Arxiv:2008.08609V1 [Gr-Qc] 19 Aug 2020 Spheres Will Entangle Only If the Gravitational ﬁeld Ex- Tational ﬁeld (Sec

Loss of coherence of matter-wave interferometer from ﬂuctuating graviton bath

Marko Toroš,1 Anupam Mazumdar,2, 3, ∗ and Sougato Bose1, † 1University College London, Gower Street, WC1E 6BT London, United Kingdom. 2University of Groningen PO Box 72, 9700 Groningen, The Netherlands. 3Van Swinderen Institute, University of Groningen, 9747 AG Groningen, The Netherlands. In this paper we consider non-relativistic matter-wave interferometer coupled with a quantum graviton bath – and discuss the loss of coherence in the matter sector due to the matter-graviton vertex. First of all, such a process does not lead to any entanglement, but nonetheless the on-shell scattering diagram can lead to loss of coherence as we will show.√ Importantly, we will show that graviton emission is the only one-vertex Feynman-diagram ∼ G which is consistent with the conservation of energy and momentum at the dominant order ∼ O(c−2). We will ﬁnd that the resulting dephasing is extremely mild and hardly places any constraints on matter-wave interferometers in the mesoscopic regime. In particular, the show that the corresponding loss of coherence in the recently proposed experiment which would test quantum aspects of graviton – via entanglement of two matter-wave interferometers – is completely negligible.

I. INTRODUCTION scattering diagram which also yields the gravitational ∼ 1/r potential. The virtual graviton is a non-classical The interface between quantum mechanics and grav- entity, it does not satisfy the classical equations of mo- ity poses significant mathematical challenges and severe tion, while the two vertices satisfy the conservation of 2 conceptual problems [1]. The first discussions about the energy-momentum tensor . In principle, we can witness nature of the gravitational field – and its relation to quan- the entanglement between the two masses in a laboratory tum mechanics – date back to Bronstein [2, 3] and Feyn- as pointed out by Ref. [7]. The entanglement witness can man [4], closely following the conceptual development of be read by measuring the spin correlations of the two quantum mechanics and quantum field theory (QFT). masses within a Stern-Gerlach (SG) setup. Since then this has lead to a multitude of field theoreti- One might wonder what is the effect of the graviton cal considerations about quantum gravity [5], and more fluctuations inside the experimental-box where the gravi- recently to an increased interest from the community to tons can interact with the matter-wave interferometer. develop the corresponding phenomenology [6]. Indeed, we can think about gravitons as an environmen- However, the experimental challenges to discern be- tal bath which will dephase/decohere the matter-wave tween classical models of the gravitational field and the system. In analogy to what happens with an electro- quantum one are far from trivial – gravitational effects magnetic bath one can expect to have graviton emission, are extremely weak for light masses, while quantum absorption, and scattering processes, all leading to a new features are completely suppressed for heavy objects. channel for the loss of coherence. However, the coupling Nonetheless, it has recently been proposed to test the of the gravitational field to matter presents subtleties, quantum nature of linearised gravity by using matter- and several different theoretical approaches have been wave systems in the mesoscopic regime [7, 8]. In partic- considered [13–24]. ular, we can consider two massive spheres, each placed In this work we discuss the loss of coherence for a in a spatial superposition, and located in proximity – non-relativistic matter-wave when coupled to the grav- such that they can interact gravitationally – but still itational field. We will start by considering a labora- far enough that all other interactions, e.g. electromag- tory frame and obtain the coupling of a matter-wave sys- netic interactions are negligible 1. Importantly, the two tem to the dynamical degrees of freedom of the gravi- arXiv:2008.08609v1 [gr-qc] 19 Aug 2020 spheres will entangle only if the gravitational field ex- tational field (Sec. II). In particular, we will investigate hibits bonafide quantum features which are absent in all the matter-wave dynamics when the gravitational field classical models of gravity. is eliminated from the description – this will result in The two quantum masses can only entangle via off- Lindblad master equation and the corresponding char- shell/virtual exchange of a graviton between them [11]. acteristic rate for the decay of coherence. We will work In the QFT parlance, this is nothing but a non-relativistic in the linearised quantum gravity framework and show that the dominant contribution for the loss of coherence is related to a graviton emission – the only QFT process at order O(c−2) and compatible with energy con- ∗ [email protected] servation (Sec. III). Finally, we will show that simple † [email protected] 1 There are numerous experimental challenges, discussed in the original reference [7], and subsequently in Refs. [9, 10]. These challenges are pertinent to any matter-wave experiment per- formed in a mesoscopic regime, and do not pose any fundamental 2 The off-shell graviton has 6 degrees of freedom, while on-shell challenges. graviton has 2 dynamical degrees of freedom [12]. 2

¨ ¨ (here h11 ≡ h11(t, 0) denotes only a number evaluated on the reference FNC curve). Using Eqs. (1) and (2) we then readily ﬁnd the interaction Lagrangian between graviton and matter degrees of freedom: m L = h¨ x2. (3) int 4 11

√ B. Matter-graviton Hamiltonian Figure 1. Feynman diagram for the vertex ∼ G defined in Eqs. (10) and (11). In words: a matter-wave systems of We now consider the gravitational field, see Refs. [20, initial energy ωi emits an on-shell graviton of frequency 2ωm 34]: resulting in a final frequency ωf = ωi − 2ωm for the matter- wave system. The graviton frequency 2ωm arises from the r 2 Z G quadratic position coupling of the matter-wave system, ∼ xˆ , ˆ ~ λ −i(ωkt−k·x) hij(t, x) = dk 2 2 gˆk,λeij(n)e + H.c., to the the gravitational field, ∼ hˆ11 (see Eq. (3)). π c ωk (4) where G is the Newton’s constant, ωk = kc, k = kkk, stochastic models of the gravitational field [25] are in- n = k/kkk, and gˆk,λ is the annihilation operator. In compatible with energy conservation and that they are Eq. (4) we also implicitly assume the summation over the P λ closely related to the spontaneous wavefunction collapse polarizations, λ, where ejk denote the basis tensors for models [26, 27] (Sec. IV). the two polarizations, λ = 1, 2. The basis tensors satisfy the completeness relation:

X λ λ II. LINEARISED QUANTUM GRAVITY eij(n)ekl(n) = PikPjl + PilPjk − PijPkl, (5) λ In this section, our working hypothesis is linearised where Pij ≡ Pij(n) = δij − ninj. From Eq. (3) and quantum gravity, for a review see [28]. We further assume λ (4) we however see that only e11(n) is relevant for the matter to be non-relativistic, i.e. slowly moving. We find matter-wave system. For later convenience we write the the dominant interaction between matter and graviton in integral: the Fermi normal coordinates (FNC) which does not have Z any remnant gauge freedom [29, 30] and is commonly 32π dn P11(n)P11(n) = . (6) used to describe laboratory experiments [31]. 15 As we will see this latter expression quantifies the average effect (on the x-axis matter-wave state) of the gravitons A. Fermi normal coordinates emitted in arbitrary directions. We can readily write also the corresponding kinetic For simplicity, we will assume that the relevant motion term for the mass-less graviton field to be: of the particle is along the x-axis and consider the FNC µ Z coordinates, x = (ct, x, y, z), of an ideal observer follow- H = dk ω g† gˆ . (7) ing a geodesic trajectory (the situation of an observer fol- grav ~ k k,λ k,λ lowing a generic time-like curve can be analysed in a sim- In addition, we assume for the purpose of illustration ilar fashion without affecting the final results). We start that matter is harmonically trapped and described by a from the general relativistic point-particle Lagrangian: simple Hamiltonian 2p µ ν L = −mc gµν x˙ x˙ , (1) ˆ†ˆ Hm = ~ωmb b, (8) where m is the mass of the system, c is the speed of light, ˆ and gµν is the metric expressed in FNC coordinates. In where ωm is the harmonic frequency, and b is the an- particular, we write the metric as gµν = ηµν + hµν , ηµν is nihilation operator. We also introduce the matter-zero- the Minkowski metric, and hµν the spacetime curvature point-fluctuations to be: perturbation near the geodesic up to order O(x2) [32]. r ~ Assuming that matter is moving slowly, the dominant δmzpf = . (9) contribution to the dynamics will be given by [33]: 2mωm 1 The interaction Hamiltonian can be derived from the in- g = −(1 + h¨ x2), (2) teraction Lagrangian in Eq. (3), and using Eq. (4) we 00 2c2 11 find: ¨ 2 where h11 = 2c R0101 is the “+” component of the Z X λ ˆ 2 gravitational waves usually discussed in the transverse- Hint = dk Gk gˆk,λX +H.c., (10) traceless (TT) coordinates, and R is the Riemann tensor λ 3 where the coupling is given by We denote the total statistical operator of the problem as ρˆ(tot) (the matter-wave system and the gravita- s (g) G 3ω3 tional field), and by ρˆ (ρˆ ) is the reduced statistical Gλ = − ~ k eλ (n), (11) k 64π2c2ω2 11 operator for the matter-wave system (the gravitational m field), obtained by tracing away the gravitational field (the matter-wave system). The von-Neumann equation and we have defined the (adimensional) amplitude can be expressed in the interaction picture as: quadrature d i † (tot) (int) (tot) Xˆ = ˆb + ˆb (12) ρˆt = − [Ht , ρˆt ], (17) dt ~ q ~ ˆ where (the position observable is xˆ = 2mω X). For later con- m Z venience we define also the (adimensional) phase quadra- (int) X Hˆ = dk Gλgˆ e−iωkt Xˆ + H.c. (18) ture t k k,λ t λ is the interaction Hamiltonian from Eq. (10) transformed ˆ† ˆ Pˆ = i(b − b) (13) into the interaction picture. The amplitude quadrature q (the adimensional position observable) in the interaction ~mωm ˆ picture is given by: (the momentum observable is pˆ = 2 P ). Impor- tantly, we note that the coupling Gλ in Eq. (11) does k Xˆ = ˆbe−iωmt + ˆb†eiωmt. (19) not depend on the mass of the matter-wave system, but t only on graviton and matter-wave frequencies, ωk and , The dynamics in Eq. (17) can be formally solved: respectively – as such, the effect on matter-wave system in the mesoscopic is precisely the same as, say, on atomic Z t (tot) (tot) i ˆ (int) (tot) systems. Of course, for a given value of the position am- ρˆt =ρ ˆ0 − ds [Hs , ρˆs ]. (20) √ ~ 0 plitude Xˆ, since xˆ ∼ X/ˆ m, the lighter system will be in a much larger spatial superposition in comparison to By then inserting Eq. (20) into Eq. (17), and tracing over the heavier one. the bath (the gravitational field), we obtain: In summary, the total Hamiltonian is now given by: d 1 Z t ρˆ = − ds tr [Hˆ (int), [Hˆ (int), ρˆ(tot)]], (21) Hˆ = Hˆ + Hˆ + Hˆ , (14) t 2 g t s s tot grav m int dt ~ 0 where Hˆ , Hˆ , and Hˆ are given in Eqs. (7),(8), and (int) (tot) grav m int where the first-order term, trg[Ht , ρˆ0 ], vanished as (10). † hgˆk,λi = hgˆk,λi = 0. On the other hand, the second- order term on the right-hand side of Eq. (21) is non-zero – as it depends on the value of the vacuum fluctuations, III. DEPHASING DUE TO GRAVITON BATH – † QFT MODEL hgˆk,λgˆk0,λ0 i, defined in Eq. (16) (see also (18)). Eq. (21) is, however, still a formal (exact) relation, containing the net effect of all Feynman diagrams with any number of In this section we estimate the loss of matter-wave co- vertices. We will now discuss the approximations that herences induced by the coupling to the quantum field will lead to the more familiar Lindblad form of the quan- model of the gravitational field (see the previous Sec. II). tum master equation, describing the effect of the dom- The interaction between graviton and a matter wave sys- inant tree-level Feynman diagram contribution only, we tem will give rise to a vertex diagram, as shown in Fig. 1. √ will not take higher order contributions in coupling G. The interaction vertex conserves the energy-momentum,√ (tot) and the coupling strength is given by G. Specifically, We will first impose the Born approximation, ρˆs ≈ (g) we will assume that the graviton field is in the ground ρˆs ⊗ρˆ , on the right hand-side of Eq. (21). Importantly, state, i.e. without any excitations (i.e. an initially empty the Born approximation precludes from the analysis any bath): entanglement between the matter-wave system, and the gravitational field as we are explicitly assuming a factor- † izable state. hgˆk,λgˆk0,λ0 i = 0, (15) (g) Furthermore, the state of the graviton bath, ρˆs = † (3) 0 (g) hgˆk,λgˆ 0 0 i = δ (k − k )δλ,λ0 (16) k ,λ ρˆ0 , is always the same as far as the system is concerned – here we are assuming that the interaction between the † † and hgˆk,λgˆk0,λ0 i = hgˆk,λgˆk0,λ0 i = 0. system and the gravitational field is weak, with negligi- We construct the quantum master equation for the ble effect on the latter, which is an excellent assumption. matter-wave system – by tracing out the gravitational Note that the entanglement and decoherence are the two field – closely following the derivation from Ref. [35]. sides of the same coin: the graviton bath can decohere 4 the matter-wave system only if the two first entangle – Eq. (11), to obtain: a process which can be expressed using Feynman diagrams containing loops, or bringing another matter-wave d Z ∞ Z G ω3 X ρˆ = − ds dk ~ k eλ (k)eλ (k) interferometer, such as in the case of Ref. [7, 11]. In dt t 64π2c2ω2 11 11 0 m λ the latter case, it was the tree level off-shell graviton ex- n −iωks ˆ 2 ˆ 2 iωks ˆ 2 ˆ 2 change lead the entanglement/decoherence between the × e Xt Xt−sρˆt − e Xt ρˆtXt−s two matter-wave systems. However, in our case, within o iωks ˆ 2 ˆ 2 −iωks ˆ 2 ˆ 2 the Born approximation, we are thus left with only tree- e ρˆtXt−sXt − e Xt−sρˆtXt . (25) level classical processes of momentum exchange (just a vertex diagram) which will lead to dephasing, but not The summation can be evaluated using the completeness decoherence. Nonetheless, dephasing can still lead to a relation from Eq. (5) and the relation in Eq. (6) – we loss of coherence, as this is well known in the community then integrate over the solid angle by first expressing the 2 of open quantum systems [36]. 2 ωk integration measure as dk = k dkdn = c3 dωkdn, where We will further impose the Markov approximation by k = kkk and n = k/kkk, we obtain: replacing ρˆs with ρˆt in the right-hand side of Eq. (21), Z ∞ Z ∞ 5 where we are assuming that the dynamics at time t de- d G~ωk ρˆt = − ds dωk . pends only on the state at time t, i.e. ρˆt. We will also re- dt 30πc5ω2 ˆ (int) ˆ (int) 0 0 m place Hs with Ht−s in the right-hand side of Eq. (10), n −iωks ˆ 2 ˆ 2 iωks ˆ 2 ˆ 2 i.e. only when |s − t| is much less than the bath corre- × e Xt Xt−sρˆt − e Xt ρˆtXt−s lation time. Here, we have assumed infinitesimal |s − t| o iωks ˆ 2 ˆ 2 −iωks ˆ 2 ˆ 2 with respect to the time-scale of the system – we will e ρˆtXt−sXt − e Xt−sρˆtXt , (26) have only a non-vanishing contribution ∼ O(G). In addition, the latter assumption also allows us to extend the We now finally insert the position amplitude observable upper integration limit to infinity, i.e. R t ds → R ∞ ds. from Eq. (19), and apply the rotating wave approxima- 0 0 ˆ ˆ† In summary, applying the approximations from the tion, i.e. we keep terms with equal number of b and b , previous two paragraphs to Eq. (21), we find the follow- and neglect the other fast rotating terms which typically ing Markovian master equation: give only a small correction, see [37]: Z ∞ 5 Z ∞ Z ∞ d G~ωk d 1 (int) (int) (g) ρˆt = − dωk ds ρˆt = − ds trg[H , [H , ρˆt ⊗ ρˆ ]]. (22) dt 30πc5ω2 dt 2 t t−s 0 m 0 ~ 0 −i(ωk−2ωm)s ˆ†2ˆ2 ˆ2 ˆ†2 e b b ρˆt − b ρˆtb We have inserted the interaction Hamiltonian from i(ωk−2ωm)s ˆ†2ˆ2 ˆ2 ˆ†2 Eq. (18) into Eq. (22) to eventually find + e ρˆtb b − b ρˆtb (27)

Z ∞ Z Z d 1 0 where we have kept only the non-zero contribution, ∼ ρˆt = − 2 ds dk dk dt ~ 0 δ(ωs − ωk), while we have omitted the other contribu- 0 tions. Specifically, the terms ∼ δ(ω + ω ) are zero as † λ λ −i(ωk−ωk0 )t s k × hgˆ gˆ i G G 0 e k,λ k0,λ0 k k the graviton cannot have negative frequency, while the n 5 −iωk0 s ˆ 2 ˆ 2 iωk0 s ˆ 2 ˆ 2 terms ∼ ω δ(ωk) vanish. We then finally integrate over × e Xt Xt−sρˆt − e Xt ρˆtXt−s k all possible out-going graviton frequencies, ωk, using the o R ∞ −i(ω −2ω )s iωks ˆ 2 ˆ 2 −iωks ˆ 2 ˆ 2 fact that ds e k m = πδ(ω − 2ω ). Eventu- e ρˆtXt−sXt − e Xt−sρˆtXt , (23) 0 k m ally, we find a simple Lindblad equation (in Schrödinger picture): where we have already used the fact that there are no † excitations of the gravitational field, hgˆk,λgˆk0,λ0 i = 0 (see d 1 ρˆ = γ ˆb2ρˆ ˆb†2 − {ˆb†2ˆb2, ρˆ } , (28) Eq. (15)). dt t grav t 2 t We now insert the non-zero value for the vacuum fluc- † 0 tuations, hgˆk,λgˆk0,λ0 i ∼ δ(k − k )δλ,λ0 (see Eq. (16)): where the emission rate is given by

3 Z ∞ Z 32G~ωm d 1 X λ 2 γ = , (29) ρˆ = − ds dk (G ) grav 5 dt t 2 k 15c ~ 0 λ n and {· , ·} denotes the anti-commutator. −iωks ˆ 2 ˆ 2 iωks ˆ 2 ˆ 2 × e Xt Xt−sρˆt − e Xt ρˆtXt−s Note that γgrav depends only on the frequency of the o matter-wave system, ω , and on the fundamental con- eiωksρˆ Xˆ 2 Xˆ 2 − e−iωksXˆ 2 ρˆ Xˆ 2 , (24) m t t−s t t−s t t stants of nature. Importantly, it is independent of the mass or any other intrinsic or extrinsic property of the λ and inserting the expression for the coupling Gk from system. 5

ωm (Hz) quanta of frequency 2ωm. On the other hand, the last 107 109 1011 1013 1 ˆ†2ˆ2 47 term, 2 {b b , ρˆt}, can be related to the probability con- 10 d servation, i.e. it enforces that dt tr[ˆρt] = 0, where the QFT model trace is over the matter degrees of freedom. It is also interesting to consider a thermal state for 27 10 the graviton ﬁeld in place of the vacuum state. In particular, a thermal state would give a nonzero value for † stochastic model hgˆk,λgˆk0,λ0 i( instead of Eq. (15)). This would allow the 107 absorption of on-shell gravitons by the matter-wave system and would give a second channel for the loss of coher-

( s ) time dephasing / decoherence ence (we would also have a graviton-absorption Feynman

10-13 diagram). 7 9 10 10 1011 1013

ωcut-off (Hz) IV. STOCHASTIC GRAVITON BATH Figure 2. Estimate of dephasing/decoherence time, td, for the QFT (blue) and stochastic (orange) model discussed in the text; specifically, we estimate the decoherence time, td, using In this section, we will consider a stochastic graviton 2 −1 tdec = (γ(∆x/δmzpf) ) , where γ is replaced by expression in bath, where we will treat the gravitational field as an ef- Eqs. (29) and (44), respectively, ∆x is the superposition size, fective noise. In particular, we will construct a stochastic and δmzpf denote the matter-wave zero-point-fluctuations (see model, where the gravitational field is not a dynamical Eq. 9). For concreteness, we have considered a matter-wave degree of freedom, but only an external noise field. Fur- −14 system of mass m ∼ 10 kg placed in a superposition size thermore, we will show the analysis from the recent work ∆x = 250µm. We note that the QFT model gives a very long in Ref. [25] – where the authors have heuristically ob- decoherence time – which depends on the exchanged quanta, tained a decoherence rate starting from the coupling in ~ωm, between the gravitational field and the matter-wave system (top blue-colored axis). On the other hand, the stochastic Eq. (3) – is to be understood as a stochastic model. Im- model gives a significantly reduced decoherence time which portantly, we will show that the latter is not compatible depends on a cut-off frequency (bottom orange-colored axis). with energy-momentum conservation. The former strictly conserves energy at each vertex, while the latter induces a constant flow of energy from the gravitational field to the matter-wave system. The green dashed line A. Stochastic model denotes a reference threshold dephasing/decoherence time of 10s. We will start the construction of the stochastic model from the linearised quantum gravity, (see Sec. II B), by Further note that the quantum master equation in replacing the mode function, gˆk,λ, with a classical C- Eq. (28) is valid for a wide range of particle masses, from valued stochastic process, g˜k,λ. In particular, we will the microscopic, e.g. neutrons and atoms, to the meso- assume that the noise has a zero-mean, E[˜gk,λ] = 0, and scopic scale and beyond, e.g. nanoparticles and micro- completely identified by the two point correlation func- sized objects. Furthermore, the rate in Eq. (29) can tions: be used to estimate the gravitational dephasing for any matter-wave system. Specifically, we define the gravita- ∗ 0 1 (3) 0 0 E[˜gk,λ(t)˜gk0,λ0 (t )] = δ (k − k )δλ,λ0 τδ˜ (t − t ), (31) tional dephasing time as 2

2 and [˜g g˜ 0 0 ] = 0, where [ · ] denotes the average δmzpf E k,λ k ,λ E td = , (30) over different noise realization, and we have introduced a γ ∆x2 grav time-scale, τ˜, which is a free parameter of the noise field where δmzpf denotes the matter zero-point-fluctuations (one can see that such a parameter needs to be introduced defined in Eq. (9). We have estimated the gravitational already from dimensional analysis). Eq. (31) can be seen dephasing for the proposal [7] in Fig. 2. as a symmetrized stochastic version of the quantum bath We can now intuitively understand the gravitational correlation functions in Eqs. (15) and (16). The time-scale τ˜ can be related to the decay of tem- dephasing in terms of Feynman vertex diagram√ (see Fig. (1). The graviton-matter interaction is ∼ G which poral correlations. For example, let us consider Gaussian can be seen from the interaction Hamiltonian in Eq. (10). correlations with spread τ˜, and suppose that they decay It is also instructive to look at the first term on the right- on a time-scale which is much faster than the character- −1 ˆ2 ˆ†2 istic time-scale of the system, e.g. τ˜ ωm , then we hand side of Eq. (28), b ρˆtb , and consider an initial pure 0 2 − (t−t ) state, ρˆ = |ψ ihψ |. The latter term can be understood 2˜τ2 0 0 0 0 can approximate√ e ∼ τδ˜ (t−t ) (any constant pre- as the effect of graviton emission: one applies ˆb2 to the factor, say 2π, can be absorbed in τ˜ by redefinition of ˆ2 state, i.e. b |ψ0i, corresponding to the removal of two the correlations). 6

Following the above prescription, and considering the where ρˆt ≡ E[|ψihψ|] denotes the average density matrix. 2 interaction Lagrangian in Eq. (3), we are then led to pos- We first note that the rate τγ˜ h – unlike the corresponding tulate the following stochastic interaction Hamiltonian: one of the QFT model in Eq. (29) – depends on two free parameters, namely on the UV cut-off frequency ωcut-off ˆ (s) ˜ ˆ 2 Ht = ~ωmhtX . (32) (which keeps γh finite), and on the parameter τ˜ (which can be interpreted as the noise correlation time). ˆ ˆ ˆ† We recall that X = b + b is the (adimensional) posi- In this scenario, the gravitational field continuously tion amplitude, which is a rescaled position observable, pumps energy into the matter system, similar to the ˆ xˆ, i.e. xˆ = δmzpfX, where δmzpf denotes the zero-point case of the position-basis wavefunction collapse mod- fluctuations defined in Eq. (9). The adimensional fluc- els [26, 27] (for a quantitative analysis see Appendix A 2). ˜ tuations , ht, are obtained from the fluctuations of the The stochastic model considered in this section is, how- ¨ gravitational field – given by the component h11 – by ever, substantially different from the latter ones. We first normalizing with respect to the frequency of the system, note that the localization operator ∼ xˆ2 is unable to dis- ˜ ¨ 2 ωm, i.e. ht ∼ h11/(8ωm) . In particular, we can define tinguish between the two states located symmetrically the adimensional fluctuations of the gravitational field in with respect to the origin, say, ∼ |−xi+|xi, and as such the following compact notation: cannot be turned into a fully satisfactory model of the Z λ quantum-to-classical transition as advocated in Ref. [38]. ˜ Gk ht ≡ dk g˜k,λ+H.c. , (33) Furthermore, to obtain the usual mathematical struc- ~ωm ture of spontaneous wavefunction collapse models, one λ would need to also postulate an anti-Hermitian coupling exploiting the coupling Gk defined in Eq. (11). Using Eqs. (11), (31), (33) we then readily find: to the gravitational field (in place of the one in Eq. (32)) as well as construct a nonlinear stochastic equation which Z 3 0 G~ωk X λ λ induces the collapse for the wavefunction [39]. Specifi- [h˜ h˜ 0 ] =τδ ˜ (t − t ) dk e (n)e (n). E t t 2 2 4 11 11 cally, the Hamiltonian in Eq. (32) is still Hermitian, i.e. 64π c ωm λ a stochastic Hamiltonian, and as such preserves the norm (34) of the wavefunction (for the derivation using stochastic The summation can be evaluated using the completeness calculus see Appendix A 1). relation from Eq. (5). We then integrate over the the solid angle by first expressing the integration measure 2 2 ωk as dk = k dkdn = c3 dωkdn, where k = kkk and n = B. Hybrid quantum-stochastic model k/kkk, and then using Eq. (6). The resulting expression reduces to Let us now discuss a recent calculation, see Ref. [25], Z 5 0 G~ωk where the authors have heuristically obtained a decoher- [h˜ h˜ 0 ] =τδ ˜ (t − t ) dω , (35) E t t k 5 4 ence rate starting from the coupling in Eq. (3). In par- 30πc ωm ticular, they have calculated which can be readily evaluated by introducing a UV fre- q quency cut-off ωcut-off (to keep the expression finite). In ¨ ¨ m hhˆ hˆ iδ2 particular, we express the final result as: 11 11 mzpf γs ≡ , (39) 4~ ˜ ˜ γh 2 0 E[htht0 ] =τ ˜( ) δ(t − t ), (36) ωm where where for later convenience we have introduced the rate Z r ˆ¨ G~ λ 2 h11 = dk 2 2 gˆk,λeij(n)ωk + H.c. (40) s 6 π c ωk G~ωcut-off γh ≡ 5 2 . (37) 180πc ωm Eq. (40) can be obtained from Eq. (4) by differentiating twice with respect to time, t, and evaluating the resulting The quantum master equation can be obtained starting expression at t = 0 and x = 0 (t = 0 can be set as the from the stochastic Hamiltonian in Eq. (32) and from the two-point correlations do not change over time, and x = noise correlation function in Eq. (36). In particular, com- √ 0 is the location of the reference geodetic curve traced by ˆ (s) ˆ 2 bining the two expression we note that Ht ∼ τγ˜ hX , the observer describing the experiment). and since the master equation is a second order effect, To compute explicitly the expectation value in ˆ (s) ˆ (s) ∼ [Ht [Ht0 , · ]] we readily see that the decoherence rate Eq. (39), we will insert the expression from Eq. (40), 2 ˆ 2 which gives: is ∼ τγ˜ h with the Lindblad operator given by X . In particular, the master equation is given by (for the derivation Z Z r r using stochastic calculus see Appendix A 1): ˆ¨ ˆ¨ 0 G~ G~ 2 2 hh11h11i = dk dk 2 2 2 2 ωkωk0 π c ωk π c ωk0 † X λ λ0 0 d 1 × hgˆk,λgˆ 0 0 i eij(n)eij (n ), (41) ρˆ =τγ ˜ 2 Xˆ 2ρˆ Xˆ 2 − {Xˆ 4, ρˆ } , (38) k ,λ dt t h t 2 t λ,λ0 7 where we have already used the fact that there are no V. DISCUSSION † excitations of the gravitational field, hgˆk,λgˆk0,λ0 i = 0 (see Eq. (15)). We then insert the non-zero value for the We will now discuss the above results in light of the † vacuum fluctuations from Eq. (16), i.e. hgˆk,λgˆk0,λ0 i ∼ experimental protocol which aims to test the graviton as 0 δ(k − k )δλ,λ0 , which gives: a quantum mediator. The theorem of LOCC (local op- eration and classical communication) would forbid any Z 3 entanglement between the two quantum systems (if they ¨ ¨ G~ωk X λ λ hhˆ hˆ i = dk e (n)e (n). (42) were not entangled to begin with) via classical communi- 11 11 π2c2 ij ij λ cation/channel. However, the two quantum systems can interact via off-shell/virtual quanta to entangle the two The summation can be evaluated using the completeness quantum systems [11]. This is the essence of the QGEM relation from Eq. (5). We will integrate over the solid (quantum gravity induced entanglement of masses) pro- angle by first expressing the integration measure as dk = tocol [7, 11]. 2 2 ωk k dkdn = c3 dωkdn, where k = kkk and n = k/kkk, and In the original proposal, there are two quantum masses then use the relation in Eq. (6). To keep the expression whose centre of mass is separated by a distance d, while finite we then introduce a frequency cut-off, ωcut-off, and their spatial superposition size is assumed to be ∆x. In evaluate the final integral over the momenta k, to find: order to obtain an entanglement phase of order one – due to exchange of virtual gravitons – the masses (assumed to 6 be the same in the simplest case) were taken to be m = ˆ¨ ˆ¨ 16G~ωcut-off 1 hh11h11i = . (43) −14 45πc5 m2 ∼ 10 kg, d − ∆x = 200 µm, and ∆x ∼ 250µm. The masses are kept in a well preserved vacuum at low Finally, inserting the correlation from Eq. (43) back into temperature to eliminate strong sources of decoherence Eq. (39), we readily find: mediated via electromagnetic interactions, and the entire setup is assumed to be in a free fall to minimize the effect s of classical noise sources [10]. G ω6 ~ cut-off In light of the discussion in the previous sections, we γs = 5 2 . (44) 180πc ωm can envisage that our matter-wave interferometers are in a graviton bath – for the sake of argument, we may It is instructive to compare the decoherence rate, γs, in consider just one interferometer without loss of general- Eq. (44) with that of the dephasing rate, γh , obtained for ity. In Ref. [25] the authors estimated that there will the stochastic model in Sec. IV. We immediately observe be a decoherence due to a stochastic graviton bath to that one of the interferometers (see Sec. IV). They estimated that the decoherence rate would be given by Eq. (44),

γs = γh, and found that the decoherence time for the above men- tioned parameters to be given by the orange curve in Fig. 2. However, here we have noted that their compu- where γh is given in Eq. (37). Furthermore, we recall that the dephasing rate for the stochastic model was given by tation violates the energy-momentum conservation for a 2 graviton-matter interaction. τγ˜ h (see Eq. 38), where the temporal correlation, τ˜, is a free parameter of the model. If we now set the correlation Instead, we have provided the computation within lin- 2 earised treatment of quantum gravity – the QFT model time to τ˜ = 1/γh, we immediately find γs =τγ ˜ h, which is the "decoherence" rate considered in Ref. [25]. (see Sec. II B). The process of energy exchange is well ex- In other words, the decoherence rate discussed in plained by the vertex diagram given by Fig. 1, for which Ref. [25] is the same as the dephasing rate obtained for there is no entanglement, but there is a loss of coherence the stochastic model from Sec. IV3. We will compare in due to the graviton-matter interaction, and the rate of more detail the heuristic model from Ref. [25] with that loss of coherence is given by Eq.(29). In particular, the of the QFT treatment of linearised gravity in the next corresponding dephasing time is given by the blue curve section. in Fig. 2. We can now compare the two results: the loss of coherence is extremely slow for the QFT model, while following the heuristic estimate from Ref. [25], i.e. the stochastic model, the loss of coherence is much faster. The dephasing rate (decoherence rate) from the QFT 3 It is now clear that the rate in Eq. (39) implicitly couples all (stochastic) model depends on a frequency cut-off for frequencies of the gravitational field to the matter-wave system, the graviton bath (on the frequency of the matter-wave regardless of the frequency of the latter, ωm. Indeed, the decoherence rate postulated in Eq. (39) is related to the gravitational system). Specifically, to estimate the decoherence time ¨ ¨ field only through the vacuum fluctuations, ∼ hhˆ hˆ i. As such in Ref. [25] the authors assumed a frequency cut-off, 11 11 12 the underlying process for the decoherence rate in Ref. [25] is not ωcut-off ∼ c/∆x ∼ 10 Hz, where ∆x ∼ 250µm denotes described by an energy conserving vertex, in stark contrast with the superposition size, thus relating the frequency cut- the QFT model discussed in Sec. II B. off, ωcut-off, with the spatial size of the system. To es- 8 timate the corresponding dephasing rate for the QFT 1. Stochastic dynamics model we suppose in first instance that the frequency of 12 the system is of the same order, i.e. ωm ∼ 10 Hz. We We can readily write a stochastic evolution for any can then see from Fig. 2 that the stochastic model from state |ψti. In particular, we can start from the following Ref. [25] overestimates the actual dephasing/decoherence Schrödinger equation time by more than ∼ 30 orders of magnitude. However, the characteristic frequencies of the system will be sig- d i ˆ (s) |ψti = − Ht |ψti, (A1) nificantly lower, and the corresponding gravitational de- dt ~ (s) phasing time, td, will be even larger. For example, at where Hˆ is the stochastic Hamiltonian defined in 73 t ω ∼ 1Hz one finds td 10 s, and thus the difference be- Eq. (32). Eq. (A1) is however only a formal expres- tween the predictions? of the QFT and stochastic model sion – due to the white-noise – and is to be interpreted might be as large as ∼ 80 orders of magnitude. with the Stratonovich definition for the stochastic inte- The main reason for such a large discrepancy is that gral, which naturally emerges as a white-noise limit of a the stochastic model violates the energy-momentum con- smooth noise [40, 41]. servation – in stark contrast to the QFT model which We would like to derive the corresponding master equa- strictly preserves it. The rate of the former depends on a tion for the average state, ρˆt ≡ E[|ψtihψt|]. However, this somewhat arbitrary frequency cut-off – with all the fre- proves to be difficult in the Stratonovich formalism. In quencies below the cut-off continuously heating the sys- ˆ (s) particular, the state |ψti and Ht in (A1) are not in- tem – while the latter depends on the frequency of the dependent and thus we are faced with the problem of exchange quanta with the matter-wave system. In this ˆ calculating the expectation values, e.g. E[Hint|ψtihψt|]. regard, we believe that the main channels of decoher- We can, however, transform the Stratonovich Eq. (A1) ence rate remains that of the electromagnetic interactions to the corresponding form: and their decoherence rates have been discussed for the ˆ 2 ˜ 2 ˆ 4 QGEM experiment in Refs. [7, 9, 10]. There might be d|ψti = −iωmX dht − τγ˜ hX dt |ψti, (A2) unknown 5th force beyond the Standard Model (BSM) where we have formally introduced the stochastic incre- of physics which remains on card to be modelled for the ˜ ˜ ˜ decoherence rate. We will constrain such BSM physics in ment dht = htdt. Importantly, the increment, dht, and a separate paper. the state, |ψti, are now independent, which greatly sim- plifies the evaluation of the expectation values. In the To summarise, our estimation shows that a quantum following, we will also use: graviton bath in a vacuum (inside the capsule) cannot ˜ ˜∗ γh 2 place any significant constraint in performing the exper- E[dhtdht ] =τ ˜( ) dt, (A3) iment – the main sources of dephasing and decoherence ωm remains the electromagnetic channel. which can be obtained from Eq. (31), for example, following the formal prescription δ(dt)dt ∼ 1 from Ref. [42]. The price to pay is that the Itô calculus does not follow the usual rules of ordinary calculus to which we are ac- VI. ACKNOWLEDGEMENTS customed. Specifically, in the following we will use the Itô product rule [41]: We would like to thank Jiro Soda for excit- d (ab) = (da)b + a(db) + (da)(db), (A4) ing discussion regarding the results of their pa- where a and b denote |ψ i or hψ |. The final term in per. MT and SB would like to acknowledge EPSRC t t the right-hand side of Eq. (A4), (da)(db), arises as the grant No.EP/N031105/1, and SB the EPSRC grant noise increment, dh˜ , is of order ∼ O(dt1/2) as can be EP/S000267/1. AM’s research is funded by the Nether- t seen from Eq. (A3)), and hence (da)(db) ∼ O(dt) is lands Organisation for Science and Research (NWO) non-negligible. The second term in the right-hand side grant number 680-91-119. of Eq. (A2) naturally emerges when moving from the Stratonovich to the Itô definition of the stochastic integral and can be obtained from the second order noise term ω Xˆ 2dh˜ ω Xˆ 2dh˜∗and applying Eq. A3. Alterna- Appendix A: Unraveling of stochastic model m t m t tively, the latter term can be also viewed as the term that enforces the usual probabilistic interpretation, i.e In this appendix we discuss the mathematical details E[d(hψt|ψti)] = 0, as can be explicitly checked using of the stochastic model introduced in the main text and Eqs. (A2), (A3), and (A4) by setting a = hψt| and its relation to the master equation for the average sta- b = |ψti. tistical operator in Eq. (38) (Sec. A 1). We then discuss One can also readily calculate the average density ma- the energy divergence of the stochastic model and com- trix, ρˆt ≡ E[|ψihψ|]. We use again the Itô product rule pare it to the same effect which arises in spontaneous in Eq. (A4), with a = |ψi and b = hψ|, and evaluate the wavefunction collapse models (Sec. A 2). average E[ · ] using Eq. (36), to readily find Eq. (38). 9

2. Energy gain where h · it = tr[ · ρt] . Speciﬁcally, using Eq. (38), and the cyclic property of the trace, we readily ﬁnd:

We estimate the rate of the energy gain of the matter- 2 d ~ωmτγ˜ h 2 2 1 4 2 wave system – induced by the coupling to the gravita- hHˆmit = hXˆ PˆXˆ − {Xˆ , Pˆ }it, (A8) tional field – according to the stochastic model. We first dt 4 2 rewrite the matter Hamiltonian in Eq. (8) as Using the commutation relation in Eq. (A6), we then eventually find: ω Hˆ = ~ m (Xˆ 2 + Pˆ2), (A5) m 4 d hHˆ i = 2˜τγ2 ω hXˆ 2i , (A9) dt m t h~ m t where Xˆ = (ˆb† +ˆb) and Pˆ = i(ˆb† −ˆb) are the adimensional position and momentum operators defined in Eqs. (12) where we note the positive sign of the expression in the and (13), which satisfy the following commutation rela- right-hand side. tion The energy signature in Eq. (A9) could be detected by non-interferometric experiments [43] (for interferometeric [X,ˆ Pˆ] = 2i, (A6) tests see [44–46]). The constant energy gain in stochastic models can be, however, avoided by considering energy- basis couplings, see [47–52], or it can be mitigated by de- as can be explicitly checked starting from the canonical vising dissipative variants [53, 54] – this would amount commutation relation [ˆx, pˆ] = i . ~ to postulating a different Hamiltonian in Eq. (32). For We then calculate spontaneous wave function collapse models motivated by d d considerations about gravity, see [39, 55, 56], and for al- hHˆ i = tr Hˆ ρˆ , (A7) ternative models of gravity, see [57]. dt m t m dt t

[1] Roger Penrose. On gravity’s role in quantum state reduc- A, 101(5):052110, 2020. tion. General relativity and gravitation, 28(5):581–600, [12] Tirthabir Biswas, Tomi Koivisto, and Anupam Mazum- 1996. dar. Nonlocal theories of gravity: the flat space propaga- [2] Matvei P Bronstein. Quantentheorie schwacher gravita- tor. In Barcelona Postgrad Encounters on Fundamental tionsfelder. Phys. Z. Sowjetunion, 9(2–3):140–157, 1936. Physics, pages 13–24, 2013. [3] G. Gorelik. Matvei bronstein and quantum gravity: 70th [13] Esteban Calzetta and BL Hu. Noise and fluctuations anniversary of the unsolved problem. Physics-Uspekhi, in semiclassical gravity. Physical Review D, 49(12):6636, 48(10):1039, 2005. 1994. [4] R. Feynman. Chapel Hill Conference Proceedings. 1957. [14] C Anastopoulos. Quantum theory of nonrelativistic [5] Daniele Oriti. Approaches to quantum gravity: Toward a particles interacting with gravity. Physical Review D, new understanding of space, time and matter. Cambridge 54(2):1600, 1996. University Press, 2009. [15] C Anastopoulos and BL Hu. A master equation for grav- [6] Sabine Hossenfelder. Experimental search for quantum itational decoherence: probing the textures of spacetime. gravity, 2010. Classical and Quantum Gravity, 30(16):165007, 2013. [7] Sougato Bose, Anupam Mazumdar, Gavin W Morley, [16] C Jess Riedel. Evidence for gravitons from decoher- Hendrik Ulbricht, Marko Toroš, Mauro Paternostro, An- ence by bremsstrahlung. arXiv preprint arXiv:1310.6347, drew A Geraci, Peter F Barker, MS Kim, and Gerard 2013. Milburn. Spin entanglement witness for quantum grav- [17] MP Blencowe. Effective field theory approach to gravi- ity. Physical review letters, 119(24):240401, 2017. tationally induced decoherence. Physical review letters, [8] Chiara Marletto and Vlatko Vedral. Gravitationally in- 111(2):021302, 2013. duced entanglement between two massive particles is suf- [18] Fumika Suzuki and Friedemann Queisser. Environmental ficient evidence of quantum effects in gravity. Physical gravitational decoherence and a tensor noise model. In review letters, 119(24):240402, 2017. Journal of Physics: Conference Series, volume 626, page [9] Thomas W. van de Kamp, Ryan J. Marshman, Sougato 012039. IOP Publishing, 2015. Bose, and Anupam Mazumdar. Quantum Gravity Wit- [19] VA De Lorenci and LH Ford. Decoherence induced ness via Entanglement of Masses: Casimir Screening. 6 by long wavelength gravitons. Physical Review D, 2020. 91(4):044038, 2015. [10] M. Toroš, T.W. van de Kamp, R.J. Marshman, M.S. [20] Teodora Oniga and Charles H-T Wang. Quantum gravi- Kim, A. Mazumdar, and S. Bose. Relative Accelera- tational decoherence of light and matter. Physical Review tion Noise Mitigation for Entangling Masses via Quan- D, 93(4):044027, 2016. tum Gravity. 7 2020. [21] Teodora Oniga and Charles H-T Wang. Quantum coher- [11] Ryan J. Marshman, Anupam Mazumdar, and Sougato ence, radiance, and resistance of gravitational systems. Bose. Locality & Entanglement in Table-Top Testing of Physical Review D, 96(8):084014, 2017. the Quantum Nature of Linearized Gravity. Phys. Rev. [22] Diego A Quiñones, Teodora Oniga, Benjamin TH Var- 10

coe, and Charles H-T Wang. Quantum principle of sens- ferential equations, volume 82. American Mathematical ing gravitational waves: From the zero-point fluctuations Soc., 2012. to the cosmological stochastic background of spacetime. [42] Howard M Wiseman and Gerard J Milburn. Quantum Physical Review D, 96(4):044018, 2017. measurement and control. Cambridge university press, [23] Vlatko Vedral. Decoherence of massive superpositions 2009. induced by coupling to a quantized gravitational field. [43] A Vinante, R Mezzena, P Falferi, Matteo Carlesso, and arXiv preprint arXiv:2005.14596, 2020. Angelo Bassi. Improved noninterferometric test of col- [24] Qidong Xu and MP Blencowe. Toy models for grav- lapse models using ultracold cantilevers. Physical review itational and scalar qed decoherence. arXiv preprint letters, 119(11):110401, 2017. arXiv:2005.02554, 2020. [44] Stefan Nimmrichter, Klaus Hornberger, Philipp [25] Sugumi Kanno, Jiro Soda, and Junsei Tokuda. Noise Haslinger, and Markus Arndt. Testing spontaneous and decoherence induced by gravitons. arXiv preprint localization theories with matter-wave interferometry. arXiv:2007.09838, 2020. Physical Review A, 83(4):043621, 2011. [26] Angelo Bassi and GianCarlo Ghirardi. Dynamical reduc- [45] Marko Toroš and Angelo Bassi. Bounds on quantum col- tion models. Physics Reports, 379(5-6):257–426, 2003. lapse models from matter-wave interferometry: calcula- [27] Angelo Bassi, Kinjalk Lochan, Seema Satin, Tejinder P tional details. Journal of Physics A: Mathematical and Singh, and Hendrik Ulbricht. Models of wave-function Theoretical, 51(11):115302, 2018. collapse, underlying theories, and experimental tests. Re- [46] Yaakov Y Fein, Philipp Geyer, Patrick Zwick, Filip Ki- views of Modern Physics, 85(2):471, 2013. ałka, Sebastian Pedalino, Marcel Mayor, Stefan Gerlich, [28] C.P. Burgess. Quantum gravity in everyday life: General and Markus Arndt. Quantum superposition of molecules relativity as an effective field theory. Living Rev. Rel., beyond 25 kda. Nature Physics, 15(12):1242–1245, 2019. 7:5–56, 2004. [47] Nicolas Gisin. Stochastic quantum dynamics and relativ- [29] Eric Poisson, Adam Pound, and Ian Vega. The motion ity. Helvetica Physica Acta, 62(4):363–371, 1989. of point particles in curved spacetime. Living Reviews in [48] Ian Colin Percival. Primary state diffusion. Proceedings Relativity, 14(1):7, 2011. of the Royal Society of London. Series A: Mathematical [30] Charles W Misner, Kip S Thorne, John Archibald and Physical Sciences, 447(1929):189–209, 1994. Wheeler, et al. Gravitation. Macmillan, 1973. [49] LP Hughston. Geometry of stochastic state vector re- [31] Clifford M Will. The confrontation between general rela- duction. Proceedings of the Royal Society of London. Se- tivity and experiment. Living reviews in relativity, 9(1):3, ries A: Mathematical, Physical and Engineering Sciences, 2006. 452(1947):953–979, 1996. [32] Matt Visser. Post-newtonian particle physics in curved [50] Dorje C Brody and Lane P Hughston. Efficient simula- spacetime. arXiv preprint arXiv:1802.00651, 2018. tion of quantum state reduction. Journal of Mathematical [33] Malik Rakhmanov. Fermi-normal, optical, and wave- Physics, 43(11):5254–5261, 2002. synchronous coordinates for spacetime with a plane [51] Stephen Louis Adler, DC Brody, TA Brun, and LP Hugh- gravitational wave. Classical and Quantum Gravity, ston. Martingale models for quantum state reduc- 31(8):085006, 2014. tion. Journal of Physics A: Mathematical and General, [34] Steven Weinberg. Gravitation and Cosmology: Princi- 34(42):8795, 2001. ples and Applications of the General Theory of Relativity. [52] Heinz-Peter Breuer, Ertan Göklü, and Claus Läm- John Wiley and Sons, New York, 1972. merzahl. Metric fluctuations and decoherence. Classical [35] Heinz-Peter Breuer, Francesco Petruccione, et al. The and Quantum Gravity, 26(10):105012, 2009. theory of open quantum systems. Oxford University Press [53] Andrea Smirne and Angelo Bassi. Dissipative continuous on Demand, 2002. spontaneous localization (csl) model. Scientific reports, [36] Maximilian A Schlosshauer. Decoherence: and the 5(1):1–9, 2015. quantum-to-classical transition. Springer Science & Busi- [54] Marko Toroš, Giulio Gasbarri, and Angelo Bassi. Col- ness Media, 2007. ored and dissipative continuous spontaneous localiza- [37] Girish S Agarwal. Quantum optics. Cambridge Univer- tion model and bounds from matter-wave interferometry. sity Press, 2012. Physics Letters A, 381(47):3921–3927, 2017. [38] John S Bell and John Stewart Bell. Speakable and un- [55] Lajos Diósi. Models for universal reduction of macro- speakable in quantum mechanics: Collected papers on scopic quantum fluctuations. Physical Review A, quantum philosophy. Cambridge university press, 2004. 40(3):1165, 1989. [39] Giulio Gasbarri, Marko Toroš, Sandro Donadi, and An- [56] Antoine Tilloy and Lajos Diósi. Principle of least de- gelo Bassi. Gravity induced wave function collapse. Phys- coherence for newtonian semiclassical gravity. Physical ical Review D, 96(10):104013, 2017. Review D, 96(10):104045, 2017. [40] Ludwig Arnold. Stochastic differential equations. New [57] Angelo Bassi, André Großardt, and Hendrik Ulbricht. York, 1974. Gravitational decoherence. Classical and Quantum Grav- [41] Lawrence C Evans. An introduction to stochastic dif- ity, 34(19):193002, 2017.