Arxiv:2002.06990V1 [Quant-Ph] 17 Feb 2020 and Postselected Quantum Systems

Footprints of quantum pigeons

Gregory Reznik,1 Shrobona Bagchi,1 Justin Dressel,2, 3 and Lev Vaidman1, 2 1Raymond and Beverly Sackler School of Physics and Astronomy, Tel-Aviv University, Tel-Aviv 69978, Israel 2Institute for Quantum Studies, Chapman University, Orange CA 92866, USA 3Schmid College of Science and Technology, Chapman University, Orange CA 92866, USA (Dated: February 18, 2020) We show that in the mathematical framework of the quantum theory the classical pigeonhole principle can be violated more directly than previously suggested, i.e., in a setting closer to the traditional statement of the principle. We describe how the counterfactual reasoning of the paradox may be operationally grounded in the analysis of the tiny footprints left in the environment by the pigeons. After identifying the drawbacks of recent experiments of the quantum pigeonhole eﬀect, we argue that a deﬁnitive experimental violation of the pigeonhole principle is still needed and propose such an implementation using modern quantum computing hardware: a superconducting circuit with transmon qubits.

I. INTRODUCTION A somewhat weaker failure of the classical pigeonhole principle can be obtained when the holes are also quan- Quantum paradoxes describe phenomena that would tum (e.g. spin states), since it is less surprising that be impossible if Nature strictly obeyed classical intrinsically quantum features do not follow classical physics. Quantum mechanics presents many para- rules. doxes. A particular class of quantum paradoxes arises Even if the “holes” in such a scenario are classical, when we consider quantum systems between an initial we still have to clarify the meaning of a “quantum pi- preparation and final measurement. Notable examples geon being in a hole”. Standard quantum mechanics of such pre- and postselection paradoxes include the does not have a clear answer to the question: Where three-box paradox [1] where it is inferred that a parti- was a particle in between a preselection and postselec- cle with certainty has been in two distinct locations si- tion? In classical physics, the statement “this pigeon is multaneously, and the Hardy paradox [2] where it is in- in that hole” can be tested in parallel by many differ- ferred that each particle of a particle-antiparticle pair ent measurements that do not affect the situation. We has traveled through the same region of space with- do not assume this for quantum pigeons, because mea- out appearing there together. A more recent example surements performed on a quantum object generally is the quantum pigeonhole paradox [3, 4] where one change its state. A quantum pigeon can be prepared places a number of particles into a smaller number of in a superposition of several spatial locations, which boxes and infers that no two particles had occupied the also makes statements about such a pigeon occupying same box. This latter paradox has prompted extensive a particular hole not clearly defined. The exception is discussion and several experimental implementations when a quantum pigeon is described by a well-localized [5–9]. We revisit this pigeonhole paradox and propose wave packet with support only in one hole, in which a conceptually stronger variation. We also suggest case no paradoxical behavior arises. So, we need to that the existing experimental implementations have carefully define what we mean by a quantum pigeon not yet definitively demonstrated the paradox. occupying a particular hole. We will use the following The classical pigeonhole principle states that if one definition [11]: puts N pigeons into M pigeonholes, such that N > M, If we can infer with certainty [that] the re- then there must be at least one pigeonhole that con- sult of a measurement at time t of an obtains more than one pigeon. It was formulated by servable C equals to c, then C = c is an Dirichlet in the 19th century [10] and is widely used element of reality. in number theory and combinatorics. The principle In our case: If we can infer with certainty that the seems obvious and formalizes the fundamental concept measurement at time t of the presence of the pigeon of counting, yet it can be apparently violated by pre- in a particular hole would yield a positive result, then arXiv:2002.06990v1 [quant-ph] 17 Feb 2020 and postselected quantum systems. the pigeon was in the hole at time t. For a quantum system that is only preselected, a measurement outcome C = c will be obtained with II. ELEMENTS OF REALITY certainty only if the system is prepared in an eigen- state of C. However, when the system is both pre- To demonstrate a quantum violation of the classical and postselected, the condition for obtaining measure- pigeonhole principle one prepares a particular super- ment outcome C = c with certainty is different. position of N (quantum) pigeons distributed into M For a system preselected in a state |Ψi and postse- (classical) holes, then later measures another particu- lected in a state |Φi, the probability for a particular lar superposition of the N pigeons. In between the result of an intermediate measurement is given by the preparation and a successful postselection one then Aharonov-Bergmann-Lebowitz (ABL) formula [12] predicts with certainty that any particular hole does |hΦ|P |Ψi|2 not contain more than one pigeon. Moreover, this C=c Prob(C = c) = 2 2 . surprising prediction may be checked experimentally |hΦ|PC=c|Ψi| + |hΦ|PC6=c|Ψi| by placing a probe to count the pigeons in any box. (1) 2

Thus, the requirement for C = c to be an element of or zero pigeons: reality, i.e. Prob(C = c) = 1, becomes: ≤1 Y (m) X (j) Y (m) PX = P + PX P . (4) ( X X C = c is an element hΦ|P |Ψi= 6 0 m {j} m6=j ⇐⇒ C=c . of reality hΦ|PC6=c|Ψi = 0 For the pigeonhole principle to fail, the observ- ≤1 (2) able C = PX should be inferred to have the value c = 1 with certainty for either choice of X. Since ≤1 >1 Provided that the postselection becomes impossible PC=1 = PX and PC6=1 = PX , (1) produces the fol- when C 6= c, we can infer that C would be measured lowing requirements: to be c with certainty when the postselection succeeds. ≤1 >1 The measurement in this definition is understood hΦ|PX |Ψi= 6 0, hΦ|PX |Ψi = 0. (5) as counterfactual, i.e., it did not necessarily happen. However, it is assumed that if the measurement of C had been performed, then it must have been the only A. How to place four pigeons in two holes with not more than one pigeon in each hole measurement on the system between the pre- and postselection. Making more than one measurement would change the scenario and disrupt the inference. Even if We now demonstrate the failure of the pigeonhole the parts of the system are far away, performing mea- principle with four pigeons in two pigeonholes, mod- surement on one part can influence possible outcomes eled as four particles in two boxes. (We do not expect of the measurement of other parts. to perform experiments with real quantum pigeons.) A single measurement of the presence of more than one particle in any of the holes, yields with certainty >1 PX = 0. We prepare the particles in the initial state III. VIOLATING THE PIGEONHOLE PRINCIPLE WITH N PIGEONS IN TWO 1 |Ψi = √ |Ai1|Ai2|Ai3|Ai4 + |Ai1|Ai2|Bi3|Bi4+ HOLES 3 |Bi |Bi |Bi |Bi , The classical pigeonhole principle is a global state- 1 2 3 4 ment about all holes: there should exist at least one (6) hole with a particular property (having more than one pigeon). We will now show for quantum mechanics then postselect the particles in the final state: that given a particular pre- and postselection scenario 1 we can infer with certainty that we will not find more |Φi = √ |Ai1|Ai2|Ai3|Ai4 − |Ai1|Ai2|Bi3|Bi4+ than one pigeon in a single hole that we check. The 3 paradoxical situation is that we are certain not to find |Bi1|Bi2|Bi3|Bi4 . more than one pigeon in any one of the holes we try, no (7) matter how many times we try to find a hole contain- ing more than one pigeon. Nature seems to conspire Our requirements for (2) are then satisfied: against the experimenter by always hiding multiple pigeons from view, provided that the experimenter only 1 1 hΦ|P>1|Ψi = (1 − 1) = 0, hΦ|P≤1|Ψi = . (8) checks one box at a time and obtains a successful post- A 3 A 3 selection. We consider N pigeons placed in two pigeonholes A Similarly, and B. The pigeons may be partitioned into subsets of 1 labeled pairs {j, k}, triples {j, k, l} and so forth. The hΦ|P>1|Ψi = 0, hΦ|P≤1|Ψi = . (9) B B 3 statement that pigeonhole X contains more than one pigeon then corresponds to the projection operator If we were to try to find more than one particle in box A between pre- and postselection, then we would be X Y (m) Y (m) certain to fail. Similarly, if we were to try to find more P>1 = P P + X X X than one particle in box B, we would be certain to {j,k} m=j,k m6=j,k fail. No matter how many times we attempt to find X Y (m) Y (m) Y (m) P P + ... + P , (3) multiple particles in any single box, we would fail. X X X {j,k,l} m=j,k,l m6=j,k,l m In fact, our example demonstrates even stronger violation of classical reasoning. We put four particles (j) in two boxes such that there are no particles at all in where PX = |XijhX|j denotes the projection on a 0 =0 every box! That is, an observable C = PX testing state in which pigeon j is present in hole X, P(j) = X whether there are zero particles in box X will show I(j) − P(j) denotes the complementary projection on with certainty that there are none, c0 = 1. Indeed, the X >0 =0 0 the state in which pigeon j is not present in hole X, complement PX = I − PX corresponding to c 6= 1 and the summations are over the possible subsets of has the form >1 two or more pigeons. The negation of PX is that pi- X (j) Y (m) P>0 = P P + P>1, (10) geonhole X does not contain more than one pigeon, X X X X ≤1 >1 {j} m6=j PX = I − PX , i.e., pigeonhole X contains either one 3 and we obtain our requirements for (2) where |niX denotes the Fock state with n identical particles in the box X. Similarly, the pre- and postse- 1 hΦ|P>0|Ψi = 0, hΦ|P=0|Ψi = . (11) lection states in (6) and (7) become X X 3 1 Note that these results strongly depend on the exact |Ψi = √ |4iA|0iB + |2iA|2iB + |0iA|4iB , (15) definition of the measurements (3) and (10). If we ask 3 a different question, “Are there exactly four particles in and box X?”, then the outcome will be yes with certainty, P=4 = 1 for both boxes X. 1 X |Φi = √ |4iA|0iB − |2iA|2iB + |0iA|4iB , (16) 3

B. How to place N pigeons in two holes with not and lead to the same situation. The measurement of more than K pigeons in a hole the presence of more than one pigeon in any hole X >1 yields PX = 0 with certainty. Moreover, the measure- Let us consider how to generalize this result. We ment of more than zero pigeons in each hole also yields >0 discussed cases with no particles in a box and with PX = 0 with certainty. And it can be shown in the no more than one particle in a box. Classically, it is same way that the generalization of section III.B for possible to distribute N particles between two boxes K pigeons in the hole and more than two holes hold with no more than K particles in each box only if N ≤ for indistinguishable pigeons too. 2K. We find that in quantum mechanics it is also possible when N > 2K, except for one special case in which N = 2K + 1. Indeed, When N > 2K + 1 we can IV. PREVIOUS PROPOSALS TO VIOLATE THE PIGEONHOLE PRINCIPLE use the same method. We prepare the particles in the initial state We have presented a method of violating the pigeon- N K+1 N N 1 Y Y Y Y hole principle with quantum pre- and postselected par- |Ψi = √ |Ain + |Ain |Bim + |Bin , 3 ticles. Our example logically fits the classical pigeon- n=1 n=1 m=K+2 n=1 hole principle definition more directly than previous (12) proposals [3, 4]. However, our proposal has a seri- ous weakness for experimental verification. As men- then postselect the particles in the final state: tioned above, the meaning of an observable C that N K+1 N N asks whether there is more than one pigeon in a par- 1 Y Y Y Y |Φi = √ |Ain − |Ain |Bim + |Bin . ticular hole is that there is a measuring device capable 3 n=1 n=1 m=K+2 n=1 of displaying only one of two readings: ‘yes’, there is (13) more than one pigeon, or ‘no’, there is no more than one pigeon. That is, the quantum measurement should In case N = 2K + 1 this method does not work for not provide K, the exact number of pigeons in the hole, box B and straightforward calculation shows that no but instead only two readings: K > 1 and K ≤ 1. successful method exists. For arbitrary pre and post- The physical implementation of such a measurement >K >K selection both conditions of PA = 0 and PB = 0 requires that the measuring device must be affected can only be satisfied if the whole preselected state is exactly in the same way when we have two pigeons in orthogonal to the postselected state, which is impos- the hole and when we have three or four pigeons in sible. Therefore, there is no example of placing three the hole. Similarly, it must be affected exactly in the particles in two boxes such that no box contains more same way for either one or zero pigeons. While it is than one particle. not unthinkable to arrange an effective interaction that Note, that there is no limitation when number of achieves a similar response for two or more quantum boxes M > 2. If N ≤ KM then there is even a pigeons, most basic physical interactions are bi-particle classical solution for putting particles such that not couplings so it is challenging to ensure the needed in- more than K particles are present in any box. When sensitivity to particle number. Thus, the previous pro- N > KM, then the quantum solution is the preselec- posals for demonstrating the failure of the pigeonhole tion of state (12) and postselection of state (13). The principle (which are based on bi-particle interactions) only exception is N = 1 and K = 0. are still attractive from an experimental point of view even if their definitions do not fit the exact wording of the classical pigeonhole principle. C. How to place indistinguishable pigeons in two holes with not more than one pigeon in each hole A. How to place N pigeons in two holes such The failure of the pigeonhole principle can be that no hole contains two pigeons demonstrated also for quantum indistinguishable particles. In case of identical particles, using a Fock state The pigeonhole principle tells us that after placing representation is more convenient. For example, the N > 2 pigeons in two holes there should be at least one projection in (3) becomes hole with more than one pigeon. More than one is at least two, so a slightly weaker test is to check whether >1 PX = |2iX h2|X + |3iX h3|X + ... + |NiX hN|X , (14) there is at least one hole with two pigeons. Classically, 4 there is no difference, since one can always find two To explain this, recall a theorem connecting strong pigeons as a subset of more than two, so it is sufficient and weak measurements [1]. If the result of a strong to show that no holes have two pigeons to demonstrate measurement of some variable obtains a particular a violation of the pigeonhole principle. However, in eigenvalue with certainty, then the weak value is equal quantum mechanics there can be a difference between to this eigenvalue. Thus, for all pairs of particles and asking for exactly two pigeons and asking for two or {j,k} for both boxes PX = 0 holds. Weak values more. w characterize effective weak coupling and since weak In [3] a situation in which N > 2 particles are placed coupling does not disturb significantly the two-state into two boxes such that no box contains a pair of vector description of the pre- and postselected parti- particles was presented. Since this situation should cles, these null weak values remain small even when not occur classically, this weaker test still implies a all (weak) couplings are present. This is arguably the failure of the pigeonhole principle. To achieve this the most interesting physical implication of the quantum following states were pre- and postselected: pigeonhole effect.

In [4] another proposal for the failure of the pigeon- N 1 Y hole principle (which attracted significantly more at- |Φi = √ (|Ain + |Bin). (18) tention) was presented. This variation showed that N 2 n=1 the failure of the pigeonhole principle can occur even in systems without entangled pre- and postselections In this situation we can claim that every pair of par- (see also [5, 6]). The lack of entanglement makes this ticles {j, k} is not present together in any particular variation particularly attractive for experimental im- box X. For every pair {j, k}, the probability to find plementation. Consider the following pre- and postse- the pair in any box X vanishes. Indeed, we obtain: lected states: N−2 (1 − i) N hΦ|P{j,k}|Ψi = (12 + (−i)2) = 0, (19) 1 X N− 1 Y 2 |Ψi = √ (|Ain + |Bin), (22) 2 N 2 n=1 −i(1 − i)N−2 hΦ|I − P{j,k}|Ψi = 6= 0, (20) X N− 3 N 2 2 1 Y |Φi = √ (|Ain + i|Bin). (23) N {j,k} (j) (k) 2 n=1 where PX = PX PX . While this test is classically equivalent to testing The pre- and postselected states are completely separa- that there is certainly no more than one particle in ble; nevertheless, the probability to find any particular each box, this is not true quantum mechanically. In- pair of particles in the same box is zero. deed, if we perform a similar test for the presence of As before, this statement is correct only when one exactly three particles in a particular box using the pair is tested. Moreover, unlike the previous example same pre- and postselections, then we have nonvanish- it is correct only if the boxes A and B are not dis- ing probability to find them. The ABL formula (1) tinguished. The projection operator corresponding to yields this measurement is

N−4 {j,k} {j,k} (1−i) 2 P{j,k} = P + P . (24) | N− 1 | same A B Prob(P{j,k,l} = 1) = 2 2 X (1−i)N−4 2 (1−i)N−4 2 It tells us whether or not the particles j and k are | N− 1 | + | − 5 N− 1 | 2 2 2 2 present in the same box without providing information 1 = . (21) about which box they are in. 26 For our pre- and postselected states we obtain for This is why this example is formally not as strong every pair j, k: as our first example, even though the classical pigeon- N−2 hole principle is violated in both. Nevertheless, this {j,k} (1 − i) 2 2 hΦ|Psame |Ψi = (1 + i ) = 0, (25) example has an intriguing physical meaning in quan- 2N tum mechanics. The implied phenomenon is that if particles j and k would normally interact with each −i(1 − i)N−2 other when both present in box X, then in the specified hΦ|I − P{j,k}|Ψi = 6= 0. (26) same 2N−1 pre- and postselected situation the particles would apparently not interact. Moreover, provided the particle Similarly to the previous example, and unlike clas- interactions are weak enough, the particular pre- and sical physics, even if we are sure not to find any pair postselection effectively switches off all bi-particle in- in the same box, we might still find three particular teractions while preserving interactions between larger particles being in the same (without knowing which) numbers of particles. box. The ABL formula (1) for such a case yields: 5

A common weakness of the existing pigeonhole experiments is that the “holes” are usually spin or polar- N−4 (1−i) 2 ization states. These degrees of freedom are manifestly {j,k,l} | 2N−1 | Prob(Psame = 1) = N−4 N−4 quantum concepts, so it is not so strange that they (1−i) 2 (1−i) 2 | 2N−1 | + | − 3 2N−1 | fail to fulﬁll a classical principle. Nevertheless, these 1 demonstrations do show the conceptual failure of the = . (27) 10 principle. Spin can be up or down. If we have more than two particles, classical counting logic still tells us that there should be at least one pair of particles with V. EXPERIMENTS DEMONSTRATING THE the same spin state. PIGEONHOLE PARADOX

Testing for the presence of particles is challenging, A. Demonstration of the failure of the so the most promising experimental implementation pigeonhole principle with neutrons for violating the pigeonhole principle is that of example [4], since it tests the particle pair interactions Let us first discuss experiment [8], which uses the z rather than the locations of the particles. Moreover, component of a neutron’s spin to encode which “box” this implementation has pre- and postselected sepa- it occupies, with σz = 1 signifying hole A and σz = −1 rable states that are more easily arranged. Still, the signifying hole B. The experiment includes a source experiment is very difficult, since the natural coupling of individual neutrons and devices that prepare and between pairs of particles is very weak. postselect the required spin-polarization states, so task There are now several experimental papers that (i) is achieved. claim to demonstrate the violation of the pigeonhole No direct demonstration of task (ii) for pairs of neu- principle. In [7] “NMR investigation of pigeonhole ef- trons was performed. Instead, a careful measurement fect” quantum gates that schematically simulate the of the weak value of σ was performed for each neutron. pigeonhole experiment were implemented. Quantum z The following argument that this weak measurement is simulation, i.e. performing a sequence of quantum sufficient for demonstrating the failure of a pigeonhole gates that formally model the pigeonhole experiment, principle was provided: is not a compelling demonstration. In NMR exper- a) The weak measurement provided the weak value iment there is no direct connection between logical of the spin component of the pre- and postselcted neu- qubit and physical local system. More physical im- trons, (σ ) = i. plementations were performed with neutrons [8], and, z w b) For a product of variables related to separable more recently, with photons [9]. We argue that all particles (non-entangled pre and postselection states), these experiments are not yet satisfactory for defini- the weak value of a product is a product of weak values: tively demonstrating the quantum pigeonhole effect. (O(j)O(k)) = (O(j)) (O(k)) . Thus, A direct demonstration of [4] can be generally di- w w w vided into the following tasks. (j) (k) (j) (k) 2 i) Prepare N particles in the prescribed state (22) (σz σz )w = (σz )w(σz )w = i = −1. (28) and then postselect the particles in the state (23). ii) Add a strong interaction between randomly cho- c) For dichotomic variables, if a weak value is equal to sen pair of particles conditioned on their presence in an eigenvalue, then this eigenvalue, if measured, will be the same box. Upon pre- and postselection according obtained with certainty, i.e. it is an element of reality. to (i), show that this interaction is suppressed. (This theorem appeared first in [1].) Therefore, for every pair of particles j and k we have iii) An alternative to (ii) that is closer to the spirit of (j) (k) the original classical pigeonhole principle is to strongly an element of reality σz σz = −1. The interpretation measure, using external devices, that a randomly cho- of this is that these two particles have opposite spin z sen pair does not share the same box (without distin- components, which corresponds to particles “being in guishing the boxes). different holes”. This inference thus achieves task (iii). iv) Replace a strong interaction as in (ii) by a weak The difficulty with this experiment is that the joint bi-particle interaction, but make it between all pairs spin measurement was not actually performed since of particles. Show that upon pre- and postselection as no two neutrons ever coexisted in the measurement in (i), the effect of the interactions almost disappears apparatus. Only single-particle weak measurements of (j) (k) (becomes second-order in the weak disturbance). (σz)w = i were performed. The result σz σz = −1 Task (i) for [4] is simple and there is no doubt that was only inferred, which was only possible because the it was demonstrated, even if it was not specifically re- pre- and postselections were known to be separable. A ported in the experimental papers on the quantum pi- more convincing demonstration should be performed geonhole effect. However, it is clear that (i) by itself is without relying on any prior information about the not sufficient. From a physics point of view, task (iv) pre- and postselected states of the particles. Without might be the most interesting experiment; however, we this information, Eq. (28) of step b) does not hold, so have not seen a convincing implementation of it (de- the measured weak values do not directly provide a spite some claims made in [9]). Performing task (ii) demonstration of the failure of the pigeonhole princi- or task (iii) is the most important to be able to claim ple. that the pigeonhole effect was demonstrated. We will Indeed, the information that the pre- and postse- now analyze to which extent they were achieved. lected states are not entangled is crucial. Consider the 6 following pre- and postselected states of N particles: are the same, but they would reveal the polarizations when they are different. This difference, however, is N N ! 1 Y Y not crucial, since it does not remove the paradoxical |Ψi = √ |↑in + |↓in , (29) feature of the original proposal. 2 n=1 n=1 b) The paradox is defined for distinguishable particles, but the experiment was an interference of identi- N N ! cal particles. 1 Y Y The pigeonhole principle failure in [4] is defined for |Φi = √ |↑in + i |↓in . (30) 2 n=1 n=1 distinguishable particles. For any particular pair (j, k) (j) (k) we know that σz σz = −1. If the particles are identi- For this pre- and postselected state of N particles for cal and cannot be distinguished, then there is no mean- (n) every particle n we have (σz)w = i, but for every pair ing for this statement. Instead of coupling between (j) (k) we have (σz σz )w = 1. This example, in which there distinct particles, the core mechanism of the measure- is no failure of the pigeonhole principle despite every ment in experiment [9] is Hong-Ou-Mandel interfer- particle having value (σz)w = i, shows that the experi- ence, which requires identical bosons. The example ment [8] does not provide an unconditional demonstra- [4] can be designed with identical particles if they are tion of the pigeonhole principle. distinguishable by some degrees of freedom, such as a mode in the interferometer which identifies them. Indeed, experiment [9] starts with three identical pho- B. Demonstration of the failure of the tons, but in different modes which could identify the pigeonhole principle with photons particles of the pigeonhole paradox. The problem is that their measuring device, the polarization beam Let us now turn to experiment [9]. From the ab- splitter, mixes the modes, i.e. scrambles the identity. stract one can understand that it achieves almost ev- How can we demonstrate that particular particles do erything, including weak measurements, i.e., task (iv). not have the same polarization when the measurement However, in the summary it was admitted that there loses the identity of the particles? was no direct demonstration: “We implement the de- c) The demonstrated lack of disappearance of the sired measurement indirectly by analyzing the mea- effect of the interactions in the pigeonhole setup for surement effects order by order and reveal the paradox strong measurements of two pairs of particles is not will not survive under high-order measurement.” It is relevant, since it is expected only for weak coupling. not clear what it might mean “order by order” since The experiment [9] was also performed with coupling only strong couplings were described in this experi- (interference) of three particles, testing the presence of ment. As such, we see mainly an attempt to perform all three particles sharing the same spin state. How- task (ii), but we feel that even this was not done in a ever, in [4] it was never claimed that there are no three fully satisfactory way. We see the following weaknesses pigeons in the same hole, the claim was only about of the experiment. pairs (see also discussion in [5, 6]). a) Instead of showing that properly pre- and postse- This fact is clearly demonstrated by (27). The prob- lected particles have the property of “not being in the ability of finding three particles in the same box is 1 same hole”, it was shown that particles with the desired nonzero and is equal to 10 . The experiment indeed property were never found with proper postselection. demonstrated that there are sets of three photons with In task (iii) we are supposed to randomly choose two the same polarization, confirming (27), but this is a properly pre- and postselected particles and perform a demonstration of a situation which is not paradoxical. (j) (k) strong measurement to show that σz σz = −1. In- stead, the experiment was done in such a way that (j) (k) VI. IMPLEMENTATION PROPOSALS the particles fulfilling σz σz = 1 had zero probability of being properly postselected. Indeed, this is what the experiment showed: none of the properly In the previous section we provided more criticism pre- and postselected particles with this property were of existing experiments than solutions. A convincing found in the experiment. This finding supports the demonstration of the pigeonhole principle is a very dif- claim that properly postselected particles do not fulfill ficult task. A proper implementation of the example (j) (k) σz σz = 1, but does not demonstrate the paradoxi- [4] requires the measurement of nonlocal variables like cal situation of the failure of the pigeonhole principle. parity [13]. For example, the setup of the recent imple- There were no observed and detected photons pre and mentation [14] of nonlocal measurement [15] could al- postselected in the states corresponding to the failure low a more satisfactory experiment showing the failure of the pigeonhole principle. of the pigeonhole principle in the version with “holes” Note that adding number-resolving photon detectors still being spin states. Indeed, in [14] there was the ex- to allow observing the properly postselected photons perimental realization of a pre- and postselected pair that both arrive at the same port (corresponding to of particles fulfilling a requirement similar to the type (j) (k) different polarizations) could solve the problem. In we are looking for: σx σz = −1. Importantly, this this case, however, the detection would not just tell us property was measured, so this setup could be adapted that the photons do, or do not have the same polar- to demonstrate task (iii) of the quantum pigeonhole ef- ization, as was proposed in [4]. Instead, the detector fect. clicks will not reveal which polarization they are if they Modern superconducting quantum computation cir- 7

transmons coupled to the same resonator produce the same shifts for the same energies. Therefore, the res- onances for the odd transmon subspace (01 and 10) will produce no net frequency shift χj,k − χj,k = 0. In contrast, the even subspace (00 and 11) will produce distinct shifts of ±2χj,k that are detuned beyond the resonator linewidth. Therefore, pumping the readout resonators on the common resonance for the odd subspace allows the two parity subspaces to be distinguished. Pumping for a short duration produces an integrated signal I¯j,k corresponding to a weak mea- (j) (k) surement of the parity σz σz , while pumping for a long duration produces a projective measurement. To achieve task (iii) of [4], one would select a random readout resonator and pump it for a long duration. We expect that for all successful postselections only (j) (k) the parity eigenvalue σz σz = −1 will be observed. FIG. 1. Proposed setup for demonstrating the quantum pi- This procedure can then be repeated for random pairs geonhole paradox with superconducting transmons. Each as many times as desired. Note that this procedure pair (j, k) of three transmons j, k = 1, 2, 3 is coupled disper- will work only if a single pair is measured at a time. If sively to parity readout resonators. The transmon states multiple resonators are pumped simultaneously, then determine the dispersive shifts χj,k of the resonator fre- no definite parity results will be obtained. quencies. Pumping the readout resonators will produce To achieve task (iv), i.e. to demonstrate that signals Ij,k(t). For task (iii), one resonator is pumped (j) (k) for a long duration to projectively measure a parity eigen- σz σz = −1 for all pairs simultaneously, one would (j) (k) instead pump all three resonators for a very short du- value σz σz = −1. For task (iv), all three resonators are pumped for a short time to measure the parity weak values ration and record all signals. After averaging many of -1. successful postselections together, one would obtain (j) (k) the parity weak values (σz σz )w = −1 for all three pairs. The short pump duration prevents appreciable cuits could enable a demonstration of not only task disturbance so that these weak values can be observed. (iii) for [4], but also task (iv). One would encode the As indicated before, a weak value (of our dichotomic “boxes” as distinct energy states of an anharmonic os- (j) (k) variable) (σz σz )w = −1 implies that if this variable cillator, such as a transmon [16]. Since these distinct were strongly measured, then the measurement would states correspond to mesoscopic collective charge os- report the same value with certainty. cillations along superconducting wires at the micron An experiment of this type is on the cutting edge scale, their oscillation energies are somewhat more de- of current technology. Tuning the dispersive shifts of fensible classical boxes than the intrinsic spin states the six transmon-resonator couplings χj,k while allow- of individual quantum particles. Different oscillators ing the individual qubit control to perform the needed need not share identical energies, but their lowest two pre- and postselections is a nontrivial engineering task. energy states can be arranged to fall within the same Moreover, high-fidelity weak measurements would re- energy intervals, and thus “share the same energy quire quantum efficiency higher than what has been boxes” in a way analogous to how two classical parti- obtained in previous demonstrations of weak contin- cles with slightly different positions can share the same uous measurement [18–20]. Preliminary experiments spatial box. of the direct parity measurements that are needed Direct parity measurements are possible for pairs of here have been attempted [21, 22], but have not yet transmons and are already being experimentally im- achieved the fidelity required for a convincing demon- plemented for the purpose of quantum error correc- stration of the quantum pigeonhole effect. We consider tion [17], where the parity measurements are used to this proposal as a challenge to the experimental com- stabilize entangled code spaces and track bit errors. munity in the near term. Thus, one could directly measure the needed negative (j) (k) parities, corresponding to σz σz = −1, for randomly chosen pairs of pre- and postselected transmons. This VII. THE PAST OF QUANTUM PARTICLES experiment will have a conceptual advantage relative to [14] since here the pointer is an external measur- The failure of the pigeonhole principle for quantum ing device, and not another degree of the measured particles happens when they are pre- and postselected, photons. so the failure takes place in the past. However, stan- Figure 1 illustrates how to implement the needed dard quantum theory does not provide a clear picture parity measurements for the quantum pigeonhole para- for the past of quantum particles. In our discussions of dox with superconducting transmons. Each pair (j, k) the quantum pigeonhole principle, we adopted a defi- of three transmons j, k = 1, 2, 3 is coupled dispersively nition of a counterfactual character: if it was inferred to parity readout resonators. The resonator frequen- using the ABL formula (1) that the particle would be cies shift by amounts ±χj,k conditioned on specific found in a particular box with certainty when searched, transmon energies. The shifts are tuned such that then we said that the particle was in the box. This 8 definition has a conceptual difficulty that we must ad- two-particle trace criterion, there is no failure of the dress: we want to make claims about the presence or pigeonhole principle after checking all pairs in this ex- absence of quantum particles in a box even when we ample. This fact helps clarify the limited meaning of do not check the box. One could argue that a strong this example of the paradox: we have to consider only a measurement, in fact, changes the quantum state and special measurement interaction in which the environ- thus disrupts the ability to make the desired inference. ment does not distinguish between situations in which Vaidman [23, 24] proposed an alternative way to rea- different nonzero numbers of particles are present in son about the past of a quantum particle that can be the box. Only in that case will all the traces cancel to grounded more operationally: if the particle was there, correspond to the conclusion from the ABL rule (1). it should weakly interact with the environment and In contrast, direct local traces demonstrate the fail- leave a trace behind. Thus, the question of a particle’s ure of the pigeonhole principle of example [3] well. presence in a box can be answered by checking for the If we arrange a situation in which only two par- presence of a trace left in the box after the postse- ticular particles, j and k, leave local traces in the lection. This will not be a robust trace in which the boxes according to (31), then after the postselection environment changes its state to an orthogonal state, there will be zero amplitude for the two-particle trace since that would disrupt the evolution of the particle 2 ⊥ ⊥ |χj iA|χk iA and zero amplitude for the two-particle too sharply. The appearance of a small amplitude of an 2 ⊥ ⊥ trace |χj iB|χk iB. In a realistic case, when all par- orthogonal component for the environment is sufficient ticles leave amplitude traces, the two-particle trace to establish a suitable trace. More precisely, the ampli- will survive the postselection, however only as a part tude of the orthogonal component in the environment 3 ⊥ ⊥ ⊥ of three-particle trace, such as |χj iA|χk iA|χl iA. should be of the same order as the amplitude of the Since the amplitude is of the third order in (and component that would appear if a single well-localized not the second), such a trace is neglected. particle were placed in the box directly. In the example [4], similarly to our four-particle ex- To simplify the analysis, we assume that for each 2 ⊥ ⊥) particle there is a particular (different) position at ample, local two-particle traces |χj iA|χk iA persist which it could reside in the box, so any trace it could for every pair of particles in box A (with similar traces leave in the box will be independent of traces from in box B). This example thus represents a failure of other particles. Every particle that is definitely in the the pigeonhole principle with a limited meaning. We box leaves a trace with a small amplitude , i.e. the can claim that the particles in every pair do not oc- local environment of a particle j makes the following cupy the same box, but we can only check this state- evolution: ment with a measurement that does not tell us which specific box. The measurement thus cannot be just ⊥ some local measurement; it requires entanglement of |χji → η |χji + |χj i. (31) the environment in box A and box B such that the We assume that nothing happens to the environment local couplings leave the box identities uncertain. To if there is no particle. this end a nonlocal parity measurement is needed for With these assumptions for the trace, all discussed a proper demonstration. setups for the quantum pigeonhole paradox will show The measurement should arrange local couplings in a trace of order for every particle in every box after box A and B such that particle j in box A affects a the postselection. That is, there will be amplitudes composite entangled system I of local environments in ⊥ of order for every orthogonal component state |χj i. A and B exactly in the same way as particle k in box Thus, according to Vaidman’s weak trace criterion, ev- B, and also that particle j in box B affects another ery particle indeed was present in every box. This is composite entangled system II of the environment ex- already a somewhat paradoxical situation, but it is not actly in the same way as particle k in box A. In this a demonstration of the failure of the pigeonhole prin- situation, particles j and k in box A will create trace 2 ⊥ ⊥ ciple. As is well-known from Hardy’s paradox [2, 25], in the environment |χj iI|χk iII, but this trace will we cannot conclude that the two particles are present be exactly the same as the trace left by the two parti- together in a box solely from the evidence of single- cles present in box B, and thus the environment will particle traces. If we put a pair of particles j and k know that the two particles are in the same box, but into box A, the interaction with the environment de- will not know in which. For such a specially tailored 2 ⊥ ⊥ scribed by (31) leads to the appearance of a component environment there will be no trace |χj iI|χk iII after ⊥ ⊥ with the product state |χj i|χk i with an amplitude of the postselection, in correspondence to the claim that order 2. This is the proper criterion for particles j there are no particles in the same box. Parity mea- and k both occupying a box together. surement procedures of this type were described in the Let us analyze our examples according to this cri- previous section. terion. Are particles 1 and 2 present in box A after The trace approach to the past of the particle allows pre- and postselection in the example of four particles us to understand the failure of the pigeonhole princi- in two boxes, with Eqs. (6, 7)? The component of ple for pre- and postselected quantum systems from an 2 ⊥ ⊥ the state of the environment in box A, |χ1 iA|χ2 iA, operational perspective, but it requires careful reason- is created due to the interaction, but it disappears af- ing. For quantum systems we cannot apply classical ter the postselection. However, even after the post- arguments according to which if both particles j and selection, we are still left with a trace of order 2 in k are each in box A, then they are both together in box A of particles 1 and 3 and, in fact, of any other A. The operational meaning of “the particle was in pair except for 1 and 2. Thus, according to the naive the box” is that the particle left a single-particle trace 9 in the box. Similarly, the meaning of “two particles experimentally. were in the box together” is that there is a two-particle Although several experiments exist already for these trace left in the box (as opposed to two single-particle simpler examples, our careful examinations reveal sev- traces). When the particles are pre- and postselected, eral shortcomings that prevent them from being defini- there are cases with two single-particle traces in the tive demonstrations of the paradox. To address these box without a corresponding two-particle trace in the shortcomings, we suggested alternative experimental box. (Note that this possibility requires some entan- implementations that would more convincingly demon- glement between the two single-particle traces.) These strate the failure of the pigeonhole principle. In par- cases correspond to the failure of the pigeonhole prin- ticular, using direct parity measurements of supercon- ciple and is in agreement with the original approach ducting transmons seems like a promising way to com- according to which we decide that a particle is in the pellingly demonstrate the paradox with mesoscopic box if it could be found there with certainty, and two quantum hardware at the near-classical micron scale. particles in the box if they could jointly be found there The quantum pigeonhole paradox traditionally uses together with certainty. the ABL rule to establish elements of reality in the past. We argue that these inferences correspond per- fectly to a more empirically grounded test, namely the VIII. CONCLUSIONS identification of weak environmental traces. That is, pigeons will leave behind footprints in the boxes that We have carefully revisited in what sense the clas- one can later detect. Since such weak traces are de- sical pigeonhole principle may be violated by quan- tectable in experiments, this criteria gives a firm oper- tum systems. We introduced a variation of the tradi- ational meaning for the paradox. tional quantum pigeonhole paradox that corresponds more directly to the statement of the original pigeonhole principle. However, our variation will be chal- ACKNOWLEDGMENTS lenging to implement experimentally due to the need for threshold-type detection that is largely insensitive This work has been supported in part by the to pigeon number. The existing examples, that show U.S.-Israel Binational Science Foundation Grant No. failures of consequences of the pigeonhole principle in- 735/18, and by PBC PostDoctoral Fellowship at Tel stead, have the advantage of being easier to implement Aviv University.

[1] Y. Aharonov and L. Vaidman, Complete description of [11] L. Vaidman, Lorentz-invariant “elements of reality” a quantum system at a given time, J. Phys. A: Math. and the joint measurability of commuting observables, Gen. 24, 2315 (1991). Phys. Rev. Lett. 70, 3369 (1993). [2] L. Hardy, Quantum mechanics, local realistic theo- [12] Y. Aharonov, P. G. Bergmann, and J. L. Lebowitz, ries, and lorentz-invariant realistic theories, Phys. Rev. Time symmetry in the quantum process of measure- Lett. 68, 2981 (1992). ment, Phys. Rev. 134, B1410 (1964). [3] Y. Aharonov, S. Nussinov, S. Popescu, and L. Vaid- [13] G. S. Paraoanu, Non-local parity measurements and man, Peculiar features of entangled states with posts- the quantum pigeonhole effect, Entropy 20, 100405(1) election, Phys. Rev. A 87, 014105 (2013). (2018). [4] Y. Aharonov, F. Colombo, S. Popescu, I. Sabadini, [14] X.-Y. Xu, W.-W. Pan, Q.-Q. Wang, J. Dziewior, D. C. Struppa, and J. Tollaksen, Quantum violation L. Knips, Y. Kedem, K. Sun, J.-S. Xu, Y.-J. Han, of the pigeonhole principle and the nature of quantum C.-F. Li, G.-C. Guo, and L. Vaidman, Measurements correlations, Proc Natl Acad Sci USA 113, 532 (2016). of nonlocal variables and demonstration of the failure [5] B. E. Y. Svensson, Even quantum pigeons may thrive of the product rule for a pre- and postselected pair of together, Proc Natl Acad Sci USA 113, E3052 (2016). photons, Phys. Rev. Lett. 122, 100405 (2019). [6] Y. Aharonov, F. Colombo, S. Popescu, I. Sabadini, [15] Y. Aharonov, D. Z. Albert, and L. Vaidman, Mea- D. C. Struppa, and J. Tollaksen, Reply to Svensson: surement process in relativistic quantum theory, Phys. Quantum violations of the pigeonhole principle, Proc Rev. D. 34, 1805 (1986). Natl Acad Sci USA 68, E3053 (2016). [16] J. Koch, M. Y. Terri, J. Gambetta, A. A. Houck, [7] A. V.S., S. S. Hegde, and T. Mahesh, NMR investiga- D. Schuster, J. Majer, A. Blais, M. H. Devoret, S. M. tion of the quantum pigeonhole effect, Phys. Lett. A Girvin, and R. J. Schoelkopf, Charge-insensitive qubit 380, 577 (2016). design derived from the cooper pair box, Phys. Rev. A [8] M. Waegell, T. Denkmayr, H. Geppert, D. Ebner, 76, 042319 (2007). T. Jenke, Y. Hasegawa, S. Sponar, J. Dressel, and [17] R. Mohseninia, J. Yang, I. Siddiqi, A. N. Jordan, and J. Tollaksen, Confined contextuality in neutron inter- J. Dressel, Always-on quantum error tracking with ferometry: Observing the quantum pigeonhole effect, continuous parity measurements, arXiv:1907.08882 Phys. Rev. A 96, 052131 (2017). (2019). [9] M.-C. Chen, C. Liu, Y.-H. Luo, H.-L. Huang, B.-Y. [18] K. Murch, S. Weber, C. Macklin, and I. Siddiqi, Ob- Wang, X.-L. Wang, L. Li, N.-L. Liu, C.-Y. Lu, and serving single quantum trajectories of a superconduct- J.-W. Pan, Experimental demonstration of quantum ing quantum bit, Nature 502, 211 (2013). pigeonhole paradox, Proc Natl Acad Sci USA 116, [19] S. Weber, A. Chantasri, J. Dressel, A. N. Jordan, 1549 (2019). K. Murch, and I. Siddiqi, Mapping the optimal route [10] P. G. L. Dirichlet and R. Dedekind, Lectures on num- between two quantum states, Nature 511, 570 (2014). ber theory, 16 (American Mathematical Society, 1999). [20] Z. Minev, S. Mundhada, S. Shankar, P. Reinhold, 10

R. Gutiérrez-Jáuregui,R. Schoelkopf, M. Mirrahimi, jectories of remote superconducting qubits, Phys. Rev. H. Carmichael, and M. Devoret, To catch and reverse Lett. 112, 170501 (2014). a quantum jump mid-flight, Nature 570, 200 (2019). [23] L. Vaidman, Past of a quantum particle, Phys. Rev. A [21] D. Riste, M. Dukalski, C. Watson, G. De Lange, 87, 052104 (2013). M. Tiggelman, Y. M. Blanter, K. W. Lehnert, [24] J. Dziewior, L. Knips, D. Farfurnik, K. Senkalla, R. Schouten, and L. DiCarlo, Deterministic entan- N. Benshalom, J. Efroni, J. Meinecke, S. Bar-Ad, glement of superconducting qubits by parity measure- H. Weinfurter, and L. Vaidman, Universality of lo- ment and feedback, Nature 502, 350 (2013). cal weak interactions and its application for interfero- [22] N. Roch, M. E. Schwartz, F. Motzoi, C. Macklin, metric alignment, Proc Natl Acad Sci USA 116, 2881 R. Vijay, A. W. Eddins, A. N. Korotkov, K. B. (2019). Whaley, M. Sarovar, and I. Siddiqi, Observation of [25] L. Vaidman, Lorentz-invariant “elements of reality” measurement-induced entanglement and quantum tra- and the joint measurability of commuting observables, Phys. Rev. Lett. 70, 3369 (1993).