Lattice-Regularized Quantum Gravity

• Lattice-regularized quantum gravity is a family of nonperturbative methods that discretize spacetime using simplicial geometries, causal triangulations, and lattice gauge variables.
• It employs various frameworks such as Regge calculus, EDT, and CDT to reformulate the Einstein-Hilbert action and explore UV fixed points, semiclassical behavior, and continuum limits.
• These approaches enable the study of scaling properties, geometric observables, and phase transitions, offering practical insights into asymptotic safety and emergent four-dimensional spacetime.

Lattice-regularized quantum gravity denotes a family of nonperturbative constructions in which the formal path integral over geometries is replaced by a discrete sum or integral over simplicial geometries, causal triangulations, lattice gauge variables, or fermionic and bosonic lattice fields. In two dimensions, such lattice theories indeed provide regularizations of continuum well defined quantum gravity theories; in four dimensions, they are used to search for a non-perturbative UV fixed point and a continuum limit in the spirit of asymptotic safety (Ambjorn, 2022).

1. Conceptual basis

The common continuum starting point is the Einstein-Hilbert action. In the Lorentzian form used in CDT, it is written as

$$S[g_{\mu\nu};G,\Lambda] = \frac{1}{16 \pi G}\int_M d^4 x \sqrt{-g(x)} \, \big( R(x) - 2 \Lambda\big),$$

with the corresponding formal path integral over metrics modulo diffeomorphisms. The motivation for lattice regularization is that ordinary perturbative quantization around a fixed background is nonrenormalizable, while gravity differs from nongravitational lattice field theory because spacetime geometry itself is dynamical. For this reason, the lattice cannot be a rigid hypercubic scaffold carrying gravitational variables; the lattice itself must represent the geometry (Ambjørn et al., 7 Apr 2026).

In simplicial formulations this is achieved by replacing smooth manifolds with piecewise flat ones. Regge’s formula expresses curvature as concentrated on subsimplices of codimension two, while the volume term becomes a sum over simplices. For equilateral triangulations this yields especially simple actions, such as $$S_T(\kappa_2,\kappa_4) = - \kappa_2 N_2(T) + \kappa_4 N_4(T)$$ in four-dimensional EDT, where the path integral becomes a weighted counting problem over triangulations rather than a functional integral over continuum metrics (Ambjorn, 2022).

A second conceptual distinction concerns signature. Euclidean Dynamical Triangulations (EDT) sums over Euclidean triangulations from the outset, whereas Causal Dynamical Triangulations (CDT) begins with Lorentzian geometries carrying a preferred proper-time foliation and only afterwards performs a well-defined Wick rotation. The CDT literature treats this causal restriction as physically important rather than merely technical, because it changes the entropy of configurations and the resulting universality class (Loll, 2019).

2. Principal lattice frameworks

The major frameworks differ in their microscopic variables, the rôle assigned to causality, and the mechanism by which continuum geometry is expected to emerge.

Framework	Fundamental variables	Characteristic claim
Regge calculus	Squared edge lengths $l_{ij}^2$ on a simplicial lattice	A strong-coupling/anti-screening phase with a nontrivial ultraviolet fixed point at finite Newton coupling
EDT	Equilateral 4-simplices; geometry encoded in connectivity	A tuned measure term can recover semiclassical four-dimensional geometries
CDT	Causal Lorentzian triangulations built from $(4,1)$, $(1,4)$, $(3,2)$, and $(2,3)$ simplices	An extended de Sitter phase $C_{\rm dS}$, a nonperturbative Wick rotation, and second-order transitions

In the Regge–Wheeler simplicial formulation, the lattice path integral is

$Z_L \;=\; \int d\mu[l^2]\, e^{-I[l^2]},$

with squared edge lengths $l_{ij}^2$ as the fundamental degrees of freedom. The induced metric inside each simplex is reconstructed from edge lengths via

$$S_T(\kappa_2,\kappa_4) = - \kappa_2 N_2(T) + \kappa_4 N_4(T)$$0

and curvature is localized on hinges $$S_T(\kappa_2,\kappa_4) = - \kappa_2 N_2(T) + \kappa_4 N_4(T)$$1, which in four dimensions are triangles. The pure Regge–Einstein action,

$$S_T(\kappa_2,\kappa_4) = - \kappa_2 N_2(T) + \kappa_4 N_4(T)$$2

was used to extract scaling exponents and invariant correlation functions in four dimensions (Hamber, 2015).

In EDT, smooth geometries are replaced by triangulated four-manifolds built from equilateral 4-simplices of fixed edge length $$S_T(\kappa_2,\kappa_4) = - \kappa_2 N_2(T) + \kappa_4 N_4(T)$$3, and geometry is encoded in the connectivity rather than in fluctuating link lengths. The partition function takes the form

$$S_T(\kappa_2,\kappa_4) = - \kappa_2 N_2(T) + \kappa_4 N_4(T)$$4

where $$S_T(\kappa_2,\kappa_4) = - \kappa_2 N_2(T) + \kappa_4 N_4(T)$$5 is the order of a triangle and $$S_T(\kappa_2,\kappa_4) = - \kappa_2 N_2(T) + \kappa_4 N_4(T)$$6 couples a nontrivial local measure term. After the Regge reduction, the action is written in the simulation form

$$S_T(\kappa_2,\kappa_4) = - \kappa_2 N_2(T) + \kappa_4 N_4(T)$$7

A central claim of the 2016 EDT analysis is that fine-tuning the measure coupling $$S_T(\kappa_2,\kappa_4) = - \kappa_2 N_2(T) + \kappa_4 N_4(T)$$8 is necessary because the lattice regulator breaks continuum general coordinate invariance; with that tuning, the model exhibits semiclassical four-dimensional behavior and no obvious barrier to a continuum limit (Laiho et al., 2016).

CDT regularizes the Lorentzian gravitational path integral by summing over causal triangulations built from flat Minkowskian four-simplices. The allowed four-dimensional building blocks are of types $$S_T(\kappa_2,\kappa_4) = - \kappa_2 N_2(T) + \kappa_4 N_4(T)$$9, $l_{ij}^2$0, $l_{ij}^2$1, and $l_{ij}^2$2, with spacelike edges of squared length $l_{ij}^2$3 and timelike edges of squared length $l_{ij}^2$4. The Lorentzian path integral

$l_{ij}^2$5

admits a nonperturbative Wick rotation $l_{ij}^2$6, leading to a Euclidean action used in Monte Carlo simulations. In practice this is written as

$l_{ij}^2$7

with $l_{ij}^2$8 as bare lattice couplings (Ambjørn et al., 7 Apr 2026).

3. Continuum limit, critical behavior, and renormalization

The Wilsonian logic of renormalization remains central in lattice gravity, but its implementation is altered by the fact that all scales must be established on the basis of the dynamics. Cooperman’s general analysis emphasizes that one cannot simply import the usual RG picture from lattice field theory on a fixed background: one needs a continuum model, a finite-size scaling ansatz, physical observables defining constant physics, and a dynamically justified standard unit of length (Cooperman, 2014).

In Regge calculus this logic is implemented through the divergence of a correlation length,

$l_{ij}^2$9

with the four-dimensional lattice estimate

$(4,1)$0

consistent with the conjecture $(4,1)$1. The corresponding physical phase is the strong-coupling phase $(4,1)$2, interpreted as anti-screening, and the UV fixed point is characterized by $(4,1)$3 (Hamber, 2015).

In EDT, the continuum-limit discussion is framed by the claim that fine-tuning is not an embarrassment but the expected signature of a regulator that breaks a target symmetry. The parameter that must be tuned is the local-measure coupling $(4,1)$4. After accounting for regulator-induced symmetry breaking, the 2016 analysis argues that pure gravity has only one nonredundant relevant coupling, namely the cosmological constant in Planck units $(4,1)$5, so that the continuum theory would have a one-dimensional UV critical surface. The same work further argues that the subtracted $(4,1)$6 is of order unity in the ultraviolet and decreases toward zero in the infrared (Laiho et al., 2016).

For CDT, the renormalization program has been pursued both conceptually and numerically. Cooperman’s CDT-specific scheme proposes lines of constant physics by matching ensembles through quantities such as the total four-volume, the effective de Sitter volume profile, and the spectral dimension, while recognizing that lattice spacing and standard units of length must be defined internally rather than assumed (Cooperman, 2014). More recent CDT work states that second-order phase transitions have been found and that the $(4,1)$7-$(4,1)$8 transition line may act as a UV critical line. In particular, approaching that line from within $(4,1)$9 gives

$(1,4)$0

which is presented as evidence compatible with $(1,4)$1 and therefore with a nontrivial continuum limit, while remaining explicitly suggestive rather than definitive (Ambjørn et al., 7 Apr 2026).

4. Geometric observables and emergent spacetime

The most developed lattice observables are geometric. In EDT, finite-size scaling of the volume-volume correlator and shell volume profiles yields a Hausdorff dimension

$(1,4)$2

(statistical error only), and the central shell-volume profile is well fit by the Euclidean de Sitter form. The same analysis reports that the long asymmetric tail of the profile decreases toward the continuum limit and interprets it as a cutoff artifact associated with baby universes rather than a property of the continuum theory. Diffusion observables further give a running spectral dimension with extrapolated values

$(1,4)$3

and

$(1,4)$4

so that the ultraviolet spectral dimension is consistent with $(1,4)$5 (Laiho et al., 2016).

In CDT, the principal large-scale observable is the spatial volume profile $(1,4)$6. In phase $(1,4)$7 it fits

$(1,4)$8

with fluctuations

$(1,4)$9

These scalings are those of an extended four-dimensional universe. The associated effective minisuperspace action,

$(3,2)$0

has Euclidean de Sitter space as its classical solution. CDT also employs diffusion observables, with

$(3,2)$1

and a fit

$(3,2)$2

as well as quantum Ricci curvature based on average distances between geodesic spheres (Ambjørn et al., 7 Apr 2026).

Matter probes provide a more direct long-distance test. In quenched EDT with free scalar fields on the dual lattice, the two-particle binding energy is fitted as

$(3,2)$3

with continuum and infinite-volume extrapolation yielding

$(3,2)$4

consistent with the Newtonian four-dimensional prediction $(3,2)$5, and

$(3,2)$6

in fiducial lattice units. The same analysis infers

$(3,2)$7

so that the fiducial lattice spacing is about $(3,2)$8 and the finest lattice spacing is about $(3,2)$9 (Dai et al., 2021).

5. Alternative microscopic formulations

Not all lattice regularizations start from simplicial metrics or triangulations. Wetterich’s construction of lattice diffeomorphism invariance distinguishes between an abstract lattice and its positioning $(2,3)$0 in a continuous manifold. A lattice action is lattice diffeomorphism invariant if, after rewriting it in terms of cell averages and lattice derivatives, it is independent of this positioning. In two dimensions this is expressed as

$(2,3)$1

and the claim is that this guarantees diffeomorphism invariance of the continuum limit and of the quantum effective action, with the metric emerging as a collective field rather than a fundamental degree of freedom (Wetterich, 2011).

Spinor-based lattice gravity pushes this logic further by making geometry composite. In Wetterich’s and Diakonov’s fermionic proposals, the microscopic variables are spinors and local Lorentz gauge fields, not a fundamental metric or vierbein. The composite frame field is built from fermion bilinears, while local Lorentz symmetry is exact on the lattice and diffeomorphism invariance is expected in the continuum limit. Diakonov’s hypercubic construction uses the covariant composite tetrad

$(2,3)$2

and motivates it by the “Sign Problem” of quantum gravity, namely the sign-indefiniteness of local diffeomorphism-invariant Euclidean actions (Wetterich, 2012). Vladimirov and Diakonov’s lattice spinor gravity develops a mean-field phase diagram in which second-order transition surfaces appear, a chiral-symmetry-breaking phase is found, and the Goldstone action takes the explicitly diffeomorphism-invariant form

$(2,3)$3

That work expects that Einstein gravitation is achieved at the phase transition and states that “a bonus is that the cosmological constant is probably automatically zero” (Vladimirov et al., 2012).

A different non-simplicial route appears in the tensor-network formulation of two-dimensional gravity. There the starting point is a lattice gauge theory with compact $(2,3)$4 link variables and an adjoint scalar, chosen so that the naive continuum limit reproduces a first-order gauge formulation of two-dimensional Euclidean gravity with positive cosmological constant. Character expansion produces an exact tensor network,

$(2,3)$5

and reveals a series of critical points at pure imaginary gravitational coupling associated with first-order transitions. After augmentation by a Yang-Mills term,

$(2,3)$6

the lattice spacing can be controlled and the scaling limit leads to a topology-dependent critical law

$(2,3)$7

where $(2,3)$8 is the Euler characteristic. The same paper also proposes a natural lattice generalization yielding Polyakov/Liouville-type effective dynamics (Asaduzzaman et al., 2019).

Within canonical loop quantum gravity, a different use of lattice regularization appears in the improved holonomy representation of the Ashtekar connection. The standard first-order link approximation is replaced by the second-order symmetric relation

$(2,3)$9

equivalently

$C_{\rm dS}$0

which becomes exact for linear links and leads to a cleaner cell-based regularization of the Hamiltonian constraint (Bilski, 2020).

6. Controversies, limitations, and current status

The central controversy is not whether lattice regularization is mathematically useful, but whether any given lattice model has a physically acceptable continuum limit. In EDT, the 2016 asymptotic-safety picture is internally coherent but not settled. The authors do not provide a definitive numerical proof of a second-order critical point; the spectral-dimension analysis relies on fitting ansätze such as the AJL form; the subtraction procedure for $C_{\rm dS}$1 is physically motivated but not uniquely derived; and the claim that pure gravity has only one relevant coupling differs from many continuum functional RG truncations, which often find three relevant directions (Laiho et al., 2016).

CDT has stronger semiclassical evidence, but its ultraviolet status is still open. Recent work explicitly states that the evidence for a UV fixed point is suggestive, not conclusive; that observables are still under development; that local propagating degrees of freedom are not yet directly isolated; and that current simulations typically involve universes of linear size only about $C_{\rm dS}$2–$C_{\rm dS}$3 Planck lengths, so finite-size effects and discretization artifacts remain substantial (Ambjørn et al., 7 Apr 2026).

The alternative formulations are even more exploratory. In spinor quantum gravity, the emergence of Einstein gravity is argued rather than demonstrated, the explicit phase analysis is largely two-dimensional, and the evidence is strongly tied to mean-field methods and criticality arguments (Vladimirov et al., 2012). In the tensor-network gauge-theory construction, everything concrete is two-dimensional, the critical points of the simplest model appear only after analytic continuation to pure imaginary coupling, and the naive lattice gravity action yields only first-order transitions rather than a useful continuum limit (Asaduzzaman et al., 2019).

An unconventional recent example makes these interpretive issues especially clear. A phenomenological bridge between loop quantum gravity and lattice physics models a spin network as a regular tetrahedral lattice with lattice constant

$C_{\rm dS}$4

a vertex Hamiltonian built from a Lennard–Jones potential, zero-point energy, and harmonic vibrations, and graviton-like excitations treated as spin-0 bosons. The construction is explicit about being phenomenological rather than a derivation of standard LQG dynamics, and it therefore illustrates how far the label “lattice quantum gravity” can stretch once one moves away from regulator-based definitions of the gravitational path integral (MacKay, 19 Jul 2025).

Taken together, these programs show that lattice-regularized quantum gravity is not a single method but a family of nonperturbative strategies. The shared objective is a continuum quantum theory obtained from a discrete regulator, but the microscopic variables, symmetry principles, and criteria for success differ sharply. What is already established is that discrete formulations can probe Planck-scale geometry nonperturbatively and can, in some cases, recover semiclassical four-dimensional behavior. What remains unsettled is which, if any, of these regularizations yields the continuum quantum theory of gravity in four dimensions.