Path 5: Causal Set Theory

Spacetime as a network of events: points with only before/after relations.

Inventory

Rationale: Causal set theory posits that spacetime is fundamentally a discrete set of elementary “events” partially ordered by causality (i.e. for any two events, either $x \prec y$ (x comes before y) or they are unrelated). This simple poset structure encodes all that is needed to approximate a continuous spacetime: the number of elements corresponds to volume and the order gives the light-cone structure. If one “sprinkles” points randomly (with a Poisson process) into a 4D continuum, with the order inherited from the continuum’s light-cones, one gets a causal set that, for large number of points, has high probability of approximating that continuum. The hope is that at Planck scale, the universe is such a causal set (not a continuum), thereby automatically avoiding infinities (no infinitely small regions) and explaining deep puzzles like: Why does spacetime have 4 dimensions? (Answer: maybe only 4D causal sets produce continuum-like behavior when large.) Why is the cosmological constant so small? (Perhaps a large causal set’s dynamics lead to a fluctuating “effective $\Lambda$” that averages out small.) Causal sets bring a genuine atomistic view of spacetime: “atoms” of spacetime with only causality and number exist, and geometry (distance, curvature) is an emergent, approximate concept.

Prerequisite Themes: Basic order theory and graph theory (to understand partial orders, transitive relations, etc.), Lorentzian geometry (so as to connect order to light-cone structure), and some relativity (especially concepts of time, light-cones, and maybe Lorentz invariance – remarkably, a random sprinkling preserves local Lorentz invariance on average). Also, some quantum mechanics and perhaps knowledge of simple quantum transition amplitudes because one wants eventually to sum over or compute dynamics on these sets.

Dependencies: Causal set theory is philosophically close to Path 2 (LQG) in spirit (discreteness, background independence) but technically very different. It doesn’t need the continuum like Path 4 (CDT) does at the start; instead, it builds upward from discreteness. Dynamically, one avenue is the “sequential growth” models by Rideout and Sorkin, where the causal set grows one element at a time with certain probabilities (maintaining causality) – this is a unique dynamical law not obviously related to other approaches, though one can attempt to derive an action principle (Sorkin proposed a discrete analog of the Einstein action for causal sets). There are light overlaps with Path 7 (Emergent): one might see a causal set as a kind of emergent structure from an even deeper level (some have considered quantum causal sets or combining causal sets with quantum information ideas). It’s relatively self-contained, though to connect with reality it would eventually need to show how fields and particles live on the causal set (work is done on defining a “d’Alembertian” operator on a causal set for a scalar field, for example).

Signs of Progress: The first big sign was showing that a causal set can produce near-continuum behavior. Ongoing work includes calculating the “spectral dimension” of a causal set (an indicator of effective dimension at various scales) – a successful theory should show 4 at large scales, which preliminary simulations have indicated. Another positive sign would be if causal sets can naturally suppress the formation of curvature singularities (so far, no “spacetime” of a causal set is perfectly singular in the GR sense, because the fundamental structure is different). If causal sets can generate phenomena like cosmological expansion or even approximate FRW cosmologies when the set is grown in certain ways, that would be a striking validation. On the flip side, making contact with concrete physics – e.g. deriving the neutron star mass limit or gravitational wave propagation from a causal set model – remains distant; any step in that direction (like successfully defining and simulating how gravitational waves might disperse or not in a causal set) would mark progress. Because it’s a radical approach, even disproving it decisively (e.g. showing that no continuum can come out in the large-scale limit except in wrong dimensions or with unwanted behavior) would be important. As long as it isn’t disproved, the main progress markers are theoretical consistency and approximate agreement with known physics emerging from the simplicity of “order + number.”

Base Camp 5.1: Partially Ordered Sets and Causal Structure

Scope: Brush up on order theory: a poset is a set with a partial order $\prec$ that is transitive, antisymmetric, etc. Understand how in a Lorentzian spacetime, events with a time-ordering define a natural partial order (causal order: $x \prec y$ if $x$ is in the causal past of $y$).

Stepping Stones: (1) Basic definitions: a poset, chains (totally ordered subsets), antichains (no two elements comparable). Understand that a chain in a causal set might correspond to a timelike path, an antichain to a set of spacelike-separated events (a “slice”). (2) The concept of past and future sets: for any element, define $I^-(x) = \{y: y \prec x\}$ and similarly future. This is like the causal past of an event. (3) Note that not every poset corresponds to a nice spacetime – we restrict to those that are locally finite (meaning no element has an infinite number of predecessors in a finite interval, which in practice means no accumulation of points, matching the idea of a discrete spacetime). (4) Introduce the concept of height or rank if any (like levels in a layered poset, though a general causal set doesn’t have a global notion of time without extra structure – but can define a longest chain, etc.).

Base Camp 5.2: Basics of Causal Set Theory

Scope: State the foundational conjecture: “Order + number = geometry.” A causal set is a locally finite poset that, if suitably “sprinkled” into a continuum, reproduces the continuum’s causal structure and volume (cardinality) approximately. Understand how one can approximate a continuum by randomly selecting points in it (Poisson process) to become the causal set, thus capturing both the order (causal) and density (volume).

Stepping Stones: (1) Define a causal set (causet) formally: $(\mathcal{C}, \prec)$ where $\prec$ is a partial order that is (i) acyclic (no $x \prec x$ through a loop), (ii) transitive, (iii) locally finite (for any $x \prec z$, the set $\{y | x \prec y \prec z\}$ is finite). (2) Explain random sprinkling: for a given spacetime region of volume $V$, take $N$ points according to Poisson distribution with mean density $\rho$ so that $N \approx \rho V$. For each pair of points, relate them by $\prec$ if one is to the past of the other in the continuum. The resulting random poset is one candidate for a causal set approximating that spacetime. (3) State that in the limit of dense sprinkling ($\rho \to \infty$ and points $\to$ continuum), the probability distribution of the poset yields with high probability the same causal relations as the continuum, and volume is recovered because number of elements $\approx \rho V$. (4) Emphasize that causal set theory takes this approach as fundamental: spacetime is fundamentally such a discrete poset, not embedded in any continuum (the continuum is an emergent approximation for large number). (5) Note one of key properties: randomness in sprinkling ensures Lorentz invariance in expectation – there’s no preferred lattice, unlike a regular grid which breaks Lorentz invariance, a random Poisson sprinkling is statistically invariant because distribution is same in any Lorentz frame (Poisson process is parameterized by volume only, which is Lorentz-invariant concept).

Base Camp 5.3: Reconstructing Spacetime Properties from Causal Sets

Scope: Given only the poset, how to define approximate dimension, volumes, distances? This covers the idea of calculating the degree distribution or abundance of certain sub-orders to get dimension, and the concept of distance via longest chain approximating proper time.

Stepping Stones: (1) Define dimension estimation: one method is the Myrheim–Meyer dimension, using the ratio of $M_2$ (number of pairs) to $M_3$ (number of triples) or similar, derived from expectation in continuum. Another is using the spectral dimension (take the causal set’s graph of links within certain reach and run a random walk). (2) Rough idea: In $d$ dimensions, a region of volume $V$ has expected number of elements $n=\rho V$. Also the number of relations (comparable pairs) grows as $n^{2/d}$ roughly (because volume of causal intervals scales with time separation in a way related to dimension). So by counting relations one can solve for $d$. (3) Volume from counting: trivial, $|I^-(x)|$ is number of elements in past of $x$, which for a sprinkling corresponds to continuum volume of past of $x$ times density. So volume, up to fluctuations, can be recovered. (4) Distance: define Alexandrov interval in a causal set as $\{z | x \prec z \prec y\}$ for a causal pair $(x,y)$. The longest chain between $x$ and $y$ (the largest set of elements $x = e_0 \prec e_1 \prec ... \prec e_k = y$) corresponds to the longest proper time path. In a causal set that came from a flat continuum, the chain length $\approx \tau (x,y)/\delta$ where $\delta$ is roughly the discreteness scale. For curved, one might approximate proper time by chain length times some average spacing. (5) These reconstructions show that in principle, all geometric info is latent in the poset. It’s difficult in practice to extract curvature etc directly, but ongoing work tries to define e.g. Ricci scalar from counting how many elements in a certain distance away (like use layered sets as an analog to small volumes).

Base Camp 5.4: Dynamics of Causal Sets (Growth Models)

Scope: One big question: how does the causal set evolve? Since it’s not embedded in a continuum, we need a rule for adding elements that ideally recovers something like Einstein’s equations in expectation. Learn about the sequential growth models by Rideout and Sorkin: elements are born one-by-one, choosing a past link structure with certain probability, under conditions of causality and bell causality.

Stepping Stones: (1) The concept of a Markov process of growing a poset: start empty, at each step add a new maximal element (one with no future yet). The probability of attachment (which existing elements are in its past) is governed by some coupling constants that we can interpret as related to “cosmological constant” etc. (2) Rideout-Sorkin model: new element chooses to link to each existing element independently with probability $p$, such that the result is transitive closure of those links. Actually, the simplest model (the transitive percolation or uniform model) corresponds to a zero cosmological constant solution where any two orders are equally likely. With a parameter one can favor having certain numbers of relations – this parameter can be thought to tune the cosmological constant. They found a family of models labeled by a coupling $p$ that produce different “cosmic expansion” behaviors (like number of elements as function of layer). (3) A condition called “Bell causality” or “internal temporality”: it basically ensures the probability rule doesn’t depend on anything but the structure to which we attach (no hidden variables). Combined with “covariance” (no preferred labeling of events aside from order) restricts models strongly – basically yields the sequential growth approach. (4) Discuss what these models achieve: they produce causal sets that, in continuum limit, could approximate cosmological spacetimes (e.g. the preferred model yields dimension 4 and de Sitter-like expansion at large $p$). However, so far no model uniquely stands out as Einstein’s eq analog – that’s open.

Base Camp 5.5: Phenomenology and Particles in Causal Sets

Scope: See how causal sets could in principle make contact with physics: e.g. the idea of a fluctuating cosmological constant (Sorkin’s observation that in a finite causal set, $\Lambda$ gets an uncertainty $\sim 1/\sqrt{N}$), or how matter fields might be embedded (nontrivial, since fields usually require additional structure, but e.g. a scalar field could be a function on the set).

Stepping Stones: (1) Fluctuations of $\Lambda$: since volume = number, a finite Poisson process gives Poisson fluctuations in $N$ around expectation, leading to an induced uncertainty in measured $\Lambda$. Sorkin pointed this as a way quantum gravity might explain observed small $\Lambda$ as an averaging effect. (2) Spectral dimension running was considered (some predictions like a bounce from 2 at small scales – in causal sets actually the dimension is fixed by construction but one can see effective diffusions. If time: mention that CDT saw spectral dimension ~2 at Planck scale, causal sets generically have spectral dimension ~2 in the UV due to underlying discreteness scale – a nice agreement across approaches). (3) Matter coupling: a big open area, but e.g. to put a scalar field, one can define a discretized d’Alembertian on the causal set (various proposals exist, using the fact that in continuum $\Box f(x)$ can be approximated by integrals over small volumes – translate to sums over causal set neighborhoods). Dowker et al. have constructed an operator that converges to continuum $\Box$ when the causet approximates Minkowski. This allows defining field equations on a causet. (4) Mention any experimental search: The main suggestion has been the “cosmic rain” of causet discreteness signals – e.g. the fluctuations in arrival times of cosmological photons (unlike a crystal lattice which yields scattering, a random lattice yields noise). There is a model by Dowker for “swerves” where massive particles undergo a kind of diffusion in momentum due to underlying discrete structure. They try to constrain that with cosmic ray propagation data, etc. (5) Causal sets have not produced a falsified prediction yet; rather they give an overall plausible framework where one might interpret dark energy’s smallness as due to large number of atoms of spacetime (~Poisson $\frac{1}{\sqrt{N}}$ rule).

← Back to Gasherbrum I – Quantum Gravity