Extracting Geometry from Quantum Spacetime: Obstacles down the road
aa r X i v : . [ g r- q c ] J un Extracting Geometry from Quantum Spacetime:Obstacles down the road
Yuri Bonder, Chryssomalis Chryssomalakos, and Daniel SudarskyInstituto de Ciencias Nucleares, Universidad Nacional Aut´onoma de M´exicoApartado Postal 70-543, Ciudad de M´exico 04510, M´exicobonder; chryss; [email protected] 27, 2017
Abstract
Any acceptable quantum gravity theory must allow us to recover the classical spacetime in the appro-priate limit. Moreover, the spacetime geometrical notions should be intrinsically tied to the behavior ofthe matter that probes them. We consider some difficulties that would be confronted in attempting suchan enterprise. The problems we uncover seem to go beyond the technical level to the point of questioningthe overall feasibility of the project. The main issue is related to the fact that, in the quantum theory, itis impossible to assign a trajectory to a physical object, and, on the other hand, according to the basictenets of the geometrization of gravity, it is precisely the trajectories of free localized objects that definethe spacetime geometry. The insights gained in this analysis should be relevant to those interested inthe quest for a quantum theory of gravity and might help refocus some of its goals.
The recovery of classical spacetime from theories involving pregeometrical concepts, like Loop QuantumGravity [1], Spin Networks [2], Causal Sets [3], Causal Dynamical Triangulations [4], etc., is a highly nontrivialtask. Here, we consider the problem of the recovery, at least at an effective level, of the standard geometricalnotions of General Relativity (GR) from such quantum geometrical structures. Naturally, the spacetimemetric that would emerge from a generic fundamental spacetime theory, should realize the equivalenceprinciple, as that is the essence of GR. Namely, it should describe the motion of free physical objectswithout postulating it in an ad hoc manner, that is, one should recover the world lines of the effective point-like objects, which must coincide with the geodesics of the metric. Otherwise we could not argue justifiablythat we have recovered GR from a more fundamental theory.Even if one learns how to build the space of physical states of the fundamental spacetime theory, andhow to construct the operators associated with geometrical quantities, like light cones, four volumes, spatialvolumes, areas, or lengths, it is essential that the spacetime metric derived form such notions plays the rolethat it has in the classical theory. In other words, the emerging spacetime geometry should be at the basis ofthe behavior of the physical objects, and it is precisely this feature, presumably a direct consequence of thefundamental theory, what should give empirical meaning to the notion of a spacetime metric. In a detailedanalysis by Reichenbach [5], it is shown that without such a clear connection between spacetime geometryand the matter one uses to probe it, the statements made about the former are vacuous. This, in turn, pointstowards an essentially relational approach [6]. Something similar ought to hold, at a more fundamental levelof description, if the standard classical characterization of spacetime is to emerge naturally from it. Thus,we take the view that the description of the matter fields, and their role in providing a characterization ofthe spacetime geometry, should be an essential aspect of such a theory.The guiding principle of the analysis that we undertake is to associate suitable world lines to physicalobjects, which are taken to represent the geodesics of an emergent metric, assuming this is at all possible.1nce this is accomplished, one could attempt to read off the effective metric from the collection of such worldlines. We focus on the difficulties faced in such an enterprise which are ultimately tied to the fact that, inthe underlying fundamental theory, one cannot expect to have a notion of point particles. In fact, the matterexcitations are expected to have extended support since, when going from the fundamental theory towards aneffective description, one would need to pass through a phase where matter is represented by quantum fields.The problem is then tied to the need to associate world lines to such extended excitations. We assumethat such world lines, in the intermediate stages connecting the fundamental and effective descriptions,correspond to something like centers of mass (CoM) associated with relatively localized distributions ofmatter (i.e., regions with nonvanishing energy-momentum tensor).The program thus involves associating to any extended matter distribution of a certain scale, a CoMworld line, and using those world lines to extract the effective metric at that scale. Here we note that, as itis well known [7, 8], even in GR, in general, the motion of extended objects does not correspond to geodesicsof the underlying spacetime. This fact, in turn, is what suggests the idea that there might be a new effectivemetric for which such world lines were geodesics. However, as we shall see, trying to recover an effectivespacetime metric from these CoM world lines is, at best, problematic.It is worth mentioning that the question we explore is related to the problem of spacetime averaging. Theissue of characterizing the spacetime average of relevant quantities in GR is an open problem (for a reviewsee [9]). This problem was initially recognized in reference [10] (an English version can be found in [11]) andfurther attention was generated with the appearance of [12]. Although there are proposals that work forparticular cases, for example on scalars [13], or when the average is done on spatial hypersurfaces [14], andeven some covariant versions for tensors [15,16], the issue is still unsettled. One of the open issues is the fateof the Einstein equations under such averaging procedure, since, in general, the Einstein tensor associatedwith the average metric is not equal to the average Einstein tensor, and thus, the connection of such objectswith the average energy-momentum tensor is unclear. However, the issue of the relationship between theaveraged metric and the average motion of the physical objects in the corresponding spacetime geometryhas received less attention.One natural proposal which, as far as we know, has not been considered before, is to define the averagedeffective spacetime metric, at a given scale L , as that for which the world lines of the CoM of suitable testobjects of proper size L are geodesics — section 3.1 clarifies what such objects should be. The appealingfeature of this proposal is that an effective equivalence principle is built in ab initio for the effective metrics.On the other hand, if we take that approach too literally, we face a clear breaking of that principle, reflectedin the fact that there is, in general, no universal scale-independent metric. The point, of course, is that weshould consider that the “true” metric corresponds to the smallest possible scale, as this is dictated by theavailable probes. As far as we can see, in most applications to cosmological and astrophysical problems, thisapproach does not face any serious drawbacks, as one can never go to scales below those characteristic of,say, a grain of dust. Moreover, one should have a clear understanding that the effective metric can only beconsidered as a useful auxiliary tool.We should also stress that the idea of a spacetime that emerges from more fundamental degrees of freedomis not new. There are many models in the literature proposing a realization of such a process [17–30], however,our approach is different since we look for possible fundamental obstructions in this program (for studies inthe same spirit see references [31, 32]).The initial exploration of the issues described above is the goal of the present manuscript, which isorganized as follows: In section 2 we list the main steps required to obtain an emergent geometrical descriptionfrom the fundamental quantum theory of gravity and we specify in which step we perform our analysis. Next,in section 3 we deal with the program of extracting an effective geometry at a certain coarse-grained scale,from a smaller scale, focusing on those aspects that already appear in the classical description. We alsoexplore a number of approaches for the extraction of an effective geometry using extended but classicalobjects. Section 4 is devoted to further difficulties including some aspects of the problem which, we believe,could have implications in any attempt to extract a classical spacetime from a more fundamental theory ofquantum gravity. Finally, in section 5 we discuss the significance of our findings and future prospects forthis line of research. 2 The emergence of general relativity
We now frame the problem as a sequence of regimes, and of steps connecting them, providing a path torecover GR from a fundamental quantum gravity description.1.
Full Planck-scale regime.
The fundamental spacetime degrees of freedom are characterized in termsof an appropriate quantum gravity language. These degrees of freedom, when excited in appropriatemanners, correspond to what we call spacetime, which however, at this point is represented in terms ofpregeometrical notions, such as the excitations of a spin network, events of the causal sets, simplices indynamical triangulations, etc. , and the matter degrees of freedom are characterized by secondary andsubsidiary mathematical objects intrinsically tied to the former. Thus, the matter degrees of freedommust be thought of as “attached” to the fundamental pregeometrical degrees of freedom as a naturalway to represent, at the basic level, the general idea that matter “lives” in spacetime.2.
Hard semi-classical gravity regime.
Spacetime geometry is characterized by a metric and the mat-ter degrees of freedom are described in the language of quantum fields in curved spacetime. However,in this regime, the matter states are highly nonclassical and the curvature variations over the regionsof interest are large, although all measures of curvature are below the Planck scale. We think of thisregime as representing situations where the changes in curvature are large even over regions where thefundamental excitations of the matter fields have nontrivial support.3.
Soft semi-classical gravity regime.
The spacetime geometry is characterized by a metric and thematter degrees of freedom are described in the language of quantum fields in curved spacetime. Herewe assume that the matter fields are in states that result in a classical-like behavior and the metrichas been smoothed out by averaging over suitable regions, leading to strong bounds on curvatures andtheir variations (e.g., with variations of, say, the curvature scalar, ∆ R , over spacetime volumes of size L bounded by | ∆ R | < /L ).4. GR with point particles . There is a classical description of spacetime in terms of a smooth metricand where we have a classical description of matter in terms of collection of particles moving alongsome specific world lines.5.
GR with fluids . Spacetime is described in terms of a smooth metric and matter is modelled as afluid, which describes, in an effective way, a large collection of classical particles.It is worth noting that the regime 5 is, in fact, the regime for which GR was formulated and where thetheory has been significantly tested, leading to excellent agreement between observations and predictions [33].In the regime 4 we believe that GR is reliable, but, in fact, the theory has only been partially tested usingtest point particles (say elementary particles as gamma rays, neutrinos, and perhaps the protons and heaviernuclei constituting the high-energy cosmic rays) in the context of astrophysics [34] and cosmology [35], as wellas simple test of Newtonian gravity in the lab with elementary particles [36]. However, as far as we know,there are no tests of the role of such elementary particles as sources of gravitation. Most GR precision testsare actually made with extended objects such as torsion pendula [37] or smooth spheres as in Gravity ProbeB [38]. Regarding regime 3, we can point only to the studies of inflation and the generation of primordialcosmic perturbations [39] as examples where empirical data supports the results of the theoretical analysis.Needless to say that the regimes 1 and 2 remain, up to day, empirically unexplored.According to our current theoretical understanding, we have some ideas on how to connect the differentregimes. For regime 1 we have essentially no control and, although there are several promising proposalsthat are being actively pursued, there is no theory that can be thought of as definite, reliable, and fullyworkable. Thus, it is clear that the step connecting 1 to 2 is not available either. In fact, as is well known, ifthe regime 1 is based on a canonical approach, what one faces is a timeless theory, and the task of recoveringtime is nontrivial [40].Regime 2 is where semiclassical gravity together with quantum field theory in curved spacetime are meantto be valid. The implicit connection between the two theories is straight forward: the quantum field theoryis constructed over the spacetime metric g ab in the standard fashion (say, using the algebraic approach [41])and the matter fields are supposed to be characterized by a particular quantum state ω such that Einstein’s3quations hold for the renormalized expectation value of the energy-momentum tensor, ω [ T Renab ], which actsas the gravitational source. Here one faces the issue of fully characterizing the renormalization and specifyingthe finite counter-terms but, otherwise, things are under control. A similar description holds for regime 3,and thus, passing from regime 2 to 3 should be straightforward. To do that, one would start with theexpectation value ω [ T Renab ]( x ) on some appropriate state characterizing a situation in regime 2 and averageit over spacetime regions of a suitable size to obtain a smeared object ¯ ω [ T Renab ]( x ), which would then be usedto define an effective metric g (3) ab through some compatibility conditions. However, this part of the programis technically demanding and it is not studied here.Regime 4 is where we seem to face no fundamental questions and every aspect of the theory is, in principle,under control. Free point particles follow the geodesics of the effective metric in this regime, g (4) ab , and thematter energy-momentum tensor is simply T ab (4) ( x ) = X i m i Z dλu ai u bi δ ( x − γ i ) , (1)where m i and u ai ( λ ) are the i th particle mass and four-velocity, respectively. For simplicity we are consideringthese particles to be noninteracting and we denote the corresponding geodesics by γ i ( λ ), where λ is thecorresponding parameter. Consider the step connecting the regimes 3 and 4, i.e., the transition from localizedfield excitations to effective point particles. This step involves caclulating < ω | T Renab | ω > , identifying thelocalized excitations, and replacing them with effective point particles of mass m i , and world lines γ i . Giventhese world lines, we would need to identify them as the geodesics of g (4) ab . That is, we would need to constructa tensor field g (4) ab such that, for all i , u ai ( λ ) ∇ a u bi ( λ ) = 0 where ∇ a is the derivative operator associated with g (4) ab .Regime 5 is the one for which GR was constructed and which is used in almost any relevant application.Note that we often go back to regime 4 when considering things like planetary motion in the solar system, butwe take that just as a very good approximation, and when dealing with high precision studies, we introducecorrections that take into account complexities of the finite sizes of the objects under consideration. However,the transition from regime 4 to 5 is also nontrivial. To actually achieve it, we would need, in principle, arecipe to smear the nonsmooth tensor T ab (4) over suitable spacetime regions, leading to a smooth averagedtensor field T ab (5) . Then, the construction of the effective metric suitable in this regime, g (5) ab would need toensure that the corresponding conservation equation holds. Again, in most practical examples, the inputis probably not enough to completely determine g (5) ab , but it is certainly a consistency requirement for asituation where we expect GR to hold.The task at hand would be to explore some of the issues that arise in the various steps, primarily thequestion of geometry extraction from the behavior of matter probes, and the characterization of the energy-momentum tensor of the effective bodies from the immediate previous level. The task is clearly a formidableone, and thus, here we address it using a simplified version of the problem. The simplified analysis we carryout contemplates the passage from regime 4 to a regime 4 ′ corresponding simply to one characterized bya larger scale. The idea is that starting from regime 3 we may use different averaging scales and arrive atvarious versions of regime 4. We want thus to consider starting from the regime 4 tied to a certain scale L and then going to the regime 4 ′ corresponding to the scale L ′ with L ′ ≫ L . In other words, we wantto consider the steps leading to the generation of the effective world lines, effective metric, and effectiveenergy-momentum tensor of the regime 4 ′ assuming we are given these same elements at the level of theregime 4. It is worth noting that we do not consider the very important question of the degree to whichEinstein’s equations for regime 4 ′ are satisfied. We want to study the change on the geometrical quantities that can be read off from the motion of extendedprobes when the probing objects scale is modified. For this purpose we need to extract a world line char-acterizing the probe’s evolution, which is to be identified as a geodesic of the effective spacetime metric.Clearly, these world lines have to be defined in a canonical and background independent way, otherwise wemight introduce structures that are in conflict with the underlying diffeomorphism invariance ruining any4hance of recovering GR. The most natural option available corresponds to choosing a covariant center ofmass associated with the matter distribution given in terms of the classical energy-momentum tensor (or bythe expectation value of the energy-momentum tensor operator in the quantum state of the matter fields).There could be alternative methods to assign world lines to extended objects not involving the center ofmass, but they involve, in some way or another, similar problems. However, the method we are advocatingseems to be the one that better realizes the equivalence principle in the sense of establishing, by construction,that free objects move along geodesics.We assume that we are presented with a background spacetime with metric g ab and we try to extractan effective geometry by analyzing the center of mass trajectory for some extended test objects. Observethat the basic aspects of our method are already present in the simple setting where the object’s internalstructure is characterized in terms of a multipolar expansion [42]. In fact, in the limit where the size of thatobject is taken to zero but the expansion is truncated at the second (dipolar) term, the evolution of thesystem can be described by the so-called Papapetrou equations [7] T c ∇ c S ab + T c T b T d ∇ d S ac + T c T a T d ∇ d S bc = 0 , (2) T d ∇ d ( M T a + T b T c ∇ c S ba ) − S bc T d R bcda = 0 , (3)where T a is the unit tangent to the object’s world line, S ab and M stand, respectively, for the total spin andtotal mass of the object (whose definitions are given below), and ∇ a and R abcd are the derivative operatorand the Riemann curvature tensor associated to g ab . All tensors are evaluated on the corresponding pointon the object’s world line. Evidently, the presence of spin and a nonconstant mass produce a deviation ofthe CoM trajectory from a geodesic of the background geometry. Thus, the procedure we describe abovewould fail to recover the true backgound geometry. Nonetheless, we can still ask if for a given class ofextended objects, there is an effective geometry such that its geodesics coincide with the corresponding CoMtrajectories, and which depends on the extended objects’ characteristics. This subsection is based on the definition of the covariant (i.e., observer independent) CoM in GR originallyformulated by Dixon [42]. This definition applies to an extended object described in terms of its energy-momentum tensor, however, for simplicity we focus in the particular case of a collection of N free test massivepoint-like particles; all the steps we describe generalize straightforwardly to continuous matter distributionsby properly changing summations to integrals. To be certain that the CoM can be defined, we assume thehypotheses of reference [8] which essentially imply that the spacetime curvature at the position of the objectis small in comparison with (the inverse of the square of) the object’s size, in such a way that the object’sworld tube is contained in a normal convex hull U .Our intention is to generalize the special relativistic (SR) notion of the centroid of a system of N particles,with respect to an observer at x with four velocity U , given byΞ a ( x , U ) = P Ni =1 E i ( U )Ξ ai ( x , U ) P Nj =1 E j ( U ) , (4)where Ξ ai ( x , U ) is the i th particle position in the observer’s frame, and E i ( U ) = U · P i is the correspondingenergy, as well as the covariant CoM definition obtained by applying the above centroid recipe in the restframe of the extended object, where the total momentum has vanishing spatial components, i.e. , it is parallelto the frame’s four velocity. Now, there are three basic questions that we need to answer before we are ableto generalize a special relativistic definition to the GR case:1. What is to replace the SR simultaneity hyperplane, which is orthogonal to an observer’s four velocity?2. How can positions be treated as vectors in a general spacetime, like they are in SR?3. How are vectors in distinct events to be summed together, as is needed in computing, e.g. , the totalmomentum of a system? 5he short answers to these are, respectively:1. Let each time-like four velocity define a simultaneity hypersurface generated by geodesics orthogonalto it.2. Use the (inverse of) the exponential map to give vector coordinates to all points in the region of interest.3. Use parallel transport along the unique geodesic connecting a general point to some fixed base point,to bring all vectors in the same tangent space there, where they can be meaningfully summed.We now elaborate on these answers: let x be a point in U and U a a time-like, future-directed unit vectortangent to U at x , i.e. , belonging to the tangent space V x . We first construct the simultaneity surfacewith respect to U a , denoted by Σ( x , U ), which is given by all events that can be reached from x by ageodesic passing through x and whose tangent, at this point, is orthogonal to U a . Note that Σ( x , U ) may,in general, not be a hyper-surface, but the hypotheses we make are such that, its restriction to U is indeed aspace-like hyper-surface. Denote by y i ( x , U ) the point where the i th particle world line intersects Σ( x , U );since we have assumed that U is a normal convex hull, there is only one geodesic connecting y i ( x , U ) and x for every particle. This allows to unequivocally find the vectors Ξ ai ( x , U ) ∈ V x that “point in the directionof the particles”, at the instant defined by U a , by imposing two conditions: (i) The Ξ ai ( x , U ) are tangentto Γ( x , U ), namely, U · Ξ i ( x , U ) = 0, where X · Y ≡ g ab X a Y b , and (ii) the geodesic that passes through x with tangent Ξ ai ( x , U ) intersects the i th particle world line at an affine distance 1, i.e. , in terms of theexponential map, exp x [Ξ ai ( x , U )] = y i ( x , U ) . (5)We also parallel transport the momentum P ai of the i th particle from y i ( x , U ) to x , along the uniquegeodesic joining these two points, and denote the result of this operation by ˜ P ai ( x , U ). The total momentumat x , with respect to U a , is then defined as P a ( x , U ) = N X i =1 ˜ P ai ( x , U ) . (6)Note that the sum in the right hand side is well defined, as all P ai belong to V x , and that P a ( x , U ), thusdefined, is a time-like vector. Finally, observer and position-dependent energies E i ( x , U ) may be assignedto the particles by E i ( x , U ) = U · ˜ P i ( x , U ) . (7)We have now at our disposal all the elements necessary to define the GR version of the centroid,Ξ a ( x , U ) = P Ni =1 E i ( x , U )Ξ ai ( x , U ) P Nj =1 E j ( x , U ) , (8)where we use the same symbol for the GR centroid as in the SR case, given that the latter is a specialinstance of the former.Finally, to arrive at a covariant GR CoM recipe, we need to define a “rest frame” for our extendedobject. For some fixed base point x , we look for that special frame velocity U ( x ) that results in the totalmomentum of (6) being parallel to U ( x ), P a ( x , U b ( x )) = µ U a ( x ) , µ ∈ R . (9)It has been shown [8] that for a given extended object and under the work hypotheses, U a ( x ) exists andit is unique. This is then taken to uniquely define the rest frame of the extended object, with origin at theevent x . We define the x -centered rest-frame momentum of the i th particle and total momentum of thesystem, as ˜ P ai ( x ) = ˜ P ai ( x , U b ( x )) , (10) P a ( x ) = P a ( x , U b ( x )) . (11)6imilarly, the x -centered rest-frame energy of the i th particle is E i ( x ) = U ( x ) · ˜ P i ( x ) , (12)so that a reasonable attempt at a GR version of the covariant SR CoM recipe isΞ a ( x ) = P Ni =1 E i ( x )Ξ ai ( x ) P Nj =1 E j ( x ) , (13)(where Ξ ai ( x ) = Ξ ai ( x , U b ( x ))) with the understanding that the actual CoM is the exponential map of theabove vector (see Fig. 1). We are almost done, but not quite at the end yet, as the above definition depends b bb b π π Ξ ( x ) Ξ ( x )Ξ( x ) P ( x ) P ( x ) P ( x )˜ P ( x ) ˜ P ( x ) Σ x exp[Ξ( x )]Figure 1: Schematics of Dixon’s x -centered CoM definition for a two-particle system. For the CoM x should be such that Ξ vanish.on the base point x . We emphasize that this is not obvious from (13): even an x -independent CoM wouldbe described by an x -dependent normal coordinate, as moving, say, x further away from the CoM worldlinewould result in a larger vectorial coordinate, the exponential map of which would land one on exactly thesame worldline as before. The non-obvious fact is that the vectorial normal coordinate Ξ a ( x ) in (13) failsto exactly compensate for the movement of x , as described above, so that the CoM worldlines computedin frames with distinct origins do not, in general, coincide. The last element in our series of definitions isobtained by finding those special points x = ¯ x such that Ξ a (¯ x ) = 0 — the locus of the ¯ x gives the CoMworldline, independent of any observer characteristics. It is proven in [8] that the above CoM worldline is aunique continuous and everywhere time-like curve.Having at our disposal the vector fields Ξ ai ( x ), ˜ P ai ( x ), we define a total angular momentum tensor fieldvia J ab ( x ) = N X i =1 Ξ ai ( x ) ˜ P bi ( x ) − Ξ bi ( x ) ˜ P ai ( x ) . (14)Contracting this with the total momentum field gives a new vector field, W a ( x ) = g bc ( x ) P b ( x ) J ca ( x ), whichcan be shown to vanish on the CoM worldline, W (¯ x ) = 0. This CoM property was crucial in the first covariantdefinition in curved spacetimes [42] and it is sometimes used to motivate alternative CoM definitions [43].7he total spin S ab of the extended object is then defined as the total angular momentum with respect to theCoM, that is, S ab = J ab (¯ x ). In addition, the total mass is defined as the total momentum norm, evaluatedon the CoM world line, that is M = q g ab (¯ x ) P a (¯ x ) P b (¯ x ) . (15)It turns out that, in the limit when the object size tends to zero, the CoM world line is obtained from thePapapetrou equations (2)-(3), and thus, it generally fails to be a geodesic of the background metric. Onthe other hand, in flat spacetime, the objects we consider as probes always follow geodesics [44]. We nowexamine two different methods to extract information about the effective geometry by identifying the CoMworldline with the geodesics of an effective metric. In this section we assume that we can monitor the center of mass position of a given class of extended objects,and thus, we have the manifold crisscrossed by world lines. From these world lines we seek to extract somenotions of an effective geometry. In particular, to extract an effective connection we use the property thatthe vector field W a ( x ) vanishes along the CoM worldline. Let T a be the unit tangent vector of the CoMworldline, then the second derivative of W a along this line is0 = T c ∇ c ( T b ∇ b W a ) | ¯ x = ˙ T b ∂ b W a | ¯ x + T b T c (cid:0) ∂ c ∂ b W a + 2Γ abd ( Z ) ∂ c W d (cid:1) | ¯ x . (16)The above equation may be cast into the form [45]0 = ˙ T m + T r T s γ mrs (¯ x ) , (17)where γ mrs are the effective connection components.A detailed examination of this method reveals a serious obstacle: from equation (16) one can only readthe effective connection components along the CoM world line and not in the perpendicular directions. Toextract these latter components we would have to probe the same spacetime region in other directions withprecisely the same extended object. However, in a general spacetime it is not possible to ensure that twoobjects in different regions are identical. This issue seems to be related with the well-known obstacles fordefining rigid bodies in GR. Therefore, in general, these effective connection components would dependon the details of the extended objects that are used, calling into question the possibility of recovering anobject-independent spacetime geometry. Another approach that offers more direct and relevant information about the effective geometry is to readoff directly the effective sectional curvature. Recall that this curvature can be measured through the relativeacceleration of infinitesimally close geodesics. We can then define the effective sectional curvature as codifyingthe relative acceleration of the CoM world lines of neighboring free extended objects.More concretely, we use a uniparametric family of extended objects where the parameter κ selects theextended object. Also, we denote by ¯ x ( κ, λ ) the CoM world line of the object κ , which is, in turn, param-eterized by λ , and by T a the unit tangent of the CoM world line of each object, that is T a ∝ ( ∂ ¯ x/∂λ ) a .Then, the deviation vector of neighboring geodesics is X a = (cid:18) ∂ ¯ x∂κ (cid:19) a + bT a , (18)where b is fixed by the condition that X a be orthogonal to T a . Moreover, the relative acceleration, a a = T c ∇ c ( T b ∇ b X a ), is linked to the effective curvature tensor in the probed region, r abcd , via a a = − r bcda T b X c T d . (19)A series of calculations, using this approach, have been carried out in [46], in the case where the underlyingspacetime has constant curvature, and considering, for simplicity, the centroid of theprobe rather than the8ctual CoM. We found that, in general, the effective geometry depends on the details of the extended objectplaying the role of the free particle. In fact, in extreme cases, e.g., if the object is large enough or hotenough (i.e., if it has sufficient internal energy), the effective sectional curvature can be of opposite sign ascompared to the one of the underlying metric. This, by itself, seems to cast serious doubts on the possibilityof recovering an object-independent effective spacetime geometry. In addition, in reference [46] it is shownthat further complications arise when attempting to fulfill the program outlined above. First, the effectivesectional curvature depends on the relative acceleration of the center of mass world lines of two extendedobjects. Therefore, the issue of whether these objects are identical is also present in this case. In reference [46]such problems were avoided by using the background spacetime symmetry, but it is clearly impossible tobypass this issue in general. Second, the effective curvature can be read off only in the plane defined bythe vectors X a and T a ; the other components of r abcd cannot be defined without considering additionalobjects, and of course, that brings back the issue of defining identical extended objects in curved spacetimes.Remarkably, all these complications do not appear in flat spacetime, because, in that case, the CoM alwaysfollows a geodesic. Note that, as discussed in that work, the main features and problems encountered inreference [46] are expected to be present when using the CoM, instead of the centroid, of the probe.The fact that, in this approach, the dependence of the CoM world line on the object’s internal structuregoes beyond that specified by the total spin of the extended object, involving higher multipole momenta,could, in fact, become an obstacle for defining canonical probes with which to carry out our program. Thisissue needs further investigation, which is left out of the present work. Despite its long history, the CoM recipe presented in section 3.1 suffers from a consistently overlookedshortcoming: in a sense we make precise shortly, it fails to be associative. Associativity is a property of somesort of product, so, it is not surprising that our more precise take on this concept, which we now undertake,starts with defining a product: given two point particles π A and π B , call T A , T B , their respective energy-momentum tensors, that have support on the particles’ world lines. A center of mass recipe, in its simplestform, should accept T A , T B as input, and produce an effective point particle, call it π AB , described by theenergy-momentum tensor T AB , with support on the world line of π AB . Thus, we may introduce a ∗ -productbetween point-particle energy-momentum tensors, and write for the above operation T AB = T A ∗ T B . Wecertainly want the operation to be commutative, given that there is no intrinsic way to define the order ofthe factors, so that we demand T A ∗ T B = T B ∗ T A .What other properties are needed for the product ∗ to be satisfactory? A basic question emerges naturallywhen considering n -particle systems: is the above binary ∗ -product sufficient to determine the center of massby repeated application, or does such a system require a new, n -ary product, that cannot be reduced tobinary operations? The former option is undoubtedly preferable, on both aesthetic and practical grounds.But if this is going to be the case, the ∗ -product has better be associative, so as to define a unique result,regardless of the bracketing of the factors used. For example, for a three-particle system, one could compute T ( AB ) C = ( T A ∗ T B ) ∗ T C and, in a similar notation, T A ( BC ) or T ( AC ) B — it would certainly be desirable thatall these results coincide.How does the Dixon center of mass definition perform with respect to the above criteria? First, it is easyto see that the application of this recipe to a two-particle system defines a commutative but non-associative ∗ -product. Furthermore, its application to an n -particle system, with n >
2, gives rise to an n -ary productthat is not related in any discernible way to repeated applications of the ∗ -product. Thus, for example, if thecenter of mass of an n -particle system { π , . . . , π n } has been computed `a la Dixon, and an additional particleis subsequently included in the system, the calculation of the center of mass of the resulting ( n + 1)-particlesystem has to start from scratch — there is simply not enough information in the energy-momentum tensor T ...n to adequately represent the n -particle system in the ( n + 1)-particle calculation.A similar problem occurs in flat spacetime when the effective intrinsic spin of the objects is not takeninto account [47]. The recipe considers point like particles which have no intrinsic spin, and representsan extended object by its center of mass. When the spin of the composed system is not taken into theaccount, the resulting recipe fails to be associative. The proposal of reference [47] is to include the spin9f every component of the extended object when calculating its total angular momentum, and, with thismodification, the CoM definition in flat spacetime becomes associative. However, this simple solution doesnot cure all the causes that make the GR CoM nonassociative.If Dixon’s CoM definition fails to be associative, what particular characteristic of the system causes thefailure? To answer this question fully, in the general case, is a rather complicated matter — what we attempthere is a simplified treatment, aiming to give us a hint of the nature of the phenomenon. We consider athree-particle system { π , π , π } , and ask under what circumstances is the center of mass description ofthe system at its most accurate? A true point particle has a sharp position, and a sharp momentum,corresponding to a single point in phase space. A system of particles will be well approximated by its CoM ifits extension both in position and momentum space is small. For the position, the scale is set by the radiusof curvature of spacetime; for the momentum, by the masses of the particles. Assuming both of these scalesto be of order unity, we expect simplifications to occur, when the system may be fit inside a parallelepiped ofsides ǫ , η ≪ ǫ and η — one could hope that, forexample, the leading term(s) in such an expansion would be both informative and computationally accessible.The particular quantity we compute along these lines is the difference δ Ξ (12)3 in the Dixon CoM normalcoordinates for the above three-particle system, when computed in two different ways: first, we put the1 2 3 b A b B b C b O P P P PδP ˜ P ˜ P A O Figure 2: Failure of associativity in the CoM calculation of a three-particle system (see text).origin of our normal coordinate system on the 123-CoM worldline, so that the CoM normal coordinate Ξ vanishes. We then apply the recipe to the (12) subsystem, composing the result with particle 3, to get toΞ (12)3 = Ξ + δ Ξ (12)3 = δ Ξ (12)3 . It should be clear, from our definition above of the CoM of a system, thatΞ (12)3 , when exponentiated, gives a good approximation for the (12)3-CoM worldline, but, strictly speaking,it is not the exact result — the latter would be found if, instead, we moved the origin of our frame so as tomake Ξ (12)3 vanish. The two approaches though coincide in the order of perturbative calculation we workwith.We identify now the dominant term in δ Ξ (12)3 . When computing X , the four-momenta P i are parallel-transported to the center of mass O of the system along the geodesics connecting the vertices of the triangle123 in figure 2 with the center of mass O , and are summed there to give P . Computing Ξ (12)3 , on the other10and, involves parallel-transporting P , P to the midpoint A of 12, along the segments 1 A , 2 A , respectively,and then transporting their sum to O , along A O , where it is added to the transported P to give P (12)3 .The difference between the total momenta P (12)3 and P thus computed, clearly involves the holonomy ofthe parallel transport of P , P along the closed circuits 1- A - O -1 and 2- A - O -2 respectively. Since the linearsize of these two triangular circuits is of order ǫ , the above holonomy is quadratic in ǫ , and some detailedargumentation, that we will not invoke explicitly here, implies that this will fix the order of the leading termin δ Ξ (12)3 to O ( ǫ ). Accordingly, we may set, for the moment, η = 0 in our calculation, which means thatthe parallel transported four-momenta ˜ P i , i = 1 , , O , ˜ P i = ˜ P = E U , i = 1 , , , (20)where U a = P a / | P | is the 123-CoM four-velocity (we denote by | v | ≡ √ v · v the norm of a time-likefour vector). In practice then we decide what the common four-momentum ˜ P is at O for all three particles,and then parallel-transport it back to the vertices of the triangle, along the geodesics that connect them to O , to find what the various P i should be in order for η to vanish. The total four-momentum, under thesecircumstances, is given by P = 3 ˜ P = 3 E U , while the energies E i are all equal, E i = ˜ P i · U = | ˜ P | = E . (21)When computing the (12)3-CoM, the fact that P (12)3 is slightly tilted with respect to P , implies thatthe simultaneity hypersurface Σ (12)3 , orthogonal to P (12)3 , will also be tilted with respect to Σ . Theintersection point of Σ (12)3 with the world line of particle i will then be different from that of Σ , which willaffect its vectorial coordinate Ξ i . To simplify our notation, we distinguish by a prime all quantities relevantto the (12)3-computation, while unprimed quantities refer to the 123 one — in particular, P → P , and P (12)3 → P ′ . Then δ Ξ i = Ξ ′ i − Ξ i = λP i , since P i is tangent to the particle i world line. Orthogonality ofthe Ξ’s to the total momentum implies0 = Ξ ′ i · P ′ = (Ξ i + λP i ) · ( P + δP ) ⇒ λ = − Ξ i · δPP i · P , (22) i.e. , δ Ξ i = Ξ i (12)3 − Ξ i = − Ξ i · δPP i · P P i = − Ξ i · δP E P i . (23)In the above expression, δP ≡ P (12)3 − P is given by δP = ˜ P A O + ˜ P A O + ˜ P O − ˜ P O − ˜ P O − ˜ P O = 12 R O ( A, O )( P ) + 12 R O ( A, O )( P ) , with ˜ P A O denoting the parallel transport of P along 1 A O , etc. , while R O is the Riemann tensor at O , and R O ( X, Y )( Z ) m ≡ R mabc X a Y b Z c . Noting that the matrix R ( X, Y ) only depends on the plane of X , Y , andthe (signed) area of the parallelogram they span, we may infer from the above expression that δP = 112 R O ( X, Y )( P + P ) ≡ R △ ( P + P ) , (24)where X ≡ Y ≡
32 are sides of the triangle, and we used the fact that, due to our particular simplifyingassumptions, O is the barycenter of the 123-triangle (the last equality defines a convenient notation for thematrix R O ( X, Y ) ≡ R △ ). One last result that will be needed for the calculation of δ Ξ is the change in theenergy of the i -th particle, δE i = E ′ i − E i . We compute δE i = δ ˜ P i · P | P | ! = δ ˜ P i · P | P | + ˜ P i · δP | P | − ˜ P i · P | P | P · δP = O ( ǫ η ) , δX of the parallel transport of a vector X alonga closed circuit is orthogonal to X (this is the geometrical content of the antisymmetry of R abcd in its lasttwo indices), and also that P k ˜ P i . Note that in using this result for δE i in what follows, we will maintainthe second term of the middle line, even though it does not contribute to the corresponding perturbativeorder. We conclude that individual particle energies do not change, to this order, when switching from 123to (12)3, and, hence, neither does the total energy E . We may now compute the change δ Ξ in the CoMvectorial coordinates, δ Ξ m = δ E − X i E i Ξ mi ! = E − X i δE i Ξ mi + E i δ Ξ mi = E − X i ( P i · δP )Ξ mi − (Ξ i · δP ) P mi = E − S mn δP n , where, in this order of our perturbative expansion, the tildes may be removed, and we have introduced thespin S of the system, S mn = P i Ξ mi P ni − Ξ ni P mi and the total energy E = P i E i = | P | . Combining with (24)we arrive at δ Ξ m = 112 E − S mn R abcn X a Y b ( P + P ) c . (25)A more symmetrized expression results if we average the above result over all possible ways of computingthe CoM, δ Ξ m = 118 E − S mn R abcn X a Y b P c , (26)where δ Ξ m = ( δ Ξ m (12)3 + δ Ξ m + δ Ξ m ) is essentially the CoM coordinate of the three CoMs, that stilldeviates from Ξ m . Thus, the leading term responsible for the failure of associativity of the Dixon CoMdefinition involves the spin of the system interacting with the curvature of spacetime, a result certainlyreminiscent of Papapetrou’s equations. That the above result is cubic in ǫ is evident in that X , Y are bothof order ǫ , as is S mn , which involves Ξ ai linearly.Eq. (26) may look appealing in its simplicity, but fails to deliver in one important sense: with oursimplifying assumptions, the spatial part of δ Ξ m is identically zero. This is due to the fact that, for η = 0,the constituent particles of our probe are at rest in the rest frame, and so the spatial part of S mn vanishes.All we get from (26) then is the change in the time coordinate of the CoM, which is not exactly what ourinitial motivation was. To compute the change in the CoM position along the simultaneity hypersurface Σ,we have to push the above calculation to the O ( ǫ η ) regime, by assuming small, O ( η ), spatial componentsfor the constituent particles’ momenta, i.e. , by taking P mi = E ( U m + ηV mi ) , with V i · U = 0 , X i V mi = 0 . (27)Sparing the reader the uninspiring details, we find that the term in the right hand side of (25) also contributesto this higher order, but two more terms show up, proportional to X i R △ ( P i , P )Ξ mi , X i E − ( P i · P ) R △ ( P + P , P i )Ξ mi (28)respectively (with R △ ( A, B ) ≡ R abcd X a Y b A c B d )— their particular contribution to this order may be com-puted by substituting (27) in the above equation.It is worth pointing out that, in most astrophysical situations, the O ( ǫ η ) displacements in (26), (28), arequite small, compared to the typical size of the objects involved. However, the situation may be dramaticallyreversed if we contemplate extending an analysis similar to this even to the regime 2. Moreover, there aresome fundamental issues that emerge from the discussion of this section, one of which is discussed next.12 .1.1 Implications for the renormalization program We should emphasize the fundamental role that associativity plays in many of the standard theoreticaltechniques in physics. For instance, the renormalization program within the Wilsonian approach [48], isbased on the idea of integrating up the degrees of freedom that are associated with very high energies, orvery short scales, to obtain an effective characterization that is relevant at lower energies. Evidently, for thisprocess to have any hope of leading to trustable results, it is essential that when integrating from the highenergy scale Λ (1) directly to the lower energy scale Λ (3) , we obtain the same result as that one would reach byfirst integrating from Λ (1) to the intermediate scale Λ (2) , and then integrating the resulting effective theoryfrom Λ (2) to the low energy scale Λ (3) . In fact such path independence lies at the core of the renormalizationgroup techniques which are an essential aspect of the whole effective theory approach. It should be clearthat, when associativity fails, all bets are off regarding such path independence.To see this more clearly imagine we have a spacetime with metric g ab , and a set of 6 N particles withmasses m i and world lines γ i , where the subscript i labels the particle, and we proceed to integrate out theshort scale by dividing the collection into 2 N groups of 3 particles. This leads to 2 N world lines representingthe CoM of effective particles, on a spacetime with effective metric ˜ g ab , that have masses ˜ m j and that movealong their CoM world lines ˜ γ j , where now j = 1 , . . . , N . Observe that we can associate to these situationsan effective energy-momentum tensor using, for example, equation (1). Moreover, we can reiterate theprocess dividing the 2 N effective particles into N sets of particle pairs and integrating out another scale ofthe system. The point is that one could instead start by dividing the collection in N sets of 6 particles each,integrating out and representing the situation by N composite particles and the two procedures lead to twoinequivalent energy-momentum tensors which, however, are supposed to describe the same physical situationat the same coarse graining level. Note that this issue, as described here, concerns the scale transition ofclassical systems, however, something similar can be expected to occur when dealing with such change ofscale in quantum mechanical regimes such as quantum field theory.The issue that needs to be considered is the following: suppose we start with a characterization of thematter degrees of freedom at the quantum gravity scale Λ Planck , and assume we integrate out the fluctuationsassociated with very high energy scales. In that way we recover a characterization of the situation at handin terms of an effective spacetime metric and a quantum field theory on that curved spacetime, with aneffective energy-momentum tensor and with characteristic cutoff scale Λ
GUT . On the other hand, we wouldlike to recover the effective description at, say, the electroweak scale Λ EW . The question is whether we wouldobtain the same result by integrating directly from Λ Planck to the low energy scale Λ EW . We do not, atthis point, carry out any detailed calculation but rather argue on the basis of what has been found that theanswer is likely going to be negative. Also, we want to point out that similar concerns have been raised inreferences [49, 50] within the cosmological context.The lack of associativity of the CoM that we have found, and which shows that, in general, the effectivegeodesics one obtains depend on the manner one integrates out the fundamental constituents of our testobjects, suggests that the resulting metrics that would be obtained in the two ways, are generically different,and that those differences depend strongly on the characteristics of the fluctuations, or more specifically, onthe behavior of the fundamental degrees of freedom at the Λ Planck scale.One of the fundamental suitability tests for a quantum gravity theory is the recovery, in an appropriatelimit, of GR. Our findings above indicate that perhaps the test itself needs to be substantially reformulated.Of course, the problem could be bypassed if one could define an observer-independent CoM world line whichis associative, but at the moment we see no path in this direction. To recover associativity one would haveto sacrifice observer independence and take a relational point of view, as it is done in reference [46]. Inthis case one uses the simultaneity hypersurface associated with some particular observer, regardless of theobject under study. However, such an observer dependent CoM would not be a suitable starting point forrecovering an observer-independent metric characterization of spacetime.
There are well known arguments that combine quantum mechanics and GR which conclude that there mustbe fundamental localizability limits associated with black hole formation once sufficient energy is confined ina small enough region [51]. In this regard we would like to put forward a simple argument suggesting thatthe issue ought to be revised. Suppose we manage to concentrate a given amount of energy in a region that,13s seen in frame S , is close to the black hole forming regime. Now, for an observer boosted with respect to S , with boost factor γ , the size of the region where the energy is confined is shrank by γ while the energyconcentrated in that region increases by the same factor. Thus, if γ is large enough, the black hole formationbound is reached, for the boosted observer, and the two observers disagree in their expectations of blackhole formation. This, in our view, illustrates the need to investigate in more detail issues that, like this one,lie at the interface between gravity and quantum theory, with observer independent methods.On the other hand, as it is well known, the position operator plays a very important role in quantummechanics, and there is no satisfactory definition of this operator in quantum field theories. Now, sincethe CoM we have been considering seems to be applicable to situations involving the matter fields energy-momentum tensor, it is tempting to generalize the recipe to quantum field theories, where the energy-momentum tensor is an operator, and to try to define a position operator following that CoM construction.We observe that, in generic situations, this proposal requires dealing with expectation values of the energy-momentum tensor, therefore, determining the CoM world line faces several additional complications ascompared to the situation involving the energy-momentum tensor of a classical N -particle system [52, 53].In this subsection we consider the manner in which the discussions in the previous sections impinge onquantum aspects of the matter description. Our focus is on the fact that the quantization process is designedto reproduce the symplectic structure with the replacement of Poisson brackets by quantum mechanicalcommutators. Thus, modifications of the Poisson brackets would have repercussions in the construction ofthe matter quantum theory. We now argue that the Poisson structure should be modified if we use thevariables characterizing the CoM of the extended objects.In reference [44] it was pointed out that, in special relativity, the observer-independent CoM has non-vanishing Poisson brackets among its different components. In fact, these brackets depend on the totalmomentum and spin of the extended object. This nontrivial expression stems from the fact that the rela-tivistic CoM weighs the particles’ position by their energies rather than their masses, which, appropriatelyallows massless particles to contribute to the location of the CoM. As the particles’ energies depend ontheir momenta, the CoM can be thought of as a function of the position and momenta of the constitutiveparticles of the extended object. Moreover, the resulting form of the Poisson brackets is remarkably similarto expressions found in rather different approaches [54]. It is also noteworthy that, in contrast to most workson noncommutative geometry, this noncommutativity arises naturally and it is fully compatible with Lorentzinvariance. The point is that the noncommutativity here is tied to the objects used in probing spacetimerather than to the underlying spacetime itself.Coming back to the Poisson brackets, at first sight it seems straightforward to generalize the calculationof reference [44] to GR for an extended object comprised by N free point-like particles by using the cor-responding Poisson brackets of the point-like particles. However, to actually perform such a calculation istechnically daunting and we can only sketch some qualitative conclusions. We can separate the issues relatedto this calculation into two classes: fundamental and technical. The fundamental difficulties stem from thefact that the Poisson brackets are related, by definition, with phase space functions. In turn, phase spaceis associated with a particular spacetime foliation (i.e., a notion of time) that allows one to describe theevolution of such phase space functions. In contrast, the construction of CoM in curved spacetime involvesquantities living in different leaves of a given foliation. However, if the evolution of each particle conformingthe extended object is known, we can translate this information to phase space variables to calculate Poissonbrackets.Moreover, it is unclear what quantities can be taken to be the canonical pairs. In this regard, it seemsnatural to use the foliation associated with the rest frame itself, put the origin of the coordinates at theactual CoM, and use a normal Riemann frame associated to that point. Then, the spatial components ofthe vectors whose exponential maps intersect the particles’ world lines and the spatial components of themomenta transported to this point are good candidates to be canonical pairs.Regarding the technical difficulties, note that when calculating Poisson brackets one needs to considerchanges of the state of the system since these brackets involve taking derivatives with respect to the canonicalvariables. However, working with the CoM recipe we are using, these changes, in turn, imply changes in thehypersurface on which the parallel transport is done. Furthermore, one has to be careful to ensure one isworking with truly independent variables, namely, if one’s coordinate frame is tied to the CoM, then, whenconsidering a variation in the state of the system corresponding to, say, one of the constitutive particles, thelocation of the CoM is modified, thus changing the vectors pointing to the other particles. Therefore, to14ctually perform the calculation one would need to use a particular fiducial foliation and a fixed coordinatesystem where all the degrees of freedom are independent. A choice would be to use the foliation associatedwith the CoM rest frame and to place the origin at the CoM world line, but allowing it to change (i.e., tobe displaced from the origin) when variations are considered. In this sense it seems interesting to perform aperturbative analysis similar to that presented in section 4.1 where only the first corrections due to curvatureare sought — this is still technically challenging since one has to make sure that the approximations areproperly implemented especially after taking derivatives, and it is thus left as the subject of a future work.We hope that it is still possible to use symmetry arguments and argue a priori what is the dominant curvaturecorrection.Nevertheless, even before calculating explicitly the Poisson brackets we are interested in, we can say that,as a correction to Pryce’s result, there will be terms proportional to the spacetime curvature. This, we expect,is due to the fact that, as we described above, when taking variations of the particles’ positions and momenta,both the CoM and the rest-frame hypersurface are changed, and, when comparing with the state before thevariation, one needs to bring the objects from those surfaces back onto the original foliation, resulting ina holonomy due to the parallel transport involved. The outcome is that the curvature of the underlyingspacetime will modify the Poisson structure with respect to that of the classical theory in nontrivial ways.Then some interesting questions arise. For instance, what is the full Poisson bracket algebra satisfied by theeffective degrees of freedom, and what type of action would give rise to such symplectic structures? We have considered some obstacles that will be faced when trying to recover classical geometry from afundamental theory of quantum gravity. We argued that, when recovering the classical spacetime notions ofGR, one would have to rely on the identification of matter states that could play the role of free particles and,moreover, that the effective geometry should be read off from the covariant CoM world lines characterizingsuch objects. This has lead us to examine the characteristics of the CoM of extended objects, which wasdone in a simpler setting where the background spacetime is assumed as given.We pointed out the relevance for such a program of the well-known fact that the CoM world lines donot follow the geodesics of the background spacetime. This, together with the requirement that we shouldhave a scheme where the equivalence principle is built in at the fundamental level into the very notion ofspacetime geometry, suggests that, in extracting the effective spacetime geometry, one should identify a priori the CoM world line of a free extended object as a geodesic of an effective geometry. However, as we haveshown, even this relatively simple step in the recovery of an effective geometry is filled with complications.In particular, we noted that there are difficulties in finding all the components of the effective connectionbecause several identical extended objects are needed to probe the given region in all spacetime directions.In addition, we have also described a method for obtaining an effective curvature tensor by considering therelative acceleration between the CoM world lines of two nearby extended objects, and found that the resultsdepend strongly on the detailed characteristics of such objects.On the other hand, the fact that the dominant part of many of the effects we considered depends on thetotal spin of the extended object, as already noted by Papapetrou, suggests that one should use extendedobjects with zero total spin to characterize the effective metric geodesics. However, when the dominantterm is insensitive to such a choice, higher multipolar momenta of the extended body would come into playin an essential way. One may then chose to use as probes, at a given scale, objects with vanishing, up toa certain order, higher multipolar momenta. This suggests that, at least in the context of the averagingproblem and the quest to obtain effective metrics at a given scale, one could use the following approach:given an underlying spacetime metric, an averaging scale, and an accuracy, consider extended objects withproper spatial extent in that scale, and determine what are the minimal number of energy-momentum tensormultipolar momenta that need to vanish to ensure that the center of mass does not depend on the probeswithin the corresponding subclass of test extended bodies. One may then use such world lines to definethe averaged spacetime metric at the corresponding scale. In practice, this would be a reasonable recipe toaddress the averaging problem in cosmology and astrophysics.Moreover, the difficulties in extracting a spacetime geometry with extended objects suggest that the taskwould be even harder when dealing with quantum probes. In fact, when considering the possible application15f the prescription involving extended bodies with vanishing multipolar momenta, as described above, severalproblems appear right away. First and foremost, the fact that we need to start with an underlying spacetimemetric to define the probe’s multipolar momenta, makes it evident that this prescription is unsuitable toconnect the regimes 1 and 2. And even in other regimes, we have to deal with the fact that, given the lackof associativity of the construction, it is impossible to have a recipe that works regardless of the initial scale.This suggests that one needs to rely on something like a smallest possible scale. However, since it is hard topoint out to a fundamental length scale without coming into conflict with Lorentz Invariance [55–57], thisapproach seems unsuitable at the fundamental level. In other words, the proposal seems inadequate in thequest to recover the GR spacetime from a quantum gravity theory. This raises the possibility that, in aregime where the quantum properties of the particles cannot be neglected, the geometrical language itselfmight cease to have meaning, since it may be impossible to properly define it.We have also shown that the covariant definition of the CoM is not associative in the sense that it cannotbe used to represent the parts into which one divides an extended object. That is, in order to calculate theCoM of the complete extended object, all the individual particles have to be identified as such, as neglectingtheir internal compositeness would lead to incompatible results. This can have practical consequences whenusing the CoM to characterize an extended object in regions with high curvature, but it is also interestingat the conceptual level because it suggests that the CoM is only well defined if fundamental particles exist.This nonassociativity of the CoM construction also has important consequences at the quantum level. Inparticular, it would cast serious doubts on the whole renormalization program for a quantum field theory thatcan be considered as effective and emergent from the underlying theory of quantum gravity since the notionthat one can integrate over certain degrees of freedom and then consider the effective variables obtained fromsuch integration as new fundamental variables, is at the basis of the renormalization group approach.We also considered briefly the issue of how to generalize the calculation of the Poisson brackets of thecomponents of the center of mass to curved spacetimes and how this indicates a problem when recoveringthe quantum characterization of the matter degrees of freedom. Interestingly, this problem seems to persistin situations where one believes that the underlying quantum nature of spacetime can be neglected. Theexpected nontrivial appearance of the spacetime curvature in the commutators of the free particles impliesthat there could be curvature effects on the quantum theory of “point particles” extracted from the underlyingtheory of quantum gravity. This indicates, in turn, that at the quantum level one can expect obstacles whenrecovering the GR prescription of minimal coupling. Effects characterizing a granular spacetime structurecould then show up [58–61]. Also, since the fundamental matter degrees of freedom cannot be taken asinfinitely localized, it is unclear how to operationally state the equivalence principle.There is still a lot of work to be done along the lines described in this manuscript. In particular, itis intriguing to see if it is possible to generalize the construction of the center of mass to situations wherematter is described in the realm of quantum field theory in curved spacetimes. It seems that this would bea required first step in order to define a covariant position operator which could be used to determine somecanonical world lines that could be used to define an effective geometry. Naturally, such a program wouldencounter similar difficulties to those we have described in this work. In fact, it can be expected that thosedifficulties would appear aggravated simply because, in that context, one needs to regularize and renormalizethe energy-momentum tensor. In view of the nonassociativity we have discussed above, those steps indicatethat we should question the meaning of an effective geometry which would be tied to the motion of realphysical matter probes. On the other hand, it is doubtful whether something that is not intimately relatedwith the behavior of such matter probes should be rightfully described as the physical spacetime geometry.After facing all these problems, we are tempted to conclude that the underlying quantum gravity theoryshould be such that, when extracting classical GR, nothing like what we presented here arises. In fact, itseems that such an underlying theory has to be of a rather different nature from the usual approaches and,in particular, it should not involve any fundamental spacetime discreetness. We do not want, at this point,to argue that the approach we have described in this paper represents the only possible path, or that theobstacles we have found are truly insurmountable. On the other hand, we don’t see an alternative approachthat both, bypasses the difficulties we have encountered, and still provides a satisfactory recovery of GRfrom quantum gravity. 16 cknowledgements
We acknowledge P. Aguilar for insightful comments and T. Koslowski and the late J. D. Bekenstein forpointing out some interesting references.
References [1] C. Rovelli, “Loop Quantum Gravity,”
Living Reviews in Relativity , vol. 1, no. 1, p. 1, 1998.[2] J. Engle, E. Livine, R. Pereira, and C. Rovelli, “LQG vertex with finite Immirzi parameter,”
NuclearPhysics B , vol. 799, no. 1, pp. 136 – 149, 2008.[3] R. D. Sorkin, “Causal Sets: Discrete Gravity,” in
Lectures on Quantum Gravity , pp. 305–327, Springer,2005.[4] J. Ambjørn, J. Jurkiewicz, and R. Loll, “Dynamically Triangulating Lorentzian Quantum Gravity,”
Nuclear Physics B , vol. 610, pp. 347 – 382, 2001.[5] H. Reichenbach,
The Philosophy of Space and Time . Dover, 1958.[6] A. Corichi, M. P. Ryan, and D. Sudarsky, “Quantum geometry as a relational construct,”
ModernPhysics Letters A , vol. 17, no. 09, pp. 555–567, 2002.[7] A. Papapetrou, “Spinning test-particles in General Relativity. I,”
Proceedings of the Royal Society ofLondon A: Mathematical, Physical and Engineering Sciences , vol. 209, no. 1097, pp. 248–258, 1951.[8] W. Beiglb¨ock, “The center-of-mass in Einsteins theory of gravitation,”
Communications in MathematicalPhysics , vol. 5, no. 2, pp. 106–130, 1967.[9] R. J. Van Den Hoogen, “Averaging space-time: Where do we go from here?,” in
Twelfth Marcel Gross-mann Meeting on General Relativity (A. H. Chamseddine, ed.), pp. 578–588, World Scientific, 2012.[10] M. Shirokov and I. Fisher, “Isotropic space with discrete gravitational-field sources. On the theory of anonhomogeneous isotropic universe,”
Soviet Astronomy , vol. 6, p. 699, 1963.[11] M. F. Shirokov and I. Z. Fisher, “Isotropic space with discrete gravitational-field sources. On the theoryof a nonhomogeneous isotropic universe,”
General Relativity and Gravitation , vol. 30, no. 9, pp. 1411–1427, 1998.[12] G. F. R. Ellis, “Relativistic cosmology: Its nature, aims and problems,” in
General Relativity andGravitation: Invited Papers and Discussion Reports of the 10th International Conference on GeneralRelativity and Gravitation, Padua, July 3–8, 1983 (B. Bertotti, F. De Felice, and A. Pascolini, eds.),pp. 215–288, Springer, 1984.[13] A. K. Raychaudhuri, “Theorem for nonrotating singularity-free universes,”
Physical Review Letters ,vol. 80, no. 4, p. 654, 1998.[14] T. Buchert, “On average properties of inhomogeneous fluids in general relativity: Dust cosmologies,”
General Relativity and Gravitation , vol. 32, no. 1, pp. 105–125, 2000.[15] R. A. Isaacson, “Gravitational radiation in the limit of high frequency. I. The linear approximation andgeometrical optics,”
Physical Review , vol. 166, no. 5, p. 1263, 1968.[16] R. A. Isaacson, “Gravitational radiation in the limit of high frequency. II. Nonlinear terms and theeffective stress tensor,”
Physical Review , vol. 166, no. 5, p. 1272, 1968.[17] T. Jacobson, “Thermodynamics of spacetime: The Einstein equation of state,”
Physical Review Letters ,vol. 75, no. 7, p. 1260, 1995. 1718] S. Liberati, M. Visser, and S. Weinfurtner, “Naturalness in an emergent analogue spacetime,”
PhysicalReview Letters , vol. 96, no. 15, p. 151301, 2006.[19] N. Seiberg, “Emergent spacetime,” in
The Quantum Structure Of Space And Time , pp. 163–213, WorldScientific, 2007.[20] D. Oriti, “A quantum field theory of simplicial geometry and the emergence of spacetime,”
Journal ofPhysics: Conference Series , vol. 67, no. 1, p. 012052, 2007.[21] I. Bars, S.-H. Chen, and G. Qu´elin, “Dual field theories in ( d −
1) + 1 emergent spacetimes from aunifying field theory in d + 2 spacetime,” Physical Review D , vol. 76, no. 6, p. 065016, 2007.[22] R. Loll, “The emergence of spacetime or quantum gravity on your desktop,”
Classical and quantumgravity , vol. 25, no. 11, p. 114006, 2008.[23] H. S. Yang, “Emergent spacetime and the origin of gravity,”
Journal of High Energy Physics , vol. 2009,no. 05, p. 012, 2009.[24] B.-L. Hu, “Emergent/quantum gravity: Macro/micro structures of spacetime,”
Journal of Physics:Conference Series , vol. 174, no. 1, p. 012015, 2009.[25] O. Dreyer, “Emergent Relativity,” in
Approaches to Quantum Gravity: Towards a New Understandingof Space, Time and Matter (D. Oriti, ed.), pp. 99–110, Cambridge University Press, 2009.[26] H. Steinacker, “Emergent geometry and gravity from matrix models: An introduction,”
Classical andQuantum Gravity , vol. 27, no. 13, p. 133001, 2010.[27] T. Padmanabhan, “Equipartition of energy in the horizon degrees of freedom and the emergence ofgravity,”
Modern Physics Letters A , vol. 25, no. 14, pp. 1129–1136, 2010.[28] S. El-Showk and K. Papadodimas, “Emergent spacetime and holographic CFTs,”
Journal of HighEnergy Physics , vol. 2012, no. 10, pp. 1–72, 2012.[29] R.-G. Cai, “Emergence of space and spacetime dynamics of Friedmann-Robertson-Walker universe,”
Journal of High Energy Physics , vol. 11, p. 016, 2012.[30] N. Huggett and C. W¨uthrich, “Emergent spacetime and empirical (in)coherence,”
Studies in Historyand Philosophy of Science Part B: Studies in History and Philosophy of Modern Physics , vol. 44, no. 3,pp. 276–285, 2013.[31] H. Salecker and E. Wigner, “Quantum limitations of the measurement of space-time distances,”
PhysicalReview , vol. 109, no. 2, p. 571, 1958.[32] S. Carlip, “Challenges for emergent gravity,”
Studies in History and Philosophy of Science Part B:Studies in History and Philosophy of Modern Physics , vol. 46, pp. 200–208, 2014.[33] C. M. Will, “The confrontation between General Relativity and experiment,”
Living Reviews in Rela-tivity , vol. 9, no. 1, p. 3, 2006.[34] E. Berti et al. , “Testing General Relativity with present and future astrophysical observations,”
Classicaland Quantum Gravity , vol. 32, no. 24, p. 243001, 2015.[35] S. F. Daniel et al. , “Testing General Relativity with current cosmological data,”
Physical Review D ,vol. 81, no. 12, p. 123508, 2010.[36] D. Schlippert et al. , “Quantum test of the universality of free fall,”
Physical Review Letters , vol. 112,no. 20, p. 203002, 2014.[37] C. D. Hoyle et al. , “Submillimeter test of the gravitational inverse-square law: A search for “large”extra dimensions,”
Physical Review Letters , vol. 86, no. 8, p. 1418, 2001.1838] C. F. Everitt et al. , “Gravity probe B: Final results of a space experiment to test General Relativity,”
Physical Review Letters , vol. 106, no. 22, p. 221101, 2011.[39] A. R. Liddle and D. H. Lyth,
Cosmological inflation and large-scale structure . Cambridge UniversityPress, 2000.[40] J. Barbour, T. Koslowski, and F. Mercati, “The solution to the problem of time in Shape Dynamics,”
Classical and Quantum Gravity , vol. 31, no. 15, p. 155001, 2014.[41] R. Brunetti, C. Dappiaggi, K. Fredenhagen, and J. Yngvason,
Advances in Algebraic Quantum FieldTheory . Springer, 2015.[42] W. Dixon, “A covariant multipole formalism for extended test bodies in General Relativity,”
Il NuovoCimento , vol. 34, no. 2, pp. 317–339, 1964.[43] J. Madore, “The equations of motion of an extended body in General Relativity,”
Annales de l’InstitutHenri Poincar´e Physique Th´eorique , vol. 11, no. 2, pp. 221–237, 1969.[44] M. H. L. Pryce, “The mass-centre in the restricted theory of relativity and its connexion with thequantum theory of elementary particles,”
Proceedings of the Royal Society of London A: Mathematical,Physical and Engineering Sciences , vol. 195, no. 1040, pp. 62–81, 1948.[45] Y. Bonder, “Quantum particles and an effective spacetime geometry,”
AIP Conference Proceedings ,vol. 1473, no. 1, pp. 178–183, 2012.[46] P. Aguilar, Y. Bonder, C. Chryssomalakos, and D. Sudarsky, “Operational geometry on de Sitter space-time,”
Modern Physics Letters A , vol. 27, no. 23, p. 1250130, 2012.[47] C. Chryssomalakos, H. Hernandez-Coronado, and E. Okon, “Center of mass in Special and GeneralRelativity and its role in an effective description of spacetime,”
Journal of Physics: Conference Series ,vol. 174, no. 1, p. 012026, 2009.[48] K. G. Wilson and J. Kogut, “The renormalization group and the ǫ -expansion,” Physics Reports , vol. 12,no. 2, pp. 75 – 199, 1974.[49] S. Matarrese and M. Pietroni, “Resumming cosmic perturbations,”
Journal of Cosmology and Astropar-ticle Physics , vol. 2007, no. 06, p. 026, 2007.[50] S. Matarrese and M. Pietroni, “Baryonic acoustic oscillations via the renormalization group,”
ModernPhysics Letters A , vol. 23, no. 01, pp. 25–32, 2008.[51] C. A. Mead, “Possible connection between gravitation and fundamental length,”
Phys. Rev. , vol. 135,pp. B849–B862, Aug 1964.[52] T. A. Roman, “Quantum stress-energy tensors and the weak energy condition,”
Physical Review D ,vol. 33, no. 12, p. 3526, 1986.[53] M. S. Morris, K. S. Thorne, and U. Yurtsever, “Wormholes, time machines, and the weak energycondition,”
Physical Review Letters , vol. 61, no. 13, p. 1446, 1988.[54] C. Chryssomalakos, H. Hern´andez, E. Okon, and P. V. Montejo, “Towards a first-principles approachto spacetime noncommutativity,”
Journal of Physics: Conference Series , vol. 68, no. 1, p. 012003, 2007.[55] J. Collins, A. Perez, D. Sudarsky, L. Urrutia, and H. Vucetich, “Lorentz invariance and quantum gravity:An additional fine-tuning problem?,”
Physical Review Letters , vol. 93, no. 19, p. 191301, 2004.[56] F. Dowker, J. Henson, and R. D. Sorkin, “Quantum gravity phenomenology, Lorentz invariance anddiscreteness,”
Modern Physics Letters A , vol. 19, no. 24, pp. 1829–1840, 2004.[57] D. Mattingly, “Modern tests of Lorentz invariance,”
Living Reviews in Relativity , vol. 8, no. 1, p. 5,2005. 1958] A. Corichi and D. Sudarsky, “Towards a new approach to quantum gravity phenomenology,”
Interna-tional Journal of Modern Physics D , vol. 14, no. 10, pp. 1685–1698, 2005.[59] Y. Bonder and D. Sudarsky, “Quantum gravity phenomenology without Lorentz invariance violation:A detailed proposal,”
Classical and Quantum Gravity , vol. 25, no. 10, p. 105017, 2008.[60] Y. Bonder and D. Sudarsky, “Unambiguous quantum gravity phenomenology respecting Lorentz sym-metry,”
Reports on Mathematical Physics , vol. 64, no. 1-2, pp. 169–184, 2009.[61] P. Aguilar, D. Sudarsky, and Y. Bonder, “Experimental search for a Lorentz invariant spacetime gran-ularity: Possibilities and bounds,”