Causal Set Dynamics and Sequential Growth

Classical sequential growth asks how a causal set can be built one maximal element at a time without making birth labels physical. Internal temporality, normalization, discrete general covariance, and a Bell-type spectator condition constrain the transition probabilities to a family governed by nonnegative couplings. This is a classification of covariant stochastic growth laws. It is neither a completed quantum dynamics nor evidence that typical histories approximate four-dimensional general relativity.

Helpful background. Flat-Spacetime Quantum Field Theory supplies amplitudes and continuum limits; Einstein Equations supplies the target dynamics; Quantum Information supplies interference and decoherence functionals; Quantum Gravity supplies evidence standards; Causality and Global Structure supplies causal order; and Computational Relativity supplies finite-ensemble checks.

Covariant growth conditions

Begin with the empty causal set. At stage $n\to n+1$, add a new maximal element $x$. Its precursor $P\subseteq C_n$ must be a down-set: if $y\in P$ and $z\prec y$, then $z\in P$. This guarantees transitivity without changing relations among existing elements.

Four conditions constrain a Markov law.

Internal temporality permits the newborn to lie to the future of, or be unrelated to, an existing element; it cannot lie to its past. Markov normalization makes the probabilities of all allowed next births sum to one. Discrete general covariance requires the product of transition probabilities along a naturally labeled birth history to be independent of the natural labeling of the same final order, so birth order is gauge. Bell causality requires ratios of two competing transition probabilities to be unaffected by spectator elements belonging to neither precursor. The last condition is a causal screening principle, not the Bell-locality condition of laboratory quantum theory.

Rideout and Sorkin show that, with mild nonvanishing qualifications, these conditions yield classical sequential growth models parameterized by couplings $t_k\ge0$ Rideout and Sorkin 2000, §§II–IV. Set $t_0=1$ to remove an overall scale and define

$$\lambda(\varpi,m)= \sum_{k=m}^{\varpi} {\varpi-m\choose k-m}t_k.$$

If an elementary birth from an $n$-element causet uses a specified precursor of cardinality $\varpi$ with $m$ maximal elements, its probability is

$$\Pr(C_n\longrightarrow C_{n+1}) =\frac{\lambda(\varpi,m)}{\lambda(n,0)}.$$

The formula depends on order-theoretic data, not the names of labels. To obtain the probability of an unlabeled child isomorphism class, sum over the distinct precursors that produce it. Those multiplicities matter: the two singleton precursors of a two-element antichain produce the same unlabeled child but are separate elementary births. Different nonnegative sequences $\{t_k\}$ give genuinely different dynamics; covariance does not select one.

A finite normalization check

For a one-element causet, the new element can be unrelated, with precursor $\varnothing$, or above the old element. Their weights are $t_0$ and $t_1$, and $\lambda(1,0)=t_0+t_1$ normalizes them.

For a two-element antichain, allowed precursors are $\varnothing$, either singleton, and the full pair. Their unnormalized weights are

$$t_0,\qquad t_1,\qquad t_1,\qquad t_2,$$

whose sum is $\lambda(2,0)=t_0+2t_1+t_2$. For a two-chain, the down-sets are $\varnothing$, the bottom singleton, and the full chain. Their weights are $t_0$, $t_1$, and

$$\lambda(2,1)=t_1+t_2,$$

giving the same denominator. This checks normalization in two distinct orders without referring to their labels.

The down-set assumption is discriminating. If the top element of a two-chain were allowed as a precursor without its bottom ancestor, transitive closure would nevertheless place the newborn above both. That illegal subset would double-count the same unlabeled transition and spoil the normalization just checked. Thus internal temporality plus transitivity is structural, not a convenient enumeration rule.

A familiar subfamily is transitive percolation, $t_k=t^k$. Then

$$\lambda(\varpi,m)=t^m(1+t)^{\varpi-m}, \qquad \lambda(n,0)=(1+t)^n.$$

It is exactly soluble enough for checks, but typical orders at generic parameters need not be manifoldlike.

From probabilities to quantum dynamics

A quantum sum over causal sets cannot be obtained by replacing every classical transition probability with its square root. Quantum interference is naturally described by a decoherence functional $D(A,B)$ on sets of histories, with quantum measure $\mu(A)=D(A,A)$. For three disjoint events it obeys the grade-two rule

$$\mu(A\cup B\cup C)= \mu(A\cup B)+\mu(A\cup C)+\mu(B\cup C) -\mu(A)-\mu(B)-\mu(C),$$

rather than ordinary additivity Sorkin 1994, §§2–3. Positivity, label invariance, extension to infinite histories, and a physically useful event algebra must all be controlled.

A second route assigns an action. The Benincasa–Dowker construction combines counts of small order intervals so that its sprinkling expectation approximates the scalar-curvature term in a suitable continuum regime Benincasa and Dowker 2010, Eqs. (3)–(7). A formal sum

$$Z=\sum_{[C]}\frac{1}{\lvert\operatorname{Aut}(C)\rvert} e^{iS_{\mathrm{cs}}[C]}$$

still needs a definition of the class of orders, convergence or contour prescription, measure, boundary data, and continuum limit. The enormous entropy of nonmanifoldlike orders is an adverse competitor to action suppression.

Results and open questions as of 2026-08-20

Approach or position	Strongest controlled result and source	Scope, adverse result, or unresolved question
Classical sequential growth	Covariance, Bell causality, and Markov assumptions classify a nonnegative $t_k$ family; Rideout and Sorkin 2000, §§II–IV	Conditions do not select couplings, give quantum interference, or make typical causets four-dimensional
Transitive percolation and posts	Exact stochastic subfamily and analytically tractable large-scale questions; Surya 2019, §4.1	Generic histories can be nonmanifoldlike; cosmological interpretation depends on special conditioning
Causal-set action	Interval-counting operators recover continuum d’Alembertian/curvature expectations under sprinkling assumptions; Benincasa and Dowker 2010, Eqs. (3)–(7)	Fluctuations, nonlocality scale, entropy, Lorentzian sum, and boundary prescription remain
Quantum-measure programme	Grade-two interference gives a covariant generalization of probability; Sorkin 1994, §§2–3	No unique full causal-set decoherence functional or accepted continuum phenomenology is selected

The permissive position treats classical covariance, a causal-set action, and quantum-measure theory as compatible pieces of a future microscopic dynamics. The conservative position regards the growth theorem as complete only for its classical axioms and requires a normalized quantum dynamics with manifoldlike dominance and Einstein observables. A skeptical position emphasizes entropic dominance by nonmanifoldlike orders. These interpretations are not equally evidenced: the finite classical classification is rigorous, while the four-dimensional quantum continuum remains open.

The conclusions could change after a consistent quantum sequential-growth classification, a controlled Lorentzian sum with manifoldlike dominance, a continuum limit reproducing Einstein correlations, a counterexample to the covariance or Bell-causality assumptions, or evidence after 2026-08-20 changes the comparison.

Exercises

Take $t_0=1$, $t_1=a$, and $t_2=b$, with $a,b\ge0$. List the transition probabilities from a two-element antichain and verify normalization. What happens if $b=0$?

Solution

The denominator is $\lambda(2,0)=1+2a+b$. The empty precursor has probability $1/(1+2a+b)$; each of the two singleton precursors has probability $a/(1+2a+b)$; and the full two-element precursor has probability $b/(1+2a+b)$. Their sum is

$$\frac{1+a+a+b}{1+2a+b}=1.$$

When $b=0$, a newborn cannot have both unrelated existing elements as its complete precursor in this stage, though singleton and empty births remain possible if $a\gt0$. This is a genuine dynamical restriction, not a gauge choice: covariance removes birth labels but does not erase the coupling dependence.

References

• Benincasa, Dionigi M. T., and Fay Dowker. “The Scalar Curvature of a Causal Set.” Physical Review Letters 104 (2010): 181301. DOI.
• Rideout, David P., and Rafael D. Sorkin. “A Classical Sequential Growth Dynamics for Causal Sets.” Physical Review D 61 (2000): 024002. Open manuscript.
• Sorkin, Rafael D. “Quantum Mechanics as Quantum Measure Theory.” Modern Physics Letters A 9 (1994): 3119–3127. Open manuscript.
• Surya, Sumati. “The Causal Set Approach to Quantum Gravity.” Living Reviews in Relativity 22 (2019): 5. Open review.