Arxiv:1405.0051V2 [Quant-Ph] 3 Dec 2014 Due to the Many Proposals on the Characterization of Microscopic and Macroscopic Domain

Linking Measures for Macroscopic Quantum States via Photon-Spin Mapping

F. Fröwisa,∗, N. Sangouarda,b,, N. Gisina

aGroup of Applied Physics, University of Geneva, CH-1211 Geneva 4, Switzerland bDepartment of Physics, University of Basel, CH-4056 Basel, Switzerland

Abstract We review and compare several measures that identify quantum states that are “macroscopically quantum”. These measures were initially formulated either for photonic systems or spin ensembles. Here, we compare them through a simple model which maps photonic states to spin ensembles. On the one hand, we reveal problems for some spin measures to handle correctly photonic states that typically are considered to be macroscopically quantum. On the other hand, we find significant similarities between other measures even though they were differently motivated.

1. Introduction for problems concerning the validity and interpretation of quantum mechanics. As well, they immediately yield a The first experiments that triggered the development practical aspect, for instance, in view of efforts to realize of quantum mechanics were conducted by relatively sim- large-scale quantum computing. ple means. After the theoretical framework has been es- Against this background, it is somehow astonishing tablished, the effort to experimentally verify some much that a commonly accepted framework of “macroscopic quan- more demanding predictions like entanglement and non- tum physics” is still lacking. The famous gedanken ex- locality increased significantly. Nowadays, we master experiment of Schrödinger [18] ponders on the existence of perimental techniques that even led to commercial prod- large objects in a superposition of two classical, distinct ucts such as secure communication and true random num- states like a cat being dead and alive. Complementary, ber generators. Furthermore, it is by now possible to en- Leggett [19, 20] argued that, in large systems, there is ter the quantum regime of “large” systems; large either difference between an accumulated quantum effect origi- in terms of mass, energy or number of involved micro- nated on a microscopic scale and a “true” quantum effect scopic constituents. Among other contributions, experi- on a macroscopic scale. While the former is undoubtedly menters brought superconducting devices [1–3] and mas- an experimental challenge due to the complexity and the sive mechanical oscillators [4, 5] to the quantum regime; large number of degrees of freedom, only the latter is sup- one observed interference effects with giant molecules [6], posed to provide insight into the aforementioned problems. entangled diamonds [7], cells [8], doped crystals [9] and Based on these and other contributions [21], many physi- large spin ensembles [10–12]; we also witnessed entangle- cists came up with measures to quantify how “macroscopi- ment between optical modes including hundreds of pho- cally quantum” a state is [22–30]. Such mathematical def- tons [13, 14]. Arguable, all these experiments show quan- initions potentially provide a clear view on macroscopic tum behavior in large systems. But how could one com- quantum effects. Furthermore, an established definition is pare them? In which sense is one more macroscopically the basis for theoretical conclusions of, for example, the quantum than another one? Answers to these and simi- stability of macroscopic quantum states with respect to lar questions would allow us to challenge old but still un- noise [22, 31] and measurement imperfections [32]. solved problems. One of those is the transition between arXiv:1405.0051v2 [quant-ph] 3 Dec 2014 Due to the many proposals on the characterization of microscopic and macroscopic domain. How and in which macroscopic quantum states, it is clearly necessary to com- sense do large systems become “classical” — after all, they pare those measures in order to understand the similari- are composed of microscopic particles that are quantum ties and differences. First attempts have been made in mechanically in nature? Some proposals try to answer Refs. [28], where several measures suitable for spin mea- suchlike questions within quantum theory. For instance, sures [22, 23, 25, 26, 28] have been classified. Another work the decoherence program [15] provides a mechanism for [33] applies some measures [21, 25, 26, 28] to a specific the loss of quantum correlations that typically becomes multi-mode photon state. The ultimate goal is to provide stronger for larger systems. Other ideas suggest a solution a general framework for macroscopic quantum effects that by extending the theory as it is done, for example, in col- covers all important physical systems. With this, one is lapse models [16, 17]. Clearly, these efforts are important able to directly compare different systems. For instance, ∗Corresponding author one could then compare experiments on trapped ions with Email address: [email protected] (F. Fröwis) massive objects in the superposition of spatial positions

Preprint submitted to Elsevier November 8, 2018 (see additional remarks in Sec. 5). (some proposals are even defined for mixed states). The In this paper, we aim to continue this research line larger f(ψ), the more macroscopically quantum |ψi is. Of- by bridging measures that were formulated for spin sys- ten, f is called the effective size of |ψi. The qualitative dis- tems [22, 23, 25, 26, 28] and for single photonic modes tinction between macroscopic and non-macroscopic quan- [27, 30] (some of them are valid for both systems). To this tum states based on f is to some extent arbitrary. end, we use a simple model of a photon-spin mapping, in System size.— All published proposals agree that a particular, the absorption of a photonic state into a spin quantum state can only be macroscopically quantum if the ensemble. Under the assumption that the properties of respective system size is in some sense “large”. For systems the photonic state are completely mapped to the spin en- composed of microscopic particles, it is necessary to have semble, we have a tool at hand to analyze and compare in many constituents. If one considers one (or few) bosonic which sense states are macroscopically quantum according modes, we require to have high excitation numbers or high to different measures. masses. The exact values for having a “large” system are In the following, we draw some conclusions for several not crucial for the present discussion. As we are concerned measures based on this mapping. As we will see later, it with spin and photonic systems in this paper, the system is necessary that the mean photon number, N, of the con- size is defined as the number of spin-1/2 particles, M, or sidered state is much smaller than the number of the spins the mean photon number, N, respectively. in the ensemble, M. After the mapping, N corresponds to Schrödinger-cat state vs. macroscopic quantum state.— the excitation level of the spins. In this regime, we observe A first distinction of the current literature can be made by that some measures for spin states behave differently than the basic form of the states considered to be macroscopic. in the case where N is comparable with M, which is the Some authors [23, 25, 26, 30] consider superpositions of regime where they have been studied so far. Apparently, two (or a few) “classical” states like through this work we also understand better the present proposals and learn about the implications of their initial |ψi ∝ |ψ0i + |ψ1i . (1) intuition. On the other side, we find that there are tight mathe- In a simplified way, one can say that one seeks a mathematical connections between certain measures, even though matical definition for the verbal characterization that the the physical motivation for introducing them is apparently states |ψ0i and |ψ1i are “macroscopically distinct” (Leggett very different. We conclude therefore that, at least par- [19]). In the remainder of this paper, we call macroscopic tially, there exists already some consensus on the charac- superpositions of the form (1) a Schrödinger-cat state. terization of macroscopic quantum states among the present On the other side, some proposals [22, 24, 27–29] do not proposals. require a specific form of the quantum state and may even This paper is structured as follows. In Sec. 2, we set allow for mixed states. States that are macroscopically the basic nomenclature and summarize the existing pro- quantum due to these definitions are here called macro- posals in the field of macroscopic quantum states. We also scopic quantum states (see examples below). If any confu- review some established implications on the stability. In sion is excluded, we also use this term as an umbrella term Sec. 3, we introduce and elaborate on the model for the that includes the concept of a Schrödinger-cat state. photon-spin interaction we use to link different measures. Scaling versus absolute numbers.— Experiments that Some implications are discussed in Sec. 4. Conclusions and aim to generate macroscopic quantum states may be in- outlook are given in Sec. 5. terested in states |ψi with large f(ψ). In such a case, the absolute value of the normalized f(ψ) is the figure of merit. In turn, in many theoretical studies, one does not only con- 2. Review of measures for macroscopic quantum sider a fixed system size but more abstractly state families. states A state family is a prescription that assigns to any system size a quantum state |ψi. Then, one can investigate, for In this section, we first clarify some subtle but im- example, the scaling of f(ψ) with the system size.1 The portant points for the discussion of macroscopic quantum larger the scaling for f(ψ), the more macroscopically quan- physics (Sec. 2.1). Then, in Secs. 2.2 and 2.3, we give a tum is the respective state family |ψi. In this paper, we rough overview on some measures for macroscopic quan- focus on how the considered measures scale. In particular, tum states that have been proposed so far. In Sec. 2.4, we we say that |ψi is macroscopically quantum with respect discuss some implications on the stability in the presence to f(ψ) if the respective state family scales linearly with of noise and imperfections. the system size. Furthermore, two measures are said to be compatible if they identify the same class of macroscopic 2.1. Prelimiary discussion quantum states. Common goal of the measures.— The common feature of all works that are considered in this paper is to identify among all quantum states those that are macroscopically 1The scaling of a function f with N is defined as f = O(g(N)) :⇔ quantum. This is done by defining a function f(ψ) ≥ 0 limN→∞ f/g exists and is strictly positive.

2 Subjectivity.— Finally, we add a comment that was, in maximal value σ such that |ψ0i and |ψ1i are distinguish- similar words, already made in several other works. In or- able with a success probability Ps of at least Pg ∈ (1/2, 1], der to identify macroscopic quantum states, one has to which is free parameter of the measure. Thus, the effective introduce some kind of restrictions or some “structure” size is reads within the Hilbert space. This is similar to entanglement √ −1 theory, where one has to define a tensor structure in a high- SizePg = max 2 2erf (Pg − 1)σ (3) σ:Ps≥Pg dimensional Hilbert space to distinguish between separable and entangled states. for general Pg, where erf is the error function. As an The restrictions for the measures of macroscopic quan- example, one finds that Size2/3 ≈ maxσ:Ps≥2/3 0.86σ. tum states are to some extent subjective. This fact could Let us examine the measure for the photonic state be used to criticize the measures. However, in many pro- 1 posals, the predefined structure is motivated by actual lim- |ψi = √ (|0i + |2Ni) , (4) itations we encounter in nature (and therefore in the lab). 2 For example, the restriction to certain (local) measure- which is often considered to be a Schrödinger-cat state. It ments [22, 25, 30], local interactions [28] or certain noise is straightforward to see that σ can be up to O(N) for any models [27] reflect such strong limitations. We believe that value P ∈ (1/2, 1); hence it identifies |ψi as macroscopi- they are necessary in order to identify a qualitative differ- g cally quantum. ence between macroscopic and accumulated microscopic As a second example, consider the archetypal macro- quantum effects. scopic photonic superposition Note that, for spin systems, a measurement or interaction is called local, if the respective operator can be written 1 as a sum of terms where each term has nontrivial support |Ψαi = (|αi + |−αi) , (5) p2(1 + exp(−2|α|2)) only on one particle. For example, the interaction with an external magnetic field oriented in z direction is generated where | ± αi = D±α |0i is a coherent state. The operator by † ∗ Dα = exp(αa − α a) (with a as the annihilation operator M X and α ∈ C) is the so-called displacement operator and |0i J = σ(i), (2) z z is the vacuum state. The states |αi and | − αi are not i=1 distinguishable with photon number measurements. How- (i) ⊗i−1 ⊗M−i where σz ≡ id C2 ⊗ σz ⊗ id C2 is a Pauli operator ever, its macroscopic quantum nature can be revealed by acting on the ith spin. For the sake of simplicity, we extending the definition of the measure to all “state of the only consider operators that are symmetric with respect to art” measurements, like homodyne measurements. This qubit permutations. Furthermore, all local operator con- modification enlarges the class of Schrödinger-cat states sidered in this paper are normalized to a spectral radius to states like |Ψαi. that equals the number of qubits.2 We note that many of We emphasize that this measure is not invariant under the following concepts can be generalized to “quasi-local” displacement operations. This can be illustrated by a third operators in the sense that each addend has nontrivial sup- example, which is the displacement of a single photon in port on O(1) qubits (e.g., nearest-neighbor interaction). the superposition of two different modes [34],

2.2. Measures for photonic systems − 1 ψα = √ Dα ⊗ id (|0, 1i − |1, 0i) . (6) We now review two measures that are naturally defined 2 for photonic systems. In order to apply the coarse-grained-based measure, con- Coarse-grained-based measure [30].— Consider two states sider a basis change to |±i ∝ |0i±|1i. The question is thus |ψ0i and |ψ1i. Assume there exists a measurement that how well can we distinguish between Dα |+i and Dα |−i. distinguishes them with high success probability. To call As shown in Ref. [30], the effective size reads the two states macroscopically distinct, in Ref. [30] it is re- quired that they are distinguishable even with very coarse- s − −1 4 grained (i.e., “classical”) detectors. SizePg ( ψ ) = 2αerf (2Pg − 1) − 2, (7) α π(2P − 1)2 The specific definition is based on photon-number de- g tection in one-mode states. The coarse-graining is intro- that is, it scales with the square root of the photon num- duced via a Gaussian distribution that fixes the probabil- ber in the regime where P ∈ (1/2, 1/2(1 + p2/π)) ≈ ity to measure m photons when indeed there have been n g √ (0.5, 0.8989). present, p(m|n) = 1/( 2πσ) exp(−(n−m)2/(2σ2)), where Besides the obvious intuition to define macroscopic dis- σ ≥ 0 is the parameter that adjusts the measurement ac- tinctness by realistic detectors, there is a more abstract curacy. The effective size of superposition (1) is then the motivation to connect this measurement to the notion of Schrödinger-cat states. The specific noise model is such 2This is also the reason why we omit the factor 1/2 in Eq. (2). that nearby eigenstates of the photon number operator 3 are less distinguishable. The distinctness of two states Index p [22, 24].— Consider a system of M compo- that are hardly distinguishable in the presence of noise nents and a local measurement A (e.g., Jz). The moti- is hence based on “microscopic details”. The noise model vation for introducing the index p is the observation that helps to reveal how much their distinguishability depends a typical classical probability distribution√ exhibits a stan- on microscopic and macroscopic differences, respectively. dard deviation in the order of M. The ratio of standard Counting oscillations in the Wigner function [27].— deviation and spectral radius vanishes in the limit of large The motivation for this measure is most easily gathered by N. However, there exists quantum states and correspond- the following example. Compare the superposition |Ψαi ing local observables where the standard deviation is also with the incoherent mixture 1/2 |αihα| + 1/2 |−αih−α|. in the order of M, meaning that even for very large system The inspection of the respective Wigner functions shows sizes, one observes persistent fluctuations in the measure- that both states exhibit strong peaks around the values ment statistics. ±α. But while for the incoherent mixture one has a sum The index p is therefore defined as of two (positive) Gaussian functions, the Wigner function p of |Ψ i shows strongly fluctuating oscillations between the max Vψ(A) = O(M ), (12) α A:local peaks. 2 2 The authors of Ref. [27] systematized this insight by where Vψ(A) = hA iψ − hAiψ is the variance of A with re- defining the effective size of a state ρ that measures the spect to |ψi. The maximization is carried out over the set frequency and the amplitude of the Wigner function. They of symmetric, normalized local operators. If p = 2, the |ψi find that this quantity can be calculated without directly is called a macroscopic quantum state (in Ref. [22], these invoking the Wigner function. One has states are called anomalously fluctuating states). The GHZ state yields V(J ) = M 2, which is the maximal value. 2 z I(ρ) = −Tr[ρL(ρ)] + Trρ /2 (8) The extension to mixed states is nontrivial. An incoherent mixture |0ih0|⊗M + |1ih1|⊗M leads to the same with variance, but is typically not considered as a macroscopic X 1 1 quantum state. In Ref. [36] (see also Ref. [24]), the so- L(ρ) = − a ρa† − ρa† a − a† a ρ , (9) m m 2 m m 2 m m called index q is introduced as m q where m is the index for the different modes of ρ. Note that max(M, max k[A, [A, ρ]]k1) = O(M ) (13) A:local the last term was proposed later [35] in order to guarantee √ † positivity of the measure. For pure, single-mode states, where kXk1 = Tr XX is the trace norm of X. A tight |ψi,Eq. (8) reduces to connection to the index p was shown in the case of pure ρ. Measure based on “increasing the interferometric util- † 2 1 I(ψ) = ha aiψ − |haiψ| + . (10) ity” [23].— Another way to characterize Schrödinger-cat 2 states is to ask how much more useful is superposition (1) One easily sees that the Fock state |Ni thus is a macro- compared to its components |ψ0i and |ψ1i. The choice scopic quantum state as I = N + 1/2. The superposition of the specific task is clearly a crucial aspect of this kind |0i + |2Ni exhibits the same effective size and is hence not of characterization. Nevertheless, there does not seem a more macroscopically quantum as |Ni. conceptual difference between choosing a certain task and † 2 2 For |Ψαi, one has that I = ha ai+1/2 = |α| tanh |α| + restricting oneself to certain (local) Hamiltonians and mea- 1/2, while the incoherent sum is zero-valued. surements (as done for other measures). The authors of Note that I(ρ) is invariant under displacing ρ, that is, Ref. [23] focus on interferometry experiments with a given † − generator A under idealized conditions and introduce the I(DαρDα) = I(ρ). Therefore, the state |ψα i is not considered to be macroscopically quantum by this measure. quantity M = θsing/θsup, where θsing and θsup are the sensitivities of the single states |ψii (i = 0, 1) and the su- 2.3. Measures suitable for spin systems perposed state |ψi, respectively. In principle, this idea is In this subsection we mention the measures that are applicable to photonic and spin systems. suitable for spin systems (even if they are not exclusively In Ref. [28], it is argued that one should square M in defined for spins). This part is compactly written since a order to compare this measure to others like [25, 26, 28]. more extended review is already given in either Refs. [28, As shown in Ref. [23], one then finds 33]. As a simple example for the measures, we consider (hAi − hAi )2 the Greenberger-Horne-Zeilinger (GHZ) state M2 ≈ ψ1 ψ0 . (14) Vψ0 (A) + Vψ1 (A) 1 ⊗M ⊗M |GHZi = √ |0i + |1i , (11) The quantity M is defined only when |ψ i are not 2 i eigenstates of A. The later happens, for instance, to the with |0i and |1i as the eigenstates of σz with eigenvalue 1 GHZ state (11) with the generator A = Jz. This issue and −1, respectively. can be fixed by specifying the maximum benefit of |ψi 4 compared to maximum benefit of the single components, is defined as average distance of |ψ1i to |ψ0i via where the maximum is taken over the all “feasible” genera- ∞ tors (where the term feasible is clearly to be defined). The X 2 D¯ = |λ | d. (17) maximal utility for |0i⊗N is given –in the case of the re- d P (i) d=0 striction to local operators– for A = Jx := i σx . Then, 2 ⊗M ⊗M we find for the GHZ state M = 2M. For |ψ0i = |0i , one easily sees that |1i ∈ HN . Measurement-based measure [25].— The analysis of Therefore, D¯ = M, which coincides with the intuition. Schrödinger’s gedanken experiment reveals that a macro- While this definition is in general not limited to qubit scopic superposition is composed of two states that are dif- systems, one might disagree on the right microscopic step ferent in a macroscopically large number of “sites”. Taking in other systems. For example, the annihilation operator the cat example literally, the dead and the alive cat dif- a may be a good candidate in one-mode photonic systems, fer in every biological cell. Translated to a mathemat- but this is not clear at all. ical definition, the authors of Ref. [25] call the many- Measure based on quantum Fisher information [28].— particle superposition (1) macroscopically quantum if one The goal of this proposal is to identify a macroscopic quan- can find a macroscopic number of distinct groups such tum effect that is clearly distinct from any accumulated that the measurement of one such group let one distin- microscopic quantum effect. To this end, consider the uni- guish between |ψ0i and |ψ1i. For a more flexible char- tary time evolution and ask, for given initial state and acterization, one introduces a minimal success probability system Hamiltonian H, what is the minimal time in order Pmin = 1−δ, which has to be overcome by each group. The that the evolved state becomes orthogonal to the initial one-shot probability for a successful distinction is given by state. This time is called orthogonalization time τ. It the trace norm of the reduced density operators. Define turns out that, for pure states |ψi, the variance of H gives (k) p ρi = TrM\k |ψiihψi|, where k is a subset of all qubits. a lower bound on τ, that is, τ ≥ π/(2 Vψ(H)) [37]. As The success probability then equals in the discussion of the index p, one observes that, for local Hamiltonians, some entangled states like the GHZ 1 1 (k) (k) (k) state show a different scaling of the variance than sepa- PS = + kρ0 − ρ1 k1. (15) 2 4 rable states. It is therefore possible to reach much lower The effective size is defined as values of τ by using entangled states. Concerning mixed states, one notices that an improve- Cδ(ψ) = M/nmin, (16) ment of the bound on τ is by using the so-called quantum Fisher information (QFI) Fρ(H); see, for example, where nmin is the minimal number of qubits in order to Ref. [38] for the precise mathematical definition. Here, we (nmin) distinguish between |ψ0i and |ψ1i with PS ≥ 1 − δ. just note that for pure states, one has Fψ(H) = 4Vψ(H) As an instance, consider the GHZ state. The measure- and in general Fρ(H) ≤ 4Vρ(H). One can show [39] that p ment of each qubit in the basis |0i and |1i gives perfect dis- τ ≥ π/ Fρ(H). tinguishability. Therefore, nmin = 1 and C1(GHZ) = M. The definition of macroscopic quantum states based on Measure based on the notion of a “microscopic step” the QFI is now the following. Given a state ρ, one maxi- [26].— Another way to characterize macroscopically dis- mizes Fρ(H) within the set of all “feasible” Hamiltonians. tinct states is to introduce the concept of a microscopic In Ref. [28], the set of local operators is chosen. The ef- step. In the case of qubits, the natural unit is the creation fective size is then determined as and annihilation of a spin excitation, that is, σx |0i = |1i. Fρ(H) Then, one “counts” how many of those steps one has to Neff (ρ) = max , (18) H:local 4M apply to, say, |ψ0i on average in order to reach |ψ1i. The larger this number, the more macroscopically distinct are where the normalization ensures that maxρ Neff (ρ) = M. these two states. For example, the two components of the Again, the maximum is attained by the GHZ state. The GHZ state are macroscopically far apart, since one has to interpretation of Neff is the following. A state that ex- apply qubit flip operations to all qubits of |0i⊗M to reach ⊗M hibits a large QFI gives rise to more rapid changes in the |1i . expectation values of some observables, for example, the To make this idea mathematically more rigorous, start projection onto the initial state. Its experimental verifica- with the subspace H0 = span {|ψ0i}. For d > 0, itera- tion is a macroscopic quantum effect, since it cannot be tively construct H˜d as the span of all states that can be (i) reproduced by separable states nor by states that exhibit generated by applying local operators like σx to all states only few-particle entanglement [39–41]. ˜ |φi ∈ Hd−1 and define Hd = Hd\Hd−1. In words, Hd Based on Eq. (18), one can also formulate a measure contains all states that are one (nontrivial) microscopic for Schrödinger-cat states (1) by defining step way from Hd−1 (and also from Hd+1). Then, represent |ψ1i in terms of this Hilbert space structure, that is, rF Neff (ψ) P∞ Neff (ψ) = . (19) |ψ1i = λd |vdi with |vdi ∈ Hd. The effective size D¯ 1 1 d=0 2 Neff (ψ0) + 2 Neff (ψ1) 5 This quantity is called relative Fisher information. implies that one cannot efficiently verify the experimen- Classification.— The measures that are presented in tal generation of Schrödinger-cat states in a strict sense.3 Sec. 2.3 were classified in Ref. [28]. Under certain restric- Note that this result was –so far– not shown for more gen- tions, which are necessary for some measures to be well- eral macroscopic quantum states identified by the index p defined (see Appendix A and Ref. [28]), one finds the gen- and the QFI. eral result that Questions regarding the connection of Schrödinger-cat

2 rF states and measurement imperfections are considered in M = O(M) ⇔ Neﬀ = O(M) ⇒ Ref. [32]. There, it is shown that the more distinguish- ¯ Cδ = O(M) ⇔ D = O(M) ⇒ (20) able two one-mode photonic states are with respect to

p = 2 ⇔ Neff = O(M). coarse-grained measurements (see Sec. 2.2), the more fragile is the detection of the relative phase of their superposi- Relations (20) should be read in the following way. Sup- tion. Measurement imperfections such as insensitivity and pose, for a given quantum states |ψi as in Eq. (1), one coarse-graining renders an experimental verification of the 2 rF finds that M = O(M). This holds if and only if Neff = Schrödinger-cat state more challenging as the effective size O(M) and it implies that, for example, Cδ = O(M). The of the state increases. first strict implication (from the first to the second line The results of Refs. [31, 32] show a certain duality of in Eq. (20)) distinguishes between two different concepts Schrödinger-cat states. The more two states are macro- within the set of Schrödinger-cat states. The implication scopically distinct, the harder it is to observe the relative from the second to the third line in the same relation shows phase between the respective superposition. The loss of that the concept of macroscopic quantum state is strictly phase information, however, reduces the coherent super- more general and includes Schrödinger-cat states defined position to an incoherent mixture, which is in general not via the proposals [23, 25, 26]. Examples of such a macro- macroscopically quantum anymore. scopic quantum state that are not a Schrödinger-cat state (according to Refs. [23, 25, 26]) are states generated by 3. Mapping between photonic states and spin en- optimal covariant cloner [42], or, more generally, certain sembles spin-squeezed states. In this section, we are going to introduce and derive 2.4. Implications on stability of Schrödinger-cat states the main tool for the comparison between measures for While archetypal Schrödinger-cat states like the GHZ photonic and spin systems, respectively. It is a mapping state has been identified as important quantum states in of photonic states to spin states. The spins represent two- large systems, it also became more and more evident that level atoms, which are initially in the ground state. They there seems to be a tight connection between Schrödinger- fully absorb the incoming light field and are thus excited. cat states and their fragility under noise. This intuition This allows us to investigate the effective size for photonic was recently supported by mathematical statements [22, states with measures that are a priori not formulated for 31, 32]. this system. Macroscopic quantum states in the sense that p = 2 The total Hilbert space is the tensor product of a pho- were shown to be fragile with respect to certain corre- tonic field and M two-level systems, that is, H = Hph ⊗ 2 2⊗M lated noise that is generated by local operators [22]. With Hs ≡ ` (C) ⊗ C . The ground state and the excited Eq. (20), this implies instability also with respect to all state of a single spin are denoted as |gi and |ei, respec- other measures presented in Sec. 2.3. On the contrary, tively. We model the Hamiltonian as states that exhibit p = 1 are stable in the sense that there † does not exist any such noise model where the total deco- H = χ(a ⊗ J+ + a ⊗ J−), (21) herence rates exceeds the sum of the individual noise rates R PM (i) [22]. Note that there are similarities to the statements where χ ∈ is the coupling strength, J+ = i=1 |eihg| † of Ref. [27] concerning the stability of macroscopic states and J− = J+. The operators J+ and J− create and annihi- according to Eq. (8). late atomic excitations, respectively, which are symmetri- In Ref. [31], the effect of uncorrelated noise models on cally distributed within the atomic ensemble. This Hamil- Schrödinger-cat states are investigated. Consider the set tonian does not take certain details into account such as of states (1) whose effective sizes according to Refs. [25] the spatial distribution of the spins and the incoming pho- scale as O(M). One can show that if a state from this set tons, or any propagation effects. Nevertheless, it is in first is subject to specific local noise channels (like local depo- order a good approximation to the absorption step in a larization noise), the success probability to distinguish this quantum memory [43]. noisy state from initially orthogonal states decreases exponentially in M. In particular, one cannot efficiently distin- 3See Ref. [31] for details how to modify the Schrödinger-cat states guish |ψ0i+|ψ1i from the mixture |ψ0ihψ0|+|ψ1ihψ1|. This to avoid this kind of instability. For the prize of reducing the effective size, one can gain stability with respect to this criterion.

6 It is convenient to introduce the so-called Dicke states initial state |ki ⊗ |M, 0i yields

X ⊗k ⊗N−k |M, ki ∝ πl |ei ⊗ |gi , (22) U |ki ⊗ |M, 0i ≈

πl k s X k (26) cosk−l(g)(−i sin g)l |k − li ⊗ |M, li , where one sums over all particle permutations πl. The l l=0 states |M, ki are properly normalized and form a basis in the subspace of permutation symmetric states. The which corresponds to a binomial distribution in excitation operators J± together with Jz = [J+,J−] generate a so- number basis. We fix g = π/2 in order to fully absorb the called SU(2) algebra with [Jz,J±] = ±2J±. One can easily photonic mode by the spin ensemble. Then, one has that k show that Jz |M, ki = (−M + 2k) |M, ki and J± |M, ki = U |ki ⊗ |M, 0i ≈ (−i) |0i ⊗ |M, ki. Hence, the physical 2 C± |M, k ± 1i with C± = M/2(M/2+1)−(M/2−k)(M/2− interpretation of our approximation is that, due to the k∓√1). With this, it is easy to see that, e.g., H |1i⊗|M, 0i = large number of spins, the probability that a single atom χ N |0i ⊗ |M, 1i. is “hit by two photons” is negligible. If we now write the P We now derive an approximate formula for the prop- initial photonic state in the Fock basis, |ψi = k ck |ki agator U = exp(−iHt), which becomes exact in the limit with ck ∈ C, we simply find of infinite number of spins. We remark that dealing with X k large spin ensembles is necessary in order to ensure that U |ψi⊗|M, 0i ≈ |0i⊗ (−i) ck |M, ki =: |0i⊗|φi . (27) the mapping from the photonic mode to the spins pre- k serves all properties of the light mode. Another impor- We denote the set of symmetric states with hJ+J−iφ ≤ tant point is that the final results are independent of M N M by SN . for M N, which is a posteriori verified, see Sec. 4. In the same way as photonic can be mapped to spin For the derivation, we first consider the operators X+ states, photonic operators have their spin counterparts. † and X− = X+ and X3 := [X+,X−]. If the relation The annihilation operator a is transformed via 2 [X3,X±] = ±2c X±, c ∈ R, is fulfilled, the three oper- † † i ators generate a SU(2) algebra. Then, one can show for −iπ/2(aσ++a σ−) iπ/2(aσ++a σ−) e ae ≈ −√ J−. (28) V = eλ(X++X−) that M

2 V = etanh(cλ)/cX− cosh(cλ)X3/c etanh(cλ)/cX+ . (23) This can be shown using Eq. (24) with A = −ig(a ⊗ σ++ † a ⊗ σ−) and B = a; noticing that [A, B] = −iπ/2σ− and This is done by the ansatz V = exp(f(λ)X−) exp(g(λ)X3) [A, σ−] ≈ −iπ/2a. exp(h(λ)X+). Differentiating V with respect to λ and Finally, let us calculate an interesting example, namely successive application of the identity a coherent state |αi as the photonic input state. Its ampli-√ 2 k tudes in the Fock basis read ck = exp(− |α| /2)α / k!, ∞ X 1 which is the square root of a Poisson distribution. Note eABe−A = [A, . . . [A,B] ... ], (24) n! that even though the coherent state contains Fock states n=0 | {z } n times |ki that clearly do not satisfy the requirement k M, 2 the arising differential equations are solved with the bound- their total contribution is negligible in the limit M |α| . ary conditions f(0) = g(0) = h(0) = 0. We therefore expect only an exponentially small error for † the approximation summarized in Eq. (25). Then, the Now, the problem is that X+ = a ⊗ J− does not give rise to a SU(2) algebra, because of this assignment it fol- Poisson distribution can be very well approximated by the † binomial distribution, that is lows that X3 = −a a ⊗ Jz − J+J− does not fulfill the necessary commutation relation with X±. However, if we s !(M−k)/2 M |α|2 α k consider only states |Ψi ∈ H with a comparatively low ex- c ≈ 1 − √ , (29) † k citation number ha a+J+J−iΨ = N M, we can restrict k M M ourselves to an effective subspace spanned by |M, ki where 2 (M−k)/2 2 k M. In this limit, one has hJzi|M,ki = −M + 2k ≈ if we approximate (1 − |α| /M) ≈ exp(− |α| /2) k k −M, that is, Jz acts approximately as an identity and we and M!/(M − k)! ≈ M . Including now the factors (−i) † find X3 ≈ Ma a − J+J−. Then, the commutation√ rela- as in Eq. (27), we observe that this is exactly the distri- tions are approximately√ fulfilled,√ where c = M. With bution of the product state σ := J / M and g = χ Mt, we find the main result of ± ± s ⊗M this section |α|2 α |φαi = 1 − |gi − i√ |ei . (30) † †   a a−σ+σ− M U ≈ e−i tan(g)aσ+ (cos g) e−i tan(g)a σ− . (25) M

A straightforward calculation of the action of U on the This is an interesting result, since it shows that the coherent state, which is often considered as the “most classical” pure photonic state, does not generate entanglement 7 4 within the spin ensemble. Instead, every spin (being ini- In order to maintain the implication Cδ = O(N) ⇒ Neﬀ = tially in the ground state) is locally rotated. O(N), we have to constrain the measurements to be local In contrast, a pure Fock state |Ni induces much entan- (or quasi-local). glement between the spins, since the respective spin state Secondly, it is necessary to revise the index p. The N is the Dicke state |φi = (−i) |M,Ni. Again with the ap- maximal variance for state from SN scales with O(MN), proximation M N, a simple expression for the Schmidt that is, no p = 2 scaling can be achieved. One can still decomposition of a 50 : 50 splitting between the spins is deﬁne macroscopic quantum states via the index p for found to be |φi ∈ SN by dividing the maximal variance by the num-

N ber of qubits. Then, the state is called macroscopic if this N X √ number scales with the maximal scaling, which is here N. |M,Ni = (−i) pl |M/2, li ⊗ |M/2,N − li , (31) For pure states, this is mathematically equivalent to the l=0 QFI approach (18). In the following, we refer to the “mod- N N ified index p”. Notice that now the quantum fluctuations with pl ≈ l /2 . The entanglement between these equally- sized block can then be measured, for example, by the –for pure states measured by the standard deviation of P entanglement entropy E = − l pl log pl [44]. One finds a local observable– are not anymore in the order of the with further approximations that E = O(log N). spectral radius, which was the initial motivation for the definition of the index p. Hence, this slightly modifies the interpretation of the measure [22]. Now, the quantum fluc- 4. Applying the photon-spin mapping to measures tuations are persistent on a scale which is N times larger of macroscopic quantum states than what is expected from “classical states” like (spin) co- In this section, we use the mapping (27) to compare herent states instead of being persistent on the scale linear measures for photonic and spin systems. Generally, if we in M. use Eq. (27) to map photonic states to spin ensembles, we To illustrate the relations (32), we study four simple examples of photonic states and calculate their effective consider symmetric states from a “low-energy sector” SN , size for the spin measures of Sec. 2.3. These are the su- which is just a small part of the entire Hilbert space HS. The maximal scaling of the effective size is supposed to be perposition of two coherent states with opposite phases, − in the order of the system size of the photonic state, which |Ψαi, the displaced single photon, |ψα i, the superposition is the average photon number, N. This is a different order of two Fock states |0i + |2Ni and a single Fock states |Ni. than the total number of spins, M. The results are summarized in table 1. We start by discussing whether the relations (20) apply The state |Ψαi translates via the mapping (27) into a superposition of two spin-coherent states |φ i + |φ i also if we restrict ourselves to SN . In addition, we also √ α −α 2 M study four basic examples of photonic states examined by with hφα|φ−αi = (1 − ) and = 2|α|/ M 1 (see the “spin measures” reviewed in Sec. 2.3 to illustrate our Eq. (30)). This state has been intensively investigated in findings. Finally, we discuss direct links between measures the literature; for instance in Refs. [21, 23, 25, 26, 28]. for photonic and spin systems. All these measures agree in assigning an effective size of approximately M2 = 2 |α|2. 4.1. Spin measures used for photonic states: four examples The next example is the displaced single photon state, − |ψα i (see Eq. (6)). It is easy to see that all spin measures The restriction to the subset SN immediately poses the question whether similar relations like (20) hold. Indeed, are invariant under local rotations. Since the displacement as shown in Appendix A, one finds that operator is mapped to such a local rotation, the effective size according to these measures is the same as for a Bell 2 rF state, which is –due to the low number of spin excitations– M = O(N) ⇔ Neff = O(N) ⇒ necessarily a microscopic state. Cδ = O(N) ⇔ D¯ = O(N) ⇒ (32) 00 The third example is |0i + |2Ni (see Eq. (4)), which “Modified index p ⇔ Neff = O(N) is mapped to |φi ∝ |M, 0i + |M, 2Ni. The interference- based measure [23] has some problems dealing with this for states |φi ∈ S (see also Fig. 1). In addition to the N state. With the original definition (14), it diverges for the assumptions needed for relations (20), one has to require choice A = a†a, which correspond to J J /M in the spin the following. + − case. If we instead maximize the denominator of M2 as An effective size C = O(N) implies that the number δ suggested in Ref. [28], the measure becomes vanishingly of spins per measured group is in the order of M/N 1. small, since V|2Ni(Jx) ≈ M(4N + 1) while the difference in the expectation value of a†a scales with N. 4We note that an numerically exact simulation of the interaction The relative Fisher information [28] performs similarly. (21) of a coherent state with the spin ensemble indicates the gener- The reason is that the maximal Fisher information is al- ation of a small amount of entanglement within the ensemble. We found numerical evidence that, for g = π/2 and |α|2 M, the ready large for high-excitation Dicke states. We find that negativity between a fifty-fifty splitting of the spin ensemble equals Neff (φ) = 2N + 1, while one has Neff (|M, 0i) = 1 and |α|2 /(4M). 8 rF Neff (|M, 2Ni) = 4N +1. With Eq. (19), one has Neff (φ) = access to only a subset of all particles. The latter select in- 1. The relative Fisher information does not identify the sensitive photon number detectors as a realistic constraint state of Eq. (4) as a Schrödinger-cat state since already in high-photon-number scenarios. one constituent is considered to be a macroscopic quan- As we have seen in Eq. (33), the reduced density ma- tum state. trix of Dicke state is a binomially distributed, incoherent The measurement-based measure [25], on the other side, mixture of Dicke states. If we translate the Dicke states assigns to this state a large effective size. To see this, di- back to photonic Fock states, we see that this is equiv- vide the spin ensemble into x groups. The reduced density alent of considering the Fock state after a beam splitter x matrices of one such group read ρ0 = |M/x, 0ihM/x, 0| with transmitivity 1/x, where the reflected part is traced (M/x) P2N (M/x) out. The photon number detection right after the beam and ρ2N = k=0 pk |M/x, ki hM/x, k|, with splitter therefore corresponds to an inefficient detector. It (M/x) 2N k 2N−k is an open question how far reaching the similarities of the pk ≈ (1/x) (1 − 1/x) , (33) k two measures are. In particular, does Cδ = O(N) imply

that SizePg = O(N) and vice-versa? Even if this is not respectively. Therefore, the success probability (15) to the case, there is clearly a strong connection between the (M/x) x distinguish the two states equals PS = 1 − p0 /2 ≈ measures (and the microscopic-step-based measure [26]). 1 − (1 − 1/x)2N /2. In order to distinguish with sufficiently We thus conclude that the two measures are equivalent up high probability for any N, one can maximally set x = cN to the choice of the noise model for the detection. It is with proper value of c depending on the given threshold δ. clear that this choice depends on the experimental situ- Hence, we find Cδ = O(N). ation and, for theoretical studies, one may even consider Similarly, the “microscopic-step”-based measure [26] as- both measures in parallel. signs a large effective size to the state (4), since one needs exactly 2N step in “units” of J+ in order to reach |2Ni 4.3. Comparing measures for general macroscopic quan- from the vacuum. tum states The QFI-based measure and the (modified) index p Although the Wigner-function-based measure [27], the recognizes this state as macroscopic, since, as mentioned index p [22] and the QFI-based measure [28] are motivated before, Dicke states themselves are macroscopic quantum by different arguments, we will see in this section that states provided N is large enough. there exist tight mathematical connections between them. The last example, the single Fock state |Ni can only be To this end, we map the√ entire measure [27] to the treated by the Fisher information and the index p. Again, spin ensemble via a 7→ −i/ MJ− (see Eq. (28)). With one finds that Neff ≈ 2N + 1. the identity J± = 1/2(Jx ± iJ−), it is straightforward to see that − Eff. size Eq. |Ψαi |ψα i |0i + |2Ni |Ni 2 1 M (14) O(N) O(1) O(1/M) n.d. I(ρ) 7→ (h[Jx, [Jx, ρ]]iρ + h[Jy, [Jy, ρ]]iρ) , (34) rF 8M Neff (19) O(N) O(1) O(1) n.d. Cδ (16) O(N) O(1) O(N) n.d. which reduced in the case of pure states to D¯ (17) O(N) O(1) O(N) n.d. √ 1 Size (3) O(N) O( N) O(N) n.d. I(ψ) 7→ (V (J ) + V (J )) . (35) Pg 4M ψ x ψ y Index p(∗) (12) O(N) O(1) O(N) O(N) Neff (18) O(N) O(1) O(N) O(N) We compare this to the QFI-based measure (18) for pure I (8) O(N) O(1) O(N) O(N) states, which is the maximal variance for local observables divided by M. The local operator leading to the maximal Table 1: Four examples of photonic states and their effective sizes variance must be located on the equator, that is, Jmax = according to the discussed measures. The second column guides the cos ϕJx + sin ϕJy for a specific value of ϕ, because the reader to the definitions of the measures. The remaining columns 2 summarize the effective sizes for the four examples discussed in contribution from Jz is at most in the order of N , which Sec. 4.1. “n.d.” stands for “not defined”. (∗) The index p is used is significantly lower than the maximal variance, which in the modified version (see text). can be in the order of MN. Thus, it is straightforward to bound Neff (φ)/4 < I(φ) ≤ Neff (φ)/2. We find the same result for the index p, given the modification of the 4.2. Measurement-based vs. coarse-grained-based measures effective size discussed in Sec. 4.1. Let us now come to a direct comparison between the In contrast to the measures for Schrödinger-cat states measurement-based measure for spin system [25], Cδ, and such as Eq. (1), the measures discussed in this section iden- the coarse-grained-based measure formulated for photonic tify Fock states (Dicke states) and (spin) squeezed states modes [30], SizePg . The similarity between the measures as macroscopic quantum states, if the mean photon num- † is clearly the motivation to call two states |φ0i and |φ1i ber, ha ai, is high enough. The effective size of Fock states macroscopically distinct if they stay distinguishable under |Ni and squeezed vacuum states is ha†ai + 1/2 [27], which special restrictions. In the former case, it is the limited is the maximal number. 9 In the case of mixed states, we still find some con- we have a simple tool at hand to compare photonic states nection between the measures. First note that I(ρ) is a with spin states. More specifically, we are able to contrast convex function, that is, for ρ = aρ1 + (1 − a)ρ2, one has measures for macroscopic quantum states for photonic and that I(ρ) = aI(ρ1) + (1 − a)I(ρ2) − a(1 − a)I(ρ1 − ρ2). spin systems. To exemplify the usefulness of the mapping, Next, it was conjectured [45] and later proven [46] that the we presented three main results. QFI is the convex roof of the variance (up to a factor of First, we showed that measures for spin systems essen- four). This implies that any convex function that reduces tially keep the relations among them even if we restrict to to the variance for pure states can be bounded from above spin states that result from the absorption of the photonic by the QFI. Thus, one has excitations. It is important to stress that operating in the limit N M requires a “redefinition” of the notion of 1 I(ρ) ≤ Neff (ρ). (36) “macroscopic quantum state”. The effective size of these 2 states can never go beyond O(N), while quantum states Furthermore, the trace norm in the definition of index q from the full Hilbert space may exhibit effective sizes scal- (see Eq. (13)) is also an upper bound on I(ρ), since for ing as O(M). any operator A, B it holds that Tr(AB) ≤ kBkkAk1; in Second, we found similarities between the coarse-grained- our case, B = ρ with operator norm kρk smaller or equal based measure for photonic modes [30] and the measurement- unity. based measure mainly for spin ensembles [25] as both mea- Even though the full equivalence between the measures sures directly or indirectly connect the notion of macro- for mixed states is an open question, they agree on the scopic distinctness to distinctness under macroscopically effective size for the following state realistic conditions. They differ in the exact modeling of this idea, but they are tightly related on a conceptual level 1 + d 1 − d − − and also lead to comparable results for the discussed ex- ρ = |ΨαihΨα| + Ψ Ψ , (37) 2 2 α α amples.

− Finally, we identified strong mathematical connections where |Ψα i ∝ |αi−|−αi and d ∈ [0, 1]. All three measures 2 2 between measures for general macroscopic quantum states assign a value in the order of d |α| to this quantum state [22, 24, 27, 28]. This is interesting as the motivation for (for I(ρ), see Ref. [27]). these measures is partially very different and neverthe- The results of this section are summarized in Fig. 1. less they give rise to a similar classification of macroscopic quantum states. Spins Photons One may ask how unambiguous this mapping is and

(1) whether, for example, more realistic models could lead to diﬀerent conclusions. It is clear that there exists indeed such mappings; for instance, if every Fock state is not mapped to a coherent but to an incoherent mixture (2) of excitations. Then, the photonic state does not induce entanglement within the spin ensemble and the measures do not assign any large eﬀective size. However, this choice

(3) for the mapping does not preserve the properties of the photonic state since it is not reversible. We further note that similar results as presented in this paper can be ob- tained by more realistic Hamiltonians. For example, one may respect some geometrical aspect like the spatial dis- Figure 1: Summary of the relations between the measures for states tribution of the spins and the direction of the incoming |φi ∈ SN as discussed in Sec. 4. We identify three different levels of macroscopic quantum states, where each level includes the previous light beam. As long as the photons potentially generate one: (1) superpositions of “classical” states like |Ψαi; (2) superposi- sufficiently large entanglement between the spins, all con- tions of “macroscopically distinct” states like |0i + |2Ni; (3) general clusions of this paper are still valid. Even more, we hope macroscopic quantum states like |Ni. The implications are either mathematically strict (⇒, ⇔) or there is a conceptual connection that the presented mapping is useful for other theoretical (!). The assumptions and restrictions that are used for the rela- studies to characterize the nonclassical properties of some tions are discussed in the text. Compare also to table 1. photonic states, since notions like entanglement or nonlo- cality are not directly defined for a single-mode states (cp., e.g., to Refs. [47, 48]). 5. Summary and discussion Through this mapping, we have discovered strong connections between the discussed measures. However, one In this paper, we considered an idealized photon-spin clearly encounters conceptual differences. For example, interaction to fully map one-mode photonic states onto the coarse-grained-based measure [30] for single modes is spin ensembles in a regime where the spin number, M, not invariant under displacement, while all other men- is much larger than the mean photon number, N. Thus tioned measures are. In particular, for spin ensembles, 10 the displacement operator translates to a local rotation of Acknowledgments.— We thank Pavel Sekatski for help- the spins. Even though this seems to be a rather small ful discussions and comments. This work was supported by detail, it is important to clarify such open questions. Ul- the Austrian Science Fund (FWF), grant number J3462; timately, it is desirable to develop a “resource theory for the National Swiss Science Foundation (SNSF), grant num- macroscopic quantumness”, similar as it was done for en- ber PP00P2-150579; and the European Research Council tanglement [49]. This theory would be able to answer ques- (MEC). tions like: Under which unitary operations is the effective invariant (like local rotations for entangled states)? It is Appendix A. Restrictions for relations (20) and (32) a worthwhile observation that, for spin systems, one can increase the effective size by local measurements. For ex- Sufficient restrictions for relations (20).— As discussed ample, a two-dimensional cluster state [50] has a small in Ref. [28], the following assumptions are proceeded on variance with respect to any (quasi-)local operator and to show the relations of Eq. (20). For M2, we maximize therefore a small effective size. However, by measuring numerator and denominator independently over all local every second qubit, one can deterministically generate a operators in order maintain finite effective sizes. For Cδ, GHZ state (up to known local unitaries) with an effective we assume that the scaling of the effective size Cδ does not size equaling M/2. depend on the specific choice of δ. A sufficient condition It is further an interesting question whether this ap- therefore is to demand that there are negligible correla- proach of comparing measures for macroscopic quantum tions in the measurement statistics between the subgroups. states can be extended to compare, for instance, massive For D¯, we restrict ourselves to states where D¯ is indepen- bosonic modes with light modes. In this case, archetypal dent whether one counts the steps from |ψ0i to |ψ1i or optomechanical setups could be used for a mapping be- vice-versa. tween light and massive objects. However, there are also Additional arguments for relations (32).— To show limitations and it is likely that even on a qualitative level the relations (32), it suffices to slightly modify the proof there could remain certain differences between proposals for (20) in Ref. [28]. First, we show that for states |φi ∈ for macroscopic quantum states. In Ref. [29], for exam- SN , the maximal variance is bounded by O(MN). To ple, the notion of macroscopic quantum states is linked to see this, notice that symmetric states are eigenstates of states whose generation is potentially used for a falsifica- 2 2 2 2 the total angular momentum, J = Jx + Jy + Jz , with tion of collapse models like [16]. However, the influence eigenvalue M(M + 2). Since hJziφ = −M + 2N, we find of the mass of the object onto measures for photonic sys- 2 that maxJ:local Vφ(J) ≤ Vφ(Jx)+Vφ(Jt)+Vφ(Jz) ≤ hJ iφ− tems is an open question and thus a direct comparison is 2 hJziφ ≈ 4M(N + 1/2). So one finds nontrivial. Altogether, this work shows that the discussed mea- max max Vφ(J) = O(NM). (A.1) sures share common features. In particular, there ap- φ∈SN J:local pears a hierarchy of three different levels of macroscopic This observation also motivates the modification of the quantum states. There are superpositions of “classical” index p for states from SN . states, superpositions of “macroscopically distinct” states Similarly, one can show that the maximal expectation and general macroscopic quantum states that show some value for any local operator of the form Jϑ = cos ϑJx + sort of macroscopic quantum effect. We also see that it √ sin ϑJy is in the order of MN. Consider the reduced is sometimes not straightforward to “transfer” the mea- density matrix for one spin, which equals ρ1 = 1/2(1+~n~σ). sures to other physical systems. It may happen that the For states |φi ∈ SN , we thus have Tr(ρ1σz) ≤ −1+2N/M. resulting effective sizes are to some extent surprising, as Hence, one finds that nz = 1 − 2N/M and, due to the for example states like |0i + |2Ni are not considered to 2 2 positivity of ρ1, that n +n ≤ 4N/M. So, we can estimate be macroscopically quantum by some measures. The re- x y √ max hJ i ≤ Mp4N/M = O( MN). Furthermore, this striction to the low-energy sector reveals conflicts of some ϑ ϑ φ shows that max |hJ i − hJ i | = O(N). measures with our intuition. Although this does not imply |φ0i,|φ1i∈SN z φ0 z φ1 The estimates on the maximal variance and the maxi- that these measures are not well-defined, it may show us mal expectation values are important to bound the effec- that measures based on some intuition are mainly applica- tive sizes of M2, N rF and N . With these findings, one ble to a regime where the intuition comes from. However, eff eff argues that the effective size of these measures are at most the usage or adaption of measures to diversity of different O(N). scenarios and experimental setups helps to understand the In addition, we have to consider the following issue. For features as well as the limitations of each approach. Even- every state family where C does not scale linearly in M, it tually, one may wish to modify the proposals to be able to δ is clear that the minimal group size n depends on M. In handle a wider range of experiments. We therefore hope min particular, for C = O(N), one has that n = O(M/N). that this work contribute to a better understanding of the δ min If we allow for general measurements on such large groups, structure of the different measures and it might be one we cannot establish a direct link to the QFI approach N , step towards a commonly accepted theory of macroscopic eff because a nonlocal measurement is translated to a nonlocal quantum physics. 11 Hamiltonian, which is not the class that is considered for [21] W. Dür, C. Simon, J. I. Cirac, Effective size of certain macro- the maximization of V(H). Therefore, in order to main- scopic quantum superpositions, Phys. Rev. Lett. 89 (21) (2002) tain this relation, we have to restrict ourselves to collective 210402. [22] A. Shimizu, T. Miyadera, Stability of quantum states of finite measurements. macroscopic systems against classical noises, perturbations from environments, and local measurements, Phys. Rev. Lett. 89 (27) (2002) 270403. References [23] G. Björk, P. G. L. Mana, A size criterion for macroscopic superposition states, J. Opt. B: Quantum Semiclass. Opt. 6 (11) [1] M. R. Andrews, C. G. Townsend, H.-J. Miesner, D. S. Durfee, (2004) 429. D. M. Kurn, W. Ketterle, Observation of interference between [24] A. Shimizu, T. Morimae, Detection of macroscopic entangle- two bose condensates, Science 275 (5300) (1997) 637–641. ment by correlation of local observables, Phys. Rev. Lett. 95 (9) [2] J. R. Friedman, V. Patel, W. Chen, S. K. Tolpygo, J. E. Lukens, (2005) 090401. Quantum superposition of distinct macroscopic states, Nature [25] J. I. Korsbakken, K. B. Whaley, J. Dubois, J. I. Cirac, 406 (6791) (2000) 43–46. Measurement-based measure of the size of macroscopic quan- [3] T. Hime, P. A. Reichardt, B. L. T. Plourde, T. L. Robertson, C.- tum superpositions, Phys. Rev. A 75 (4) (2007) 042106. E. Wu, A. V. Ustinov, J. Clarke, Solid-state qubits with current- [26] F. Marquardt, B. Abel, J. von Delft, Measuring the size of controlled coupling, Science 314 (5804) (2006) 1427–1429. a quantum superposition of many-body states, Phys. Rev. A [4] J. D. Teufel, T. Donner, D. Li, J. W. Harlow, M. S. Allman, 78 (1) (2008) 012109. K. Cicak, A. J. Sirois, J. D. Whittaker, K. W. Lehnert, R. W. [27] C.-W. Lee, H. Jeong, Quantification of macroscopic quantum Simmonds, Sideband cooling of micromechanical motion to the superpositions within phase space, Phys. Rev. Lett. 106 (22) quantum ground state, Nature 475 (7356) (2011) 359–363. (2011) 220401. [5] N. Kiesel, F. Blaser, U. Delić, D. Grass, R. Kaltenbaek, M. As- [28] F. Fröwis, W. Dür, Measures of macroscopicity for quantum pelmeyer, Cavity cooling of an optically levitated submicron spin systems, New Journal of Physics 14 (9) (2012) 093039. particle, PNAS (2013) 201309167. [29] S. Nimmrichter, K. Hornberger, Macroscopicity of mechanical [6] S. Gerlich, S. Eibenberger, M. Tomandl, S. Nimmrichter, quantum superposition states, Phys. Rev. Lett. 110 (16) (2013) K. Hornberger, P. J. Fagan, J. Tüxen, M. Mayor, M. Arndt, 160403. Quantum interference of large organic molecules, Nat Commun [30] P. Sekatski, N. Sangouard, N. Gisin, Size of quantum superposi- 2 (2011) 263. tions as measured with classical detectors, Phys. Rev. A 89 (1) [7] K. C. Lee, M. R. Sprague, B. J. Sussman, J. Nunn, N. K. Lang- (2014) 012116. ford, X.-M. Jin, T. Champion, P. Michelberger, K. F. Reim, [31] F. Fröwis, M. v. d. Nest, W. Dür, Certifiability criterion for D. England, D. Jaksch, I. A. Walmsley, Entangling macroscopic large-scale quantum systems, New J. Phys. 15 (11) (2013) diamonds at room temperature, Science 334 (6060) (2011) 1253– 113011. 1256. [32] P. Sekatski, N. Gisin, N. Sangouard, How difficult it is to prove [8] J. Hald, J. L. Sørensen, C. Schori, E. S. Polzik, Spin squeezed the quantumness of macroscropic states?, arXiv:1402.2542 atoms: A macroscopic entangled ensemble created by light, [quant-ph]. Phys. Rev. Lett. 83 (7) (1999) 1319–1322. [33] T. J. Volkoff, K. B. Whaley, Measurement- and comparison- [9] I. Usmani, C. Clausen, F. Bussières, N. Sangouard, M. Afzelius, based sizes of schrödinger cat states of light, Phys. Rev. A 89 (1) N. Gisin, Heralded quantum entanglement between two crystals, (2014) 012122. Nature Photonics 6 (4) (2012) 234–237. [34] P. Sekatski, N. Sangouard, M. Stobińska, F. Bussières, [10] M. F. Riedel, P. Böhi, Y. Li, T. W. Hänsch, A. Sinatra, P. Treut- M. Afzelius, N. Gisin, Proposal for exploring macroscopic en- lein, Atom-chip-based generation of entanglement for quantum tanglement with a single photon and coherent states, Phys. Rev. metrology, Nature 464 (7292) (2010) 1170–1173. A 86 (6) (2012) 060301. [11] N. Behbood, F. M. Ciurana, G. Colangelo, M. Napolitano, [35] H. Jeong, M. Kang, C.-W. Lee, Reply to comment on "Quan- G. Toth, R. J. Sewell, M. W. Mitchell, Generation of macro- tification of macroscopic quantum superpositions within phase scopic singlet states in a cold atomic ensemble, arXiv:1403.1964 space", arXiv:1108.0212 [quant-ph]. [quant-ph]. [36] T. Morimae, Superposition of macroscopically distinct states [12] B. Lücke, J. Peise, G. Vitagliano, J. Arlt, L. Santos, G. Tóth, means large multipartite entanglement, Phys. Rev. A 81 (1) C. Klempt, Detecting multiparticle entanglement of dicke (2010) 010101. states, arXiv:1403.4542 [cond-mat, physics:quant-ph]. [37] L. Mandelstam, I. Tamm, The uncertainty relation between en- [13] N. Bruno, A. Martin, P. Sekatski, N. Sangouard, R. T. Thew, ergy and time in non-relativistic quantum mechanics, J. Phys. N. Gisin, Displacement of entanglement back and forth between (UDSSR) 9 (4) (1945) 249. the micro and macro domains, Nat Phys 9 (9) (2013) 545–548. [38] S. L. Braunstein, C. M. Caves, Statistical distance and the ge- [14] A. I. Lvovsky, R. Ghobadi, A. Chandra, A. S. Prasad, C. Simon, ometry of quantum states, Phys. Rev. Lett. 72 (22) (1994) 3439. Observation of micro-macro entanglement of light, Nat Phys [39] F. Fröwis, Kind of entanglement that speeds up quantum evo- 9 (9) (2013) 541–544. lution, Phys. Rev. A 85 (5) (2012) 052127. [15] W. H. Zurek, Decoherence, einselection, and the quantum ori- [40] G. Tóth, Multipartite entanglement and high-precision metrol- gins of the classical, Rev. Mod. Phys. 75 (3) (2003) 715–775. ogy, Phys. Rev. A 85 (2) (2012) 022322. [16] G. C. Ghirardi, A. Rimini, T. Weber, Unified dynamics for mi- [41] P. Hyllus, W. Laskowski, R. Krischek, C. Schwemmer, W. Wiec- croscopic and macroscopic systems, Phys. Rev. D 34 (2) (1986) zorek, H. Weinfurter, L. Pezzé, A. Smerzi, Fisher information 470–491. and multiparticle entanglement, Phys. Rev. A 85 (2) (2012) [17] A. Bassi, K. Lochan, S. Satin, T. P. Singh, H. Ulbricht, Models 022321. of wave-function collapse, underlying theories, and experimental [42] F. Fröwis, W. Dür, Are cloned quantum states macroscopic?, tests, Rev. Mod. Phys. 85 (2) (2013) 471–527. Phys. Rev. Lett. 109 (17) (2012) 170401. [18] E. Schrödinger, Die gegenwärtige situation in der quanten- [43] K. Hammerer, A. S. Sørensen, E. S. Polzik, Quantum interface mechanik, Naturwissenschaften 23 (48) (1935) 807–812. between light and atomic ensembles, Rev. Mod. Phys. 82 (2) [19] A. J. Leggett, Macroscopic quantum systems and the quantum (2010) 1041–1093. theory of measurement, Progress of Theoretical Physics Supple- [44] R. Horodecki, P. Horodecki, M. Horodecki, K. Horodecki, Quan- ment 69 (1980) 80–100. tum entanglement, Rev. Mod. Phys. 81 (2) (2009) 865–942. [20] A. J. Leggett, Testing the limits of quantum mechanics: moti- [45] G. Tóth, D. Petz, Extremal properties of the variance and the vation, state of play, prospects, Journal of Physics: Condensed quantum fisher information, Phys. Rev. A 87 (3) (2013) 032324. Matter 14 (15) (2002) R415–R451. 12 [46] S. Yu, Quantum fisher information as the convex roof of variance, arXiv:1302.5311. [47] J. K. Asbóth, J. Calsamiglia, H. Ritsch, Computable measure of nonclassicality for light, Phys. Rev. Lett. 94 (17) (2005) 173602. [48] J. S. Ivan, N. Mukunda, R. Simon, Generation of NPT entanglement from nonclassical photon statistics, arXiv:quant- ph/0603255. [49] M. Skotiniotis, Private communication, 2014. [50] H. J. Briegel, R. Raussendorf, Persistent entanglement in arrays of interacting particles, Phys. Rev. Lett. 86 (5) (2001) 910–913.