Causal Set Kinematics

Causal-set kinematics replaces a differentiable Lorentzian manifold by a locally finite partial order. The order records which events can causally precede which; the number of elements records spacetime volume. This terse prescription becomes geometrical only after one states a faithful-embedding criterion, a discreteness density, and diagnostics for dimension and manifoldlikeness. It does not imply that a generic finite order resembles a spacetime.

Helpful background. Flat-Spacetime Quantum Field Theory supplies field and propagator limits; Einstein Equations supplies the continuum geometry to recover; Quantum Information supplies order and correlation language; Quantum Gravity supplies regime distinctions; Causality and Global Structure supplies causal relations; and Computational Relativity supplies numerical estimator checks.

Locally finite order

Use an irreflexive relation $\prec$ on a countable set $C$. It obeys

$$x\not\prec x,\qquad x\prec y\ \text{and}\ y\prec z\Longrightarrow x\prec z,$$

and local finiteness requires

$$\lvert I(x,z)\rvert\lt\infty, \qquad I(x,z)=\{y\in C:x\prec y\prec z\}$$

whenever $x\prec z$. Irreflexivity plus transitivity also forbids causal cycles. A link $x\prec^\ast y$ is a relation with no intervening $z$; links form the Hasse diagram, while their transitive closure recovers the causal order. A chain is a totally ordered subset and models a timelike sequence. An antichain contains no related pair and is a discrete analogue of a spacelike set.

The definition is label independent. Relabeling elements changes a matrix representation or drawing but not the causal set. Local finiteness supplies a microscopic replacement for finite spacetime volume, yet it imposes no fixed valence: a sprinkled element can have infinitely many links in an unbounded Lorentzian spacetime.

Faithful embedding and Poisson sprinkling

An injective map $\phi:C\to(M,g)$ is faithful at density $\rho$ only if it preserves causal order for distinct elements,

$$x\prec y\quad\Longleftrightarrow\quad \phi(x)\in J^-\!\left(\phi(y)\right), \qquad x\ne y,$$

and the image has the density and statistical properties of a Poisson sprinkling, with geometry varying slowly on the discreteness scale $\rho^{-1/d}$. The restriction $x\ne y$ is necessary because the continuum relation $J^-$ includes its base point, whereas $\prec$ is irreflexive. For a measurable region of spacetime volume $V$,

$$\Pr(N=n)=e^{-\rho V}\frac{(\rho V)^n}{n!}, \qquad \mathbb E[N]=\rho V, \qquad \operatorname{Var}(N)=\rho V.$$

Thus $N/\rho$ estimates volume with fractional Poisson noise $N^{-1/2}$. Because the probability depends only on invariant volume, sprinkling into Minkowski space selects no preferred inertial lattice. Individual realizations are irregular, but their law is Lorentz invariant Bombelli et al. 1987, pp. 521–523.

For time-oriented, past- and future-distinguishing Lorentzian spacetimes in dimension greater than two, the chronological structure fixes the metric up to a conformal factor; a volume measure fixes that remaining scale Malament 1977, Theorem 2. “Distinguishing” means that distinct events have distinct chronological pasts and distinct chronological futures. This continuum theorem motivates “order plus number.” It does not prove that every partial order has a unique continuum approximation. The causal-set Hauptvermutung—that one causal set cannot faithfully approximate two macroscopically inequivalent spacetimes—remains open in the required generality Surya 2019, §2.3.

Dimension and distance estimators

Let an Alexandrov interval contain $N$ points, and let $R$ be the number of related unordered pairs. Its ordering fraction is

$$r=\frac{2R}{N(N-1)}.$$

For a high-density sprinkling into a $d$-dimensional flat Alexandrov interval,

$$\mathbb E[r]= \frac{\Gamma(d+1)\Gamma(d/2)} {2\Gamma(3d/2)}.$$

It equals $1/2$ for $d=2$ and $1/10$ for $d=4$. Inverting this monotone relation gives the Myrheim–Meyer dimension estimator. Longest-chain length scales as $(\rho V)^{1/d}$ and estimates timelike separation after a dimension-dependent calibration. Counts of small order intervals support curvature and discrete d’Alembertian estimators. Definitions and calibration regimes are reviewed in Surya 2019, §§2.4–2.7.

Several independent estimators should agree over a scale window. At small $N$, Poisson and order fluctuations dominate. Near curvature radii, flat calibration fails. A Kleitman–Rothschild-type or highly layered order may have a numerical ordering fraction resembling some dimension while failing locality, interval abundance, or topology tests. Dimension is therefore a diagnostic, not a manifoldlikeness theorem.

Checks and boundaries

The dimensions are consistent: $[\rho]=L^{-d}$, $[V]=L^d$, and $N$, $R$, and $r$ are dimensionless. In the dense flat limit, relative volume fluctuations vanish as $N^{-1/2}$. In the opposite limit $N=0$ or $1$, $r$ is undefined because there are no pairs; assigning a dimension would be false precision.

Order reversal $x\prec y\mapsto y\prec x$ preserves interval cardinalities and dimension estimates but reverses the time orientation encoded by the order. These counting diagnostics alone therefore cannot recover its direction. Topological information is subtler still: it can be inferred from suitable thickened antichains under manifoldlike assumptions, not from one antichain alone.

These are kinematical statements. A dynamics must explain why manifoldlike orders receive appreciable weight, how quantum interference is defined, and why an Einsteinian continuum phase emerges. Classical sequential growth, quantum measures, and causal-set actions address pieces of that next problem.

Exercises

An interval contains $N=5$ elements and $R=5$ related unordered pairs. Compute its ordering fraction and compare it with the flat $d=2$ and $d=4$ expectations. If it arose by sprinkling at density $\rho=20L^{-2}$, estimate its area and one-standard-deviation Poisson uncertainty. State why the answer does not prove the interval is two-dimensional Minkowski space.

Solution

The ordering fraction is

$$r=\frac{2R}{N(N-1)} =\frac{10}{20}=\frac12,$$

which matches the mean flat two-dimensional value and differs from the four-dimensional value $1/10$. The volume estimate is $\widehat V=N/\rho=0.25L^2$. The exact sampling standard deviation is $\sqrt{\rho V}/\rho$; replacing the unknown $\rho V$ by the observed count gives the plug-in estimate $\widehat\sigma_V=\sqrt5/(20L^{-2})\simeq0.112L^2$. At this small count, this is not an exact symmetric 68-percent confidence interval. Moreover, one statistic does not test interval abundances, topology, locality, curvature, or faithful embedding; many nonmanifoldlike orders can accidentally give $r=1/2$.

References

• Bombelli, Luca, Joohan Lee, David Meyer, and Rafael D. Sorkin. “Space-Time as a Causal Set.” Physical Review Letters 59 (1987): 521–524. DOI.
• Malament, David B. “The Class of Continuous Timelike Curves Determines the Topology of Spacetime.” Journal of Mathematical Physics 18 (1977): 1399–1404. DOI.
• Surya, Sumati. “The Causal Set Approach to Quantum Gravity.” Living Reviews in Relativity 22 (2019): 5. Open review.