Arxiv:1704.00346V2

Home , Cluster state

arXiv:1704.00346v2 [quant-ph] 6 Jul 2017 ilb ocre ihtefloigquantity following we the motivation with concerned this be properties With will correlation their measurements. study random to [5]. under is nonclassicality questions states the open complex analyze several of to with possibility problem interesting One hard sys- a multiparticle is of tems features nonclassical of acterization employed. com- be where can examples systems outstanding quantum are plex multipartite [4] and sharing [3] secret computing applica- concerns quantum what one-way In tions, [2]. between parties signalling separated superluminal spatially for allow in- would more causal fluences superluminal and, finite-speed [1] that illus- show inequalities to to recently, Bell employed without been nonlocality have trate g., level, fundamental e. the systems, multipartite On new completely applications. two- about and bring by phenomena they displayed Rather, effects systems. nontrivial party the of amplification ino oa realism. local of tion n aihsohrie oeta,we rprynor- properly when realism, that, local Note in malized, violations otherwise. to vanishes lead and that settings for only ees n a eetapriua elinequality Bell particular a select can One levels. tr htcnb aidwti elseai and, scenario Bell a within varied param- be all can to that correspond eters variables integration the where fqbt,e . h orsodn ubro inequivalent of number corresponding the number g., modest e. relatively qubits, a For of the grows. as parties difficulties of increasing number face pro- would Collins-Gisin-Linden- This nonlocality however, the of 11]. cedure, of [10, measure inequality context a (CGLMP) the Massar-Popescu as in considered defined applied quantity been the and where has [9], (1) ref. works. of in [8] case experimental the also and is 7] adopted This [6, approach theoretical the previous mainly in was This experiment. Bell integrate si h aefrmliatt nageet h char- the entanglement, multipartite for case the is As mere a provide not do systems multiparticle Quantum h probability The 3 P nttt o ula eerh ugra cdm fScien of Academy Hungarian Research, Nuclear for Institute f V I nad Rosier, de Anna vralpsil etnso h corresponding the of settings possible all over ehv bevdti ffc ob yia ntesneta te it that pres sense also entanglement. the We multipartite in genuine coefficients. typical of drawn witness be reali randomly to with local some values effect including small this relatively observed for have We even and settings or o admysmldosrals upiigy trapidl it Surprisingly, nonclassicality of observables. indicator sampled an randomly As for states. quantum partite a eitrrtda rbblt fviola- of probability a as interpreted be can epeeta xasienmrclaayi fvoain o violations of analysis numerical exhaustive an present We 2 .INTRODUCTION I. eatmnod ´sc,FdrlUiest fPernambuco of University F´ısica, Federal de Departamento P P V V ( ρ 1 a eue tdffrn context different at used be can = ) utpriennoaiyadrno measurements random and nonlocality Multipartite nttt fTertclPyisadAtohsc,Facult Astrophysics, and Physics Theoretical of Institute hsc n nomtc,Uiest fGank 038Gda Gdańsk, 80-308 of University Informatics, and Physics Z 1 ae Gruca, Jacek f (Ω) d Ω , 1 ennoParisio, Fernando I f and 1 = (1) ahosre a hoeamong choose ables can observer Each evr efrigmaueet nagvnsaeof state given a on N measurements performing servers hgnlprojections thogonal ics hi ancneune.I h atscinwe perspectives. section some last and the remarks In final and our tables consequences. give several of main form their the in discuss results section our In present we programming). numeric III (linear the employed of be description to brief method a provide we section next also and qubits five qutrits. to of states up sev- bipartite involving of of states nonclassicality probability relevant of the degree use eral the we evaluate that to so violation here, approach adopt the will is the we This being that party. measure- required per measurements possible information of context all number only to In refers the (space integration ments, inhabit. behaviors the polytopes case, of local this which space inequal- probabilities), the particular level joint a consider of deeper of directly a choice and the On ity with prohibitive. dispense become can would inequalitywe settings one time of addressing a thus, number at and, 2) large (say very fixed already is a with inequalities Bell rlBl xeietwith experiment Bell eral eea ntr transformations unitary general U for angles three qubits: by parametrized are transformations tary j 2 i hswr speetdi h olwn a.I the In way. following the in presented is work This nornmrclaayi ecnie h otgen- most the consider we analysis numerical our In ( qu φ Tamás Vértesi, 1 i,j e,H40 ercn ..Bx5,Hungary 51, Box P.O. Debrecen, H-4001 ces, mi iltdframs l observables. all almost for violated is sm d I ECITO FTEMETHOD THE OF DESCRIPTION II. { nrae ihtenme fparties of number the with increases y ψ , t with its O eepo h rbblt fviolation of probability the employ we 1 i 1 i,j O , oa els yfmle fmulti- of families by realism local f n h rbblt fvoaina a as violation of probability the ent χ , eie E56091 Brazil 50670-901, PE Recife, , egdfralivsiae states investigated all for merged 2 i .,O ..., , 1 i,j = ) d fMathematics, of y 3 qbt)and (qubits) 2 = m i

n islwLaskowski Wies law and ´nsk, Poland O i } j − i ( = cos sin i P 1 = φ φ N 1 i,j d r 1 i,j i − =0 , 1 e ptal eaae ob- separated spatially e 2 | iψ v − r .,N ..., , j i i iχ 1 i,j m | i v 1 i,j = i j i ih d rirr observ- arbitrary endb or- by deﬁned ) U v cos sin j i j i (qutrits). 3 = | | r ikdb the by linked φ i φ 1 i 1 i,j 1 i,j h uni- The . e e − iχ iψ 1 i,j 1 i,j (2) ! , 2 and eight angles for qutrits: III. RESULTS AND ANALYSIS

U i (φi,j , ψi,j ,χi,j , φi,j , ψi,j ,χi,j , φi,j , ψi,j )= j 1 1 1 2 2 2 3 3 We applied the numerical method to prominent fami- i,j i,j i,j iψ1 i,j iχ1 cos φ1 e sin φ1 e 0 lies of quantum states: i,j i,j i,j −iχ1 i,j −iψ1  sin φ1 e cos φ1 e 0  (1) the generalized N qubit GHZ state [1, 16] − 0 01  i,j i,j  GHZ(α) N = sin α 0...0 N + cos α 1...1 N, i,j iψ2 i,j iχ2 cos φ2 e 0 sin φ2 e | i | i | i 0 1 0 and for α = π/4, GHZ(α) N GHZ N; ×  i,j i,j i,j i,j  | i ≡ | i sin φ e−iχ2 0 cos φ e−iψ2 − 2 2 (2) the four qubit singlet state [17, 18]  10 0  i,j i,j i,j iψ3 − 1 0 cos φ3 e sin φ3 . (3) ψ = ( 0011 + 1100 ) ×  i,j  4 i,j i,j −iψ3 | i √3 | i | i 0 sin φ3 cos φ3 e − 1   ( 0101 + 0110 + 1001 + 1010 ); A local realistic description of an experiment is equiv- − √12 | i | i | i | i alent to the existence of a joint probability distribution p (r1, ..., r1 , ..., rN , ..., rN ), where ri = 0, 1, ..., d 1 lr 1 m1 1 mN ji (3) the N qubit Dicke state with e excitations [19] denotes the result of the measurement of{ the ith− ob-} i −1/2 server’s Oj observable. If the model exists, quantum pre- N De = 0...01...1 0...0 , dictions for the probabilities are given by the marginal | N i e | iN permutations sums: X e 1 N 1 N where the special case e = 1 is referred|{z} to as the N P (r , ..., r Ok1 , ..., OkN ) 1 qubit W state W N D [20]; 1 1 N N | i ≡ | Ni = Tr(ρ vk1 vk1 vkN vkN ) (4) | ih | ⊗ · · · ⊗ | ih | d−1 (4) the four qubit cluster state [21] 1 1 N N = plr(r1 , ..., rm1 , ..., r1 , ..., rmN ) 1 N 1 rj1 ,...,rj =0 Cluster4 = ( 0000 + 0011 + 1100 1111 ); P N | i 2 | i | i | i − | i where P (r1, ..., rN O1 , ..., ON ) denotes the probability k1 kN (5) the generalized N qutrit GHZ state of obtaining the result ri by the ith observer while mea- i suring observables O and ji = ki (i =1, ..., N). It can d=3 1 ki 6 GHZ (α) N = sin α 0...0 + cos α( 1...1 + 2...2 ); be shown that for some quantum entangled states the | i | i √2 | i | i marginal sums cannot be satisfied, which is an expression of Bell’s theorem. (6) the three qutrit singlet state (Aharonov state) [22] Our task is to find, for a given state ρ and a set of observables Oi (i = 1, ..., N; k = 1, ..., m ), whether − 1 ki i i A 3 = ( 012 + 120 + 201 the local realistic model exists, i.e., all the equations (4) | i √6 | i | i | i can be satisfied. This can be done by means of linear 011 101 110 ); programming (see e.g. [12–14]). It is worth mentioning − | i − | i − | i that the method allows us to reveal nonclassicality even (9) the three qutrit Dicke states with the sum of exci- without direct knowledge of Bell inequalities for the given tations equals to e [23] experimental situation. Finally, we check how many sets of settings (in per- 1 Q1 = π 001 , cents) lead to violation of local realism. We introduce | 3i √ {| i} 3 π a frequency pV (ρ) which for a sufficiently large statistics X converges to the probability of violation P (ρ). We pro- 2 1 V Q = 2 π 011 + π 002 , vided sufficient statistics to not observe changes in results | 3i √ {| i} {| i} 15 π π ! on the third decimal place. X X 1 The measurement operators are sampled according to Q3 = 2 111 + π 012 . Haar measure [15]. The angles ψ and χ are taken from | 3i √ | i {| i} r r 10 π ! uniform distributions on the intervals: 0 ψr < 2π and X ≤ π 0 χr < 2π. To generate φr in interval 0 φr it ≤ ≤ ≤ 2 We calculated the frequencies pV (ρ) for an increasing is convenient to use an auxiliary random variable ξr dis- 1/2 number of different settings per site. All results are pre- tributed uniformly on 0 ξr < 1 and φr = arcsin(ξr ) sented in Tables I, II and III. Some states which appear ≤ 1/4 for r 1, 2 and φ3 = arcsin(ξr ). Of course, all vari- on the tables are not listed above. They will be defined ables∈{ are generated} independently for each observer i and in the appropriate paragraphs. Our results lead to the measurement setting j. following observations. 3

1. Comparison with known results This discrepancy, though small, is due to the fact that our criterion is a necessary but not a suﬃcient one. The probability of violation was previously examined in several contexts. The only analytical result on tight inequalities was obtained in [6] for the simplest scenario of two settings and two outcomes, where the probabil- 3. Non-additivity and multiplicative features of PV (ρ) ity of violation of diﬀerent versions of the CHSH inequality [24] has been obtained by the two qubit GHZ The question of additivity seems to be better posed state (the Bell state). In this case our numerical method in terms of P (ρ) than in terms of maximal viola- gives the same value (No. 1) as the analytical expression V CHSH tions of a Bell inequality. Consider the example of the pV (GHZ2)= PV (GHZ2)=2(π 3) 0.283183 with − ∼ state GHZ2 GHZ2 , for which probability of vio- accuracy to four decimal places. | i ⊗ | i lation is non-additive, since p ( GHZ2 GHZ2 ) For N > 2, the GHZ state has been studied only nu- V | i ⊗ | i ≈ 1.7pV( GHZ2 ), and is a bit less than half of p ( GHZ4 ). merically. In [6] the state was analyzed in the context of | i V | i WWWZB˙ inequality for N 6. In [7] the analysis was Therefore instead of additivity, we should consider the ≤ multiplicative features of P (ρ). Concerning p ( GHZ2 ) extended to N = 15 qubits (WWWZB˙ inequality) and V V | i and p ( GHZ2 GHZ2 ), the probabilities that mea- N = 6 (using a similar linear programming method). In V | i ⊗ | i all cases, the results agree with our numerical method. surement results admit a local realistic description, PLR = 1 PV , should be multiplied. In this particular case, − 2. Genuine tripartite entanglement criterion

PLR( GHZ2 )=1 PV( GHZ2 )=1 2(π 3), (5) | i − | i − 2 − 2 We note that for any two-qubit state and two mea- PLR( GHZ2 GHZ2 )= PLR( GHZ2 ) = (7 2π) . surement settings per party, the probability of violation | i ⊗ | i | i − of local realism cannot be greater than 2(π 3), i.e., the − Hence, p (ρ ρ )=1 (7 2π)2 =0.486176 two-qubit GHZ state gives the highest probability. The V |GHZ2i ⊗ |GHZ2i − − analytical proof is deferred to the Appendix. which ﬁts our numerical results up to displayed digits Then, it is straightforward to prove that for any (No. 53). biproduct state ψ12 ψ3 the two-qubit quantum prob- We also examined the product of the two qubit GHZ | i⊗| i ability P (r1, r2 A ,B ) is described by a local realistic state with a state that does not violate any two setting | i j theory if and only if P (r1, r2, r3 A ,B , C ) does. Hence, Bell inequality, namely the Werner state: ρWerner2 = | i j k in the examined cases of entangled states of NE par- 1/√2 GHZ 2 GHZ + (1 1/√2)11/4. In this case the ticles, multiplied by the product state 0 ⊗N0 , the full probability| i ofh violation| for− the resulting state is the same | i (NE + N0)-particle state has, as expected, exactly the as for GHZ2 , what can be explained by the above mul- | i same probability of violation as its entangled component tiplicative feature, since PLR(ρWerner2 ) = 1. alone. The above property comes along with the fact that biseparable states (i.e. convex mixtures of biproduct states) can only lower the probability of violation compared to biproduct states. So we can argue that for 4. Non-maximal probability of violation for GHZ states of any 3-qubit state (including mixed states) with two mea- more than 3 particles surement settings per party, if PV (ρ) > 2(π 3), this certiﬁes that the three qubit state is genuinely− tripartite entangled, that is, it can not be written in any of the We observe a surprising feature, which emerges if the forms ψ12 0 , ψ13 0 and 0 ψ23 and con- number of qubits is larger than three. It is well known vex combinations| i ⊗ | i of| thesei ⊗ | states.i | Indeed,i ⊗ | datai in the that the N-qubit GHZ state maximizes many entangle- table I indicates that both GHZ3 and W3 states are gen- ment conditions and measures [26]. However, already for uinely tripartite entangled as the respective probabilities: N = 4 the probability of violation for the cluster state 74.688% (No. 28) and 54.893% (No. 36) are much higher (No. 94) is greater than for the GHZ state (No. 57). than 28.319%. The situation is even more dramatic for N = 5, where One could construct a similar condition for higher the probability is greater for any out of 10 randomly sam- number of parties (N > 3) but in this case one may pled pure states (Nos. 133-142). give only numerical bounds for the critical probability, There is a particular entanglement measure which is in because analytical results are not known in these cases. pace with the above observations, namely the generalized We also considered the probability of violation for the Schmidt Rank (SR) [21], corresponding to the minimal state ψ3(θ) = cos θ 111 + sin θ W3 (Nos. 42-50). For all number of product states required to represent a given values of angle θ >| 25.975i ◦ one| cani prove that the state state. The SR of a GHZ state is two for any number of is genuinely three-partite entangled [25], whereas our nu- qubits, and it has been shown in [21] that the SR behaves merical method reveals the threshold slightly below 30◦. as 2⌊N/2⌋ for cluster states of N qubits. 4

5. All typical states of ﬁve or more qubits violate local 100 realism for almost all settings a) 90

Even with only two observables per party it becomes 78 almost impossible not to detect non-classicality for states 68 with 5 qubits or more. Any of the studied states (Nos. 108-117) including random 5-qubit states (Nos. number of settings for the 2nd observer: 52 133-142) lead to nearly 100% probability of violation. 2 settings In fact, the numbers are so close, that one can not dis- 3 settings tinguish the states by means of the violation probability. probability of violation (%) 4 settings This amounts to an enhancement of the content of Gisin’s 28 theorem in the sense that not only all entangled states 2 3 4 5 6 7 8 9 10 seem to be nonclassical, but they violate local realism number of settings for the 1st observer for almost all experimental situations. That is, given an 100 entangled state it is very likely that one can prove its non- b) classicality on a ﬁrst try by choosing random observables 90 (note also related recent results in Ref. [29]). This is to be contrasted with the original demonstration [27, 28], 78 involving two qubits, where the settings have to be care- fully selected. Of course, one can always ﬁnd some states 68 with a pV (ρ) which is much smaller than 100% (e.g. Nos. number of settings for the 2nd observer: 114, 115), but they are strictly less entangled. 52 2 settings 3 settings

probability of violation (%) 4 settings 28

2x2 3x2 4x2 3x35x2 7x2 5x3 7x3 6x45x5 7x4 6. pV (ρ) rapidly increases with the number of settings 6x2=4x3 8x2=4x49x2=6x3 10x2=5x4 product of the number of measurement settings The probability of violation increases significantly also with the number of settings per party. For the two qubit FIG. 1. The probability of violation for the two qubit GHZ GHZ state, and five measurement settings per site, the states vs. (a) the number of measurement settings for the corresponding violation probability is almost equal to 1. first observer; (b)a product of the number of settings for both This means that almost all randomly sampled settings observers. lead to a conflict with local realistic models and to the violation of some Bell inequality. 7. Nonclassicality of bound entangled states This rapid growth is more pronounced than it is for robustness against white noise admixture. An increase A bound entangled state (BES) is entangled but undis- is also observed in the resistance to noise, but it is usu- tillable [33]. However, in [34] it was shown that the 4- ally a much less evident effect and visible particularly in qubit bound entangled Smolin state [35] can maximally multipartite cases [13]. For example, due to the recent violate a 2-setting Bell inequality similar to the standard work [30], an increase of 0.58% in the noise resistance CHSH inequality. In accordance with this finding, when of the two-qubit maximally entangled state required 30 numerically investigating the possibility of a local real- settings (see also a previous work [31]). It is also con- istic description for the Smolin state, even with 2 set- jectured that the above improvement in the noise resis- tings per party, we get small, but nonzero probability tance could not be attained with fewer settings. Note of violation, p (ρ )=0.023% (No. 101). Although also that one cannot go beyond the increase of 3.682% V Smolin4 this value is three orders of magnitude smaller than that in noise resistance using an infinite number of projective for other examined entangled states, it grows very fast measurements [32]. (faster than for other entangled states) with the number The dependence of pV (ρ) as a function of the number of settings (Nos. 102-107) and the growth seems to be of settings can be approximated by 1 ae−bx, where a,b exactly exponential. − are constant parameters and x can be either the number In general, if we investigate the PV of a PPT state [36, of settings referring to one party (with the other number 37] and find it to be non-vanishing, then the state must of settings fixed) (Fig. 1a) or a product of the number of be entangled. Note that this conclusion can be reached possible measurement settings (Fig. 1b). Of course there even without the knowledge of which Bell inequality is are other possible combinations involving the number of to be violated. This may be particularly useful when settings. the state involves many subsystems. In general, if we 5

25 investigate the PV of a PPT state [36, 37] and find it to be non-vanishing, then the state must be entangled. Note that this conclusion can be reached even without 20 the knowledge of which Bell inequality is to be violated. This may be particularly useful when the state involves 15 many subsystems. 2×2×2 The three qubit BES ρBES introduced in [38] also 10 violates some Bell inequality, but this seems to be sta- 10 tistically very rare since we had not observed any viola- 5 tion of local realism for two settings per party. Never- 9.8 0 2 4 6 8 10 theless, when the same measurement is applied to every probability of violation (%) particle, we observed a nonzero probability of violation, 0 2×2×2 0 10 20 30 40 50 60 70 80 90 pV (ρBES )=0.008%. The last considered example of bound entangled states a (deg) 3×3 is the two qutrit state ρBES that was used to disprove the famous Peres conjecture [39, 40]. Despite the fact that FIG. 2. Probability of violation for 2 qutrit generalized this state does not admit a local realistic model, the vio- GHZ(α) state vs. α lation is proved only for judiciously specified observables and inequality, which occur seldom enough, so that we 10 did not find violations in any of the 10 randomly chosen 10 1 2 |GHZ2i 2 × 2 10 28.318 settings. 9 2 2 |GHZ2i 3 × 2 10 52.401 9 3 2 |GHZ2i 4 × 2 10 68.654 9 4 2 |GHZ2i 5 × 2 10 78.947 8. Two qutrits: coincidence of maximal entanglement and 9 5 2 |GHZ2i 6 × 2 10 85.391 maximal nonclassicality 9 6 2 |GHZ2i 7 × 2 10 89.482 9 7 2 |GHZ2i 8 × 2 10 92.150 8 Entanglement and nonclassicality are distinct re- 8 2 |GHZ2i 9 × 2 10 93.945 8 sources. The former corresponds to the purely mathe- 9 2 |GHZ2i 10 × 2 10 95.198 9 matical concept of state nonseparability while the later 10 2 |GHZ2i 3 × 3 10 78.219 9 amounts to its manifestation in experiments. It is ac- 11 2 |GHZ2i 4 × 3 10 89.545 9 knowledged that a clear illustration of this point is 12 2 |GHZ2i 5 × 3 10 94.658 9 13 2 |GHZ2i 6 × 3 10 97.085 the unexpected difference between maximally entangled 9 14 2 |GHZ2i 7 × 3 10 98.303 states and states that maximally violate a Bell inequal- 8 15 2 |GHZ2i 8 × 3 10 98.953 ity. In [9] it is suggested that this anomaly may be 9 16 2 |GHZ2i 4 × 4 10 96.169 an artifact of almost all measures that have been used 8 17 2 |GHZ2i 5 × 4 10 98.460 to quantify nonclassicality. Our numerical results show 8 18 2 |GHZ2i 6 × 4 10 99.321 that, according to the probability of violation, there is 8 19 2 |GHZ2i 7 × 4 10 99.672 no anomaly in the nonclassicality of two qutrit gener- 8 20 2 |GHZ2i 5 × 5 10 99.504 alized GHZ state. The maximal probability of viola- ◦ 10 21 2 |GHZ2(1 )i 2 × 2 10 0.00000025 tion p =24.011% (No. 150) is attained for the symmet- ◦ 9 V 22 2 |GHZ2(10 )i 2 × 2 10 0.093 d=3 ◦ ◦ 9 ric state GHZ2 (35.26 ) instead of the asymmetric one: 23 2 |GHZ2(20 )i 2 × 2 10 2.826 d=3 ◦ ◦ 9 GHZ (29.24 )[pV = 22.317% (No. 149)], which maxi- 24 2 |GHZ2(30 )i 2 × 2 10 14.796 2 ◦ 9 mally violates the CGLMP inequality [11]. A little sur- 25 2 |GHZ2(40 )i 2 × 2 10 26.599 9 26 3 |GHZ2i ⊗ |0i 2 × 2 × 2 10 28.317 prising is the behavior of the probability of violation 9 27 3 |GHZ2i ⊗ |0i 3 × 2 × 2 10 52.399 around α = 0, where we observe a small local mini- 9 ◦ 28 3 |GHZ3i 2 × 2 × 2 10 74.688 mum for α = 6 (see Fig. 2). The minimum remains 9 29 3 |GHZ3i 3 × 2 × 2 10 90.132 even if the number of settings per party is increased to 9 30 3 |GHZ3i 4 × 2 × 2 10 95.357 three. A possible explanation of this feature could be the 9 31 3 |GHZ3i 3 × 3 × 2 10 97.245 fact that there are two relevant Bell inequalities for the 8 32 3 |GHZ3i 4 × 3 × 2 10 98.926 considered case – CHSH and CGLMP inequalities with 8 33 3 |GHZ3i 4 × 4 × 2 10 99.590 different functions representing the violation probability. 9 34 3 |GHZ3i 3 × 3 × 3 10 99.542 The total probability of violation is a combination of the 9 35 3 |W3i 1 × 2 × 2 10 15.244 9 probabilities for those particular inequalities, what may 36 3 |W3i 2 × 2 × 2 10 54.893 9 result in several extremes. 37 3 |W3i 3 × 2 × 2 10 76.788 9 38 3 |W3i 4 × 2 × 2 10 87.287 TABLE I: Frequencies of violation of local realism pV observed statisti- 9 39 3 |W3i 5 × 2 × 2 10 92.465 cally with random measurements on qubit states 9 40 3 |W3i 3 × 3 × 2 10 91.366 9 No. N State Settings Stat. pV 41 3 |W3i 3 × 3 × 3 10 97.797 6

◦ 9 42 3 |ψ3(15 )i 2 × 2 × 2 10 4.941 TABLE II. Frequencies of violation of local realism pV ob- ◦ 9 43 3 |ψ3(20 )i 2 × 2 × 2 10 10.327 served statistically with random measurements on random ◦ 8 44 3 |ψ3(25 )i 2 × 2 × 2 10 18.762 qubit states ◦ 9 45 3 |ψ3(25.975 )i 2 × 2 × 2 10 20.786 N V ◦ 9 No. Settings Stat. p 46 3 |ψ3(30 )i 2 × 2 × 2 10 30.323 8 ◦ 9 118 3 2 × 2 × 2 10 12.396 47 3 |ψ3(45 )i 2 × 2 × 2 10 64.382 8 ◦ 9 119 3 2 × 2 × 2 10 33.893 48 3 |ψ3(60 )i 2 × 2 × 2 10 74.689 8 ◦ 9 120 3 2 × 2 × 2 10 38.959 49 3 |ψ3(75 )i 2 × 2 × 2 10 65.377 8 ◦ 9 121 3 2 × 2 × 2 10 45.186 50 3 |ψ3(90 )i 2 × 2 × 2 10 54.893 8 8 122 3 2 × 2 × 2 10 43.505 51 4 |GHZ2i ⊗ |00i 2 × 2 × 2 × 2 10 28.318 8 7 123 3 2 × 2 × 2 10 4.812 52 4 |GHZ2i ⊗ |00i 3 × 2 × 2 × 2 10 52.407 8 8 124 3 2 × 2 × 2 10 59.824 53 4 |GHZ2i ⊗ |GHZ2i 2 × 2 × 2 × 2 10 48.617 8 7 125 3 2 × 2 × 2 10 35.197 54 4 |GHZ2i ⊗ |GHZ2i 3 × 2 × 2 × 2 10 65.887 9 8 126 3 2 × 2 × 2 10 43.602 55 4 |GHZ3i ⊗ |0i 2 × 2 × 2 × 2 10 74.683 8 7 127 3 2 × 2 × 2 10 43.747 56 4 |GHZ3i ⊗ |0i 3 × 2 × 2 × 2 10 90.134 7 8 128 4 2 × 2 × 2 × 2 10 95.016 57 4 |GHZ4i 2 × 2 × 2 × 2 10 94.240 7 8 129 4 2 × 2 × 2 × 2 10 93.104 58 4 |GHZ4i 3 × 2 × 2 × 2 10 98.352 7 7 130 4 2 × 2 × 2 × 2 10 95.630 59 4 |GHZ4i 4 × 2 × 2 × 2 10 99.339 7 7 131 4 2 × 2 × 2 × 2 10 90.957 60 4 |GHZ4i 3 × 3 × 2 × 2 10 99.624 7 6 132 4 2 × 2 × 2 × 2 10 92.616 61 4 |GHZ4i 4 × 3 × 2 × 2 10 99.867 7 5 133 5 2 × 2 × 2 × 2 × 2 10 99.862 62 4 |GHZ4i 4 × 4 × 2 × 2 10 99.937 7 7 134 5 2 × 2 × 2 × 2 × 2 10 99.857 63 4 |GHZ4i 3 × 3 × 3 × 2 10 99.934 7 5 135 5 2 × 2 × 2 × 2 × 2 10 99.900 64 4 |GHZ4i 4 × 3 × 3 × 2 10 99.981 7 5 136 5 2 × 2 × 2 × 2 × 2 10 99.889 65 4 |GHZ4i 4 × 4 × 3 × 2 10 99.989 7 5 137 5 2 × 2 × 2 × 2 × 2 10 99.913 66 4 |GHZ4i 4 × 4 × 4 × 2 10 99.993 7 6 138 5 2 × 2 × 2 × 2 × 2 10 99.878 67 4 |GHZ4i 3 × 3 × 3 × 3 10 99.995 7 5 139 5 2 × 2 × 2 × 2 × 2 10 99.884 68 4 |GHZ4i 4 × 3 × 3 × 3 10 99.999 7 5 140 5 2 × 2 × 2 × 2 × 2 10 99.880 69 4 |GHZ4i 4 × 4 × 3 × 3 10 99.999 7 5 141 5 2 × 2 × 2 × 2 × 2 10 99.861 70 4 |GHZ4i 4 × 4 × 4 × 3 10 100.00 7 4 142 5 2 × 2 × 2 × 2 × 2 10 99.878 71 4 |GHZ4i 4 × 4 × 4 × 4 10 100.00 8 72 4 |W4i 2 × 2 × 2 × 2 10 85.920 7 73 4 |W4i 3 × 2 × 2 × 2 10 95.129 7 74 4 |W4i 4 × 2 × 2 × 2 10 97.969 7 6 103 4 ρSmolin4 4 × 2 × 2 × 2 10 0.127 75 4 |W4i 5 × 2 × 2 × 2 10 99.013 7 7 104 4 ρSmolin4 5 × 2 × 2 × 2 10 0.195 76 4 |W4i 3 × 3 × 2 × 2 10 98.757 7 7 105 4 ρSmolin4 3 × 3 × 2 × 2 10 0.197 77 4 |W4i 3 × 3 × 3 × 2 10 99.767 7 6 106 4 ρSmolin4 3 × 3 × 3 × 2 10 0.601 78 4 |W4i 3 × 3 × 3 × 3 10 99.966 7 5 107 4 ρSmolin4 3 × 3 × 3 × 3 10 2.009 79 4 |W4i 4 × 4 × 4 × 2 10 99.999 7 2 8 108 5 |GHZ5i 2 × 2 × 2 × 2 × 2 10 99.601 80 4 |D4i 2 × 2 × 2 × 2 10 83.577 6 2 7 109 5 |GHZ5i 3 × 2 × 2 × 2 × 2 10 99.900 81 4 |D4i 3 × 2 × 2 × 2 10 94.065 7 2 7 110 5 |W5i 2 × 2 × 2 × 2 × 2 10 98.311 82 4 |D4i 4 × 2 × 2 × 2 10 97.315 2 7 2 7 111 5 |D5i 2 × 2 × 2 × 2 × 2 10 99.254 83 4 |D4i 3 × 3 × 2 × 2 10 98.428 1 7 2 7 ρ 2 ρ 3 112 5 2 ( D5 + D5 ) 2 × 2 × 2 × 2 × 2 10 0.047 84 4 |D4i 3 × 3 × 3 × 2 10 99.716 7 2 6 113 5 |GHZi4|0i 2 × 2 × 2 × 2 × 2 10 94.240 85 4 |D4i 3 × 3 × 3 × 3 10 99.964 7 2 5 114 5 |GHZi3|00i 2 × 2 × 2 × 2 × 2 10 74.688 86 4 |D4i 4 × 4 × 4 × 2 10 99.996 7 − 8 115 5 |GHZi2|000i 2 × 2 × 2 × 2 × 2 10 28.318 87 4 |ψ4 i 2 × 2 × 2 × 2 10 74.943 6 − 7 116 5 |L5i [41] 2 × 2 × 2 × 2 × 2 10 99.782 88 4 |ψ4 i 3 × 2 × 2 × 2 10 89.604 6 − 7 117 5 |R5i [41] 2 × 2 × 2 × 2 × 2 10 99.957 89 4 |ψ4 i 4 × 2 × 2 × 2 10 94.918 − 7 90 4 |ψ4 i 3 × 3 × 2 × 2 10 96.621 − 7 91 4 |ψ4 i 3 × 3 × 3 × 2 10 99.344 − 6 92 4 |ψ4 i 3 × 3 × 3 × 3 10 99.908 TABLE III: Frequencies of violation of local realism pV observed statis- − 5 93 4 |ψ4 i 4 × 4 × 4 × 2 10 99.991 tically with random measurements on qutrit states 8 94 4 |Cluster4i 2 × 2 × 2 × 2 10 97.283 8 No. N State Settings Stat. pV 95 4 |Cluster4i 3 × 2 × 2 × 2 10 99.275 d=3 ◦ 9 7 143 2 |GHZ (0 )i2 2 × 2 10 9.925 96 4 |Cluster4i 4 × 2 × 2 × 2 10 99.705 d=3 ◦ 9 7 144 2 |GHZ (5 )i2 2 × 2 10 9.801 97 4 |Cluster4i 3 × 3 × 2 × 2 10 99.884 d=3 ◦ 8 7 2 98 4 |Cluster4i 3 × 3 × 3 × 2 10 99.976 145 2 |GHZ (10 )i 2 × 2 10 10.021 6 d=3 ◦ 7 99 4 |Cluster4i 3 × 3 × 3 × 3 10 99.997 146 2 |GHZ (15 )i2 2 × 2 10 11.609 5 d=3 ◦ 8 100 4 |Cluster4i 4 × 4 × 4 × 2 10 99.999 147 2 |GHZ (20 )i2 2 × 2 10 15.057 8 d=3 ◦ 8 ρSmolin 148 2 |GHZ (25 )i2 2 × 2 10 19.363 101 4 4 2 × 2 × 2 × 2 10 0.023 d=3 ◦ 9 8 2 102 4 ρSmolin4 3 × 2 × 2 × 2 10 0.068 149 2 |GHZ (29.24 )i [asym] 2 × 2 10 22.317 7

d=3 ◦ 9 of the obtained results is that either for many particles 150 2 |GHZ (35.26 )i2 [sym] 2 × 2 10 24.011 d=3 ◦ 7 or many measurement settings we observe a conﬂict with 151 2 |GHZ (35.26 )i2 [sym] 3 × 3 10 78.667 d=3 ◦ 7 local realism for almost any choice of observables (the 152 2 |GHZ (35.26 )i2 [sym] 4 × 4 10 98.229 d=3 ◦ 8 probability of violation is greater than 99%) for typical 153 2 |GHZ (40 )i2 2 × 2 10 22.980 d=3 ◦ 8 families of quantum states. 154 2 |GHZ (45 )i2 2 × 2 10 19.763 d=3 ◦ 8 Concerning the nonclassicality of two qutrits, our re- 155 2 |GHZ (50 )i2 2 × 2 10 15.054 d=3 ◦ 8 sults are compatible with those presented in [9], that is, 156 2 |GHZ (55 )i2 2 × 2 10 10.153 d=3 ◦ 8 157 2 |GHZ (60 )i2 2 × 2 10 6.329 maximally entangled and maximally nonclassical states d=3 ◦ 8 158 2 |GHZ (65 )i2 2 × 2 10 3.638 coincide. It is worth mentioning that, in addition, we d=3 ◦ 8 159 2 |GHZ (70 )i2 2 × 2 10 1.818 addressed the apparently paradoxical result obtained in d=3 ◦ 8 160 2 |GHZ (75 )i2 2 × 2 10 0.714 [42]. It amounts to the observation, that the products of d=3 ◦ 8 161 2 |GHZ (80 )i2 2 × 2 10 0.174 k-qubit GHZ states and (N k) pure single qubit states d=3 ◦ 8 − 162 2 |GHZ (85 )i2 2 × 2 10 0.012 are more nonclassical than the N qubit GHZ state, if d=3 ◦ 8 163 2 |GHZ (90 )i2 2 × 2 10 0.000 we employ the robustness of correlations against white d=3 ◦ 8 164 3 |GHZ3 (0 )i 2 × 2 × 2 10 53.360 noise admixture as a measure of nonclassicality. Our nu- d=3 ◦ 8 165 3 |GHZ3 (35.26 )i 2 × 2 × 2 10 82.720 merical method shows that the probability of violation − 8 166 3 |A i3 2 × 2 × 2 10 72.328 of local realism for such product states (for N = 3, 4, 5 1 8 167 3 |Q3i 2 × 2 × 2 10 31.371 and k = 1, ..., N) is the same as for k-qubit GHZ state 2 8 168 3 |Q3i 2 × 2 × 2 10 48.506 and thus strictly smaller than for the N qubit GHZ state. 3 8 169 3 |Q3i 2 × 2 × 2 10 48.564 This suggests that resistance against noise, although relevant, is not a good quantiﬁer of nonclassicality.

IV. CLOSING REMARKS V. ACKNOWLEDGEMENTS In this paper we employed linear programming as a useful tool to analyze the nonclassical properties of quan- A.R., J.G., and W.L. are supported by NCN Grant tum states. We checked how many randomly generated No. 2014/14/M/ST2/00818. F.P. thanks the ﬁnancial sets of observables allow for violation of local realism. support from CAPES, CNPq, and FACEPE. T.V. is sup- Most of the conclusions were presented in the previous ported by the Hungarian National Research Fund OTKA sections. Here we want to stress that the overall message (Grant No. K111734).

[1] D. M. Greenberger, M. A. Horne, and A. Zeilinger, Bells [11] A. Ac´ın, T. Durt, N. Gisin, and J. I. Latorre, Phys. Rev. Theorem, Quantum Theory and Conceptions of the Uni- A 65, 052325 (2002). verse, edited by M. Kafatos, Vol. 69 (Kluwer Academic, [12] D. Kaszlikowski, P. Gnaciński, M. Zukowski,˙ W. Mik- Dordrecht, 1989). laszewski, and A. Zeilinger, Phys. Rev. Lett. 85, 4418 [2] J.-D. Bancal, S. Pironio, A. Ac´ın, Y.-C. Liang, (2000). V. Scarani, and N. Gisin, Nature Physics 8, 867870 [13] J. Gruca, W. Laskowski, M. Zukowski,˙ N. Kiesel, (2012). W. Wieczorek, C. Schmid, and H. Weinfurter, Phys. [3] R. Raussendorf and H. J. Briegel, Phys. Rev. Lett. 86, Rev. A 82, 012118 (2010). 5188 (2001). [14] J. Gondzio, J. A. Gruca, J. A. J. Hall, W. Laskowski, [4] Y.-C. Liang, F. J. Curchod, J. Bowles, and N. Gisin, and M. Zukowski,˙ Journal of Computational and Applied Physical review letters 113, 130401 (2014). Mathematics 263, 392 (2014). [5] N. Brunner, D. Cavalcanti, S. Pironio, V. Scarani, and [15] K. Zyczkowski˙ and M. Ku´s, Journal of Physics A: Math- S. Wehner, Rev. Mod. Phys. 86, 419 (2014). ematical and General 27, 4235 (1994). [6] Y.-C. Liang, N. Harrigan, S. D. Bartlett, and [16] V. Scarani and N. Gisin, Journal of Physics A: Mathe- T. Rudolph, Phys. Rev. Lett. 104, 050401 (2010). matical and General 34, 6043 (2001). [7] J. J. Wallman, Y.-C. Liang, and S. D. Bartlett, Phys. [17] H. Weinfurter and M. Zukowski,˙ Phys. Rev. A 64, 010102 Rev. A 83, 022110 (2011). (2001). [8] P. Shadbolt, T. Vértesi, Y.-C. Liang, C. Branciard, [18] A. Cabello, Phys. Rev. A 68, 012304 (2003). N. Brunner, and J. L. O’Brien, Scientific Reports 2 [19] R. H. Dicke, Phys. Rev. 93, 99 (1954). (2012), 10.1038/srep00470. [20] W. Dür, G. Vidal, and J. I. Cirac, Phys. Rev. A 62, [9] E. A. Fonseca and F. Parisio, Phys. Rev. A 92, 030101 062314 (2000). (2015). [21] H. J. Briegel and R. Raussendorf, Phys. Rev. Lett. 86, [10] D. Collins, N. Gisin, N. Linden, S. Massar, and 910 (2001). S. Popescu, Phys. Rev. Lett. 88, 040404 (2002). [22] A. Cabello, Phys. Rev. Lett. 89, 100402 (2002). 8

[23] W. Laskowski, J. Ryu, and M. Zukowski,˙ Journal of α = π/4 recovers the two-qubit maximally entangled Physics A: Mathematical and Theoretical 47, 424019 state. Let PV (α) denote the probability of violation cor- (2014). responding to the state GHZ(α) 2. Note that mixed [24] J. F. Clauser, M. A. Horne, A. Shimony, and R. A. Holt, two-qubit states cannot provide| higheri probability of vi- Phys. Rev. Lett. 23, 880 (1969). olation, therefore we can restrict our attention to the [25] D. Collins, N. Gisin, S. Popescu, D. Roberts, and probabilities PV (α). V. Scarani, Phys. Rev. Lett. 88, 170405 (2002). Firstly we prove the following lemma: if the CHSH [26] R. Horodecki, P. Horodecki, M. Horodecki, and inequality is violated using a state GHZ(α) 2 and some K. Horodecki, Rev. Mod. Phys. 81, 865 (2009). | i [27] N. Gisin, Physics Letters A 154, 201 (1991). projective measurements, at least the same violation oc- curs with the maximally entangled state GHZ 2 using [28] N. Gisin and A. Peres, Physics Letters A 162, 15 (1992) | i . the same measurements. [29] C.E. González-Guillén, C.H. Jiménez, C. Palazuelos, I. Villanueva, Comm. Math. Phys. 344, 141 (2016); C.E. Proof. Let us write the measurement observables A and González-Guillén, C. Lancien, C. Palazuelos, I. Vil- B as lanueva, arXiv preprint arXiv:1607.04203 (2016). [30] S. Brierley, M. Navascues, and T. Vertesi, arXiv preprint A = axσx + ayσy + azσz, arXiv:1609.05011 (2016). B = bxσx + byσy + bzσz, (A2) [31] T. Vértesi, Phys. Rev. A 78, 032112 (2008). [32] F. Hirsch, M. T. Quintino, T. Vértesi, M. Navascués, and where σx,y,z denote Pauli matrices, and the coefficients of N. Brunner, arXiv preprint arXiv:1609.06114 (2016) Alice measurements ax,ay,az square to 1 (and similarly . for Bob). Then we have the joint correlator [33] M. Horodecki, Pawel Horodecki, and Ryszard Horodecki, Phys. Rev. Lett. 80, 5239 (1998). AB = a b + sin 2α(a b a b ). (A3) [34] R. Augusiak and P. Horodecki, Phys. Rev. A 74, 010305 h i z z x x − y y (2006). [35] J. A. Smolin, Phys. Rev. A 63, 032306 (2001). On the other hand, the CHSH expression reads [36] A. Peres, Phys. Rev. Lett. 77, 1413 (1996). CHSH = A1B1 + A1B2 + A2B1 A2B2 . (A4) [37] M. Horodecki, P. Horodecki, and R. Horodecki, Phys. h i h i h i − h i Lett. A 223, 1 (1996). [38] T. Vértesi and N. Brunner, Phys. Rev. Lett. 108, 030403 Using formula (A3), we get (2012). CHSH(α)= C + sin(2α)(C C ), (A5) [39] A. Peres, Foundations of Physics 29, 589 (1999). z x − y [40] T. Vértesi and N. Brunner, Nat. Commun. 5, 5297 where α 0,π/4 and (2014). ∈{ } [41] G. Tóth, O. Gühne, and H. J. Briegel, Phys. Rev. A 73, C = a1 b1 + a1 b2 + a2 b1 a2 b2 , (A6) 022303 (2006). i i i i i i i − i i [42] W. Laskowski, T. Vértesi, and M. Wie´sniak, Journal of Physics A: Mathematical and Theoretical 48, 465301 where i can take x, y, and z. Notice that Cz 2, there- ≤ (2015). fore a CHSH value greater than 2 in equation (A5) implies that Cx Cy > 0. This in turn implies that in case of violation of the− CHSH inequality (that is CHSH > 2), we have CHSH(π/4) CHSH(α) for all α 0,π/4 . Appendix A: Proof ≥ ∈{ } Given the above lemma, it is not difficult to see that P (α) P = 2(π 3) for all α 0,π/4 . Indeed, The following observation is proven below: The prob- V V notice≤ that the classically− attainable∈ { region} of the 2- ability of violation P for any two-qubit state and two V setting 2-outcome scenario is completely characterized binary-outcome measurements cannot be greater than by eight different versions of the CHSH expressions (see 2(π 3). − e.g. [6]). However, after suitable relabeling of the in- In order to prove it, first recall that P = 2(π 3) for V − puts and flipping of the outcomes they all end up in the GHZ2 state using two binary-outcome settings per | i the standard CHSH defined by equation A4. Hence, vi- party [6]. Let olation of any of the versions of the CHSH inequality using a partially entangled state (A1) along with some GHZ(α) 2 = sin α 00 + cos α 11 (A1) | i | i | i projective measurements entails at least the same violation of this version using the maximally entangled state stand for the two-qubit pure partially entangled state and the same measurements. This implies the relation with α 0,π/4 written in the Schmidt bases. Clearly, p (α) p = 2(π 3) we set out to prove. ∈{ } V ≤ V −