Why is our universe four dimensional and Lorentzian?

---

Why Four-Dimensional Lorentzian Spacetime?

Our universe has three large dimensions of space and one of time, and its large-scale geometry is Lorentzian (one negative sign in the metric). A good part of the reason lies in the mathematics of partial differential equations: which kinds of equations support genuine time evolution with finite signal speed, and which do not.

This essay develops the case in undergraduate-friendly prose, explaining every symbol used. Even if parts of this post are too hard, consider it as a pointer to what you need to understand to properly address the issue. Implicit is the idea that there imay be some underlying physical process capable of generating 'universe manifolds' with a distribution of dimensionalities and metric signatures.

This post was put together with GPT5.2 and Gemini.

---

1. Static fields: Laplace’s equation

In n spatial dimensions Laplace’s equation (a special case of Poisson’s equation) is

∇²φ = Σ_i=1ⁿ ∂²φ/∂x_i² = 0.

Here φ(x) is a scalar field (electric potential, temperature, etc.). The Laplacian ∇² (“del squared”) sums second spatial derivatives, measuring the total local curvature of the field.

Domain quantification for Laplace’s equation

Let φ: ℝⁿ → ℝ be twice differentiable and let Ω ⊆ ℝⁿ be the region of interest. Writing “∇²φ = 0” means

∀ x ∈ Ω,   ∇²φ(x) = 0.

In words: φ is harmonic at every point of Ω. In electrostatics, for example, with charge density ρ, the potential satisfies Poisson’s equation

∇²V(x) = −ρ(x)/ε₀   in ℝ³,

and therefore Laplace’s equation holds only in charge-free subregions (where ρ = 0).

Physical meaning

Think of φ as temperature. The sign of ∇²φ compares φ at a point to the average of φ over a small surrounding sphere:

• ∇²φ > 0 means φ is locally below its neighbourhood average (so, under diffusion dynamics, it would tend to increase with time).
• ∇²φ < 0 means φ is locally above its neighbourhood average (so it would tend to decrease with time).
• ∇²φ = 0 means perfect local balance: the value equals the neighbourhood average (the mean value property).

It is important to stress: ∇²φ = 0 does not mean φ is uniform. Harmonic functions can vary smoothly. For example, φ(x,y) = x in two dimensions has Laplacian zero but is clearly not constant. What Laplace’s equation really guarantees is that there are no interior maxima or minima: extrema occur only on the boundary.

Thus Laplace’s equation describes equilibrium in source-free regions. Mathematically it is an elliptic PDE: it is controlled by boundary data and does not describe time evolution.

---

2. Smoothing in time: the diffusion equation

The diffusion (heat) equation is

∂u/∂t = D ∇²u,

where u(x,t) is temperature or density, t is time, and D is the diffusion coefficient (units length²/time). Diffusion smooths irregularities. Why do we say it has “infinite propagation speed”?

Point disturbance. Start with u(x,0) = δ(x) (a spike at the origin). The solution for t > 0 is a Gaussian (the heat kernel)

u(x,t) = (4πDt)^−n/2 exp(−|x|²/(4Dt)).

The standard deviation in each coordinate is √(2Dt) (so the typical radius grows like √(2nDt)). For any t > 0 the Gaussian is nonzero for every x (though tiny far away), so the disturbance has instantaneous support everywhere. Diffusion is therefore parabolic: it smooths, but it does not impose a finite signal speed. (In real materials diffusion is an effective, coarse-grained description; microscopic physics remains causal.)

---

3. Signals and causality: the wave equation

The one dimensional wave equation is

∂²u/∂t² − c² ∂²u/∂x² = 0,

with wave speed c. Solutions are travelling waves u(x,t) = f(x − ct) + g(x + ct). Disturbances propagate at finite speed c: if the initial data are localised, the solution vanishes outside the region that the wave has had time to reach.

The minus sign matters because it makes the equation hyperbolic. Hyperbolic PDEs have real characteristic curves (here x ± ct = constant), which define a finite domain of dependence: the value at a point depends only on data in its past light-cone (in 1D, its past interval). Elliptic equations (like Laplace’s) have no such real characteristics; they are controlled globally by boundary conditions instead of evolving locally in time.

---

4. The box operator and why only one time dimension

Relativistic wave equations use the d’Alembertian (“box”)

□ = g^μν∂_μ∂_ν.

In flat spacetime (or locally in a freely falling frame), with Lorentzian signature (−,+,+,+), this becomes

□ = −(1/c²)∂²/∂t² + ∇²,

and equations like □φ = 0 are hyperbolic: they support finite-speed waves (electromagnetism, gravitational waves) and a well-posed initial value problem.

With all plus signs (Euclidean signature), the operator becomes elliptic: there are no light cones and no genuine wave propagation.

With two or more minus signs (multiple time dimensions), the problem is deeper than “the energy goes negative”. One typically loses a natural notion of a Cauchy surface (a spacelike “snapshot” carrying enough data to determine evolution), and the initial value problem is generically ill-posed. In field theory language, it becomes difficult (often impossible) to maintain both stability and unitarity: there is no clean, positive-definite conserved energy that plays the role of a stable Hamiltonian generating time evolution.

So: no time gives no dynamics; multiple times tend to destroy well-posed evolution and stability; exactly one time gives hyperbolic dynamics, finite signal speed, and a coherent causal structure.

---

5. A brief note on “why three space dimensions?”

The argument above already explains why a Lorentzian signature with one time dimension is the minimal structure needed for causal propagation. The “why three space dimensions?” question is partly separate, but there is at least one clean geometric fact worth stating. In n spatial dimensions, the field of a point source spreads over the surface area of an (n−1)-sphere, so long-range forces scale as

F(r) ∝ 1/rⁿ⁻¹   (for n > 2),

giving the inverse-square law only when n = 3. This does not “prove” that only three spatial dimensions can support complexity, but it does show that the familiar hierarchy “stable atoms - stable chemistry - long-lived planetary systems” is not generic as n varies: change the dimensionality and you change the basic fall-off of the forces on which bound structures depend; the inverse-square law in three spatial dimensions uniquely supports robust bound orbits, while in other dimensionalities bound systems are typically far more fragile or non-generic (i.e. they require precise tuning, and a small perturbation typically causes escape or collapse).

---

6. Why ‘one time dimension’ and a Lorentzian metric

We can now answer the title question directly.

If there were no time dimension at all, spacetime would reduce to a purely spatial (positive-definite) geometry - a metric with signature (+,+,...,+). The natural second-order field operators would then be elliptic (Laplace/Poisson-type): they would impose global constraints fixed by boundary data, rather than generating time evolution and finite-speed propagation.Those equations describe static balance: solutions are fixed by boundary conditions and source distributions. There would be no evolution, no wave propagation, and no causal unfolding of events - a universe “frozen” into equilibrium.

If there were two or more time dimensions, the basic wave operator would carry two or more minus signs. Then the usual framework of physics strains or breaks: there is no privileged notion of “time evolution from initial data”, because one generally lacks a natural Cauchy surface (a 'snapshot' of the universe at one instant of time that contains enough information to predict the future) and the associated well-posed initial value problem. In quantum field terms one typically encounters negative-norm states or energy functionals not bounded below, undermining stability.

With exactly one time dimension, the box operator is hyperbolic. That hyperbolicity yields light cones: the set of events that can influence (or be influenced by) a given event at speed ≤ c. Light cones divide spacetime into past, future, and “elsewhere”, providing the mathematical backbone of finite signal speed and causality. Given suitable initial data on a spacelike slice, the future evolution is (in the relevant sense) determined.

So the Lorentzian metric with exactly one negative sign is not an arbitrary convention. It is the minimal signature that supports stable, causal dynamics rather than static constraints or ill-posed evolution.

---

7. In summary

Elliptic equations (Laplace/Poisson) describe static constraints. Parabolic equations (diffusion) smooth disturbances but have instantaneous support. Hyperbolic equations (waves) give finite signal speed and well-posed evolution from initial data.

A Lorentzian signature supplies exactly the sign structure needed for a hyperbolic box operator and therefore light cones: finite-speed propagation and a coherent causal order. Remove time and you remove dynamics; add extra time directions and you generically lose well-posed evolution and stability. In that sense, 3+1 Lorentzian spacetime sits in a narrow “window” where predictable causal physics is possible, and with it cosmological structure and life.

---

Addendum: Greg Egan’s Orthogonal as a stress test

Greg Egan’s Orthogonal trilogy is a useful imaginative stress test. He explores a universe with a purely Euclidean metric (+,+,+,+). It is mathematically consistent, but it becomes physically alien: the usual hyperbolic wave structure, invariant signal speed, and light-cone causality are not available without substantial compensating assumptions. Read as fiction with working mathematics, it is a good way to feel how tightly “Lorentzian” is entangled with “signals, dynamics, and causality”.

Egan escapes stasis in 'Orthogonal' by decoupling “time” from the metric: although the geometry is positive-definite, he introduces a preferred non-metric evolution parameter along which fields evolve. This produces dynamics and propagation, but only by abandoning geometric causality and invariant signal speed - showing that, without a Lorentzian signature, time and causality must be imposed artificially rather than arising naturally from the geometry.

---

.