These physicists think we're totally wrong about entropy

A new line of argument attacks the usual status of the second law of thermodynamics by tying it to the “Boltzmann brain” problem—an unsettling scenario in which random fluctuations at extremely late times could produce self-aware brains far more often than ordinary observers in a long-lived universe. The core claim is that, once the logic is traced carefully, the second law may not be a standalone law of nature at all, but a circular inference: entropy is expected to rise because the past is assumed to have had low entropy, and that low-entropy past is trusted largely because we have records consistent with it—records that themselves rely on the same entropy-increase assumption.

The Boltzmann brain problem starts with the second law’s implication that entropy in an isolated system tends to increase, driving the universe toward more mixed, featureless states. Yet high-entropy states are not perfectly smooth. They exhibit spontaneous fluctuations, meaning small, temporary clumps of matter can appear by chance. Over sufficiently long times, those fluctuations can become arbitrarily large: not just molecules or cells, but even entire brains. In that picture, the universe we experience could itself be one such fluctuation—an entropy “bubble” in which entropy is rising again. The most disturbing twist is probabilistic: smaller fluctuations are more common, so it becomes vastly more likely that observers are brief, self-contained “brains” arising from fluctuations rather than products of a structured cosmic history.

The new paper revisits whether this conclusion can be avoided using established physics. It scrutinizes assumptions Boltzmann originally relied on—especially the idea that particles behave like a random gas that only “bump” into each other. Real physics, including the Standard Model, involves strong clustering: quarks don’t appear alone but in bound states such as protons and neutrons, and gravity further promotes clumping. Even after incorporating those effects, the authors conclude there is still no fully rigorous argument that rules out Boltzmann brains using only established physics.

From there, the argument turns to the second law itself. The usual justification runs in a loop: entropy increases, so past records are reliable; but the expectation that entropy increases is grounded in the assumption that entropy was lower in the past. The paper argues there is no clean logical break in that chain. It even proposes an alternative that is presented as equally legitimate: instead of assuming entropy was minimal at the big bang, assume it was minimal 1,000 years ago. Under that “1,00 CE hypothesis,” entropy would have been higher before that point, implying time effectively ran in the opposite direction until the minimum, then flipped. If that kind of reversal is logically on par with the big-bang low-entropy assumption, then the second law’s usual “arrow of time” justification becomes suspect.

The takeaway is not that entropy demonstrably decreases, but that the foundations behind why it should—especially the low-entropy boundary condition—may be underdetermined by logic alone. The result is a challenge to how confidently physics can claim the universe’s observed time direction without smuggling in the very conclusion it seeks to justify.