Being able to coarse-grain and having something interesting happen are two different stories
Episode 1 said "entropy increases = discarded information does not come back on its own." That explanation, in fact, has a hole in it. The microscopic laws are perfectly symmetric under time reversal ── play the video backwards and the equations do not complain at all. So why does reality only run one way? The arrow of time does not come out of coarse-graining alone. One more card is needed ── the assumption that the universe began with extraordinarily low entropy (the past hypothesis). And how extraordinary? By Penrose's estimate, \(1/10^{10^{123}}\). This bonus episode confirms that underneath the "throw it away, same answer" the main series has leaned on all along, there sits another enormous assumption.
Boltzmann tried to derive the increase of entropy from coarse-graining (the H-theorem). He immediately met two devastating objections.
| Objection | Content | Effect |
|---|---|---|
| Loschmidt's reversal | Reverse every particle's velocity and the system runs toward decreasing entropy. Given that the laws are time-symmetric, that solution is exactly as legitimate | deriving "increases" from the laws is impossible |
| Zermelo's recurrence | A finite system must, if you wait long enough, return arbitrarily close to its initial state (Poincaré recurrence) | increasing forever is impossible in principle |
Both are correct. Boltzmann's reply was an honest one ── entropy increases not because of the laws but because it is low right now. Start from a low state and the only place to go is the overwhelmingly higher ones (the counts are too lopsided). So it increases. The reversed and recurrent solutions exist; they are just fantastically rare.
But this only shifts the question by one step. So why is it low now? Because yesterday it was lower. And why yesterday? ── Follow the chain and you run into the beginning of the universe.
You can check it in the box below. The particles only bounce off walls; the law is perfectly time-symmetric.
The third is the crucial one. Same law, same coarse-graining ── no arrow of time unless the initial condition is special. The arrow is not a property of the laws but a property of the initial condition.
How rare was the universe's initial condition? Let's follow Penrose's famous estimate.
Gather all the matter in the observable universe into one giant black hole and its entropy (from the Bekenstein–Hawking formula) is
$$S_{\max}\sim 10^{123}\,k_B$$That is the largest entropy this universe could have. The actual entropy of the early universe, from photon counts and so on, was about \(10^{88}\,k_B\). Inverting Episode 1's \(S=k\log W\) gives \(W=e^{S/k}\), so
$$\frac{W_{\text{initial}}}{W_{\max}}\sim\frac{e^{10^{88}}}{e^{10^{123}}}\ \approx\ \frac{1}{10^{10^{123}}}$$The denominator is a 1 followed by \(10^{123}\) zeros. Try to write it out and you will run out of elementary particles in the universe before you run out of digits. Penrose put it as: the Creator had to hit a target of \(1/10^{10^{123}}\) in phase space.
Here something counterintuitive happens. Photographs of the early universe (the cosmic microwave background) show it was almost perfectly uniform. Normally "uniformly mixed" means high entropy ── coffee with milk stirred in. So why is this extraordinarily low entropy?
| Low entropy | High entropy | |
|---|---|---|
| a gas (no gravity) | bunched in one corner | spread out uniformly |
| matter with gravity | spread out uniformly | gathered into lumps (ultimately a black hole) |
Gravity only attracts, so gathering is "more likely" ── more configurations. Hence for gravity, uniform is the anomalous state. The early universe being uniform means the gravitational degrees of freedom were entirely unexcited ── which Penrose formalised as the Weyl curvature hypothesis (the Weyl curvature, i.e. the gravitational degrees of freedom themselves, vanished in the early universe). The history of the universe is the process by which gravity slowly eats that entropy budget, turning uniformity into stars, galaxies and black holes. What we receive from the Sun is not heat so much as low entropy.
Here is the most important point of Bonus ①. Being able to coarse-grain is not enough to make a world interesting.
Had the universe started in thermal equilibrium, coarse-graining would work perfectly. Same temperature everywhere, same density. A handful of macroscopic variables would suffice and predictions would be 100% right. And nothing would happen. No structure, no layers, no chemistry, no life, no observers.
The renormalization group guarantees "throw it away, same answer," but it does not guarantee that anything happens. The world has layers, and interesting things happen in them, because the universe began with \(10^{123}\) bits of headroom and still has not used it up. We live inside that headroom.
Established: that the microscopic laws of mechanics are time-reversal symmetric (with the small exception of CP violation); that Loschmidt's reversal paradox and Zermelo's recurrence paradox follow; that the second law of thermodynamics therefore cannot be derived from mechanical laws alone; that in gravitating many-body systems the uniform state is low-entropy and the concentrated state high-entropy (negative heat capacity of gravitating systems); that black hole entropy is proportional to area; and that the maximum entropy of the observable universe is estimated at \(\sim10^{123}k_B\) ── all established physics and standard understanding.
Unsolved / under debate: (1) The past hypothesis itself is a posit, not an explanation. It does not answer "why was the initial condition like that?" (2) It is claimed that inflation explains the low-entropy initial condition, but the persistent objection that "inflation itself needs a low-entropy state to get going" has not been settled. (3) The Weyl curvature hypothesis is Penrose's proposal, not an established law. (4) Penrose's \(10^{10^{123}}\) depends on a particular way of estimating; treat it as a device for feeling the order of magnitude. (5) Whether "the entropy of the whole universe" can even be defined (how to count the entropy of the gravitational field; the system is not isolated) is itself debated in both physics and philosophy. ── This episode is a confirmation that the main series rests on another enormous assumption, not a resolution of it.
Because microscopic laws are time-reversal symmetric, the second law does not follow from coarse-graining alone (Loschmidt's reversal, Zermelo's recurrence). Entropy increases because "it is low now," and following that chain backwards runs into the beginning of the universe ── the past hypothesis. The specialness required is \(1/10^{10^{123}}\) by Penrose's estimate. And with gravity the story inverts: being uniform is what counts as low entropy (the Weyl curvature hypothesis). The history of the universe is gravity slowly eating that headroom.
And the connection back to the main series. Coarse-grainable + a special initial condition = layers, structure, an arrow of time. The renormalization group guarantees "throw it away, same answer" but does not guarantee that anything happens. In a universe that started at equilibrium, coarse-graining works perfectly and nothing occurs. We live in a layered, interesting world because the universe began with \(10^{123}\) bits of headroom and has not used it up.
Print / make a PDF: ⌘+P (Ctrl+P on Windows). On screen, "Start from the left half," "Reverse time" and "Start at equilibrium" let you confirm that the arrow of time comes from the initial condition. "See the answer" opens each solution.