betaFeedback

WorksPhysics and Astronomy

Gravitational observables from a post-Carrollian expansion

Abstract

Einstein gravity, with a few exceptions, only admits nonlocal observables. This feature is often attributed to diffeomorphism invariance, since a gravitational field can thereby locally be transformed away. We show this attribution to be incorrect by constructing infinitely many exact ultralocal Dirac observables in (velocity dominated aka electric) Carroll gravity. The latter describes Einstein gravity's ultrarelativistic limit, is fully diffeomorphism invariant, and of independent interest in a variety of contexts. Starting from these Carroll observables a systematic construction of Dirac observables for Einstein gravity is outlined. The resulting "post-Carrollian" expansion is organized by the number of spatial gradients or fractional inverse powers of Newton's constant, where each partial sum is local.

Introduction. The paucity of local observables in Einstein gravity hampers further progress in gravitational physics by limiting options to: compare different approaches [1, 2], isolate subsystems [2], or control the nonperturbative continuum limit of discretizations [3]. In spatially closed geometries no local observables exist in vacuum on the basis of NoGo theorems [4, 5, 6]. This feature is thought to persist for coupling to generic matter, and is witnessed by perturbative computations [2, 27, 28, 29, 13]. It is often attributed to diffeomorphism invariance: since a gravitational field can locally be transformed away, there is little to observe pointwise. The NoGo results sharpen this to the absence of local observables, i.e. ones built from the dynamical variables and their derivatives at any finite order. Conversely, the two results reported here should allow one to reframe these traditional obstructions: in fully diffeomorphism invariant Carroll gravity there exist infinitely many ultralocal observables. These enter as the zeroth order of a systematic construction of – seemingly impossible – local approximants to Einstein observables.

Carroll gravities describe Einstein gravity’s ultrarelativistic limit and together with their geometry are an active area of current research [22, 23, 24, 25]. There are several variants of which we focus here on the original, velocity dominated (aka electric) limit as introduced by Isham [17] and Henneaux [18]; see [19] for the terminology and [21, 20] for a first order formulation. For short, we shall refer to it as Velocity Dominated Carroll (VDC) gravity. For our purposes its second order Hamiltonian formulation is advantageous. The action minimally coupled to a scalar field reads

S=∫titfdt∫Σd3x{pab∂tqab+pφ∂tφ−νaHa−nH},S = \int_{t_i}^{t_f} dt \int_{\Sigma} d^3x \left\{p^{ab}\partial_t q_{ab} + p^\varphi\partial_t\varphi- \nu^a\mathcal{H}_a - n\mathcal{H}\right\},
Ha=−2∇bpab+pφ∂aφ,\mathcal{H}_a = -2\nabla_b p^b_a + p^\varphi\partial_a\varphi,
H=2κ(pbapab−12p2+14(pφ)2)+1κq(U(φ)+Λ).(1)\mathcal{H} = 2\kappa\left(p^a_b p^b_a - \frac{1}{2}p^2 + \frac{1}{4}(p^\varphi)^2\right) + \frac{1}{\kappa}q\left(U(\varphi)+\Lambda\right). \tag*{(1)}

Here Ha\mathcal{H}_a, a=1,2,3a = 1, 2, 3, is the diffeomorphism constraint and H\mathcal{H} is the Hamiltonian constraint. Both are built from the canonical pairs (pab,qab)(p^{ab}, q_{ab}), (pφ,φ)(p^\varphi, \varphi), which obey standard Poisson brackets. Indices are moved with qabq_{ab}, ∇a\nabla_a is the associated covariant derivative, and q=det⁡(qab)q = \det(q_{ab}), p=paap = p^a_a. The scalar field φ\varphi has been rendered dimensionless by the rescaling φ↦2κφ\varphi\mapsto\sqrt{2\kappa}\varphi and is selfinteracting via the potential U(φ)U(\varphi). The Lagrange multipliers (n,ν)(n,\nu) play the role of lapse and shift, and κ,Λ\kappa,\Lambda are the Newton and cosmological constants. Finally, the underlying manifold M=R×ΣM = \mathbb{R} \times\Sigma is foliated in terms of spacelike hypersurfaces, which we assume to be compact and closed, for simplicity. The action (1) differs from its counterpart SES_{\mathrm{E}} for Einstein gravity merely by the absence of a term q(−R(q)+qab∂aφ∂bφ)/(2κ)q(-R(q) + q^{ab}\partial_a\varphi\partial_b\varphi)/(2\kappa) in the Hamiltonian constraint, where R(q)R(q) is the Ricci scalar built from the spatial metric qabq_{ab}. The action SS can be obtained from SES_{\mathrm{E}} by S=lim⁡ℓ→∞SE∣ℓS = \lim_{\ell\to\infty} S_{\mathrm{E}|\ell}, where SE∣ℓS_{\mathrm{E}|\ell} has spacelike distances enhanced via qab∣ℓ=ℓ2qabq_{ab|\ell} = \ell^2 q_{ab}, p∣ℓab=ℓ−2pabp^{ab}_{|\ell} = \ell^{-2}p^{ab}, ℓ≫1\ell\gg1, combined with κ∣ℓ=ℓ3κ\kappa_{|\ell} = \ell^3\kappa [10]. Since κ∝GN/c3\kappa\propto G_{\mathrm{N}}/c^3, this can alternatively be viewed as a limit of large Newton constant GNG_{\mathrm{N}} or small speed of light cc.2

Bibliography

Dirac observables in VDC gravity. Given the textbook-advertised uniqueness properties of the Einstein-Hilbert action (from which the Hamiltonian version SES_{\mathrm{E}} derives) it is surprising that (1) is also fully diffeomorphism invariant. This is implicit in [18] and can be understood through the redundancy [10] of the so-called Ehresmann connection. Concretely, one has δϵHS=0\delta^H_\epsilon S=0 [26], where the Hamiltonian gauge variations δϵHX\delta^H_\epsilon X, X=(n,νa),(qab,pab),(φ,pφ)X=(n,\nu^a),(q_{ab},p^{ab}),(\varphi,p^\varphi), are parameterized by descriptors ϵ(t,x),ϵa(t,x)\epsilon(t,x),\epsilon^a(t,x) that are in one-to-one correspondence to linearized diffeomorphisms along the standard lines [14]. To match nn we take the temporal descriptor ϵ\epsilon to be a spatial −1-1 density that vanishes at t=ti,tft=t_i,t_f. The gauge variations other than δϵHn,δϵHνa\delta^H_\epsilon n,\delta^H_\epsilon\nu^a are canonically generated via δϵHF={F,∫d3x[Hϵ+Haϵa]}\delta^H_\epsilon F=\{F,\int\mathrm{d}^3x[\mathcal{H}\epsilon+\mathcal{H}_a\epsilon^a]\}. Related to its full diffeomorphism invariance the action (1) also propagates the same number of physical degrees of freedom as its Einstein counterpart.

Dirac observables for (1) are essentially gauge invariant quantities, δϵHO≈0\delta^H_\epsilon\mathcal{O}\approx0, where ‘≈\approx’ means equality modulo Ha=0=H\mathcal{H}_a=0=\mathcal{H}. For the solutions we find, the vanishing holds exactly; moreover they are built from (qab,pab),(φ,pφ)(q_{ab},p^{ab}),(\varphi,p^\varphi) only. Our Dirac observables for VDC gravity therefore satisfy

{Ha,O}=0={H,O},(2)\{\mathcal{H}_a,\mathcal{O}\}=0=\{\mathcal{H},\mathcal{O}\}, \tag*{(2)}

which is a strong version of the standard characterization [15]. Without further ado we introduce now the building blocks for the construction of such O\mathcal{O}’s:

πab:=qac(pcb−13qcbp),\pi^b_a:=q_{ac}\left(p^{cb}-\frac{1}{3}q^{cb}p\right),
E:=κ3p2−κ2(pφ)2−1κq(U(φ)+Λ).(3)E:=\frac{\kappa}{3}p^2-\frac{\kappa}{2}(p^\varphi)^2-\frac{1}{\kappa}q\left(U(\varphi)+\Lambda\right). \tag*{(3)}

These are spatial densities of weight +1+1, +2+2, respectively. Further H=2κpabpba−(2κp2/3+E)\mathcal{H}=2\kappa p^b_a p^a_b-(2\kappa p^2/3+E), so that we take E≥0E\geq0 in the following (in case one would want to solve H=0\mathcal{H}=0, which will not be done here). The interest in the quantities (3) lies in the fact that they are invariant under temporal gauge transformations: δϵHπab=Lϵ⃗πab\delta^H_\epsilon\pi^b_a=\mathcal{L}_{\vec{\epsilon}}\pi^b_a, δϵHE=Lϵ⃗E\delta^H_\epsilon E=\mathcal{L}_{\vec{\epsilon}}E, where Lϵ⃗\mathcal{L}_{\vec{\epsilon}} is the spatial Lie derivative, acting according to the tensor density type under consideration. This holds off-shell, i.e. without using any equations of motion. For a scalar +1+1 density FF one has in particular Lϵ⃗F=∂a(ϵaF)\mathcal{L}_{\vec{\epsilon}}F=\partial_a(\epsilon^aF), so that its spatial integral over a closed spacelike hypersurface Σ\Sigma vanishes. In addition to EE there are two other building blocks Q1,P1Q_1,P_1 locally built from q,p,φ,pφq,p,\varphi,p^\varphi. These also obey δϵHQ1=Lϵ⃗Q1\delta^H_\epsilon Q_1=\mathcal{L}_{\vec{\epsilon}}Q_1, δϵHP1=Lϵ⃗P1\delta^H_\epsilon P_1=\mathcal{L}_{\vec{\epsilon}}P_1, and form a canonical pair. However, since for nontrivial potential U(φ)U(\varphi) no explicit expressions are available for them, we refrain from using them here directly.

The strategy in the following is simply to form nonlinear combinations of (3) that are spatial +1+1 densities FF. These combinations will still be invariant under temporal gauge variations but obey δϵHF=∂a(ϵaF)\delta^H_\epsilon F=\partial_a(\epsilon^aF). Aiming at such combinations we start with the traces of the matrix powers of πa3\pi^3_a. By the Cayley-Hamilton theorem π3−(tr⁡π2)π/2−1det⁡π=0\pi^3-(\operatorname{tr}\pi^2)\pi/2-\mathbb{1}\det\pi=0 holds; so one may take tr⁡π2\operatorname{tr}\pi^2 and tr⁡π3\operatorname{tr}\pi^3 as independent traces. Augmenting EE, one can build ratios of homogeneous polynomials in the three density +1+1 variables (tr⁡π2)1/2(\operatorname{tr}\pi^2)^{1/2}, (tr⁡π3)1/3(\operatorname{tr}\pi^3)^{1/3}, E1/2E^{1/2}. Any ratio of total density weight +1 qualifies as an FF with the above properties. For illustration, we note as simple examples

F=(tr⁡π3)n/3(tr⁡π2)(n−1)/2,F=En/2(tr⁡π3)(n−1)/3,n∈N.(4)F = \frac{(\operatorname{tr}\pi^3)^{n/3}}{(\operatorname{tr}\pi^2)^{(n-1)/2}}, \qquad F = \frac{E^{n/2}}{(\operatorname{tr}\pi^3)^{(n-1)/3}}, \qquad n \in\mathbb{N}. \tag*{(4)}

Result: Velocity dominated Carroll gravity admits infinitely many strong, off-shell Dirac observables of the form OF=∫Σd3xFO_F = \int_\Sigma d^3x F, where FF is a rational function of net spatial density weight +1 in (tr⁡π2)1/2(\operatorname{tr}\pi^2)^{1/2}, (tr⁡π3)1/3(\operatorname{tr}\pi^3)^{1/3}, E1/2E^{1/2}.

These observables exist both in vacuum (where E1/2∝∣p∣E^{1/2} \propto|p|) and upon minimal coupling to a scalar field φ\varphi. They are ultralocal in the sense that FF is built from the phase space variables (qab,pab)(q_{ab}, p^{ab}), (φ,pφ)(\varphi, p^\varphi) at a spacetime point (t,x)(t,x). Additional expressions like (4) could be formed using the before-mentioned Q1Q_1, P1P_1 building blocks. As a step towards a general NoGo theorem Kuchar [8] performed a systematic search for Dirac observables in vacuum VDC gravity that are linear in the momenta, and found only ∫d3x p\int d^3x\,p. This is consistent with the above result in that all other admissible FF’s are nonlinear.

Additional symmetries and evasion of NoGo results. From the viewpoint of Einstein gravity the existence of these ultralocal Dirac observables is puzzling because there such a quantity cannot exist by the NoGo theorems [7, 4, 5, 6]. Somehow VDC gravity must evade the NoGo results, and we are led to inquire: how? It cannot be blamed on a breaking of diffeomorphism invariance because, as stressed before, VDC gravity is still fully diffeomorphism invariant. The key step in establishing the NoGo results is to classify all so-called generalized symmetries, i.e. operations that map a solution of the linearized field equations into another one. The result is [4, 5, 6] that all such operations are produced either by linearized diffeomorphisms or by rescalings. There are no hidden symmetries. Though initially shown in a Lagrangian formulation, the result can be transferred to phase space and then entails the absence of local Dirac observables in Einstein gravity [7]. In view of this line of reasoning one is led to suspect that VDC gravity evades the NoGo results by having hidden symmetries. Indeed, this turns out to be the case. Moreover, these additional symmetries are off-shell symmetries of the nonlinear action, not just symmetries of the linearized field equations.

To formulate the result, let FF be an ultralocal solution of {F,H}=0\{F,\mathcal{H}\}=0. Explicitly, F=F(qab,pab,φ,pφ)F=F(q_{ab},p^{ab},\varphi,p^\varphi) is a function of phase space variables but not their derivatives that solves

∂H∂pab∂F∂qab−∂H∂qab∂F∂pab+∂H∂pφ∂F∂φ−∂H∂φ∂F∂pφ=0.(5)\frac{\partial\mathcal{H}}{\partial p^{ab}}\frac{\partial F}{\partial q_{ab}}-\frac{\partial\mathcal{H}}{\partial q_{ab}}\frac{\partial F}{\partial p^{ab}}+\frac{\partial\mathcal{H}}{\partial p^\varphi}\frac{\partial F}{\partial\varphi}-\frac{\partial\mathcal{H}}{\partial\varphi}\frac{\partial F}{\partial p^\varphi}=0. \tag*{(5)}

If FF is also a spatial +1 density we introduce local gauge variations by

δFqab=∂F∂pab,δFpab=−∂F∂qab,\delta^F q_{ab}=\frac{\partial F}{\partial p^{ab}}, \qquad\delta^F p^{ab}=-\frac{\partial F}{\partial q_{ab}},
δFφ=∂F∂pφ,δFpφ=−∂F∂φ.(6)\delta^F\varphi=\frac{\partial F}{\partial p^\varphi}, \qquad\delta^F p^\varphi=-\frac{\partial F}{\partial\varphi}. \tag*{(6)}

and δFn=0\delta^{F}n=0, δFνa=0\delta^{F}\nu^{a}=0. Then δF\delta^{F} is a local symmetry of VDC gravity in the sense that

δFS=∫Σd3x(pab∂F∂pab+pφ∂F∂pφ−F)∣titf.(7)\delta^{F}S=\int_{\Sigma}d^{3}x\left(p^{ab}\frac{\partial F}{\partial p^{ab}}+p^{\varphi}\frac{\partial F}{\partial p^{\varphi}}-F\right)\bigg|_{t_i}^{t_f}. \tag*{(7)}

Since the action is preserved up to a boundary term the gauge variations preserve the Hamiltonian equations of motion. For the constraints H\mathcal{H}, Ha\mathcal{H}_{a} this holds exactly, while for the evolution equations it holds to linear order in δF\delta^{F}. Importantly, these symmetries are unrelated to diffeomorphisms and we shall refer to them as “additional local symmetries” of VDC gravity. A further analysis shows that they also give rise to generalized symmetries in the sense of [4, 5, 6], which thereby explains how the NoGo results for Einstein gravity are evaded in VDC gravity.

Of course (7) is only relevant if ultralocal solutions of (5) exist. But just such solutions have been found in the earlier result, for example (4). Ultralocal Dirac observables for VDC gravity therefore give rise to additional local symmetries and vice versa. The reason for spelling out (5) is that it highlights the key difference to Einstein gravity. Since H\mathcal{H} is ultralocal in VDC gravity, the equation (5) is for each spacetime point (t,x)(t,x) a linear first order partial differential equation. As such it is soluble by the method of characteristics. The characteristics are governed by a system of nonlinear ordinary differential equations, which turn out [34] to coincide, pointwise in xx, with the evolution equations for Bianchi I spacetimes in general relativity [16]. The pointwise counterpart of (5) thereby defines conservation laws for the Bianchi I evolution. From this viewpoint, one can recover the previously found FF’s via a suitable lift (functions of tt to functions of (t,x)(t,x)) of Bianchi I conservation laws.

From Carroll to Einstein. So far, we treated VDC gravity as a gravity theory in its own right with properties intriguingly different from Einstein gravity. One can alternatively regard Carroll gravity as the leading term in a suitable decomposition of an action principle for Einstein gravity. With different initial actions such decompositions have been presented in [35, 10, 19]. These can be used as the starting point for the construction or expansion of solutions in a tradition dating back to the analysis of Belinski-Khalatnikov-Lifshitz, see [35, 9]. In the present context, this raises the possibility of constructing observables for Einstein gravity in a new type of expansion. For the standard perturbative expansion the computation of observables for Einstein gravity has been pursued by several teams in different (also quantum) settings and encompasses weak field, small Newton constant, small clock time deviation, see e.g. [28, 29, 27, 13]. In view of the above results one is led to search for a complementary expansion of gravitational observables, where VDC gravity and its simple ultralocal observables are taken as the zeroth order. If well-posed, such an expansion could be viewed alternatively as one for large Newton constant or small speed of light (“post-Carrollian”), see after Eq. (1). It should complement resummation attempts for the weak coupling series [30, 31] and shed light on earlier notions of quasilocality [11, 12]. In a cosmological context it is an expansion away from the Big Bang. As a proof of principle a well-posed expansion of this type has been obtained for Gowdy cosmologies using integrable systems methodology [32].

In a Hamiltonian framework one will search for Dirac observables in Einstein gravity in the form OE=∫d3xFEO_{\mathrm{E}}=\int d^{3}xF_{\mathrm{E}}, with FEF_{\mathrm{E}} a spatial +1+1 density. Then only weak commutativity with the Hamiltonian constraint needs to be imposed, i.e. {HE,OE}∣HE=Ha=0=0\{\mathcal{H}_{\mathrm{E}},O_{\mathrm{E}}\}|_{\mathcal{H}_{\mathrm{E}}=\mathcal{H}_{a}=0}=0. Performing the scaling mentioned after Eq. (1) one has ℓ−3HE∣ℓ=H+κˇV\ell^{-3}\mathcal{H}_{\mathrm{E}}|_{\ell}=\mathcal{H}+\check{\kappa}\mathcal{V}, where

V=q2κ(−R(q)+qab∂aφ∂bφ),(8)\mathcal{V}=\frac{q}{2\kappa}\left(-R(q)+q^{ab}\partial_{a}\varphi\partial_{b}\varphi\right), \tag*{(8)}

is the gradient potential, and κˇ=ℓ−2\check{\kappa}=\ell^{-2} can also be read as (κˉ/κ)2/3∝κ−2/3∝cG2GN−2/3(\bar{\kappa}/\kappa)^{2/3}\propto\kappa^{-2/3}\propto c_{\mathrm{G}}^{2}G_{\mathrm{N}}^{-2/3} if κˉ=κℓ−3\bar{\kappa}=\kappa\ell^{-3} is held fixed. An expansion in powers of κˇ\check{\kappa} can therefore be viewed alternatively as one in the number of spatial gradients, in powers of c2c^{2}, or in powers of GN−2/3G_{\mathrm{N}}^{-2/3}. Taking one of the ultralocal Dirac observables O\mathcal{O} of VDC gravity as the zeroth order, O0=O\mathcal{O}_{0}=\mathcal{O}, we consider an ansatz of the form OE=∑j≥0κˇjOjO_{\mathrm{E}}=\sum_{j\geq0}\check{\kappa}^{j}\mathcal{O}_{j}. It is convenient to average HE\mathcal{H}_{\mathrm{E}} with the descriptor ϵ\epsilon of a temporal gauge variation. The defining equation for a Dirac observable for Einstein gravity then reads {HE(ϵ),OE}=HE(A(ϵ))+H⃗(B⃗(ϵ))\{\mathcal{H}_{\mathrm{E}}(\epsilon),O_{\mathrm{E}}\}=\mathcal{H}_{\mathrm{E}}(A(\epsilon))+\vec{\mathcal{H}}(\vec{B}(\epsilon)), for some coefficient functions A(ϵ)A(\epsilon), B⃗(ϵ)\vec{B}(\epsilon) that may depend on the phase space variables. In the present framework these will be expanded according to A(ϵ)=∑j≥0κˇjAj(ϵ)A(\epsilon)=\sum_{j\geq0}\check{\kappa}^{j}A_{j}(\epsilon), B⃗(ϵ)=∑j≥0κˇjB⃗j(ϵ)\vec{B}(\epsilon)=\sum_{j\geq0}\check{\kappa}^{j}\vec{B}_{j}(\epsilon), where each term may depend on the VDC phase space variables and their spatial derivatives. Inserted into the above defining equation, the lowest order reads {H(ϵ),O0}=H(A0(ϵ))+H⃗(B⃗0(ϵ))\{\mathcal{H}(\epsilon),\mathcal{O}_{0}\}=\mathcal{H}(A_{0}(\epsilon))+\vec{\mathcal{H}}(\vec{B}_{0}(\epsilon)), and the strong vanishing of the Poisson bracket for the VDC Dirac observables found earlier amounts to A0(ϵ)=0A_{0}(\epsilon)=0, B⃗0(ϵ)=0\vec{B}_{0}(\epsilon)=0. For the subleading terms one is led to a system of recursive equations

{H(ϵ),Oj}=−{V(ϵ),Oj−1}+V(Aj−1(ϵ))+H(Aj(ϵ))+H⃗(B⃗j(ϵ)),j≥1.(9)\{\mathcal{H}(\epsilon),\mathcal{O}_{j}\}=-\{\mathcal{V}(\epsilon),\mathcal{O}_{j-1}\}+\mathcal{V}(A_{j-1}(\epsilon))+\mathcal{H}(A_{j}(\epsilon))+\vec{\mathcal{H}}(\vec{B}_{j}(\epsilon)),\qquad j\geq1. \tag*{(9)}

At each order, the basic structure of (9) is that of a so-called “homological equation”, {H(ϵ),X}=Y\{\mathcal{H}(\epsilon),X\}=Y, with YY known. For perturbations around a finite dimensional integrable system several solution techniques are available, see [33] and the references therein. These techniques do, however, not readily carry over to VDC gravity.

Solution of the recursion. Instead, we developed a custom technique (partly drawing on [34]) that invokes the following ingredients: (a) A spatial −1-1 density TT called a “clock field”, characterized by {T,H(ϵ)}=ϵ\{T,\mathcal{H}(\epsilon)\}=\epsilon. (b) The solution of a single set of ordinary differential equations, the previously encountered characteristics of (5). (c) An explicit recursive formula for the coefficients Aj(ϵ)A_{j}(\epsilon), B⃗j(ϵ)\vec{B}_{j}(\epsilon), j≥1j\geq1. (d) A master solution formula relating Oj\mathcal{O}_{j} to Oj−1\mathcal{O}_{j-1}. We now first comment consecutively on each of these ingredients and then present the solution of (9) in terms of them.

(a) There are several ways to construct a clock field. One that works both in vacuum and for a generic interaction potential U(φ)U(\varphi) is as follows. Assume that the mixed tensor π ba\pi^{a}_{\ b} from (3) is diagonalizable over the reals. Let mab:=q−1/3qabm_{ab}:=q^{-1/3}q_{ab} be the unimodular metric canonically conjugate to πab=macπc b\pi^{ab}=m^{ac}\pi_{c}^{\ b}. Then

T=tr⁡[πLog⁡(mmˉ−1)]4κtr⁡π2,(10)T=\frac{\operatorname{tr}\left[\pi\operatorname{Log}(m\bar{m}^{-1})\right]}{4\kappa\operatorname{tr}\pi^{2}}, \tag*{(10)}

is a spatial −1-1 density and satisfies {T,H(ϵ)}=ϵ\{T,\mathcal{H}(\epsilon)\}=\epsilon. Here mˉab\bar{m}_{ab} is separately invariant, {mˉab,H(ϵ)}=0\{\bar{m}_{ab},\mathcal{H}(\epsilon)\}=0, and built from the components of πab\pi^b_a. The logarithm is that for symmetric 3×33 \times3 matrices. TT is a variant of a clock field in the sense of [13], but plays a different role here.

(b) The next ingredient is the characteristic flow of the Hamiltonian vector field associated with H\mathcal{H}. Writing z=(qab,φ,pab,pφ)z=(q_{ab},\varphi,p^{ab},p_\varphi) for a generic phase space point, the flow consists of solutions zf(z;τ)z^f(z;\tau) of the evolution equations ∂τzf={zf,H}\partial_\tau z^f=\{z^f,\mathcal{H}\}, zf(z;0)=zz^f(z;0)=z, pointwise in x∈Σx\in\Sigma, with respect to a (unphysical) flow parameter τ=τ(x)\tau=\tau(x) that is a spatial −1-1 density. Backpropagating a generic point zz with TT gives rise to coordinates zˉ:=zf(z,−T)\bar{z}:=z^f(z,-T) that are invariant under temporal gauge transformations, {zˉ,H}=0\{\bar{z},\mathcal{H}\}=0. These we use to define an interpolating flow z(s)=zf(z,(s−1)T(z))z^{(s)}=z^f(z,(s-1)T(z)), 0≤s≤10\leq s\leq1, that connects the original phase space point zz at s=1s=1 to its gauge invariant counterpart zˉ\bar{z} at s=0s=0. We write z(s)=(qab(s),φ(s),p(s)ab,p(s)φ)z^{(s)}=(q^{(s)}_{ab},\varphi^{(s)},p^{ab}_{(s)},p^{\varphi}_{(s)}) for the components and use F∣sF|_s to denote evaluation of a phase space function FF on z(s)z^{(s)}.

(c) For j=1j=1 the coefficients entering the solution of (9) read

A1(ϵ)=∫01ds s(1−s)Lγ⃗(α,ϵ),O0∣s,α=T,A_1(\epsilon)=\int_0^1 ds\,s(1-s)\mathcal{L}_{\vec{\gamma}(\alpha,\epsilon),\mathcal{O}_0}\big|_{s,\alpha=T},
B⃗1(ϵ)=−∫01ds s{γ⃗(α,ϵ),O0}∣s,α=T,(11)\vec{B}_1(\epsilon)=-\int_0^1 ds\,s\{\vec{\gamma}(\alpha,\epsilon),\mathcal{O}_0\}\big|_{s,\alpha=T}, \tag*{(11)}

where γ⃗(α,ϵ):=qab(α∂bϵ−ϵ∂bα)\vec{\gamma}(\alpha,\epsilon):=q^{ab}(\alpha\partial_b\epsilon-\epsilon\partial_b\alpha), and Lξ⃗T=ξa∂aT−∂aξaT\mathcal{L}_{\vec{\xi}}T=\xi^a\partial_aT-\partial_a\xi^aT is the Lie derivative of a spatial −1-1 density. A recursive formula, not needed here, then determines all other Aj(ϵ)A_j(\epsilon), B⃗j(ϵ)\vec{B}_j(\epsilon), j≥2j\geq2.

(d) The master formula relating the different orders of the expansion has the concise form

Oj=∫01ds{V(α),Oj−1}∣s,α=T,j≥1.(12)\mathcal{O}_j=\int_0^1 ds\{V(\alpha),\mathcal{O}_{j-1}\}\big|_{s,\alpha=T},\qquad j\geq1. \tag*{(12)}

Technically, the key assertion is that for all j≥1j\geq1 there exist spatial −1-1 densities Aj(ϵ)A_j(\epsilon) and vectors B⃗j(ϵ)\vec{B}_j(\epsilon) (determined by (c)), such that (12) solves (9). Since VV contains two spatial derivatives, it entails that the Oj\mathcal{O}_j are local quantities. Specifically they are of the form

Oj=∫d3x Gj(x),(13)\mathcal{O}_j=\int d^3x\,G_j(x), \tag*{(13)}

where GjG_j is a differential polynomial in the phase space variables (and a spatial +1+1 density) containing at most 2j2j spatial derivatives.

Result. Let TT be a clock field as in (a), let z(s)z^{(s)} be the interpolating flow from (b), and take the Oj\mathcal{O}_j as computed from (d). Then, the partial sums of the series ∑j≥0κ~jOj\sum_{j\geq0}\tilde{\kappa}^j\mathcal{O}_j provide systematic local approximants to Dirac observables O~\widetilde{\mathcal{O}} in Einstein gravity. The leading-order term O0\mathcal{O}_0 is one of the ultralocal Dirac observables for VDC gravity described earlier. The corrections are organized by the number of spatial gradients or by powers of κ~∝c2GN−2/3\tilde{\kappa}\propto c^2G_{\mathrm{N}}^{-2/3}.

This is very different from the usual weak coupling (in powers of GN1/2G_{\mathrm{N}}^{1/2}) or post-Newtonian (in powers of 1/c1/c) expansions, where all orders are highly nonlocal. The result initiates a new approach to the intrinsic physics of diffeomorphism invariant systems. It may be hoped that the obstructions mentioned in the introductory paragraph can be alleviated thereby.

The first post-Carrollian correction

Here we derive the formulas providing the first post-Carrollian correction for Dirac observables. Explicitly, we claim that given an ultralocal VDC Dirac observable O0\mathcal{O}_0, and a clock field TT, the expression

O1=∫01ds {V(α),O0}∣s,α=T,(14)\mathcal{O}_1=\int_0^1 ds\,\{\mathcal{V}(\alpha),\mathcal{O}_0\}\big|_{s,\alpha=T}, \tag*{(14)}

solves {H(ϵ),O1}=−{V(ϵ),O0}+H(A1(ϵ))+H⃗(B⃗1(ϵ))\{\mathcal{H}(\epsilon),\mathcal{O}_1\}=-\{\mathcal{V}(\epsilon),\mathcal{O}_0\}+\mathcal{H}(A_1(\epsilon))+\vec{\mathcal{H}}(\vec{B}_1(\epsilon)), with A1(ϵ)A_1(\epsilon), B⃗1(ϵ)\vec{B}_1(\epsilon) from (11). We prepare the following

Lemma A.1. Let FF be a spatial differential polynomial in the basic phase space variables and a spatial +2+2 density. Write F(α)=∫d3x F(x)α(x)F(\alpha)=\int d^3x\,F(x)\alpha(x) for a spatial −1-1 density α\alpha. For a given clock field TT define OF:=∫01dsF∣s(T)\mathcal{O}_F:=\int_0^1 ds F|_s(T). Then OF\mathcal{O}_F satisfies

δϵ,0HOF=F(ϵ)+∫01ds s[δϵ,0HF(α)−δα,0HF(ϵ)]∣s,α=T.(15)\delta^H_{\epsilon,0}\mathcal{O}_F=F(\epsilon)+\int_0^1 ds\,s\left[\delta^H_{\epsilon,0}F(\alpha)-\delta^H_{\alpha,0}F(\epsilon)\right]\big|_{s,\alpha=T}. \tag*{(15)}

In particular, $[\delta^H_{\epsilon,0}\mathcal{O}_F-F(\epsilon)]|_{\epsilon=T}=0.

Proof. The derivation is based on the interplay between the gauge variation and the ss derivatives of flowed quantities. From the definitions of the interpolating flow z(s)=zf(z,(s−1)T(z))z^{(s)}=z^f(z,(s-1)T(z)), 0≤s≤10\le s\le1, with ∂τzf={zf,H}\partial_\tau z^f=\{z^f,\mathcal{H}\}, zf(z;0)=zz^f(z;0)=z, one obtains the relations

s[δϵ,0HF(α)]∣s=δϵ,0H[F(α)∣s],dds[F(ϵ)∣s]=[δα,0HF(ϵ)]∣s,α=T.(16)s\left[\delta^H_{\epsilon,0}F(\alpha)\right]\big|_s=\delta^H_{\epsilon,0}\left[F(\alpha)\big|_s\right],\qquad \frac{d}{ds}\left[F(\epsilon)\big|_s\right]=\left[\delta^H_{\alpha,0}F(\epsilon)\right]\big|_{s,\alpha=T}. \tag*{(16)}

Then

δϵ,0HOF=∫01ds(F(ϵ)∣s+δϵ,0H[F(α)∣s]∣α=T)=∫01ds(F(ϵ)∣s+s[δα,0HF(ϵ)]∣s,α=T+s[δϵ,0HF(α)−δα,0HF(ϵ)]∣s,α=T)=∫01ds dds(sF(ϵ)∣s)+∫01ds [δϵ,0HF(α)−δα,0HF(ϵ)]∣s,α=T.(17)\begin{aligned} \delta^H_{\epsilon,0}\mathcal{O}_F &=\int_0^1 ds\left(F(\epsilon)\big|_s+\delta^H_{\epsilon,0}\left[F(\alpha)\big|_s\right]\big|_{\alpha=T}\right)\\ &=\int_0^1 ds\left(F(\epsilon)\big|_s+s\left[\delta^H_{\alpha,0}F(\epsilon)\right]\big|_{s,\alpha=T} +s\left[\delta^H_{\epsilon,0}F(\alpha)-\delta^H_{\alpha,0}F(\epsilon)\right]\big|_{s,\alpha=T}\right) \\ &= \int_0^1 ds\,\frac{d}{ds}\left(sF(\epsilon)\big|_s\right)+\int_0^1 ds\,\left[\delta^{\mathrm H}_{\epsilon,0}F(\alpha)-\delta^{\mathrm H}_{\alpha,0}F(\epsilon)\right]\big|_{s,\alpha=T}. \tag*{(17)} \end{aligned}

The first term evaluates to sF(ϵ)∣s=1−sF(ϵ)∣s=0=F(ϵ)sF(\epsilon)\big|_{s=1}-sF(\epsilon)\big|_{s=0}=F(\epsilon), where z(s=1)=zz^{(s=1)}=z (the original phase space point) and z(s=0)=zˉz^{(s=0)}=\bar z (its δϵ,0H\delta^{\mathrm H}_{\epsilon,0} invariant counterpart) enter. ■

One application of the Lemma is to the j=1j=1 instance of (9). Since A0(ϵ)=0A_0(\epsilon)=0 the assertion reads δϵ,0HO1=F1(ϵ)−H(A1(ϵ))−H⃗(B⃗1(ϵ))\delta^{\mathrm H}_{\epsilon,0}\mathcal{O}_1=F_1(\epsilon)-\mathcal{H}(A_1(\epsilon))-\vec{\mathcal{H}}(\vec{B}_1(\epsilon)), where F1(ϵ)={V(ϵ),O0}F_1(\epsilon)=\{\mathcal{V}(\epsilon),\mathcal{O}_0\} plays the role of FF in the Lemma. In order to evaluate the integral term on the right-hand side of (15) we note δϵ,0HF1(α)−δα,0HF1(ϵ)={δϵ,0HV(α)−δα,0HV(ϵ),O0}\delta^{\mathrm H}_{\epsilon,0}F_1(\alpha)-\delta^{\mathrm H}_{\alpha,0}F_1(\epsilon)=\{\delta^{\mathrm H}_{\epsilon,0}\mathcal{V}(\alpha)-\delta^{\mathrm H}_{\alpha,0}\mathcal{V}(\epsilon),\mathcal{O}_0\}. It can be simplified using

δϵ,0HV(α)−δα,0HV(ϵ)=H⃗(γ⃗(α,ϵ)),γa(α,ϵ)=qqab(α∂bϵ−ϵ∂bα),(18)\delta^{\mathrm H}_{\epsilon,0}\mathcal{V}(\alpha)-\delta^{\mathrm H}_{\alpha,0}\mathcal{V}(\epsilon)=\vec{\mathcal{H}}(\vec{\gamma}(\alpha,\epsilon)),\qquad\gamma^a(\alpha,\epsilon)=q q^{ab}(\alpha\partial_b\epsilon-\epsilon\partial_b\alpha), \tag*{(18)}

This can be verified either by direct computation or by matching terms of order κ~\tilde{\kappa} in the constraint algebra of Einstein gravity. So far,

δϵ,0HF1(α)−δα,0HF1(ϵ)=H⃗({γ⃗(α,ϵ),O0}).(19)\delta^{\mathrm H}_{\epsilon,0}F_1(\alpha)-\delta^{\mathrm H}_{\alpha,0}F_1(\epsilon)=\vec{\mathcal{H}}\left(\{\vec{\gamma}(\alpha,\epsilon),\mathcal{O}_0\}\right). \tag*{(19)}

Upon evaluation on the interpolating flow we use

H⃗∣s(ξ⃗)=H⃗(ξ⃗)+(s−1)H(Lξ⃗T),(20)\vec{\mathcal{H}}\big|_s(\vec{\xi})=\vec{\mathcal{H}}(\vec{\xi})+(s-1)\mathcal{H}(\mathcal{L}_{\vec{\xi}}T), \tag*{(20)}

where Lξ⃗T=ξa∂aT−∂aξaT\mathcal{L}_{\vec{\xi}}T=\xi^a\partial_aT-\partial_a\xi^aT is the Lie derivative of a spatial −1-1 density. This follows from the d/dsd/ds differentiated version, which in turn reflects the constraint algebra of VDC gravity, i.e. {H(ϵ1),H(ϵ2)}=0\{\mathcal{H}(\epsilon_1),\mathcal{H}(\epsilon_2)\}=0, {H⃗(ξ⃗),H(ϵ)}=H(Lξ⃗ϵ)\{\vec{\mathcal{H}}(\vec{\xi}),\mathcal{H}(\epsilon)\}=\mathcal{H}(\mathcal{L}_{\vec{\xi}}\epsilon). Combining (19) with (20) the integral term in the Lemma evaluates to

∫01ds s[δϵ,0HF1(α)−δα,0HF1(ϵ)]∣s,α=T=−H(A1(ϵ))−H⃗(B⃗1(ϵ)),(21)\int_0^1 ds\,s\left[\delta^{\mathrm H}_{\epsilon,0}F_1(\alpha)-\delta^{\mathrm H}_{\alpha,0}F_1(\epsilon)\right]\big|_{s,\alpha=T}=-\mathcal{H}(A_1(\epsilon))-\vec{\mathcal{H}}(\vec{B}_1(\epsilon)), \tag*{(21)}

with A1(ϵ)A_1(\epsilon), B⃗1(ϵ)\vec{B}_1(\epsilon) as in (11). In summary, (14) indeed solves the j=1j=1 instance of (9), with the specific A1(ϵ)A_1(\epsilon), B⃗1(ϵ)\vec{B}_1(\epsilon) given.

Finally, we check that O1\mathcal{O}_1 is a local functional on phase space, specifically one of the form asserted in (13), for j=1j=1. Mostly, this follows from the fact that V(α)\mathcal{V}(\alpha) has regular functional derivatives given by

δV(α)δqab=αVab−q2κ(∇a∇b−qab∇c∇c)α,\frac{\delta\mathcal{V}(\alpha)}{\delta q_{ab}}=\alpha V^{ab}-\frac{q}{2\kappa}\left(\nabla^a\nabla^b-q^{ab}\nabla^c\nabla_c\right)\alpha,
δV(α)δϕ=−1κqqab∇b(α∇aϕ),(22)\frac{\delta\mathcal{V}(\alpha)}{\delta\phi}=-\frac{1}{\kappa}q q^{ab}\nabla_b(\alpha\nabla_a\phi), \tag*{(22)}

where we defined

Vab:=q2κ(Rab−qabR)−q2κ(qadqbc−qabqcd)∂cϕ∂dϕ.(23)V^{ab}:=\frac{q}{2\kappa}\left(R^{ab}-q^{ab}R\right)-\frac{q}{2\kappa}\left(q^{ad}q^{bc}-q^{ab}q^{cd}\right)\partial_c\phi\partial_d\phi. \tag*{(23)}

By (14) we need to evaluate {V(α),O0}\{\mathcal{V}(\alpha),\mathcal{O}_0\}. Since O0=∫d3x F0(x)\mathcal{O}_0 = \int d^3x\,F_0(x) for some ultralocal F0F_0 its functional derivatives reduce to pointwise ones, δO0/δqab(x)=∂F0(x)/∂qab(x)\delta\mathcal{O}_0/\delta q_{ab}(x) = \partial F_0(x)/\partial q_{ab}(x), etc.. Further, as V\mathcal{V} does not depend on the momenta, the functional derivatives of V\mathcal{V} with respect to the momenta vanish. Together

{V(α),O0}=∫d3x(δV(α)δqab(x)∂F0(x)∂pab(x)+δV(α)δϕ(x)∂F0(x)∂pϕ(x)),(24)\{\mathcal{V}(\alpha),\mathcal{O}_0\} = \int d^3x \left(\frac{\delta\mathcal{V}(\alpha)}{\delta q_{ab}(x)}\frac{\partial F_0(x)}{\partial p^{ab}(x)}+\frac{\delta\mathcal{V}(\alpha)}{\delta\phi(x)}\frac{\partial F_0(x)}{\partial p^\phi(x)}\right), \tag*{(24)}

is a differential polynomial of order 2 in the phase space variables. Since T(x)T(x) is an ultralocal −1-1 density on phase space, evaluation of (24) on α=T\alpha=T will not change this. Finally, the interpolating flow z(s)z^{(s)} arises from the pointwise in x∈Σx \in\Sigma integrated characteristic flow ∂τzf={zf,H}\partial_\tau z^f = \{z^f,\mathcal{H}\}. Hence

G1(x)=∫01ds(δV(α)δqab(x)∂F0(x)∂pab(x)+δV(α)δϕ(x)∂F0(x)∂pϕ(x))∣s,α=T,(25)G_1(x) = \int_0^1 ds \left(\frac{\delta\mathcal{V}(\alpha)}{\delta q_{ab}(x)}\frac{\partial F_0(x)}{\partial p^{ab}(x)}+\frac{\delta\mathcal{V}(\alpha)}{\delta\phi(x)}\frac{\partial F_0(x)}{\partial p^\phi(x)}\right)\bigg|_{s,\alpha=T}, \tag*{(25)}

is a spatial differential polynomial in the phase space variables of degree 2.

The generalization to higher orders is considerably more complicated, but the above presentation of the j=1j=1 case gives a good flavor of the reasoning involved.

References

  1. [1]J. Ambjørn, J. Gizbert-Studnicki, A. Goerlich, and D. Nemeth, Is lattice quantum gravity asymptotically safe? Making contact between causal dynamical triangulations and the functional renormalization group, Phys. Rev. D110 (2024) 126006.
  2. [2]W. Donnelly and S. Giddings, Observables, gravitational dressing, and obstructions to locality and subsystems, Phys. Rev. D94 (2016) 104038.DOI
  3. [3]B. Dittrich and P. Höhn, From covariant to canonical formulations of discrete gravity, Class. Quant. Grav. 27 (2010) 155001.DOI
  4. [4]I. Anderson and C. Torre, Symmetries of the Einstein equations, Phys. Rev. Lett. 70 (1993) 3525.
  5. [5]I. Anderson and C. Torre, Classification of local generalized symmetries for the vacuum Einstein equations, Commun. Math. Phys. 176 (1996) 479.DOI
  6. [6]L. Marchildon, Lie symmetries of Einstein’s vacuum equations in N dimensions, J. Nonlin. Math. Phys. 5 (1998) 68.DOI
  7. [7]C. Torre, Gravitational observables and local symmetries, Phys. Rev. D48 (1993) R2373.DOI
  8. [8]K. Kuchar, General relativity: dynamics without symmetry, J. Math. Phys. 22 (1981) 2640.DOI
  9. [9]H. Ringström, Cosmology, the Big Bang, and the BKL conjecture, Comptes Rendus. Mécanique, 353 (2025) 53.DOI
  10. [10]M. Niedermaier, Nonstandard action of diffeomorphisms and gravity’s anti-Newtonian limit, Symmetry 12 (2020) 752.DOI
  11. [11]I. Khavkine, Local and gauge invariant observables in gravity, Class. Quant. Grav. 32 (2015) 185019.DOI
  12. [12]C. Goeller, P. Höhn, and J. Kirlin, Diffeomorphism-invariant observables and dynamical frames in gravity: reconciling bulk locality with general covariance, arXiv:2206.01193.arxiv.org/abs/2206.01193
  13. [13]B. Dittrich and J. Tambornino, A perturbative approach to Dirac observables and their spacetime algebra, Class. Quant. Grav. 24 (2007) 757.DOI
  14. [14]J. Pons, D. Salisbury, and K. Sundermeyer, Revisiting observables in generally covariant theories in the light of gauge fixing methods, Phys. Rev. D80 (2009) 084015.DOI
  15. [15]M. Henneaux and C. Teitelboim, Quantization of gauge systems, Princeton UP, 1992.
  16. [16]G. Ellis, R. Maartens, M. MacCallum, Relativistic Cosmology, Cambridge UP, 2012.DOI
  17. [17]C. Isham, Some quantum field theory aspects of the superspace quantization of general relativity, Proc. R. Soc. A351 (1976) 209.DOI
  18. [18]M. Henneaux, Geometry of zero signature spacetimes, Bull. Soc. Math. Belg. 31, 47 (1979).
  19. [19]M. Henneaux and P. Salgado-Rebolledo, Carroll contractions of Lorentz invariant theories, JHEP 11 (2021) 180.DOI
  20. [20]S. Pekar, A. Perez, and P. Salgado-Rebolledo, Cartan-like formulation of electric Car- rollian gravity, JHEP 09 (2024) 059.DOI
  21. [21]J. Hartong, Gauging the Carroll algebra and ultrarelativistic gravity, JHEP 08 (2015) 69.DOI
  22. [22]E. Bergshoeff, J. Figueroa-O’Farrill, and J. Gomis, A non-lorentzian primer, SciPost Phys. Lect. Notes 69 (2023).
  23. [23]L. Donnay, A. Fiorucci, Y. Herfray, and R. Ruzziconi, Carrollian perspective on Celestial Holography, Phys. Rev. Lett. 129 (2022) 071602.DOI
  24. [24]R. Ruzziconi, Carrollian Physics and Holography, Phys. Rept. 1182 (2026) 1-87.
  25. [25]L. Ciambelli, R. Leigh, C. Marteau, and M. Petropoulos, Carroll structures, Null geom- etry, and conformal isometries, Phys. Rev. D100 (2019) 046010.DOI
  26. [26]M. Niedermaier, The gauge structure of strong coupling gravity, Class. Quant. Grav. 32 (2015) 015007.DOI
  27. [27]M. Fröb, C. Rein, and R. Verch, Graviton corrections to the Newtonian potential using invariant observables, JHEP 01 (2022) 180.DOI
  28. [28]S. Giddings and J. Perkins, Perturbative quantum evolution of the gravitational state and dressing in general backgrounds, Phys. Rev. D110 (2024) 026012.DOI
  29. [29]C. Cheung, A. Sivaramakrishnan, J. Wilson-Gerow, and Lihang Zhou, Perturbatively dressed observables, Phys. Rev. D114 (2026) 086013.DOI
  30. [30]P. Damgaard and K. Lee, Schwarzschild Black Hole from Perturbation Theory to all orders, Phys. Rev. Lett. 132 (2024) 251603.DOI
  31. [31]M. Saavedra, The classical gravitational impulse at high energies, arXiv:2605.18544.
  32. [32]M. Niedermaier and M. Sedighi Jafari, Dirac observables for Gowdy cosmologies regular at the Big Bang, SIGMA 22 (2026) 097.DOI
  33. [33]A. Nikolaev, Kato perturbative expansion in classical mechanics and an explicit expres- sion for the Deprit generator, Theor. Math. Phys. 182 (2015) 407.DOI
  34. [34]M. Niedermaier, Canonical Trivialization of gravitational gradients, Class. Quant. Grav. 34 (2017) 115013.DOI
  35. [35]D. Hansen, N. Obers, G. Oling, and B. Sogaard, Carroll expansion of general relativity, SciPost Phys. 13 (2022) 055.DOI

Paper details

vlnc:W_pmc4mk8x
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents