The geometry of spacetime is forced

What classical mechanics says

Classical mechanics combines velocities by adding them. If I am on a train moving at fifty kilometers per hour and I throw a ball forward at ten, an observer on the platform sees the ball fly past at sixty. The arithmetic is so basic that it does not feel like physics.

Apply the same idea to light. If a flashlight on the train emits a beam forward, the platform observer should see the beam at light-speed plus the train’s speed. The Michelson-Morley result says the platform observer sees the beam at exactly light-speed, not a hair faster. The arithmetic of plain addition is wrong about light, which means the arithmetic of plain addition was never quite right about anything. It was approximately right because in everyday life nothing moves anywhere near the speeds at which the discrepancy would be noticeable. Earth’s orbital speed of 30 kilometers per second sounds enormous in human terms, and it is one ten-thousandth of the speed at which the discrepancy becomes obvious.

So either the experiment was wrong, or the way we add velocities was wrong. By the early 1900s, the experiment had been confirmed often enough that physicists had to take the second option seriously. The question was not whether classical addition broke. The question was what would replace it.

The geometry was never optional

In 1905, Einstein took the speed-of-light invariance as a postulate rather than something to be derived, and he asked what mechanics has to look like if that postulate is true (Einstein 1905). The answer turned out to be geometric. Hermann Minkowski put the geometry in its cleanest form three years later (Minkowski 1908). Spacetime is a four-dimensional manifold whose distance function differs from Euclidean distance by a single minus sign. That minus sign is what does all the work.

Let me show what I mean.

In ordinary Euclidean geometry, the distance between two points in space is given by the Pythagorean theorem. If I label the separation along three perpendicular axes as \(\Delta x\), \(\Delta y\), \(\Delta z\), the squared distance is

\[ \Delta s^2 = \Delta x^2 + \Delta y^2 + \Delta z^2. \]

This is the metric [the rule that tells you how far apart two points are in a given geometry] of three-dimensional space. It does not care which way you orient your coordinate system. Rotate the axes in any direction and the formula stays the same, because a distance is a distance regardless of how you label its components. That coordinate-independence is the geometric fact that makes Pythagoras a real fact about space rather than a quirk of one particular drawing.

To extend this to spacetime, the natural move is to add a fourth term for the time separation. There are two ways to do it. The first is to add a plus sign and treat time as just another spatial dimension. That gives you a four-dimensional Euclidean space, and physics built on top of it would look very different from the physics we observe. The second is to put a minus sign in front of the time term, treating time as a dimension of the opposite kind. That gives you the Minkowski metric [the spacetime distance function, with one negative term for time and three positive terms for space]:

\[ \Delta s^2 = -c^2 \Delta t^2 + \Delta x^2 + \Delta y^2 + \Delta z^2. \]

The factor of c is there to make the units match; the minus sign is the load-bearing structural difference. Most ways of writing down a four-dimensional metric would put a plus there. Minkowski put a minus, and the experiment is what forced him to.

Here is why. Imagine a light pulse leaving the origin at time zero and arriving at some point a time \(\Delta t\) later, having traveled a spatial distance \(r = \sqrt{\Delta x^2 + \Delta y^2 + \Delta z^2}\). The pulse moves at speed c, so \(r = c\,\Delta t\). Plug into the Minkowski formula:

\[ \Delta s^2 = -c^2\,\Delta t^2 + r^2 = -c^2\,\Delta t^2 + c^2\,\Delta t^2 = 0. \]

The interval for any light pulse is exactly zero. Now let a different observer measure the same pulse. They see different values for \(\Delta t\) and \(r\), because they are moving relative to the first observer and their clocks and rulers disagree. If they too measure the speed of light to be c, then for them, \(r' = c\,\Delta t'\) must also hold, and their \(\Delta s^2\) also evaluates to zero. The minus sign is what allows different observers to disagree about the time and distance of the same event while still agreeing on the interval. The interval is preserved because the minus sign creates exactly one speed, c, where the time term and the space term cancel. That speed is invariant in the precise sense that every observer measures it the same.

If the metric had a plus sign in front of the time term, no such cancellation would be possible. Every speed would give a unique non-zero interval, and there would be no special speed for all observers to agree on. The minus sign is the precise mathematical statement of c-invariance. Pick any other sign and the geometry collapses back to four-dimensional Euclidean space, and four-dimensional Euclidean space disagrees with the Michelson-Morley result.

The geometry is forced by experiment, not chosen for elegance. The minus sign is the smallest change that turns Euclidean four-space into a geometry where light’s invariant speed is possible at all.

What the shape of the metric does

The minus sign also changes what “rotation” means in the geometry, and that is where the word hyperbolic enters.

In Euclidean space, the transformations that preserve distance are ordinary rotations. Pick a point at distance r from the origin and rotate; the point stays at distance r. If you mark every point sitting at distance r from the origin, those points trace out a circle in two dimensions, or a sphere in three. The equation is \(\Delta x^2 + \Delta y^2 = r^2\). Rotation slides points around that circle, which is why rotation preserves distance.

In Minkowski space, the transformations that preserve the interval include the ordinary spatial rotations, plus a new operation called a boost [a change of velocity, the analogue of rotation but acting between a space dimension and the time dimension instead of between two space dimensions]. A boost mixes a space coordinate with the time coordinate the way a rotation mixes two space coordinates. The set of all boosts and ordinary rotations together is the Lorentz transformation [the relativistic replacement for plain velocity addition; it tells you how a moving observer’s measurements of position and time relate to your own, and it reduces to plain addition in the slow limit].

Because the metric has a minus sign, the analogous shape is different. If you mark every point sitting at the same interval from the origin in a \(t\)-\(x\) plane, the points trace out a hyperbola instead of a circle. The equation is

\[ \Delta x^2 - c^2\,\Delta t^2 = \text{constant}, \]

the equation of a hyperbola rather than the equation of a circle. A boost slides points along that hyperbola the same way a rotation slides them around a circle. That is what “the geometry of spacetime is hyperbolic” means in concrete terms. In the full four-dimensional picture the shape becomes a hyperboloid, the way it would be a sphere in Euclidean four-space.

The trigonometry that describes circular rotation is the trigonometry of sine and cosine. The trigonometry that describes hyperbolic rotation is the trigonometry of sinh and cosh, the hyperbolic functions [the analogues of sine and cosine for hyperbolas instead of circles, related to the exponential function in the same way the ordinary trig functions are related to circles]. The familiar identity \(\cos^2\theta + \sin^2\theta = 1\) keeps you on a circle as \(\theta\) varies. Its hyperbolic counterpart \(\cosh^2\varphi - \sinh^2\varphi = 1\) keeps you on a hyperbola as \(\varphi\) varies. A boost is to spacetime what a rotation is to space, and the function that does the job has changed from cos to cosh because the shape we slide along has changed from a circle to a hyperbola. That is the entire content of hyperbolic geometry for the purpose of this paper.

Why c is an asymptote, not a wall

The picture so far still leaves something puzzling. If I am on a fast spacecraft moving at ninety percent of c, and I fire a probe forward from inside the craft at ninety percent of c relative to me, what does an outside observer see? Plain addition would say one hundred and eighty percent of c. The Lorentz transformation says about ninety-nine point four. The combined velocity is closer to c than either piece, but it never reaches c.

There is a cleaner variable for thinking about this. It is called rapidity [a measure of relativistic motion that adds linearly the way ordinary velocity does in everyday physics, but which can grow without bound as velocity approaches c]. Rapidity is the parameter of a hyperbolic rotation, the way an angle is the parameter of an ordinary rotation. If two motions are described by their rapidities rather than their velocities, the rapidities just add. The arithmetic of plain addition that broke for velocities works perfectly for rapidities. The reason velocities saturate at c is that the function relating the two is bounded:

\[ v = c \tanh(\varphi). \]

The variable \(\varphi\) is the rapidity. As \(\varphi\) runs from zero to infinity, \(\tanh(\varphi)\) runs from zero to one. Multiply by c and the velocity runs from zero to c without ever reaching it. Stack two motions by adding their rapidities; the resulting velocity is what the Lorentz transformation gives. That is what the addition formula is doing. It is hyperbolic addition wearing a velocity disguise.

That changes what c is. Classical thinking about the speed of light tends to picture a wall, as if the universe contained a rule that says “you may not exceed this speed” and somehow enforced it. The geometric reading is different. There is no wall. There is an asymptote. The geometry has the shape of an unbounded variable mapped onto a bounded one by a smooth function, and the speed of light is the value the bounded one approaches but never reaches. You cannot break c the way you cannot reach the horizon by walking. The horizon is not a barrier; it is what far away looks like when the geometry is the shape it is.

I find this clarifying every time I come back to it. The mystery of why nothing exceeds c dissolves the moment c is reframed as a limit of the geometry rather than a rule on top of it. The geometry was forced by Michelson-Morley. The asymptote is what the geometry looks like from the inside.

Where Euclidean space lives in this picture

The next question that comes up almost immediately is whether Euclidean space and Minkowski spacetime are separate kinds of geometry or whether one is part of the other. The honest answer is that Euclidean space is part of the larger Minkowski structure in two specific ways, not a parallel alternative to it.

The first way: Euclidean three-space is what you get by taking a slice of Minkowski spacetime at constant time. Pick an observer, freeze their clock at some particular instant, and look at all the points they consider simultaneous with that instant. On that slice, \(\Delta t = 0\), and the Minkowski metric reduces to

\[ \Delta s^2 = \Delta x^2 + \Delta y^2 + \Delta z^2. \]

That is Pythagoras. Each observer’s spatial “now” is Euclidean. Different observers’ slices through spacetime are tilted relative to one another, because simultaneity is not preserved under boosts, but each slice taken on its own is internally a flat three-dimensional Euclidean space. Euclidean three-space is the geometry of any one observer’s spatial cross-section through the four-dimensional object.

The second way: Euclidean kinematics is the low-velocity limit of relativistic kinematics. The hyperbolic functions that govern boosts reduce to linear ones when the rapidity is small. In that limit, \(\sinh(\varphi)\) approaches \(\varphi\), \(\cosh(\varphi)\) approaches one, and the Lorentz transformation reduces to plain addition of velocities. The geometry does not look hyperbolic from inside the slow corner because the bending is not visible at those scales. It looks Euclidean. It is Euclidean to as many decimal places as you can measure, until you push to speeds where the curvature matters.

Both readings end at the same place. Euclidean geometry is a slice and a limit of Minkowski geometry, not a peer to it. The relationship is not “two different geometries side by side.” It is “the larger geometric object, and the corner of it where rotation, distance, and motion all approximate one another well enough that you can pretend the geometry is flat.” Our entire perceptual and intuitive history happened in that corner. Euclidean space is the geometry of the corner. Minkowski spacetime is the geometry of the object that contains the corner.

Where Minkowski space lives in the larger picture

The picture I have built so far is still simpler than what physics actually says about the universe. Minkowski space is flat. There is no gravity in it. The geometry of an actual region of spacetime, with stars and planets and the bending of light around the Sun, is curved, and the description gets more complex. I want to walk up that complexity in stages, because the chain of inference this paper rests on is not “the universe is hyperbolic” but “the universe is a richer geometric object than any one level can capture, and we are observers somewhere on its surface.”

The first step up is general relativity. Einstein’s 1915 theory replaces flat Minkowski spacetime with a Riemannian manifold [a smooth space whose distance function can vary from point to point, instead of being given by a single fixed formula]. Strictly, the spacetime of GR is Lorentzian (one minus sign in the metric, the way Minkowski is) and pseudo-Riemannian, but the point is that the curvature of the manifold is a real, point-to-point feature of the geometry, and gravity is what we call that curvature when matter and energy are present. Minkowski space is then the local tangent picture, the geometry you see when you zoom in close enough to any one point that the curvature looks negligible. The universe is not flat. It is curved, and the curvature is doing the work that nineteenth-century physics ascribed to a force.

That move is the one that lets gravity work without anything pulling on anything. Mass and energy curve the geometry around them, and other mass and energy travel along the straightest available paths through that curvature. From inside, those paths look like falling, like attraction, like one body pulling on another. From outside, they are geodesics [the natural straight-line paths through a curved geometry, what a free particle’s trajectory becomes when no force is acting on it]. Two stars orbiting each other are not exchanging an invisible force. They are following the lines that curvature carves around their combined mass-energy. The pattern shows up more generally in field theories: configurations in a field shape the geometry the field’s other excitations follow, and configurations that look similar tend to evolve toward each other along the resulting curvature. Like attracts like, in this geometric reading, because like configurations bend the geometry in the same way and the curvature converges them.

There is an opposite-signed effect at very large scales. Dark energy [the ambient repulsion or cosmological-constant-like pressure that drives the universe’s accelerating expansion] pushes regions of spacetime apart faster than gravity pulls them together at the cosmological scale. Whatever dark energy turns out to be, the empirical fact is that the universe has both a clustering attraction at one scale and a dispersing repulsion at another, and the structure we observe is what those two opposing dynamics produce together. I come back to this pattern in the architecture section, because it has implications for how a substrate-faithful representation should treat similar regions of meaning, and an analogous pattern of attraction-plus-repulsion shows up there.

The second step is topological. A Riemannian manifold has a metric, a notion of distance. If you forget the metric and keep only the structure that says which points are “near” which others without committing to a measure, you have a topological space [an object with a notion of continuity and neighborhood, but no notion of distance]. Topology captures things like connectedness, the presence of holes, the way regions glue together into a whole. Riemannian manifolds are topological spaces with extra structure on top. The same topological space can carry many different metrics, and the metric is in some sense an additional choice the universe (or our model of it) has to make.

The third step starts to look stranger. There are mathematical objects that describe the relations between spaces themselves rather than the points within a space. Categories and higher categories [mathematical structures that take “things and the relations between them” as the primary data, rather than “spaces and the points inside them”] are the working vocabulary. Topological quantum field theory and certain programs in quantum gravity use this language, because at the scales they care about, the metric structure has stopped being the right primitive and the relational structure has taken over. Whether the universe ultimately wears one of these categorical descriptions as its native form is a contested research question. The fact that physics programs reach for this language at all is a clue.

Before going further, a pause is worth taking. Every step in this stack so far has been built outward from our four-dimensional slice. Minkowski is what the slice looks like locally. Riemannian curvature is what happens when the slice bends. Topology is what survives if you forget the metric on the slice. Categorical descriptions abstract from the points of the slice. The story has been told outward from where we stand, and that structure is the same move physics has already had to undo three times in its history. Ptolemy put Earth at the center of the heavens because Earth was where the observers were. Copernicus moved the center to the Sun because the planetary orbits came out simpler that way, and for a few centuries the Sun was the new center of the universe. Then the Sun turned out to be one star in a galaxy, the galaxy one in a cluster, the cluster one of countless others. Each correction was the discovery that what looked like the center was just the edge of what we could see. The honest version of the geometric stack admits that we are probably doing the same thing again. Our 4D Lorentzian slice is one feature of a larger object, and other parts of that object might have different signatures, different causal structures, or no metric at all. What we cannot see from inside the slice is presumably what we end up calling outside-the-stack influence. The geometric stack as I have laid it out is a map of what we can build outward from where we are. It is probably not a map of the whole host.

The fourth step is the most speculative and the most exciting. Since around 2010, a family of programs sometimes called It-from-Qubit has been arguing that spacetime geometry itself is emergent from non-geometric primitives. Van Raamsdonk’s 2010 paper showed that the spatial geometry of certain holographic models can be literally built up from patterns of quantum entanglement in a lower-dimensional boundary theory (Van Raamsdonk 2010). Swingle showed that the tensor networks used in such constructions are structurally the same object as the entanglement-renormalization networks of condensed-matter physics (Swingle 2012). Maldacena and Susskind have argued that wormholes and quantum entanglement are the same phenomenon viewed through different mathematical lenses (Maldacena and Susskind 2013). On these readings, geometry is not fundamental. Geometry is a coarse-grained description of correlation patterns in a more primitive structure. The universe, at the bottom, may not be a manifold at all.

Each level of the stack contains the previous as a special case. Riemannian manifolds reduce to Minkowski when curvature vanishes locally. Topological spaces include Riemannian manifolds as a subclass with extra structure. Categorical descriptions can recover topological spaces by appropriate restriction. Pre-geometric correlation patterns can give rise to geometry as an emergent phenomenon. Conversely, our perceptual access lives near the simpler end of this stack, and most of what we directly experience is even simpler than that, because we never resolve the curvature, never test the topology, and never see anything but the local Minkowski tangent.

There is one more kind of structure the universe presents, and it does not fit cleanly into the geometric stack but runs through every level of it. Thermodynamic structure: states evolve toward higher entropy, gradients dissipate, and the stable patterns we observe arise as flows through dissipation rather than as static endpoints. A river is not a thing; it is what gravity and a temperature difference look like over a watershed. A living cell is not a thing; it is what a chemical gradient looks like once it has found a way to keep itself running. Even the cosmological story has a thermodynamic spine: the universe is unwinding from a low-entropy initial condition toward a higher-entropy distribution, and what we call structure is what the unwinding produces along the way. The free-energy framing in neuroscience is the same shape applied to brains, where stable percepts and behaviors are the patterns that minimize prediction error within the organism’s available variables (Friston 2010). Geometry tells you what the space looks like. Thermodynamics tells you how the contents of the space organize themselves over time, and the answer is “into whatever flows the gradients allow.” Both layers are present in the universe. Neither one alone is enough.

The point of laying this stack out is not to claim any particular level as the “real” one. Physics is still working that out. The point is that the structure of the universe is something like the whole assembly: a flat slice we perceive directly, sitting inside a curved Riemannian manifold that gravity reveals, sitting inside a topology whose structure cosmology and quantum gravity probe, sitting inside a richer relational object the most contemporary programs are trying to articulate, and run through at every level by the thermodynamic flows that turn that geometry into history. We do not live in any one of these levels alone. We live on the surface of all of them at once, and our intuitions are calibrated to the simplest level, which is the smallest visible part of the assembly.

That is the geometric setup the rest of the paper rests on. The chain that ends at “model meaning after the universe” is not “model after hyperbolic geometry.” It is “model after the whole stack the universe presents, with the flows that run through it.”

Settled physics, the door it opens

A note on the status of everything above. The geometry of spacetime is hyperbolic in the sense I have described, and this is not a contested claim. Working physicists have not argued about it for a hundred years. Special relativity has been confirmed to extraordinary precision in particle accelerators, in atomic clocks flown on aircraft, in the synchronization corrections required for GPS, in the lifetimes of cosmic ray muons reaching the ground. The Lorentz transformation is one of the best-tested pieces of physics ever written down. The Minkowski metric is settled. None of the spine claims of this section are speculative.

What this paper does with the result is something physics does not usually do. The geometry is forced. That much is fixed. The next question is what it implies about us, the observers who happen to live at velocities four orders of magnitude below the asymptote and who built our intuitions out of the slow corner. I have walked through how strange it is that the speed of light comes out the same regardless of the observer’s motion. The strangeness is real, and the resolution is geometric, but the resolution carries an implication that is easy to miss if you stop reading at the equation. We were never going to see the geometry directly. Everything in our perceptual and evolutionary history happened in the regime where rapidity and velocity look like the same variable, where Lorentz transformations look like plain addition, where hyperbolas look like straight lines. The hyperbolic structure of spacetime is something we know about by inference, not by living inside it the way we live inside the three dimensions of our daily experience.

That observation is the door to the next section. If we are observers calibrated to a slow slice of a richer geometric object, we should ask carefully what we can and cannot see from inside that slice. Several things that look from the slice like features of the world may instead be features of the slice. That is the question I take up next.