The Astrophage Thermodynamics Problem

Page content

Notes on why Andy Weir’s astrophage (from Project Hail Mary*) violates the second law of thermodynamics, prompted by watching Veritasium’s “The Most Misunderstood Concept in Physics”.*

The claim

In the book, astrophage feeds by absorbing heat directly — infrared/thermal energy from its surroundings (a star’s corona, or the dense atmosphere of Rocky’s homeworld Erid) — and stores that energy in a compact, fully recoverable form (described as mass added to held neutrinos). Critically, it does this without radiating anything back out during absorption. It behaves as a perfect heat sink: energy goes in, nothing thermal comes out, and the energy can later be released as clean, usable energy (propulsion).

A note on how this was written. These notes started as mine, but what follows went through several rounds with Claude: checking the thermodynamics, hunting down sources, and rewriting whole sections. It found a real error in my original argument about melting ice. I twice had to correct it about what the book actually says — it invented a heat sink for astrophage that Weir never gives it, and then graded him against it. The result is far enough from my draft that claiming it as my own writing would be a stretch. The remaining errors are still mine.

It is also worth noting that most of what is wrong about the physics is only described in the book. The movie happily glosses over enough details to mostly avoid actually saying anything wrong.

Why this is a second-law violation, not just an efficiency problem

Ordinary heat is disordered energy — it’s the energy of particles jiggling incoherently. You can’t convert it entirely into ordered, useful energy (work) without a colder reservoir to dump some of it into. This is the Kelvin-Planck statement of the second law: no process can have as its sole result the absorption of heat from a reservoir and its complete conversion into work. A Carnot engine gets around this by only converting part of the heat into work and dumping the rest into a cold sink — that’s the whole reason Carnot efficiency is capped at 1 − Tc/Th rather than 100%.

Astrophage, as described, has no cold sink. It absorbs heat and emits nothing — no waste radiation, no rejected heat, nothing. It then claims to recover that energy later as fully usable stored energy (propulsion). That’s a machine that takes 100% of absorbed heat and converts it to 100% recoverable work, with no entropy cost anywhere. It isn’t an unusually efficient engine, or even an impossibly efficient one that beats Carnot’s limit. It’s a perpetual motion machine of the second kind: a device that denies there is a limit to beat.

Why ice melting is different (and legitimate)

Ice melting looks superficially similar — it absorbs heat and “stores” it — but the comparison shows exactly what astrophage is missing, and the difference turns out to be bookkeeping rather than efficiency.

When ice melts, the absorbed latent heat goes into increasing the entropy of the water: liquid water has vastly more accessible microstates than a crystal lattice at the same temperature, and the extra internal energy is soaked up exactly in proportion to that new disorder — ΔS = ΔH/T at the melting point. For water that’s 333.55 kJ/kg of latent heat, about 22 J/mol·K of entropy gained at 273.15 K.

The crucial detail is that at the melting point this exchange is reversible. The ice gains ΔH/T; whatever supplied the heat loses exactly ΔH/T. The books balance to zero. Melting destroys no usable energy at all. Usability is destroyed only when heat crosses a finite temperature gap — ice sitting in a warm room, rather than ice in equilibrium with a bath at 0 °C.

Nor is melted ice worthless. Ice water at 273 K in a 298 K room is a genuine store of extractable work: it serves as the cold reservoir, the room is the hot one, and an engine run between them produces real output. This isn’t a thought experiment — ice thermal storage is a commercial technology, banking cold overnight and cashing it out against peak air-conditioning demand.

So ice tells us something sharper than “heat storage costs you usability.” It tells us what the ledger has to look like. Ice absorbs Q at temperature T and its own entropy rises by exactly Q/T, precisely enough to pay for what the source lost. That balance is why the process is allowed.

Astrophage absorbs the same Q at the same T and its entropy rises by essentially nothing, because the energy goes into a compact, ordered, fully recoverable form. The source still loses Q/T. Nothing gains it. The net entropy change of the universe is −Q/T, and that is the violation, in one line.

The Maxwell’s Demon angle

Storing heat while extracting its full exergy (usable-work potential) with no accompanying entropy increase is functionally identical to Maxwell’s Demon: sorting/concentrating disordered energy into an ordered, usable form without paying an entropy price. The modern resolution (Landauer’s principle / Bennett) shows why even a literal demon can’t do this: the demon must eventually erase or reset the information it used to do the sorting, and that erasure is itself an irreversible, entropy-producing, heat-dissipating act. The demon has to dump waste heat into some reservoir, just like an engine does — there’s no bookkeeping trick that avoids it.

Astrophage, absorbing heat and radiating nothing, has no analogous release valve. There’s nowhere for the entropy that should accompany that absorbed heat to go.

The one legitimate escape hatch (and why it doesn’t apply here)

There is a real physical process that looks like “absorb energy, do useful work, no obvious waste heat visible”: photosynthesis and solar panels. But this works because sunlight isn’t ambient-temperature heat — it’s radiation characteristic of the sun’s ~5800 K surface arriving at a much cooler (~300 K) environment. That mismatch is itself a form of low entropy (few high-energy photons in, many more low-energy photons radiated back out — the same “20 photons out for every 1 in” accounting from the Veritasium video). Plants and panels exploit that gradient, and they do reject waste heat to the cool surroundings; the entropy books balance.

Astrophage is not described this way. It isn’t written as capturing concentrated, high-color-temperature radiation and doing work with the mismatch. It’s written as absorbing thermal energy from a hot environment: the atmosphere on Erid, or — when it’s bred in quantity on Earth — an enclosure heated by the sun, not sunlight itself. And it is written as never rejecting anything at all.

Weir does get one thing right that the cheapest version of this objection misses. Astrophage maintains an internal temperature of 96.415 °C and can only absorb from a reservoir hotter than itself, which is exactly why it can’t feed in Earth’s open atmosphere. There is always a genuine temperature difference between astrophage and whatever it’s eating.

It’s tempting, then, to score this as an efficiency problem — put the source temperature on top, astrophage’s 96.415 °C on the bottom, and ask how far short of Carnot it falls. That calculation is a mistake, and it’s worth spelling out because it’s an easy one to make.

Carnot’s Tc is the temperature of a reservoir the engine exhausts into. Astrophage’s interior is not that. It’s the working body — where the energy is banked — and it’s held at a constant 96.415 °C precisely because nothing flows out of it. A cold reservoir has to accept rejected heat and stay cold, which means passing that heat onward to something else. Astrophage passes nothing onward. It has an internal temperature; it does not have a cold side.

With no cold reservoir, Kelvin-Planck doesn’t hand you a reduced efficiency. It hands you zero. A cyclic device in contact with a single reservoir produces no net work whatsoever, and this holds regardless of how hot that reservoir is — the corona, the photosphere, Erid’s 210 °C atmosphere, a warm room. All identical. The permitted conversion is 0%. Astrophage delivers 100%.

So the gap can’t be narrowed by picking a hotter feeding ground, and it isn’t wider on Erid than in the corona. The missing ingredient was never a temperature gradient; astrophage has one everywhere it feeds. The missing ingredient is an exhaust. Nor does the 4.26 μm Petrova emission supply it — that’s the release of stored energy as directed thrust, the paying-out step, not the taking-in one.

And the book never contemplates one. Nowhere does the question “where does the waste heat go” come up, because in Weir’s model there is no waste heat. Absorption is total and one-directional by construction. Weir didn’t miscalculate an efficiency here. There’s no rejection step in his design to calculate one from.

Bottom line

Ice melting shows what a legal heat-absorption ledger looks like: the absorber’s entropy rises by exactly what the source’s falls, and nothing is created or destroyed. Astrophage is written to absorb heat while its own entropy stays flat, banking 100% of what it takes in as recoverable work. The source’s entropy still falls. Nothing rises to match it.

Carnot’s limit barely enters into it. With no cold reservoir, astrophage’s permitted output is zero, and that figure holds just as firmly in the corona as it does on Erid, since the source temperature plays no part in the argument. Drawing heat from a single reservoir and yielding nothing but work is the exact thing Kelvin-Planck rules out. It’s a perfect Maxwell’s Demon running with no erasure cost, which the second law forbids outright rather than as a matter of engineering difficulty.

Absorbing heat is fine. Absorbing it efficiently is fine too. The question the book never asks is where the heat goes afterward, and every answer thermodynamics allows involves it going somewhere.

References

The prompt

Second law and heat engines

  • HyperPhysics (C. R. Nave, Georgia State University), “Second Law: Heat Engines”. — Kelvin-Planck statement: heat from a single reservoir cannot be converted entirely into work; some heat must be exhausted to a cold reservoir.
  • HyperPhysics, “Carnot Cycle”. — The 1 − Tc/Th ceiling as the maximum efficiency between two reservoirs.
  • Yan, Introduction to Engineering Thermodynamics, §6.4, Kelvin-Planck and Clausius statements, Engineering LibreTexts. — Equivalence of the two statements.

Exergy and the ambient-heat problem

  • “Exergy”, Wikipedia. — The dead state: matter or energy at ambient temperature, pressure, and composition “contains energy but no exergy.” This is the crux of the ambient-infrared objection.
  • Purdue ME 300, Thermodynamics II, Packet 2: Availability (Exergy). — Dead state and the availability balance, in course-notes form.

Ice as a cold reservoir

  • A. Bhatia, Air Conditioning with Thermal Energy Storage, Course M04-028, CED Engineering. — Latent heat of fusion (144 Btu/lb) used commercially to bank cooling capacity off-peak. The stronger citation is ASHRAE’s Design Guide for Cool Thermal Storage, 2nd ed. (2019), which is paywalled.
  • “Ice storage air conditioning”, Wikipedia. — One tonne of water stores ~334 MJ of cooling, ~93 kWh.
  • “Enthalpy of fusion”, Wikipedia (citing CRC Handbook, 62nd ed.). — Water: 333.55 kJ/kg, i.e. 6.01 kJ/mol, giving ΔS_fus ≈ 22 J/mol·K at 273.15 K. Note: the NIST Chemistry WebBook entry for water does not publish an enthalpy of fusion, despite being the obvious place to look.

Maxwell’s Demon and the information cost

  • Owen Maroney, “Information Processing and Thermodynamic Entropy”, Stanford Encyclopedia of Philosophy. — The demon and its modern information-theoretic resolution.
  • R. Landauer, “Irreversibility and Heat Generation in the Computing Process,” IBM Journal of Research and Development 5(3), 183–191 (1961). doi:10.1147/rd.53.0183 · full text (PDF). — Erasing a bit dissipates at least kT ln 2.
  • C. H. Bennett, “The Thermodynamics of Computation—A Review,” International Journal of Theoretical Physics 21(12), 905–940 (1982). doi:10.1007/BF02084158. — Logical vs. thermodynamic reversibility; erasure as the demon’s unavoidable cost.
  • A. Bérut et al., “Experimental verification of Landauer’s principle linking information and thermodynamics,” Nature 483, 187–189 (2012). doi:10.1038/nature10872 · open access · PubMed. — The bound is measured, not merely derived.

The solar escape hatch

  • NASA, “Sun Facts”. — Photosphere at ~5,500 °C (~5,800 K). The precise effective temperature of 5772 K normally lives on the NSSDCA Sun Fact Sheet, which is currently not resolving.
  • X.-G. Zhu, S. P. Long & D. R. Ort, “What is the maximum efficiency with which photosynthesis can convert solar energy into biomass?”, Current Opinion in Biotechnology 19(2), 153–159 (2008). doi:10.1016/j.copbio.2008.02.004 · PubMed. — Thermodynamic ceilings on photosynthesis: ~4.6% for C3, ~6% for C4.
  • On the “20 photons out for every 1 in” figure: this is asserted in the Veritasium video and is easy to derive from the temperature ratio (5772 K / 255 K ≈ 22.6), but I could not trace it to a specific paper. Wu & Liu, “Radiation entropy flux and entropy production of the Earth system,” Reviews of Geophysics 48, RG2003 (2010), doi:10.1029/2008RG000275, is the usual pointer for radiation entropy budgets, but it is paywalled and I have not read it.

The novel

  • Andy Weir, Project Hail Mary (Ballantine, 2021). The astrophage mechanics — energy stored as mass, an internal temperature of 96.415 °C, emission at the “Petrova frequency” of 4.26 μm — are described in the text; no reliable secondary source was available. Likewise the conditions on Erid (~210 °C, ~29 atm) and the fact that astrophage cannot feed in Earth’s atmosphere because it is cooler than astrophage itself. (4.26 μm is genuinely the CO₂ asymmetric-stretch band, so the spectroscopy underlying the premise is real.)