to obtain a
152: necessary and sufficient condition for a set of general quantum
153: states to be probabilistically distinguishable by LOCC \cite{CHE04},
154: see also Ref. \cite{BW06}. Nathanson showed that the number of
155: $d\otimes d$ maximally entangled states distinguishable by separable
156: operations is at most $d$ \cite{NAT05} and the same result was
157: independently obtained by Owari and Hayashi using a slightly
158: different method \cite{OH06}, which extensively generalized the
159: results obtained by Ghosh {\it et al.} \cite{GKR+01} and by Fan
160: \cite{FAN04}. Watrous found a bipartite subspace with interesting
161: property that no bases distinguishable by separable operations, and
162: employed it to explicitly construct a class of quantum channels with
163: suboptimal environment-assisted capacity \cite{WAT05}, which settled
164: an open problem suggested by Hayden and King \cite{HK05}. Hayashi
165: {\it et al.} obtained a connection between the number of
166: distinguishable states by separable operations and the average
167: entanglement degree of the quantum states to be discriminated
168: \cite{HMM+06}. In Ref. \cite{DFJY07} we studied the local
169: distinguishability of an arbitrary basis of a multipartite state
170: space and provided a universal tight lower bound on the number of
171: locally unambiguously distinguishable members in an arbitrary basis.
172: All these results suggest the following question:``What kind of
173: quantum states can be perfectly distinguishable by separable
174: operations?" The answer to this question will not only lead to a
175: better understanding of the nature of separable operations, but also
176: accelerate the present research of quantum nonlocality.
177:
178: The purpose of this paper is to study the strength and the
179: limitation of separable operations by considering the
180: distinguishability of quantum states under separable operations. We
181: present a necessary and sufficient condition for the
182: distinguishability of a set of general multipartite quantum states.
183: Assisting with this condition, we completely solve the problem of
184: distinguishing $(D-1)$ multipartite pure states by separable
185: operations, where $D$ is the dimension of the multipartite state
186: space under consideration. Two consequences are of particular
187: interests. First we show that there exists a large class of
188: separable operations performing on a $2\otimes 2$ quantum system
189: cannot be realized by LOCC. This is somewhat surprising as it
190: indicates that the capability of separable operations is much more
191: powerful than that of LOCC operations even when only the simplest
192: composite quantum system is under consideration, which settles an
193: open problem recently posed by Gheorghiu and Griffiths \cite{GG07}
194: and highlights the nonlocal nature of separable operations. It is
195: also worth noting that before our work only nonlocal separable
196: operations acting on $3\otimes 3$ systems are known \cite{BDF+99}.
197:
198: Second we show that any basis of the orthogonal complement of a
199: multipartite pure state is indistinguishable by separable operations
200: if and only if this state cannot be a superposition of $1$ or $2$
201: orthogonal product states, i.e., has an orthogonal Schmidt number
202: not less than $3$, thus provide a generalization of the result by
203: Watrous \cite{WAT05}, who proved that any basis of the orthogonal
204: complement of a $d\otimes d$ maximally entangled state is
205: indistinguishable by separable operations when $d\geq 3$.
206: Furthermore, we also present an explicit construction of
207: indistinguishable bipartite (or tripartite) subspaces of dimension
208: $7$ (resp. $6$), which reduces the minimal dimension of
209: indistinguishable bipartite subspace from $8$ (given by Watrous in
210: Ref. \cite{WAT05}) to $7$.
211:
212: Throughout this paper we consider a multipartite quantum system
213: consisting of $K$ parts, say $A_1,\cdots, A_K$. We assume part $A_k$
214: has a state space $\mathcal{H}_k$ with dimension $d_k$. The whole
215: state space is given by $\mathcal{H}=\otimes_{k=1}^K\mathcal{H}_k$
216: with total dimension $D=d_1\cdots d_K$. Sometimes we use the
217: notation $d_1\otimes \cdots \otimes d_K$ for $\mathcal{H}$. With
218: these assumptions, the rest of the paper is organized as follows. In
219: Section II we give some necessary preliminaries, including a brief
220: review of the notion of separable operations, the concept of
221: generalized Schmidt number of a quantum state, and some technical
222: lemmas. In Section III we present a necessary and sufficient
223: condition for the distinguishability of a set of orthogonal quantum
224: states on $\mathcal{H}$ by separable operations (Theorem
225: \ref{exactdis}). Many immediate but interesting consequences of this
226: condition are discussed in details. Sections IV and V devote to the
227: following specific problem: Let $\ket{\Phi}$ be a pure state on
228: $\mathcal{H}$ and let
229: $\mathcal{B}=\{\ket{\Psi_1},\cdots,\ket{\Psi_{D-1}}\}$ be an
230: orthonormal basis of $\{\ket{\Phi}\}^\perp$(i.e., the orthogonal
231: complement of $\ket{\Phi}$), determine the distinguishability of
232: $\mathcal{B}$ by separable operations. We show this problem can be
233: solved analytically. Two special cases are most notable. The first
234: case is concerned with two qubits, i.e., $K=2$ and $d_1=d_2=2$. We
235: find a necessary condition for the distinguishability of
236: $\{\ket{\Psi_1},\ket{\Psi_2},\ket{\Psi_3}\}$ by separable operations
237: is that the summation of the concurrences of $\ket{\Psi_k}$ is equal
238: to that of $\ket{\Phi}$ (Theorem \ref{2otimes2sep}). When
239: $\ket{\Phi}$ is maximally entangled, such a condition is also
240: sufficient (Corollary \ref{2ME}). The second case of interest is
241: that any basis $\mathcal{B}$ of $\{\ket{\Phi}\}^\perp$ is
242: indistinguishable by separable operations if and only if the
243: orthogonal Schmidt number of $\ket{\Phi}$ (the least number of
244: orthogonal product states that can linearly express $\ket{\Phi}$) is
245: not less than $3$ (Theorem \ref{indissubspace}). Section VI is of
246: independent interest. We give an explicit construction of an
247: indistinguishable subspace of dimension $7$ (or $6$) when
248: considering a composite quantum system consisting of two qutrits
249: (resp. three qubits). We conclude the paper in Section VII, and put
250: some additional proofs in Appendix.
251:
252: \section{Preliminaries} A general quantum state $\rho$ is
253: characterized by its density operator, which is a positive
254: semi-definite operator with trace one on $\mathcal{H}$. The density
255: operator of a pure state $\ket{\psi}$ is simply the projector
256: $\op{\psi}{\psi}$. The support of $\rho$, denoted by $\supp(\rho)$,
257: is the vector space spanned by the eigenvectors of $\rho$ with
258: positive eigenvalues. Alternatively, suppose $\rho$ can be
259: decomposed into a convex combination of $\op{\psi_k}{\psi_k}$, say,
260: \begin{equation}\label{essemble}
261: \rho=\sum_{k=1}^n p_k\op{\psi_k}{\psi_k},
262: \end{equation}
263: where $0<p_k\leq 1$ and $\sum_{k=1}^n p_k=1$. Then
264: $\supp(\rho)=span\{\ket{\psi_k}:1\leq k\leq n\}$. This fact will be
265: extensively used without being explicitly stated. In particular,
266: $\rho$ is said to be separable if $\ket{\psi_k}$ can be chosen as
267: product states.
268:
269: A nonzero positive semi-definite operator $E$ on $\mathcal{H}$ is
270: said to be separable if the normalized form ${E}/{\tr(E)}$ is a
271: separable quantum mixed state. A separable positive operator-valued
272: measure (POVM) is a collection of semi-definite positive operators
273: $\Pi=\{\Pi_1,\cdots, \Pi_n\}$ such that $\sum_{k=1}^n
274: \Pi_k=I_{\mathcal{H}}$ and $\Pi_k$ is separable for each $1\leq
275: k\leq n$, where $I_{\mathcal{H}}$ is the identity operator on
276: $\mathcal{H}$. Now we are ready to introduce the notion of separable
277: operation. Intuitively, a separable operation is a trace-preserving
278: completely positive linear map which sends separable states to
279: separable ones. For completeness, we state a formal definition as
280: follows.
281: \begin{definition}\label{so-def}\upshape
282: A separable operation is a quantum operation with product Kraus
283: operators. More precisely, a separable operation $\E$ acting on a
284: multipartite state space $\mathcal{H}=\otimes_{j=1}^K\mathcal{H}_j$
285: should be of the following form:
286: \begin{equation}
287: \E(\rho)=\sum_{k=1}^N E_k\rho E_k^\dagger,
288: \end{equation}
289: where $E_k=\otimes_{j=1}^K E_{kj}$ satisfies $\sum_{k=1}^N
290: E_{k}^\dagger E_k=I$, and $E_{kj}$ is linear operator on
291: $\mathcal{H}_j$.
292: \end{definition}
293: The set of separable operations is a rather restricted class of
294: quantum operations. It is not difficult to show that any LOCC
295: operation is separable, see Ref. \cite{BDF+99} for a detailed
296: discussion. However, the converse part is not true as there exists
297: some separable operation acting on $3\otimes 3$ cannot be
298: implemented locally \cite{BDF+99}.
299:
300: Let $\mathcal{S}=\{\rho_1,\cdots, \rho_n\}$ be a collection of $n$
301: quantum states. We say that $\mathcal{S}$ is perfectly
302: distinguishable by separable measurements if there is a separable
303: POVM $\Pi$ such that $\tr(\Pi_k\rho_j)=\delta_{k,j}$ for any $1\leq
304: k,j\leq n$. Interestingly, if a set of quantum states
305: $\{\rho_1,\cdots, \rho_n\}$ can be perfectly distinguishable by a
306: separable POVM $\Pi$, then we can transform this set of quantum
307: states into another set of LOCC distinguishable states by applying a
308: suitable separable operation $\E$ on the quantum system. More
309: precisely, let $\{\sigma_1,\cdots, \sigma_n\}$ be any set of
310: separable quantum states that can be perfectly distinguishable by
311: LOCC, then the mentioned separable operation can be constructed as
312: follows:
313: \begin{equation}\label{separableoperation}
314: \mathcal{E}(\rho)=\sum_{k=1}^n \tr(\Pi_k\rho)\sigma_k.
315: \end{equation}
316: One can easily verify that $\mathcal{E}(\rho_k)=\sigma_k$. The
317: function of $\mathcal{E}$ can be intuitively understood as
318: performing a separable POVM $\{\Pi_k\}$ on $\rho$ and then preparing
319: a separable state $\sigma_k$ according to the outcome $k$, finally
320: forgetting the classical information. Clearly, $\mathcal{E}$ is a
321: separable quantum operation and can be used to perfectly
322: discriminate the given states. Due to the above reason, when a set
323: of states are perfectly distinguishable by a separable POVM, we
324: directly say they are perfectly distinguishable by separable
325: operations.
326:
327: The rest of this section is to present some useful notations and
328: technical lemmas.
329: \begin{lemma}\label{tech}\upshape
330: Let $E$ be a positive semi-definite operator such that $0\leq E\leq
331: I_{\mathcal{H}}$, and let $\rho$ be a density matrix on
332: $\mathcal{H}$. Then $\tr(E\rho)=1$ if and only if $E-P\geq 0$, where
333: $P$ is the projector on the support of $\rho$.
334: \end{lemma}
335:
336: \textit{Proof.} We first show that if $E-P\geq 0$ and $0\leq E\leq
337: I$, then $\tr(E\rho)=1$. To see this, let $\ket{\psi}$ be any
338: normalized vector in $\supp(\rho)$. Then $\braket{\psi}{P}{\psi}=1$.
339: It follows from $E-P\geq 0$ that $\braket{\psi}{E}{\psi}\geq 1$. On
340: the other hand, $E\leq I$ implies $\braket{\psi}{E}{\psi}\leq 1$. So
341: for any $\ket{\psi}\in \supp(\rho)$ we have
342: $$\tr(E\op{\psi}{\psi})=\braket{\psi}{E}{\psi}=1.$$
343: Noticing that $\rho$ can be decomposed into $\sum_{k=1}^n p_k
344: \op{\psi_k}{\psi_k}$ such that $\ket{\psi_k}\in \supp(\rho)$ and
345: $\sum_{k=1}^n p_k=1$, we have
346: $$\tr(E\rho)=\sum_{k=1}^n p_k\tr(E\op{\psi_k}{\psi_k})=1.$$
347:
348: Conversely, $\tr(E\rho)=1$ and $\braket{\psi}{E}{\psi}\leq 1$ imply
349: that $\tr(E\op{\psi}{\psi})=1$ for any normalized vector
350: $\ket{\psi}\in \supp(\rho)$. So $E\ket{\psi}=\ket{\psi}$ or
351: $E\op{\psi}{\psi}=\op{\psi}{\psi}$. Noticing that $P=\sum_{k=1}^n
352: \op{\psi_k}{\psi_k}$ such that $\{\ket{\psi_k}\}$ is an orthonormal
353: basis for $\supp(\rho)$, then $EP=P$ and $PE=P$. Let $Q=I-P$. Then
354: $$E=(P+Q)E(P+Q)=P+QEQ,$$
355: which implies $E-P=QEQ\geq 0$.\hfill $\blacksquare$
356:
357: We also need the following concept of Schmidt number of a quantum
358: state.
359:
360: \begin{definition}\label{sch-def}\upshape
361: The {\it Schmidt number} of a multipartite quantum state $\rho$,
362: denoted by $\sch(\rho)$, is the minimal number of product pure
363: states whose linear span contains $\supp(\rho)$. The {\it orthogonal
364: Schmidt number} of $\rho$, denoted by $\sch_{\perp}(\rho)$, can be
365: defined similarly with the additional requirement that these product
366: pure states should be mutually orthogonal.
367: \end{definition}
368:
369: For pure state $\ket{\Psi}$ the (orthogonal) Schmidt number is
370: exactly reduced to the minimal number of (orthogonal) pure product
371: states needed to express $\ket{\Psi}$ as a superposition and this is
372: just the definition given by Eisert and Briegel \cite{EB01}. It is
373: difficult to determine the Schmidt number of a general multipartite
374: state. Only for bipartite pure states and some special multi-qubit
375: states the Schmidt number can be analytically determined \cite{EB01,
376: AACJ+00}. We should point out that the definition of Schmidt number
377: given here does not coincide with the one given by Terhal and
378: Horodecki \cite{TH00}.
379:
380: By definition, we have the following useful fact:
381: \begin{equation}\label{lowerbound}
382: \sch(\rho)\geq \max\{\sch(\Psi): \ket{\Psi}\in \supp(\rho)\}.
383: \end{equation}
384: This inequality can be used to provide a lower bound for
385: $\sch(\rho)$.
386:
387: It is interesting that the Schmidt number provides a simple
388: criterion for the separability. In fact, when $\rho$ is separable,
389: $supp(\rho)$ can be spanned by a set of product states. Thus we
390: have:
391: \begin{lemma}\label{schmidtsep}\upshape
392: For any quantum state $\rho$, $R(\rho)\leq \sch(\rho)$, where
393: $R(\rho)$ is the rank of $\rho$. The equality holds if $\rho$ is
394: separable.
395: \end{lemma}
396:
397: A typical use of Lemma \ref{schmidtsep} is to show a quantum state
398: $\rho$ is entangled. Usually this can be achieved by finding a
399: vector $\ket{\Psi}$ in $\supp(\rho)$ such that $\sch(\Psi)>R(\rho)$.
400:
401: The Schmidt decomposition of a bipartite pure state is not unique.
402: However, when multipartite pure states with Schmidt number $2$ are
403: under consideration, we do have a unique decomposition. Let
404: $\ket{a}=\otimes_{k=1}^K\ket{a_k}$ and
405: $\ket{b}=\otimes_{k=1}^K\ket{b_k}$ be product vectors on
406: $\mathcal{H}=\otimes_{k=1}^K\mathcal{H}_k$. Then $\ket{a}$ and
407: $\ket{b}$ are said to be different at the $k$-th entry if
408: $\ket{a_k}\neq z\ket{b_k}$ for any complex number $z$. Let
409: $h(\ket{a},\ket{b})$ be the total number of different entries
410: between $\ket{a}$ and $\ket{b}$. Then we have the following
411: interesting result.
412: \begin{lemma}\label{uniquedecom}\upshape
413: Let $\ket{\Phi}=\ket{a}+\ket{b}$ be an entangled state of
414: $\mathcal{H}$ such that $h(\ket{a},\ket{b})\geq 3$. Then there
415: cannot exist product vectors $\ket{c}$ and $\ket{d}$ such that
416: $\ket{\Phi}=\ket{c}+\ket{d}$ and $\{\ket{a},\ket{b}\}\neq
417: \{\ket{c},\ket{d}\}$. In other words, $\ket{\Phi}$ has a unique
418: Schmidt decomposition.
419: \end{lemma}
420:
421: \textit{Proof.} Without loss of generality, assume
422: $h(\ket{a},\ket{b})=3$. By contradiction, suppose there exist
423: another two product vectors $\ket{c}$ and $\ket{d}$ which also yield
424: a decomposition of $\ket{\Phi}$. That is,
425: \begin{equation}\label{schmidtdecom}
426: \ket{a_1a_2a_3}+\ket{b_1b_2b_3}=\ket{c_1c_2c_3}+\ket{d_1d_2d_3},
427: \end{equation}
428: where $a_k$ is not equal to $b_k$ for each $1\leq k\leq 3$. Let
429: $\ket{a_1^\perp}\in span\{\ket{a_1},\ket{b_1}\}$ such that
430: $\ip{a_1^\perp}{a_1}=0$ and $\ip{a_1^\perp}{b_1}\neq 0$. Then
431: multiplying $\bra{a_1^\perp}$ on the both sides of the above
432: equation we have that $\ket{b_2b_3}$ is contained in
433: $S=span\{\ket{c_2c_3},\ket{d_2d_3}\}$. Similarly we can also show
434: $\ket{a_2a_2}\in S$. On the other hand, there are only two unique
435: product vectors $\ket{c_2c_3}$ and $\ket{d_2d_3}$ in $S$ (up to some
436: factors). Then it follows that $\{\ket{a_2a_3}, \ket{b_2b_3}\}$ is
437: essentially $\{\ket{c_2c_3}, \ket{d_2d_3}\}$. Without loss of
438: generality, assume $\ket{a_2a_3}=\alpha \ket{c_2c_3}$ and
439: $\ket{b_2b_3}=\beta\ket{d_2d_3}$ for some complex numbers $\alpha$
440: and $\beta$, then
441: \begin{equation}\label{contra}
442: (\alpha\ket{a_1}-\ket{c_1})\ket{c_2c_3}+(\beta\ket{b_1}-\ket{d_1})\ket{d_2d_3}=0,
443: \end{equation}
444: which is possible if and only if $\ket{a_1}=\alpha^{-1}\ket{c_1}$
445: and $\ket{b_1}=\beta^{-1}\ket{d_1}$. That also means
446: $\ket{a}=\ket{c}$ and $\ket{b}=\ket{d}$. A contradiction with our
447: assumption.\hfill $\blacksquare$
448:
449: We also need the following result concerning with the separability
450: of rank $2$ positive semi-definite operators.
451:
452: \begin{lemma}\label{2sumsep}\upshape
453: Let $\ket{\Psi}$ and $\ket{\Phi}$ be pure states on $\mathcal{H}$
454: and $\lambda\geq 0$. Then
455: $\rho(\lambda)=\op{\Psi}{\Psi}+\lambda\op{\Phi}{\Phi}$ is separable
456: if and only if one of the following cases holds:
457:
458: (i) Both $\ket{\Psi}$ and $\ket{\Phi}$ are product states and
459: $\lambda\geq 0$;
460:
461: (ii) $\ket{\Psi}$ is a product state and $\lambda=0$;
462:
463: (iii) There are two product states $\ket{a}$ and $\ket{b}$ and
464: positive real numbers $\alpha$, $\beta$, $\gamma$, $\delta$ such
465: that $\ket{\Psi}=\alpha\ket{a}+\beta\ket{b}$,
466: $\ket{\Phi}=\gamma\ket{a}-\delta\ket{b}$, and
467: $\lambda=\alpha\beta/\gamma\delta$. (There may be some global phase
468: factors before $\ket{\Psi}$ and $\ket{\Phi}$.)
469: \end{lemma}
470:
471: \textit{Proof.} By a direct calculation one can readily verify that
472: (i), (ii) and (iii) are sufficient for the separability of
473: $\rho(\lambda)$. We only need to show that they are also necessary.
474: The case that both states are unentangled is trivial, corresponding
475: to (i). We shall consider the following two cases only:
476:
477: (1) One of $\ket{\Psi}$ and $\ket{\Phi}$ is entangled. First assume
478: that $\ket{\Phi}$ is entangled. Then for any $\lambda>0$.
479: $\rho(\lambda)$ is just a mixture of an entangled state and a
480: product state, and should be entangled for any $\lambda>0$, as shown
481: by Horodecki {\it et al.} \cite{HSTT03}. So the only possible case
482: is $\lambda=0$, i.e., (ii) holds. Second, when $\ket{\Psi}$ is
483: entangled we can show by a similar argument that $\rho(\lambda)$
484: cannot be separable for any $\lambda\geq 0$.
485:
486: (2) Both states are entangled. We shall show that condition (iii) is
487: necessary. Suppose that $\rho(\lambda)$ is separable, then by Lemma
488: \ref{schmidtsep}, it follows that
489: $\sch(\rho(\lambda))=R(\rho(\lambda))=2$. In other words, the
490: support of $\rho(\lambda)$ can be spanned by two product states
491: $\ket{a}$ and $\ket{b}$. Noticing that both $\ket{\Psi}$ and
492: $\ket{\Phi}$ are contained in the support of $\rho(\lambda)$, we can
493: write
494: \begin{eqnarray}
495: \ket{\Psi}&=&\alpha\ket{a}+\beta\ket{b},\\
496: \ket{\Phi}&=&\gamma\ket{a}+\delta\ket{b}
497: \end{eqnarray}
498: for nonzero complex numbers $\alpha,\beta,\gamma, \delta$.
499: Furthermore, $\ket{a}$ and $\ket{b}$ are also the only product
500: states in $\supp(\rho(\lambda))$. So $\rho(\lambda)$ should be a
501: mixture of $\op{a}{a}$ and $\op{b}{b}$. Substituting $\ket{\Psi}$
502: and $\ket{\Phi}$ into $\rho(\lambda)$ and setting the coefficients
503: of $\op{a}{b}$ and $\op{b}{a}$ to be zero we have
504: \begin{equation}
505: \alpha\beta^*+\lambda\gamma\delta^*=0.
506: \end{equation}
507: By adding suitable phase factors before $\ket{\Psi}$ and
508: $\ket{\Phi}$, we can make $\alpha$ and $\gamma$ positive. The
509: condition $\lambda>0$ implies that $\beta=|\beta|e^{i\theta}$ and
510: $\delta=-|\delta|e^{i\theta}$ for some real number $\theta$.
511: Absorbing the phase factor $e^{i\theta}$ into $\ket{b}$ and
512: resetting $\beta$ and $\delta$ to $|\beta|$ and $-|\delta|$
513: respectively, we know that condition (iii) is satisfied. \hfill
514: $\blacksquare$
515: \section{Conditions for the distinguishability of quantum states by separable
516: operations}It would be desirable to know when a collection of
517: quantum states is perfectly distinguishable by separable operations.
518: Orthogonality is generally not sufficient for the existence of a
519: separable POVM for discrimination. A rather simple but useful
520: necessary and sufficient condition is as follows.
521: \begin{theorem}\label{exactdis}\upshape
522: Let $\mathcal{S}=\{\rho_1,\cdots,\rho_n\}$ be a collection of
523: orthogonal quantum states of $\mathcal{H}$. Then $\mathcal{S}$ is
524: perfectly distinguishable by separable operations if and only if
525: there exist $n$ positive semi-definite operators $\{E_1,\cdots,
526: E_n\}$ such that $P_k+E_k$ is separable for each $1\leq k\leq n$,
527: and $\sum_{k=1}^n E_k=P_0$, where $P_k$ is the projector on
528: $\supp(\rho_k)$, and $P_0=I_{\mathcal{H}}-\sum_{k=1}^n P_k$.
529: \end{theorem}
530:
531: \textit{Proof.} Sufficiency is obvious. Suppose that (ii) holds, let
532: us define a POVM $\Pi=\{\Pi_1,\cdots, \Pi_n\}$ as follows:
533: $\Pi_{k}=P_k+E_k$ for each $1\leq k\leq n$. It is easy to verify
534: that $\Pi$ is a separable measurement that perfectly discriminates
535: $\mathcal{S}$.
536:
537: Now we turn to show the necessity. Suppose $\mathcal{S}$ is
538: perfectly distinguishable by some separable POVM, say
539: $\{\Pi_1,\cdots,\Pi_n\}$. Take $E_k=\Pi_k-P_k$ for each $1\leq k\leq
540: n$. Then $\sum_{k=1}^n E_k=P_0$. To complete the proof, it suffices
541: to show $E_{k}\geq 0$. By the assumption, we have
542: $\tr(\Pi_k\rho_k)=1$. Then the positivity of $E_k$ follows directly
543: from Lemma \ref{tech}. \hfill
544: $\blacksquare$\\
545:
546: In some sense Theorem \ref{exactdis} is only a reformulation of our
547: discrimination problem. One may doubt it can provide any valuable
548: results. However, this seemingly trivial reformulation does give us
549: much more operational forms of the measurement operators for
550: discrimination and connects them to the separability problem, which
551: is one of the central topics in quantum information theory and has
552: been extensively studied since the last two decades. So many well
553: developed tools for the separability problem may be applicable in
554: the context of discrimination. As a result, Theorem \ref{exactdis}
555: is unexpectedly powerful.
556:
557: Some special but interesting cases of Theorem \ref{exactdis} deserve
558: careful investigations. When the supports of the states in
559: $\mathcal{S}$ together span the whole state space, i.e.,
560: $supp(\sum_{k=1}^n\rho_k)=\mathcal{H}$, $\mathcal{S}$ is perfectly
561: distinguishable by separable operations if and only if $P_k$ is
562: separable for each $1\leq k\leq n$. In particular, an orthonormal
563: basis of $\mathcal{H}$ is perfectly distinguishable by separable
564: operations if and only if it is a product basis, which is a well
565: known result first obtained by Horodecki \textit{et al.} using a
566: rather different method \cite{HSSH03}, see also \cite{CHE04} for
567: another simple proof.
568:
569: A slightly more general case is when there exists $k$ such $P_0+P_k$
570: and $P_j$ are all separable for any $j\neq k$. In this case we say
571: that $\{P_j:j\neq k\}$ is completable \cite{NOTE1}. It has been
572: shown in Ref. \cite{BDM+99} that a sufficient condition for a set of
573: orthogonal product states to be completable is that they are exactly
574: distinguishable by LOCC. Combining this with Theorem \ref{exactdis},
575: we have the following interesting corollary:
576: \begin{corollary}\label{UPB-sep}\upshape
577: A set of LOCC distinguishable product states together with any state
578: in its orthogonal complement is always perfectly distinguishable by
579: separable operations.
580: \end{corollary}
581:
582: We notice that DiVincenzo and co-workers have shown in Ref.
583: \cite{DMS+00} that any three (or four) orthogonal multipartite
584: (resp. bipartite) product states are distinguishable by LOCC. Hence
585: we conclude immediately by the above corollary that any four (or
586: five) orthogonal multipartite (resp. bipartite) states including
587: three (resp. four) product states are perfectly distinguishable by
588: separable operations. The case when $\mathcal{S}$ is a set of
589: product states is of particular interest and has been studied
590: carefully in Ref. \cite{DMS+00} using a rather different method.
591: Specifically, let
592: $\mathcal{S}=\{\ket{\psi_1},\cdots,\ket{\psi_n}\}$. Then it has been
593: shown that if $S_k=S-\{\ket{\psi_k}\}$ is completable for each
594: $1\leq k\leq n$, then $S$ is perfectly distinguishable by separable
595: operations. Interesting examples include any orthogonal UPB
596: consisting of four $2\otimes 2\otimes 2$ (or five $3\otimes 3$)
597: product states. Clearly we have refined the results by DiVincenzo
598: {\it et al.}
599:
600: As another important consequence of Theorem \ref{exactdis}, we shall
601: show that the problem of distinguishing $(D-1)$ orthogonal pure
602: states by separable operations can be completely solved. Notice that
603: if $\ket{\Phi}$ is a product state, then the only form of
604: $\mathcal{B}$ that can be distinguishable by separable operations is
605: a product basis. For simplicity, in what follows we shall assume
606: $\ket{\Phi}$ is an entangled state. We shall consider two cases
607: $\sch_\perp(\Phi)=2$ and $\sch_\perp(\Phi)\geq 3$ respectively in
608: the next two sections.
609:
610: \section{Distinguishability of arbitrary orthonormal basis of
611: $\{\ket{\Phi}\}^\perp$ with $\sch_\perp(\Phi)=2$}
612:
613: We start with the simplest $2\otimes 2$ case and provide an
614: analytical solution. It is well known that any $2\otimes 2$ state
615: $\ket{\Psi}$ can be uniquely represented by a $2\times 2$ matrix
616: $\Psi$ as follows:
617: \begin{equation}\label{isomorphism}
618: \ket{\Psi}=(I\otimes \Psi)\ket{\Phi_+},
619: \end{equation}
620: where $\ket{\Phi_+}=1/\sqrt{2}(\ket{00}+\ket{11})$ is a maximally
621: entangled state. The \textit{concurrence} of $\ket{\Psi}$ is given
622: by the absolute value of the determinant of $\Psi$, i.e.,
623: $C(\Psi)=|\det(\Psi)|.$ This rewriting form coincides with the
624: original definition given by Wootters \cite{WOO98}. So $0\leq
625: C(\Psi)\leq 1$ for any $2\otimes 2$ state $\ket{\Psi}$. In
626: particular, $C(\Psi)$ vanishes for product states while attains one
627: for maximally entangled states. We say complex numbers $z_1$ and
628: $z_2$ are anti-parallel if there exists a negative real number $a$
629: such that $z_1=az_2$. First we need the following simplified version
630: of Lemma \ref{2sumsep} in the $2\otimes 2$ case.
631: \begin{lemma}\label{2sep}\upshape
632: Let $\ket{\Psi}$ and $\ket{\Phi}$ be two $2\otimes 2$ entangled pure
633: states, and let $\lambda\geq 0$. Then
634: $\rho(\lambda)=\op{\Psi}{\Psi}+\lambda\op{\Phi}{\Phi}$ is separable
635: if and only if $\Psi\Phi^{-1}$ has two anti-parallel eigenvalues and
636: $\lambda=C(\Psi)/C(\Phi)$. Note that $\ket{\Phi}$ is entangled
637: implies that $\Phi$ is invertible.
638: \end{lemma}
639:
640: Now the distinguishability of three $2\otimes 2$ orthogonal states
641: by separable operations is characterized as follows.
642: \begin{theorem}\label{2otimes2sep}\upshape
643: Let $\ket{\Phi}$ be a $2\otimes 2$ pure state, and let
644: $\mathcal{B}=\{\ket{\Psi_1},\ket{\Psi_2},\ket{\Psi_3}\}$ be an
645: orthonormal basis for $\{\ket{\Phi}\}^{\perp}$. Then $\mathcal{B}$
646: is perfectly distinguishable by separable operations if and only if
647: (i) $\Psi_k\Phi^{-1}$ has two anti-parallel eigenvalues for each
648: entangled state $\Psi_k$; and (ii)
649: $C(\Psi_1)+C(\Psi_2)+C(\Psi_3)=C(\Phi)$.
650: \end{theorem}
651:
652: \textit{Proof.} By Theorem \ref{exactdis}, the POVM element that can
653: perfectly identify $\ket{\Psi_k}$ should have the form
654: $\Pi_k=\op{\Psi_k}{\Psi_k}+\lambda_k\op{\Phi}{\Phi},$ where $0\leq
655: \lambda_k\leq 1$ and $\lambda_1+\lambda_2+\lambda_3=1$. If
656: $\ket{\Psi_k}$ is a product state, then it follows from Lemma
657: \ref{2sumsep} that $\Pi_k$ is separable if and only if
658: $\lambda_k=0$. Otherwise by Lemma \ref{2sep} we have that $\Pi_k$ is
659: separable if and only if ${\Psi_k}{\Phi}^{-1}$ has two anti-parallel
660: eigenvalues and $\lambda_k=C(\Psi_k)/C(\Phi)$. So by
661: $\lambda_1+\lambda_2+\lambda_3=1$ we have
662: $C(\Psi_1)+C(\Psi_2)+C(\Psi_3)=C(\Phi)$. That completes the
663: proof.\hfill $\blacksquare$
664:
665: The most interesting part in the above theorem is Condition (ii). It
666: reveals a precise relation between the concurrences and the
667: distinguishability by separable operations. By a careful
668: observation, we notice that a similar condition has been previously
669: obtained by Hayashi {et al.} in Ref. \cite{HMM+06}, where a very
670: general relation between distinguishability by separable operations
671: and average entanglement degree of the states to be discriminated
672: was presented. When only $2\otimes 2$ states were involved, this
673: relation can be reduced to an inequality similar to condition (ii).
674: The key difference is that here what we obtained is an exact
675: identity that (almost) completely characterizes the
676: distinguishability by separable operations.
677:
678: Suppose now ${\ket{\Phi}}$ is maximally entangled. Then $\Phi=U$ for
679: some $2\times 2$ unitary matrix. Noticing that
680: $$\tr(\Psi_k\Phi^{-1})=\tr(U^\dagger {\Psi_k})=2\ip{\Phi}{\Psi_k}=0$$
681: for $k=1,2,3$, we conclude that $\Psi_k\Phi^{-1}$ should have two
682: anti-parallel eigenvalues (assume that $\ket{\Psi_k}$ is entangled).
683: Furthermore, the concurrence of a $2\otimes 2$ maximally entangled
684: state is one. Collecting all these facts together we have the
685: following interesting corollary:
686: \begin{corollary}\label{2ME}\upshape
687: Let $\ket{\Phi}$ be a $2\otimes 2$ maximally entangled state, and
688: let $\mathcal{B}=\{\ket{\Psi_1},\ket{\Psi_2},\ket{\Psi_3}\}$ be an
689: orthonormal basis for $\{\ket{\Phi}\}^{\perp}$. Then $\mathcal{B}$
690: is perfectly distinguishable by separable operations if and only if
691: $C(\Psi_1)+C(\Psi_2)+C(\Psi_3)=1$.
692: \end{corollary}
693:
694: Theorem \ref{2otimes2sep} implies that there exists a large class of
695: $2\otimes 2$ separable operations that cannot be implemented
696: locally. We shall exhibit an explicit construction of such separable
697: operations. We achieve this goal by identifying a set of states that
698: is distinguishable by separable operations but is not LOCC
699: distinguishable.
700:
701: Let $\ket{\Psi(\theta)}=\cos\theta\ket{01}+\sin\theta\ket{10}$ and
702: $\ket{\Phi(\theta)}=\cos\theta\ket{00}+\sin\theta\ket{11}$ for
703: $\theta\in\mathcal{R}$. Then for any $0<\alpha\leq \beta\leq \pi/4$,
704: let $\ket{\Phi}=\ket{\Phi(\beta)}$, and let
705: $\ket{\Psi_1}=\ket{\Psi(\alpha)},$
706: $\ket{\Psi_2}=\cos\gamma\ket{\Psi(\alpha-\frac{\pi}{2})}+\sin\gamma\ket{\Phi(\beta-\frac{\pi}{2})}$,
707: $\ket{\Psi_3}=\sin\gamma\ket{\Psi(\alpha-\frac{\pi}{2})}-\cos\gamma\ket{\Phi(\beta-\frac{\pi}{2})}$
708: be an orthonormal basis of $\{\ket{\Phi}\}^\perp$. Choose $\gamma$
709: such that
710: \begin{equation}\label{gamma}
711: \tan^{-1}\sqrt{\frac{\sin 2\alpha}{\sin 2\beta}}\leq \gamma\leq
712: \tan^{-1} \sqrt{\frac{\sin 2\beta}{\sin 2\alpha}}.
713: \end{equation}
714: By direct calculations we can easily verify that the assumptions of
715: Theorem \ref{2otimes2sep} are fulfilled, thus
716: $\mathcal{B}=\{\ket{\Psi_1},\ket{\Psi_2},\ket{\Psi_3}\}$ is
717: perfectly distinguishable using separable operations. However, it is
718: clear whenever $0<\alpha<\beta\leq \frac{\pi}{4}$, both
719: $\ket{\Psi_1}$ and $\ket{\Psi_2}$ are entangled, thus $\mathcal{B}$
720: is indistinguishable by LOCC, as shown by Walgate and Hardy
721: \cite{WH02}. Interestingly, $\ket{\Psi_3}$ is reduced to a product
722: state when $\gamma$ takes one of the extreme values in Eq.
723: (\ref{gamma}).
724:
725: The above example also implies that the distinguishability by
726: separable operations has some continuous property. We formalize this
727: intuitive observation as follows:
728: \begin{theorem}\label{continuous}\upshape
729: Let $c_1,c_2,c_3$ be nonnegative real numbers such that $0\leq
730: c_1+c_2+c_3\leq 1$. Then there exists a set of three $2\otimes 2$
731: states $\{\ket{\Psi_1},\ket{\Psi_2},\ket{\Psi_3}\}$ satisfying (i)
732: $C(\Psi_k)=c_k$ for $k=1,2,3$; and (ii)
733: $\{\ket{\Psi_1},\ket{\Psi_2},\ket{\Psi_3}\}$ is perfectly
734: distinguishable by separable operations.
735: \end{theorem}
736:
737: There is a nice geometrical way to illustrate the
738: distinguishability of quantum states by different class of
739: operations. To see this, we identify a triple of orthogonal quantum
740: states $(\ket{\psi_1},\ket{\psi_2},\ket{\psi_3})$ by a point
741: $(C(\psi_1),C(\psi_2),C(\psi_3))$ of $R^{3}$. We say such a point is
742: distinguishable by global (or separable, LOCC) operations if the
743: related states are distinguishable by the corresponding operations.
744: By a careful analysis, we can show any point in the regular
745: tetrahedron in Fig. 1 corresponds to some orthonormal basis of
746: $\{\ket{\Phi_+}\}^{\perp}$. A detailed proof for this interesting
747: fact is given in the Appendix A. So the points in the regular
748: tetrahedron represent a region that can be perfectly distinguishable
749: by global quantum operations. The yellow triangle $BCD$ is exactly
750: the set of points distinguishable by separable operations. Finally,
751: three vertices $B$, $C$, and $D$ are the only points that are
752: distinguishable by LOCC. From such a representation we can clearly
753: see that the class of $2\otimes 2$ separable operations strictly
754: lies between the classes of LOCC and global operations.
755:
756: \begin{figure}[ht]
757: \centering
758: \includegraphics[scale=0.7]{demo}
759: \caption{Illustration of the distinguishability of the basis of $\{\ket{\Phi_+}\}^\perp$, where $\ket{\Phi_+}$ is a $2\otimes 2$
760: maximally entangled states and we identify an orthonormal basis $(\ket{\psi_1},\ket{\psi_2},\ket{\psi_3})$
761: by a point $(C(\psi_1),C(\psi_2),C(\psi_3))$ in $R^3$. Any point in the regular tetrahedron ABCD corresponds a legal orthonormal basis
762: of $\{\ket{\Phi_+}\}^\perp$, thus is distinguishable by global operations. Vertices B, C, and D are the only points corresponding
763: to LOCC distinguishable basis. Interestingly, the triangle $\bigtriangleup BCD$ represents bases that are distinguishable by separable operations. }
764: \label{2}
765: \end{figure}
766:
767:
768: Now we consider the multipartite setting. There are two possible
769: cases: $\ket{\Phi}$ is reduced to a bipartite entangled state only
770: between two parties; or $\ket{\Phi}$ is an entangled state at least
771: between three parties. The following two theorems deal with these
772: two cases separately. Surprisingly, the first case is essentially
773: reduced to $2\otimes 2$ case.
774: \begin{theorem}\label{schmidt2}\upshape
775: Let $\ket{\Phi}=\ket{a_1\cdots a_{K-2}}\ket{\Phi'}$ be a pure state
776: on $\mathcal{H}$ such that ${\ket{\Phi'}}$ is an entangled state
777: between parties $A_{K-1}$ and $A_{K}$. Then $\mathcal{B}$ is
778: perfectly distinguishable by separable operations if and only if
779: each entangled state $\ket{\Psi}$ in $\mathcal{B}$ can be written
780: into the form $\ket{\Psi}=\ket{a_1\cdots a_{K-2}}\ket{\Psi'}$ for
781: some bipartite entangled state $\ket{\Psi'}$ between $A_{K-1}$ and
782: $A_{K}$ such that (i) all ${\ket{\Psi'}}$ and ${\ket{\Phi'}}$ can be
783: embeded into a $2\otimes 2$ subspace $\mathcal{H}'$ of
784: $\mathcal{H}_{K-1}\otimes \mathcal{H}_{K}$; and (ii)
785: $\Psi'{\Phi'}^{-1}$ has two anti-parallel eigenvalues and
786: $\sum_{\Psi'} C(\Psi')=C(\Phi')$, where $\Psi'$ and $\Phi'$ are
787: $2\otimes 2$ matrices that are the representations of $\ket{\Psi'}$
788: and $\ket{\Phi'}$ on $\mathcal{H}'$ in the sense of Eq.
789: (\ref{isomorphism}), respectively.
790: \end{theorem}
791:
792: \textit{Proof.} Let $\ket{\Psi}$ be any entangled state of
793: $\mathcal{B}$. Then there is some $0<\lambda\leq 1$ such that
794: $\op{\Psi}{\Psi}+\lambda\op{\Phi}{\Phi}$ is separable. According to
795: the assumptions on $\ket{\Phi}$ and employing Lemma \ref{2sumsep},
796: we can easily see that $\ket{\Psi}$ should be of the form
797: $\ket{b_1\cdots b_{K-2}}\otimes \ket{\Psi'}$. Second, we have
798: $\ket{a_1\cdots a_{K-2}}$ should be identical to $\ket{b_1\cdots
799: b_{K-2}}$ up to some factors. Otherwise
800: $span\{\ket{\Psi},\ket{\Phi}\}$ cannot contain product states. So we
801: need $\op{\Psi'}{\Psi'}+\lambda\op{\Phi'}{\Phi'}$ being separable.
802: Now the problem is reduced to the $2\otimes 2$ case and the result
803: follows directly from Theorem \ref{2otimes2sep}. \hfill
804: $\blacksquare$
805:
806: The case that $\ket{\Phi}$ is at least entangled among three parties
807: is much more simple. Any distinguishable basis should contain a
808: unique entangled state with a predetermined form.
809: \begin{theorem}\upshape
810: Let $\ket{\Phi}=\cos\theta\ket{a}+\sin\theta\ket{b}$ be an entangled
811: pure state of $\mathcal{H}$ such that $\ip{a}{b}=0$ and
812: $h(\ket{a},\ket{b})\geq 3$. Then $\mathcal{B}$ can be perfectly
813: distinguishable by separable operations if and only if there is a
814: unique entangled state $\ket{\Psi}$ in $\mathcal{B}$ with the form
815: $\ket{\Psi}=\sin\theta\ket{a}-\cos\theta\ket{b}$, and the other
816: states should form an orthogonal product basis of
817: $\{\ket{a},\ket{b}\}^\perp$.
818: \end{theorem}
819: \textit{Proof.} Sufficiency is obvious. We only consider the
820: necessity. Suppose that $\mathcal{B}$ is distinguishable by
821: separable operations. Noticing that $\ket{\Phi}$ is entangled, we
822: conclude that $\mathcal{B}$ should contain at least one entangled
823: state. By Lemma \ref{uniquedecom}, one can easily show that any
824: entangled state $\ket{\Psi}$ such that
825: $\op{\Psi}{\Psi}+\lambda\op{\Phi}{\Phi}$ is separable should be
826: contained in $span\{\ket{a},\ket{b}\}$. Noticing that $\ket{\Psi}$
827: is orthogonal to $\ket{\Phi}$, we can deduce that $\ket{\Psi}$ is
828: uniquely determined and is given by
829: $\sin\theta\ket{a}-\cos\theta\ket{b}$. \hfill $\blacksquare$
830: \section{$\{\ket{\Phi}\}^\perp$ has no bases distinguishable by
831: separable operations when $\sch_\perp(\Phi)\geq 3$} Now we consider
832: the case when $\sch_\perp(\Phi)\geq 3$. Our result is a
833: generalization of Watrous's result. It is remarkable that the
834: orthogonal Schmidt number of $\ket{\Phi}$ completely characterizes
835: the distinguishability of the subspace $\{\ket{\Phi}\}^\perp$.
836:
837: \begin{theorem}\label{indissubspace}\upshape
838: Let $\ket{\Phi}$ be an entangled pure state on $\mathcal{H}$. Then
839: $\{\ket{\Phi}\}^{\perp}$ having no orthonormal basis perfectly
840: distinguishable by separable operations if and only if
841: $Sch_{\perp}(\Phi)>2$. In particular, when $Sch_{\perp}(\Phi)=2$,
842: there always exists an orthonormal basis $\mathcal{B}$ of
843: $\{\ket{\Phi}\}^{\perp}$ that is perfectly distinguishable by LOCC.
844: \end{theorem}
845:
846: \textit{Proof.} Necessity: Suppose $Sch_\perp(\Phi)=2$. Then the
847: existence of a distinguishable basis by separable operations follows
848: directly from Theorem \ref{schmidt2}. Here we further construct a
849: basis for $\{\ket{\Phi}\}^\perp$ that is distinguishable by LOCC.
850: Notice that $\ket{\Phi}$ can be written into the form
851: $\ket{\Phi}=\cos\theta\ket{\Phi_0}+\sin\theta\ket{\Phi_1},$ where
852: $\ket{\Phi_0}$ and $\ket{\Phi_1}$ are orthogonal product states on
853: $\mathcal{H}$ and $0<\theta<\pi/2$. Then we can extend
854: $\ket{\Phi_0}$ and $\ket{\Phi_1}$ into a complete orthogonal product
855: basis $\{\ket{\Phi_0},\ket{\Phi_1},\cdots, \ket{\Phi_{D-1}}\}$ of
856: $\mathcal{H}$. We can further assume this basis is distinguishable
857: by local projective measurements \cite{BDM+99}. Replacing
858: $\ket{\Phi_0}$ and $\ket{\Phi_1}$ with
859: $\ket{\Psi}=\sin\theta\ket{\Phi_0}-\cos\theta\ket{\Phi_1}$ we obtain
860: a basis for $\{\ket{\Phi}\}^\perp$ that is distinguishable by the
861: same local projective measurements.
862:
863: Sufficiency: In this case we have $Sch_{\perp}(\Phi)\geq 3$. If
864: $Sch(\Phi)\geq 3$. Then for any $\lambda_k>0$ we know from Lemma
865: \ref{2sumsep} that
866: $\Pi(\lambda_k)=\op{\Psi_k}{\Psi_k}+\lambda_k\op{\Phi}{\Phi}$ should
867: be entangled. That implies any basis of $\{\ket{\Phi}\}^\perp$ is
868: indistinguishable by separable operations.
869:
870: Suppose now that $Sch_{\perp}(\Phi)\geq 3$ and $Sch(\Phi)=2$. By
871: contradiction, assume there exists a basis $\mathcal{B}$
872: distinguishable by separable operations. Then applying Lemma
873: \ref{uniquedecom}, we know there are two unique product vectors
874: $\ket{a}$ and $\ket{b}$ such that $\ket{\Phi}=\ket{a}+\ket{b}.$ So
875: any state $\ket{\Psi_k}$ with $\lambda_k>0$ should also be a
876: superposition of $\ket{a}$ and $\ket{b}$. As there are only two
877: orthogonal states in $span\{\ket{a},\ket{b}\}$, we conclude the only
878: possibility is that there is a unique state $\ket{\Psi_k}$ with
879: $\lambda_k=1$ and all the other states are product states. On the
880: other hand, let $\ket{\Psi}$ be the entangled state in
881: $span\{\ket{a},\ket{b}\}$ such that $\ip{\Psi}{a}=0$. Then
882: \begin{equation}
883: \op{\Psi_k}{\Psi_k}+\op{\Phi}{\Phi}=\op{a}{a}+\op{\Psi}{\Psi}
884: \end{equation}
885: is separable. But the right hand side of the above equation is a
886: summation of an entangled state and a product state, it should be
887: entangled \cite{HSTT03}. A contradiction. \hfill $\blacksquare$
888:
889: Let us check some interesting examples. Take $\ket{\Phi}$ to be a
890: $W$-type state, $\ket{\Phi}=a\ket{001}+b\ket{010}+c\ket{100},$ where
891: $|a|^2+|b|^2+|c|^2=1$, and $abc\neq 0$. Then
892: $Sch_\perp(\Phi)=Sch(\Phi)=3$. It follows from Theorem
893: \ref{indissubspace} that any orthonormal basis of
894: $\{\ket{\Phi}\}^{\perp}$ cannot be perfectly distinguishable by
895: separable operations. This yields an indistinguishable subspace with
896: dimension $2^3-1=7$, which is a slight improvement over the
897: bipartite case, where a $3\otimes 3$ indistinguishable subspace of
898: dimension $8$ was given by Watrous \cite{WAT05}. But if we choose
899: the $GHZ$-type state $\ket{\Phi}=\cos\theta
900: \ket{000}+\sin\theta\ket{111},$ where $0\leq \theta\leq
901: \frac{\pi}{2}$. Then $\{\ket{\Phi}\}^{\perp}$ does have an
902: orthonormal basis that is perfectly distinguishable by LOCC.
903: Interestingly, if we take
904: $\ket{\Phi}=\alpha\ket{000}+\beta\ket{+++},$ where $\alpha\beta\neq
905: 0$ and $\ket{+}=(\ket{0}+\ket{1})/\sqrt{2}$. Then we can easily see
906: that $Sch(\Phi)=2$ but $\sch_\perp (\Phi)\geq 3$, as $\ket{\Phi}$
907: cannot be written into a superposition of two orthogonal product
908: states. Thus it follows from the above theorem that
909: $\{\ket{\Phi}\}^\perp$ does not have any orthonormal basis
910: distinguishable by separable operations.
911:
912: \section{Indistinguishable subspaces with smaller dimensions}
913: The dimension of indistinguishable subspaces can be further reduced.
914: We shall show that there exist a $3\otimes 3$ indistinguishable
915: subspace with dimension $7$ and a $2\otimes 2\otimes 2$
916: indistinguishable subspace with dimension $6$. We first consider
917: bipartite case. Let
918: $\ket{\Phi_1}=(\ket{00}+\ket{11}+\ket{22})/\sqrt{3}$ and
919: $\ket{\Phi_2}=\ket{01}$. Let $S=span\{\ket{\Phi_1},\ket{\Phi_2}\}$.
920: It is clear that $S^{\perp}$ is a bipartite subspace with dimension
921: $7$. Surprisingly, we have the following interesting result:
922: \begin{theorem}\upshape\label{dim7}
923: $S^{\perp}$ is indistinguishable by separable operations.
924: \end{theorem}
925:
926: \textit{Proof.} We need the following easily verifiable properties
927: of $S$ to complete the proof:
928:
929: P0: $\ket{\Phi_2}$ is the unique product vector (up to a scalar
930: factor) in $S$;
931:
932: P1. $\sch(\Psi)=3$ for any entangled state $\ket{\Psi}\in S$;
933:
934: P2. Any positive operator $\rho$ such that $\supp(\rho)=S$ satisfies
935: $\sch(\rho)=4$.
936:
937: The validity of P0 can be directly verified. We only consider P1 and
938: P2. We first show the validity of P1. Actually, any entangled state
939: $\ket{\psi}$ from $S$ can be written into a superposition of
940: $\ket{\Phi_1}$ and $\ket{\Phi_2}$. That is,
941: \begin{equation}
942: \ket{\psi}=\alpha\ket{\Phi_1}+\beta\ket{\Phi_2},
943: \end{equation}
944: where $\alpha\neq 0$ and $|\alpha|^2+|\beta|^2=1$. Substituting
945: $\ket{\Phi_1}$ and $\ket{\Phi_2}$ into the above equation, we have
946: \begin{equation}
947: \ket{\psi}=\frac{1}{\sqrt{3}}(\ket{0}(\alpha\ket{0}+\sqrt{3}\beta\ket{1})+\alpha\ket{11}+\alpha\ket{22}),
948: \end{equation}
949: which is clearly an entangled state with Schmidt number $3$ for any
950: $\alpha\neq 0$.
951:
952: Now we turn to show the validity of P2. By contradiction, suppose
953: for some $\rho$ such that $\supp(\rho)=S$ we have $Sch(\rho)=3$.
954: Then there should exist three unnormalized product states
955: $\ket{a_1b_1}$,$\ket{a_2b_2}$, and $\ket{a_3b_3}$ such that
956: \begin{eqnarray}
957: \ket{\Phi_1}&=&\ket{a_1b_1}+\ket{a_2b_2}+\ket{a_3b_3},\\
958: \ket{\Phi_2}&=&\alpha_1\ket{a_1b_1}+\alpha_2\ket{a_2b_2}+\alpha_3\ket{a_3b_3}.
959: \end{eqnarray}
960: Without loss of generality, we may assume $\alpha_1\neq 0$. Then it
961: is clear that
962: \begin{equation}
963: \ket{\Phi_1}-\alpha_1^{-1}\ket{\Phi_2}=(1-\frac{\alpha_2}{\alpha_1})\ket{a_2b_2}+(1-\frac{\alpha_3}{\alpha_1})\ket{a_3b_3}.
964: \end{equation}
965: By P1, the left hand side of the above equation has Schmidt number
966: $3$, while the right hand side of the above equation has Schmidt
967: number at most $2$. That is a contradiction.
968:
969: Now let $\mathcal{B}=\{\ket{\Psi_k}:1\leq k\leq 7\}$ be any
970: orthogonal basis of $S^\perp$. By Theorem \ref{exactdis}, the POVM
971: element identifying $\ket{\Psi_k}$ is of the form
972: $\Pi_k=\op{\Psi_k}{\Psi_k}+E_k$, where $\sum_{k=1}^7
973: E_k=\op{\Phi_1}{\Phi_1}+\op{\Phi_2}{\Phi_2}$ and $E_k\geq 0$.
974:
975: So we can choose $E_k$ such that $E_k\neq 0$ and $E_k\neq
976: \op{01}{01}$ (up to some factor). We shall show that $\Pi_k$ should
977: be entangled. There are two cases. If $R(E_k)=1$ then by P1 we have
978: $\sch(E_k)=3$, which follows that $\sch(\Pi_k)\geq 3>R(\Pi_k)$.
979: Similarly, if $R(E_k)=2$ then by P2 we have $\sch(\Pi_k)\geq
980: \sch(E_k)=4>R(\Pi_k)$. In both cases $\Pi_k$ is entangled. That
981: completes the proof. \hfill $\blacksquare$
982:
983: Using the very same method, we can construct a $2\otimes 2\otimes 2$
984: indistinguishable subspace with dimension $6$. For instance, take
985: $\ket{\Phi_1}=(\ket{001}+\ket{010}+\ket{100})/\sqrt{3}$ and
986: $\ket{\Phi_2}=\ket{000}$. Then $\{\ket{\Phi_1},\ket{\Phi_2}\}^\perp$
987: is indistinguishable by separable operations.
988:
989: \section{Conclusions} In summary, we provided a necessary and
990: sufficient condition for the distinguishability of a set of
991: multipartite quantum states by separable operations. A set of three
992: $2\otimes 2$ pure states that is perfectly distinguishable by
993: separable operations but is indistinguishable by LOCC then was
994: explicitly constructed. As a consequence, there exists a large class
995: of nonlocal separable operations even for the simplest composite
996: quantum system consisting of two qubits. We also showed that the
997: orthogonal complement of a pure state has no bases distinguishable
998: by separable operations if and only if this state has an orthogonal
999: Schmidt number not less than $3$. We believe these results would be
1000: useful in clarifying the relation between separable operations and
1001: LOCC.
1002:
1003: There are still a number of unsolved problems that are worthwhile
1004: for further study. We have mentioned some in the previous context.
1005: Here we would like to stress two of them. The first one is
1006: concerning with the distinguishability of orthogonal product pure
1007: states. It a simple fact that any set of orthogonal product pure
1008: states can be perfectly distinguishable by some
1009: positive-partial-transpose preserving (PPT) operation. Is this also
1010: true for separable operations? We have seen that separable
1011: operations are powerful enough to distinguish any set of orthogonal
1012: product pure states in $3\otimes 3$ and $2\otimes 2\otimes 2$. If
1013: the answer is affirmative in general, then how to prove? Otherwise a
1014: counterexample would be highly desirable. With little effort we can
1015: easily show that it is sufficient to verify whether any UPB is
1016: perfectly distinguishable by separable operations. The difficulty we
1017: met is that the structure of UPB on a multipartite state space
1018: remains unknown except for some special cases.
1019:
1020: Another problem is to find more applications of locally
1021: indistinguishable subspaces (LIS). The work of Watrous suggests that
1022: bipartite LIS can be used to construct quantum channels with
1023: sub-optimal environment-assisted capacity. It would be of great
1024: interest to employ LIS as a tool to give tighter upper of the
1025: capacity. In the asymptotic setting similar problem has been
1026: thoroughly studied by Winter \cite{WIN05}. Hopefully, these efforts
1027: would reveal some deep properties of LIS and testify the richness of
1028: the mathematical structure of this notion.
1029: \section*{Appendix: Geometric representation of the concurrences of orthogonal bases of $\{\ket{\Phi}\}^\perp$}
1030:
1031: First we recall a useful representation of the concurrence of a
1032: two-qubit pure state. Let $\{\ket{\Phi_k}: 1\leq k\leq 4\}$ be the
1033: magic basis \cite{WOO98}, and let $\ket{\psi}$ be any pure state
1034: such that $\ket{\psi}=\sum_{k=1}^4 \lambda_k\ket{\Phi_k}$. Then the
1035: concurrence of $\ket{\psi}$ is given by the following formula:
1036: \begin{equation}\label{con-2}
1037: C(\psi)=|\lambda_1^2+\lambda_2^2+\lambda_3^2+\lambda_4^2|.
1038: \end{equation}
1039: Suppose now we choose $\ket{\Phi}$ to be one of
1040: $\{\ket{\Phi_k}:1\leq k\leq 4\}$, say $\ket{\Phi_4}$. Then any
1041: vector from $\{\ket{\Phi}\}^\perp$ should be a linear combination of
1042: $\{\ket{\Phi_1},\ket{\Phi_2},\ket{\Phi_3}\}$. As a direct
1043: consequence, we have the following interesting lemma which connects
1044: the concurrences of orthonormal bases of $\{\ket{\Phi}\}^\perp$ to
1045: $3\times 3$ unitary matrices.
1046:
1047: \begin{lemma}\label{con-unitary}\upshape
1048: Let $\{\ket{\psi_1},\ket{\psi_2},\ket{\psi_3}\}$ be an orthogonal
1049: basis for $\{\ket{\Phi}\}^\perp$, then there exists a $3\times 3$
1050: unitary matrices $U=[u_{kl}]_{3\times 3}$ such that
1051: \begin{eqnarray}\label{con-3}
1052: C(\psi_1)=|u_{11}^2+u_{12}^2+u_{13}^2|,\\
1053: C(\psi_2)=|u_{21}^2+u_{22}^2+u_{23}^2|,\\
1054: C(\psi_3)=|u_{31}^2+u_{32}^2+u_{33}^2|.
1055: \end{eqnarray}
1056: Conversely, for any unitary matrix $U=[u_{kl}]_{3\times 3}$, there
1057: exists an orthogonal basis for $\{\ket{\Phi}\}^\perp$, say
1058: $\{\ket{\psi_1},\ket{\psi_2},\ket{\psi_3}\}$ such that Eq.
1059: (\ref{con-3}) holds.
1060: \end{lemma}
1061: Motivated by the above lemma, let $\mathcal{P}_1$ be the set of
1062: $x\in \mathcal{R}^3$ such that there exists a unitary matrix
1063: $U=[u_{kj}]_{3\times 3}$ such that
1064: \begin{eqnarray}\label{P-1}
1065: x_1&=&|u_{11}^2+u_{12}^2+u_{13}^2|,\\
1066: x_2&=&|u_{21}^2+u_{22}^2+u_{23}^2|,\\
1067: x_3&=&|u_{31}^2+u_{32}^2+u_{33}^2|.
1068: \end{eqnarray}
1069: Then it is clear that $\mathcal{P}_1$ is exactly the set of points
1070: corresponding to the concurrences of the orthogonal bases of
1071: $\{\ket{\Phi}\}^\perp$. Let $\mathcal{P}_2$ be the set of $x\in
1072: \mathcal{R}^3$ satisfying the following equations:
1073: \begin{eqnarray}\label{P-2}
1074: x_1+x_2+x_3\geq 1,\\
1075: x_1+x_2-x_3\leq 1,\\
1076: x_2+x_3-x_1\leq 1,\\
1077: x_3+x_1-x_2\leq 1,\\
1078: 0\leq x_1,x_2,x_3\leq 1.
1079: \end{eqnarray}
1080: Obviously, $\mathcal{P}_2$ is just a unit regular tetrahedron. We
1081: shall show that $\mathcal{P}_2$ is contained by $\mathcal{P}_1$.
1082: That means any point of the regular tetrahedron corresponds to some
1083: orthonormal basis of $\{\ket{\Phi}\}^\perp$.
1084: \begin{theorem}
1085: $\mathcal{P}_2\subseteq \mathcal{P}_1$.
1086: \end{theorem}
1087:
1088: \textit{Proof.} For any point $x\in \mathcal{P}_2$, we shall
1089: construct a $3\times 3$ unitary operation $U$ such that Eq.
1090: (\ref{P-1}) holds. Without loss of generality, we may assume
1091: $x_1\geq x_2\geq x_3\geq 0$. So Eq. (\ref{P-2}) is reduced to the
1092: following equations:
1093: \begin{eqnarray}\label{simP-2}
1094: x_1+x_2+x_3\geq 1,\\
1095: x_1+x_2-x_3\leq 1,\\
1096: 0\leq x_3\leq x_2\leq x_1\leq 1.
1097: \end{eqnarray}
1098: Construct a unitary matrix $U$ as follows:
1099: $$U=\left(
1100: \begin{array}{ccc}
1101: u_{11} & u_{12} & u_{13} \\
1102: u_{21} & u_{22} & u_{23} \\
1103: u_{31} & u_{32} & u_{33} \\
1104: \end{array}
1105: \right).\left(
1106: \begin{array}{ccc}
1107: 1 & 0 & 0 \\
1108: 0 & e^{i\theta} & 0 \\
1109: 0 & 0 & e^{i\theta} \\
1110: \end{array}
1111: \right),
1112: $$
1113: where $[u_{kl}]_{3\times 3}$ is a real orthogonal matrix, and $0\leq
1114: \theta\leq \pi$. By a simple calculation, we have
1115: $$U=\left(
1116: \begin{array}{ccc}
1117: u_{11} & u_{12}e^{i\theta} & u_{13}e^{i\theta} \\
1118: u_{21} & u_{22}e^{i\theta} & u_{23}e^{i\theta} \\
1119: u_{31} & u_{32}e^{i\theta} & u_{33}e^{i\theta} \\
1120: \end{array}
1121: \right).
1122: $$
1123: Our purpose is to choose suitable real numbers $u_{11}$, $u_{21}$,
1124: $u_{31}$, and $\theta$ such that
1125: \begin{eqnarray}
1126: u_{11}^2+(1-u_{11}^2)e^{2i\theta}&=&x_1,\\
1127: u_{21}^2+(1-u_{21}^2)e^{2i\theta}&=&x_2,\\
1128: u_{31}^2+(1-u_{31}^2)e^{2i\theta}&=&x_3,\\
1129: u_{11}^2+u_{21}^2+u_{13}^2&=&1.
1130: \end{eqnarray}
1131: From the first three equations, we have
1132: \begin{equation}\label{u1k}
1133: u_{1k}^2=\frac{\sin\theta\pm\sqrt{x_k^2-\cos^2\theta}}{2\sin\theta},~k=1,2,3.
1134: \end{equation}
1135: We shall choose a suitable $\theta$ such that the last normalized
1136: relation holds. That is, choose $\theta$ such that
1137: $$\sin\theta\pm\sqrt{x_1^2-\cos^2\theta}\pm\sqrt{x_2^2-\cos^2\theta}\pm\sqrt{x_3^2-\cos^2\theta}=0,$$
1138: where $\cos^{-1}(x_3)\leq \theta\leq \pi/2$. The difficulty here is
1139: how to choose suitable signs for $u_{1k}$. Let
1140: \begin{eqnarray}
1141: f(\theta)=\sin\theta-\sqrt{x_1^2-\cos^2\theta}-\sqrt{x_2^2-\cos^2\theta}-\sqrt{x_3^2-\cos^2\theta},\nonumber\\
1142: g(\theta)=\sin\theta-\sqrt{x_1^2-\cos^2\theta}-\sqrt{x_2^2-\cos^2\theta}+\sqrt{x_3^2-\cos^2\theta}.\nonumber
1143: \end{eqnarray}
1144: It is clear that
1145: \begin{eqnarray}
1146: f(\frac{\pi}{2})&=&1-x_1-x_2-x_3\leq 0,\\
1147: g(\frac{\pi}{2})&=&1+x_3-x_1-x_2\geq 0,\\
1148: f(\cos^{-1}(x_3))&=&g(\cos^{-1}(x_3))\in \mathcal{R},
1149: \end{eqnarray}
1150: from which we have
1151: \begin{equation}
1152: f(\frac{\pi}{2}).g(\frac{\pi}{2})\leq 0~\mbox{and}
1153: ~f(\cos^{-1}(x_3)).g(\cos^{-1}(x_3))\geq 0.
1154: \end{equation}
1155: Noticing that $f(\theta)g(\theta)$ is a real-valued function of
1156: $\theta$, we conclude by the intermediate value theorem that there
1157: exists $\cos^{-1}(x_3)\leq \theta_0\leq \frac{\pi}{2}$ such that
1158: $f(\theta_{0})g(\theta_0)=0$. So $f(\theta_0)=0$ or $g(\theta_0)=0$.
1159: Without loss of generality, assume $g(\theta_0)=0$, then we can
1160: choose $u_{11}$, $u_{21}$, $u_{31}$ as follows:
1161: \begin{eqnarray}
1162: u_{11}=\sqrt{\frac{\sin\theta_0-\sqrt{x_1^2-\cos^2\theta_0}}{2\sin\theta_0}},\\
1163: u_{21}=\sqrt{\frac{\sin\theta_0-\sqrt{x_2^2-\cos^2\theta_0}}{2\sin\theta_0}},\\
1164: u_{31}=\sqrt{\frac{\sin\theta_0+\sqrt{x_2^2-\cos^2\theta_0}}{2\sin\theta_0}}.
1165: \end{eqnarray}
1166: By extending $(u_{11},u_{21}, u_{31})$ into an real orthogonal
1167: matrix we obtain other entries $u_{kl}$. With that we complete the
1168: proof of $\mathcal{P}_2\subseteq \mathcal{P}_1$.\hfill
1169: $\blacksquare$
1170:
1171: We strongly believe that it also holds $\mathcal{P}_1\subseteq
1172: \mathcal{P}_2$, thus $\mathcal{P}_1=\mathcal{P}_2$. However, we
1173: don't know how to give a rigorous proof for this up to now.
1174:
1175: \smallskip
1176:
1177: \section*{Acknowledgement}
1178:
1179: We thank Z.-F. Ji, G.-M. Wang, J.-X. Chen, Z.-H. Wei, and C. Zhang
1180: for helpful conversations. R. Duan acknowledges J. Walgate for
1181: interesting discussions at QIP 2007.
1182:
1183: \begin{thebibliography}{99}
1184: \bibitem{BDF+99} C. H. Bennett, D. P. DiVincenzo, C. A. Fuchs, T.
1185: Mor, E. Rains, P. W. Shor, J. A. Smolin, and W. K. Wootters,
1186: \newblock ``Quantum nonlocality without entanglement,"
1187: \newblock {\it Phys. Rev. A}, vol. 59, no. 2, pp. 1070--1091, 1999.
1188:
1189: \bibitem{BDM+99} C. H. Bennett, D. P. DiVincenzo, T. Mor, P. W.
1190: Shor, J. A. Smolin, and B.M. Terhal,
1191: \newblock ``Unextendible product bases and bound entanglement,"
1192: \newblock {\it Phys. Rev. Lett.}, vol. 82, no. 26, pp. 5385--5388, 1999.
1193:
1194: \bibitem{DMS+00} D. P. DiVincenzo, T. Mor, P. W. Shor, J. A. Smolin,
1195: and B.M. Terhal,
1196: \newblock ``Unextendible product bases, uncompletable product bases and bound
1197: entanglement,"
1198: \newblock {\it Comm. Math. Phys.}, vol. 238, pp. 379--410, 2003.
1199:
1200: \bibitem{WSHV00} J. Walgate, A. J. Short, L. Hardy, and V. Vedral,
1201: \newblock ``Local Distinguishability of multipartite orthogonal quantum states,"
1202: \newblock {\it Phys. Rev. Lett.}, vol. 85, no. 23, pp. 4972--4975, 2000.
1203:
1204: \bibitem{VSPM01} S. Virmani, M. F. Sacchi, M. B. Plenio, and D. Markham,
1205: \newblock ``Optimal local discrimination of two multipartite pure states,"
1206: \newblock {\it Phys. Lett. A}, vol. 288, no. 2, pp. 62--68, 2001.
1207:
1208: \bibitem{CY02} Y.-X. Chen and D. Yang,
1209: \newblock ``Optimally conclusive discrimination of nonorthogonal entangled states by local operations and classical communications,"
1210: \newblock {\it Phys. Rev. A}, vol. 65, no. 2, 022320, 2002.
1211:
1212: \bibitem{HM03} M. Hillery and J. Mimih,
1213: \newblock ``Distinguishing two-qubit states using local measurements and restricted classical communication,"
1214: \newblock {\it Phys. Rev. A}, vol. 67, no. 4, 042304, 2003.
1215:
1216: \bibitem{JCY05} Z. Ji, H. Cao, and M. Ying,
1217: \newblock ``Optimal conclusive discrimination of two states can be achieved locally,"
1218: \newblock {\it Phys. Rev. A}, vol. 71, no. 3, 032323, 2005.
1219:
1220:
1221: \bibitem{ABB+05} A.~Ac\'{\i}n, E. Bagan, M. Baig, Ll. Masanes, and R.
1222: Mu$\tilde{{\rm n}}$oz-Tapia,
1223: \newblock ``Multiple-copy two-state discrimination with individual measurements,"
1224: \newblock {\it Phys. Rev. A}, vol. 71, no. 3, 032338, 2005.
1225:
1226: \bibitem{OGA06} Y. Ogata,
1227: \newblock ``Local distinguishability of quantum states in infinite-dimensional systems,"
1228: \newblock {\it J. Phys. A: Math. Gen.}, vol. 39, no. 12, pp. 3059-3069, 2006.
1229:
1230: \bibitem{WH02} J. Walgate and L. Hardy,
1231: \newblock ``Nonlocality, asymmetry, and distinguishing bipartite states,"
1232: \newblock {\it Phys. Rev. Lett.}, vol. 89, no. 14, 147901, 2002.
1233:
1234: \bibitem{TDL01} B. M. Terhal, D. P. DiVincenzo, and D. W. Leung,
1235: \newblock ``Hiding bits in Bell states,"
1236: \newblock {\it Phys. Rev. Lett.}, vol. 86, pp. 5807--5810, 2001.
1237:
1238: \bibitem{DLT02} D. P. DiVincenzo, D. W. Leung, and B. M. Terhal,
1239: \newblock ``Quantum data hiding,"
1240: \newblock {\it IEEE Trans. Inf. Theory}, vol. 48, no. 3, pp. 580--598, 2002.
1241:
1242: \bibitem{EW02} T. Eggeling and R. F. Werner,
1243: \newblock ``Hiding Classical Data in Multipartite Quantum States,"
1244: \newblock {\it Phys. Rev. Lett.}, vol. 89, no. 9, 097905, 2002.
1245:
1246: \bibitem{CHE04} A. Chefles,
1247: \newblock ``Condition for unambiguous state discrimination using local operations and classical communication,"
1248: \newblock {\it Phys. Rev. A}, vol. 69, no. 5, 050307(Rapid communications), 2004.
1249:
1250: \bibitem{BW06} S. Bandyopadhyay and J. Walgate,
1251: \newblock ``Local distinguishability of any three quantum states,"
1252: \newblock quant-ph/0612013, 2006.
1253:
1254: \bibitem{NAT05} M. Nathanson,
1255: \newblock ``Distinguishing bipartitite orthogonal states using LOCC: best and worst cases,"
1256: \newblock {\it J. Math. Phys.}, vol. 46, no. 6, 062103, 2005.
1257:
1258: \bibitem{OH06} M. Owari and M. Hayashi,
1259: \newblock ``Local copying and local discrimination as a study for non-locality of a set,"
1260: \newblock {\it Phys. Rev. A}, vol. 74, no. 3, 032108, 2006. See also quant-ph/0411143 for a preliminary version.
1261:
1262: \bibitem{GKR+01} S. Ghosh, G. Kar, A. Roy, A. Sen(De), and U. Sen,
1263: \newblock ``Distinguishability of Bell states,"
1264: \newblock {\it Phys. Rev. Lett.}, vol. 87, no. 27, 277902, 2001.
1265:
1266: \bibitem{FAN04} H. Fan,
1267: \newblock ``Distinguishability and indistinguishability by local operations and classical communication,"
1268: \newblock {\it Phys. Rev. Lett.}, vol. 92, no. 17, 177905, 2004.
1269:
1270: \bibitem{WAT05} J. Watrous,
1271: \newblock ``Bipartite subspaces having no bases distinguishable by local operations and classical communication,"
1272: \newblock {\it Phys. Rev. Lett.}, vol. 95, no. 8, 080505, 2005.
1273:
1274: \bibitem{HK05} P. Hayden and C. King,
1275: \newblock ``Correcting quantum channels by measuring the environment,"
1276: \newblock {\it Quantum Inf. Comput.}, vol. 5, no. 2, 156--160, 2005.
1277:
1278: \bibitem{HMM+06} M. Hayashi, D. Markham, M. Murao, M. Owari, and S. Virmani,
1279: \newblock ``Bounds on multipartite entangled orthogonal state discrimination using local operations and classical communication,"
1280: \newblock {\it Phys. Rev. Lett.}, vol. 96, no. 4, 040501, 2006.
1281:
1282: \bibitem{DFJY07} R. Y. Duan, Y. Feng, Z. F. Ji, and M. S. Ying,
1283: \newblock ``Distinguishing arbitrary multipartite basis unambiguously using local operations and classical communication,"
1284: \newblock {\it Phys. Rev. Lett.}, vol. 98, no. 23, 230502, 2007.
1285:
1286: \bibitem{HSSH03} M. Horodecki, A. Sen(De), U. Sen, and K. Horodecki,
1287: \newblock ``Local indistinguishability: More nonlocality with less entanglement,"
1288: \newblock {\it Phys. Rev. Lett.}, vol. 90, no. 4, 047902, 2003.
1289:
1290: \bibitem{CL03} P.-X. Chen and C.-Z. Li,
1291: \newblock ``Orthogonality and distinguishability: Criterion for local distinguishability of arbitrary orthogonal states,"
1292: \newblock {\it Phys. Rev. A}, vol. 68, no. 6, 062107, 2003.
1293:
1294: \bibitem{COH07} S. M. Cohen,
1295: \newblock ``Local distinguishability with preservation of entanglement,"
1296: \newblock {\it Phys. Rev. A}, vol. 75, no. 5, 052313, 2007.
1297:
1298: \bibitem{FS07} Y. Feng and Y. Shi,
1299: \newblock ``Characterizing locally distinguishable orthogonal product states,"
1300: \newblock arXiv:0707.3581[quant-ph], 2007.
1301:
1302: \bibitem{GG07} V. Gheorghiu and R. B. Griffiths,
1303: \newblock ``Entanglement transformations using separable operations,"
1304: \newblock {\it Phys. Rev. A}, vol. 76, no. 3, 032310, 2007, a preliminary version is available at arXiv: 0705.0369v1[quant-ph].
1305:
1306: \bibitem{EB01} J. Eisert and H. J. Briegel,
1307: \newblock ``Schmidt measure as a tool for quantifying multiparticle entanglement,"
1308: \newblock {\it Phys. Rev. A}, vol. 64, no. 2, 022306, 2001.
1309:
1310: \bibitem{AACJ+00} A. Acin, A. Andrianov, L. Costa, E. Jane, J.I. Latorre, R. Tarrach,
1311: \newblock ``Generalized Schmidt Decomposition and Classification of Three-Quantum-Bit States,"
1312: \newblock {\it Phys. Rev. Lett.}, vol. 85, no. 7, 85, pp. 1560 - 1563, 2000.
1313:
1314: \bibitem{TH00} B. M. Terhal and P. Horodecki,
1315: \newblock ``Schmidt number for density matrices,"
1316: \newblock {\it Phys. Rev. A}, vol. 61, no. 4, 040301(Rapid communications), 2000.
1317:
1318: \bibitem{NOTE1} It was intrdouced in Ref. \cite{BDM+99} that a set of orthogonal product
1319: states $\mathcal{S}$ is completable if it can be extended into an
1320: orthogonal product basis of $\mathcal{H}$ (or local extension
1321: $\mathcal{H}_{ext}=\otimes_{k=1}^K(\mathcal{H}_k\oplus
1322: \mathcal{H}_k')$ of $\mathcal{H}$). Here we generalize this notion
1323: by allowing $\mathcal{S}$ is an arbitary set of orthogonal separable
1324: projecors such that the projector on the orthogonal complement of
1325: $\mathcal{S}$ is again separable.
1326:
1327: \bibitem{HSTT03} P. Horodecki, J. A. Smolin, B. M. Terhal, and A. V.
1328: Thapliyal,
1329: \newblock ``Rank two bipartite. bound entangled states do not exist,"
1330: \newblock {\it Theor. Comput. Sci.}, vol. 292, no. 3, pp. 589--596, 2003.
1331:
1332: \bibitem{WOO98} W. K. Wootters,
1333: \newblock ``Entanglement of formation of an arbitrary state of two qubits,"
1334: \newblock {\it Phys. Rev. Lett.}, vol. 80, no. 10, pp. 2245--2248, 1998.
1335:
1336: \bibitem{WIN05} A. Winter,
1337: \newblock ``On environment-assisted capacities of quantum channels,"
1338: \newblock arXiv:quant-ph/0507045v1, 2005.
1339: \end{thebibliography}
1340: \end{document}
1341: