PHYSICAL REVIEW D 96, 105004 (2017) Quantum estimation of parameters of classical spacetimes
T. G. Downes,1,* J. R. van Meter,2,3 E. Knill,3 G. J. Milburn,1 and C. M. Caves1,4 1Centre for Engineered Quantum Systems, School of Mathematics and Physics, The University of Queensland, St. Lucia, Queensland 4072, Australia 2Department of Mathematics, University of Colorado, Boulder, Colorado 80309, USA 3National Institute of Standards and Technology, Boulder, Colorado 80305, USA 4Center for Quantum Information and Control, University of New Mexico, MSC07–4220, Albuquerque, New Mexico 87131-0001, USA (Received 17 November 2016; published 6 November 2017) We describe a quantum limit to the measurement of classical spacetimes. Specifically, we formulate a quantum Cram´er-Rao lower bound for estimating the single parameter in any one-parameter family of spacetime metrics. We employ the locally covariant formulation of quantum field theory in curved spacetime, which allows for a manifestly background-independent derivation. The result is an uncertainty relation that applies to all globally hyperbolic spacetimes. Among other examples, we apply our method to the detection of gravitational waves with the electromagnetic field as a probe, as in laser-interferometric gravitational-wave detectors. Other applications are discussed, from terrestrial gravimetry to cosmology.
DOI: 10.1103/PhysRevD.96.105004
I. INTRODUCTION connection between local changes in the metric and relative changes in the states of quantum fields that live on the The geometry of spacetime can be inferred from physical spacetime. These changes, used to sense the spacetime measurements made with clocks and rulers or, more gen- parameters, can be characterized in terms of an evolution of erally, with quantum fields, sources, and detectors. We the state driven by a stress-energy integral with respect to assume that the ultimate precision achievable is determined ˆ by quantum mechanics. In this paper we obtain parameter- the change in the metric, which gives the operator P. based quantum uncertainty relations that bound the preci- For the universal connection between changes in states sion with which we can determine properties of spacetime in of quantum fields and the stress-energy integrals, we rely terms of stress-energy variances. Such uncertainty relations on the locally covariant formalism for quantum fields on might become increasingly relevant to empirical observation curved spacetime backgrounds [5]. For this purpose, we treat gravity classically as in general relativity, determined as, for example, laser-interferometric gravitational-wave − detectors are expected to approach quantum-limited sensi- by a metric with signature ð ; þ; þ; þÞ on a spacetime manifold. This is treated as a fixed background on which tivity across a wide bandwidth in the near future. — An informative, high-level way to quantify the precision the quantum fields used for measurements we call these “ ”— of a parameter measurement is by the inverse variance probe fields propagate. In particular, we do not con- sider back action from the quantum fields on the metric. In hðδθ~Þ2i of an estimator θ~. The best precision with which we any case, we expect that such back action would transfer can measure a parameter is determined by the quantum uncertainty in the quantum field being measured to the Fisher information [1,2]. For pure states, the Fisher infor- metric and reduce the achievable precision, for otherwise, mation reduces to a multiple of the variance hðΔPˆÞ2i of an ˆ by an argument given in Refs. [6,7], the uncertainty evolution operator P that describes how the quantum state principle for matter would be violated (see also Ref. [8] changes with changes in the parameter. This determines a for a study of the problem of back action when measuring parameter-based uncertainty relation [3,4], the structure of spacetime). We allow for the presence of classical fields that can ℏ2 hðδθ~Þ2ihðΔPˆÞ2i ≥ ; ð1:1Þ propagate on the spacetime background. We only consider 4 those fields that play a direct role as sources for the quantum fields used for measurement. These sources are whose form is reminiscent of the Heisenberg uncertainty determined by classical devices needed to implement the relations. measurement. For the types of measurement considered Such parameter-based uncertainty relations can be here, the parametrized changes of the metric are indepen- applied to parameters associated with local changes of dent of these equipment-related classical fields. In particu- the spacetime metric. They are derived from a universal lar, these classical fields contain no information about the parameter of interest, and for this reason we do not model *[email protected] them explicitly.
2470-0010=2017=96(10)=105004(23) 105004-1 © 2017 American Physical Society DOWNES, VAN METER, KNILL, MILBURN, and CAVES PHYSICAL REVIEW D 96, 105004 (2017) Our work in this paper relies on the parameter-based This approach was used by Braunstein, Caves, and uncertainty relation (1.1). To put such uncertainty relations Milburn [4] to develop optimal quantum estimation for in context, we consider now how the familiar Heisenberg spacetime displacements in flat Minkowski spacetime. In uncertainty relations generalize to parameter-based uncer- spacetime, not only can one move a fixed proper distance or tainty relations. The Heisenberg uncertainty relation for time, one can also boost and rotate. Quantum parameter position and momentum states that the product of the estimation was thus also developed for the parameters uncertainties in position and momentum, for any quantum corresponding to these actions [4]. The results were state, must be greater than a constant. In terms of the developed with the quantized electromagnetic field as variances of position and momentum, the Heisenberg the probe. These results show that estimates of a spacetime uncertainty relation is written as translation can be made increasingly accurate as the uncertainty in the operator that generates the translation, ℏ2 boost, or rotation is made very large. Δ ˆ 2 Δ ˆ 2 ≥ hð xÞ ihð pÞ i 4 ; ð1:2Þ In this paper we are interested in limits on the precision of estimates of parameters of the classical gravitational field. In general relativity the gravitational field is a where ℏ is the reduced Planck’s constant. The Heisenberg manifestation of the geometry of spacetime, which is uncertainty relation is derived in standard quantum described by a metric. The metric determines the length mechanics, where position and momentum are both rep- of the invariant (proper) interval between two spacetime resented as Hermitian operators. events according to [9] A similar relation, albeit with a different interpretation, exists between time and energy: 2 μ ν ds ¼ gμνðxÞdx dx ; ð1:4Þ ℏ2 δ 2 Δ ˆ 2 ≥ μ ν 0 hð tÞ ihð HÞ i 4 : ð1:3Þ where gμνðxÞ is the metric tensor, with indices , ¼ ,1, 2, 3 for time and the three spatial coordinates. The dxμ are Unlike position, time in standard quantum mechanics is a infinitesimal coordinate differences. We assume the classical parameter. The time-energy uncertainty relation is Einstein summation convention, where repeated upper an example of a quantum limit on parameter estimation. and lower indices are summed over. One tries to estimate a classical parameter, in this case time, Before we describe the relevant quantum parameter by making measurements on a quantum system, a “clock,” estimation, we ask the following: can any insight be found whose evolution depends on time. In the time-energy by applying the Heisenberg uncertainty relation, in its 2 parameter-based form, directly to a proper distance? It was uncertainty relation (1.3), hðδtÞ i is the classical variance along these lines that Unruh [10] derived an uncertainty of the estimate of t; this classical variance arises ultimately relation for a component of the metric tensor. Once from quantum uncertainties in clock variables conjugate to coordinates are chosen, there should only be quantum the Hamiltonian H. uncertainty in the proper time and the proper distance. As The most common way to make quantum mechanics these are in turn related to the metric via Eq. (1.4), any compatible with classical relativity is to demote position to uncertainty in the proper distance is equivalent to uncer- a parameter, just like time in the previous example. Physical tainty in the metric. By applying the Heisenberg uncer- systems can then be thought of as living on, and intera- tainty relation to a proper distance, Unruh found a simple, cting with, the classical spacetime manifold. In this yet insightful uncertainty relation for a single component of relativistic context, the position-momentum uncertainty the metric. The Unruh uncertainty relation for the g11 relation naturally becomes a quantum limit on parameter component of the metric (assuming particular Cartesian- estimation [3]. like coordinates), in terms of variances, is This parameter-based approach is the natural, opera- tional way to think of uncertainty relations. Nothing in the ℏ2 traditional Heisenberg uncertainty relation (1.2) refers 2 ˆ 11 2 hðδg11Þ ihðΔT Þ i ≥ ; ð1:5Þ directly to a measurement of position or momentum. In V2 the parameter-based approach, one considers measure- ments of any sort whose results are used to estimate where here, and henceforth, we use units such that changes in a parameter; quantum mechanics then says, G ¼ c ¼ 1. The conjugate variable to g11 is the corre- via the quantum Fisher information, that the uncertainty of sponding component of the quantized stress-energy tensor, 11 the estimate is limited by the uncertainty in the operator that in this case the pressure Tˆ in the x1 direction. The key generates changes in the parameter, in a way that looks like feature of this uncertainty relation is the inverse propor- (but by being operational is more powerful than) a tradi- tionality to V2, the square of the four-volume of the tional Heisenberg uncertainty relation. measurement.
105004-2 QUANTUM ESTIMATION OF PARAMETERS OF CLASSICAL … PHYSICAL REVIEW D 96, 105004 (2017) We provide a general framework for deriving such an optimistic in the sense that they suggest a higher-than- uncertainty relation, by formulating it as a problem in achievable precision, we generally wish to make the quantum estimation theory. The metric gμνðxÞ is defined relationship between the measurement region, stress- for each point x on the manifold. If the quantum probe energy variance, and measurement precision tighter. For (measurement device) occupies some four-volume V, then this we show that the metric change can be localized, the probe’s state depends on the metric at every point in provided that the sensitivity of our measurement to the that region. If we consider the metric to be an arbitrary parameter of interest is not affected. Because the locally function on the manifold, then we need to estimate an covariant framework requires compactly supported regions, infinite number of parameters to define it completely. the localization is necessary when the parameter is a global Instead, we consider regions of spacetime that can be property of spacetime. described by metrics characterized by a single parameter θ. Another important issue—perhaps the most important For example, the Schwarzschild metric, which describes issue for the interpretation and relevance of our results—is the spacetime around a static nonrotating black hole, is that computations of the relevant stress-energy variances defined by the single parameter M, the mass of the black can be difficult. In most situations, however, the probe hole. The task is to estimate this mass parameter by devices introduce fields that have large mean values making measurements on physical systems living on the compared to a zero-mean reference state, which is typically spacetime manifold. the vacuum state. In these cases, the calculation can be There has recently been some promising work in this simplified by recognizing that mean-field-independent direction [11–13], focusing on quantum probes consisting contributions become negligible. of scalar fields in Gaussian states. Here we present a Once we have determined the general parameter-based general formalism for relativistic quantum metrology, using uncertainty relations connecting measurement precision to arbitrary fields and states. In so doing, we address several stress-energy variances, we apply them to several situations related issues, which have, we believe, not previously of interest. The first involves estimating constant metric received enough attention in this context. Ensuring that components and recovers the Unruh uncertainty relations. quantum observables in different spacetimes measure “the By considering specific metric components, we also obtain same” physical parameter is nontrivial. If the spacetimes the parametrized form of the Heisenberg uncertainty differ by a global perturbation, the positions of measure- relations. Next we study in detail the problem of interfero- ment devices therein might correspondingly differ, further metric gravitational-wave detection with light. This complicating the issue; more generally, spacetime points in requires the full power of our approach. Here we take two such spacetimes cannot unambiguously be identified advantage of localization by means of a “bump” function with each other. Care must also be taken to ensure supported in the region of the measurement devices and use coordinate independence. the large-mean-field property to enable the explicit calcu- Fortunately, a framework exists for comparing quantum lation of a bound on precision that depends on the observables in perturbed, classical spacetimes in a coor- amplitude of the light fields used. This recovers the dinate-independent manner [5]. This locally covariant well-known shot-noise limit, but goes beyond this limit framework, which was developed in the context of alge- in two ways. First, for the case of a wideband mean field on braic quantum field theory, serves as our starting point. As top of vacuum fluctuations, we obtain a general shot-noise the current work is geared more towards physical experi- bound that does not make an assumption of narrow ments than most literature invoking algebraic quantum field bandwidth detection of the probe field; in particular, we theory, it is worth a comment. The aim of algebraic find a wideband shot-noise limit in terms of a frequency- quantum field theory is to put quantum field theory on weighted integration over mean photon numbers. Second, rigorous mathematical footing, while the aim of what might we find that wideband squeezing gives the optimal, be termed pragmatic quantum field theory is to make sub-shot-noise sensitivity under the assumptions we are experimental predictions [14,15]. Strides toward connect- making. We briefly discuss other examples, including ing the two have been made recently [5,16–19], and this cosmological parameters and gravimetry. progress makes the current work possible. Our work is organized as follows. In Sec. II we review The locally covariant framework directly connects the quantum estimation theory including, in particular, the stress-energy to the change in state associated with a quantum Cram´er-Rao bound, which is the expression of compactly supported change in the metric. The connection parameter-based uncertainty principles. In Sec. III we is via the concept of relative Cauchy evolution developed in review the locally covariant approach to quantum field Ref. [5] and leads directly to our bounds on measurement theory in perturbed, classical spacetime developed by precision. An issue is that the bounds obtained are with Brunetti, Fredenhagen, and Verch [5]. In Sec. IV we show respect to the best observable supported in a region that can that the application of the Cram´er-Rao bound discussed in be much bigger than the region containing the probes. Sec. II to such a perturbed spacetime results in a coordinate- While this means that the bounds are guaranteed to be independent uncertainty relation between a local spacetime
105004-3 DOWNES, VAN METER, KNILL, MILBURN, and CAVES PHYSICAL REVIEW D 96, 105004 (2017) property and a quantum operator that depends on the probe The variance of an unbiased estimate of the parameter θ, field. In Sec. V we generalize this uncertainty relation to based on the distribution of measurement outcomes to be global spacetime properties. In Sec. VI we consider the observed, is bounded by the classical Cram´er-Rao lower application of the formalism to the estimation of metric bound [4], components, proper time, and proper distance in a certain class of perturbed spacetimes. In Sec. VII we derive a 2 1 hðδθ~Þ i ≥ ; ð2:3Þ quantum uncertainty bound on the detection of gravita- FðθÞ tional waves using the electromagnetic field as a probe, as in laser-interferometric gravitational-wave detectors. Then, where FðθÞ is the classical Fisher information for the in Sec. VIII, we consider additional applications, and measurement, given by finally we conclude in Sec. IX. Z 1 ∂pðωjθÞ 2 FðθÞ¼ dω : ð2:4Þ II. QUANTUM ESTIMATION THEORY pðωjθÞ ∂θ
In this paper we consider estimating an individual In this paper we consider the special case where ρˆ is parameter of a spacetime metric, so we only need the differentiable at θ0, with differential generated by the self- theory of single-parameter quantum estimation. A general adjoint operator hˆ, as expressed by scheme for quantum parameter estimation is depicted in Fig. 1. A quantum state, represented by a density operator dρˆ i ˆ ˆ ¼ − ½h; ρˆ : ð2:5Þ ρˆ0, undergoes a unitary transformation UðθÞ that depends dθ ℏ on the parameter θ of interest, producing a one-parameter ˆ ˆ † family of states, ρˆðθÞ¼UðθÞρˆ0U ðθÞ. Measurements are Informally, we write made on the system, with results ω, which are fed into an ρˆ θ θ −idθh=ˆ ℏρˆ θ idθh=ˆ ℏ estimator θ~ðωÞ of the parameter. ð 0 þ d Þ¼e ð 0Þe : ð2:6Þ We consider generalized measurements, described by positive-operator-valued measures (POVMs). For simplicity It can then be shown that for any POVM, the classical ˆ 2 2 we consider such POVMs given by a positive-operator- Fisher information satisfies FðθÞ ≤ 4hðΔhÞ =ℏ [1–3], ˆ 2 valued density EˆðωÞ that satisfies the completeness property where hðΔhÞ i is the quantum variance of the generator ˆ Z h; moreover, there is a POVM that saturates this bound when ρˆðθ0Þ is pure [3]. Applying this bound to Eq. (2.3) dωEˆ ω 1ˆ; 2:1 ð Þ¼ ð Þ yields the quantum Cram´er-Rao lower bound [4],
1ˆ ℏ2 where is the identity operator. The outcomes of a particular hðδθ~Þ2ihðΔhˆÞ2i ≥ : ð2:7Þ measurement follow a probability distribution pðωjθÞ con- 4 ditional on the parameter θ. The probability distribution for the outcomes ω can be calculated as One can now construct examples by identifying parameters and their corresponding generators. For example, since the Hamiltonian is the generator of time translations, this gives pðωjθÞdω ¼ TrðEˆðωÞρˆðθÞdωÞ: ð2:2Þ the time-energy uncertainty relation (1.3) presented in the θ Introduction. Since the momentum operator generates The problem of estimating the parameter is essentially displacements, using it as the generator gives the θ~ θ that of choosing a value to make a good estimate of by parameter-based version of the Heisenberg uncertainty ω considering the observed in relation to the known relation. Another important example is provided by the ω θ ω probability distributions pð j Þd . A common example number operator and phase, which is the basis of is the maximum likelihood estimator, which is the choice of Heisenberg-limited phase estimation [20–22]. θ~ that retrospectively maximizes the probability of the observed measurement outcomes. III. RELATIVE CAUCHY EVOLUTION We employ the locally covariant formulation of quantum field theory in curved spacetime as developed by Brunetti, Fredenhagen, and Verch [5]. The approach has been used to develop a notion of “identical physics” on different space- times [23]. A key result is a method for calculating how quantum observables respond to local changes in the FIG. 1. Scheme for quantum parameter estimation. background spacetime. Say we believe some particular
105004-4 QUANTUM ESTIMATION OF PARAMETERS OF CLASSICAL … PHYSICAL REVIEW D 96, 105004 (2017)
ðsÞ ðsÞ region of the universe to be well described by a metric gμν (b) Each ðM; g Þ is a globally hyperbolic that depends on a parameter s.Ifs is assumed to para- spacetime. ðsÞ metrize a compactly supported perturbation, the locally (c) C is also a Cauchy surface for each ðM; g Þ. covariant approach can be used to calculate how any The geometric assumptions listed here can be seen in observable Eˆ s responds to such a change. The response greater mathematical detail in Sec. 4.1 of Ref. [5]. ð Þ ˆ is evaluated as the rate of change of the observable with Consider now a Hilbert-space operator Að0Þ defined on 0 respect to the parameter. As we noted in the Introduction, ðM; gð ÞÞ; this is an operator acting on a representation of this is just what we need to calculate the quantum Cram´er- the algebra generated by the quantum fields on ðM; gð0ÞÞ. Rao lower bound. This operator could, for example, be a POVM element for a We emphasize, however, that only compactly supported particle detector and belong to the algebra of operators perturbations in spacetime were considered in Ref. [5]. The localized to the spatiotemporal extent where the detector is motivation for this restriction is similar to that for restrict- active. ing the domain of distributions to test functions and ensures It was shown in Ref. [5] that Aˆð0Þ unitarily transforms that relevant quantities are well defined. We consider how under an s-parametrized metric perturbation into a new to approach more general perturbations in Sec. IV. operator AˆðsÞ. The functional derivative of the action of this The locally covariant approach is formulated in a unitary transformation with respect to the metric is category-theory framework. It involves the category of defined as globally hyperbolic spacetimes and the mapping of each to Z an algebra of observables. By this formalism, which is δ ˆ d ˆ ∘ AðsÞ d ðsÞ summarized in Appendix A, the evolution of an observable AðsÞ¼ dμðxÞ gμν ðxÞ; ð3:1Þ ds 0 δgμνðxÞ ds 0 in response to a spacetime perturbation is made math- M qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ematically well defined. ∘ ð0Þ 0 1 2 3 Note that here a spacetime is a pair ðM; gÞ, where M is a where dμðxÞ¼ j det g jdx dx dx dx is the proper four-manifold admitting a Lorentzian metric and g is a volume element for the metric gð0Þ. The interpretation of Lorentzian metric. The additional property of global hyper- this is as follows: fields are prepared in N−, they then scatter bolicity is a restriction on the causal structure on the off an intermediate region (the compact subset of geometric manifold. It removes the possibility of closed time-like assumption 3) and are measured in Nþ; s controls a localized curves and ensures the spacetime can be foliated into perturbation within this region, and Eq. (3.1) gives the Cauchy surfaces. This in turn ensures that any hyperbolic infinitesimal movement of the observables in the Hilbert- field equation (Klein-Gordon, Maxwell, etc.) has a well- space representation of AðM; gð0ÞÞ due to an infinitesimal posed initial-value formulation. perturbation ds around s ¼ 0. Further properties and inter- A one-parameter family of spacetimes fM; gðsÞg was pretations of this functional derivative and the relative considered in Ref. [5], all sharing “initial” and “final” Cauchy evolution can be found in Refs. [5,23,24]. Cauchy surfaces, as well as respective neighborhoods N− For the case of the Klein-Gordon field, with its corre- and Nþ of those Cauchy surfaces. These spacetimes differ sponding Weyl algebra of observables, it can be shown that only in their geometry, and only within a compact region both elements of this algebra and polynomials of field ðsÞ between N− and Nþ. We refer to the metric g as operators constructed from it obey the following relation perturbed when s ≠ 0 and fiducial when s ¼ 0. (Theorem 4.3 from Ref. [5]): To be more precise, we make the following geometric δ ˆ assumptions: AðsÞ i ˆ ˆ μν 0 ¼ ½Að0Þ; T ðxÞ : ð3:2Þ (1) ðM; gð ÞÞ is a globally hyperbolic spacetime. δ 2ℏ gμνðxÞ s¼0 (2) We choose a Cauchy surface C in ðM;gð0ÞÞ and two ð0Þ ˆ μν open subregions ðN ; g Þ with the following prop- Here T ðxÞ is the renormalized stress-energy tensor on the N erties: relevant Hilbert space as discussed in Ref. [5] and satisfy- ing Theorem 4.6.1 of Ref. [25]. We assume, more specifi- (a) Nþ is within the future and N− is within the past causal regions of the Cauchy surface C. cally, that it is normally ordered in accordance with the ð0Þ procedure advocated by Brown and Ottewill [26]. (b) ðN ; gN Þ are globally hyperbolic spacetimes. Inserting Eq. (3.2) into Eq. (3.1), we have (c) N contain Cauchy surfaces for the whole ð0Þ Z spacetime ðM; g Þ. ðsÞ d ∘ i μν d (3) fg g ∈ −1 1 is a set of Lorentzian metrics on M with ˆ μ ˆ 0 ˆ ðsÞ s ½ ; AðsÞ¼ d ðxÞ 2ℏ ½Að Þ; T ðxÞ gμν ðxÞ; ð3:3Þ the following properties: ds 0 M ds 0 (a) Each gðsÞ deviates from gð0Þ only on a compact ˆ subset of the region in the past of N and the where we emphasize that the operator A does not depend on þ ˆ future of N−. x. By defining the operator P as
105004-5 DOWNES, VAN METER, KNILL, MILBURN, and CAVES PHYSICAL REVIEW D 96, 105004 (2017) Z 1 ˆ ∘ ˆ μν d ðsÞ from the measurement device. We can take advantage of a P ¼ dμðxÞT ðxÞ gμν ðxÞ; ð3:4Þ 2 M ds 0 flexibility built into quantum estimation theory whereby we can obtain an uncertainty bound from any parameter that we have our estimator θ~ is sensitive to. To see this, let Θ denote the global parameter of interest d ˆ i ˆ ˆ AðsÞj ¼ ½Að0Þ; P : ð3:5Þ for a spacetime with metric gμνðΘÞ, with Θ0 being the ds s¼0 ℏ fiducial value of the parameter and θ~ denoting its estimator. The above relative Cauchy evolution equation has also We can choose a smooth, compactly supported “bump” 1 – been shown to hold for spin-2 and spin-1 fields [16 18], function, 0 ≤ χ ≤ 1, with χðxÞ¼1 on the measurement with an appropriately defined stress-energy tensor. For region. We consider the localized perturbation example, for the electromagnetic field, which we consider gμνðθÞ ≡ gμνðθ0 þðθ − θ0ÞχÞ, parametrized in terms of θ, in Sec. VII, the Weyl algebra of the Klein-Gordon field is with θ0 ¼ Θ0.Ifχ has sufficiently large extent and simply replaced by the Weyl algebra of gauge-equivalence transitions to 0 sufficiently slowly (say, adiabatically), classes of the vector potential. Then by Theorem 3.2.9 in we can argue that the sensitivity of θ~ to θ, given by Ref. [17], the functional derivative with respect to the ~ ~ ðdhθθi=dθÞjθ θ approaches that of θ to Θ, which is perturbed metric of an operator Aˆ is again given by the ¼ 0 d θ~ =dΘ 1, where the latter identity follows commutator with the stress-energy tensor, as in Eq. (3.2). ð h iΘ ÞjΘ¼Θ0 ¼ from the assumption that θ~ is an unbiased estimator to first ~ IV. ESTIMATION OF SPACETIME order in Θ − Θ0. This means that θ is also an unbiased PERTURBATION estimator of θ, to first order in θ − θ0, so that the Cram´er- ~ Due to the dual nature of operators and states, a Rao bound applies to θ with gμνðθÞ, giving consequence of Eq. (3.5) is that we can write ℏ2 ~ 2 ˆ 2 ˆ ˆ hðδθÞ ihðΔPÞ i ≥ ; ð4:3Þ ρˆð0 þ dsÞ¼e−idsP=ℏρˆð0ÞeidsP=ℏ; ð4:1Þ 4 where ρˆðsÞ is a density operator in the Gelfand-Neimark- where – ð0Þ Z Segal [27 29] representation of the algebra on ðM;g Þ 1 ˆ ˆ ∘ ˆ μν d after the action of UðsÞ, the unitary in the Hilbert-space P ¼ dμT gμνðθÞ: ð4:4Þ 2 dθ representation corresponding to the relative Cauchy evo- M θ0 lution (more precisely to a unit-preserving automorphism β How the bump function transitions from 1 on the on the algebra of observables, called gðsÞ in Appendix A). measurement region to 0 is arbitrary. The choice affects Then, noting the equivalence of Eqs. (4.1) to (2.6),we the bound, however, through excess contributions to the obtain [30] variance hðΔPˆÞ2i in the bump function’s support outside ℏ2 the measurement region. To get the best bounds on δ~ 2 Δ ˆ 2 ≥ hð sÞ ihð PÞ i 4 ð4:2Þ precision, we choose bump functions that minimize this excess variance while achieving the desired sensitivity. In where s~ is the estimator for the perturbation parameter. the examples to be considered, this excess variance can be A limitation of the quantum Cram´er-Rao bound (4.2) is attributed to contributions from a reference state such as the that the spacetime perturbation is restricted to compact Minkowski vacuum. This is because the state associated support, whereas physically interesting perturbations are with the measurement and from which the bound is typically not so restricted. For example, a variation in the computed is a localized deviation from the reference state mass of the Earth would vary the metric at unbounded and the bump function necessarily extends beyond the distances away. To apply our formalism to this situation, we region of localization. We observe that the shape of the approximate global perturbations compactly. As a first transition of the bump function from 1 to 0 affects attempt, one might consider using compact perturbations the excess variance [31–33]. In particular, the excess can that approach the global one in some limit, but this might be reduced by ensuring that the transition from 1 to 0 is ˆ 2 lead to unbounded values of ðΔPÞ , which would trivialize slow. This is analogous to the adiabatic limit. For the case the bound (4.2). To avoid this we take advantage of the fact of the Minkowski vacuum, the reference-state contribution that the measurements that yield an estimate of θ~ are of can be made arbitrarily small by this method, as demon- observables that are accessible in a compact measurement strated for example in Ref. [31]. region determined by the measurement device. Given this, The use of slowly varying bump functions is expected the bounds above are necessarily conservative for pertur- to reduce excess variance from the reference state, but bations with large extent: they apply to all observables in does not necessarily lead to readily computable bounds. For the region of the perturbation, even those causally separated this, we observe that informative measurements rely on
105004-6 QUANTUM ESTIMATION OF PARAMETERS OF CLASSICAL … PHYSICAL REVIEW D 96, 105004 (2017) deviations from the reference state with relatively large two-point correlators [16,19,25]. Such a reference state localized mean fields. This is both out of necessity and to allows us to perform normal ordering via the point-splitting maximize the signal-to-noise. Typical measurements are approach. A corresponding stress-energy tensor can then be designed not to detect the reference-state contributions to constructed, with a well-defined expectation value [25], the variance, but rather to detect an effect in the presence of which is unique up to terms which cancel in the commu- a strong mean field, which greatly enhances the signal we tator (3.5) and in the variance (and is thus sufficient for our are looking for. Indeed, for arbitrary curved spacetimes, it is purposes). Moreover, while Hadamard reference states are not known how to calculate the reference-state contribu- generally not unique, their contribution to the relevant tions, nor is it known how to design a measurement on the variance is mean field independent. probe field that detects the corresponding mean-field- independent effects. The neglect of reference-state contri- V. COORDINATE INDEPENDENCE AND butions can then be regarded as a way of finding quantum COMPACT PERTURBATIONS limits on the kinds of measurements we know how to do, which involve large mean fields. The formulation given so far is generally covariant. For the case of free fields, the large-mean-field scenario Consider two compactly supported perturbations of the is formalized by considering the measurement state as a metric where one is obtained from the other by a local displacement by a local Weyl unitary of a reference state isometry, that is one acting as the identity except on a with mean field zero. In Minkowski space, these displace- compact region in the past of Nþ and the future N−. Then ments are enacted by conventional modal displacement both perturbations induce the same relative Cauchy evo- operators, where the displacement is by an amount deter- lution, as shown in Ref. [5] (see also Appendix B). mined by the mean field. We show in Sec. VII that the A complication to this obvious conclusion of general variance hðΔPˆÞ2i has terms that grow with the mean field as covariance arises, however, when the parameter of interest well as mean-field-independent terms. We identify the is expressed in a coordinate-dependent way and we require mean-field-independent terms as the reference-state con- the use of a bump function to localize the associated metric tribution to the variance. For a fixed bump function, but perturbation. This is the situation when estimating global large mean field, the reference-state contribution becomes parameters. An example is the invariant mass of a black negligible. This is the main strategy used for the analysis of hole, where there are a number of different standard gravitational-wave detection in Sec. VII. coordinate systems to choose from. When the bump For the above discussion, we assumed that the meas- function is expressed in the first coordinate system so as urement device is contained in a finite measurement region, to be independent of the invariant mass, in the second it can where an incoming reference state such as the vacuum state depend on the invariant mass. This means that in the second is temporarily modified for the measurement. This modi- coordinate system, the bump-function-modified metric fication is usually necessary to enhance the signal that we perturbation includes a term coming from the derivative are looking for. In the case of an interferometric measure- of the mass-dependent bump function with respect to mass, ment using light, the modification is accomplished by and this leads to discrepancies in the values of the variances introducing a large-amplitude light field confined between and Cram´er-Rao bounds depending on which coordinate mirrors. To accommodate these modifications in the gen- system is used to define the bump function. If our erally covariant formalism, the background includes exter- sensitivity argument for the choice of bump function is nally introduced classical sources with fixed relationships valid, we should obtain valid bounds regardless of coor- to the manifold, meaning that these relationships are dinate system. It is desirable, however, to choose bump unchanged by the perturbation of the metric under inves- functions for which the dependence on coordinate system is tigation. The formalism and relative Cauchy evolution still negligible. In this section, we show that such is the case for applies, as suggested in Ref. [5]. For the example of the variance of Pˆ in an arbitrary spacetime, assuming a gravitational-wave detection, the fixed relationship can sufficiently large mean field. To demonstrate the basic be justified by the observation that the classical sources mechanism by which coordinate independence is achieved, follow geodesics for the original as well as the perturbed we first consider as an illustrative example the expectation metric and are thus manifestly independent of the pertur- of Pˆ in Schwarzschild spacetime. bation in Gaussian normal coordinates. The fiducial metric in Schwarzschild coordinates is While our examples involving large mean-field devia- tions are on flat backgrounds, we expect that the justifi- 2 2 −1 S μ ν m0 2 m0 2 gμνdx dx ¼ − 1 − dt þ 1 − dr cation for neglecting reference-state contributions based on S S r r large mean-field deviations also applies to general curved 2 2 backgrounds. For any free field in a (globally hyperbolic) þ r dΩ ; ð5:1Þ spacetime, there exist zero-mean-field reference states, called Hadamard states, characterized by well-defined and in isotropic coordinates it is
105004-7 DOWNES, VAN METER, KNILL, MILBURN, and CAVES PHYSICAL REVIEW D 96, 105004 (2017) 2 μ 1 − m0=2ρ d d I ν − 2 S ≠ I gμνdxI dxI ¼ dt gμνðmÞ gμνðmÞ; ð5:6Þ 1 þ m0=2ρ dm dm 0 0 4 m0 þ 1 þ ðdρ2 þ ρ2dΩ2Þ: ð5:2Þ nor are these two tensor fields related by any coordinate 2ρ S transformation. To understand this, observe that gμνðmÞ − S I I gμνð0Þ and gμνðmÞ − gμνð0Þ also represent distinct tensor By our prescription for approximating a global pertur- fields; the first terms in the two tensor fields can be made bation by a compact perturbation, we add to every instance equal by an m-dependent coordinate transformation, but of the fiducial mass m0 a bump function of the form not without making the second terms unequal. It should ðm − m0Þχðt; r; θ; ϕÞ or ðm − m0Þχðt; ρ; θ; ϕÞ: ˆ ˆ come as no surprise, then, that h∶PS∶i and h∶PI∶i appear to 2 − χ be unequal: S μ ν ðm0 þðm m0Þ Þ 2 gμνdx dx ¼ − 1 − dt Z S S r 1 ∶ ˆ ∶ ∶ ˆ tt∶ ∶ ˆ rr∶ 2 ϑ ϑ φ −1 h PS i¼ ðh T iþh T iÞr sin dtdrd d ; 2ðm0 þðm − m0ÞχÞ K r þ 1 − dr2 þ r2dΩ2; r ð5:7Þ Z ð5:3Þ 1 h∶Pˆ ∶i¼ ðh∶Tˆ tt∶iþh∶Tˆ rr∶iþr2h∶Tˆ ϑϑ∶i I r 2 K μ 1 − m0 m − m0 χ =2ρ I ν − ½ þð Þ 2 2 2ϑ ∶ ˆ φφ∶ 2 ϑ ϑ φ gμνdxI dxI ¼ dt þ r sin h T iÞr sin dtdrd d : ð5:8Þ 1 þ½m0 þðm − m0Þχ=2ρ 4 m0 þðm − m0Þχ This appearance is deceptive, however, as we see from þ 1 þ ðdρ2 þ ρ2dΩ2Þ: 2ρ Z ˆ αr 2 ð5:4Þ h∶∇αT ∶ir sin ϑdtdrdϑdφ K Z Z S I ˆ αr ˆ ϑϑ The resulting metrics gμνðmÞ and gμνðmÞ are not related ¼ dλnαh∶T ∶i − ðrh∶T ∶i ∂ by a coordinate transformation and thus are no longer K K 2 ˆ φφ 2 physically equivalent. This reflects the fact that there is no þ rsin ϑh∶T ∶iÞr sin ϑdtdrdϑdφ; ð5:9Þ unique way to approximate a global perturbation with a λ compact perturbation. Our particular choice depends on our where d is the surface element induced on the boundary ∂ initial coordinates, out of convenience. These metrics are, K of K and nμ is the corresponding surface normal. Since ˆ μν ˆ μν however, locally related by a coordinate transformation on h∶T ∶i vanishes on ∂K and assuming ∇μT ¼ 0 (as a patch restricted to the region where χ ¼ 1. Therein, both required for a properly defined stress-energy tensor [25]), metrics are locally indistinguishable from that of a black which implies that the left-hand side of Eq. (5.9) vanishes hole with mass m and are related by the m-dependent identically, we conclude that coordinate transformation that relates Schwarzschild and isotropic coordinates. h∶Pˆ ∶i − h∶Pˆ ∶i I Z S Now letting m0 ¼ 0 for simplicity, consider the (nor- ˆ ϑϑ 2 ˆ φφ 2 mally ordered) expectation value of Pˆ, where the nonzero ¼ ðrh∶T ∶iþrsin ϑh∶T ∶iÞr sin ϑdtdrdϑdφ ¼ 0: expectation value of the stress-energy is assumed to be K ð5:10Þ confined to a region K in which χ ¼ 1, i.e., K ¼ ∶ ˆ μν∶ χ 1 ∈ suppðh T iÞ and ðxÞ¼ for x K. Note that K is Note the critical role played by the vanishing divergence of strictly contained in the interior of suppðχÞ, since as μν Tˆ in the above demonstration of coordinate independ- discussed in the previous section we assume the transition ence. This is no coincidence; for more discussion of the of the bump χ from 1 to 0 is both smooth and gradual. We relevance of the stress-energy tensor to diffeomorphism have invariance in the context of relative Cauchy evolution, Z see Ref. [5]. 1 ˆ ∘ ˆ μν d ∶P∶ dμ ∶T ∶ gμν m : 5:5 In the above example, we only considered the first h i¼2 h i ð Þ ð Þ ˆ K dm 0 moment of P, whereas the Cram´er-Rao bound involves the variance of Pˆ. We deal with this question now by S I Notice that gμνð0Þ¼gμνð0Þ, since both Schwarzschild and considering an arbitrary s-dependent coordinate transfor- isotropic coordinates reduce to standard spherical coordi- mation from unprimed coordinates to primed coordinates, nates in this limit. Yet on a coordinate patch that is assumed to cover the support
105004-8 QUANTUM ESTIMATION OF PARAMETERS OF CLASSICAL … PHYSICAL REVIEW D 96, 105004 (2017) ∂ ∂ K of the mean stress-energy (for a more general treatment ðsÞ ðsÞ 0 gαβ ðxÞ and gα0β0 ðx Þ; ð5:12Þ see Appendix B). In classic index notation, the metric ∂s 0 ∂s 0 components in the two systems are related by where we now write the s derivatives as partial derivatives ðsÞ 0 μ ν ðsÞ gα0β0 ðx Þ¼L α0 ðsÞL β0 ðsÞgμν ðxÞ; ð5:11Þ to emphasize that the respective coordinates are held fixed while taking the s derivative. Because the coordinate μ μ 0 α0 where L α0 ðsÞ¼∂x ðx ;sÞ=∂x . We are interested in the transformation is s dependent, these two tensors are not tensor fields the same, but are related by