A change of perspective: switching quantum reference frames via a perspective-neutral framework
Augustin Vanrietvelde1, Philipp A. Höhn1,2, Flaminia Giacomini1,2, and Esteban Castro-Ruiz1,2
1Institute for Quantum Optics and Quantum Information, Austrian Academy of Sciences, Boltzmanngasse 3, 1090 Vienna, Austria 2Vienna Center for Quantum Science and Technology (VCQ), Faculty of Physics, University of Vienna, Boltzman- ngasse 5, 1090 Vienna, Austria
Treating reference frames fundamentally tional language also inspires a new inter- as quantum systems is inevitable in quan- pretation of Dirac and reduced quantized tum gravity and also in quantum foun- theories within our model as perspective- dations once considering laboratories as neutral and perspectival quantum theo- physical systems. Both fields thereby face ries, respectively, and reveals the explicit the question of how to describe physics rel- link between them. In this light, we sug- ative to quantum reference systems and gest a new take on the relation between a how the descriptions relative to different ‘quantum general covariance’ and the dif- such choices are related. Here, we ex- feomorphism symmetry in quantum grav- ploit a fruitful interplay of ideas from both ity. fields to begin developing a unifying ap- proach to transformations among quantum reference systems that ultimately aims at 1 Introduction encompassing both quantum and gravita- Reference frames are essential in our description tional physics. In particular, using a grav- of physical phenomena. Every time we measure a ity inspired symmetry principle, which en- physical quantity or describe a physical event, we forces physical observables to be relational do so with respect to a reference frame. In prac- and leads to an inherent redundancy in tice, reference frames are physical objects that are the description, we develop a perspective- sufficiently decoupled from the system we want neutral structure, which contains all frame to describe. When the reference frame does not perspectives at once and via which they influence the system of interest for all practical are changed. We show that taking the purposes, we treat it like an external entity in perspective of a specific frame amounts to our theoretical analysis. For example, in our cur- a fixing of the symmetry related redun- rent most successful theories, general relativity dancies in both the classical and quan- and quantum theory (incl. quantum field theory), tum theory and that changing perspec- such reference frames are taken as idealized clas- tive corresponds to a symmetry transfor- sical systems that are non-dynamical and neither mation. We implement this using the lan- back-react on spacetime itself, nor on other fields guage of constrained systems, which natu-
arXiv:1809.00556v4 [quant-ph] 22 Jan 2020 contained in it. rally encodes symmetries. Within a simple However, a more fundamental approach to one-dimensional model, we recover some physics should dispose of such an idealization and of the quantum frame transformations of take seriously the fact that reference frames are [1], embedding them in a perspective- always physical systems themselves and thereby neutral framework. Using them, we il- subject to interactions and the laws of physics. In lustrate how entanglement and classical- particular, if we accept the universality of quan- ity of an observed system depend on the tum theory, we have to face the question of how to quantum frame perspective. Our opera- describe physics with respect to a quantum frame of reference and, subsequently, how the descrip- Philipp A. Höhn: [email protected], corresponding author tions relative to different such quantum frames
Accepted in Quantum 2019-12-26, click title to verify. Published under CC-BY 4.0. 1 are related to one another. Classical frame trans- ativity this is the diffeomorphism symmetry); we formations will not suffice to switch from the per- shall see a toy version of this below. spective of one quantum frame to that of another, This entails the paradigm of relational local- as already epitomized, in its most extreme, by the ization: some dynamical matter or gravitational Wigner’s friend scenario. So what will take their degrees of freedom (in the full theory these will place in a fully quantum formulation? be fields) serve as temporal or spatial reference The answer to this question will, of course, de- systems for others and these relations are invari- pend on the concrete physics at hand. But our ant under diffeomorphisms. That is, these rela- aim here and in [2–5] will be to initiate the devel- tions are invariant under the gauge symmetry of opment of a novel and systematic method for an- general relativity and thereby physically mean- swering this question, that ultimately can encom- ingful. Such gauge invariant relations are usu- pass quantum reference frames in both quantum ally referred to as relational observables. This and gravitational physics. That is, this method applies not only to classical general relativity, but shall be applicable in both fields to produce also to (background independent) quantum grav- the sought-after transformations among quantum ity approaches where one aims at quantizing all frame perspectives. dynamical degrees of freedom, while retaining (a Reference frames (or, more generally, reference quantum version of) diffeomorphism invariance. systems) indeed provide a natural arena for an in- As such, quantum reference systems and rela- terplay of quantum and gravitational physics, ap- tional observables appear ubiquitously in quan- pearing ubiquitously in both fields. Their recog- tum gravity and, given their indispensability, nition as quantum systems themselves dates back they have been studied extensively [7,8,11–21]. to at least 1967 when, in a historical coincidence, However, owing to the challenges in a field the- two seminal papers, by Aharonov and Susskind ory context, what has not been studied exten- [6] and DeWitt [7], separately brought this recog- sively is how to switch among different choices nition to center stage in the foundations of quan- of such relational quantum reference systems in tum theory and quantum gravity, respectively. quantum gravity. For systems with finitely many The ensuing study and usage of quantum refer- phase space degrees of freedom (such as quantum ence systems took, however, rather different di- cosmological models), the first systematic frame- rections in the two fields and our goal will be to work for changes of quantum reference systems unify some of these developments. was developed in [22–24] and applied to tem- In (quantum) gravity, there is even a necessity poral references called relational clocks, but re- to employ physical systems as references with re- stricted to sufficiently semiclassical states. The spect to which to describe the remaining physics. crucial feature of this framework is a ‘perspective- On account of the diffeomorphism symmetry of neutral’ quantum theory which contains all clock general relativity, which is a consequence of gen- choices at once and each clock choice corresponds eral covariance, physical systems are not local- to a gauge fixing. ized and oriented relative to some absolute spa- In quantum information, on the other hand, tiotemporal structure, but with respect to one quantum reference frames have been extensively another [8]. This is spacetime relationalism and, discussed mainly with the purpose of devising for later purpose, we note that this is closely re- communication tasks with physical systems serv- lated to Mach’s principle, which roughly states ing as detectors. In the seminal papers [6, 25] it that what is an inertial frame is not determined was shown that it is possible to overcome super- with respect to an absolute space (as in Newto- selection rules (such as the charge superselection nian physics), but by the other dynamical con- rule in Ref. [6]) via the introduction of a quan- tent of the universe [8–10]. Physics is purely rela- tum reference frame. In Ref. [26] it was shown tional: intuitively, if one moves around the entire that quantum mechanics can be consistently for- dynamical content of the universe while keeping mulated without appealing to abstract reference the relations among its constituents intact, it will frames of infinite mass. The subsequent liter- not change the physics. Mach’s principle thereby ature [27–34] has then focused on different as- implies a symmetry principle and a correspond- pects of the introduction of quantum systems as ing redundancy in the description (in general rel- reference frames, and mainly on a) the lack of
Accepted in Quantum 2019-12-26, click title to verify. Published under CC-BY 4.0. 2 a shared reference frame, b) bounded-size refer- principle [41], stating that it is always possible to ence frames, and c) the possibility of overcoming find a quantum reference frame having a definite general superselction rules by employing quan- causal structure in the local vicinity of any point. tum reference frames (e.g., see [27] for a review). Our main ambition will be to synthesize these These approaches resort to an encoding of quan- developments in quantum gravity and quantum tum information into relational degrees of free- information into a unifying method for switch- dom. The latter are invariant under an averaging ing perspectives in the quantum theory that in- over the external symmetry group, defining deco- cludes both spatial and temporal quantum refer- herence free subsystems. A relational approach ence systems and applies in both fields. Indeed, to quantum reference frames has been consid- in the course of this work, here and in [2–5], we ered also in [35–37]. The transformation between shall show how (some of) the quantum reference two quantum reference frames is in general not frame transformations of [1] and the method of considered in this applied quantum information relational clock changes [22–24] can be accom- context, with the important exception of [30]. modated and reproduced within one framework. More foundationaly, quantum reference frames This will be achieved by adopting key ingredients have also been used to derive the Lorentz group from both sides. from operational conditions on quantum commu- In particular, we shall adopt a gravity inspired nication without presupposing a specific space- symmetry principle to develop, as proposed in time structure [38]; this exemplifies how quantum [42], a perspective-neutral super structure that information protocols can constrain the space- encodes, so to speak, all perspectives at once time structures in which they are feasible. and requires additional choices to ‘jump’ into the A suitable starting point for establishing a con- perspective of a specific frame. Technically, we nection between these efforts in quantum infor- will achieve this by availing ourselves of the tools mation and in quantum gravity is the approach to and concepts of constrained Hamiltonian systems quantum reference frames developed in [1], which [43, 44] that also play a key role in the canoni- we shall further exploit in the present paper. The cal formulation of general relativity and quantum main idea of Ref. [1] is to formulate changes be- gravity [8,19] and that were also used for the re- tween quantum reference frames in an operational lational clock changes in [22–24]. and fully relational way, without referring to any The symmetry principle will, as mentioned external or absolute entity. Within this formu- above, entail two related key features: (i) an lation, Ref. [1] investigates the extension of the inevitable redundancy in the description of the covariance of physical laws under such reference physics (gauge freedom and constraints), and (ii) frame transformations, paving the way for a for- that the physically meaningful (i.e. gauge invari- mulation of a notion of “quantum general covari- ant) information is purely relational. The inher- ance”. Such developments give concrete meaning ent redundancy will permit us to treat all pos- to the idea of describing physics from the point sible reference frames as part of a larger phys- of view of a quantum frame of reference. This ical system at once and on an equal footing; approach might be particularly relevant in the a priori no choice of frame is preferred and no context of quantum gravity, where a fixed no- frame is described externally. However, in order tion of spacetime (spacetime metric) is not avail- to make operational sense out of physical phe- able. As an indication of the concrete possibility nomena, we must make additional choices to fix of formulating physics on a non-fixed spacetime these redundancies. We will show that choosing metric, Ref. [1] develops an extension of Galileo’s a system from the perspective-neutral picture to weak equivalence principle for quantum reference serve as our reference frame is equivalent to fix- frames, which holds when the reference frame is ing these redundancies and that classically this is a system falling in a superposition of accelera- a choice of gauge. Accordingly, (at least classi- tions. This approach is in resonance with works cally) switching from the internal perspective of aiming at formulating physics on indefinite causal one frame to another will amount to a symme- structures from an observer-dependent perspec- try transformation as in [22–24]. Our approach tive [39–41]. In particular, it is closely related thereby connects with, but also extends the dis- to Hardy’s proposal for a quantum equivalence cussion in [45], where it is argued that the redun-
Accepted in Quantum 2019-12-26, click title to verify. Published under CC-BY 4.0. 3 dancy in gauge theories is not just a mathemat- linking Dirac with a given reduced quantization ical artifact, but expresses the fact that physics constitutes a quantum symmetry reduction pro- is relational and provides the ‘handles’ through cedure, i.e. the quantum analog of phase space which systems can couple (and relate to one an- reduction. This reduction is always formulated other) in different ways.1 relative to a choice of quantum reference frame. Conversely, the operational language of quan- In this simple model, all these transformations tum foundations and, specifically, the approach will be valid globally on phase and Hilbert space, to quantum reference frames in [1] will supply so that no technical subtleties cloud our main ar- our operational interpretation of the formalism. guments and interpretation. In particular, it will inspire compelling new in- However, in generic systems, it will not al- sights into the quantization of constrained sys- ways be true that the quantum theory obtained tems. These insights will thereby be of relevance by applying the quantum symmetry reduction to for quantum gravity. Indeed, there exist two main Dirac quantization coincides with a specific re- strategies in the literature for canonically quan- duced phase space quantization, in line with the tizing constrained systems: observations in [20, 21, 51–59]. This will not be a problem for our approach, as we explain in more Reduced quantization: One solves the con- detail later: in general, we will interpret the result straints first at the classical level and only of applying the quantum symmetry reduction to quantizes non-redundant gauge invariant de- Dirac quantization, which removes redundancy in grees of freedom. that description, as the perspective of the associ- Dirac quantization: One quantizes first all ated frame. (incl. redundant and gauge) degrees of free- Furthermore, in generic systems, globally valid dom and solves the constraints in the quan- perspectives of quantum reference frames will be- tum theory. come impossible (this is analogous to the Gribov problem in gauge theories) and, accordingly, the There has been an ample discussion in the lit- transformations between them will likewise not erature as to how these two quantization strate- be of global validity. This is illustrated in the gies are related – with the general conclusion that companion paper [2], where we extend our discus- ‘constraint imposition and quantization do not sion to the three-dimensional N-body problem, commute’ – and about when one or the other and in [3,4], where it will be shown how to change should be applied [20, 21, 47–59]. Adopting the relational quantum clocks, using our new method. operational language of [1], we will shed some In particular, in contrast to [22–24], these clock new light on this discussion, both technically and changes will also be valid beyond a semiclassical conceptually. regime. In this article, we will begin with a techni- cally rather simple model, subject to a linear In connection with discussions of relative and constraint, on a finite dimensional phase space. global states in the literature, we emphasize that Within this model both quantization methods our perspective-neutral structure itself will not are necessary for a complete relational interpre- admit an immediate operational interpretation, tation. As we shall see, Dirac quantization will only the description relative to a given perspec- yield a perspective-neutral quantum theory, con- tive. Concretely, this means that there will be taining all quantum reference frame perspectives global quantum states for the entire physics at at once, while reduced quantization produces the once, namely those of the perspective-neutral quantum physics as seen by a specific frame. We Dirac quantized theory. However, there will be no will also provide the transformations that take global operational states and only states relative us from one to the other and will exploit this to to a frame (which will not include its own degrees establish switches between different quantum ref- of freedom) will admit an operational interpreta- erence frames. In particular, the transformation tion. This will be exploited in [4] to develop a novel take on the ‘wave function of the universe’ 1A complementary extension of these ideas, which does in quantum cosmology, as proposed in [42, 60], not rely on gauge fixings to define frames, has also recently and suggests a new conception of the relative been put forward for the field theory context in [46]. states of relational quantum mechanics [61, 62]
Accepted in Quantum 2019-12-26, click title to verify. Published under CC-BY 4.0. 4 and their interrelations. which we will encounter in the course of our work. Quantum foundations and (quantum) gravity The aim of physics is to describe the phys- are usually considered independently. However, ical world, or at least a subset thereof. Usu- our results are a clear testimony to how a fruitful ally, this is done by choosing a perspective from interplay of their tools and perspectives can lead which to describe the physical situation at hand. to new conceptual and technical insights in both Abstractly, choosing a perspective is thus tanta- fields. mount to choosing a map from the physics of in- The rest of this article is organized as follows. terest to some suitable mathematical description In sec.2, we explain the interplay of perspective- thereof. More precisely, denote by Sphys the set neutral structures and internal perspectives in of possible physical situations one wishes to de- physics more carefully; a quick reader can skip scribe (and could, in principle, measure) and by it on a first reading. Subsequently, in sec.3, Sdes the set of mathematical objects used for the we introduce a toy model of N particles in one- description of these situations. Then choosing a dimensional Newtonian space in which we impose perspective defines a map a symmetry principle, namely global translation ϕ : Sphys → Sdes invariance, which will serve as a toy version of Mach’s principle. Here we show how frame per- from the actual physics to its description. spectives are related to gauge choices. In sec.4 we The important point to notice is that the actual quantize the classical model and explicitly reveal physics, encoded in Sphys, is, in fact, perspective- the conceptual and technical relation between the neutral. For instance, suppose the physical situ- Dirac and reduced quantization of our toy model, ation is that a billiard ball flies through space so which here give equivalent expectation values. Fi- that Sphys denotes the set of all its possible spa- nally, in sec.5 we analyze some of the operational tial velocities. The statement of such a physical consequences of describing physics from the point situation per se does not require the perspective of view of a quantum system. In particular, a con- of some reference frame, but it can be described crete example will illustrate the quantum frame from many different perspectives. Indeed, sup- dependence of the degree of entanglement of an pose there is an observer Alice who measures the observed system. Finally, we conclude in sec.6 (components of the) velocity of the ball in three with an outlook on further applications of our spatial directions and reads it off the scales of approach. Details have been moved into appen- her measurement device. Then Alice would usu- 3 dices. ally take Sdes to be R and ϕA associates to each physical velocity a three-dimensional vector, cor- responding to the three real numbers she reads off 2 A meta-perspective on perspectives her measurement device, thereby specifying the velocity relative to her frame of reference. This The quick reader can skip this section and proceed is a second point to notice: the choice of a map, directly to sec.3. i.e. a perspective, ϕA is (usually) associated with The purpose of this section is to motivate a choice of reference frame, which is why we have and specify more clearly what we mean by a now attached a frame label to it. Note that only perspective-neutral theory, as proposed in [42]. a concrete perspective has an immediate opera- To this end, we shall adapt the abstract lan- tional interpretation. guage introduced in [38] to explain from a very Of course, this structure is completely general. general standpoint how different perspectives can For example, Sphys could also represent the quan- fit into one framework and how one can switch tum states associated with (possibly an ensemble between them. We shall thus revisit some of) a physical system that one can try to estimate, fairly basic questions, illustrating along the way using tomography, in a laboratory. A physically how perspective-neutral structures already ap- distinguished choice for Sdes would be the appro- pear ubiquitously in all of physics. This discus- priate set of density matrices. Clearly, a concrete sion will also highlight some peculiarities, such as description ϕA of the quantum states depends on an absence of global perspectives in most systems a choice of reference frame as it involves the choice of interest, and explain within a broad context the of a Hilbert space basis that Alice associates with structure of the sought-after perspective changes certain measurement outcomes, say, of spin in her
Accepted in Quantum 2019-12-26, click title to verify. Published under CC-BY 4.0. 5 z-direction on which another observer Bob may tures of constrained systems, incl. canonical grav- not agree. ity. In the previous two examples, while the ac- Indeed, for a system with first class constraints, tual physics is perspective-neutral, the theories we shall propose to interpret the classical con- describing it are arguably not. For example, if straint surface and the gauge invariant physi- one wrote down a standard Lagrangian for the cal Hilbert space Hphys of its Dirac quantiza- billiard ball, it would fail to be invariant under tion as the perspective-neutral physics Sphys of general time-dependent changes of coordinates in the classical and quantum theory, respectively. configuration space; it does not abide by a full Correspondingly, we shall argue that the re- symmetry principle and thereby presupposes a duced (gauge fixed) phase spaces and their re- special class of (e.g., inertial) frames with respect duced quantizations, the reduced Hilbert spaces to which it is formulated. Similarly, at least the Hred, assume the role of the descriptions Sdes of standard textbook formulation of quantum me- the classical and quantized physics, respectively. chanics implicitly assumes the frame of the ob- Hence, a perspective ϕA will define a mapping server and her measurement and preparation de- from the constraint surface/Hphys to the reduced vices at the outset. phase space/Hred and we shall only assign an By contrast, a prime example of a perspective- operational interpretation to the latter reduced neutral theory is general relativity. The Einstein- structures; these are the physics described with Hilbert action is completely independent of coor- respect to a given classical or quantum reference dinates and choices of reference frame (it is dif- frame.3 feomorphism invariant) and so the theory does For any of the above examples and theories, it not dictate the choice of perspective from which is now also clear how to switch from the perspec- to interpret and describe the physics in space- tive of, say Alice’s frame, to another, say Bob’s, time; it contains all frame perspectives at once namely through the following transformation: and on equal footing and it is up to the physi- −1 cist to pick one. For example, when considering TA→B = ϕB ◦ ϕA . (1) the dynamics in a given spacetime in general rel- Note that, while TA→B : Sdes → Sdes is a map ativity, Sphys may represent the possible physical from description to description,4 it always pro- situations happening in that spacetime.2 Given a ceeds via the perspective-neutral structure Sphys reference frame associated to some observer Alice in-between, thanks to its compositional form. (usually an orthonormal tetrad), Sdes is then nor- This is the general form of our sought-after trans- 4 mally taken to be R . Her perspective ϕA defines formations and we shall encounter it repeatedly a (usually only locally valid) coordinate descrip- throughout our work, i.e. below and in [2–5]. tion of the physics in that given spacetime, e.g., Hence, we will always switch perspectives via encoding the tangent vector corresponding to the the perspective-neutral meta-structure in both the motion of a massive object in a four-vector whose classical and the quantum theory. Notice also the components describe the velocity relative to Al- structural resemblance to coordinate changes on ice. a manifold. However, here it is more than just a In the course of our work, we shall show how coordinate transformation: it is a change of per- to embed the discussion of quantum reference spective. frames into such a meta-framework. In partic- ular, in analogy to general relativity, we will use 3There is an interesting analogy to the relation between a symmetry principle, in the form of invariant shape dynamics and general relativity [10, 63]. The two Lagrangians and (first class) constraints, to for- theories are related via a ‘linking theory’ [64] that can be regarded as the perspective-neutral theory. When re- mulate perspective-neutral theories of reference stricted to solutions admitting constant-mean-curvature frames. We shall use these theories to argue for slicings, shape dynamics and general relativity (as reduc- a novel, more general interpretation of key struc- tions of the linking theory) can be regarded as two differ- ent descriptions of the same physics.
2 4 phys By contrast, when considering the dynamics of space- Similarly, one can construct transformations TA→B := −1 time in general relativity, Sphys may represent the space of ϕB ◦ϕA : Sphys → Sphys that are actual operations on the solutions or, in the canonical formulation, the constraint physics [38]. Since we are only interested in perspectives surface (see also comments below). and their relations, such operations are not relevant here.
Accepted in Quantum 2019-12-26, click title to verify. Published under CC-BY 4.0. 6 The transformation (1) clearly assumes the For intuition: ‘jumping’ into the perspective perspective map ϕA to be invertible somewhere. of a given reference frame defines, e.g., what it The example of general relativity above makes it means to be ‘at the origin’ in position space, clear, however, that this will in general not be fixing the translational symmetry. Conversely, possible globally; ϕA need not be defined every- starting from the assumption that one can always where on the perspective-neutral physics Sphys. ‘jump’ into a frame that is ‘at the origin’, one is In other words, in general we will find that global led to a symmetry, because our ability to ‘fix’ any perspectives on the physics (with operational in- system at the origin means its ‘absolute position’ terpretation) do not exist in most interesting sys- is not physical. tems. This will also be illustrated in the compan- The technical simplicity of the model will per- ion articles [2–4]. mit us to illustrate in sec.4 the general method In consequence, the perspective changes (1) for changing perspectives via a perspective- will generally constitute non-global transforma- neutral structure in the quantum theory and, in tions and it will become a non-trivial question particular, to derive the quantum reference frame whether (and where) Bob’s perspective ϕB can transformations constructed in [1] for the one- −1 be concatenated with the inverse ϕA of Alice’s dimensional case from first principles. In this perspective. Hence, in general it will be a non- manuscript we will thus not need to worry about trivial question too whether perspective changes technical subtleties that cloud the main argu- (1) can be concatenated and constitute a group ments and which will be studied in more com- or more general structures such as a groupoid. plicated models in [2–5]. Such questions are crucial as a lot of information about the physics resides in perspectives and their 3.1 A toy model for Mach’s principle in 1D relations. For instance, the information about a space spacetime’s geometry is encoded in the relations among its reference frames and this is also where For simplicity, we shall take the N particles to symmetries reside. be of unit mass5 and the configuration space In the following, we shall now transition from as Q = RN so that the phase space is simply 2N N a perspective-neutral structure to internal per- R . We use canonical pairs (qi, pi)i=1 as coordi- spectives and study operational consequences of nates. It turns out (see AppendixA) that a La- the ensuing transformations (1). By contrast, the grangian with global translation invariance nec- constructions in [38, 65] can be regarded as pur- essarily leads to a (primary) constraint, namely suing in the opposite direction: they start with that the center of mass momentum vanishes operational conditions on relations among inter- N nal perspectives and attempt to reconstruct a P = X p ≈ 0 , (2) perspective-neutral structure. i i=1
so that the momenta of the individual particles 3 Classical reference frame perspec- are not all independent. Note that this equation tives as gauge-fixings defines a (2N − 1)-dimensional constraint surface in phase space. The symbol ≈ denotes a weak We now construct a simple model, which incorpo- equality, i.e. an equality that only holds on this rates a toy version of Mach’s principle for N in- constraint surface. (See [43, 44] for an introduc- teracting particles in one-dimensional Newtonian tion to constrained Hamiltonian systems.) space through a global translational invariance. On the constraint surface defined by (2), the We will argue that it constitutes a perspective- Hamiltonian (following from the Lagrangian of neutral theory in which no reference frame has AppendixA) will be of the form been chosen yet, and in which physical quanti- 1 ties are relational. We will then show that going H = X p2 + V ({q − q } ). (3) 2 i i j i,j to the perspective of a particular reference frame i amounts to a gauge fixing, and, correspondingly, that switches from one frame perspective to an- 5It is straightforward to generalize the model to arbi- other are gauge transformations. trary individual particle masses and we come back to this.
Accepted in Quantum 2019-12-26, click title to verify. Published under CC-BY 4.0. 7 Clearly, the constraint is preserved by the dynam- generator, for otherwise the evolution of all vari- ics {P,H} = 0 (where {., .} denotes the Pois- ables would be unambiguously determined, given son bracket) and so, in the terminology of Dirac, initial data, leaving no room for gauge freedom. no secondary constraints arise to enforce the con- Note that λ will get fixed below when fixing the servation of P and it is automatically first-class. gauge. It is evident that the equations of motion It is therefore a generator of gauge transforma- of Dirac observables (generated by Htot) will not tions [43, 44]. Indeed, in line with the symmetry depend on λ on the constraint surface; their dy- of the Lagrangian of AppendixA, it generates namics thus features no arbitrariness, given suit- global translations, infinitesimally given by able initial data. However, there is redundancy among the Dirac ( qi → qi + {qi,P } ε = qi + ε observables mentioned. Thanks to (2), only N −1 . (4) pi → pi + {pi,P } ε = pi of the pi are independent on the constraint sur- face. Similarly, only N − 1 of the relative dis- The physical interpretation is here (see Appendix tances are independent, as qi − qk is just the sum A for a discussion): the localizations qi(t) and of qi − qj and qj − qk. Altogether, we therefore motions q˙i(t) of the N particles with respect to have 2(N −1) independent gauge invariant phase the Newtonian background space have no physi- space functions. Indeed, given that P generates cal meaning, but are gauge dependent. Only the one-dimensional gauge orbits in its (2N − 1)- relative localization and motion of the particles dimensional constraint surface, the reduced (i.e. are physically relevant, thereby providing a toy gauge invariant) phase space [44] is 2(N − 1)- model for Mach’s principle. Thanks to the sym- dimensional for this model. metry, physics is here relational.6 We propose to interpret what we have de- Given the gauge symmetry, we need to find scribed thus far as a perspective-neutral super physical quantities that are gauge invariant and theory. Using this structure, we derived the gauge thus do not depend on the localization and mo- invariant degrees of freedom, but we have not de- tion relative to the Newtonian background space. scribed them from the perspective of, e.g., any Technically, these are phase space functions O, of the N particles, each of which could serve as a which Poisson-commute with the gauge genera- physical reference system. That is to say, we have tor on the constraint surface {O,P } ≈ 0 (i.e., not chosen any reference frame from which to de- are invariant under the gauge flow generated by scribe the physics. The perspective-neutral super P ) and are called Dirac observables. In this sim- structure contains, so to speak, all perspectives at ple model, there are obvious examples: For in- once and thereby does not by itself admit an im- N stance, all N momenta pi and all 2 relative mediate operational interpretation. Instead, we distances qi − qj, i, j = 1,...,N are Dirac ob- shall now argue that choosing the internal per- servables. Clearly, H is also a Dirac observable spective of a reference system on the physics is and the total Hamiltonian (sum of a gauge in- tantamount to choosing a particular gauge fix- variant Hamiltonian plus a linear combination of ing. In particular, the perspective-neutral struc- gauge generators [43, 44]) thereby reads ture tells us there is a redundancy among the ba- sic Dirac observables, but it does not by itself 1 H = X p2 + V ({q − q }) + λP , (5) choose which of the Dirac observables to consider tot 2 i i j i as the redundant ones. Gauge fixing will take care of this. where λ is a Lagrange multiplier, namely an ar- bitrary function of time which encodes the gauge freedom (eq. (58) in the Lagrangian formulation 3.2 Choosing an internal perspective = choos- in AppendixA) in the canonical equations of mo- ing a gauge tion. Intuitively, it is clear that an arbitrary func- We shall now reduce the phase space, getting rid tion of time will have to appear in the evolution of gauge freedom altogether and working only 6In this simple model, only the spatial physics is rela- with the physical quantities written in a partic- tional, while we have kept the absolute Newtonian time as ular gauge. Suppose we want to describe the physical. One can also make the temporal physics of such physics from the internal perspective of particle N particle models relational, see e.g. [9, 10, 66, 67]. A. We are free to define A as the origin from
Accepted in Quantum 2019-12-26, click title to verify. Published under CC-BY 4.0. 8 which to measure distances in coordinates, im- being a new gauge-fixed phase space, we have to posing (emphasized through the symbol !) specify the bracket structure on it that is inher- ited from the original phase space R2N . For con- ! χ = qA = 0 , (6) strained systems, this amounts to replacing the Poisson bracket with the Dirac bracket [43, 44]. which is a global gauge fixing, as {χ, P } = 17 In the present model it simply reads: and technically this implies that the constraints become second class. {F,G}D = {F,G}−{F,P }{χ, G}+{F, χ}{P,G} , This gauge choice indeed corresponds to ‘tak- (11) ing the point of view of A’, since now all relative for any phase space functions F,G, where {., .} distances between A and the other N −1 particles denotes the usual Poisson bracket. The Dirac (these are a complete set of independent configu- brackets of our basic phase-space variables are ration Dirac observables) simply become then:
qi − qA 7→ qi , i 6= A. (7) {qA, pA}D = 0 , {qi, pj}D = δij , ∀ i, j 6= A. Accordingly, we can consistently interpret the qi (12) as position measurements of the remaining parti- Hence, this reduction simply discards parti- cles relative to particle A. (The relative distances cle A’s position and momentum from among the among the remaining particles are clearly redun- physical degrees of freedom and we pick the re- dant information.) maining ones as coordinates of the reduced phase It is clear that we should also solve the redun- space. We thus end up with a theory for N − 1 dancy among the basic momenta for pA, particles – as seen by A. The corresponding re- X duced Hamiltonian can be computed from (5) and pA ≈ − pi , (8) (2,6): i6=A so that all the N − 1 pi6=A become the indepen- Hred = X p2 + X p p + V ({q } ) . (13) dent momentum variables. Note that pA is not A i i j i i6=A i6=A i6=j proportional to q˙A alone (see (61) in Appendix i,j6=A A) so that the fact that generally now pA 6= 0 does not mean the motion of A is not fixed. This Hamiltonian is of a somewhat non- In fact, we have to ensure that defining A as standard form: the usual 1/2 factor in the kinetic the origin is consistent at all times. This fixes the energy is not present and there are couplings be- Lagrange multiplier λ. Indeed, the equations of tween the pi’s. However, it encodes the relational motion are physics correctly. Indeed, restricting ourselves to the N = 3 case for clarity, the equations of mo- ∂Htot q˙i = ≈ pi + λ , (9a) tion give the accelerations (writing ∂i := ∂/∂qi): ∂pi ∂H ∂V q¨ = −2∂ V − ∂ V, (14a) p˙ = − tot ≈ − , (9b) B B C i ∂q ∂q i i q¨C = −2∂C V − ∂BV. (14b) ! so that the conservation of (6), namely q˙A = 0, Recall that the variables qB and qC encode the imposes relative positions of B and C with respect to A λ = −pA , (10) in the reduced phase space. Thus, if we take for thereby fixing any arbitrariness in the equations example (14a), the factor 2 in ∂BV stems from of motion. Inserting (10) in (9) gives us the dy- the fact that the effect of, for instance, an attrac- namics of all particles in the chosen gauge, i.e. ‘as tive force between A and B has to be counted seen by A’. twice, as it both pulls B towards A and A to- We noted above that the reduced phase space wards B. As for the presence of a ∂C V , it is due is 2(N −1)-dimensional and it is clear that it is co- to the fact that, even in the absence of an interac- tion between A and B, an interaction between A ordinatized by the (qi, pi) where i 6= A. However, and C will affect the position of A, and thus the 7This implies that χ = 0 intersects every P -generated position of B relative to it. These considerations gauge orbit once and only once. generalize to arbitrary N.
Accepted in Quantum 2019-12-26, click title to verify. Published under CC-BY 4.0. 9 As an aside, it is interesting to also look at what defined through P , by C. PBC|A canonically em- Hamiltonian (13) becomes had we permitted the beds into C as the intersection C ∩ GBC|A where particles to have differing masses mi in (57): GBC|A is the gauge fixing surface defined by the gauge χ = 0. This defines an embedding map
1 1 1 ιBC|A : PBC|A ,→ C . (16) Hred = X + p2 A 2 m m i i6=A i A whose image is C ∩ GBC|A. It is important here p p that P is equipped with the interpretation as + X i j + V ({q } ) (15) BC|A m i i6=A the physics seen by A to avoid ambiguities in the i6=j A i,j6=A embedding map. Indeed, abstractly, the reduced phase space is the space of gauge orbits (i.e., ev-
In the limit mA → ∞,(15) becomes the usual ery gauge orbit corresponds to one physical state) Hamiltonian, in agreement with the fact that a and thus simply the quotient Pred = C/ ∼, where reference system with infinite mass can be used ∼ is the equivalence relation that identifies points as an inertial frame. This limit also recovers the in the same orbit. This abstract Pred can be inter- standard situation in quantum mechanics, where preted as the perspective-neutral phase space: it the description is given with respect to a classical is gauge invariant and coordinatized by Dirac ob- reference frame. servables (which really are functions on the set of In summary, we interpret the reduced phase in orbits). It is also isomorphic to every (globally) a particular gauge as the physics described rel- gauge fixed reduced phase space. Without fur- ative to a reference frame, which corresponds to ther information, the embedding of the abstract that gauge. reduced phase space Pred into C would be highly ambiguous. However, here it is the physical in- terpretation of the gauge fixed PBC|A that singles 3.3 Switching internal perspectives out its embedding. Note that C ∩ GBC|A indeed defines a 2(N − 1)-dimensional hypersurface in It is clear that going from one reference frame to the original phase space R2N . another amounts to a finite gauge transformation Similarly, one can define a ‘projection’ on the constraint surface (and a corresponding swap of which Dirac observables to treat as re- πBC|A : C ∩ GBC|A → PBC|A , (17) dundant, achieved simply through an exchange so that πBC|A ◦ ιBC|A = IdPBC|A . This projection of A, C labels). That is, in order to switch per- essentially drops all redundant embedding infor- spective, we have to go back to the perspective- mation. The same construction can, of course, neutral structure of the original phase space, into also be carried out for the reduced phase space which the reduced phase space embeds. We shall PAB|C in C perspective, by simply exchanging A only be schematic here as the situation is geo- and C labels. In particular, GAB|C is now defined metrically evident (see fig.1); the details of the by qC = 0. following discussion can be found in AppendixB. In order to switch from A to C perspective, we The quick reader may skip the following para- now need the gauge transformation αA→C , gen- graphs and proceed directly to sec.4 . erated by the constraint P , that takes us from Denote the reduced phase space in A perspec- one embedding C ∩ GBC|A to the other C ∩ GAB|C . tive by PBC|A. As discussed above its canonical In AppendixB, we show that this defines a map coordinates are (qi, pi)i6=A. Next, denote the con- SA→C : PBC|A → PAB|C that produces the ex- straint surface in the original phase space R2N , pected result