The Problem of Time

Ordinary quantum mechanics assumes a time parameter that is not itself an operator and evolves states with a Hamiltonian. Canonical general relativity instead makes normal deformations of a spatial slice part of the gauge description and constrains the Hamiltonian. The resulting “problem of time” is not one paradox: it is a family of questions about constraints, clocks, observables, inner products, and spacetime reconstruction. Keeping those questions separate prevents a local solution from being advertised as universal.

Required background. Variational and Symplectic Geometry supplies gauge orbits, first-class constraints, and reduced phase space.

Helpful background. Flat-Spacetime Quantum Field Theory supplies Schrödinger and Heisenberg evolution; Einstein Equations supplies covariant dynamics; Quantum Information supplies conditional states and unitarity; Formulations of General Relativity supplies foliations and constraints; and Quantum Gravity supplies the common comparison criteria.

From refoliation gauge to the frozen equation

For a closed canonical system the total Hamiltonian has the form

$$H_{\rm tot}=\int_\Sigma d^3x \left(N\mathcal H_\perp+N^i\mathcal H_i\right).$$

Physical states formally satisfy $\widehat{\mathcal H}_\perp\Psi=0$ and $\widehat{\mathcal H}_i\Psi=0$, not $i\partial_t\Psi=\widehat H\Psi$ with a preferred external $t$. This is the frozen-formalism problem. It reflects refoliation invariance: changing a coordinate label for slices should not create a physical change. An asymptotic boundary can supply a nonzero Hamiltonian charge and a preferred boundary time, so the conclusion depends on boundary conditions and is not identical for closed cosmology and asymptotically flat scattering.

The classical statement is subtler than “nothing happens.” Gauge-invariant correlations can vary: a planet’s position when a clock reads noon is different from its position when the clock reads evening. The task is to define such correlations without smuggling in a preferred coordinate. Isham separates frozen formalism, global time, multiple choice, observables, and reconstruction in Isham 1992, §§5.1–5.4 and 6.2–6.4.

Five logically distinct obstructions

Global clock. A phase-space function $T$ is a good clock only where $\{T,C\}$ is nonzero and its level surfaces intersect each relevant gauge orbit once. Recollapsing cosmologies and turning points defeat monotonicity.

Multiple choice. Different admissible clocks can produce different reduced Hamiltonians, operator orderings, domains, and notions of positive frequency. A classical canonical transformation between clocks need not be unitarily implementable after quantization.

Observables. A Dirac observable commutes with all constraints, yet a useful observable must also encode change. Complete observables $F_{A,T}(\tau)$ do this by asking for $A$ where $T=\tau$; their construction is often local in phase space.

Hilbert space. A Wheeler–DeWitt solution space needs a positive physical inner product. A Klein–Gordon-like current can be indefinite, while group averaging depends on the constraint group, domain, and convergence.

Spacetime reconstruction. Equal-time commutators and constraint closure must recover refoliation-independent spacetime predictions. A theory can possess an internal clock in one truncation yet fail to reconstruct local quantum fields on a four-dimensional semiclassical geometry.

Rovelli’s partial-observable framework gives a precise correlation-based answer to the observables problem without claiming that it automatically supplies a global clock or preferred dynamics Rovelli 2002, §§2–4.

Competing strategies

One strategy reduces first: choose $T$, solve the constraint for its conjugate momentum, and quantize the remaining variables. This can yield an ordinary Schrödinger theory, but the result is tied to the clock patch. A second strategy quantizes first, constructs the physical state space, and defines relational observables there. It preserves the constraint perspective but can leave the inner product and observables technically difficult. A third treats the fundamental state as timeless and interprets records or correlations as change. A fourth recovers time only in a semiclassical Born–Oppenheimer branch.

These positions are genuinely competing when they make different claims about fundamentality and equivalence. They can also be complementary in different parameter regimes. Anderson’s broad review shows that many proposed resolutions address only selected facets and must be tested against the others Anderson 2012, §§2–4 and 7–9.

Consider the deparametrizable constraint $C=p_T+H(q,p)$. Gauge fixing $T=\tau$ gives a reduced Hamiltonian $H$ and, when $\widehat H$ is self-adjoint, unitary evolution. Replacing the constraint by $C=p_T^2-H^2$ creates two frequency branches. Choosing one branch is additional data; allowing interference between both branches makes a single Schrödinger description in $T$ incomplete. This finite model is an adverse check on any claim that “a clock variable” alone solves every time problem.

Results and open questions as of 2026-08-20

No theorem or calculation establishes one universal resolution for full four-dimensional quantum gravity. The evidence is formulation- and regime-dependent, and the remaining questions are open.

Position or formulation	Strongest result or evidence	Scope, adverse result, or missing step
Reduced quantization with an internal clock	Exact unitary evolution in deparametrizable models	Clock can have turning points; different clocks need not yield unitarily equivalent theories
Dirac quantization and complete observables	Gauge-invariant correlations can be constructed in finite systems and selected field models	Physical inner product and globally complete observable algebra remain unresolved for full gravity
Timeless records or correlations	Explains change without external time at the conceptual and toy-model level	Requires a measure and record criterion; does not by itself recover local QFT dynamics
Semiclassical WKB time	Derives approximate Schrödinger evolution for light fields along a classical branch	Fails near branch interference, turning points, or uncontrolled backreaction

The conservative conclusion is that each construction is a conditional tool: it solves a declared facet under stated clock, boundary, Hilbert-space, and semiclassical assumptions. The stronger view is that one of these tools captures the fundamental ontology and the others are approximations. Current controlled results do not settle that interpretive step, and no empirical discriminator has selected among these positions.

The conclusions could change after a proof relates broad clock choices by controlled unitary maps, after a counterexample demonstrates unavoidable inequivalence in a realistic field-theory sector, or after an anomaly-free full constraint representation yields a positive physical inner product and semiclassical spacetime observables.

A clock-patch diagnostic

For $C=p_T+P^2/(2m)=0$, define the complete observable corresponding to $Q$ at $T=\tau$. Then explain what the same construction predicts if the clock is replaced by $T'=T^2$ near $T=0$.

Solution

The gauge flow parameter $s$ gives $T(s)=T+s$ and $Q(s)=Q+(P/m)s$. Setting $T(s)=\tau$ yields $F_{Q,T}(\tau)=Q+(P/m)(\tau-T)$, which Poisson-commutes with $C$ and changes with the clock reading. For $T'=T^2$, the bracket $\{T',C\}=2T$ vanishes at $T=0$, and $T'=\tau'$ intersects a generic orbit twice, at $T=\pm\sqrt{\tau'}$. The relational value is therefore branch-dependent there. The failure is not absence of classical motion; it is failure of $T'$ to provide a single global gauge fixing and hence a single reduced quantum evolution across the turning point.

References

• Anderson, Edward. “Problem of Time in Quantum Gravity.” Annalen der Physik 524 (2012): 757–786. arXiv:1009.2157.
• Isham, Chris J. “Canonical Quantum Gravity and the Problem of Time.” In Integrable Systems, Quantum Groups, and Quantum Field Theories, 1992. arXiv:gr-qc/9210011.
• Rovelli, Carlo. “Partial Observables.” Physical Review D 65 (2002): 124013. DOI.