Loop quantum gravity

1. Motivation

General relativity (GR) makes geometry dynamical: matter and geometry influence one another. In units with $c=1$, and omitting the cosmological constant here,

$$G_{\mu \nu }=8\pi G{T}_{\mu \nu }.$$

Quantum theory successfully describes matter, but a quantum field on a prescribed curved spacetime does not yet quantize that spacetime. Near a classical singularity, treating geometry as a smooth classical background may fail. This motivates quantum gravity; it is not a proof that a particular quantization is correct, nor a theorem that every classical field must be quantized in one unique way.

Loop quantum gravity (LQG) asks whether GR can be quantized without first fixing a background metric and expanding around it. “Nonperturbative” means that the construction does not start with a small gravitational disturbance on a chosen geometry. “Background independent” means that lengths and areas come from dynamical variables, not a fixed metric. A differentiable manifold and other mathematical choices still enter. See Ashtekar and Lewandowski's foundational account.

The characteristic scale is ${\ell }_P=\sqrt{\hslash G/c^3}$, about $1.6\times 10^{-35}\mathrm{m}$. This dimensional combination indicates where quantum and gravitational effects may both matter. It does not, by itself, establish that space is a regular grid or that every distance has a minimum. LQG needs a state space, geometric operators, and consistent dynamics to make more specific statements.

2. Canonical Quantization of GR

The canonical approach describes data on a spatial slice $\Sigma$ and asks which data describe the same spacetime. In the ADM formulation the variables are a spatial metric $q_{ab}$ and its conjugate momentum ${\pi }^{ab}$. A foliation organizes the calculation; it is not a preferred physical clock.

Ashtekar Variables

We use real Ashtekar–Barbero variables for Lorentzian GR in the time gauge, with nondegenerate triads and real $\gamma >0$. The original self-dual Ashtekar connection is complex: these are related formulations, not identical variables. Barbero's real formulation and Holst's action explain this canonical change.

$$A_a^i={\Gamma }_a^i(E)+\gamma K_a^i,E_i^a=\sqrt{q}{e}_i^a,E_i^a{E}_i^b=q{q}^{ab}.$$

Here $q=\det (q_{ab})$, $e_i^a$ is an inverse triad, ${\Gamma }_a^i$ its compatible spin connection, and $K_a^i$ encodes extrinsic curvature: how the slice sits in spacetime. Spatial indices are $a,b$; $i$ labels an internal orthonormal frame. Thus $E$ carries spatial geometry and $A$ combines intrinsic connection with extrinsic curvature. Calling $E$ an “electric field” is a gauge-theory analogy; it is not electromagnetism.

With $\kappa =8\pi G$ and $c=1$,

$$\{A_a^i(x),E_j^b(y)\}=\kappa \gamma {\delta }_a^b{\delta }_j^i{\delta }^{(3)}(x,y).$$

The dimensionless Barbero–Immirzi parameter $\gamma$ labels this canonical choice. Classical vacuum equations do not fix it. Its appearance in geometric spectra makes its quantum interpretation and physical determination important; an illustrative value is not a measured constant.

3. Constraints

These variables contain redundant descriptions. In vacuum, with zero cosmological constant and an overall nonzero normalization suppressed in $\mathcal{H}$,

$${\mathcal{G}}_i=D_a{E}_i^a\approx 0,{\mathcal{V}}_a=E_i^b{F}_{ab}^i\approx 0,$$ $$\mathcal{H}=\frac{E_i^a{E}_j^b}{\sqrt{|\det E|}}\left({ϵ}^{ij}{}_k{F}_{ab}^k-2(1+{\gamma }^2)K_{[a}^i{K}_{b]}^j\right)\approx 0.$$

$F$ is the curvature of $A$ and $D$ its covariant derivative. The symbol $\approx 0$ means equality on the constraint surface, not a small error. The Gauss constraint removes arbitrary rotations of the internal frame. The vector constraint generates spatial coordinate changes up to a Gauss transformation; a pure spatial-diffeomorphism generator includes the corresponding Gauss term. The Hamiltonian constraint implements changes of slicing, rather than ordinary evolution relative to an external universal time.

Counting is informative: $A$ and $E$ give 18 phase-space components per point. Seven independent first-class constraints remove 14 phase-space components, leaving four, or two configuration degrees of freedom—the local gravitational polarizations. This is the regular classical vacuum count; boundaries and singular sectors need separate treatment.

Quantization first builds kinematics, then implements constraints and the physical inner product. A spin network is therefore not automatically a physical solution. A relational observable might be the area of a matter-defined surface when a scalar clock has a specified value. Defining these and recovering the classical constraint algebra without an anomaly are central dynamical tasks. Thiemann's quantum-constraint construction is a concrete proposal, not a unique completed phenomenological theory.

4. Holonomies and Fluxes

The basic variables are smeared along curves and across surfaces:

$$h_{\ell }[A]=\mathcal{P}\exp \left({\int }_{\ell }{A}_a^i{\tau }_{i}d{x}^a\right),E(S,f)={\int }_S{f}^i{E}_i^{a}d{S}_a.$$

${\tau }_i$ are generators of $\mathfrak{su}(2)$, $f^i$ a test field, and $\mathcal{P}$ orders noncommuting matrices along the path. A holonomy compares internal frames by parallel transport; it is an $SU(2)$ element, not a length. Flux measures oriented triad information through a surface. The sign in the exponential depends on the connection convention.

As a thought experiment, transport a frame around a small closed loop. In a smooth classical geometry the leading departure of its holonomy from the identity encodes curvature through the loop. LQG uses finite transports as fundamental variables rather than assuming an operator $A(x)$ exists point by point. Holonomies act by multiplication and fluxes by derivatives on functions of holonomies. Their algebra is sensitive to path–surface intersections.

The holonomy–flux uniqueness theorem singles out a representation under specified algebraic, covariance, and cyclic-state assumptions. It does not prove that every possible quantum gravity theory is LQG or choose the Hamiltonian.

5. Spin Networks

A gauge-invariant spin network consists of a graph, spins $j=0,\frac{1}{2},1,\dots$ on its edges, and invariant tensors called intertwiners at its vertices. Spins label $SU(2)$ representations; they are not little rotating balls. An intertwiner combines incident representations into an invariant object, implementing the Gauss constraint. Graphs specify relations in quantum geometry; they need not form a regular physical lattice.

For a surface crossing edges transversely away from vertices, the standard area operator has eigenvalues

$$A(S)=8\pi \gamma {\ell }_P^2\sum _{p\in S\cap \Gamma }\sqrt{j_p(j_p+1)}.$$

A flux at a puncture acts like an angular momentum generator, whose quadratic $SU(2)$ Casimir has eigenvalue $j(j+1)$. Taking its magnitude explains $\sqrt{j(j+1)}$. The area-operator construction establishes discreteness on the kinematic Hilbert space. Surfaces containing vertices require the more general expression. As Dittrich and Thiemann explain, a kinematic spectrum alone does not determine the spectrum of every fully relational physical observable.

A worked area measurement

Let a patch be punctured once by $j=\frac{1}{2}$ and once by $j=1$, with no vertex on it. Then

$$\frac{A}{8\pi \gamma {\ell }_P^2}=\sqrt{\frac{1}{2}(\frac{1}{2}+1)}+\sqrt{1(1+1)}=\frac{\sqrt{3}}{2}+\sqrt{2}\approx \mathrm{2.280.}$$

A single $j=\frac{1}{2}$ contributes $4\pi \sqrt{3}\gamma {\ell }_P^2$. Two such punctures give $8\pi \sqrt{3}\gamma {\ell }_P^2$, whereas one $j=1$ gives $8\pi \sqrt{2}\gamma {\ell }_P^2$: equal sums of spin labels need not give equal areas. These are local contributions within an admissible network; the rest of the graph must satisfy intertwiner conditions. This is a calculation of possible outcomes within the quantization, not an observed grain size.

The shaded patch has two transverse punctures. The geometry is schematic; the equation, not the drawn spacing, determines the area eigenvalue.

Volume is more subtle:

$$\hat{V}(R)=\sum _{v\in R\cap \Gamma }\sqrt{|{\hat{q}}_v|}.$$

${\hat{q}}_v$ contains triple products of flux generators, orientation factors, and dimensional constants. It generally acts as a matrix on vertex intertwiner spaces. Fixed edge spins need not fix volume: an appropriate intertwiner basis is also needed. The volume-operator construction explains this distinction. Discrete geometric spectra imply neither equally sized cubes nor an automatically discrete external time.

From spin networks to spin foams

A spin network describes boundary geometry. A spin foam is a labeled two-dimensional complex assigning amplitudes between such boundaries: graph edges sweep out faces and graph vertices trace edges of the complex. These are histories in a sum over quantum geometries, not nodes moving through pre-existing space.

Schematically, on a chosen complex $\mathcal{C}$,

$$Z_{\mathcal{C}}=\sum _{\{j_f,i_e\}}\prod _f{A}_f\prod _e{A}_e\prod _v{A}_v.$$

Labels are summed; face, edge, and vertex amplitudes encode a model's dynamics. Boundary data and normalization are suppressed. The EPRL construction connects a four-dimensional spin-foam vertex to LQG boundary states through simplicity constraints. Specifying a finite complex is not yet establishing the full continuum theory.

For suitable nondegenerate geometric boundary data, the large-spin asymptotics of a Euclidean EPRL four-simplex amplitude contain the Regge action, a discretized gravitational action. This semiclassical consistency result is valuable. It does not prove that every amplitude, refinement limit, or long-distance correlation function reproduces GR.

6. Loop Quantum Cosmology (LQC)

LQC applies loop-inspired quantization to symmetry-reduced cosmological systems. Imposing homogeneity before quantization discards infinitely many inhomogeneous modes. A theorem in this reduced system is not automatically a theorem in full LQG.

Consider the standard spatially flat, homogeneous, isotropic model with a massless scalar and sharply peaked semiclassical states in the improved-dynamics prescription. Its effective equations include

$$H^2=\frac{8\pi G}{3}\rho \left(1-\frac{\rho }{{\rho }_c}\right),\dot{\rho }=-3H(\rho +P),$$

where $H=\dot{a}/a$, $P$ is pressure, and ${\rho }_c$ a model-dependent critical density of Planck order. The low-density limit recovers the ordinary flat Friedmann equation. The negative quadratic correction gives $H=0$ at ${\rho }_c$. The improved-dynamics analysis states its assumptions and studies the bounce analytically and numerically.

Deriving a bounce in the effective model

Differentiation and energy conservation give, away from $H=0$ and then by continuous extension,

$$\dot{H}=-4\pi G(\rho +P)\left(1-2\frac{\rho }{{\rho }_c}\right).$$

At $\rho ={\rho }_c$, matter with $\rho +P>0$ has $\dot{H}>0$: contraction can pass to expansion. Merely finding $H=0$ would not prove a bounce. For a massless scalar, $P=\rho$ and $\rho \propto a^{-6}$. Choosing the bounce at $t=0$ gives the exact solution of these effective equations

$$a(t)=a_B{\left(1+\frac{t^2}{t_B^2}\right)}^{1/6},\rho (t)=\frac{{\rho }_c}{1+t^2/t_B^2},t_B^{-2}=24\pi G{\rho }_c.$$

Substitution verifies both equations: $H=t/[3(t_B^2+t^2)]$. At $t=\pm t_B$, $\rho ={\rho }_c/2$ and $a=2^{1/6}{a}_B$. The numerical value of $a_B$ depends on spatial-coordinate normalization; the finite density and transition from contraction to expansion carry the physical point.

Horizontal axis: $t/t_B$. Upper curve: $a/a_B$; lower curve: $\rho /{\rho }_c$. These are the effective massless-scalar solution, not observational data or a universal prediction for every cosmology.

Exactly solvable LQC permits stronger statements about classes of quantum states than this effective calculation alone: see Ashtekar, Corichi and Singh. Extending conclusions to anisotropies, perturbations, different matter or quantizations, and the full theory needs additional arguments. A model bounce does not establish what preceded our observed hot early universe.

7. Challenges and Open Questions

• Dynamics and constraints: control the quantum Hamiltonian, its regularization choices, physical states, and inner product while preserving the appropriate gauge symmetries.
• Continuum and semiclassical limits: derive smooth geometry and low-energy GR with matter from many excitations. Large spins on a fixed graph and refinement to many graph elements are different limits; both need consistent treatment.
• The Immirzi parameter: determine which dependence on $\gamma$ survives in physical predictions. Matching black-hole entropy under stated counting assumptions is not an independent laboratory measurement of $\gamma$.
• Observations: a CMB feature or gravitational-wave modification requires a specific effective model, initial state, and treatment of perturbations. Its robustness and distinction from ordinary astrophysics or other early-universe models must be demonstrated before calling it an LQG test.

A snapshot of ongoing work is Assanioussi and Livine's September 2026 revision: it analyzes classical Hamiltonian dynamics on small “candy” graphs as building blocks, explicitly leaving the detailed quantum regime for later work. This is progress on a controlled subsystem, not completion of full quantum dynamics or an empirical detection.

Common misconceptions

“The loops are strings.” Holonomies are parallel-transport variables and network labels describe geometry, not the propagating strings of string theory. “Background independence removes every assumption.” Topology, variables, quantization, and boundaries still matter. “A discrete area spectrum proves space is a pixel screen.” The spectrum concerns an operator; states can be superpositions, with no preferred regular grid. “The bounce proves there was a previous universe.” It establishes a conditional model result, with no such observation supplied by the calculation.

Status

Research checked through 13 October 2026. LQG has precise kinematic constructions and substantial work on dynamics and reduced models. The linked papers establish mathematical results and model predictions; none supplies experimental confirmation that our spacetime has this proposed quantum structure. A controlled route from full dynamics to distinctive measured predictions remains a research objective.

LQG emphasizes quantizing gravitational geometry without a fixed background metric. String theory begins from different degrees of freedom and also has nonperturbative formulations; saying that it is invariably background dependent oversimplifies the comparison. Assess each proposal by its assumptions, internal consistency, recovery of known physics, and observable consequences.