Arxiv:1908.10390V1 [Quant-Ph] 27 Aug 2019 Il Xml Fsc Lsicto Oe Through Comes Classiﬁcation a Such Pos- a of Considered

arXiv:1908.10390v1 [quant-ph] 27 Aug 2019 il xml fsc lsicto oe through comes classification a such pos- A of considered. example applications sible the on depending ment reductions. party two in the contained entanglement while average [12], en- three-tangle three-party state called important an measure to tanglement respect with entangled aiie ydffrn tts[0 1.Frisac,the instance, For 11]. state [10, three–qubit are states and different orderings by en- different maximized different as re- induce is entanglement measures of there quantify tanglement field subsystems to more active way or unique an three an no for still be Already is to [8]. and search proven subject topic has rich the systems extremely entan- many-particle particular, the of quantifying In glement and much classifying intricate. characterizing, is of and set-up multipartite complex is other the more the systems in On bipartite entanglement [5]. in quantified hand, entanglement and understood that well say already to fair is its 7]. for [6, entangle- methods proposed were quantum different detection and of experimental [5] aspects investigated various were computation Therefore ment quantum and 4]. [2] [3, dense codes of correcting stone error distribution [1], pillar key the quantum teleportation, is quantum It coding, technologies. quantum for ∗ † [email protected] [email protected] hrfr,i sraoal ocaatrz entangle- characterize to reasonable is it Therefore, fe oeta w eae fitniersac,it research, intensive of decades two than more After resources fundamental most the of one is Entanglement ailoinUiest,Mra mlcosiIsiuef Institute Smoluchowski Marian University, Jagiellonian | W ailoinUiest,Mra mlcosiIsiuef Institute Smoluchowski Marian University, Jagiellonian etrfrTertclPyis oihAaeyo Sciences of Academy Polish Physics, Theoretical for Center u-eitn ik n utprieetnlmn resist entanglement multipartite and links Cut-resistant nttt eTlcmnç˜e,Pyiso Informations Telecomunica¸cões, de of Physics Instituto etod srfıiaeGaiaca ETA Departamen Gravita¸cão Astrof´ısica CENTRA, e de - Centro i ( = fdsgigatplgclln ossigof W consisting m. link than larger topological particles a of designing number of a losing after separable h eann ig eoedsonce.W rsn const distinguished a several present exhibit We to us disconnected. allows become which rings remaining the nageetrssac oapril loss. particle a to resistance entanglement nageetrssatt atce os eitouethe introduce we loss, particles to resistant entanglement tt is state | nti ok eepoetesaeo unu ttscomposed states quantum of space the explore we work, this In 100 .INTRODUCTION I. | GHZ i m + rssati trmisetnldatrlsn narbitrar an losing after entangled remains it if -resistant nvriaed iba-U,A.RvsoPi ,14-0 Lis 1049-001 1, Pais Rovisco Av. UL, - Lisboa de Universidade | i 010 ( = i | + 000 | 001 i + i | 111 ) / √ i aiie the maximizes 3 ) / √ stemost the is 2 Dtd uut2,2019) 27, August (Dated: o¸ aoM Quinta M. Gon¸calo Karol dmBurchardt Adam u André Rui N Zyczkowski ig uhta,atrctigay( any cutting after that, such rings ˙ rPyis Ljseiz 1 038Ka´w oadand Kraków, Poland 30-348 11, Lojasiewicza Physics, or N n unu ehooisGop iba Portugal Lisboa, Group, Technologies Quantum and u ro orcincds[18]. codes quan- are correction of construction error states for and tum AME and [17] teleportation The sharing secret multipartite quantum 16]. for bi- [15, relevant possible system particularly all the to of respect partitions with entangled which maximally concept states, are the (AME) to Entangled are a leads Maximally it procedure Absolutely how of a of parts Such analyzing, some ignored. if by or affected altered domain is state multipartite multipartite given the to ceed multipartite for the out applied wipe directly to amountsystems. be needed minimal can states the separable entanglement, as with [14] mixing systems de- of For originally bipartite parties. entanglement, for of of fined be number robustness can larger the a subsystems instance, for two generalized for directly proposed entanglement. multipartite approaches shedding of Some years, aspects different recent into in light developed were entanglement finite alternative find many different to schemes. on order infinitely classification focus in to states are necessary quantum thus out there of is traits worked It qubits been [13]. four classes have such for classes already SLOCC for [10], all although then However, motivates qubits may which classification. three protocols, of class scheme same SLOCC very the this perform same to the used be in communications. obtained belonging classical be only with States using can assisted probability they operations finite if some local with equivalent another be one consid- to from one states [9], two approach ers this In Com- (SLOCC). Classical munications with assisted Operations Local Stochastic rPyis Ljseiz 1 038Kraków, Poland 30-348 11, Lojasiewicza Physics, or prilssae ihtedsrdpoet of property desired the with states -particles trigfo h oino h elsaeoecnpro- can one state Bell the of notion the from Starting quantum of features complementary and useful Many l onko 24,0-6 asaa Poland Warszawa, 02-668 Lotników 32/46, Al. , † od ´sc,IsiuoSpro ćio-IST, Técnico - Superior F´ısica, Instituto de to oinof notion ∗ sals naaoyt h problem the to analogy an establish e usto atce,btbecomes but particles, m of subset y utv ouint hsproblem, this to solution ructive of m N rssatstates -resistant atce.T netgt the investigate To particles. o,Portugal boa, n opril loss particle to ant m )o them, of 1) + quantum A . 2

A natural generalization of AME states are the so Making use of the analogy between entangled states called k-uniform states, whereby the partial tracing of and linked rings [28] we investigate first a topological any subsystems down to k qudits will result in a max- question of linking N closed rings in such a way that imally mixed state [19, 20]. Such states can be con- after cutting any m of them the remaining rings are constructed with help of combinatorial tools like orthogonal nected, whereas cutting any additional ring separates the arrays [21], which allow for a coarse-grained classifica- remaining rings. This problem is solved for any numbers tion of multipartite entanglement [22]. Another related N and m with use of algebraic techniques developed in measure, called the persistence of entanglement [23], is [30] which allow one to associate to any link of N rings defined by the minimal number of local measurements with a polynomial of N variables. A formalism developed such that the state becomes completely disentangled for for the study of the links can be used not only to gain any measurement outcome. This property provides a new intuition and pictorial visualization of the entangle- somewhat intuitive notion of entanglement strength, by ment involved, but also to find representative states for looking into the minimal requirements to destroy it com- each class of m-resistant quantum states, in a way which pletely. bypasses a great deal of algebra. A similar idea was used first by Aravind [24], who pro- This work is organized as follows. In Section II the posed to relate some N-partite quantum states with a topological notion of m-resistant link of N rings is intro- link composed of N closed rings. After any measurement duced while the analogy to multipartite quantum entan- performed on any subsystem of the state GHZ the re- glement resistant to a loss of a subsystem is presented in maining bipartite state becomes separable.| Thusi this Sec. III. Exemplary constructions of m-resistant states state can be associated with a particular configuration of of three and four qubits are provided in Section IV. In three rings, called Borromean rings – see Fig. 2, distin- the case of pure states symmetric with respect to per- guished by the fact that if one ring is cut, the remaining mutation of subsystems it is useful to use the stellar two become disconnected. Usage of other tools borrowed representation of Majorana [31–33]. The search for m- from knot theory to analyze multipartite entanglement resistant states of N subsystems of local dimension d each was further advocated in [25, 26]. In fact, Borromean is discussed in Section V, while the last Section presents rings and their generalizations interacted with various concluding remarks and lists some open questions. In branches of physics, being both an inspiration and a trig- Appendix we list exemplary quantum states obtained ger of their development [27]. with use of the combinatorial notion of orthogonal ar- An alternative interpretation of cutting the rings was rays and prove a topological proposition on existence of proposed by Sugita [28], who suggested associating in- m-resistant links of N rings. dividual subsystems with rings, entangled states with linked rings, and the partial trace over a given subsystem with the act of cutting and removing the associated II. M-RESISTANT LINKS ring. In such a way, the physical process corresponding to cutting the ring does not depend on the outcomes The question in how many different ways one may link of a measurement and is defined uniquely [29]. Such an any N rings has been addressed in [30]. We ignore (or analogy between quantum entanglement and linked rings, cut) each subset of rings and ask whether the remaining used later in [30], will be further explored here. link is connected. We do not take into account other The aim of this work is to refine the classification superfluous details of the link. In this sense, two rings of multipartite entanglement by introducing the notion can only be connected in one way because they are either of m-resistant states. Entanglement of these N-partite connected or not, while from a knot theory perspective states is preserved even if any m subsystems are traced this could be done in infinitely many different ways. away, while removing any further subsystem makes the To formalize the above idea, a particular configura- state of the remaining N m 1 subsystems completely tion of any number of linked rings is characterized by a − − polynomial, denoted as . The construction of this poly- separable. Despite sharing the spirit of looking into what P happens when certain parts of a state are ignored, as for nomial starts by associating a variable to each ring and AME states and k-uniform states, the above definition then interpreting the product between variables to be does not impose the restriction of maximal entanglement, equivalent to the associated rings being linked. Taking a which consequently increases the complexity of the prob- variable to 0 is then interpreted as cutting or ignoring the lem. associated ring. Thus, a given link can be characterized by the remaining links produced from all possible cuts. On the other hand, this approach shares attributes As an example, consider the polynomial with persistence and robustness of entanglement, in that it measures how the entanglement of a given state resists (33)= ab + bc . (1) ignoring a certain number of non–accessible subsystems, P which corresponds to taking the partial trace over them. We use the notation ni, introduced in [30], where n is This process is not as invasive as performing local mea- the number of rings and i is a cataloging index denoting surements on selected subsystems or mixing the analyzed the specific link class. Eq. (1) implies that the ring a is state with locally prepared separable states. linked to the ring b, which is also linked with the ring c, 3 while the rings a and c are not directly connected. This III. LINKS AND QUANTUM STATES can be seen by ignoring the ring a, i.e. taking a = 0, which leaves us with bc, meaning the two rings b and c Picking up from the previous section, an analogy can are still linked. The same thing happens when the ring now be made by identifying a ring to a particle and at- c is ignored, which leaves the term ab, implying that the tributing the act of ignoring a ring to the operation of rings a and b remain connected. The link associated with tracing out the particle [28], as if that subsystem has not this configuration is represented in Fig. 1. been detected. Linked rings are then associated with entangled particles, while separated rings represent separable particles. Distinct links thus represent distinct entanglement classes. Since the trace is used as the connecting ingredient, the entire analogy is basis independent. In addition, since the partial trace is defined for arbitrary Hilbert spaces, we may consider composite quantum sys- Figure 1: An element of the link class 33, classified by tems of any dimensionality. In this paper, we start work- the polynomial ab + bc, given in (1). ing with N-qubit systems, but in Section V we discuss also systems with a larger local dimension d. This line of interpretation naturally induces a set of The “link - quantum entanglement” analogy goes both rules for the construction of valid polynomials, which are ways: one may find the link class associated with a cer- summarized as follows: tain state, or; one may look for a state possessing the entanglement properties of a specific link. The former 1) There must not be any repeated terms, i.e. no ring problem is straightforward to answer: one adds a term variable can have a power greater than 1 (e.g. aab with k variables to the polynomial whenever the subsys- is the same as ab); tem with the associated k particles is entangled. The 2) Each ring variable must appear at least once; latter problem, as we shall show, is much more intricate but can usually be solved using maps between polynomi- 3) There must not be first order terms (i.e. we want als and quantum states. any ring to be linked with at least another one); The notion of m-resistant links of N rings introduced 4) Relabeling of variables is irrelevant; in the previous section naturally motivates to distinguish the following class of entangled quantum states of N 5) An n-variable monomial M is irrelevant if all of subsystems. its variables are already present as an n-ring link of lesser order monomials, built only with the vari- Definition: (m-resistance) An entangled state of N ables of M (e.g. abc+ac+ab gives the same results parties is called m-resistant if: as ac + ab after each variable is taken to 0, so we It remains entangled as any m of its N subsystems keep only ac + ab). • are traced away; These rules for polynomial construction can then be used It becomes separable if a partial trace is performed to count the number of distinct ways to link a given num- • ber of rings. For N = 2, 3, 4 and 5 rings, there are over an arbitrary set of m + 1 subsystems. respectively 1, 4, 40 and 6900 ways to form a link [30], The first nontrivial example of m-resistance arises for while for an arbitrary N the corresponding number has N = 3, which is simple enough to study on intuitive not been found yet. grounds. The case m = 0 is exemplified by the GHZ A particularly interesting subclass of links with N rings state [10–12] is one where all rings remain connected after any m < N cuts, but become fully separated if m + 1 cuts are 1 GHZ = ( 000 + 111 ) , (2) performed. Links with this property will be called m- | i √2 | i | i resistant, since they resist up to m cuts before becoming fully disconnected. Clearly, 0 mBrunnian link, diately implies that the polynomial must be the sum of with general polynomial all possible N m letter terms. For N rings, there will be N 1 classes− of different resistances. (N 1)= abcd , (4) − P ··· 4

any partial trace, it is clear that at least these states can be symmetric regarding exchanges of subsystems. This is not strictly necessary, as symmetry under qubit permutation is not invariant under local operations, but it does provide one possible direction. However, the requirement of separability, or lack of it, is not simple to achieve using any standard algebraic techniques. In this section we will demonstrate how the analogy between links of rings and entangled quantum states may be used in order to find representative mixed states for Figure 2: The Borromean link, an example of a all m-resistant cases up to N = 7. For pure states the 0-resistant link of 3 rings, classified by the polynomial problem has not been fully solved yet but nevertheless abc and given in Eq. (3). we will present a few examples as well as some geometric intuition behind them. so if a single variable is set to zero, the polynomial van- A. A general rule for mixed qubit states ishes. The case m = 1 has different properties, since if any party is ignored the remaining two parts are still con- The fundamental aspect of the analogy between links nected. From the point of view of quantum entanglement of rings and entangled states is that one can solve cer- a three-qubit state, called W [10], tain entanglement problems using the polynomials that | i characterize the links, and apply a map converting poly- 1 nomials into quantum states. As we show below, this W = ( 100 + 010 + 001 ) (5) | i √3 | i | i | i heuristic reasoning proves to be useful to identify interesting examples of quantum states resistant to loss of possesses this property, since partial trace over any of its subsystems. subsystems will result in an entangled mixed state. A The method to obtain m-resistant mixed states of N link which mimics this property is represented in Fig. 3. qubits starts from the polynomial of N variables that The polynomial which describes this link is symmetric, characterizes the corresponding m-resistant links and consists of J terms. For each term ti of the polynomial, ⊗N we associate a state ti d of the form | i ∈ H2 ⊗ H

ti = φ φ χ , (7) | i | enti | sepi | Di

where φsep stands for a separable state of m qubits, while φ| idenotes an entangled state of the remaining | enti N m qubits. An additional state, χD of a single system− D with d levels, plays an auxiliary| i role, as it will be traced over. An unnormalized superposition state, J ψN ,m = i=1 ti , allows us to generate the resulting mixed| statei ρ ˆ(N,m| )i of N qubits by partial trace over the d-dimensionalP environment, Figure 3: An example of a 1-resistant link of 3 rings,

classiﬁed by the polynomial ab + ac + bc, given in trD [ ψN,m ψN,m ] Eq. (6). ρˆ(N,m)= | i h | . (8) ψN,m ψN,m h | i To form an m-resistant statep we choose the separable m- (34)= ab + ac + bc , (6) partite state φ in (7) to be 0 ... 0 , the extra qudit P sep will assume the| indexi value of the| currenti term, and the and the generalization to any number of rings is given by entangled part φent will have the form the sum of all two-variable terms. | i Em = (m + 1) 0 ... 0 + 1 ... 1 , (9) | iijk... | iijk... | iijk... IV. IN SEARCH FOR M-RESISTANT STATES where ijk . . . are the variables appearing explicitly in the OF N QUBITS corresponding term of the polynomial. Note that for any m > 0 the above superposition is not symmetric, while The problem of ﬁnding m-resistant states for a general for m = 0 the state E0 denotes just the standard | iijk number N of qubits is not trivial. Since we are only inter- GHZ state (2) among subsystems labeled by indices i, j ested in whether a reduced state is entangled or not, for |and ki. 5

Putting the above recipe into practice, we have obtained m-resistant representative mixed states for all m for the cases N = 3 up to 7. In order to test for entanglement/separability of the reduced density matrices after each partial trace, we used a code available in [34] which looks iteratively for the nearest separable state and its decomposition into pure product states, returning the distance from the target matrix.

1. N=3 Figure 4: An example of a 0-resistant link of 4 rings, which is a 4-component Brunnian link, corresponding to the polynomial abcd – see Eq. (14). Notice that after Although representative pure states have already been cutting a single ring all three remaining rings become provided in the previous section for N = 3, we show here disconnected. an alternative answer in terms of mixed states. They are of the general form

tr [ ψ ,m ψ ,m ] The case m = 1 is associated with the polynomial ρˆ(3,m)= D | 3 i h 3 | , (10) ψ ,m ψ ,m h 3 | 3 i = abc + abd + acd + bcd (16) P with ψ3,m speciﬁed casep by case. For| m =i 0, we have a state with the entanglement which leads to the link of Fig. 5. The polynomial (16) is properties under partial trace represented in the link of Fig. 2. Its polynomial given by Eq. (3) has a single term only, which leads to the product state

ψ , = E 0 . (11) | 3 0i | 0iabc | id Hence the resulting state (8) obtained by partial trace is pure, and it is shown merely for completeness. The remaining case m = 1 is depicted in Fig. 3 and is characterized by the polynomial of Eq. (6). As such, we are lead to the following superposition of three states, entangled with respect to diﬀerent splittings, Figure 5: An example of a 1-resistant link of 4 rings, classiﬁed by the polynomial abc + abd + acd + bcd, given ψ , = E 0 0 + E 0 1 + E 0 2 . | 3 1i | 1iab | ic | id | 1iac | ib | id | 1ibc | ia | id in Eq. (16). After cutting any single ring one arrives at (12) the Borromean link.

2. N = 4 then mapped into the mixed state with

ψ , = E 0 0 + ... + E 0 3 (17) Using exactly the same procedure of N = 3, we ob- | 4 1i | 1iabc | id | ie | 1ibcd | ia | ie tain representative m-resistant mixed states for N = 4, where the ... entail all the other states associated with written in general as the remaining terms of the polynomial. Finally, the case m = 2 is described by tre [ ψ ,m ψ ,m ] ρˆ(4,m)= | 4 i h 4 | , (13) ψ4,m ψ4,m = ab + ac + ad + bc + bd + cd , (18) h | i P with ψ ,m speciﬁed casep by case. It is also worth men- | 4 i whose associated link class can be represented by the tioning that the links for more than three rings have been link of Fig. 6. The corresponding mixed state is readily drawn using a technique developed in [30] that makes di- obtained from rect use of the polynomials associated with them. For m = 0, the corresponding monomial reads ψ , = E 00 0 + ... + E 00 5 . (19) | 4 2i | 1iab | id | ie | 1icd | iab | ie (41)= abcd (14) P so the state associated with the link shown in Fig. 4 has 3. N = 5 the product structure, The case of N = 5 can be analyzed in analogy to the ψ , = E 0 . (15) | 4 0i | 0iabcd | ie previous cases. The general m-resistant mixed state will 6

set with all p letters. If N and m are positive integers, with N >m, then the mixed state

tr [ ψN,m ψN,m ] ρˆ(N,m)= Λ | i h | (29) ψN,m ψN,m h | i with p

ψN,m = Em 0 ... 0 i (30) Figure 6: An example of a 2-resistant link of 4 rings, | i | igi | iBN \gi | iΛ gi∈SN−m classified by the polynomial ab + ac + ad + bc + bd + cd, X given in Eq. (18). After cutting any two rings the is an m-resistant state of N qubits. This result has been remaining two are still connected. checked for N up to 7. For a larger number of qubits the size of the system grows as 2N so the computing time required to investigate separability of reduced den- be written as sity matrices increases accordingly. Therefore studies of systems with a larger number of qubits, although possi- trf [ ψ5,m ψ5,m ] ble, becomes impracticable. ρˆ(5,m)= | i h | , (20) ψ5,m ψ5,m Regarding the representation of links with more than h | i 6 rings, it becomes less elucidating, so we will refrain p with ψ5,m specified case by case. For m = 0 we have from providing them in those cases. Nevertheless, using | i the respective polynomials, it is a straightforward task = abcde (21) to draw these links. P so the respective state will be B. Pure qubit states ψ , = E 0 . (22) | 5 0i | 0iabcde | if Unlike the case with mixed states, identification of a For m = 1, we have pure state of the N-qubit system associated with a given m-resistant link of N rings is more intricate, if at all pos- = abcd + ... + bcde (23) P sible. We will provide here some partial answers following two different approaches: one which makes use of the which gives rise to the mixed state obtain from polynomial formalism for links; and another one based on the Majorana representation [31, 32] of symmetric states ψ , = E 0 0 +...+ E 0 4 . (24) | 5 1i | 1iabcd | ie | if | 1ibcde | ia | if of several qubits. In the spirit similar to the previous sections we may For m = 2, the polynomial is inquire if there exists a straightforward map between = abc + ... + cde , (25) the polynomials of m-resistant links of N rings, and m- P resistant pure states of N-qubits. Taking into account which is mapped into the state Eqs. (29) and (30), the most immediate ansatz for a pure m-resistant N-qubit state, denoted as (N,m) , would be | i ψ5,2 = E2 abc 00 de 0 f + ... + E2 cde 00 ab 9 f . | i | i | i | i | i | i | i (N,m) = Em 0 ... 0 . (31) (26) | i | igi | iBN \gi g S − Finally, for m = 3, we have i∈XN m Hence, we remove now the extra qudit in Eq. (30) and = ab + ... + de , (27) P need not perform the partial trace. The state still possesses the same symmetries as the corresponding polyno- which gives the state mial under particle permutations, so it is as least intuitive as a candidate. This ansatz, tested up to N = 7, gives a ψ , = E 000 0 + ... + E 00 9 . | 5 3i | 3iab | icde | if | 3ide | iabc | if correct answer only for m = 0, N 1 and N 2. The (28) first failure thus occurs for N = 5 and− m = 2,− where no corresponding 2–resistant pure state of 5 qubits has yet been found. Other choices could be made for Em in 4. N > 5 Eq. (31), for example, but as it turns out, no change| i in the coefficients of Eq. (9) will provide the correct answer. Taking into account the examples presented in the pre- Different approaches can be used to search for m- vious sections, it becomes tentative to conjecture the fol- resistant pure qubit states. Since the states can be sym- lowing statement. Let Sp−q be the set containing all metric under particle permutation, one may use the stel- subsets of p letters with p q elements, and let p be the lar representation of Majorana [31, 35], which provides − B 7 an alternative intuition on the geometry of these states. The Majorana representation of a N-qubit pure state, fully symmetric under permutations of its i1,i2,...,iN qubits, is given by 1 ψ = η1 ... ηN (32) | i √ | ii1 | iiN K σ X∈SN where the sum runs over all permutations σ SN of ∈ particles indices, K is a suitable normalization constant a) 0-resistant state b) 1-resistant state and Figure 7: Constellations defining m-resistant states of 3 θj iφj θj ηj = cos 0 + e sin 1 . (33) qubits. The middle arrow serves as a reference pointing | iX 2 | iX 2 | iX to the North pole. The constellation in a) corresponds to the state GHZ and is illustrated by the Borromean The N pairs (θj , φj ) are called Majorana points and | i uniquely define a point in the Bloch sphere. In this way, link in Fig. 2, and b) is related to the link in Fig. 3. one may define a fully symmetric N-qubit state by fixing N points on the sphere. For this reason, the name stellar representation is also common, as each point rep- where the weight of the 000 is increased with respect to | i resents a star in the sky, while a group of stars forms a Eq. (34), as one of the stars sits at the North pole. constellation. One of the advantages of the Majorana representation is that it allows one to ascribe to entanglement a degree 2. N ≥ 4 of geometrical intuition. If one fixes all the stars at a single point the corresponding state is separable. For instance N degenerated stars on the North pole represent The example of N = 3 suggests that m-resistance may the state 0 0 . However, as the degeneracy is lifted, be directly related to the number of stars on the North any non-trivial| ··· constellationi of stars corresponds to an pole, with all others evenly distributed along the equa- entangled state of N qubits. Thus, in a sense, the dis- tor. Indeed, for N = 4 this is in fact the case, where tance between the stars is directly related to the degree pure qubit states for all m-resistant can be constructed of entanglement [35, 36], although such a criterium to exactly in this manner, with the constellations depicted quantify entanglement is not uniquely defined. in Fig. 8. The pattern is thus to place m stars at the

1. N = 3

In order to investigate possible patterns, one should start with the lowest qubit case N = 3 and look for the simplest choices of constellations which give expected results. It is straightforward to show that the constellations of Fig. 7 give the desired answer for all m-resistant states a) 0-resistant state b) 1-resistant state c) 2-resistant state of 3 qubits. The m-resistant states of N-qubits constructed with help of the stellar representation and Ma- Figure 8: Constellations deﬁning m-resistant states of 4 jorana points will henceforth be denoted as (N,m)MP . qubits. The middle arrow serves as a reference pointing The constellation corresponding to m = 0 results| in thei to the North pole. Constellation a) corresponds to the 4 GHZ state ring Brunnian link from Fig. 4, b) to the link in Fig. 5, 1 and c) to the one in Fig. 6. (3, 0)MP = ( 000 + 111 ) , (34) | i √2 | i | i North pole and distribute the remaining stars evenly at for which the distance between the points at the equator the equator, where for convention we choose the ﬁrst star is the greatest. From this perspective, the GHZ state π | i at the equator with ( 2 , 0). This construction gives rise could be considered as the most entangled, although from to a family of states which, after expanding Eq. (32), can the point of view of resistance, entanglement of this 0- be expressed in general as resistant state is not resistant to the loss of a subsystem. The case of m = 1 corresponds to the state N 1 N ⊗N N+m N 1 ψm = 0 ( 1) Dm (3, 1)MP = (3 000 + 011 + 101 + 110 ) , | i N s m | i − − | i! | i √12 | i | i | i | i 1+ m (35) q (36) 8

N−i where we introduce the Dicke states [37] where we define αij . The partially transposed ≡ j matrix has one of the eigenvalues equal to N 1 ⊗m ⊗(N−m) Dm = Pj 0 1 , (37) 2 | i N | i | i N +3N 4 (45) j m X n o − q which is negative for any N. Thus, by the positive par- with the sum going through all permutations Pj . For tial transpose (PPT) test, we confirm that each 2-qubit N = 4, we have reduced density matrix is always entangled, and the state N ψ4 = (4,m) . (38) ψN−2 is (N 2)-resistant for any N. | mi | i | i − N Finally, focusing on the state ψN−3 , we perform the The construction described above breaks down at N = partial trace over any set of N | 2 qubitsi and obtain − 5. The first exception appears at (5, 1), where it is found 2 5 α23 + (α03) 0 0 0 that ψ1 = (5, 2) . All other cases, i.e. m = 0, 2 and 3, give| purei states| i with the desired resistance. One may N 0 α22 α22 0 trN−2 ρˆN−3 =   . (46) naturally inquire whether other symmetric positions of 0 α22 α22 0 stars may give rise to a (5, 2) , other than 1 star on the  0 00 α21 North pole and 4 stars| distributedi in a square at the   The partially transposed matrix of size four has all pos- equator. We have tested combinations where two stars itive eigenvalues for N > 1. In this case the PPT test are lifted by a general latitude, as well as four stars, and guarantees that the resulting state is always separable. it is possible to show that no 1-resistant state exists for We must then look into the reduced density matrix with these families of states. Evidently, this does not prove one less partial trace, in order to check if it is entangled. that a pure 1-resistant state of 5 qubits does not exist, Such density matrix has the form although it is tantalizing to conjecture it, given that pure states with certain entanglement properties under partial N trN−3 ρˆN−3 = trace do not exist. For instance, it is known [38] that 2 there are no AME states for four qubits. α33 + (α03) 000000 α03 0 α32 α32 0 α32 0 0 0 For N > 5 until 7, we have verified that the ansatz in   Eq. (36) fails in all situations except for m = 0, N 2 0 α32 α32 0 α32 0 0 0 −  0 00 α31 0 α31 α31 0  and N 1. It is in fact possible to show for general N   . that these− cases are always satisfied, i.e. that  0 α32 α32 0 α32 0 0 0     0 00 α31 0 α31 α31 0  N  0 00 α31 0 α31 α31 0  ψ0 = (N, 0) , (39)   | i | i  α03 0000001  ψN = (N,N 3) , (40)   | N−3i | − i  (47) N ψN−2 = (N,N 2) . (41) | i | − i By partially transposing any qubit, we will obtain a ma- N A proof that ψ0 is zero resistant is straightforward. trix which possesses an eigenvalue equal to Indeed, by Eq.| (36i), we have 13 7N + N 2 193 202N + 79N 2 14N 3 + N 4 N 1 ⊗N N ⊗N − − − − ψ0 = 0 ( 1) 1 . (42) p (48) | i 2 | i − − | i which is negative for all N. PPT criterion implies that This state is a generalization of the GHZ state for N all the subsystems are entangled. This proves that the qubits, which, after any partial trace, returns a density state ψN is (N 3)-resistant. | N−3i − matrix of the form

N 1 ⊗(N−k) ⊗(N−k) V. IN SEARCH FOR m-RESISTANT QUDIT trk ρˆ = 0 0 + 1 1 , (43) 0 2 | i h | | i h | STATES where trk denotes the trace over any set consisting of k qubits. This density matrix is separable for any k > 0, As already mentioned in Section IVB, we were not able N to establish a general construction of an m-resistant pure and thus the state ψ0 is 0-resistant for any N. | i N state of N qubit system. Therefore, we have expanded Regarding the state ψN−2 , after simple algebraic operations, one ﬁnds that| any Ni 2 partial traces will result our search for subsystems with a larger local dimension in the density matrix − d 3, sometimes called qudits. We succeeded in providing≥ formulas for a k-resistant N-qudit pure state for 2 α22 + (α02) 0 0 α02 N 2k. − ≥ N 0 α21 α21 0 An example of the AME state of six qubits [39] inspired trN−2 ρˆN−2 = ,  0 α21 α21 0  us to search for states with diagonal density matrices af-  α02 0 0 1  ter tracing the appropriate number of parts. More pre-  − (44) cisely, m + 1 parts in order to ﬁnd an m-resistant state. 9

Trivially such states are separable after tracing m +1 parts. Of course, the states we are looking for must re- 0 0000 0 0 0 0 0 |(5, 1)4i = |00000i 01111 01111 + |01111i main entangled after tracing out a smaller number of 02222 02222 + |02222i parts. In order to find such states, we present a connec- 03333 03333 + |03333i tion between a family of combinatorial designs and a fam- 10123 10123 + |10123i ily of quantum states with required properties. In par- 11032 11032 + |11032i ticular, we use the notion of orthogonal arrays (OA), and 12301 12301 + |12301i the established connection [21] between quantum states 13210 13210 + |13210i and OA. 20231 20231 + |20231i We use the following, consistent with the hitherto, no- 21320 21320 + |21320i 22013 22013 + |22013i tation (N,m)d for an m-resistant state of N qudits. | i 23102 23102 + |23102i 30312 30312 + |30312i 31203 31203 + |31203i A. Orthogonal arrays and quantum states 32130 32130 + |32130i 3 3 0 2 1 3 3 0 2 1 + |33021i Orthogonal arrays [42] are combinatorial arrange- ments, tables with entries satisfying given orthogonal Figure 9: Orthogonal array of unity index properties. A close connection between OA and codes, OA 42, 5, 4, 2 obtained from Reed-Solomon code of entangled states, error-correcting codes, uniform states length 5 over GF(4). Each subset consisting of two has been established [41]. Therefore, investigation of the columns contains all possible combination of symbols. connections between OA and resistant states seems to be Here, two such subsets are highlighted. The relevant a natural approach. Firstly, the concept of OA is briefly quantum state is obtained by forming a superposition of presented; secondly, relations between OA and resistant states corresponding to consecutive rows of the array – states are given. see the expression on the right-hand side. An orthogonal array OA (r, N, d, k) is a table composed by r rows, N columns with entries taken from 0,...,d 1 − in such a way that each subset of k columns contains all A proof of the latter statement is provided in Appendix possible combination of symbols with the same amount A. From the OA presented in Fig. 9, we obtain, for ex- of repetitions. The number of such repetitions is called ample, the following five-ququart, 1-resistant state: the index of the OA and denoted by λ. One may observe, that the index of OA is related to the previous coefficient (5, 1)4 = 00000 + 01111 + 02222 + 03333 + by the following formula: | i | i | i | i | i 10123 + 11032 + 12301 + 13210 + | i | i | i | i r 20231 + 21320 + 22013 + 23102 + λ = . (49) | i | i | i | i dk 30312 + 31203 + 32130 + 33021 . (51) | i | i | i | i Fig. 9 presents an example of an OA. A pure quantum state consisting of r terms might be associated with C. Existence of OAs and relevant m-resistant states OA (r, N, d, k), simply by reading all rows of OA [21]. The state of N qudits associated with the orthogonal ar- Construction of index unity OAs was provided by Bush ray OA (r, N, d, k) will be denoted as by φ N,k . | ( )id [40] in 1953. He used methods based on Galois fields theory in order to obtain the following result: B. Orthogonal arrays of index unity Theorem 1 (Bush, 53’). If d is a prime power, i.e d = pn for some prime number p and natural number n, then k The crucial quantity for our purpose, related to OA, we can construct the array OA d , d +1, d, k . is its index. It preserves the following information: how Combining Bush’s result with Proposition1 provides many repetitions of any sequence i ,...,i there are for 1 k the existence of m-resistant N-qudit states for N each subsystem of k rows. For λ = 1, any sequence ap- 2(m + 1). ≥ pears only once, and such an array is called index unity array. We emphasize their remarkable role in the search Corollary 1. For any N 2(m + 1) there exists the for resistant states. N-qudit state which is m-resistant.≥ The local dimension d is the smallest prime power larger than N 1. Proposition 1. For any orthogonal array of index unity − OA dk, N, d, k , where N 2k, the relevant quantum In Appendix B we present some of the m-resistant qu- ≥ state φ(N,k) is k 1-resistant. i.e dit states. A more interested reader might easily repro- | i − duce more of them with the help of available OA tables φ = (N, k 1)d . (50) [44]. | (N,k)i | − i 10

it does not produce all the required states for all values m N 4 5 6 7 8 of m =0, 1,...N 2. However, in some cases for which the above method− fails (e.g. N = 5 and m = 1) we 4 5 6 7 8 0 |ψ0 i |ψ0 i |ψ0 i |ψ0 i |ψ0 i constructed m-resistant pure states of N parties with a higher local dimension d making use of the combinatorial 4 notion of orthogonal arrays [41]. |ψ1 i |φ1,5i |φ1,6i |φ1,7i |φ1,8i 1 4 5 7 7 It is important to emphasize two defining characteristics of any m-resistant quantum state. The first one is the 4 5 2 |ψ2 i |ψ2 i AME |φ2,7i7 |φ2,8i7 presence or disappearance of entanglement after partial traces, which is equivalent to fixing which subsystems of 5 6 a quantum state are entangled, when the remaining part 3 |ψ3 i |ψ3 i |φ3,8i7 of the system is ignored or cannot be measured. In other words, it fixes which subsystems are able to successfully 6 7 4 |ψ4 i |ψ4 i perform protocols between each other. The second defining property is the symmetry under permutations of par- 7 8 ties, which brings about a notion of equality between all 5 |ψ5 i |ψ5 i subsystems. These two characteristics of quantum resistance may 8 6 |ψ6 i be combined into applications which require the simulta- neous action of any given number of parties in order to perform some concrete task. For instance, consider a dig- Table I: Search for m-resistant states of N parties at a ital locker belonging to a certain network of N parties, glance. Note that ψ4 is equivalent to GHZ state of 0 4 where each person is in possession of a qubit, which is 4 qubits. Three families:| i ψN , ψN | and ψiN of 0 N−3 N−2 part of a genuinely entangled N qubit state. One may de- m-resistant N-qubit states| discussedi | ini Section| IVi are vise this locker such that it may only be opened through presented on the differently shaded blue background. a protocol involving at least m people. In this case, an Six-qubit state AME(6,2) is also signalized. Moreover, m-resistant state of N qubits will provide exactly this m-resistant qudit states φ constructed by virtue | m,N id conditions, since any m 1 parties will always be in a of orthogonal arrays are demonstrated on the green separable state that makes− it impossible to form correla- background. The number in subscript is relevant to the tions between them. local dimension d of a state. Observe that the local Finally, we leave as a final remark a list of relevant dimension d of qudit states is usually (but not always) open problems: a) check whether there exist pure states equal to N 1. − of N qubits with the m-resistance property, for any m = 0, 1,...N 2; b) if this is not the case, for each N find the minimal− local dimension d such that there exist m- We organize all results obtained so far in table I. These resistant states of N qudits; and c) for any class of m- results encourage us to pose the following conjecture: resistant states of N qubits find a state for which its Conjecture 1. For any N and m, there exists an m- average entanglement after partial trace over any set of resistant N-qudit state in some local dimension d. m parties is the largest, if measured with respect to a given measure of entanglement.

VI. CONCLUDING REMARKS ACKNOWLEDGMENTS

Any link of N closed rings can be associated with a The authors would like to thank Miguel Reis Orcinha polynomial of N variables. This technique, developed in for helping with computational aspects of the code used [30], allowed us to construct a class of m-resistant links for testing the separability of the density matrices. It of N closed rings such that, after cutting any number m is a pleasure to thank Ole Andersson, Ingemar Bengts- of them, the remaining rings are connected, while cutting son and PawelHorodecki for numerous fruitful dicussions any of the remaining rings separates them. Making use and to Konrad Szymański for designing Fig. 12. G.Q. of the analogy between linked rings and entangled states acknowledges support from Funda¸cão para a Ciência e [28] we provided examples of m-resistant entangled quan- a Tecnologia (Portugal) through programme POCH and tum states, such that their entanglement is resistant to project UID/EEA/50008/2019. R.A. acknowledges sup- the loss of any set of m subsystems. port from the Doctoral Programme in the Physics and An explicit construction involving the partial trace Mathematics of Information (DP-PMI) and FCT through provides examples of m-resistant mixed states of N-qubit Scholarship No. PD/BD/135011/2017. A.B. and K.Z.˙ systems. An alternative method based on stellar repre- are supported by National Science Center under the grant sentation yields m-resistant pure states, but for N 5 number DEC-2015/18/A/ST2/00274. ≥ 11

Appendix A: Proof of Proposition 1 are of the form 0 0 0 0 ··· ··· Proof of Proposition 1. In order to prove the (k 1)- 0 0 1 1 . − ··· ··· resistance of φ(N,k) we need to show that it becomes k−1 N−k+1 separable after| the partiali trace of any k particles; while it remains entangled as any k 1 particles are traced Consider now the density| {z } matrix| {z } ρk−1(ψ) obtained by away. − tracing out ﬁrst (k 1) particles, and examine the eigenvalues of the partial− transpose of the last particle Observe that after tracing out any k particles, the den- ρTl (ψ). The aforementioned rows generate the follow- sity matrix becomes a diagonal matrix, hence it repre- k−1 ing elements in the matrix ρTl (ψ): sents a separable state. k−1 Checking the entanglement properties of the system 0 01 1 10 , after tracing k 1 particles is a bit more tricky. We shall |1 ··· 10i h0 ··· 01| . show that such− a system after partial transpose of one | ··· i h ··· | particle will always have a negative eigenvalue. Hence by We shall show that those elements are the only ones PPT criterion, it remains entangled. interlocking the bras

We investigate the density matrix ρa(ψ) after tracing 1 10 , 0 01 out first (k 1) particles. We will show that the partial h ··· | h ··· | − TN and kets transpose of the last particle ρk−1(ψ) have non-negative eigenvalues. To clarify, we sketch an example in Fig. OA 0 01 , 1 10 . of how the argument for the proof is developed. | ··· i | ··· i With this observation at hand, one can see that the ma- Tl trix ρk−1(ψ) has the block structure: 0 0 0 0 0 tr1 ρ1(ψ) = |0000i h1111| 0 1 0 0 0111 1 + |1111i h0000| ··· 02222 + . . . 1 0 0 0  0 0 ···  0 3333 T ρ l (ψ)= . . ·. ····... . , 1 0123 k−1  ......  11032  ......  T5  0 0  12301  ···· ·  1 3210   T5   2 0231 ρ1 (ψ) = |0001i h1110|   21320 + |1110i h0001| where we changed the order of computational basis in 22013 + . . . such a way that the first two vectors are 2 3102 3 0312 0 0 |1 ··· 1i . 31203 | ··· i 32130 N−k+1 3 3021 Tl | {z } Consequently, ρk−1(ψ) has at least one negative eigenvalue, namely 1. It only remains− to show the block form of the matrix Figure 10: Orthogonal array of unity index T ρ l (ψ). One may deduce it from the following two ob- OA 42, 5, 4, 2 . We investigate the matrix ρT5 (ψ) k−1 1 servations: obtained by tracing out the first particle and transposing the last one. The first two rows of presented 1. The first row in OA is the only one having zeros at OA differ in all positions except for the first one. They positions k,...,N among the first d rows. Indeed, generate the elements 0001 1110 and 1110 0001 in each of the first d rows has zeros at the first k 1 po- T5 | i h | |T5 i h | − ρ1 (ψ). There are no other elements in ρ1 (ψ) involving sition. Since the OA is of unity index at each other 0001 , 1110 , and 1110 , 0001 . Consequently, the position, the elements will differ between rows; | i | T5 i h | h | matrix ρ1 (ψ) has a negative eigenvalue. A similar 2. There are no other rows having only zeros or only argument holds for any other choice of a traced particle, ones at the positions k,...N 1. Indeed, since we and thus ρ1(ψ) is an entangled state. assumed N 2k, the number− of those positions is k. The≥ first and the second rows have zeros, and≥ ones, respectively, in those positions. Since the To begin, take two rows of the OA, which have zeros OA is of unity index, no other row may inherit this on the first (k 1) positions. Since the OA has strength property. k and is of index− unity, those rows differ in all other positions. Without loss of generality assume that they This ends the proof. 12

Appendix B: Examples of resistant states obtained tained through the method demonstrated in Section V A from OA with the help of available OA tables in [44].

We present some of the m-resistant qudit states ob- (6, 1) = 000000 + 011234 + 022341 + 033412 + tained using the established connection with orthogonal | 5i | i | i | i | i 044123 + 101111 + 112403 + 124032 + arrays. More states can be straightforwardly found with | i | i | i | i the help of available OA tables in [44]. 130324 + 143240 + 202222 + 214310 + | i | i | i | i The state (7, 1)7 might be obtained form (8, 1)7 sim- 223104 + 231043 + 240431 + 303333 + | i | i | i | i | i | i ply by deleting the last qudit from each summand. The 310142 + 321420 + 334201 + 342014 + number of summands grows rapidly as a function N m=1. | i | i | i | i 404444 + 413021 + 420213 + 432130 + For instance, the states (7, 2)7 and (8, 2)7 are super- | i | i | i | i positions of 73 elements| each;i while| the statei (8, 3) 441302 . (B1) 7 | i consists of 74 elements. The states can be quickly| ob-i

(8, 1) = 00000000 + 01123456 + 02234561 + 03345612 + 04456123 + 05561234 + 06612345 + | 7i | i | i | i | i | i | i | i 10111111 + 11352064 + 12520643 + 13206435 + 14064352 + 15643520 + 16435206 + | i | i | i | i | i | i | i 20222222 + 21546301 + 22463015 + 23630154 + 24301546 + 25015463 + 26154630 + | i | i | i | i | i | i | i 30333333 + 31265140 + 32651402 + 33514026 + 34140265 + 35402651 + 36026514 + | i | i | i | i | i | i | i 40444444 + 41031625 + 42316250 + 43162503 + 44625031 + 45250316 + 46503162 + | i | i | i | i | i | i | i 50555555 + 51604213 + 52042136 + 53421360 + 54213604 + 55136042 + 56360421 + | i | i | i | i | i | i | i 60666666 + 61410532 + 62105324 + 63053241 + 64532410 + 65324105 + 66241053 . (B2) | i | i | i | i | i | i | i

Appendix C: m-resistant links revisited Proposition 2. For any N and m = 0, 1 ...,N 1, there exists an m-resistant link of N rings. − The notion of m-resistant links is shown to be useful to visualize entanglement properties of a multipartite To prove this statement we use the technique of associating rings to polynomials provided in [30]. Consider a quantum state. Furthermore, it can be also relevant in N the context of knot theory. Even though from this per- polynomial of N variables, Pm(x1, x2,...,xN ), with m different terms, each being a product of N m variables. spective it is not common to analyze problems related to − cutting the links or neglecting them, the literature con- If m + 1 variables are set to zero, then Pm is equal to tains some items on this topic [45]. zero, which is taken to imply that the corresponding link We emphasize the relation between m-resistance and becomes disjoint after cutting m + 1 links. n-triviality of a link diagram. Let us begin with the pre- The procedure for constructing such links uses Brun- cise definition of an m-resistant link. nian braids spanned on the appropriate number of rings as basic building blocks. In particular, constructing an Definition 1 (m-resistant link). A connected link con- m-resistant link of N components amounts to construct- sisting of N components is called m-resistant if: N ing a polynomial of N variables with N−m terms of N m variables each, and identifying each of these it remains linked as any m components are ne- − • terms with the appropriate Brunnian building block – see glected; Fig. 11. Each Brunnian building block is a pure braid, it becomes unlinked after neglecting any m+1 com- i.e. a braid where the lines begin and end in the same • ponents. order. In addition, each Brunnian braid interlaces only the lines corresponding to the variables appearing in the The most common example of m-resistant link is the specific term of the polynomial, with the remaining lines Brunnian link, with m = 0. We recall that the Brunnian present but not interlaced. In the end, all braids are put link is a non-trivial link that becomes a set of trivial together by connecting lines of the same color to form unlinked rings if any single component is removed. a single braid, corresponding to the specific m-resistant A Brunnian link consisting of N rings is by construc- link obeying both conditions required in definition 1. The tion 0-resistant, see Fig. 4. There is a natural question, order by which they are connected does not matter. which is whether, for any N and m

For a given m-resistant link, this procedure begins by then connected in such a way that no additional cross- drawing one Brunnian link for each term of the polyno- ing are performed. A particular way to systematize this mial that characterizes the link and then cutting one end process is by building blueprints like the one described of each ring. The ends of all rings of the same color are in Fig. 12.

a) Braid representation for 3 ring Brunnian link.

b) Braid representation for 4 ring Brunnian link.

Figure 11: The braid representations for Brunnian links of 3 rings (top) and 4 rings (bottom). These braids constitute the building blocks for (N 4) and (N 3)-resistant links on N rings respectively. In the braid representation, the links are obtained by− joining each− end of the braids. It is interesting to note that, inside the dashed box in b), one has the braid representation of the 3 ring Brunnian link of a).

C D A

Figure 12: The blue print of an octahedron corresponding to the 1-resistant link of 4 rings (see Fig. 5), classiﬁed by the polynomial abc + abd + acd + bcd, given in Eq. (16). There is one face for each term of the polynomial with a Brunnian link of 3 rings with open edges of all rings. When the edges of the blue print are glued together, the edges of each ring are connected according to their color, without any extra crossings.

[1] V. Vedral, Introduction to Quantum Information Sci- [2] R. Laﬂamme, C. Miquel, J. P. Paz, W. H. Zurek, Perfect ence, (Oxford University Press, Oxford 2007). 14

Quantum Error Correction Code, Phys. Rev. Lett. 77, 86, 910 (2001). 198 (1996). [24] P. K. Aravind, Quantum Potentiality, Entanglement and [3] M. A. Nielsen and I. L. Chuang, Quantum Computation Passion-at-a- Distance: Essays for Abner Shimony, eds. and Quantum Information, (Cambridge University Press, R. S. Cohen et. al. Kluwer, Dordrecht (1997), pp53-59. Cambridge 2000). [25] L. H. Kauffman and S. J Lomonaco Jr, Quantum entan- [4] R. Jozsa and N. Linden, On the role of entanglement in glement and topological entanglement New J. Phys. 4, quantum computational speed-up, Proc. R. Soc. A 459, 73 (2002). 2011 (2003). [26] L. H. Kauffman and E. Mehrotra, Topological aspects [5] R. Horodecki, P. Horodecki M. Horodecki and of quantum entanglement, Quantum In.f Process. 18 76 K. Horodecki, Quantum entanglement, Rev. Mod. Phys. (2019). 81, 865 (2009). [27] M. Atiyah, Scalar curvature, flat Borromean [6] O. Gühne, M. Reimpell, and R. F. Werner, Estimat- rings, and the 3-body problem, preprint ing entanglement measures in experiments, Phys. Rev. https://arxiv.org/abs/1709.01539 Lett.98, 110502 (2007). [28] A. Sugita, Borromean Entanglement Revisited, [7] O. Gühne and G. Tóth, Entanglement detection, Phys. Proceedings of the Workshop on Knot Theory Rep. 474 1, (2009). for Scientic Objects, Osaka 2006, and preprint [8] I. Bengtsson and K. Zyczkowski,˙ Geometry of Quantum https://arxiv.org/abs/0704.1712 States II ed. (Cambridge University Press, Cambridge [29] U. Mishra, A. Sen De, U. Sen, Quantum superposition in 2017) composite systems of microscopic and macroscopic parts [9] T. Bastin, S. Krins, P. Mathonet, M. Godefroid, L. resistant to particle loss and local decoherence, Phys. Lamata, and E. Solano, Operational Families of Entan- Rev. A 87, 052117 (2013). glement Classes for Symmetric N-Qubit States, Phys. [30] G. M. Quinta and R. André, Classifying quantum en- Rev. Lett. 103, 070503 (2009). tanglement through topological links, Phys. Rev. A 97, [10] W. Dür, G. Vidal, J. I. Cirac, Three qubits can be entan- 042307 (2018). gled in two inequivalent ways, Phys. Rev. A 62, 062314 [31] M. Aulbach, D. Markham, and M. Murao, The maxi- (2000). mally entangled symmetric state in terms of the geomet- [11] S. Wu, Y. Zhang, Multipartite pure-state entanglement ric measure, New J. Phys. 12, 073025 (2010). and the generalized GHZ states, Phys. Rev. A 63, 012308 [32] J. Martin, O. Giraud, P. Braun, D. Braun, and T. Bastin, (2001). Multiqubit symmetric states with high geometric entan- [12] V. Coffman, J. Kundu, W. K. Wootters, Distributed En- glement, Phys. Rev. A 81, 062347 (2010). tanglement, Phys. Rev. A 61, 052306 (2000). [33] P. Ribeiro and R. Mosseri, Entanglement in the Sym- [13] F. Verstraete, J. Dehaene, B. De Moor and H. Verschelde, metric Sector of n Qubits, Phys. Rev. Lett. 106, 180502 Four qubits can be entangled in nine different ways, Phys. (2011). Rev. A 65, 052112 (2002). [34] http://phweb.technion.ac.il/∼stateseparator/ [14] G. Vidal, R. Tarrach, Robustness of entanglement, Phys. [35] W. Ganczarek, M. Ku´sand K. Zyczkowski,˙ Barycentric Rev. A 59, 141 (1999). measure of quantum entanglement, Phys. Rev. A 85, [15] P. Facchi, G. Florio, G. Parisi and S. Pascazio, Maximally 032314-10p (2012). multipartite entangled states, Phys. Rev. A 77, 060304 [36] K. Zyczkowski˙ and W. S lomczyński, Monge Metric on (2008). the Sphere and Geometry of Quantum States, J. Phys. [16] D. Goyeneche, D. Alsina, J. I. Latorre, A. Riera and A 34, 6689-6722 (2001). K. Zyczkowski,˙ Absolutely maximally entangled states, [37] R. H. Dicke, Coherence in Spontaneous Radiation Pro- combinatorial designs, and multiunitary matrices, Phys. cesses, Phys. Rev. 93, 99 (1954). Rev. A 92, 032316 (2015). [38] A. Higuchi, A. Sudbery, How entangled can two couples [17] W. Helwig, W. Cui, A. Riera, J. I. Latorre and H. Lo, get?, Phys. Lett. A 273, 213 (2000). Absolute Maximal Entanglement and Quantum Secret [39] A. Borras, A. R. Plastino, J. Batle, C. Zander, M. Casas, Sharing, Phys. Rev. A 86, 052335 (2012). and A. Plastino, Multi-Qubit Systems: Highly Entangled [18] Z. Raissi, C. Gogolin, A. Riera, A. Ac´ın, Constructing op- States and Entanglement Distribution, J. Phys. A 40, timal quantum error correcting codes from absolute max- 13407 (2007). imally entangled states, J. Phys. A 51, 075301 (2017). [40] K. A. Bush, Orthogonal Arrays of Index Unity, Ann. [19] A. J. Scott, Multipartite entanglement, quantum-error- Math. Statist. 3, 426 (1952). correcting codes, and entangling power of quantum evo- [41] A. S. Hedayat, N. J. A. Sloane, J. Stufken, Orthogonal lutions, Phys. Rev. A 69, 052330 (2004). Arrays: Theory and Applications, (Springer-Verlag, New [20] L. Arnaud and N. Cerf, Exploring pure quantum states York, 1999). with maximally mixed reductions, Phys. Rev. A 87, [42] C. R. Rao, Hypercubes of strength d leading to con- 012319 (2013). founded designs in factorial experiments, Bull. Calcutta [21] D. Goyeneche and K. Zyczkowski,˙ Genuinely multipartite Math. Soc. 38, 67 (1946). entangled states and orthogonal arrays, Phys. Rev. A 90, [43] M. Rains, Eric, Nonbinary quantum codes, IEEE Trans. 022316 (2014). Information Theory 45, 1827 (1999). [22] L. Seveso, D. Goyeneche, K. Zyczkowski,˙ Coarse-grained [44] N. J. A. Sloane, A Library of Orthogonal Arrays entanglement classification through orthogonal arrays, J. http://neilsloane.com/oadir/index.html Math. Phys. 59, 072203-28 (2018). [45] Y. Ohyama, A new numerical invariant of knots induced [23] H. J. Briegel and R. Raussendorf, Persistent entangle- from their regular diagrams, Topology and Applications ment in arrays of interacting particles, Phys. Rev. Lett. 37, 249 (1990).