Approaches to Quantum Gravity

Every other force in nature has been folded into quantum field theory, yet gravity refuses — and the obstruction is a clash of first principles, not a missing calculation. General relativity says the metric $g_{\mu\nu}$ is a dynamical field whose curvature is gravity, with no fixed stage and no preferred clock; quantum field theory says fields are operators on a fixed background, evolved by a Hamiltonian against an external time. Try to quantize the metric anyway — writing $g_{\mu\nu}=\eta_{\mu\nu}+\kappa h_{\mu\nu}$ and treating the graviton $h_{\mu\nu}$ like a photon — and the theory works beautifully at low energy, then falls apart: the coupling $\kappa=\sqrt{32\pi G}$ carries inverse-mass dimension, so each loop is more divergent than the last and no finite set of counterterms absorbs them all. Gravity is perturbatively non-renormalizable. This is not a defect to be patched but a signpost: Einstein's smooth spacetime is an effective description, valid until curvature radii shrink toward the Planck length $\ell_{\mathrm{P}}\approx1.6\times10^{-35}\,\mathrm{m}$ , where something new must take over. The information paradox, the singularity theorems, and the Big Bang all point to that edge. The leading programs — string theory, loop quantum gravity, asymptotic safety, causal set theory — are best read not as rival "theories of everything" but as different answers to one question: what replaces the classical metric at the Planck scale, and what stays fixed while you find out? This article lays out the shared obstruction, each program's signature equation, and the one quantitative target — the black-hole entropy $S=A/4\ell_{\mathrm{P}}^2$ — every candidate must hit.

Key takeaways

Quantizing the metric perturbatively fails: Newton's constant has negative mass dimension, so gravity is non-renormalizable and needs infinitely many counterterms above the Planck energy $E_{\mathrm{P}}\approx1.22\times10^{19}\,\mathrm{GeV}$ .
The deeper tension is background independence — GR has no fixed arena or external time, but canonical quantum theory presupposes both, producing the "problem of time".
The programs split by what they quantize: string theory replaces particles with extended objects ( $D=10$ ); loop quantum gravity quantizes the Ashtekar connection (discrete geometry); asymptotic safety seeks a UV fixed point (predictive continuum); causal sets discretize causal order (Lorentz-invariant atomism).
No experiment yet reaches the Planck scale — ~15 orders of magnitude beyond the LHC — so the programs are judged by consistency and theoretical benchmarks.
The sharpest shared benchmark is the Bekenstein–Hawking entropy $S=A/4\ell_{\mathrm{P}}^2$ : string theory counts D-brane microstates, LQG counts spin-network punctures, both recovering the area law.
Key idea: quantizing gravity is a confrontation with non-renormalizability and background independence, not a routine application of QFT — read each program by what it quantizes and what it holds fixed, and test it against the entropy $S=A/4\ell_{\mathrm{P}}^2$ .

The core insight

Quantizing gravity is not a routine application of quantum field theory but a confrontation with two facts: perturbative non-renormalizability and background independence. Each major program is defined by what it quantizes and what it keeps fixed — strings replace particles (UV-finite, extra dimensions), loop quantum gravity quantizes the Ashtekar connection (discrete geometry), asymptotic safety seeks a UV fixed point (a predictive continuum), causal sets discretize causal order (Lorentz-invariant atomism) — and all must reproduce the black-hole entropy $S=A/4\ell_{\mathrm{P}}^2$ .

Where this lives today

With the Planck scale out of direct reach, the field advances on three fronts. Observationally, LIGO–Virgo's gravitational-wave catalogue bounds the graviton mass below $\sim10^{-23}\,\mathrm{eV}/c^2$ , while gamma-ray-burst photons (notably Fermi-LAT's GRB 090510) push the scale of any linear, Planck-suppressed Lorentz violation above $E_{\mathrm{P}}$ — vindicating causal sets' Lorentz-invariant discreteness. Theoretically, the information paradox is the most active testing ground: the 2019–2020 "island" computations of the Page curve, using replica wormholes and quantum extremal surfaces, showed that unitarity and $S=A/4$ can be reconciled after all. And on the tabletop front, the Bose–Marletto–Vedral proposal aims to detect gravitationally induced entanglement between two masses — which, if seen, would be direct evidence that gravity is quantum. Meanwhile loop quantum cosmology predicts a Planck-density "bounce" replacing the Big Bang singularity. None has crowned a winner — but for the first time the candidates are squeezed by data, not just consistency.

The physical picture

Read Newton's constant $G$ as a length built into gravity itself. In a non-renormalizable theory each extra loop pays a factor of order $(E/E_{\mathrm{P}})^2$ , so corrections are negligible in the laboratory and become order one only as the energy $E$ approaches the Planck energy. The question "what is quantum gravity?" is really "what replaces the smooth metric when those corrections stop being small?" The first figure makes the scale concrete.

Figure

The scale gap: quantum-gravity effects grow like (E/E_P)² and reach order one only at the Planck energy E_P = 1.22×10¹⁹ GeV — some 15 orders of magnitude beyond the LHC, a range no experiment reaches

The deeper difficulty is not the scale but the stage. In ordinary quantum field theory you fix a background — the Minkowski metric — and quantize fields dancing on it; "before", "after", and the inner product of states are all defined against that fixed geometry. In gravity the stage is itself the field being quantized: there is no external clock and no fixed causal structure to hang a Hamiltonian on. The programs divide on whether to take this seriously from the first line (loop quantum gravity and causal sets build in background independence and predict a granular geometry) or to keep spacetime a smooth continuum — string theory by starting from a conventional background and recovering independence later, asymptotic safety by seeking a fixed point of the metric's own flow. The second figure maps the landscape along these two axes.

Figure

The landscape of approaches mapped by two choices — background-independent versus background-dependent, and smooth continuum versus fundamental discreteness — with the shared target S = A / 4ℓ_P² every candidate must reproduce

The core equations

Everything starts from the classical action we are trying to quantize. The Einstein–Hilbert action encodes the full dynamics of the metric, and the prefactor $c^4/16\pi G$ is where the trouble hides.

S_{\mathrm{EH}} = \frac{c^4}{16\pi G}\int \mathrm{d}^4x\,\sqrt{-g}\,(R - 2\Lambda) \qquad (1)

where: $g$ — determinant of the metric $g_{\mu\nu}$ ; $R$ — Ricci scalar; $G$ — Newton's constant, whose dimension drives non-renormalizability; $\Lambda$ — cosmological constant. The point: $G$ is dimensionful, unlike the dimensionless couplings of the Standard Model. Combining $\hbar$ , $c$ , and $G$ produces a unique energy and length — the scale at which those corrections become order one, the natural home of quantum gravity.

E_{\mathrm{P}} = \sqrt{\frac{\hbar c^5}{G}} \approx 1.22\times 10^{19}\,\mathrm{GeV},\qquad \ell_{\mathrm{P}} = \sqrt{\frac{\hbar G}{c^3}} \approx 1.62\times 10^{-35}\,\mathrm{m} \qquad (2)

where: $\hbar$ — reduced Planck constant; $c$ — speed of light; $E_{\mathrm{P}},\ell_{\mathrm{P}}$ — Planck energy and length, fixed by the three constants marking the meeting of quantum theory, relativity, and gravity. Non-renormalizability is visible before any loop: expanding the metric about flat space isolates the coupling $\kappa$ .

g_{\mu\nu}=\eta_{\mu\nu}+\kappa\,h_{\mu\nu},\qquad \kappa=\sqrt{32\pi G},\qquad [\kappa]=(\text{mass})^{-1} \qquad (3)

where: $h_{\mu\nu}$ — graviton field (mass-dimension one in four dimensions); $\kappa$ — gravitational coupling, of negative mass dimension. Every vertex carries a factor $\kappa$ — the fingerprint of a non-renormalizable interaction, made quantitative in Derivation 1. Taking background independence seriously instead leads to canonical quantization: promoting the ADM Hamiltonian constraint to an operator on the wavefunctional gives the Wheeler–DeWitt equation.

\left(16\pi G\,G_{ijkl}\frac{\delta^2}{\delta h_{ij}\,\delta h_{kl}} + \frac{\sqrt{h}}{16\pi G}\,{}^{(3)}R\right)\Psi[h_{ij}] = 0 \qquad (4)

where: $\Psi[h_{ij}]$ — wavefunctional on superspace, the space of 3-geometries; $h_{ij}$ — spatial 3-metric on a Cauchy slice; $G_{ijkl}$ — DeWitt supermetric; ${}^{(3)}R$ — intrinsic curvature of the slice. There is no $\partial_t\Psi$ term — the "problem of time": dynamics is frozen into a constraint (Derivation 3). Loop quantum gravity rewrites the same canonical theory in connection variables, whose Poisson bracket seeds the quantization.

\{A_a^i(x),\,E^b_j(y)\} = 8\pi G\gamma\,\delta^b_a\,\delta^i_j\,\delta^3(x-y),\qquad \hat{A}\,|s\rangle = 8\pi\gamma\,\ell_{\mathrm{P}}^2\sum_{p}\sqrt{j_p(j_p+1)}\;|s\rangle \qquad (5)

where: $A_a^i$ — $SU(2)$ Ashtekar–Barbero connection; $E^b_j$ — densitized triad encoding spatial geometry; $\gamma$ — Barbero–Immirzi parameter (fixed near $0.24$ by the entropy count); $\hat A$ — area operator; $|s\rangle$ — spin-network state; $j_p$ — half-integer spin on edge $p$ . The area spectrum is discrete with a nonzero smallest quantum — geometry is granular. Asymptotic safety keeps the continuum but demands the gravitational RG flow reach a nontrivial fixed point, governed by the functional flow equation.

k\,\partial_k \Gamma_k = \frac{1}{2}\,\mathrm{STr}\!\left[\left(\Gamma_k^{(2)}+R_k\right)^{-1} k\,\partial_k R_k\right] \qquad (6)

where: $\Gamma_k$ — effective average action at scale $k$ ; $\Gamma_k^{(2)}$ — its second functional derivative; $R_k$ — infrared cutoff; $\mathrm{STr}$ — supertrace. If the dimensionless couplings flow to a fixed point $g_*\neq0$ as $k\to\infty$ , gravity is nonperturbatively renormalizable. Causal set theory replaces the manifold: spacetime is a discrete partial order, and continuum volume is recovered by counting elements.

\langle N \rangle = \frac{V}{\ell^4},\qquad x\preceq y \ \text{transitive, acyclic, locally finite} \qquad (7)

where: $\langle N\rangle$ — expected number of elements in a region of 4-volume $V$ ; $\ell$ — discreteness length, of order $\ell_{\mathrm{P}}$ ; $\preceq$ — the causal order. "Order plus number equals geometry": the order fixes light cones, the count fixes volume. String theory extends particles to one-dimensional objects, and worldsheet consistency then fixes the dimension of spacetime — an output, not an input.

D=26\ (\text{bosonic}),\qquad D=10\ (\text{superstring}),\qquad V(r) = -\frac{G m_1 m_2}{r}\left[1 + \frac{3 G(m_1+m_2)}{c^2 r} + \frac{41}{10\pi}\frac{G\hbar}{c^3 r^2}\right] \qquad (8)

where: $D$ — critical dimension needed to cancel the Weyl (conformal) anomaly; $V(r)$ — the low-energy, model-independent quantum correction to Newton's law, coefficient $41/(10\pi)$ (Bjerrum-Bohr, Donoghue, Holstein). The two halves bracket the subject: a UV-complete framework fixing spacetime dimension, and a UV-agnostic prediction every framework shares.

Derivation AWhy perturbative quantum gravity is non-renormalizable

The goal is to show, without computing a single Feynman diagram, that treating the graviton like an ordinary field forces infinitely many undetermined parameters into the theory. The argument is pure power counting: the coupling's dimension fixes how a diagram's superficial degree of divergence grows with loop number, and for gravity that growth is unbounded. We expand the metric, read off the coupling dimension, count momenta, and conclude.

Step 1. Expand about flat space, $g_{\mu\nu}=\eta_{\mu\nu}+\kappa h_{\mu\nu}$ with $\kappa=\sqrt{32\pi G}$ , so $h_{\mu\nu}$ is a canonically normalized field and the action becomes a free graviton kinetic term plus an infinite tower of self-interaction vertices. Why: this makes gravity look like an ordinary field theory, so standard renormalization can be applied — and then seen to break.

Step 2. Read the coupling dimension. Since $[G]=(\text{mass})^{-2}$ , $[\kappa]=(\text{mass})^{-1}$ : every vertex carries a factor of negative mass dimension. Why: a coupling of negative mass dimension is the textbook definition of a non-renormalizable interaction — the danger sign appears before any loop integral is done.

Step 3. Count powers of momentum. Because each vertex supplies two derivatives tied to a factor of $\kappa$ , the superficial degree of divergence rises with loop order as $D_{\mathrm{div}}=2L+2$ in four dimensions, unbounded as $L$ grows. Why: a renormalizable theory has $D_{\mathrm{div}}$ bounded by a constant; here it climbs forever, so ever-higher-dimension operators diverge.

Step 4. Conclude. Each loop order needs new counterterms of ever-higher curvature dimension ( $R^2$ , $R^3$ , …), each with its own free coefficient, so predictivity is lost at the Planck scale. Explicitly, pure gravity is finite on-shell at one loop but diverges at two loops; gravity plus matter diverges already at one loop. Why: infinitely many free parameters means infinitely many measurements would be needed — not predictive. *The two-loop divergence is the Goroff–Sagnotti result $\dfrac{209}{2880}\dfrac{1}{(4\pi)^4\epsilon}\displaystyle\int\sqrt{-g}\,C_{\mu\nu}{}^{\rho\sigma}C_{\rho\sigma}{}^{\alpha\beta}C_{\alpha\beta}{}^{\mu\nu}$ , unabsorbable by field redefinition.*

Go deeper · Derivation B · 1 minThe Bekenstein-Hawking entropy as the universal target

Derivation BThe Bekenstein-Hawking entropy as the universal target

The goal is to derive the entropy that any quantum theory of gravity must reproduce, and to see why it is the single most powerful benchmark in the subject. A black hole radiates thermally, so it has a temperature; the first law then forces an entropy; integrating from zero mass yields a quantity proportional not to volume but to horizon area in Planck units. That area law is the fingerprint of the microstates, and reproducing it is the acid test each program faces.

Step 1. Start from the Hawking temperature of a Schwarzschild hole, $T_H=\dfrac{\hbar c^3}{8\pi G k_B M}$ , obtained from the surface gravity $\kappa_{\mathrm{sg}}=c^4/4GM$ via $T_H=\hbar\kappa_{\mathrm{sg}}/2\pi k_B c$ . Why: a temperature is what lets us treat the hole thermodynamically — it converts a geometry problem into a heat problem.

Step 2. Apply the first law $\mathrm{d}(Mc^2)=T_H\,\mathrm{d}S$ , using the horizon area $A=16\pi G^2M^2/c^4$ . Why: the first law identifies a thermodynamic entropy $S$ with the area, not leaving $A$ a merely geometric quantity.

Step 3. Integrate. With $\mathrm{d}S=\dfrac{c^2\,\mathrm{d}M}{T_H}=\dfrac{8\pi G k_B M}{\hbar c}\,\mathrm{d}M$ , integrating from $M=0$ gives $S=\dfrac{4\pi G k_B M^2}{\hbar c}=\dfrac{k_B c^3}{4G\hbar}A=\dfrac{A}{4\ell_{\mathrm{P}}^2}k_B$ . Why: the result is universal — one quarter of the horizon area in Planck units, independent of what the hole is made of, which makes it a clean target.

Step 4. Interpret microscopically. Since $S=k_B\ln\Omega$ , the area law asserts $\Omega=e^{A/4\ell_{\mathrm{P}}^2}$ microstates behind the horizon, which a real theory must produce from first principles. String theory does so via D-brane bound states (Strominger–Vafa); loop quantum gravity via spin-network punctures, fixing $\gamma$ . Why: matching a thermodynamic entropy to a microscopic count is the strongest evidence a candidate captures genuine quantum-gravitational degrees of freedom. *The Strominger–Vafa D1–D5 count gives $S=2\pi\sqrt{Q_1Q_5N}=A/4G$ , and the LQG puncture count fixes $\gamma\approx0.2375$ .*

Go deeper · Derivation C · 1 minFrom the ADM constraint to the Wheeler-DeWitt equation

Derivation CFrom the ADM constraint to the Wheeler-DeWitt equation

The goal is to trace how background independence turns the quantization of gravity into a single timeless constraint on a wavefunctional, exposing the "problem of time". The route runs through the ADM 3+1 split: GR in Hamiltonian form has a Hamiltonian that is pure constraint, vanishing on physical configurations. Quantizing that constraint à la Dirac makes it an operator annihilating physical states — and because it generated time reparametrizations, no ordinary Schrödinger evolution survives.

Step 1. From the ADM decomposition the Hamiltonian is a sum of constraints, $H=\int(N\mathcal{H}+N^i\mathcal{H}_i)$ , with the Hamiltonian constraint $\mathcal{H}=16\pi G\,G_{ijkl}\pi^{ij}\pi^{kl}-\dfrac{\sqrt{h}}{16\pi G}{}^{(3)}R\approx0$ and $\pi^{ij}$ conjugate to $h_{ij}$ . Why: the lapse $N$ and shift $N^i$ are Lagrange multipliers, so they impose $\mathcal{H}\approx0$ , $\mathcal{H}_i\approx0$ — the Hamiltonian vanishes on shell, the hallmark of a reparametrization-invariant theory.

Step 2. Quantize by $\pi^{ij}\to-i\hbar\,\delta/\delta h_{ij}$ , acting on a wavefunctional $\Psi[h_{ij}]$ on superspace. Why: with no fixed background there is no external time, so states are functionals of the whole 3-geometry, not fields on a slice.

Step 3. Impose the constraint on physical states, $\hat{\mathcal{H}}\Psi=0$ , giving the Wheeler–DeWitt equation (4) — with no $i\hbar\,\partial_t\Psi$ term at all. Why: because $\mathcal H$ generated time reparametrizations classically, its imposition removes time as an evolution parameter — the "problem of time". *The loop-variable version of the same constraints underlies the area spectrum (5).*

Limiting cases

The candidate theories are best distinguished by their behaviour in the characteristic regimes below — and every viable one must agree with the first and last rows.

Case	Result	Meaning
Low-energy / effective field theory	$V(r) = -\frac{G m_1 m_2}{r}\left[1 + \frac{3G(m_1+m_2)}{c^2 r} + \frac{41}{10\pi}\frac{G\hbar}{c^3 r^2}+\dots\right]$	Far below $E_{\mathrm{P}}$ every approach must reduce to GR as an effective theory; the leading quantum correction to Newton's potential is a universal, computable number, independent of the UV completion.
Planck scale	$\ell_{\mathrm{P}}=\sqrt{\hbar G/c^3}\approx1.62\times10^{-35}\,\mathrm{m}$	The smooth-manifold picture fails: LQG and causal sets predict literal discreteness; string theory smooths structure over the string length $\ell_s$ ; asymptotic safety hits a fixed point.
Ultraviolet completion	$g_i(k)\xrightarrow{k\to\infty}g_*$ (fixed point)	Asymptotic safety: the dimensionless couplings flow to a nontrivial fixed point $g_*\neq0$ , giving a finite-dimensional UV critical surface and a predictive continuum.
Semiclassical / horizon	$S_{\mathrm{BH}}=\frac{k_B c^3}{4G\hbar}A=\frac{A}{4\ell_{\mathrm{P}}^2}k_B$	Every viable approach must reproduce the Bekenstein–Hawking entropy from a microscopic count — the most quantitative check in the field.

Common trapNon-renormalizable does not mean wrong or useless

Perturbative non-renormalizability is often misread as "general relativity is inconsistent as a quantum theory". It is not. Einstein gravity is a perfectly good effective field theory at low energy, making unambiguous quantum predictions there — most famously the correction to the Newtonian potential with coefficient $41/(10\pi)$ . Non-renormalizability signals only the loss of predictivity at the Planck scale, where a UV completion is required; it says nothing about the regime we can probe. The opposite error is just as common: treating string theory, loop quantum gravity, asymptotic safety, and causal sets as interchangeable "theories of everything" — they make different commitments and none is yet sharply falsified or confirmed.

Worked examples

1 · The Planck scale from dimensional analysis (fully worked)

Using only $\hbar=1.055\times10^{-34}\,\mathrm{J\,s}$ , $c=2.998\times10^8\,\mathrm{m/s}$ , and $G=6.674\times10^{-11}\,\mathrm{m^3\,kg^{-1}\,s^{-2}}$ , find the Planck length, energy, and mass, and confirm the gap to the LHC.

Step 1. Planck length, the unique length from the three constants: $\ell_{\mathrm{P}}=\sqrt{\hbar G/c^3}$ . With $\hbar G=7.04\times10^{-45}$ and $c^3=2.69\times10^{25}$ , $\ell_{\mathrm{P}}=\mathbf{1.62\times10^{-35}\ m}$ — about $10^{20}$ times smaller than a proton.

Step 2. Planck energy: $E_{\mathrm{P}}=\sqrt{\hbar c^5/G}=1.96\times10^{9}\,\mathrm{J}=\mathbf{1.22\times10^{19}\ GeV}$ (dividing by $1.602\times10^{-10}\,\mathrm{J/GeV}$ ).

Step 3. Planck mass: $m_{\mathrm{P}}=E_{\mathrm{P}}/c^2=\sqrt{\hbar c/G}=2.18\times10^{-8}\,\mathrm{kg}=\mathbf{21.8\ \mu g}$ — a grain of dust, yet enormous for one particle, whose mass is some $10^{-20}\,m_{\mathrm{P}}$ .

Step 4. The gap. The LHC reaches $\sim1.4\times10^{4}\,\mathrm{GeV}$ , so $E_{\mathrm{P}}/E_{\mathrm{LHC}}\approx10^{15}$ and the correction is of order $(E_{\mathrm{LHC}}/E_{\mathrm{P}})^2\approx10^{-30}$ . The Planck scale sits ~15 orders of magnitude beyond direct experiment — the central obstacle of the field.

2 · The Bekenstein–Hawking entropy of a solar-mass black hole (fully worked)

Compute the horizon area, entropy, and temperature of a non-rotating black hole of mass $M=M_\odot=1.99\times10^{30}\,\mathrm{kg}$ .

Step 1. Horizon radius: $r_s=2GM/c^2=\mathbf{2.95\times10^{3}\ m}$ — a Sun squeezed to 3 km.

Step 2. Area: $A=4\pi r_s^2=\mathbf{1.10\times10^{8}\ m^2}$ .

Step 3. Entropy: $S/k_B=A/4\ell_{\mathrm{P}}^2=(1.10\times10^{8})/[4(2.61\times10^{-70})]=\mathbf{1.05\times10^{77}}$ — nearly twenty orders of magnitude more than the Sun as ordinary gas ( $\sim10^{58}k_B$ ).

Step 4. Temperature: $T_H=\hbar c^3/(8\pi G k_B M)=\mathbf{6.2\times10^{-8}\ K}$ — colder than the CMB, so the hole absorbs more than it radiates. The $10^{77}k_B$ of entropy behind a 3 km horizon — some $e^{10^{77}}$ microstates — is what every program must count.

3 · The smallest quantum of area in loop quantum gravity

Using equation (5) with the entropy-fixed value $\gamma\approx0.2375$ , find the smallest nonzero area eigenvalue.

Step 1. The smallest excitation is a single edge with minimal spin $j=\tfrac12$ , so the sum has one term: $\hat A=8\pi\gamma\ell_{\mathrm{P}}^2\sqrt{\tfrac12\cdot\tfrac32}$ .

Step 2. Evaluate: $\sqrt{\tfrac34}=0.866$ , so $\hat A_{\min}=8\pi(0.2375)(0.866)\,\ell_{\mathrm{P}}^2=\mathbf{5.17\ \ell_{\mathrm{P}}^2}$ .

Step 3. In SI units, $\hat A_{\min}=5.17\times(2.61\times10^{-70}\,\mathrm{m^2})=\mathbf{1.35\times10^{-69}\ m^2}$ . Space cannot be subdivided below this quantum — the concrete sense in which LQG makes geometry discrete.

Physics of the portrait

Quantum gravity has no single founder, but one figure did more than any other to turn "quantizing the metric" into a precise mathematical program.

Origins · 1 minBryce DeWitt and how this was discovered

Bryce DeWitt

1923–2004 · canonical quantum gravity and the Wheeler-DeWitt equation

Bryce DeWitt made quantum gravity a subject one could calculate in, not just speculate about. In a landmark trilogy of papers in 1967 he laid out the canonical quantization of general relativity: taking the ADM Hamiltonian, which is a pure constraint, and imposing it on a wavefunctional $\Psi[h_{ij}]$ over the space of 3-geometries — the equation now called the Wheeler–DeWitt equation, at the heart of this article's third derivation. It was DeWitt who confronted the "problem of time" head-on, recognizing that a background-independent theory has no external clock and that dynamics must be reconstructed from a timeless constraint. In the same work he developed the covariant background-field approach to graviton loops that later made the non-renormalizability argument precise, and the DeWitt supermetric $G_{ijkl}$ bears his name. He was also a forceful champion of Everett's "many-worlds" interpretation, the natural reading of a wavefunction of the whole universe — for what observer could stand outside the cosmos to collapse it? His legacy is the recognition that quantizing gravity forces a reckoning with the foundations of quantum theory itself: no fixed stage, no outside observer, no external time — a reckoning every program here inherits.

History

1936

— Bronstein argues gravity cannot be quantized like an ordinary field

1967

— DeWitt formulates canonical quantum gravity and its wave equation

1974

— Hawking's black-hole radiation sets the target S = A / 4

1986

— Goroff and Sagnotti prove pure gravity diverges at two loops

1997

— Maldacena's AdS/CFT realizes holographic quantum gravity

Going further

Texts. Misner, Thorne & Wheeler, Gravitation (1973), Ch. 43–44 · Wald, General Relativity (1984), Ch. 14 and Appendix E · Carroll, Spacetime and Geometry (2004), Ch. 9 · Rovelli, Quantum Gravity (Cambridge, 2004) · Polchinski, String Theory Vols. I–II (Cambridge, 1998) · Thiemann, Modern Canonical Quantum General Relativity (Cambridge, 2007) · Niedermaier & Reuter, "The Asymptotic Safety Scenario in Quantum Gravity", Living Rev. Relativity 9, 5 (2006) · Bombelli, Lee, Meyer & Sorkin, "Space-time as a causal set", Phys. Rev. Lett. 59, 521 (1987). Key numbers in this article: Planck length/energy/mass from $\hbar,c,G$ ; Hawking temperature and entropy from surface gravity; the smallest LQG area from $8\pi\gamma\ell_{\mathrm{P}}^2\sqrt{j(j+1)}$ with $\gamma\approx0.2375$ ; critical dimensions 26 and 10 from conformal-anomaly cancellation; the $41/(10\pi)$ potential-correction coefficient (Bjerrum-Bohr, Donoghue & Holstein); the two-loop coefficient $209/2880$ (Goroff & Sagnotti); and the Wetterich flow equation.

Prerequisites. ADM 3+1 Decomposition of Spacetime · Hamiltonian and Momentum Constraints · ADM Hamiltonian Formulation of General Relativity · Initial-Value (Cauchy) Problem in General Relativity. Leads to. Black Hole Thermodynamics · Holographic Principle and AdS/CFT · Loop Quantum Cosmology.

Ask the tutor: Ask me to walk through why $[\kappa]=(\text{mass})^{-1}$ dooms perturbative gravity, or how string theory and loop quantum gravity each recover the black-hole entropy $S=A/4\ell_{\mathrm{P}}^2$ from a microscopic count.

If gravity is non-renormalizable, how can anyone predict quantum corrections to Newton's law?

Because non-renormalizability limits predictivity only at the Planck scale, not at long distances. Treating GR as an effective field theory, the leading quantum correction to the Newtonian potential — the $\tfrac{41}{10\pi}\tfrac{G\hbar}{c^3 r^2}$ term, which sits below the classical post-Newtonian one — is a genuine prediction: its coefficient comes from the low-energy graviton loop, independent of whatever UV completion takes over near $\ell_{\mathrm{P}}$ . The catch is size: of order $(\ell_{\mathrm{P}}/r)^2$ , which is about $10^{-40}$ even at nuclear separations and falls as $r^{-2}$ beyond them — real, calculable, hopelessly tiny. It matters as principle: quantum gravity is calculable below the Planck scale.

Is string theory "ahead" of the other approaches?

It depends on the yardstick. String theory is the most developed mathematically, is UV-finite order by order, automatically contains a massless spin-2 graviton, and through AdS/CFT gives a nonperturbative definition of quantum gravity in certain backgrounds — an achievement none of the others matches. But it pays with heavy commitments: extra dimensions, supersymmetry, and a vast "landscape" of vacua that has resisted a unique prediction for our universe. Loop quantum gravity, asymptotic safety, and causal sets assume less and take background independence more seriously. "Ahead" presupposes a single race; they explore different tradeoffs, and none is confirmed.

What exactly is the "problem of time"?

In ordinary quantum mechanics time is an external parameter: states evolve by $i\hbar\,\partial_t\Psi=\hat H\Psi$ . In canonical quantum gravity there is no external time, because GR is invariant under relabelling of the time coordinate. The ADM Hamiltonian is a pure constraint, so the quantum equation is $\hat{\mathcal H}\Psi=0$ — Wheeler–DeWitt — with no time-derivative term. The wavefunction of the universe does not "evolve"; it just is. Recovering the appearance of time from this frozen formalism — for instance by treating an internal variable like the volume of the universe as a physical clock — is a deep, still-unsettled problem.

Why does reproducing black-hole entropy count as such strong evidence?

Because it is a nontrivial number that emerges from independent starting points. Hawking's semiclassical calculation gives $S=A/4\ell_{\mathrm{P}}^2$ from thermodynamics and quantum fields on a curved background, with no microscopic input. A candidate theory must then count the $e^{S}$ microstates from its own degrees of freedom and get the same answer, factor of $1/4$ included. String theory does this via D-brane bound states (Strominger–Vafa), loop quantum gravity via spin-network punctures. When two utterly different frameworks reproduce the same nonperturbative number, they are likely latching onto real degrees of freedom.

Do any of these approaches make predictions we could actually test?

Not yet decisively, but the situation is improving. Several predict tiny departures from exact Lorentz invariance or modified dispersion, which gamma-ray-burst timing and gravitational-wave data already constrain above the Planck scale. Loop quantum cosmology predicts a "bounce" instead of a Big Bang singularity, with possible CMB imprints. Causal set theory predicted the order of magnitude of the cosmological-constant fluctuations before its small value was measured. None is confirmed yet, but pure theory is giving way to observational pressure.

Are these approaches rivals, or could several be right?

Some may be different descriptions of the same physics rather than competitors. AdS/CFT already shows a gravitational theory and a gauge theory can be exactly dual — the same physics in two languages — so "which is fundamental?" is not always well posed. The final theory could absorb insights from several programs — string theory's UV finiteness, LQG's discrete geometry, causal sets' Lorentz-invariant atomism, asymptotic safety's predictive continuum — so they are best seen as complementary probes of one hard problem.