Signature from order and counting
The construction must obtain Lorentzian signature and null structure from a causal order plus element counting, or from a stated equivalent. Additive path cost is disqualified by construction, not by degree of fit.
Home / Relativity From Adjacency
Relativity From Adjacency
Suppose you are handed nothing but a discrete relation saying which elements can influence which. No space, no distance, no metric, no clock. Can flat spacetime be recovered from that alone? The answer is a qualified yes, but only through one specific construction, and the obvious construction is not it.
Exploratory
Physics Note
Special Relativity Only
Scope: flat-space relativistic kinematics. Gravitation is out of scope on this page and is not implied by anything on it.
The scope, stated first
This page asks one narrow question and refuses the larger one. The narrow question is whether the kinematic core of special relativity can be reconstructed from a discrete adjacency relation: the interval, the light cone, proper time, and the absence of a preferred frame. The larger question, whether general relativity can be reconstructed the same way, is not asked here and is not answered by anything below.
The reason for the wall is not modesty. General relativity says geometry is dynamical and sourced by stress-energy. An adjacency relation carries no energy, and nothing in the framework's substrate supplies one: asymmetric admissibility is explicitly not energy, force, or drive, and the framework's mass-like descriptor is explicitly stated not to interact gravitationally. With no source term available, there is nothing for a curvature equation to be sourced by. A gravitational reconstruction would therefore need machinery this page does not have, and claiming otherwise would be the single easiest place for a reader to catch the project overreaching.
Why the scope line matters. Recovering flat Lorentzian structure from a discrete substrate is already an open research problem after four decades of work. Reaching past it to gravitation, before the flat case is closed, converts a testable program into an untestable one.
The obvious construction fails, and it fails twice
The natural first move is to define distance as a shortest path. Assign each adjacency step a non-negative cost, then take the minimum total cost over all paths between two elements. It is intuitive, it is computable, and it is wrong.
Dead end 1: signature
A sum of non-negative step costs is positive definite by construction. It can never produce the minus sign in the relativistic interval, and it has no null directions, so the light cone simply does not appear. No amount of tuning the cost function fixes this, because the defect is in the form of the construction rather than in its parameters. Any proposal that defines separation as an additive path cost is disqualified before its details are examined.
Dead end 2: spacelike separation
Even setting signature aside, shortest-path constructions misbehave badly for spacelike pairs. In a discrete substrate dense enough to approximate a continuum, two spacelike-separated elements have common ancestors and common descendants nearly everywhere, so short connecting chains exist regardless of how far apart the pair is supposed to be. The construction reports near-adjacency for elements that should be remote.
Both failures point the same direction. Separation is not the primitive thing here. Influence is.
Order gives the geometry; counting gives the scale
The construction that works takes the relation seriously as a causal order rather than as a notion of nearness. Read the adjacency relation as saying which elements can influence which, and it becomes a partial order. Read it as saying which elements are close together, and it becomes a lattice, which is the reading that fails.
Under the causal reading, two established results do the work. The first is that the causal structure of a spacetime determines its metric up to a single overall conformal factor, together with its topology and differential structure. The light cone is not derived from the geometry; the light cone very nearly is the geometry, with one scalar left undetermined. The second is that the remaining factor is fixed by counting: the number of discrete elements in a region supplies its volume, and volume is exactly the information the conformal factor withholds. Sorkin's compression of this is hard to improve on: order plus number equals geometry.
Lorentzian signature then arrives for free rather than being engineered. A causal order has a future, a past, and a set of elements comparable to neither. Those three regions are the timelike future, the timelike past, and the spacelike elsewhere. The null cone is the boundary of the order's reach. Nothing has to be arranged to produce the minus sign, because the minus sign was never a feature of a distance function in the first place; it is a feature of a relation that divides the world into three parts instead of one.
Proper time is a count
This is the result most directly aligned with the framework's own instincts, and it deserves to be stated without softening.
In a discrete causal substrate, the timelike separation between two elements is the length of the longest chain connecting them: the maximum number of elements in a totally ordered sequence running from one to the other. Not the shortest path. The longest chain. Its asymptotics recover proper time in the continuum limit, which is why the result matters rather than merely being cute.
The reason is worth seeing directly. In relativity, the straight worldline between two events is the one that maximizes proper time, not the one that minimizes length; every detour costs you elapsed time rather than adding it. Chain length is the discrete shadow of that maximization. A clock, on this reading, is not an instrument measuring a background parameter. A clock is a counter, and proper time is the tally of how many elements a worldline actually passed through.
The framework has said for a long time that ordering is generated by the process rather than laid down in advance, and that a tik is a unit of sequence rather than a unit of duration. The longest-chain result is what that stance looks like when it is made to answer to measured physics: proper time as accumulated count, with the continuum limit as the thing to be recovered rather than assumed.
Lorentz invariance forbids a regular lattice
Here the reconstruction meets its sharpest constraint, and it is the one most often gotten wrong.
A regular lattice cannot be Lorentz invariant. Boost it and the spacing contracts along the direction of the boost, so the lattice looks different to different observers, which means it selects a preferred frame. Since Lorentz violation is tightly constrained by observation, a regularly spaced discrete substrate is not a viable model of flat spacetime. This kills the intuitive picture of spacetime as a fine grid, and it kills it for a reason that has nothing to do with how fine the grid is.
The resolution is that discreteness must be obtained statistically rather than regularly. If elements are scattered by a random process at uniform density, the resulting structure has no characteristic direction and no characteristic spacing to contract. Its distribution is invariant under boosts, and there is a theorem to the effect that no preferred frame can be extracted from such a configuration by any procedure that respects the symmetry. Discreteness and Lorentz invariance are compatible, but only through randomness. Regularity is what breaks the symmetry, not discreteness.
This bears on how the framework's own adjacency constraint should be read. The canonical text says influence is restricted to “bounded neighbor relations” and establishes “finite locality of effect,” while stating that the constraint is not space, distance, metric, geometry, topology, or continuity.
A Lorentz-invariant discrete structure is bounded in measure, meaning finite volume per element, but unbounded in count, meaning no finite limit on how many elements are adjacent to a given one. Whether “bounded” admits the measure reading is an open question about the mapping and is not settled on this page. It is flagged here rather than assumed, because assuming it is exactly the kind of quiet substitution the framework's own review discipline exists to catch.
The consequence nobody wants: locality does not survive
Statistical discreteness buys Lorentz invariance at a price, and the price is severe enough that it should be stated plainly rather than buried.
If elements are scattered in a Lorentz-invariant way, then the elements nearest to a given element in the causal order do not sit in a small ball around it. They sit along a hyperboloid, and there are unboundedly many of them, because the hyperboloid has infinite volume while the density is uniform. Each element therefore has infinitely many nearest neighbors in the infinite-volume limit. This is not a pathology to be engineered away; it is a direct consequence of insisting on Lorentz invariance in a discrete setting.
The practical symptom is that the discrete wave operator built on such a structure is radically non-local, and recovering local physics from it requires damping or averaging that has to be justified rather than assumed. So the honest summary of the flat-space reconstruction is inverted from what one would expect. Discreteness does not deliver locality. Discreteness plus Lorentz invariance delivers non-locality, and locality is the thing that has to be recovered as an approximation afterward.
What maps, and what does not
| Relativistic notion | Discrete counterpart | Status |
|---|---|---|
| Light cone | The boundary of the causal order's reach | Recovered, and structural rather than derived |
| Metric, up to a conformal factor | The causal order itself | Recovered by established theorem |
| The remaining conformal factor | Element count as volume | Recovered by counting |
| Proper time | Longest chain between two elements | Recovered, with continuum asymptotics established |
| No preferred frame | Statistical, not regular, discreteness | Recovered, and only this way |
| Relativity of simultaneity | No preferred maximal antichain | Follows from the above |
| Spacelike distance | Not supplied by the order | Open. Shortest-path constructions misbehave |
| Locality of the wave operator | Unboundedly many nearest neighbors | Open. Non-locality is generic, not incidental |
| Curvature, field equations | No source term available | Out of scope on this page |
The two open rows are the honest state of the flat-space reconstruction. A proposal that reports success on the timelike structure while staying silent on spacelike separation and on the non-locality has addressed the easy half of the problem. Silence on either row is an incomplete account rather than a passing one.
What would count as success
Proper time as a count
Timelike separation must be recovered as a count along the order, with the continuum limit demonstrated rather than asserted. Getting the maximization direction wrong, by minimizing where relativity maximizes, is a failure.
Frame independence, demonstrated
Discreteness must be statistical, and the account must show that no preferred frame is extractable by any symmetry-respecting procedure. A regular structure fails here regardless of its other virtues.
The two open rows, addressed
Spacelike separation and the non-locality of the discrete wave operator must both be confronted directly. These are where the flat-space program is genuinely unfinished, and they are not optional to discuss.
Quantum questions, including whether such a substrate can reproduce the observed correlation ceiling rather than merely exceeding the classical one, are a separate matter and are not treated on this page.
Where this comes from
Nothing on this page is original to the framework, and saying so is the point. The flat-space reconstruction is an established research program with a literature, and any adjacency-based proposal has to engage it rather than rediscover it. The load-bearing results above:
- Causal structure determines the metric up to a conformal factor, along with topology and differential structure. Hawking, King and McCarthy (1976); Malament (1977).
- Order plus number equals geometry: discrete element count supplies the volume information the conformal factor omits. Bombelli, Lee, Meyer and Sorkin (1987), which introduced causal sets.
- Timelike separation as longest chain, with continuum asymptotics. Brightwell and Gregory (1991).
- Discreteness without symmetry breaking: a uniform random scattering is Lorentz invariant in distribution, and no preferred frame can be extracted from it equivariantly. Bombelli, Henson and Sorkin (2009).
- Dimension estimation from how reachability grows with scale, which is the rigorous form of the intuition that dimension should be measured rather than imposed. Myrheim (1978); Meyer (1988).
The framework's contribution here is not the mathematics. It is the insistence that the substrate constraint be stated before the geometry, and that anything geometric be earned downstream rather than assumed at the start.
This is an exploratory physics note, not canon. It reports established results from an existing research literature and asks how a discrete adjacency constraint would have to behave to meet them. It defines no construct, modifies no canonical definition, and settles no open question about how the canonical adjacency constraint should be read. Conflicts are resolved in favor of the canonical text and governance pages.