What Is Nothing?

A puzzle that flatters us

Let 0 stand for Nothing and 1 for Something. The toy form of the mystery of existence is the transition from 0 to 1, and there is a precise sense in which it is impossible.

A cardinal $\kappa$ is regular if it cannot be written as a sum of fewer than $\kappa$ cardinals each smaller than $\kappa$:

equivalently $\operatorname{cf}(\kappa)=\kappa$, where $\operatorname{cf}$ is the cofinality. The number 1 is regular: the only resource below it is 0, the empty sum is 0, and that is the end of it. So you cannot get from Nothing to Something. The number 2 is also regular, since a single term below 2 is at most 1, and one term cannot reach 2. So you cannot get from Unity to Plurality either.

Two clean slogans, both true. The puzzle flatters us because the slogans sound like a law about being. They are a law about addition, and only at the bottom of the integers.

Taking the arithmetic literally

Continue the check past the point where the slogans stop.

• $n=1$: empty sum $=0\neq 1$. Regular.
• $n=2$: one term $\le 1$, sum $\le 1<2$. Regular.
• $n=3$: $2+1=3$, two terms, each below 3, and $2<3$. Singular.
• $n\ge 3$: $n=(n-1)+1$, two terms below $n$, and $2<n$. Singular, always.

The finite regular cardinals are exactly $\{0,1,2\}$. The next regular cardinal is $\aleph_0$, then $\aleph_1$, while $\aleph_\omega=\sum_{n<\omega}\aleph_n$ is singular. So the impossibility of generation, read as regularity, is a property of two small numbers and a few infinite ones, not a property of “going from less to more.” From 2 onward, going from less to more is free.

Question the assumption nobody states. Regularity is defined against addition. Why addition? If the operation is succession, the picture inverts at once: in Peano arithmetic $1:=S(0)$, one application of one operator, Nothing to Something, no sum anywhere. The wall was never around 1. It was around addition, and we mistook the frame for the world. This is the first lesson, and the whole essay is its unfolding: the question has no answer until you name the operation, and naming the operation is the entire question in disguise.

Three zeros

The puzzle survives in conversation by letting one zero impersonate another. There are at least three.

1. $0_{\text{set}}$: the empty set, the empty sum. Here regularity is a real theorem, but it blocks only the first two rungs and only under addition.
2. $0_{\text{phys}}$: the physical vacuum. In canonical quantisation of the electromagnetic field, each mode is an oscillator with $[\hat a_{\mathbf k\lambda},\hat a^\dagger_{\mathbf k^{\prime}\lambda^{\prime}}]=\delta_{\lambda\lambda^{\prime}}\delta^{(3)}(\mathbf k-\mathbf k^{\prime})$, and the vacuum is annihilated by every lowering operator, $\hat a_{\mathbf k\lambda}\lvert 0\rangle=0$. It is a unit vector, $\langle 0|0\rangle=1$, living inside a Fock space that already carries the full algebra. It is the most structured “nothing” we know how to write.
3. $0_{\text{phil}}$: the absence of all being, all law, all Hilbert space, all rules of inference, including the metalanguage in which one would state “nothing whatsoever.” Physics never touches this.

The slogans are theorems about $0_{\text{set}}$. The vacuum energy problem is about $0_{\text{phys}}$. The mystery the puzzle gestures at is about $0_{\text{phil}}$. Conflating them is the trick; separating them is the method.

The inversion: the operation is the primitive

A one-photon state is

the successor move, not the additive one. In the only precise sense physics offers, you do get 1 from 0, by one application of a raising operator, exactly by refusing the additive frame. The object, the photon, is cheap: one operator and a vacuum to act on. What is not cheap is the existence of $\hat a^\dagger$ and the algebra it satisfies.

So invert the question. Do not ask how 1 arises. Ask where the successor comes from, where the creation operator comes from, where the commutation relation comes from. Existence questions about objects are nearly trivial. Existence questions about generators are the hard ones. The frontier is not “Something from Nothing.” It is the underivability of generators.

A tower of grounds

Once the operation is primitive, “Nothing” stops being absolute and becomes a position in a tower. At each level there is a ground (the level’s zero) and a generator (the level’s operation), and the ground is never empty. It is the previous level’s structure, seen from above.

Read upward, every level’s act of creation is the level below saying “I was already here.” The tower never bottoms out in $0_{\text{phil}}$. It bottoms out, as far as any physics can reach, in structure we did not put there and cannot derive from within the level that uses it. The rest of this essay walks the tower outward in time: first to find where the atoms of life were forged, then to ask why almost everything in the sky looks dead, then to date the window in which Earth is not.

Where the substance came from: the dead stars that built us

Trace any carbon atom in a living cell backward and it ends in the core of a star that no longer exists.

The Big Bang nucleosynthesis of the first three minutes produced hydrogen, helium-4, and a trace of lithium, and then stopped. It stopped at a real bottleneck: there is no stable nucleus at mass 5 or mass 8, so the obvious stepping stones from helium upward do not exist. The early universe could not make carbon. For roughly the first hundred million years the cosmos held the fuel of life and none of its substance.

Carbon is built later, inside stars, by the triple-alpha process:

Beryllium-8 is unstable and falls apart in about $10^{-16}$ s, so this works only because of a resonance: an excited state of carbon-12 at $7.65$ MeV, almost exactly matching the energy of ${}^8\text{Be}+{}^4\text{He}$, lets the reaction proceed before the beryllium decays. Hoyle predicted that resonance must exist on the grounds that we are here and made of carbon, and it was found. Then ${}^{12}\text{C}+{}^4\text{He}\to{}^{16}\text{O}$, and the CNO cycle, build the oxygen and nitrogen that, with carbon and hydrogen, make up almost all of you.

Fusion releases energy only up to iron. The binding energy per nucleon rises to a peak of about $8.8$ MeV at ${}^{56}\text{Fe}$ and falls beyond it, so building heavier nuclei costs energy rather than releasing it. The elements past iron, and much of the dispersal of everything below, require stellar death: slow neutron capture in the winds of aging stars (the s-process), and rapid capture in supernovae and neutron-star mergers (the r-process). Stellar winds and explosions then scatter the enriched gas into the interstellar medium, where it condenses into the next generation of stars and planets, and into us.

This is the literal answer to “where did it all come from.” The hydrogen is primordial. Every heavier atom, the carbon of every protein, the oxygen of every breath, the phosphorus of every strand of DNA, was assembled in a stellar interior and released by a star’s death. The sky is full of dead stars because the dead stars are the foundries. Life is not the opposite of the dead cosmos. It is the dead cosmos rearranged.

Dead, or? Dissipation versus memory

So set the apparent fact squarely: stars and planets, in their billions, look dead, and one small planet is alive. Is that the right description?

It is half right, and the half it gets wrong is the interesting half. There is no agreed definition of life. The most-used working definition, from NASA, is a self-sustaining chemical system capable of Darwinian evolution. Hold a star against it.

A star is not in equilibrium. It is a dissipative structure in the sense of Prigogine: an ordered pattern that exists only because energy flows through it. Free energy from gravitational contraction and fusion pours out of the core, and to carry it the star organizes itself, into convection cells, differential rotation, magnetic dynamos, the whole machinery of stellar weather. It is born, it runs a metabolism of sorts, it ages, it dies. By the criteria of thermodynamic self-organization, a star shares a great deal with a cell. The same is true, more quietly, of a planet’s mantle convection, its climate, its weather. The cosmos is not a still life. It is full of structures held open by gradients.

What the star lacks is the one thing the definition demands and the one thing Earth has: a heritable information carrier subject to selection. A star does not copy itself with variation that is then filtered by survival. It has dissipation without memory. Life is the rare case where a dissipative structure acquired a record, a digital text written in a polymer, that it copies, varies, and passes on, so that the dissipation is no longer merely sustained but cumulative. Schrödinger anticipated the carrier as an “aperiodic crystal” before anyone had seen DNA, and identified the thermodynamic role exactly: the organism feeds on a gradient and exports entropy to stay ordered.

So the clean statement, and the resolution of “dead, or?”, is this:

Dissipation is universal; every star is a worked example. Memory is, on the present evidence, almost unimaginably rare; we know of one planet that has it. The commonality across all the stars and planets is the first term. Earth is distinguished only by the second. (The one speculative attempt to give stars the missing term, Smolin’s cosmological natural selection, in which black holes spawn offspring universes with slightly varied constants, is a real idea but an unconfirmed one, and I flag it as speculation rather than lean on it.) The right picture is not life against a dead universe. It is one universal process, the riding of gradients, that on a single known world picked up the ability to remember.

The origin of life: a phase transition, not an assembly

This reframes the origin of life and dissolves the false demand built into the usual question. Abiogenesis is not “assemble a cell from sub-cells by addition,” so cardinal arithmetic, and the whole 0-to-1 intuition, is the wrong formalism for it.

Let the chemical state be $\mathbf x(t)\in\mathbb R^N_{\ge 0}$ under mass-action kinetics, $\dot{\mathbf x}=\mathbf S\,\mathbf v(\mathbf x)$, with $\mathbf S$ the stoichiometric matrix and $\mathbf v$ the rate vector. Life, in the minimal self-reproducing sense, is not a number you sum your way to. It is an attractor, a self-sustaining fixed point or limit cycle that regenerates its own catalytic support. The transition to it is a bifurcation: as a control parameter (catalytic connectivity, monomer supply, free-energy throughput) crosses a critical value, a stable autocatalytic state appears where none existed. Reflexively-autocatalytic, food-generated set theory sharpens this to a percolation threshold: in a random catalytic network with catalysis probability $p$, a self-sustaining set appears with high probability once the expected catalytic degree crosses order unity. A connectivity transition, not an addition. There is no cardinal wall anywhere in it.

The transition needs three things, and the tower has already supplied two. The atoms came from dead stars (Section VI). The free energy came from a gradient, and here the science is still open between two candidates, not mutually exclusive. One is solar and at the surface: ultraviolet-driven cyanosulfidic chemistry that builds nucleotide, amino-acid, and lipid precursors out of hydrogen cyanide and related feedstocks, the modern descendant of the Miller-Urey experiments. The other is geochemical and in the dark: alkaline hydrothermal vents, where natural proton gradients across thin iron-sulphur mineral walls drive a proto-chemiosmosis that strikingly resembles the proton-motive force every living cell still runs on. The third thing is the bifurcation itself, the moment a self-amplifying cycle crossed its threshold and locked in. Before that moment there is chemistry held open by a gradient, dissipation without memory. After it there is a network that copies, and copying plus a gradient is heredity, and heredity plus survival is selection. That instant, the appearance of the heritable term in the equation of Section VII, is the actual origin of life. Everything before it is geology; everything after it is biology; the line between them is a phase transition, thermodynamically forced by a gradient and structurally a change in the topology of a reaction network.

The window: how long life has left on Earth

If life is a transient ordered structure riding a gradient, then it lasts exactly as long as the gradient does, and on Earth the gradient has a schedule.

The Sun brightens. As hydrogen fuses to helium in the core, the mean molecular weight rises, the core must contract and heat to hold itself up against gravity (the virial relation), and because the fusion rate climbs steeply with temperature, the luminosity rises with it, by roughly one percent every hundred million years, about ten percent per billion years. This is not a distant catastrophe. It is a slow, monotonic increase that has been running the whole time and sets a deadline.

The planet has a thermostat, the carbonate-silicate cycle: warming speeds the weathering of silicate rock, which draws carbon dioxide out of the air, which cools the planet back down. The thermostat has bought billions of years. But it works by lowering CO$_2$, and CO$_2$ is also the feedstock of photosynthesis, so the same regulation that holds the temperature is starving the biosphere from below. The numbers, with the usual model uncertainties, run roughly like this:

• In about 500 to 900 million years, atmospheric CO$_2$ falls below the level C3 photosynthesis needs, around 150 ppm, and the dominant plants die. C4 plants, which tolerate lower CO$_2$, hang on toward 10 ppm.
• In about 1 billion years, a coupled biogeochemistry-climate model finds Earth’s oxygen-rich atmosphere sharply collapsing back toward Archaean levels, a mean lifespan of $1.08\pm0.14$ Gyr, and this deoxygenation is projected to arrive before the moist greenhouse, not after it [Ozaki & Reinhard 2021]. The oxygen biosignature, the thing a distant astronomer would read as “alive,” fails first.
• In about 1 to 2 billion years, rising luminosity pushes Earth toward a moist greenhouse; water vapour, itself a greenhouse gas, accumulates and is lost to space, and the surface becomes uninhabitable for complex life. Three-dimensional climate models place this transition near 2 Gyr [Wolf & Toon 2015], earlier one-dimensional estimates nearer 1 Gyr [Caldeira & Kasting 1992], and some recent work extends the complex-life window further still [Wolf et al. 2024].
• In perhaps 2 billion years or a little more, the last deep microbial refugia, the final dissipative structures with memory, are gone, and Earth is effectively sterile.

The spread across these models, roughly 1 to 2 Gyr for the end of complex surface life, is real, and I flag it rather than paper over it; but it does not soften the conclusion, it sharpens it. Two corrections to intuition follow. First, complex surface life has of order one billion years left, not the open-ended future people imagine, and the detectable biosphere, the oxygen, may have appreciably less. Second, we are not early. Life arose about 3.8 billion years ago and the biosphere’s window closes within roughly one to two billion more, so terrestrial life is already some two-thirds to four-fifths through its total run, and complex surface life, which appeared only about 0.6 billion years ago, may be near or past the midpoint of its far shorter window. We exist late in the habitable era, not at its dawn, and that lateness is itself a fact in need of explanation.

After the window: the death of the Sun and the eras of the universe

The window closes long before the Sun itself dies, but the Sun’s death sets the next horizon.

In about 5 billion years the Sun exhausts core hydrogen and leaves the main sequence. It swells through the subgiant and red-giant branch, and near the tip, about 7.6 billion years from now, its radius approaches Earth’s orbit. The Sun loses mass as it expands, which widens the orbits, but tidal drag pulls inward, and the best current models have the Earth engulfed rather than spared. After a helium flash and a brief, brighter horizontal-branch life, the Sun ascends the asymptotic giant branch, ejects its outer layers as a planetary nebula, and leaves a carbon-oxygen white dwarf of about half its present mass, which then cools for more than $10^{13}$ years toward a cold black dwarf. The foundry that made and warmed us ends as an ember.

The universe runs the same logic at vast scale. Following the standard accounting of cosmic eras, with time measured in cosmological decades $\eta=\log_{10}(t/\text{yr})$:

• Stelliferous era, to about $\eta=14$. Star formation continues until the gas is spent. The last and longest-lived stars, the small red dwarfs, finally burn out. After about $10^{14}$ years the sky goes dark.
• Degenerate era, about $\eta=15$ to $39$. What remains are white dwarfs, neutron stars, brown dwarfs. If the proton is unstable, with a lifetime of at least $10^{34}$ years, ordinary matter slowly evaporates from within into radiation and light particles.
• Black hole era, to about $\eta=100$. Black holes inherit the universe and then evaporate by Hawking radiation, stellar-mass holes in about $10^{67}$ years, the largest in about $10^{100}$.
• Dark era, beyond $\eta=100$. Only stray photons, neutrinos, electrons and positrons drift through an ever-emptier, ever-colder, exponentially expanding space.

And it is expanding ever faster. With a positive cosmological constant the future is de Sitter: distant galaxies redshift away and cross the horizon, the observable universe empties, and what is left dilutes toward a cold, near-empty stage. This is the second silence, and it is the mirror of the first. The essay began with the vacuum, the structured nothing the universe started one level above. It ends with the vacuum reasserting itself as everything else thins away.

The constant at both ends

The two silences are written with the same symbol, and this is the closing physics.

Starting at $0_{\text{phys}}$ rather than $0_{\text{phil}}$ is not free. The vacuum is not quiet. The field Hamiltonian is

divergent, and even with a Planck cutoff $\rho_{\text{vac}}\sim 10^{74}\,\text{GeV}^4$ against an observed $\rho_\Lambda\sim 10^{-47}\,\text{GeV}^4$, the factor of order $10^{120}$. That number is the size of the Something hidden inside the physical nothing, written as a density. It is the bill for beginning one level up.

The same $\Lambda$ that measures the vacuum’s hidden content is the agent of the end. An accelerating, de Sitter universe has a nonzero floor temperature,

a thermal bath set by the expansion rate $H$, below which nothing can cool. This is fatal to the one scheme by which life might have lasted forever. Dyson’s “Time Without End” showed that in an open, decelerating universe a civilisation could in principle compute indefinitely by slowing its metabolism without limit and dumping waste heat into an ever-colder sink. A positive cosmological constant removes the ever-colder sink. There is a temperature floor and a finite horizon, hence a finite total entropy and a finite total amount of information the universe can ever process. The same $\Lambda$ that is the largest discrepancy in physics is also the cap on all the thought the cosmos will ever contain.

So the cosmological constant brackets existence at both ends. It is the measure of how much Something the vacuum hides at the origin, and it is the executioner that freezes the future and bounds the total. Heat death, properly stated, is not classical equilibrium; it is eternal de Sitter dilution toward $T_{dS}$, the physical nothing approached as a limit but, like every limit, never quite the philosophical zero. Life, in this picture, is the brief interval when gradients exist: after the vacuum has been disturbed into matter and stars, and before the expansion has thinned it all back toward the floor. We are made of the first silence and headed for the second, and we are the universe’s short chance to remember the gap between them.

The logical avatar: you cannot certify your own generators

Step back below the cosmology to the claim that organises all of it. “A system cannot derive its own generators from within” is a theorem, not a metaphor.

Let $T$ be a recursively axiomatisable theory strong enough to encode arithmetic. Its generators are its axioms and its rules of inference. Godel’s second incompleteness theorem states that if $T$ is consistent, then $T\nvdash\operatorname{Con}(T)$. Read through this lens, $\operatorname{Con}(T)$ is the statement that the generators produce something rather than collapse into triviality. The theorem says a system cannot prove from inside that its own generating apparatus is even non-degenerate. Lob’s theorem tightens it: if $T\vdash(\Box\varphi\to\varphi)$ then $T\vdash\varphi$, so the system cannot affirm the soundness of its own proofs of an unproved claim without thereby proving it. Generation from within is unbounded; justification of the generators from within is impossible. Any physics we write is at least this rich, so any physics inherits it.

Gauge freedom: the surplus you cannot see

There is a controlled physical instance at the centre of electromagnetism. Under $A_\mu\to A_\mu+\partial_\mu\chi$ the field strength

is unchanged, because mixed partials commute. Every observable is gauge-invariant; it depends on $F_{\mu\nu}$, on Wilson loops, never on $A_\mu$ alone. The generator $A_\mu$ carries surplus structure that no internal observable can resolve, by construction. Align this with the tower: the generating apparatus contains a component that is real, that does work, and that no internal observation can fix. The reason we cannot derive our generators may be the same reason we cannot observe absolute gauge, that the observables are invariant under exactly the freedom the generators carry. On this reading the laws of physics behave like a cosmic gauge, and “why these laws” is not an unanswered question but an unobservable quantity, the fingerprint of a symmetry rather than a hole in the theory.

State plainly what kind of claim this is, because a careful reader will press on it and should. The two facts being aligned are not the same kind of fact. Gauge invariance is true by construction: we build the theory so that observables depend only on $F_{\mu\nu}$, and the redundancy in $A_\mu$ is a feature we put there. Generator underivability is a theorem about formal systems, an external, logical constraint of the Godel-Lob kind, not something a designer chose. So this section is interpretive, an analogy and a conjecture, not a derived isomorphism. What the two share, and what makes the analogy worth pursuing rather than dismissing, is a single precise structural feature: a part of the description that is real, that does work, and that is invisible to every quantity internal to the system. Whether “invisible by construction” and “invisible by theorem” are two faces of one principle, or merely rhyme, is exactly the open question this framing is meant to pose, not to settle.

The fixed-point temptation

If the generators cannot be justified from within, one escape is to let the system generate itself. The mathematics is a fixed point. Lawvere’s fixed-point theorem says that in a sufficiently structured setting, if there is a point-surjective map $A\to B^A$, then every endomorphism $B\to B$ has a fixed point, and this one statement is the engine behind Cantor, Godel, Russell, Turing, and the $Y$ combinator. A self-generating universe would be a fixed point of its own generating map, a structure $U$ with a generator $g\in U$ such that $g(U)\cong U$, reproducing itself with no first step to demand. This does not remove the underivability; it relocates it, trading “what generated the generator” for “does the fixed point exist and is it unique,” questions a system cannot settle from within. What it changes is the shape of the question, from a regress of causes to a condition of self-consistent closure, which can be studied where a first cause cannot. The mystery is not dissolved. It is made into mathematics.

What this opens

Walk the whole arc once, fast. Nothing is not absolute; it is a position in a tower, and the ground of each level is the structure of the level below. The substance of life was forged in stellar interiors and dispersed by stellar death, so the dead sky is the foundry, not the opposite, of the living world. Stars and planets are dissipative structures, alive in the thermodynamic sense and dead only in lacking heritable memory, so life is not a different kind of thing but the universal riding of gradients with a record added. The origin of that record was a phase transition in a reaction network, forced by a gradient, not an assembly from parts. That gradient has a schedule: complex life on Earth has of order one billion years left, all life perhaps two, and we sit late in the window, not early. After it come the death of the Sun and the long eras of the universe, ending in a de Sitter cold set by the cosmological constant, the same constant that measures the hidden Something in the vacuum and caps the total thought the cosmos can ever hold.

The possibilities this leaves open are sharper than the puzzle we began with.

1. The object was never the mystery; the operation is. Producing 1, a photon, a carbon nucleus, a cell, is one application of a generator. The hard, perhaps permanent, problem is the existence of the generator.
2. Life is dissipation plus memory. This is a usable definition, and it predicts where to look. The galaxy is full of the first term. The search for life elsewhere is the search for the second, for gradients that have learned to keep a record.
3. We are late, and finite. Terrestrial life is past its midpoint, and a positive $\Lambda$ makes even idealised eternal life impossible. The total amount of structure, computation, and memory the universe can ever contain is bounded. That bound is a physical quantity, tied to $\Lambda$, and it deserves to be computed rather than waved at.
4. The constant brackets everything. The vacuum energy catastrophe and the fate of the cosmos are the same number read at two ends of time. Any theory that explains $\Lambda$ explains both the bill for existence and the shape of its ending, and the most promising seam, a photon mass of order $\hbar H_0/c^2$ tying the infrared cutoff to the Hubble radius, is where the abstract tower could finally make that number instead of cutting it off.
5. The laws may be a cosmic gauge, and the universe a fixed point. Underivability is then not a defect but a symmetry, and the first-cause regress becomes a self-consistency condition, a question about which structures can close on themselves.

What survives every step, the regular cardinals, the creation operator, the triple-alpha resonance, the autocatalytic threshold, the de Sitter floor, the fixed point, is one sentence: you cannot derive the existence of your generators from within the system they generate. Everything in the sky and on this one warm planet is what those generators have made in the brief, bracketed interval between two silences. We are the part of it that, for a while, can see the gap and ask where it came from. The asking is the most that the window allows, and it is not small.