PHYSICAL REVIEW RESEARCH 2, 043323 (2020)
Optimal verification of stabilizer states
Ninnat Dangniam,1,2,* Yun-Guang Han,1,2 and Huangjun Zhu 1,2,3,4,† 1Department of Physics and Center for Field Theory and Particle Physics, Fudan University, Shanghai 200433, China 2State Key Laboratory of Surface Physics, Fudan University, Shanghai 200433, China 3Institute for Nanoelectronic Devices and Quantum Computing, Fudan University, Shanghai 200433, China 4Collaborative Innovation Center of Advanced Microstructures, Nanjing 210093, China
(Received 25 July 2020; accepted 30 October 2020; published 4 December 2020)
Statistical verification of a quantum state aims to certify whether a given unknown state is close to the target state with confidence. So far, sample-optimal verification protocols based on local measurements have been found only for disparate groups of states: bipartite pure states, Greenberger-Horne-Zeilinger (GHZ) states, and antisymmetric basis states. In this work, we investigate systematically the optimal verification of entangled stabilizer states using Pauli measurements. First, we provide a lower bound on the sample complexity of any verification protocol based on separable measurements, which is independent of the number of qubits and the specific stabilizer state. Then we propose a simple algorithm for constructing optimal protocols based on Pauli measurements. Our calculations suggest that optimal protocols based on Pauli measurements can saturate the above bound for all entangled stabilizer states, and this claim is verified explicitly for states up to seven qubits. Similar results are derived when each party can choose only two measurement settings, say X and Z. Furthermore, by virtue of the chromatic number, we provide an upper bound for the minimum number of settings required to verify any graph state, which is expected to be tight. For experimentalists, optimal protocols and protocols with the minimum number of settings are explicitly provided for all equivalent classes of stabilizer states up to seven qubits. For theorists, general results on stabilizer states (including graph states in particular) and related structures derived here may be of independent interest beyond quantum state verification.
DOI: 10.1103/PhysRevResearch.2.043323
I. INTRODUCTION state σ is “close” to the target ρ with some confidence. More precisely, the verification scheme accepts a density operator Engineered quantum systems have the potential to effi- σ that is close to the target state with the worst-case fidelity ciently perform tasks that are believed to be exponentially 1 − and confidence 1 − δ. In other words, the probabil- difficult for classical computers such as simulating quantum ity of accepting a “wrong” state σ with |σ | 1 − systems and solving certain computational problems. With the is at most δ. For the convenience of practical applications, potential comes the challenge of verifying that the quantum usually the verification protocols are constructed using lo- devices give the correct results. The standard approach of cal operations and classical communication (LOCC). Such quantum tomography accomplishes this task by fully char- verification protocols have been gaining traction in the quan- acterizing the unknown quantum system, but with the cost tum certification community [7–19] because they are easy exponential in the system size. However, rarely do we need to implement and potentially require only a small number to completely characterize the quantum system as we often of copies of the state. However, sample-optimal protocols have a good idea of how our devices work, and we may only under LOCC have been found only for bipartite maximally need to know if the state produced or the operation performed entangled states [2,3,12], two-qubit pure states [13], n-partite is close to what we expect. The research effort to address Greenberger-Horne-Zeilinger (GHZ) states [17], and most re- these questions have grown into a mature subfield of quantum cently antisymmetric basis states [18]. certification [1]. Maximally entangled states and GHZ states are subsumed Statistical verification of a target quantum state ρ = under the ubiquitous class of stabilizer states, which can be | | [2–6] is an approach for certifying that an unknown highly entangled yet efficiently simulatable [20,21] and ef- ficiently learnable [22] under Pauli measurements. Another notable example of stabilizer states are graph states [23,24], *[email protected] which have simple graphical representations that transform †[email protected] nicely under local Clifford unitary transformations. They find applications in secret sharing [25], error correcting codes Published by the American Physical Society under the terms of the [26,27], and cluster states in particular are resource states Creative Commons Attribution 4.0 International license. Further for universal measurement-based quantum computing [28]. distribution of this work must maintain attribution to the author(s) Stabilizer states and graph states can be defined for multiqudit and the published article’s title, journal citation, and DOI. systems with any local dimension; nevertheless, multiqubit
2643-1564/2020/2(4)/043323(36) 043323-1 Published by the American Physical Society DANGNIAM, HAN, AND ZHU PHYSICAL REVIEW RESEARCH 2, 043323 (2020) stabilizer states are the most prominent because most quantum suggests that the maximum spectral gap achievable by X information processing tasks build on multiqubit systems. In and Z measurements is 1/2. For the ring cluster state, we this paper, we only consider qubit stabilizer states and graph prove this result rigorously by constructing an explicit optimal states unless stated otherwise, but we believe that many results verification protocol. We also prove that three settings based presented here can be generalized to the qudit setting as long on Pauli measurements (or X and Z measurements) are both as the local dimension is a prime. Efficient verification of necessary and sufficient for verifying the odd ring cluster state stabilizer states have many applications, including but not with at least five qubits. limited to blind quantum computing [29,30] and quantum gate In the course of study, we generalize the concepts of verification [31–33]. canonical test projectors, admissible Pauli measurements, and While one might expect that the determination of an opti- admissible test projectors, first introduced in Ref. [17]for mal verification strategy to be difficult in general, one could GHZ states, to arbitrary stabilizer states. We also clarify their hope for the answer for stabilizer states in view of their basic properties, which are of interest to quantum state ver- relatively simple structure. Given an n-qubit stabilizer state, ification in general. Meanwhile, we introduce several graph Pallister, Montanaro, and Linden [4] showed that the optimal invariants that are tied to the verification of graph states and strategy when restricted to the measurements of nontrivial clarify their connections with the chromatic number. In ad- stabilizers (to be introduced below) is to measure all 2n − 1 dition to their significance to the current study, these results of them with equal probabilities, which yields the optimal provide additional insights on stabilizer states and graph states constant scaling of the number of samples, themselves and are expected to find applications in various other related problems. n − δ−1 δ−1 ≈ 2 1 ln ≈ 2ln . The rest of this paper is organized as follows. First, we N − (1) 2n 1 present a brief introduction to quantum state verification in One could choose to measure only n stabilizer generators at Sec. II and preliminary results on the stabilizer formalism the expense of now a linear scaling [4]: in Sec. III. In Sec. IV, we study canonical test projec- tors and admissible test projectors for stabilizer states and n ln δ−1 graph states and clarify their properties. In Sec. V, we de- N ≈ . (2) rive an upper bound for the spectral gap of verification operators based on separable measurements. Moreover, we This trade-off is not inevitable in general. Given a graph state propose a simple algorithm for constructing optimal verifica- associated with the graph G, by virtue of graph coloring, tion protocols based on Pauli measurements and provide an χ Ref. [7] proposed an efficient protocol which requires (G) explicit optimal protocol for each connected graph state up to χ −1 δ−1 measurement settings and (G) ln samples. Here seven qubits. In Sec. VI, we discuss optimal verification of χ the chromatic number (G)ofG is the smallest number of graph states based on X and Z measurements. In Sec. VII, colors required so that no two adjacent vertices share the same we consider verification of graph states with the minimum color. With this protocol, one can verify two-colorable graph number of settings. Section VIII summarizes this paper. To states, such as one- or two-dimensional cluster states, with streamline the presentation, the proofs of several technical −1 δ−1 2 ln tests. results are relegated to the Appendices, which also contain In this paper, we study systematically optimal verification Tables I and II. of stabilizer states using Pauli measurements. We prove that the spectral gap of any verification operator of an entangled stabilizer state based on separable measurements is upper II. STATISTICAL VERIFICATION / bounded by 2 3. To verify the stabilizer state within infidelity A. The basic framework and significance level δ, therefore, the number of tests re- quired is bounded from below by Let us formally introduce the framework of statistical ver- ification of quantum states. Suppose we want to prepare the 1 3lnδ−1 target state ρ =| |, but actually obtain the sequence N = ln δ ≈ . (3) σ ,...,σ ln[1 − 2/3] 2 of states 1 N in N runs. Our task is to determine whether all of these states are sufficiently close to the target Moreover, we propose a simple algorithm for constructing state on average (with respect to the fidelity, say). Following optimal verification protocols of stabilizer states and graph Refs. [4–6], we perform a local measurement with binary out- states based on nonadaptive Pauli measurements. An optimal comes {E j, 1 − E j}, labeled as “pass” and “fail” respectively, protocol for each equivalent class of graph states with respect on each state σk for k = 1,...,N with some probability p j. to local Clifford transformations (LC) and graph isomorphism Each operator E j is called a test operator. Here we demand is presented in Table I in the Appendix, and our code is avail- that the target state ρ can pass the test with certainty, which able in Ref. [34]. These results suggest that for any entangled means E jρ = ρ. The sequence of states passes the verification stabilizer state the bound in (3) can be saturated by protocols procedure iff every outcome is “pass.” built on Pauli measurements. The efficiency of the above verification procedure is deter- In addition, we study the problem of optimal verification mined by the verification operator based on X and Z measurements and the problem of verifi- cation with the minimum number of measurement settings. m = , The two problems are of interest in certain scenarios in which p jE j (4) the accessible measurement settings are restricted. Our study j=1
043323-2 OPTIMAL VERIFICATION OF STABILIZER STATES PHYSICAL REVIEW RESEARCH 2, 043323 (2020) where m is the total number of measurement settings. If the fi- which implies that delity |σk| is upper bounded by 1 − , then the maximal σ ν() min pk 1/m. (9) average probability that k can pass each test is [4,6] k max tr(σ ) = 1 − [1 − β()] = 1 − ν(). (5) Here the second inequality is saturated iff p = 1/m for all |σ | 1− k k. Here, β() is the second largest eigenvalue of the verification operator , and ν():= 1 − β()isthespectral gap from C. Verification of a tensor product the maximum eigenvalue. | Now and throughout the paper, we assume that the states Suppose the target state is a tensor product of the form σ ,σ ,...,σ | = J | | 1 2 N received in different runs are independent of j=1 j , where J 2 and each tensor factor j each other (more general situations can be dealt with using could be a multipartite state and either separable or entangled. the recipe proposed in Refs. [5,6]). Then these states can pass It is instructive to clarify the relation between verification N tests with probability at most operators of | and that of each tensor factor. Given a verification operator for | ,thereduced verifi- N N | N cation operator of for the tensor factor j is defined as tr(σ j ) [1 − ν() j] [1 − ν()¯] , (6) j=1 j=1 := || , (10) j j j = / = − |σ | where ¯ j j N with j 1 j is the average where | := | . Note that | =| ,so infidelity [5,6]. If N tests are passed, then we can ensure the j j = j j j j j j is indeed a verification operator for | .If is separable, condition< ¯ with significance level δ = [1 − ν()]N .To j then each is also separable. Reduced test operators can be verify these states within infidelity and significance level δ, j defined in a similar way. the number of tests required is [4–6] Proposition 2. Suppose is a verification operator for | = J | = , ,..., 1 1 1 j=1 j , and j for j 1 2 J are reduced N(,δ,) = ln δ ln . (7) ln[1 − ν()] ν() δ verification operators of . Then the second largest eigen- value and spectral gap of satisfy the following inequalities: If there is no restriction on the measurements, the optimal performance is achieved by performing the projective mea- β() max β( j ),ν() min ν( j ). (11) jJ jJ surement onto the target state | | itself, which yields 1 1 − ν() = 1 and N =ln δ/ln(1 − ) ln δ 1/ as the ul- Proof. timate efficiency limit allowed by quantum theory.1 β( j ) = max j| j| j | j : j | j =0 B. Minimal sets of test operators = max ( j|⊗ j|)(| j ⊗| j ) { }m | | : | =0 A set of test operators E j j=1 for is minimal if any j j j { }m | proper subset of E j j=1 cannot verify reliably because max || =β(), (12) the common pass eigenspace of operators in the subset has | : | =0 dimension larger than one. A minimal set of test operators has which implies (11). the following properties. m Conversely, suppose are verification operators for Proposition 1. Suppose = p E is a verification j j=1 j j | with spectral gap ν( )for j = 1, 2,...,J.Let = operator based on a minimal set of m test operators. Then j j J | ν() 1/m. If the inequality is saturated then p = 1/m for j=1 j; then is a verification operator for , and j j all j. are reduced verification operators of by the definition in Proof. By assumption, for each k ∈{1, 2,...,m}, there (10). Straightforward calculation shows that the spectral gap of reads exists a pure state |k that is orthogonal to the target state | and belongs to the pass eigenspace of E j for all j = k, ν() = min ν( j ), (13) that is, E j|k =|k . Therefore 1 jJ β() k||k p j k|E j|k = p j which saturates the upper bound in (11). In addition, if each j=k j=k j can be realized by LOCC (Pauli measurements), then so can . On the other hand, the number of distinct test opera- = 1 − p ∀k, (8) k tors (measurement settings) required to realize (naively as suggested by the definition) increases exponentially with the number J of tensor factors. It is of practical interest to reduce this number. 1A related certification framework by Kalev and Kyrillidis [35]for Suppose can be realized by the set of test operators stabilizer states is, in a sense, opposite to ours. In our framework, j { ( j)}m j = m j ( j) ( j) ( j) m j we are given the worst case fidelity and are asked to find an optimal Ek k=1, that is, j k=1 pk Ek , where (pk )k=1 is a | measurement, whereas in their work we are given a (stabilizer) mea- probability vector. In addition, j is the unique common − ( j) = surement and are asked to bound the worst case fidelity 1 to the eigenstate of Ek with eigenvalue 1. Let m max j m j; then desired stabilizer state within some radius r (their “”). | can be reliably verified by the following set of test
043323-3 DANGNIAM, HAN, AND ZHU PHYSICAL REVIEW RESEARCH 2, 043323 (2020) operators the symplectic form J T 0 −1n = ( j), = , ,..., . [μ, ν]:= μ Jν, J = . (19) Ek : Ek k 1 2 m (14) 1n 0 j=1 Let μx (μz) be the vector composed of the first (last) n ( j) = < μ μ = μx μz Here we choose Ek 1 for m j k m, so that each test Ek elements of ; then ( ; ), where the semicolon de- can be realized by virtue of the original test for each tensor notes the vertical concatenation. In addition, we have [μ, ν] = factor or the trivial test (with the test operator equal to the μz · νx + μx · νz. (Addition and subtraction are the same in identity). To verify this claim, first note that Ek| =| for arithmetic modulo 2.) The Weyl representation [36] of each = , ,..., | μ ∈ 2n k 1 2 m, so each Ek is a test operator for . Suppose vector Z2 yields a Pauli operator | is a common eigenstate of all Ek with eigenvalue 1, that μx·μz μx μz | | = ρ = | | g(μ) = i X Z , (20) is, Ek 1. Let j tr j¯( ), where tr j¯ denotes the partial trace over all tensor factors except for the jth factor. x μx μx z μ = 1 ⊗···⊗ n μ where X X1 Xn and similarly for Z . Each Then we have P k μ = , , , Pauli operator in n is equal to i g( )fork 0 1 2 3 and ρ ( j) = | ( j) ⊗ | | | = μ ∈ 2n = = 1 tr jEk Ek 1 Ek 1 (15) Z2 . For example, we have X g((1;0)), Z g((0;1)), and Y = g((1; 1)). By the following identity for all j, k. This equation implies that tr(ρ E ( j)) = 1, so each j k μ ν = [μ,ν] μ + ν = 2[μ,ν] ν μ , ρ ( j) g( )g( ) i g( ) i g( )g( ) (21) j is supported in the eigenspace of Ek with eigenvalue 1 for all k.Itfollowsthatρ j =| j j| and | |=| |,so g(μ) and g(ν) commute iff [μ, ν] = 0. | 2n is the unique common eigenstate of all Ek with eigenvalue The symplectic complement of a subspace W in Z2 is 1 and it can be reliably verified by the test operators in (14). defined as m Let (qk ) = be any probability vector with qk > 0for k 1 W ⊥ = μ ∈ Z2n | [μ, ν] = 0, ∀ν ∈ W . (22) all k (sufficient number of settings are measured) and = 2 m | ⊥ k=1 qkEk; then is a verification operator for with The subspace W is isotropic if W ⊂ W , in which case ν > ( ) 0 according to the above discussion. In addition, the [μ, ν] = 0 for all μ, ν ∈ W . Hence all Pauli operators associ- | reduced verification operator of for tensor factor j reads ated with vectors in W commute with each other. The maximal m dimension of any isotropic subspace is n, and such a maximal = || = ( j). = ⊥ j j j qkEk (16) isotropic subspace satisfies the equality W W and is called k=1 Lagrangian. Each isotropic subspace W of dimension k is determined by a 2n × k basis matrix over Z2n whose columns According to proposition 2,wehave 2 form a basis of W . Conversely, a 2n × k matrix M is a basis m m matrix for an isotropic subspace if the following condition ν ν = ν ( j) ν ( j) , ( ) ( j ) qkEk max qkEk holds: (q )k k=1 k k=1 T (17) M JM = 0k×k. (23) where the maximization is taken over all probability vectors Two Lagrangian subspaces W and W of Z2n are complemen- with m components. The right-hand side coincides with the 2 tary if their intersection is trivial (consists of the zero vector maximum spectral gap achievable by any verification operator only), in which case Span(W ∪ W ) = Z2n. of | that is based on the set of test operators {E ( j)}m j .Note 2 j k k=1 The Clifford group is the normalizer of the Pauli group P . q = m < k m n that k 0for j when the maximum spectral gap Up to phase factors, it is generated by phase gates, Hadamard ( j) = < is attained given that Ek 1 for m j k m. gates for individual qubits and controlled-not gates for all pairs of qubits. Its quotient over the Pauli group is isomorphic III. STABILIZER FORMALISM to the symplectic group with respect to the symplectic form in (19). A. Pauli group H = 2 ⊗n Let (C ) be the Hilbert space of n qubits. The B. Stabilizer codes Pauli group for one qubit is generated by the following three S P S matrices: A subgroup of n is a stabilizer group if is com- mutative and does not contain −1. Since S cannot contain − = 01, = 0 i , = 10. a Pauli operator with phases ±i (otherwise −1 ∈ S), every X Y Z − (18) 10 i0 0 1 element except the identity has order 2. Thus S is isomorphic k k The n-fold tensor products of Pauli matrices and the identity to an elementary abelian group Z2 of order 2 , where k n {1, X,Y, Z}⊗n form an orthogonal basis for the space B(H) is the cardinality of a minimal set of generators. Suppose of linear operators on H. Together with the phase factors that the stabilizer group S is generated by the k generators S , S ,...,S ; then the elements of S can be labeled by {±1, ±i}, these operators generate the Pauli group Pn, which 1 2 k n+1 k has order 4 . Two elements of the Pauli group either com- vectors in Z2 as follows: mute or anticommute. k y = y j , ∈ k . Up to phase factors, n-qubit Pauli operators can be labeled S S j y Z2 (24) 2n by vectors in the binary symplectic space Z2 endowed with j=1
043323-4 OPTIMAL VERIFICATION OF STABILIZER STATES PHYSICAL REVIEW RESEARCH 2, 043323 (2020)
The stabilizer code HS of S is the common eigenspace of between S and S reads eigenvalue 1 of all Pauli operators in S, which has dimen- n−k 1 sion 2 . Alternatively, it is also defined as the common S S = SS tr( ) |S|·|S | tr( ) eigenspace of eigenvalue 1 of the k generators S1, S2,...,Sk. S∈S,S ∈S The projector onto the code space reads n = 2 |S ∩ S |−| −S ∩ S | |S|·|S | ( ( ) ) k + 1 1 S j = = . n S S (25) 2 |S¯∩S | S ∩ S = S¯ ∩ S , |S| 2 |S|·|S | S∈S j=1 = (29) 0 otherwise. Conversely, S happens to be the group of all Pauli operators in S ∩ S S¯ ∩ S S ∩ P that stabilize the stabilizer code HS . So there is a one-to- Note that is a subgroup of of index 2 if n S = S¯ ∩ S one correspondence between stabilizer groups and stabilizer . Equation (29) implies the following equation codes. To later establish the relation between stabilizer groups 2n|S¯∩S | S ∩ S = S¯ ∩ S , |S|·|S | w w and isotropic subspaces, we introduce the signed stabilizer tr( S,w S ,w ) = (30) 0 otherwise. group of the stabilizer code HS to be the union k for all w ∈ Z given that S¯ = S¯ and S¯ ∩ S = S¯ ∩ S .In S¯ := S ∪ (−S), (26) 2 w w addition, we have where −S :={−S|S ∈ S}. 2n S , ,..., tr( S, S ) = tr( S ) = (31) Given the stabilizer group with generators S1 S2 Sk, w |S | for each w ∈ Zk , define S as the group generated by w 2 w − w j = , ,..., S ( 1) S j for j 1 2 k; then w is also a stabilizer thanks to the equality S, = 1. So the number of H S w w group. The stabilizer code S of w is the common ∈ n = w vectors w Z2 at which tr( S,w S ) 0 is equal to , ,..., − w j = eigenspace of S1 S2 Sk with eigenvalue ( 1) for j |S|/|S¯ ∩ S |. 1, 2,...,k and is also denoted by HS, . The projector onto w Lemma 1. Suppose S j, T j are stabilizer groups on H j for the stabilizer code reads j = 1, 2,...,J;letS = S1 × S2 ×···×SJ and T = T1 × n T ×···×T H ⊗ H ⊗···⊗H + − w j 2 J be stabilizer groups on 1 2 J 1 ( 1) S j y y ¯ S,w = = χw(S )S , (27) and let T be the signed stabilizer group associated with T . = 2 ∈ n j 1 y Z2 Then S ∩ T¯ = L × L ×···×L , where 1 2 J (32) y w·y where L = S ∩ T¯ with T¯ being the signed stabilizer groups χw(S ) = χw(y) = (−1) (28) j j j j associated with T¯j. S k can be understood as a character on or Z2 [37]. Note that Proof. To simplify the notation, here we prove (32)inthe H ∈ k = all stabilizer codes S,w for w Z2 share the same signed case J 2; the general case can be proved in a similar way. stabilizer group, that is, S¯w = S¯. Any S ∈ S has the form S = S1 ⊗ S2 with S1 ∈ S1 and S2 ∈ According to the Weyl representation in (20), each ele- S2. If in addition S ∈ T¯ , then S1 ∈ T¯1 and S2 ∈ T¯2. Therefore ment in the stabilizer group S is equal to g(μ)or−g(μ) S1 ∈ L1 and S2 ∈ L2, so that S ∈ L1 × L2, which implies that μ ∈ 2n S S ∩ T¯ ⊆ L × L for Z2 . In this way, is associated with an isotropic 1 2. ⊂ 2n ∈ L × L = ⊗ subspace W Z2 of dimension k, and there is a one-to-one Conversely, any S 1 2 has the form S S1 S2 correspondence between elements in S and vectors in W . with S1 ∈ S1 ∩ T¯1 and S2 ∈ S2 ∩ T¯2, which implies that S ∈ S ¯ Suppose the k generators S1, S2,...,Sk of S correspond to and S ∈ T¯ . Therefore L1 × L2 ⊆ S ∩ T , which confirms (32) μ ,μ ,...,μ the k symplectic vectors 1 2 k, which form a basis in view of the opposite inclusion relation derived above. y μ = in W . Then S corresponds to the vector j y j j My for ∈ k = μ ,μ ,...,μ each y Z2, where M : ( 1 2 k ) is a basis matrix C. Stabilizer states for W . Note that all the stabilizer groups S for w ∈ Zk are w 2 When the stabilizer group S is maximal, that is, |S|=2n, associated with the same isotropic subspace according to the the stabilizer code HS has dimension 1 and is represented by above correspondence, and this correspondence extends to |S S¯ a normalized state called a stabilizer state and denoted by . the signed stabilizer group . Conversely, given an isotropic |S S k Note that is uniquely determined by up to an overall subspace W of dimension k,2 stabilizer groups can be con- |S k phase factor. According to (25), the projector onto reads structed as follows. Let {μ j} = be any basis for W ; for each j 1 n k + vector a in Z , a stabilizer group can be constructed from 1 1 S j 2 |S S|= S = S = , (33) a j n the k generators (−1) g(μ j )for j = 1, 2,...,k. All these 2 2 S∈S j=1 stabilizer groups extend to a common signed stabilizer group. In this way, there is a one-to-one correspondence between where S1, S2,...,Sn are a set of generators of S. For each ∈ n S − w j signed stabilizer groups and isotropic subspaces. w Z2, define w as the group generated by ( 1) S j for
Suppose S and S are two n-qubit stabilizer groups of j = 1, 2,...,n; then Sw is also a maximal stabilizer group. k k orders 2 and 2 , respectively; let S and S be the projec- In addition, the associated stabilizer state |Sw is the com- w j tors onto the corresponding stabilizer codes. Then the overlap mon eigenstate of S1, S2,...,Sn with eigenvalue (−1) for
043323-5 DANGNIAM, HAN, AND ZHU PHYSICAL REVIEW RESEARCH 2, 043323 (2020) y n y j n S S j = 1, 2,...,n.LetS = = S for y ∈ Z as in (24) with Lemma 2. Suppose and are two maximal stabilizer j 1 j 2 k = n; then the projector onto |Sw reads groups with basis matrix M and M , respectively. Then T |S¯ ∩ S |=|ker(MT JM )|=2n−rank(M JM ),
n n−rank(MT M +MT M ) + − w j = 2 z x x z , (39) 1 ( 1) S j w·y y |S S |= S, = = (−1) S , w w w 2 j=1 y∈Zn 2 −n T 2 max | Sw|S | = 2 | ker(M JM )| ∈ n w (34) w Z2 which reduces to (33) when w = 00 ···0. The set of stabilizer − MT JM − MT M +MT M = 2 rank( ) = 2 rank( z x x z ), states |S for w ∈ Zn forms an orthonormal basis in H, w 2 (40) known as a stabilizer basis. Stabilizer bases are in one-to- 2n one correspondence with Lagrangian subspaces in Z2 . Based where Mx (Mz) denotes the submatrix of M composed of the on this observation, one can determine the total number of first (last) n rows, and Mx (Mz) is defined in a similar way. n-qubit stabilizer states, with the result [36] D. Graph states