In standard physics, everything evolves "in time." But RTSG's deepest axiom says the arrow of time is the arrow of complexification — the monotonic growth of structure. This tutorial shows you how to take that seriously: replace every $d/dt$ in the framework with $d/d\Sigma$, and let time be the derived quantity.
Comfortable: Derivatives, chain rule, the idea that $df/dx = (df/dy)(dy/dx)$.
Helpful but not required: Matrices, eigenvalues, logarithms. We'll build everything from scratch.
Not required: Quantum mechanics, differential geometry, category theory. If those appear, they'll be explained.
Every physics equation you've ever seen says "here's how things change over time." The position of a ball. The temperature of a room. The state of a quantum field. Time is the ruler everything is measured against.
But RTSG starts from a different axiom: time is not fundamental — structure is. The arrow of time is the arrow of complexification: the universe accumulates structure, and "time passing" is what that accumulation feels like from the inside.
If that's true, then time $t$ is the wrong independent variable. The right one is entropy $\Sigma$ — a measure of how much structure has been instantiated. Time is just the rate at which the entropy clock ticks:
$\dot\Sigma$ is the entropy production rate: how fast new structure enters reality per unit clock-time. It becomes the conversion factor between structural evolution and what clocks measure.
Before we can build $\Box_\Sigma$, you need to know what $\Box$ is. It's simpler than it looks: it's just the wave version of "curvature."
You already know the second derivative $d^2f/dx^2$. It measures curvature: how sharply a curve bends. If you have a function of space and time — say, the displacement of a vibrating string at position $x$ and time $t$ — then you have two kinds of curvature:
A wave happens when these two curvatures fight each other. The classic wave equation is:
This says: "temporal curvature equals spatial curvature (times wave speed squared)." When the string is bent in space, that bending accelerates it in time. That's all a wave is.
You can rewrite the wave equation by moving everything to one side:
That operator in the parentheses — the combination of second derivatives with a minus sign on time — is the d'Alembertian $\Box$. In 3D space (with $c = 1$ for simplicity):
nabla2 is your spatial Laplacian — picture it as a 3D convolution kernel
that measures "how different is this point from its neighbors" — then box = -d²/dt² + nabla2.
The box operator is just nabla2 promoted to spacetime, with time getting the
opposite sign.
The RTSG Will Field $W$ is a complex scalar field that lives on spacetime (or on the RTSG configuration space). Its equation of motion is:
Reading this term by term:
| Term | What it does |
|---|---|
| $\Box W$ | Wave propagation — the Will Field propagates like a wave through configuration space. Spatial curvature drives temporal acceleration. |
| $-\alpha W$ | Restoring force — pulls $W$ toward zero (like a spring). The coefficient $\alpha$ controls how strongly. This is the "entropic restoring coefficient." |
| $-\beta|W|^2 W$ | Self-interaction — the field's own intensity affects its dynamics. This is the cubic nonlinearity from the Ginzburg-Landau framework. (The action has $|W|^4$; the equation of motion has $|W|^2 W$ because you take one derivative off.) |
If you drop the $\alpha$ and $\beta$ terms, you get $\Box W = 0$ — a free wave. The extra terms make the wave interact with itself, which is where all the interesting physics lives: phase transitions, condensation, mass gaps.
Now we substitute entropy for time inside the d'Alembertian.
We want to replace $\partial/\partial t$ everywhere with something involving $\partial/\partial \Sigma$. The chain rule says:
This is exactly the same logic as changing variables in an integral. If $\Sigma$ is a function of $t$, then any derivative with respect to $t$ can be re-expressed in terms of a derivative with respect to $\Sigma$, multiplied by the rate of change $\dot\Sigma$.
The d'Alembertian needs $\partial^2/\partial t^2$. Applying the chain rule twice:
By the product rule, the outer $\partial/\partial t$ hits both $\dot\Sigma$ and $\partial/\partial\Sigma$:
Two terms appear:
| Term | Meaning |
|---|---|
| $\ddot\Sigma\;\partial/\partial\Sigma$ | Entropy acceleration — is the clock speeding up or slowing down? If the universe's structural production rate is changing, this term captures it. |
| $\dot\Sigma^2\;\partial^2/\partial\Sigma^2$ | Curvature in entropy-time, scaled by the square of the clock speed. This is the dominant term when $\dot\Sigma$ is roughly constant. |
Substitute into $\Box = -\partial_t^2 + \nabla^2$:
Or, written more compactly using a single "entropy-covariant" temporal operator:
This compact form absorbs both terms (the acceleration and the curvature) into a single expression. It's the standard trick for writing second derivatives in curvilinear coordinates — the same structure you see in the Laplacian in polar or spherical coordinates.
The EOM becomes:
The equation looks the same — $\alpha$, $\beta$, and the nonlinear term are untouched. All the physics of phase transitions, mass gaps, and condensation survives. What changes is inside $\Box_\Sigma$: every temporal derivative now carries the entropy production rate $\dot\Sigma$ as a dynamical coefficient.
We've been writing $\Sigma$ everywhere. Now: what is it, concretely, in RTSG? To answer that, we need to build up the concept of von Neumann entropy from first principles.
Imagine a box with 100 coins. You know 60 are heads and 40 are tails, but you don't know which coins are which. Entropy measures your ignorance: how many different arrangements of those coins are consistent with what you know?
Boltzmann's formula says:
where $p_i$ is the probability of outcome $i$. If you know the outcome perfectly (one $p_i = 1$, rest are 0), entropy is zero — no ignorance. If all outcomes are equally likely ($p_i = 1/N$ for all $i$), entropy is maximal at $\ln N$.
-log(probability). In information theory,
it's the average number of bits you need to encode an outcome. Low entropy = compressible.
High entropy = incompressible. A JPEG compresses a photo (low entropy, lots of structure)
but can't compress noise (high entropy, no structure).
In quantum mechanics, the state of a system isn't always a clean, pure state $|\psi\rangle$. Sometimes you have mixed information: the system is in state $|\psi_1\rangle$ with probability $p_1$, state $|\psi_2\rangle$ with probability $p_2$, etc. The density matrix encodes all of this in a single object:
Each $|\psi_i\rangle\langle\psi_i|$ is a "projector" — it points in the direction of state $i$ in state space. The density matrix is a weighted average of all these projectors.
Key properties of $\rho$:
| Property | Meaning |
|---|---|
| $\rho \geq 0$ | All eigenvalues are non-negative (probabilities can't be negative) |
| $\Tr(\rho) = 1$ | Probabilities sum to 1 |
| $\rho^2 = \rho$ | Pure state (perfect knowledge) — this is a projector |
| $\Tr(\rho^2) < 1$ | Mixed state (incomplete knowledge) |
The density matrix $\rho$ is Hermitian and positive, so it has real non-negative eigenvalues $\lambda_1, \lambda_2, \ldots, \lambda_N$ that sum to 1. These eigenvalues are probabilities — they tell you the weight of each independent "direction" in state space.
Von Neumann's move: apply the Boltzmann formula to these eigenvalues.
The $\ln$ of a matrix is defined through its eigenvalues: if $\rho = U\,\text{diag}(\lambda_i)\,U^\dagger$, then $\ln\rho = U\,\text{diag}(\ln\lambda_i)\,U^\dagger$. So $\rho\ln\rho$ has eigenvalues $\lambda_i\ln\lambda_i$, and the trace sums them up.
eigenvalues = np.linalg.eigvalsh(rho), then
S = -np.sum(eigenvalues * np.log(eigenvalues)) (handling zeros carefully).
That's it — von Neumann entropy is classical Shannon entropy applied to the eigenvalue
spectrum of the density matrix.
| State | Eigenvalues | $\Sigma$ | Meaning |
|---|---|---|---|
| Pure | $(1, 0, 0, \ldots)$ | $0$ | Perfect knowledge. One definite state. |
| Maximally mixed | $(1/N, 1/N, \ldots)$ | $\ln N$ | Total ignorance. All states equally likely. |
| Partially mixed | $(0.7, 0.2, 0.1)$ | $0.80$ | Some knowledge, some uncertainty. |
Now we connect this to RTSG. Recall the foundational equation:
Physical Space (PS) is Quantum Space (QS) quotiented by bisimulation equivalence. Two QS states are bisimilar if they match each other's transitions indefinitely — they're observationally indistinguishable.
Now: PS has a density matrix. The instantiation operator $C$ maps QS into PS, and not all of QS gets mapped equally. Some regions are heavily instantiated (lots of structure), others are sparse. The density matrix $\rho_{PS}$ captures this distribution:
And the RTSG entropy is:
The entropy production rate is then:
Since $\Tr(\rho_{PS}) = 1$ is constant, the second term vanishes, giving:
This is the rate at which the bisimulation quotient's structure diversifies. When $\dot\Sigma$ is large, reality is actively complexifying. When $\dot\Sigma \to 0$, the system has reached structural equilibrium — a fixed point of instantiation.
Here is every RTSG equation that changes under $d/dt \to \dot\Sigma\,d/d\Sigma$, collected in one place.
The $\sqrt{\dot\Sigma}$ in the noise term comes from Itô rescaling: Brownian motion scales as $\sqrt{dt}$, and $dt = d\Sigma/\dot\Sigma$, so $dW_t = dW_\Sigma/\sqrt{\dot\Sigma}$.
| Regime | $\lambda_\Sigma$ | $\dot\Sigma$ | Meaning |
|---|---|---|---|
| Stable attractor | $< 0$ | Moderate, steady | GL ground state. Directed agency. Structure accumulates predictably. |
| Flow / phase transition | $\approx 0$ | High | Critical point. Maximum structural throughput per entropy unit. |
| Dissolution | $> 0$ | $\to \infty$ or $\to 0$ | Either runaway complexity (structural explosion) or frozen stasis. |
Unchanged. The mass gap depends on $\alpha$ (the restoring coefficient), not on the choice of time parameterization. The GL potential shape — and therefore the gap — is coordinate-independent.
The unitary evolution operator $U_\Sigma$ now propagates in entropy-steps rather than time-steps. The Born rule $p_i = \|\Pi_i\psi\|^2$ is preserved because $L^2$ norm is invariant under unitary evolution regardless of parameterization.
Everything above treated the entropy $\Sigma$ and the GL action $S[W]$ as two separate objects that we connected by a coordinate change. Veronika Pokrovskaia observed something deeper: the complex action IS negative entropy.
The Euclidean GL action equals the negative von Neumann entropy of the bisimulation quotient.
This isn't an analogy. It's a claim that two quantities you thought were separate — the functional that governs the dynamics of the Will Field, and the entropy that measures accumulated structure in Physical Space — are literally the same number with opposite signs.
To see why, you need to understand one move that connects quantum mechanics to thermodynamics. It's called Wick rotation, and it's the single most powerful trick in theoretical physics.
In the quantum path integral, configurations of the Will Field are weighted by:
This is oscillatory — phases spin around the unit circle, and contributions from different field configurations interfere with each other. This is quantum behavior: superposition, interference, all of it.
Now perform the substitution $t \to i\tau$ (replace real time with imaginary time). This is Wick rotation. The Lorentzian action $S$ becomes the Euclidean action $S_E$, and the weight transforms:
Oscillation becomes damping. Interference becomes selection. The path integral stops being a quantum superposition and starts being a thermal partition function — the same object that governs statistical mechanics. Configurations with low $S_E$ are exponentially favored over those with high $S_E$.
forEach loop over field configurations where each
contribution is a complex phasor — they all have magnitude 1 and differ only in phase.
Cancellations are what select the classical path. After Wick rotation, $e^{-S_E}$ is a
weighted sum where big-action configs are exponentially suppressed. It's the
difference between sum(exp(i*phases)) and sum(exp(-costs)) —
interference vs. Boltzmann selection.
Now look at the Euclidean GL action. After Wick rotation, the minus sign on the temporal kinetic term flips (because $\partial_t \to -i\partial_\tau$ and squaring gives $-1$):
All plus signs. This is exactly a Ginzburg-Landau free energy functional — the standard object from condensed matter physics that governs phase transitions. And a free energy is $F = E - T\,S_{\text{thermo}}$, which at the level of the functional is negative entropy (up to a temperature factor).
The Will Field drift was defined as $\mu = -\delta S/\delta\bar{W}$. If $S = -\Sigma$:
The Will Field drifts up the entropy gradient — toward states with more instantiated structure. The arrow of time isn't added by hand; it falls out of the equation of motion. Every agent, every particle, every system governed by the Will Field is doing gradient ascent on entropy.
If the action is already $-\Sigma$, then all of Section 5's coordinate-change machinery isn't an external transformation — it's revealing the parameterization the physics was already in. The factor $\dot\Sigma$ isn't a conversion between two independent variables. It's the Jacobian of an identity that was hidden by writing things in clock-time.
The partition function weights field configurations by $e^{\Sigma}$ — configurations with more entropy (more instantiated structure) dominate exponentially. The "classical path" that dominates the path integral is the one that maximizes entropy.
Quantum interference (in real time) is entropy superposition. Decoherence — the transition from quantum to classical — is the transition from oscillatory $e^{iS}$ to damped $e^{-S_E} = e^{\Sigma}$. Decoherence is entropy selection: the universe "chooses" the maximum-entropy path by suppressing all others.
Claim. Let $W$ be the Will Field on the RTSG configuration space, $S_E[W]$ its Euclidean Ginzburg-Landau action, and $\Sigma = -\Tr(\rho_{PS}\ln\rho_{PS})$ the von Neumann entropy of the bisimulation quotient $PS = QS/\!\sim_{\text{bisim}}$. Then:
$$S_E[W] = -\Sigma + \text{const}$$up to a topological constant independent of field configuration.
Status: Theorem candidate. Requires verification that:
| Condition | Status |
|---|---|
| Wick rotation is clean — no boundary terms or anomalies | Expected (standard scalar GL theory) |
| BRST cohomology $H^0(s)$ survives $t \to i\tau$ | Needs verification |
| $\rho_{PS}$ density matrix well-defined under Euclidean continuation | Needs verification |
| The "const" is field-configuration-independent | Needs verification |
Attribution: Veronika Pokrovskaia (@B_Veronika), April 2026.
In clock-time, the "rate of structural change" is an interpretive gloss on the dynamics. In entropy-time, it's a measurable quantity: $\dot\Sigma$ appears explicitly in the equations. You can ask: what is the entropy production rate of this system? Of this brain? Of this economy? Of this conversation? And the answer is a number with units of nats/second (or bits/second) that shows up directly in the dynamics.
A system at the GL ground state satisfies $\partial W/\partial\Sigma = 0$: no further structural evolution per unit entropy. Clock-time may keep ticking, but nothing happens — no new structure enters PS. This is the formal version of "time without events is empty."
Conversely, a system with $\partial W/\partial\Sigma \neq 0$ but $\partial W/\partial t \approx 0$ is one where clock-time passes slowly but structural change is happening — the entropy clock is running fast relative to the wall clock. This describes deep creative states, where subjective time distorts.
This is the philosophical payoff. You're not saying "the universe evolves in time and entropy happens to increase." You're saying: the universe evolves in structure, and time is the shadow that structural evolution casts on clocks. The calculus now matches the ontology.
| Question | Status | Notes |
|---|---|---|
| Is $\dot\Sigma > 0$ provable from the axioms? | Open | This would make the arrow of time a theorem, not an axiom. Needs a monotonicity proof for the von Neumann entropy of the bisimulation quotient under the instantiation operator $C$. |
| Does $\ddot\Sigma$ vanish at cosmic scales? | Conjecture | If the universe's entropy production rate is approximately constant, the $\ddot\Sigma$ term in $\Box_\Sigma$ drops out and the equations simplify dramatically. Testable against CMB data. |
| What is $\dot\Sigma$ for a human brain? | Measurable in principle | EEG entropy production rates exist in the literature. Mapping them to the RTSG $\dot\Sigma$ requires identifying the right density matrix for neural state space. |
| Does the mass gap $\Delta = \sqrt{2\alpha}$ couple to $\dot\Sigma$? | Open | In principle $\alpha$ could be $\Sigma$-dependent, making the mass gap evolve with cosmic entropy. This would be a falsifiable prediction. |
| Can $\dot\Sigma = 0$ (heat death) be reached in finite $t$? | Open | If $\Sigma$ has a supremum, $t \to \infty$ maps to a finite $\Sigma_\text{max}$. The universe "ends" not in infinite time but in finite entropy. |
| Does $H^0(s)$ survive Wick rotation $t \to i\tau$? | Open | Required for the Action-Entropy Identity. If BRST cohomology is preserved under Euclidean continuation, $S_E = -\Sigma$ is exact. Standard for scalar theories but needs explicit check with bisimulation quotient structure. |
| Does decoherence = entropy selection? | Conjecture | The Identity implies the $e^{iS} \to e^{-S_E} = e^{\Sigma}$ transition is the mechanism of decoherence: interference suppression via entropy weighting. Testable against standard decoherence theory predictions. |