Insight Editions · No. 236
Cycles of Time
An Extraordinary New View of the Universe
Roger Penrose
Penrose starts from the Second Law and the special Big Bang and builds conformal cyclic cosmology, in which our Big Bang is the stretched-out future of a previous aeon.
A study guide to Cycles of Time by Roger Penrose. Shared free for personal reading; not for sale and not endorsed by the authors or publishers. Please read and buy the original.
The Second Law needs a Big Bang of fantastically low gravitational entropy (1 in 1010124). Penrose explains it with conformal cyclic cosmology: each Big Bang is the rescaled remote future of a previous aeon.
Entropy and phase space; Big-Bang cosmology; null cones, conformal geometry and black holes; Hawking evaporation and information loss; CCC, with its maths in an appendix; and CMB tests.
1010124 vs 101089 phase volumes; entropy per baryon 1021 (black holes) vs 109 (CMB); 10100 years to the last pop; Λ ≈ N−6, N ≈ 1020.
Readers who want the deepest version of "why does time have a direction", and a clear view of a famous, contested alternative to inflation.
The Book on One Page
Roger Penrose builds a new cosmology from one old puzzle: why the universe began in a state special enough to give us a Second Law. His answer, conformal cyclic cosmology (CCC), makes our Big Bang the infinitely stretched-out future of a previous universe, and predicts marks of that universe in the microwave sky.
The puzzle (Part 1). Entropy is Boltzmann's k log V, the logarithm of a phase-space volume. Because the laws are time-reversible, the statistical argument that entropy will rise into the future predicts with equal force that it was higher in the past, an absurd conclusion. Only a fantastically special initial state rescues the Second Law, so the real question is what made the Big Bang special.
The clue (Part 2). The CMB has a perfect thermal spectrum, which looks like maximum entropy. The resolution is gravity: for gravitating matter, uniformity is low entropy and clumping into black holes is high. The Big Bang was hot and in equilibrium in its matter but had its gravitational degrees of freedom switched off, with Weyl curvature zero. By the Bekenstein–Hawking formula, the odds of that by chance are about 1 in 1010124, and inflation, built on time-symmetric dynamics, cannot explain them.
The tools (Part 2). Null cones define causality and the conformal structure of space-time. Massive particles are clocks (ν = mc2/h); massless particles cannot measure time. Conformal diagrams compress infinity into finite pictures. With a positive cosmological constant Λ, future infinity ℐ+ is spacelike, the same character as a Big Bang. Black holes form generically (Penrose's 1965 singularity theorem, Nobel 2020) and eventually evaporate.
The proposal (Part 3). Near the Big Bang and in the remote future, matter is effectively massless, so physics sees only conformal geometry. Penrose identifies each aeon's ℐ+ with the next aeon's Big Bang, glued smoothly by a conformal rescaling. The price: rest-mass must fade to zero for every surviving particle; dark matter is reborn at each bang from the rescaling's "phantom field"; and the matching forces Weyl curvature to vanish at every Big Bang, delivering the low-entropy start.
The entropy accounts. Rescaling cannot change phase-space volume, so what stops entropy piling up aeon after aeon? Penrose's answer: black holes destroy information at their singularities, shrinking phase space, a currency devaluation that resets the zero of entropy. This needs the minority view that quantum evolution is not always unitary, tied to Penrose's long-held belief that gravity causes quantum state reduction.
The test. Supermassive black-hole encounters in the previous aeon should leave concentric circles of uniform temperature shift in our CMB. Hajian's 2008 WMAP analysis looked tantalising; later independent studies found no significant signal, and the later "Hawking points" variant is likewise disputed. CCC remains a bold, testable minority proposal.
| Ch | Title | One-line payload |
|---|---|---|
| Pref. | Whence the Universe Came | From the steady state to an unorthodox scheme with steady-state echoes. |
| Prol. | Tom, Aunt Priscilla and the Water Mill | Life runs on low entropy borrowed from gravity and ultimately the Big Bang. |
| 1 | The Second Law and Its Underlying Mystery | Boltzmann entropy is robust, but backwards it predicts an absurd past unless the Big Bang was special. |
| 2 | The Oddly Special Nature of the Big Bang | Matter began in equilibrium, gravity did not; Weyl curvature zero; odds 1 in 1010124; inflation cannot help. |
| 3 | Conformal Cyclic Cosmology | Massless futures glue to big bangs; mass fades; dark matter reborn; black holes reset entropy; look for CMB circles. |
| Epil. | Tom Watches the Rain | Raindrop rings as an image of the circles test. |
| App. | The Mathematics of the Crossover | 2-spinor equations: conformal invariance, the ϖ-equation, the phantom field and the K-field. |
Whence the Universe Came
Penrose opens with a confession of lineage: he learnt cosmology inside the steady-state camp, watched it fall to the microwave background, and now offers an "unorthodox" scheme that, oddly, echoes the steady state. The book's engine is a single oddity: the Second Law of thermodynamics needs a Big Bang of fantastically low entropy, and nobody has explained why it was so special.
The starting point. When Penrose arrived at Cambridge as a mathematics graduate student in the early 1950s, the ascendant theory was the steady-state model of Bondi, Gold and Hoyle: a universe with no beginning, unchanged in its overall appearance for all time, its expansion offset by the continual creation of an extremely diffuse hydrogen gas. His friend and mentor Dennis Sciama, who taught him "the thrill of so much new physics", was a strong advocate and impressed on him the beauty of the scheme.
The fall of the steady state. About ten years after Penrose arrived, Arno Penzias and Robert Wilson found an all-pervading microwave radiation coming from every direction: the cosmic microwave background (CMB). Robert Dicke identified it as the predicted afterglow of a Big Bang, an origin first seriously proposed by Monsignor Georges Lemaître in 1927 from Einstein's 1915 equations and early evidence of expansion, now placed about 14 thousand million years ago. Sciama, with what Penrose calls "great courage and scientific honesty", publicly repudiated his earlier view and backed the Big Bang from then on.
What the book does. Cosmology has since become an exact science, driven largely by detailed CMB experiments, yet many mysteries remain. Penrose sets out the main classical relativistic models, then the puzzles, chief among them a profound oddness linking the Second Law to the nature of the Big Bang. His own proposal dates from summer 2005, with most detail more recent. The main text stays geometric and nearly equation-free; the technical machinery is "banished" to the Appendices for experts. The scheme is unorthodox, he says, but rests on "very soundly based" geometry and physics, and it has "strong echoes of the old steady-state model". He closes by wondering what Sciama would have made of it.
- Dennis Sciama: Penrose's Cambridge mentor; steady-state advocate who switched to the Big Bang after the CMB.
- Penzias and Wilson: discovered the CMB (1964 to 1965), by accident.
- Robert Dicke: recognised it as Big-Bang radiation.
- Georges Lemaître: proposed the Big-Bang origin in 1927.
- Paul Tod: formulated the conformal-extension version of the Weyl curvature hypothesis, which Penrose calls the most important input to CCC's equations.
- Helmut Friedrich: analysis of conformal infinity with positive cosmological constant, supporting CCC's mathematical viability.
- Wolfgang Rindler: cosmological horizons and the long 2-spinor collaboration.
- Tsou Sheung Tsun, Chan Hong-Mo, James Bjorken: ideas on the nature of mass, later central to the "mass fading" requirement.
Penrose's route to the idea. The book distils threads from half a century of Penrose's work. The singularity theorem (1965) showed that gravitational collapse generically ends in singularities. His conformal diagrams and the notion of ℐ (1960s) gave a way to treat infinity as a place. The Weyl curvature hypothesis (from the late 1970s) proposed that the Big Bang's low entropy is the vanishing of Weyl curvature. His long-standing view that quantum state reduction is gravitational supplied the information-loss mechanism. CCC, conceived in 2005, assembles these pieces: the conformal treatment of infinity is applied to the Big Bang and the remote future at once, and the Weyl hypothesis becomes a consequence rather than an assumption.
Tom, Aunt Priscilla and the Water Mill
A short dialogue at an old working water mill states the whole problem in miniature. Every useful flow of energy on Earth traces back through the Sun to gravitational clumping, and that traces back to the Big Bang, where "nobody really gets" why there was any organisation at all.
The scene. Tom, a sceptical boy, and his aunt Priscilla, a Cambridge astrophysics professor, watch a swollen river drive a preserved woollen mill. Tom asks where the energy that lifted the water up the mountain came from. Priscilla walks him back along the chain:
- Sun to rain. Sunlight makes ocean molecules move faster; the fastest escape as vapour; their energy of motion becomes gravitational potential energy as they rise; they freeze into cloud crystals, fall as rain on high ground, and run down to drive the mill.
- Heat is everywhere, but useless. Even cold river water holds more heat energy in random molecular motion than in its rushing currents. Tom's idea of tiny cup-anemometers harvesting that motion is a Maxwell's demon, and the Second Law forbids it: useful energy cannot be drawn from the random motion of a body at one temperature.
- The Sun does not give the Earth energy. In balance, the Earth re-radiates every joule it receives (global warming holds back a little). What the Sun supplies is organisation: it is a hot spot in a cold dark sky, and that temperature imbalance is what keeps the Second Law at bay.
- Where the organisation came from. The Sun condensed from uniformly spread gas that gravity pulled into clumps, heating as it fell. And the uniform gas came from the Big Bang.
Tom objects that a giant explosion does not sound organised. Priscilla agrees that nobody understands it: it is "one of the biggest puzzles of cosmology". Tom suggests something more organised before the Big Bang; Priscilla lists the bouncing universes, black holes that spawn new universes, and false-vacuum bubbles, then mentions "another theory that I heard about recently", which is the subject of the book.
The mill scene sets the book's question. Life and machines run on low entropy, not on energy, and every reservoir of low entropy we use is borrowed from gravity's unclumped initial state. Explaining that initial state is the task of Parts 2 and 3.
What the mill teaches. The scene is chosen with care. A water mill is the classic picture of energy conversion, and the obvious lesson would be "energy flows from the Sun to the wheel". Penrose uses it to teach the less obvious lesson that energy is conserved and therefore never scarce; what is scarce is the low-entropy form of energy. Tom's intuitions (heat should lift things; the Sun gives us energy; organisation needs no explanation) are exactly the intuitions the book overturns, one per part: Part 1 replaces energy with entropy, Part 2 traces entropy back to gravity at the Big Bang, and Part 3 asks what could have set the Big Bang up.
The Second Law and Its Underlying Mystery
Part 1 builds entropy from scratch as Boltzmann's logarithm of a phase-space volume, shows why it is so robust despite its subjectivity, and then springs the trap. The same statistical argument that predicts entropy rising into the future predicts, with equal force, that it should have been higher in the past; only an extraordinarily special, low-entropy Big Bang rescues the Second Law.
1.1 The relentless march of randomness
A law that is an inequality. Most laws of physics are equalities: Newton's second law equates rate of change of momentum with force; conservation laws equate total energy, charge, momentum and angular momentum at different times; E = mc2 equates energy with mass times c squared; Newton's third law makes action equal and opposite to reaction. The First Law of thermodynamics is energy conservation with heat included, where temperature is the thermal energy per degree of freedom; a projectile slowed by air loses kinetic energy, but molecules in air and projectile gain random motion, so nothing is lost. The Second Law is different: it is an inequality, saying only that the entropy of an isolated system (a measure of disorder or randomness) is greater, or at least not smaller, later than earlier.
Weak but universal. Penrose lists three apparent weaknesses: the law is only an inequality; the definition of entropy carries vagueness or subjectivity; and most formulations admit occasional fluctuations in which entropy temporarily falls. Against this, its reach is enormous. It holds in relativity as in Newtonian theory, for Maxwell's continuous fields as for discrete particles, and even for hypothetical dynamics with no bearing on our universe. It bites hardest for deterministic, time-reversible dynamics such as Newton's, where every allowed history run backwards is another allowed history.
The egg film. A film of an egg rolling off a table and smashing obeys Newton's laws; so does the film run backwards, in which shell, yolk and albumen reassemble and the egg leaps onto the table. Both are dynamically lawful; only one is consistent with the Second Law, and the other is so improbable we reject it outright. Strictly the law says a system is overwhelmingly likely, not certain, to become more random. Penrose notes subtleties from black-hole physics and quantum mechanics that he defers to §3.4.
Life is not the mystery. Some see the sophistication of biology as a contradiction of the Second Law. Penrose says it is not (he explains why in §2.2): biology is fully consistent with overall entropy increase. The mystery he means is "of an entirely different order of scale".
Why the Second Law cannot follow from the dynamics. The entropy definition at any instant is time-symmetric (the same for the egg whichever way the film runs). If the dynamics are also time-symmetric and entropy is not constant, then for every entropy-increasing history there is a lawful entropy-decreasing one. So the Second Law is a separate principle adjoined to the dynamical laws, not a deduction from them.
Hand a projectionist a reel with no labels and ask which way to run it. A billiard break run backwards (balls converging, stopping dead in a perfect triangle) gives itself away instantly; a single frictionless collision of two balls does not. The difference is not in the laws, which are reversible in both cases, but in how many particles' worth of order the film displays at one end. The more degrees of freedom involved, the more lopsided the counting, and the more certain the verdict. Penrose's whole argument rides on this: reversibility of the laws plus lopsided counting gives a direction of time only once one end of the film is pinned to a special state.
Going deeper: the several arrows of time
Physicists distinguish several arrows of time, and Penrose's argument implies they share one root. The thermodynamic arrow is entropy's rise. The radiative arrow: waves spread outward from sources (ripples leave a dropped stone; they never converge on it spontaneously), even though Maxwell's equations allow both. The psychological arrow: we remember the past, not the future. The cosmological arrow: the universe expands. The quantum measurement arrow: state reduction produces records only toward the future. Penrose's view, developed in the next part, is that the first three at least are all downstream of a single boundary condition, the low-entropy Big Bang. Records, memories and outgoing waves are all low-entropy correlations that only a low-entropy past can supply. Whether the expansion and quantum reduction share the same root is exactly what Part 3 explores.
1.2 Entropy, as state counting
Red and blue paint. Pour red and blue paint into a pot and stir: the colours lose their identity and the pot turns uniform purple, and no amount of stirring separates them again, though the molecular processes are reversible. Separated colours mean low entropy; uniform purple means much higher entropy.
The ball-crate model. Penrose idealises the pot as an N × N × N crate of compartments, each holding one red or one blue ball. To judge colour, group the compartments into k3 boxes of n × n × n each, with N = kn and N ≫ n ≫ 1. A box with r red and b blue balls (r + b = n3) has a hue set by r/b; call the pot uniform purple if every box has r/b between 0.999 and 1.001.
- 1024 molecules in a can of paint, so N = 108.
- Pixels of 10−2 cm look smooth, so k = 103 and n = 105.
- Arrangements that look uniform purple: about 1023 570 000 000 000 000 000 000.
- Arrangements with all blue on top, red below: a "mere" 1046 500 000 000 000.
- Probability of finding the separated state at random: about 10−23 570 000 000 000 000 000 000, barely changed if only 99.9% of the blue need be on top.
Why a logarithm. These counts are too unruly to use directly, so entropy uses their logarithm. Penrose gives a primer on log10: log10 1 = 0, log10 10 = 1, log10 1000 = 3; for other whole numbers the integer part is the digit count minus one (log10 2 = 0.301..., log10 53 = 1.724..., log10 9140 = 3.960...). The key property is log(ab) = log a + log b. If one independent part can arise in P ways and another in Q ways, the whole arises in PQ ways, so defining entropy as proportional to the log of the count makes it additive for independent systems.
Configuration space and coarse graining. Real molecules have continuous positions, orientations and internal distortions, so counting gives way to volumes in configuration space: for q point particles it has 3q dimensions, one point P encoding every particle's position. Nobody can picture such spaces; 2D and 3D sketches suffice if their limits are kept in mind, and these are abstract spaces, not physical space or space-time. Configurations that "look the same" by macroscopic measures (density, colour, chemical composition, not exact atomic positions) are lumped into a coarse-graining region, and entropy is the logarithm of that region's volume. The phrase "look the same" is vague, but the definition survives because coarse-graining volumes differ by stupendous ratios.
Take a toy pot of just 100 balls, 50 red and 50 blue, in a row. The number of arrangements with all red on the left is 1; the number with any mix is the binomial coefficient C(100, 50) ≈ 1.01 × 1029. Entropies (natural log, k = 1): 0 versus about 66.8. Double the system to 200 balls and the mixed count becomes about 9.05 × 1058 (the square, roughly), while its log only doubles to about 135. That is the additivity Penrose wants: counts multiply, entropies add, and even a modest gain in log units hides an astronomical gain in probability.
| Term | Meaning in Part 1 |
|---|---|
| Configuration space | All possible positions of every particle; 3q dimensions for q point particles. |
| Phase space | Positions plus momenta; twice the dimension; one point is the whole system's state. |
| Coarse graining | Partition of phase space into regions that look macroscopically identical. |
| Boltzmann entropy | S = k log V, V the volume of the coarse-graining region containing the state. |
| Liouville's theorem | Hamiltonian evolution preserves phase-space volume. |
| Thermal equilibrium | ℛmax, the largest coarse-graining region, bigger than all others combined. |
| Past hypothesis | The universe began in an extraordinarily small region ℬ. |
1.3 Phase space, and Boltzmann's definition of entropy
Why positions are not enough. Shake a bottle of half water, half olive oil: in moments the oil separates to the top, yet entropy rose throughout. The oil molecules attract each other strongly, so as they gather their velocities increase in vigorous mutual orbits, and it is the motion part of the state space that supplies the extra volume. Motions are needed anyway, so that the state fixes the future under Newton's laws.
Phase space. The right arena is phase space, with twice the dimensions of configuration space: every position coordinate gets a momentum coordinate (angular momentum for angular coordinates), momentum rather than velocity because of the deep link to Hamiltonian theory. For a point particle that is 3 numbers of position and 3 of momentum. The whole system's instantaneous state is one point p. Under deterministic Hamiltonian laws (Newton's, or Maxwell's fields) p traces an oriented evolution curve, and the whole space is foliated by such curves "like a bale of straw": follow the arrow for the future, go against it for the unique past.
Quantum units make volumes comparable. Since quantum theory, phase-space volumes are effectively dimensionless numbers, measured in units with ħ = h/2π = 1, where ħ = 1.05457 × 10−34 J s. That lets regions of different dimension be compared, which classically makes no sense, and it is what gives phase space its granularity, the "quanta". The granularity normally goes unnoticed because the numbers are huge; the exception is Planck's 1900 black-body spectrum, where equilibrium mixes states with different photon numbers, hence phase spaces of different dimension.
- S = k log V, where V is the volume of the coarse-graining region containing the state point, measured in units of ħ.
- k = 1.3805 × 10−23 J/K, Boltzmann's constant (natural log). With log10, the constant becomes k′ = k × 2.302585... = 3.179 × 10−23 J/K.
- Coarse graining lumps states indistinguishable by macroscopic parameters: temperature, pressure, density, flow, colour, chemical composition.
- Additivity: for independent subsystems, k log(VW) = k log V + k log W.
Low entropy versus specialness. The smashed egg on the floor is itself a very special state: reverse every particle's motion and it will reassemble and hop back onto the table. But that specialness lives in subtle correlations, not in the macroscopic parameters. Low entropy means manifest specialness, visible in macroscopic values. High-entropy states that evolve to lower entropy exist, but they are a tiny minority, and that, Penrose says, is "the whole point" of entropy and the Second Law.
Liouville's theorem. Hamiltonian evolution preserves phase-space volume: a region of volume V at time 0 becomes a region of the same volume V at time t. No contradiction with the Second Law: the evolved region becomes a sprawling, messy filament that threads a far larger coarse-graining region, because the dynamics ignore the coarse-graining boundaries.
The laboratory inside the galaxy. An experimenter defines entropy using the lab's degrees of freedom, phase space 𝒫 and region volume V. Embed the lab in the Milky Way: the full space is the product space 𝒫 × 𝒳 (dimensions add, like a plane times a line giving a 3-space), and with independent subsystems the region volume is VW, so the entropy is k log V + k log W. The external part k log W is irrelevant to the experiment and can be dropped. Penrose flags that in §3.4, with black holes and the universe as a whole, such "irrelevant" external entropy acquires fundamental significance.
1.4 The robustness of the entropy concept
Boltzmann, 1875. Boltzmann's definition was a large advance because it applies to systems not in a stationary state. Its weak point is what counts as a macroscopic parameter. Future technology might track molecules individually, forcing a finer coarse graining and a lower entropy for the same state, so entropy would seem to depend on the state of technology. Some argue that finer measurement always raises entropy in the apparatus by more than it lowers it in the system; Penrose accepts the point but says it leaves the objectivity of "macroscopic" unresolved.
Maxwell's demon. James Clerk Maxwell imagined a tiny demon opening and closing a door to sort fast and slow molecules, violating the Second Law for the gas. Include the demon's own sub-microscopic body, or replace it with a minute machine, and the law should be restored for the whole. Penrose still finds the definition "somewhat enigmatic" for such systems.
Why the subjectivity barely matters. Redrawing coarse-graining boundaries changes entropy values very little because the regions differ by stupendous factors. In the paint model, demanding ten times finer colour precision (r/b between 0.9999 and 1.0001) over boxes of half the side (one eighth the volume) leaves the entropy of "uniform purple" almost unchanged.
Mix 75 litres of hot water at about 50 °C with 75 litres of cold at about 10 °C (150 litres total). The entropy rises by about 21 407 J/K, meaning the state point moves into a coarse-graining region roughly 101027 times larger. No plausible redrawing of region boundaries dents numbers of that size. (Penrose sets aside the "not inconsiderable" entropy of the actual cleansing.)
Check: with these temperatures, ΔS = mc ln(Tf2/T1T2) gives nearer 2.7 kJ/K (a factor near 101026). The printed figure looks inflated, but the point about scale survives either way.
Fuzzy boundaries. Strictly, coarse-graining regions have no sharp edges: two almost identical states on either side of a boundary would count as macroscopically distinct. Penrose's fix is a fuzzy boundary zone whose points we simply ignore; it occupies a negligible volume compared with the vast interiors.
- Spin echo (Erwin Hahn, 1950), used in NMR. Aligned nuclear spins, driven by a varying field, precess at different rates into an apparently random jumble; reverse the field pattern and the original magnetisation returns. By macroscopic measures entropy rose and then fell, a gross apparent violation.
- The glycerine cylinders. Two snug glass tubes with viscous fluid between them and a straight line of red dye. Turn the inner tube several times: the dye smears into a faint uniform pink. Turn it back the same number of times: the line reappears, nearly as sharp as before.
- The lesson. The jumbled state carries a precise hidden order, like data on a CD or DVD that crude measurement misses but a player reads easily.
Penrose's position. The usual view is that the Second Law holds and the entropy definition was not refined enough. Penrose calls the demand for a precise, objective, universally valid entropy "a can of worms" and sees no reason entropy must have a truly fundamental, objective role in physics. Entropy is useful because, in the universe we actually meet, ordinary macroscopic measures produce coarse-graining volumes that differ by stupendously large factors. Why they differ so enormously is the objective fact "out there" that the subjectivity merely clouds, and it is the book's real target.
Going deeper: what "subjective" entropy does and does not mean
Penrose's position on entropy is subtle and easily misread. He is not saying entropy is a matter of opinion. He is saying that its exact value depends on a choice, the coarse graining, and that this choice reflects what we can measure. Two observers with different instruments will assign slightly different numbers to the same pot of paint, but they will agree, to an accuracy that dwarfs any practical concern, about which states are high and which are low, and about the direction in which entropy changes. The objectivity lives in the ordering and the ratios, not in the decimals.
The spin-echo and dye-cylinder cases show where the choice can bite. There, an apparently disordered state is secretly an ordered state encoded in correlations, and an experimenter who knows the code can extract the order back. Two readings are possible. The conventional one says the "true" entropy never rose, because a fine enough coarse graining would have tracked the hidden order. Penrose's reading is more relaxed: there need be no single correct entropy for every system, and nothing in physics demands one. What demands explanation is the brute fact that the world we meet is built so that crude macroscopic measures produce wildly unequal volumes, which is another way of saying that the world is very far from equilibrium. That fact is not subjective, and its origin is the Big Bang.
This also clarifies the role of Maxwell's demon. The demon is not a threat to the Second Law once the demon itself is counted; the modern resolution (Landauer and Bennett) locates the entropy cost in erasing the demon's memory. Penrose sidesteps the details because they do not touch his main point: whatever bookkeeping one adopts for demons and devices, the bookkeeping only works if the universe supplies a large stock of low entropy to begin with.
1.5 The inexorable increase of entropy into the future
The forward argument. Start the state point p0 in a fairly small region ℛ0. Neighbouring regions differ in size enormously, and in high dimension each region has many neighbours (typically 6 in 2D, 14 in 3D, rising fast with dimension). When p leaves ℛ0 it almost certainly enters a hugely larger region ℛ1, then a larger ℛ2, then ℛ3; entering a far smaller region would be finding a needle in a haystack, only vastly harder. Entropy, the logarithm of volume, rises more modestly than the volumes but relentlessly. Decreases are not impossible, merely overwhelmingly unlikely. The deterministic dynamics behave as if random with respect to coarse graining.
Why not jump straight to equilibrium. The largest region ℛmax is thermal equilibrium, likely bigger than all others combined; p eventually reaches it and stays, apart from rare thermal fluctuations. It does not leap there at once because the evolution is continuous: the state a moment later differs little from the state now, so entropy climbs gradually.
The trap: run the argument backwards. Ask instead how the egg came to be perched on the table. Newton's laws work equally in the past direction, so apply the same reasoning: there are enormously more evolution curves entering ℛ0 from huge regions ℛ′1, ℛ′2, ℛ′3 than from the shrinking sequence ...ℛ−3, ℛ−2, ℛ−1 that actually occurred. The "most probable" past is then a mess on the floor that assembled itself into an egg and jumped onto the table, rather than a careless person putting it there. The argument that supported the Second Law forward gives "about as completely wrong" an answer as possible backwards.
Find a pond with a thin, perfectly smooth sheet of ice that has a single stone sitting in the middle. Statistics applied blindly backwards say the most probable history is that the stone leapt out of the water while a random thermal fluctuation froze the surface flat around it. Everyone instead infers that the ice formed first and someone threw the stone. That inference is not physics alone; it is physics plus a low-entropy past (a thrower, a cold night, a Sun-driven climate). Every ordinary act of reconstructing history, from forensic science to geology, silently imports Penrose's past hypothesis.
1.6 Why is the past different?
Past-teleology. The flaw is the assumption that evolution is unbiased with respect to coarse graining. Forward it is; backward it plainly is not. Viewed in reverse, the egg's history seems guided through exceedingly improbable steps toward an improbable goal. In the future direction we would call that teleology or magic; in the past direction it is everyday experience. Penrose stresses this is "hardly a physical explanation", just observational fact: the dynamics are unconcerned with coarse graining toward the future, while the past history seems to "deliberately" seek ever smaller regions.
The resolution. Suppose the universe's initial state occupied a coarse-graining region ℬ of quite exceptional tininess. Given continuity of entropy, a gradual increase follows, which is the Second Law. So "the key issue is indeed the specialness of the Big Bang", whose low entropy, Part 2 shows, has "a rather subtle character". The argument also needs no matching tiny future region ℱ imposing a goal on late evolution.
Arguments Penrose rejects
- Psychological arrow: we call "future" whichever direction entropy grows, so the Second Law is automatic.
- Anthropic: life needs the Second Law (natural selection), so we must find ourselves where it holds.
- Life's improbability as evidence that entropy was lower in the past.
Why they fail
- We know almost nothing of what consciousness physically needs; and entropy is useful only because the universe is far from equilibrium, with one end of its history pinned to a tiny region. The tininess of ℬ is untouched.
- Equally weak on what life needs, and it cannot explain why the same law holds in galaxies thousands of millions of light years away and long before life.
- Without a prior low-entropy assumption, the most probable origin of today's living Earth is a random "miraculous" assembly, not evolution; life presupposes the law rather than proving it.
Could the Second Law reverse? It is only observation that there is no low-entropy constraint in the remote future. The ~1.4 × 1010 years since the Big Bang are "as nothing" beside the projected future. A universe whose curve must end in a tiny region ℱ would, very late, show strange particle correlations and teleological behaviour like the self-assembling egg. Curves pinned at both ends are an even tinier fraction than those pinned at one, but their logical status is similar, so they cannot be dismissed. Penrose does not find a reversal plausible and his proposal does not use it, but it is "not intrinsically absurd", and the open mind it requires is the attitude he asks for in Part 3.
- The laws are time-reversible; the Second Law is an extra principle.
- Entropy is k log V of a coarse-graining region; subjective at the edges, robust because volumes differ stupendously.
- Starting in a small region, the state point almost surely wanders into ever larger regions: entropy rises.
- The same argument run backwards predicts an absurd past, so the actual past must have been constrained.
- The constraint is a Big Bang in a region ℬ of exceptional tininess; Part 2 asks what kind of tininess.
Penrose's insistence that entropy is not fundamental sits alongside his use of it as the book's central measure. His defence is that the ratios of coarse-graining volumes, not the precise values, are objective. Critics such as Sean Carroll and David Albert accept the "past hypothesis" (a low-entropy initial state) as the needed extra postulate but differ on whether it needs a dynamical explanation at all; some (Price) argue the demand for one smuggles in a time asymmetry. Penrose takes the strong view that the past hypothesis is a fact requiring physics to explain it, which is the premise everything in Parts 2 and 3 depends on.
Boltzmann's S = k log V and Shannon's H = −Σ p log p are the same idea in two settings: both measure the logarithm of the number of possibilities compatible with what is known. When every microstate in a coarse-graining region is equally likely, Shannon's formula reduces to Boltzmann's. This is why Penrose can speak of black holes destroying information and of phase space losing volume as if they were one event: lost degrees of freedom are lost possibilities, and lost possibilities are lost bits.
- Landauer's principle verified. In 2012, Bérut and colleagues (Nature) measured the heat released when a single colloidal bead "bit" is erased and found it approaches kT ln 2, the minimum Landauer predicted. Experimental Szilard engines and electronic Maxwell's demons (2010s onward) extract work from information exactly as the bookkeeping requires, confirming that demons pay for their sorting in memory erasure.
- Fluctuation theorems. The Jarzynski equality (1997) and Crooks theorem (1999) turned the "occasional" entropy decreases Penrose mentions into exact laws: for small systems, the probability of a given entropy decrease relative to the matching increase falls off as e−ΔS/k. They are routinely tested in single-molecule experiments (RNA hairpins pulled by optical tweezers).
- The past hypothesis debate. Philosophers and physicists (David Albert's Time and Chance, 2000; Sean Carroll's From Eternity to Here, 2010; the Boltzmann brain problem, which asks why we are not lone fluctuations) have sharpened Penrose's point: without a special initial state, the most probable observer is a random fluctuation, not a product of history. Most now treat a low-entropy beginning as unavoidable; disagreement is over whether it needs a further explanation.
For anyone building models of risk or AI systems, Part 1 is a lesson in base rates and conditioning. "Most probable history" depends entirely on which end you constrain: forecasting forward from a known state is easy, but inferring causes backward without a prior (here, the low-entropy past) yields absurd reconstructions. The same trap appears in back-testing, root-cause analysis and model explanations that reverse-engineer causes from outcomes.
The Oddly Special Nature of the Big Bang
Part 2 turns the abstract puzzle into a concrete one. The microwave background shows a perfect thermal spectrum, which looks like maximum entropy, yet the Second Law demands a minimum. Penrose's resolution: matter and radiation began in equilibrium, but gravity began in its lowest-entropy, unclumped state, with zero Weyl curvature, and the odds of that by chance are about one in 1010124. Along the way he builds the tools Part 3 needs: null cones, conformal geometry, conformal diagrams, black holes and Hawking evaporation.
2.1 Our expanding universe
The evidence for expansion. The case for an explosive origin began with Edwin Hubble in 1929 (anticipated by Vesto Slipher in 1917): distant galaxies recede at speeds roughly proportional to distance, so running the film backwards brings everything together at about the same time. The evidence is the red shift of spectral lines, read as a Doppler shift, larger for more distant galaxies and consistent with uniform expansion. Later work mapped the expansion history in detail and fixed a date of about 1.37 × 1010 years for the Big Bang (with "noteworthy dissenting voices" on details).
Hubble's law in numbers. The relation is v = H0d: recession velocity proportional to distance, with the Hubble constant H0 now measured at about 67 to 73 km/s per megaparsec (the two ends of the range being the present "Hubble tension"). Its reciprocal, 1/H0, is about 14 thousand million years, which is why a simple backwards extrapolation lands close to the age that detailed models give. The red shift z is the fractional stretch of wavelengths; for nearby galaxies it is roughly v/c, while the CMB was emitted at z ≈ 1100, which is why 3000 to 4000 K radiation now arrives at 2.7 K.
No centre. Following Einstein, the Big Bang encompassed all of space, not a point within it; space itself was "very tiny". The balloon analogy helps: the surface expands with no central point on it, the centre of the balloon not being part of the surface that represents the universe.
Dark matter and dark energy. The expansion matches general relativity only with two ingredients Penrose calls "somewhat unfortunate" names: dark matter, an invisible material "constituting some 70% of the material substance" of the universe, and dark energy, most plausibly Einstein's 1917 cosmological constant Λ, small and positive (which Einstein later retracted). Both matter to CCC. General relativity is superbly tested: GPS depends on it, and the binary pulsar PSR 1913+16 has been timed over some 40 years to about 10−6 s per year, an overall precision near one part in 1014.
- K > 0, positive spatial curvature: space a 3-sphere; finite in space and time, from Big Bang to Big Crunch.
- K = 0, flat: Euclidean space; expands for ever.
- K < 0, negative curvature: hyperbolic space; expands for ever.
- All are FLRW models (Friedmann 1922, 1924; Lemaître; Robertson; Walker): spatially homogeneous and isotropic. All start in a singularity of infinite density and curvature where "Einstein's equations give up".
- Temporal behaviour mirrors spatial: spatially finite means temporally finite.
| Model | Curvature K | Λ | Fate | Future infinity ℐ+ |
|---|---|---|---|---|
| Closed Friedmann | > 0 | 0 | Big Crunch | none (ends in a crunch) |
| Flat Friedmann | 0 | 0 | Expands for ever, slowing | null (45°) |
| Open Friedmann | < 0 | 0 | Expands for ever | null (45°) |
| Any of the above with Λ > 0 | any (Λ wins) | > 0 | Exponential expansion | spacelike |
| Einstein static (1917) | > 0 | tuned | Static, unstable | none |
| de Sitter | (empty) | > 0 | Exponential, no matter | spacelike |
| Steady state | half of de Sitter | creation | Unchanging appearance | spacelike; past-incomplete |
Escher's three geometries. Maurits C. Escher drew all three 2D cases as tessellations of angels and devils: elliptic (K > 0), Euclidean (K = 0) and hyperbolic (K < 0).
Acceleration since 1998. The supernova teams led by Saul Perlmutter and Brian P. Schmidt found the expansion accelerating, explained by a small positive Λ. With Λ large enough, even a closed (K > 0) universe avoids recollapse, and the sign of K becomes irrelevant to the late, exponential expansion. Observations put the spatial geometry close to flat, which Penrose finds "somewhat unfortunate": a small curvature of either sign can never be excluded, so we may never know whether space is finite.
Inflation, briefly. Many cosmologists take flatness as support for cosmic inflation: an exponential expansion between about 10−36 and 10−32 s, by a factor of perhaps 1030 or 1060 (even 10100). Penrose declares himself "not enthusiastic", despite its near-universal acceptance, and promises that CCC offers credible alternatives (§3.5).
Why Escher matters here. The Circle Limit print uses the representation of the hyperbolic plane due to Eugenio Beltrami (1868), rediscovered some 14 years later by Henri Poincaré. Penrose calls it the conformal representation: angles are true, sizes shrink toward the edge, and the bounding circle represents infinity as a smooth finite boundary. That idea returns in §2.5 and §3.2.
2.2 The ubiquitous microwave background
The steady state and its fall. Thomas Gold and Hermann Bondi (1948), joined by Fred Hoyle, all at Cambridge, proposed continual creation of hydrogen at about one atom per cubic metre per thousand million years, exactly offsetting dilution by expansion; no origin needed, many properties deduced from self-propagation. Penrose entered Cambridge in 1952 and returned as a research fellow in 1956, knew all three, and found the model appealing. Martin Ryle's radio galaxy counts at the Mullard Radio Observatory began to tell against it; the "real death-blow" was the 1964 accidental detection by Penzias and Wilson of microwave radiation from every direction.
- Late 1940s: George Gamow and Robert Dicke predict the "flash of the Big Bang", cooled by expansion from about 4000 K to a few kelvin.
- 1964: Penzias and Wilson detect radiation at about 2.725 K; Dicke explains it.
- 1978: Nobel Prize to Penzias and Wilson.
- Nov 1989: NASA's COBE launches; 2006 Nobel to George Smoot and John Mather.
- ~379 000 years after the Big Bang (about 1/36 000 of the present age): the surface of last scattering, when protons and electrons combined into neutral hydrogen (plus about 23% helium, the α-particles made in the first minutes) and the fog cleared.
Two blatant facts. Penrose focuses on two features COBE made evident, leaving finer anisotropies for §3.6.
- A perfect thermal spectrum. The CMB matches Planck's 1900 black-body curve so closely that, with error bars exaggerated 500 times, the points still sit on the line; it is the best agreement between an observed spectrum and the Planck curve in all of observational science.
- Near-perfect uniformity. After removing the dipole due to the Earth's "proper motion" (orbit, solar motion around the galaxy, the galaxy's own motion), temperature varies across the sky by only a few parts in 105.
From COBE to Planck. COBE's FIRAS instrument measured the spectrum; its DMR instrument first detected the faint temperature ripples (in 1992). WMAP (2001 to 2010) mapped them across the sky, the data Hajian would later search for circles, and Planck (2009 to 2013, final release 2018) mapped them with still finer resolution. The ripples are the seeds of all later structure, and their statistics are where CCC must eventually be tested.
A Planck spectrum is the signature of thermal equilibrium, the largest coarse-graining region, maximum entropy. Part 1 demanded the opposite: a Big Bang of extraordinarily tiny entropy. Expansion does not save us: it is adiabatic, entropy-preserving, as R. C. Tolman showed in 1934, so the expansion is a succession of maximal regions of equal volume. Worse, a hot, uniform start is just what standard relativistic cosmology predicts. "Something appears to have gone seriously wrong."
Gravity reverses the rules. Release gas from a corner compartment of a box and it spreads evenly: uniform means high entropy. Replace molecules with stars in a galactic-sized box: gravity makes them clump, moving faster as they do, so uniform means low entropy and entropy rises with clumpiness. In Newtonian theory there is no maximum: point masses can approach without limit. In Einstein's theory the clumping saturates in a black hole. Today the largest entropy contributions by far are supermassive black holes, such as the one of about 4 000 000 solar masses at the Milky Way's centre, swamping the CMB's entropy.
Why Newtonian gravity has no equilibrium. In a box of gas, collisions share energy until every molecule's average is the same; that shared state is equilibrium. Under Newtonian gravity, a pair of stars that fall closer gains kinetic energy while the pair's total energy falls, and the surplus can be handed to a third star that is flung outward. Repeat, and a core contracts and heats while a halo expands: self-gravitating systems have negative heat capacity, getting hotter as they lose energy. There is no floor to this in Newton's theory, since point masses can approach arbitrarily closely. General relativity supplies the floor by trapping the core in a black hole, which is why Penrose says the question of gravitational equilibrium can only be posed in Einstein's theory.
So the CMB's uniformity is the low-entropy fact: matter and radiation were in equilibrium, but gravitational degrees of freedom were hugely suppressed. The cosmological principle of homogeneity and isotropy, basic to FLRW models, is the universe's low initial entropy.
From the Big Bang to the egg. The everyday Second Law traces back to this. The puzzle is not the egg's smashing but how it got onto the table in a low-entropy state. A person placed it, drawing on a small part of a large reservoir of low entropy (a well-fed human in oxygen-rich air); the hen built it, part of life's organisation, maintained by natural selection and chemistry. Life does not escape the Second Law; it runs on the Sun.
The Earth returns to space as much energy as it receives; otherwise it would heat until it did. Sunlight arrives as fewer, high-frequency (yellow) photons and leaves as many more low-frequency infrared photons. By E = hν, the same energy in more photons means more degrees of freedom, so by S = k log V the incoming energy carries much lower entropy than the outgoing. Plants use this through photosynthesis (taking carbon from CO2, returning O2); animals eat plants or plant-eaters. If the whole sky were as hot as the Sun, its energy would be useless to life.
A photon's typical energy scales with the temperature of its source. Sunlight is emitted at about 5800 K; the Earth radiates at an effective 255 K. The ratio is about 23, so for every solar photon absorbed, roughly twenty infrared photons must leave to carry the same energy away. Radiation entropy scales roughly with photon number, so the Earth exports about twenty times the entropy it imports. That surplus pays for weather, rivers, photosynthesis and every living thing; the Earth is less a power station than an entropy pump driven by a hot spot in a cold sky.
Surround the Earth with a sky everywhere at the Sun's 5800 K. Energy pours in from every direction, but none can leave, and soon the Earth sits in equilibrium at 5800 K: no rain, no wind, no life, every joule useless. Now replace it with a sky everywhere at 2.7 K and no Sun: the Earth cools to near absolute zero. Life needs both: a concentrated hot source and a cold sink. That combination exists only because gravity gathered gas into stars against a dark background, which is why Penrose traces the egg back to the Big Bang's unclumped gravity.
Why the Sun is a hot spot. Thermonuclear fusion of hydrogen to helium matters, but the key is that the Sun exists, held together by gravity; without fusion it would still shine, shrinking, hotter and short-lived. Its existence comes from gravitational clumping of initially uniform material, so the ultimate source is a Big Bang whose gravitational degrees of freedom were not activated. Penrose calls this "curiously lop-sided" and digs into Einstein's geometry to understand it.
2.3 Space-time, null cones, metrics, conformal geometry
Minkowski's 4-space. In 1908 Hermann Minkowski, once Einstein's teacher at the Zurich Polytechnic, recast special relativity as a 4-dimensional geometry of events. Einstein was unenthusiastic at first, then made it the basis of general relativity. The revolutionary point is that the geometry does not split naturally into time plus a stack of Euclidean 3-spaces of simultaneous events: simultaneity depends on the observer's velocity.
Two people stroll past each other in opposite directions on a path. The events on the Andromeda Galaxy that each would call simultaneous with their passing differ by several weeks. Small speeds produce large disagreements over large distances. Minkowski's achievement was to "take the relativity out of special relativity", giving an objective geometry independent of any observer's slicing.
Null cones and causality. Minkowski's basic structure is the null cone at each event p: a double cone (past and future) tracing how light converges on and spreads from p, drawn at 45° with units where c = 1 (light-seconds of 299 792 458 m, light-years of about 9.46 × 1012 km). Massive particles have world-lines inside the cones; massless ones (photons) run along them. P can influence q only if a future-directed causal path (on or within the cones) joins them, which requires a time orientation. In general relativity the cones are no longer uniform, tilting from place to place.
Rubber sheets and covariance. Picture the cones printed on an ideal rubber sheet: smooth deformations (diffeomorphisms, general coordinate transformations) leave causality untouched. General covariance demands that physical laws be indifferent to them. Escher's devils can be slid into each other's places, a symmetry of hyperbolic geometry; such symmetries occur in GR (FLRW) but are exceptional.
Metric versus conformal structure. A manifold is smooth but otherwise bare. A metric g (a tensor) assigns lengths to curves, hence geodesics (shortest routes) and angles. In Escher's picture, hyperbolic geodesics are circular arcs meeting the boundary at right angles, and distance between p and q is C log(|qa||pb| / |qb||pa|), with C the pseudo-radius. Conformal geometry keeps only angles (equivalently, ratios of lengths in different directions at a point, so infinitesimal shapes), not lengths. Rescaling g ↦ Ω2g with any smooth positive Ω leaves conformal structure unchanged. The hyperbolic plane is conformally identical to the interior of the Euclidean disc.
- Signature 1+3. One timelike, three spacelike orthogonal directions; flat is Minkowskian, curved is Lorentzian.
- The metric measures time, not distance. A ruler is a strip in space-time; measuring pq needs p and q simultaneous in its rest frame, checked with light signals and clocks. J. L. Synge proposed calling it "chronometry".
- Every massive particle is a clock. E = mc2 and E = hν give ν = mc2/h: frequency exactly proportional to mass.
- Massless particles cannot tell time. A photon's frequency would be zero; it "would take until eternity" to reach its first tick.
- Null-cone structure equals conformal structure. Timelike geodesics maximise proper time (a reversal of Euclidean minimising); null geodesics have zero length and are fixed by the cones alone.
Fact 4 is the hinge of the whole book: without mass there is no clock, and without a clock the scale of the universe has no physical meaning.
For an electron, m = 9.11 × 10−31 kg, so ν = mc2/h = (8.19 × 10−14 J)/(6.63 × 10−34 J s) ≈ 1.24 × 1020 Hz, a few hundred thousand times faster than the optical transitions in the best atomic clocks. A proton ticks about 1836 times faster still. Halve a particle's mass and its clock runs at half the rate; let the mass fade to zero, as CCC will require, and the tick rate goes to zero with it. Mass is not merely what clocks are made of: in this precise sense mass is the clock.
Strap a wristwatch to a photon leaving a distant galaxy. Along a null world-line the metric assigns zero elapsed time, so the watch never advances. To the photon, emission and absorption, ten thousand million years apart by our clocks, are the same instant, and so is the end of time. A universe containing only such passengers has no way to register duration or size: shrink everything uniformly, or stretch it, and nothing aboard could tell. This is the seed of CCC.
| Term | Meaning in Part 2 |
|---|---|
| Null cone | Light's paths through an event; fixes causality and the conformal structure. |
| Metric g | Assigns proper time to causal curves; "chronometry". |
| Conformal structure | Angles and null cones without scale; invariant under g ↦ Ω2g. |
| Geodesic | Free-fall path; timelike geodesics maximise proper time. |
| Diffeomorphism | Smooth "rubber-sheet" relabelling that leaves physics unchanged. |
| Signature 1+3 | One time direction, three space directions. |
Going deeper: Einstein's train and the meaning of the metric
Penrose's ruler argument repays a slower reading because it is where time quietly displaces space as the primary notion. Put an observer at the front of a uniformly moving train (or ruler). Event p is at the front, event q at the back. To check that p and q are simultaneous in the train's frame, the observer sends a light signal at event r, timed to reach the back exactly at q, where it is reflected and returns at event s. Simultaneity holds if p sits exactly half-way in time between r and s. Only then does the train's length equal the space-time separation of p and q.
Two lessons follow. First, every spatial measurement in relativity is, underneath, a pair of time measurements plus light signals. The metric g is fundamentally a device for assigning proper time to world-lines of massive bodies, which is why Synge's word chronometry fits. Second, the light signals themselves measure nothing: a null curve has zero proper time. If every clock in the universe vanished, light alone could not reconstruct the metric. It could reconstruct only the null cones, hence the conformal structure. Part 3 turns this lesson into cosmology.
2.4 Black holes and space-time singularities
Collapse. When a star of perhaps 10 or more solar masses exhausts its nuclear fuel, it collapses unstoppably. When escape velocity at its surface reaches c, the future cones tilt so far inward that their outer edges stand vertical; their envelope is the event horizon. Nothing inside can escape. A distant observer sees the star's image hover just above the horizon, increasingly red-shifted, fading rapidly to blackness.
Oppenheimer and Snyder, 1939. The first exact model assumed exact spherical symmetry and pressureless dust. The collapse continues to infinite density and curvature at the centre, a space-time singularity. Black-hole singularities look like the time-reverse of the Big Bang: every causal curve inside the horizon ends at the singularity, as every causal curve in a Friedmann model begins at the Big Bang. But the Second Law says they cannot be true mirror images: one must have extraordinarily low entropy, the other high.
- Pre-1964 belief: Lifshitz and Khalatnikov argued singularities would not form in general, asymmetric collapse.
- Earlier Penrose work: testing whether the steady state could fit general relativity, he found (unpublished) that light-ray topology forces negative energy densities, even with asymmetry.
- Autumn 1964: prompted by John Wheeler, after Maarten Schmidt's discovery of a brilliant, variable quasar, Penrose invents the trapped surface, a local mark of collapse past the point of no return.
- The theorem: given a trapped surface, non-negative light-ray focusing (energy flux across light rays never negative, the weak energy condition, with or without Λ) and evolution from a non-compact Cauchy surface Σ, a singularity is unavoidable, meaning space-time cannot be continued smoothly into the future.
- Resolution: the Russians, with Vladimir Belinski, found and fixed an error; singularities are generic. The BKL work suggests chaotic "mixmaster" behaviour near them, anticipated by Charles Misner.
What a trapped surface is. Picture a closed 2-surface, such as a sphere, at one instant. Fire flashes of light from it, inward and outward. Normally the inward flash shrinks in area and the outward one grows. A trapped surface is one where both shrink: gravity is so strong that even outgoing light converges. The idea's power is that it is local, checkable on one surface at one time, whereas an event horizon is defined only by the space-time's entire future out to infinity. Once a trapped surface exists, and energy is positive and the space-time evolves from an ordinary non-compact initial slice, the theorem says that some light ray must end after finite parameter: the space-time cannot be continued smoothly. That is the precise meaning of "singularity" here, and it is why the result needs no symmetry and no detailed model of the star.
Strength and weakness of the theorem. No symmetry, no special matter: only "physically reasonable" positive energy flux. But it says almost nothing about the singularity's form, whether density or curvature become infinite, or where trouble begins.
Do trapped surfaces really form? Chandrasekhar (1931) showed that white dwarfs (Sun's mass, Earth's radius; the first was Sirius's dark companion), held up by electron degeneracy pressure, have a limit of about 1.4 M☉ once relativity is included. A Sun-like star ends as a white dwarf then a cold black dwarf. Heavier stars blow up as supernovae (briefly outshining their galaxy), possibly leaving a neutron star (say 1.5 M☉ in about 10 km), held by neutron degeneracy, sometimes seen as a pulsar, with a limit Penrose gives as about 1.5 M☉ (the "Landau limit"). Above about 10 M☉, collapse very probably proceeds to a trapped surface. The case is far stronger at galactic centres: a million white dwarfs, none touching, fit inside 106 km and would already be trapped. "Unknown physics at high density" is beside the point.
| End state | Supported by | Mass limit (Penrose) | Size |
|---|---|---|---|
| White dwarf, then black dwarf | Electron degeneracy | ~1.4 M☉ (Chandrasekhar) | about Earth's radius |
| Neutron star (pulsar) | Neutron degeneracy | ~1.5 M☉ ("Landau limit"; now known to be ≥ 2 M☉) | about 10 km |
| Black hole | Nothing | none | Schwarzschild radius 2GM/c2, 3 km per solar mass |
The Schwarzschild radius grows in proportion to mass, so the mean density inside the horizon falls as 1/M2. A 1 M☉ hole has a horizon of 3 km and a mean density about 2 × 1019 kg/m3; a 108 M☉ hole has a horizon of about 3 × 108 km (twice the Earth–Sun distance) and a mean density near that of water. This is Penrose's scaling point: trapped surfaces for large systems need no exotic densities, so "unknown physics" offers no escape.
Cosmic censorship and the evidence. That a trapped surface yields a black hole relies on cosmic censorship, still unproved: generic collapse produces no naked singularities visible from outside. With BKL, it is probably the greatest open problem of classical GR. Observationally the case is strong: binary systems with invisible companions of a few solar masses, and stars whirling around an invisible ~4 000 000 M☉ object at the galactic centre; material falling in shows no surface being heated.
- Dark matter fraction: Penrose says dark matter is about 70% of matter; current estimates (Planck 2018) put it near 84 to 85% of the matter content (about 27% of the total energy budget against about 5% ordinary matter).
- Neutron-star maximum: the "about 1.5 M☉" limit is out of date. Pulsars of about 2.0 M☉ (PSR J1614−2230, 2010; PSR J0740+6620, about 2.08 M☉) show the maximum is at least 2 M☉, with theory suggesting 2.2 to 2.3 M☉.
- Inflation factor: given as 1030 to 1060 in §2.1 and 1020 to 1060 in §2.6, an internal inconsistency; standard estimates are at least about e60 ≈ 1026.
2.5 Conformal diagrams and conformal boundaries
Strict conformal diagrams. For spherically symmetric space-times, a 2D region 𝒟 stands for the whole: each interior point is a sphere S2, and a broken-line axis marks points that are single events. Radial light rays are drawn at 45°. The great advantage: infinite space-time squeezes into a finite picture, with infinity itself drawn as a boundary.
- Broken line: symmetry axis.
- Bold line: infinity, ℐ ("scri").
- Wiggly line: singularity.
- Slightly jagged line: incompleteness.
- Dotted line: black-hole horizon.
- Black dot: a single point; white dot: a whole sphere S2.
Minkowski space as a triangle. All of flat space-time becomes a right-angled triangle bounded by past null infinity ℐ− and future null infinity ℐ+ (every light ray starts on one and ends on the other), plus three points: i− and i+ (past and future timelike infinity, where every free massive particle begins and ends) and i0 (spatial infinity, through which every spacelike geodesic closes into a loop).
Extending beyond infinity. As the hyperbolic disc extends smoothly into the Euclidean plane, Minkowski space extends smoothly beyond its boundary: it is conformally a piece of the Einstein universe ℰ, the static S3 × time "Einstein cylinder" of 1917. ℰ can be seen as an infinite succession of Minkowski spaces, each one's ℐ+ glued to the next one's ℐ−. Penrose asks the reader to remember this picture for Part 3.
The role of Λ. Friedmann models with Λ = 0 (K = 0 or K < 0) have a null ℐ+ at 45°. With Λ > 0, every case has a spacelike ℐ+, flatter than 45°, and late-time behaviour approaches de Sitter space 𝒟, empty and maximally symmetric, a Minkowskian analogue of a 4-sphere sitting in Minkowski 5-space. The steady-state model is exactly half of de Sitter space, and so is past-incomplete: some free particle histories have no time before a finite value. A positive cosmological constant makes future infinity spacelike, and that is the geometric fact CCC is built on.
Black holes in diagram form. The Oppenheimer–Snyder diagram is glued from a collapsing Friedmann piece and the Eddington–Finkelstein extension of Schwarzschild's 1916 solution. The Schwarzschild radius 2MG/c2 is about 9 mm for the Earth and 3 km for the Sun (inside the bodies, so irrelevant there). It was thought singular until Lemaître (1927) showed a non-static smooth extension; Eddington (1930) found a simpler form without noting its significance; David Finkelstein (1958) rediscovered and explained it; the maximal Kruskal–Szekeres extension had an earlier equivalent by Synge.
- Hawking (1974): quantum fields near a black hole give it temperature T ∝ 1/M.
- 10 M☉ hole: about 6 × 10−9 K, comparable to the 2006 MIT laboratory record of about 10−9 K.
- ~4 000 000 M☉ galactic-centre hole: about 1.5 × 10−14 K.
- CMB today: about 2.7 K, far hotter, so black holes now absorb more than they emit.
- Far future: once exponential expansion cools the CMB below even the largest holes' temperature, they radiate, lose mass, heat up and vanish after up to 10100 years (a googol), ending in a "pop" with roughly the energy of an artillery shell.
| Space-time | Past boundary | Future boundary | What the diagram shows |
|---|---|---|---|
| Minkowski | null ℐ−, i− | null ℐ+, i+ | a right triangle (or diamond) |
| Einstein cylinder ℰ | none | none | an infinite strip; a stack of Minkowski diamonds glued ℐ+ to ℐ− |
| Friedmann, Λ = 0 | spacelike singularity | crunch (K > 0) or null ℐ+ | rectangle or triangle |
| Friedmann, Λ > 0 | spacelike singularity | spacelike ℐ+ | flattened top edge |
| de Sitter | spacelike ℐ− | spacelike ℐ+ | a square |
| Oppenheimer–Snyder | collapsing star | singularity inside, ℐ+ outside | dotted horizon, wiggly singularity |
| Evaporating hole | as above | singularity ends at the "pop" | horizon closes off |
Schematic conformal diagrams. Without exact symmetry, Penrose uses looser schematic diagrams built on two complementary tricks: a conformal factor Ω → 0 that squashes infinite regions to finite size, and Ω → ∞ that stretches singularities out to smooth boundaries. Both will be central to Part 3.
Going deeper: how to read a conformal diagram
Penrose's diagrams look cryptic until three rules are absorbed. First, light moves at 45°. Any line steeper than 45° is a possible history of a massive body; any flatter line joins events that cannot influence each other. Second, edges are either infinity or singularity. Bold edges are infinitely far away but drawn at finite distance; wiggly edges are where curvature blows up. Whether an edge is flatter or steeper than 45° tells you whether it is spacelike (an "instant", like the Big Bang or a Λ > 0 future) or null (like ℐ± of flat space-time). Third, distances and durations are distorted, but causal relations are exact. Two events are causally connected in the diagram if and only if they are in the real space-time.
With these rules, the diagrams in this book become arguments. The Minkowski triangle shows that every light ray in flat space-time begins on ℐ− and ends on ℐ+. The Λ > 0 Friedmann diagrams show a spacelike top edge: the future has a "shape" like a Big Bang, which is CCC's opening. The Oppenheimer–Snyder diagram shows the horizon as the 45° line that separates events that can still escape to ℐ+ from those that cannot. And a stack of Minkowski diamonds glued tip to toe inside the Einstein cylinder is, visually, the template for a chain of aeons.
Cosmological event horizon
- Exists when ℐ+ is spacelike (any Λ > 0 model).
- For an immortal observer O ending at o+ on ℐ+: the past light cone of o+.
- Events outside it are never observable to O; its position depends on O's whole future history (a change of mind changes it).
Particle horizon
- Exists when the past boundary (normally the Big-Bang singularity 𝒰−) is spacelike.
- Traced by idealised galaxy world-lines starting where the past light cone of o meets 𝒰−.
- Particles from beyond it have not yet come into view; more appear as O's time passes.
Horizons as bookkeeping devices. The two cosmological horizons answer different questions. The particle horizon asks what has had time to reach us since the beginning: it grows as we wait, and it is the reason the CMB's uniformity looks puzzling, because widely separated patches of the sky were never inside each other's particle horizons. The event horizon, which exists only when ℐ+ is spacelike, asks what can ever reach us: with Λ > 0, distant galaxies are now crossing beyond it and will never be seen again, however long we wait. Both are drawn naturally on conformal diagrams, and both will matter in Part 3, the first for the CMB correlations argument, the second for the "rogue electron" problem and the disputed horizon entropy.
2.6 Understanding the way the Big Bang was special
Inflation is misconceived as an explanation. Penrose's charge: inflation is supposed to smooth out early irregularities, but it runs on time-symmetric dynamics (an inflaton field and a phase transition, entropy-raising like freezing), so it cannot touch the Part 1 problem. The low entropy still has to be put in, and it lies in the unexcited gravitational degrees of freedom.
- Suppose the matter at decoupling had been highly lumpy; inflationists must hold that the inflaton makes this improbable.
- Time-reverse it: a lumpy collapsing universe. Irregularities grow, trapped surfaces form, and the end is "a horrendous mess of congealing black holes", likely a chaotic BKL singularity.
- Far from FLRW symmetry, reversed inflation never happens: inflation's calculations need an FLRW background.
- Reverse again: an expanding universe born from multiply bifurcating white holes, with vastly higher entropy, an initial state far more probable than ours, and one the inflaton cannot iron out.
- What singles out our Big Bang is the total absence of white-hole singularities.
Going deeper: the inflationist's reply and Penrose's rejoinder
Inflation's defenders answer that inflation was never meant to explain the Second Law; it explains flatness, the horizon problem (why opposite sides of the sky share a temperature) and the spectrum of fluctuations, given a small patch that starts inflating. Penrose's rejoinder is that this concedes his point: if inflation works only when it starts in a region already smooth enough, then the improbable specialness has merely been relocated into the initial patch. Some inflationists then appeal to eternal inflation, in which inflating regions are generic and our universe is one of endlessly many bubbles, so even very improbable starting patches occur somewhere. Penrose regards this as trading one unexplained specialness for an anthropic multiverse and points out that, by phase-space counting, a universe like ours but far smaller (just one galaxy's worth of order, with the rest thermal) would be overwhelmingly more probable than the vast smooth one we see. That scaling argument, made more fully in his Road to Reality, is the strongest form of his case.
Quantifying the specialness. The Bekenstein–Hawking entropy of a non-rotating hole of mass M is SBH = (8π2kG/ch) M2, or more generally SBH = kc3A / 4Għ with A the horizon area (between this and half of it for rotating holes), which is S = A/4 in Planck units. Penrose holds there is still no fully satisfactory count of black-hole microstates, but the formula is essential for a consistent Second Law outside the hole.
- All the mass within our particle horizon collapsed into one black hole: entropy about 10124, a rough lower limit on what a collapsing universe with this matter could reach.
- Phase-space volume for that: about 1010124.
- Actual universe at decoupling (entropy dominated by the CMB): volume no more than about 101089.
- Chance of our Big Bang by accident: about 1 in 1010124, "irrespective of inflation". It "needs some completely different kind of theoretical explanation".
With S = A/4 and A = 16πM2 (Planck units, non-rotating), S = 4πM2. One solar mass is about 9 × 1037 Planck masses, so a 1 M☉ hole has S ≈ 1077, against roughly 1058 for the Sun as an ordinary star. The galactic-centre hole at 4 × 106 M☉ reaches about 1090. Because entropy goes as M2, merging two equal holes roughly doubles the total entropy; that quadratic growth is why black holes dominate the cosmic entropy budget and why gathering all observable matter into one hole yields the colossal 10124.
Is a messy singularity still a "moment"? Could a white-hole-ridden start count as an instantaneous event? The time-reversed Oppenheimer–Snyder singularity is spacelike; BKL singularities seem spacelike; and strong cosmic censorship (unproved) says generic singularities are spacelike. So yes: the high-entropy alternative is a legitimate "Big Bang" too, which makes ours all the more special.
Measuring gravitational degrees of freedom. By analogy with Maxwell's field tensor F (6 components per point: 3 electric, 3 magnetic) and its source, the charge-current vector J (4 components), the Riemann curvature tensor R (20 components) splits into the Weyl conformal tensor C (10 components, the free gravitational field, analogue of F) and the Einstein tensor E (10 components, equivalent to the Ricci tensor, the source, analogue of J). Einstein's equations: E = (8πG/c4) T + Λg, or E = 8πT + Λg in Planck units, with T the energy tensor.
Ricci / Einstein curvature (E)
- Where matter is; the gravitational analogue of the source J.
- Acts on light like a magnifying lens, enlarging the view without distortion.
- In the Big Bang: present and large (hot dense matter).
Weyl curvature (C)
- The free gravitational field, even in vacuum; analogue of Maxwell's F.
- Acts as a purely astigmatic lens: turns small circles in the background into ellipses.
- Weyl curvature hypothesis: C = 0 at initial singularities; in black-hole and BKL final singularities, C diverges.
Why astigmatism is the right picture. Imagine a small, perfectly circular bundle of light rays travelling through empty space past a star. With no matter inside the bundle, there is no Ricci curvature along it, so nothing magnifies it uniformly. But the star's field outside the bundle is Weyl curvature, which squeezes the bundle in one direction and stretches it in the perpendicular one, without changing its area to first order: a purely astigmatic effect. Fill the bundle with matter instead and the rays focus uniformly, the magnifying effect of Ricci curvature. Every gravitational lens is a mixture of the two, and the ellipticity of background galaxies measures the Weyl part. A Big Bang with C = 0 is one in which no such shearing of light bundles yet existed anywhere: gravity was present only as the uniform focusing produced by uniformly distributed matter.
Lensing in practice. Einstein's predicted gravitational lensing now maps mass distributions, including dark matter: background galaxies are already elliptical, but statistics over many reveal the imposed pattern of ellipticity, and in clusters like Abell 1689 and Abell 2218 the arcs are visible by eye. Because conformal structure is null-cone structure, C measures the departure of the cones from Minkowski's, and it shows up as ellipticity in light-ray bundles.
Tod's version of the WCH. Statements about a tensor "at" a singularity are mathematically slippery. Paul Tod's formulation: the Big Bang is a smooth spacelike 3-surface 𝒰− forming a past boundary to space-time as a conformal manifold, just as in FLRW diagrams but without assuming symmetry. That at least keeps C finite there. Stated cleanly: space-time can be continued smoothly, as a conformal manifold, a little way before 𝒰−. Before the Big Bang? "Never fear, this is just a mathematical trick." Penrose ends Part 2: "Or might it . . .?"
Tod's version versus CCC's. The two forms of the hypothesis differ in strength, and the difference matters. Tod's says only that the Big Bang can be treated as a smooth conformal boundary, which keeps Weyl curvature finite there and is a natural, mathematically clean condition. It leaves open a finite amount of gravitational "texture" at the start. Penrose's original hypothesis, and the version CCC will derive, says Weyl curvature vanishes. The stronger statement is closer to what the Second Law seems to need, since any finite Weyl curvature represents some activated gravitational degrees of freedom. Part 3 shows that if the Big Bang is the rescaled future of a previous aeon, the stronger form follows automatically from how gravitational radiation behaves at that future boundary.
- Penrose's Nobel (2020): half the physics prize, for showing that black-hole formation is a robust prediction of general relativity (the 1965 trapped-surface theorem in §2.4); the other half went to Reinhard Genzel and Andrea Ghez for the galactic-centre object, now pinned at about 4.3 million M☉.
- Black holes seen directly: LIGO detected merging black holes in 2015 (GW150914); the Event Horizon Telescope imaged M87* (2019) and Sagittarius A* (2022). Gravitational-wave data have since tested Hawking's area theorem, with the clean 2025 event GW250114 giving strong confirmation.
- CMB precision: the Planck satellite (final results 2018) gives an age of 13.787 ± 0.020 billion years and confirms a flat geometry to better than 1%, but the Hubble tension between early-universe and local expansion rates has grown past 5σ.
- Dark energy: DESI's 2024 to 2025 data hint that dark energy weakens with time (roughly 2.8 to 4.2σ, depending on the supernova sample). If confirmed, a pure constant Λ, which both Penrose's spacelike ℐ+ and CCC assume, would need revisiting.
- Hawking radiation remains unobserved for real black holes; analogue experiments in fluids and optics reproduce its kinematics.
Penrose's move here is a model for diagnosing a mis-specified metric: the CMB "said" maximum entropy only because the measure left out a whole class of degrees of freedom (gravity). In risk and AI evaluation the same error recurs, where a KPI looks saturated or healthy because the dimension carrying the real exposure (concentration, correlation, tail dependence) is not in the state space you are counting.
Conformal Cyclic Cosmology
Part 3 states the proposal. Near the Big Bang and in the remote future, the universe is dominated by massless stuff that sees only conformal geometry, so the infinitely expanded future of one universe can be glued smoothly onto the Big Bang of the next. History becomes an endless chain of aeons; black holes destroy information and reset phase space between them, and the previous aeon may be legible as circles in the present microwave sky. Penrose works through the price (mass must fade, dark matter is reborn, cosmological entropy is dismissed) and the first data.
- Paul Tod (Oxford): conformal-boundary Weyl curvature hypothesis; the template CCC strengthens.
- Helmut Friedrich: proofs that Λ > 0 universes get a smooth, stable future boundary.
- R. C. Tolman: entropy-growing cyclic models and radiation-filled universes whose sign behaviour matches CCC.
- John Wheeler: constants that might change from cycle to cycle; quantum foam.
- Lee Smolin, Gabriele Veneziano, Paul Steinhardt and Neil Turok, Abhay Ashtekar and Martin Bojowald: rival pre-Big-Bang schemes.
- Stephen Hawking, Kip Thorne, John Preskill: the black-hole information bet, conceded by Hawking in 2004.
- Paul Dirac, Robert Dicke, Brandon Carter: large numbers, stellar lifetimes and the anthropic principle.
- David Spergel and Amir Hajian (Princeton): the first CMB search for CCC circles, 2008.
3.1 Connecting with infinity
The early universe is effectively massless. Very soon after the Big Bang, kinetic energies dwarfed rest energies, so particles behaved as massless. In standard particle physics, rest-mass comes from the Higgs field through symmetry breaking; above the Higgs energy, everything is massless "like a photon". Massless fields care only about the null-cone (conformal) structure.
- Maxwell's equations (photons): fully invariant under g ↦ ĝ = Ω2g, with F ↦ F and J ↦ Ω−4J, including the coupling to charges; consistent with QFT at a basic level, photon going to photon.
- Yang–Mills equations: Maxwell with internal indices; governs the strong force (quarks and gluons, gluons massless) and the electroweak force (photon plus W+, W−, Z, which are massive via the Higgs).
- When the Higgs is switched off (temperatures near the Big Bang, roughly LHC energies at full power), full conformal invariance should return.
- Timing: down to about 10−12 s after the Big Bang, temperatures above about 1016 K, physics is blind to the scale factor Ω.
Backwards through the Big Bang. Given Tod's smooth conformal boundary 𝒰−, conformally invariant physics can be propagated backwards across it into a hypothetical pre-Big-Bang region "in a mathematically coherent way", undisturbed by the huge scale changes.
Not a bounce. The obvious candidate for the earlier region, a collapsing universe that bounces, defeats the purpose: a collapse obeying the Second Law ends in a black-hole-riddled, BKL-chaotic crunch that can never match a conformally smooth big bang; aiming a collapse precisely at such a special state would be "an immense violation of the Second Law". Letting the Second Law run backwards in that phase "goes very much against the grain" of seeking an explanation. Tolman's radiation models raise mathematical objections too (§3.3).
Look to the other end of time. With Λ > 0 the universe settles into exponential expansion toward a smooth spacelike ℐ+. Black holes, including galactic giants, all eventually pop, the largest after a googol of years or more. What remains is mostly photons (red-shifted starlight, CMB, Hawking radiation) and gravitons from close black-hole encounters, plus surviving dark matter, none of which can build a clock.
Penrose describes depressing himself with the image of eternal tedium: galaxies, planets, life and art all gone, nothing left but waiting 10100 years for the last black hole's artillery-shell "pop", then thinning and cooling for ever. Then, in summer 2005, he asked who would be around to be bored. Only photons and gravitons, and "it is pretty hard to bore a photon". For a massless particle "eternity is no big deal": it reaches ℐ+ before its first internal tick. With no mass, there is no clock and (since distance depends on time) no ruler.
- ~1010 years: now; stars still forming, Λ beginning to dominate the expansion.
- ~1012 to 1014 years: star formation ends; the last red dwarfs burn out.
- ~1015 to 1020 years: close stellar encounters scatter stars out of galaxies or into central black holes; planets detach.
- Beyond ~1034 years: proton decay, if it occurs (current experimental limits already exceed about 1034 years).
- ~1067 years: stellar-mass black holes evaporate once the CMB has cooled below their Hawking temperature.
- ~10100 years and more: the largest supermassive holes pop.
- Then: photons, gravitons, sparse leftovers, exponential expansion: the very boring era, and, in CCC, the approach to the crossover.
Friedrich's result in plain terms. Helmut Friedrich proved, for Einstein's equations with Λ > 0 and suitable matter (radiation-like fields), that space-times which start close to de Sitter remain close to it and acquire a smooth conformal boundary at future infinity, and that this behaviour is stable: small disturbances do not spoil it. For CCC, this means the smooth ℐ+ it needs is not a property of idealised symmetric models alone but the generic fate of an expanding, Λ-dominated universe filled with massless fields. It is the mathematical counterpart, at the future end, of what Tod's hypothesis asserts at the past end.
Two smooth edges. To massless particles, ℐ+ is just another place in conformal space-time, and Helmut Friedrich's results support smooth conformal extension beyond ℐ+ when Λ > 0. Tod's proposal gives a smooth extension before 𝒰−. Both edges are spacelike, both dominated by conformally governed massless content.
Rejected: ℐ+ = our own 𝒰−
- The universe "loops round": beyond our future infinity lies our own Big Bang.
- Economical, but creates closed timelike curves; since §3.6 shows information can cross the join, this risks paradoxes or nasty constraints.
Adopted: a chain of aeons
- Before our 𝒰− lies the remote future of a previous universe; beyond our ℐ+, the big bang of a next.
- Each phase from 𝒰− to ℐ+ is an aeon; the ℐ+ of each is identified with the 𝒰− of the next, with a perfectly smooth conformal join.
The worry is obvious: zero temperature and density at ℐ+ are being identified with infinite temperature and density at a big bang. But conformal stretching at the big bang brings the infinities down to finite values, and conformal squashing at infinity brings the zeros up to finite values, and the physics on both sides is blind to the rescaling. Better still, phase-space volume is conformally invariant: shrink distances and momenta grow by exactly compensating factors. That invariance is what makes the Second Law the central problem for CCC.
For one particle, a phase-space cell has volume Δx · Δp per dimension. Stretch all lengths by a factor λ: positions grow to λΔx, and for massless quanta momentum scales inversely with wavelength, so Δp shrinks to Δp/λ. The product Δx · Δp is unchanged, in every dimension, for every particle. A whole universe of radiation therefore has the same phase-space volume at any overall scale. That is why CCC cannot "dilute away" the entropy built up in an aeon: the bookkeeping is scale-proof, and something else (black-hole information loss) has to shrink the space.
Imagine one long-lived observer in the remote future of our aeon, made, impossibly, of the last surviving electrons. She builds a clock from them and measures the expansion. Now let her electrons' mass fade toward zero, as CCC requires. Her clock slows in step; distances measured by light-signal timing lose their unit; the expansion she infers from her own instruments tends to a standstill. Not because anything stopped, but because she has lost the means to say how big anything is. CCC claims that in this condition, "infinitely large and cold" and "infinitesimally small and hot" are physically the same description.
3.2 The structure of CCC
The rogue electron problem. Not all mass ends in black holes. Bodies are ejected from galaxies and clusters; white dwarfs cool to black dwarfs; dark matter may never fall in. Proton decay, if it happens, is very slow, but charge conservation means a decaying proton must leave at least a positron, and electrons must survive too. Electrons and positrons are the lightest charged particles, and there are no massless charged particles (they would show up copiously in present-day particle processes). Pairs could annihilate into photons, but many particles will end up alone inside their own cosmological event horizons, unable ever to meet a partner. Calling one trapped electron useless as a clock lacks "the kind of rigour that physical laws ought to demand".
- Ignore isolated particles. Rejected as unrigorous.
- Drop charge conservation: electrons slowly decay into neutral particles (a massless neutrino would then be needed). No evidence, "extremely unpleasant" theoretically, and it would seem to require a massive photon, which ruins the philosophy.
- Rest-mass slowly fades to zero for every particle (electrons, positrons, neutrinos, protons if they survive, dark matter). No evidence either, but far weaker theoretical objections. Penrose's choice.
Why mass is a weaker law than charge. Charge is additive and comes in integer multiples of a basic unit (the anti-down-quark's, one third of the proton's). Mass is not additive (E = mc2 adds kinetic energy to a composite's mass), the values of particle masses are completely unexplained, and mass already was not constant in the early universe. Technically, rest-mass and spin are Casimir operators of the Poincaré group (symmetries of Minkowski space), hence "good quantum numbers". With Λ > 0, the relevant group is the de Sitter group, and there rest-mass is not exactly a Casimir operator (an extra term involves Λ), so a very slow decay is "not out of the question".
Going deeper: what "mass fading" would have to look like
Penrose's mass decay is not a decay of particles into other particles but a slow drift of the rest-mass parameter itself, for every species, over times far longer than anything observed. He offers three supports. Rest-mass is not conserved additively (a hot gas weighs more than a cold one), so it is not protected the way charge is. The masses of elementary particles are unexplained inputs, so no deeper principle fixes them. And in a universe with Λ > 0, the symmetry group of empty space-time is de Sitter's, not Poincaré's, and in de Sitter's group rest-mass is not an exact conserved label. If masses fade in exact proportion, nothing local would notice except a very slow apparent weakening of gravity relative to the mass scale. If they fade at different rates, ratios like the proton-to-electron mass ratio would drift, which atomic-clock comparisons already bound to below about 10−16 per year in the present era. CCC can live with this because the fading may be negligible for 10100 years and become decisive only near ℐ+, but that same flexibility is why critics call the requirement unfalsifiable as stated.
Which clock defines the metric? If masses decay, an "electron metric" would put ℐ+ at a finite time and make the expansion halt or reverse, inconsistent with Einstein's equations, and electron, neutrino and proton metrics would disagree unless all masses fade in proportion. Penrose's fix: use Λ itself, or effectively G, to fix the scale, preserving Einstein's equations with constant Λ throughout each aeon. The CCC philosophy becomes: gravity-free (and Λ-free) physics loses track of time, not physics as a whole.
Gravity's subtlety. The free gravitational field (Weyl tensor) is conformal, but its coupling to sources is not, unlike Maxwell's.
- Write Maxwell symbolically as ∇F = 4πJ; after rescaling, ∇̂F̂ = 4πĴ, the same equation. Free fields: ∇F = 0, also the Schrödinger equation for a free photon.
- Gravitational waves obey a conformally invariant analogue ∇K = 0 (the free graviton's Schrödinger equation), where K = C in Einstein's physical metric.
- Under g ↦ Ω2g: Ĉ = Ω2C (to keep C the conformal curvature) but K̂ = ΩK (to keep the wave equation invariant), so K̂ = Ω−1Ĉ, i.e. Ĉ = ΩK̂.
- Approaching ℐ+, Ω → 0 smoothly with non-zero normal derivative, while K̂ reaches finite values that encode the outgoing gravitational radiation.
- Therefore Ĉ = 0 on ℐ+. Smooth conformal matching then forces C = 0 at the next aeon's 𝒰−.
- CCC thus delivers the strong form of the Weyl curvature hypothesis, Weyl curvature vanishing at each big bang, not merely finite as in Tod's version.
Swapping Ω for its reciprocal. Beyond the crossover, the next aeon's conformal factor must blow up at 𝒰− with Ω−1 smooth: Ω must "suddenly become its reciprocal". Penrose encodes the scale in the 1-form Π = dΩ/(Ω2 − 1), which is smooth across and unchanged by Ω ↦ Ω−1. With only massless fields before the crossover, the remaining freedom is a field ϖ obeying a non-linear, conformally invariant "ϖ-equation" (a Calabi equation: (□ + R/6)ϖ = (2/3)Λϖ3). The solution that returns Einstein's metric is the phantom field, equal to 1 in that metric and carrying no independent degrees of freedom.
Dark matter reborn. Continuing straight through would make the next aeon's effective gravitational constant negative. Using Ω−1 instead turns the phantom field into a real physical scalar field (initially infinite) just after the new big bang. The mathematics forces this extra dominant component, which inherits the degrees of freedom of the previous aeon's gravitational waves. Penrose proposes it is the initial form of dark matter, acquiring mass shortly after the Big Bang (presumably with the Higgs). The equations also do not allow all sources to stay strictly massless after the crossover; the natural constraint is to delay the appearance of rest-mass as long as possible.
- Dark energy (Λ > 0): without it ℐ+ would be null, not spacelike, and could not match the spacelike 𝒰−.
- Dark matter: the reborn ϖ-field supplies the dominant initial matter of each aeon.
- Penrose finds it "perhaps significant" that both appeared in observations in recent decades and both are necessary for CCC.
The value of Λ. Interpreting Λg as vacuum energy gives a value about 10120 too large. The "why now" coincidence (Λ's effect just matching matter's attraction today) puzzles many; Penrose less so. He suggests Λ ≈ c3/(N6Għ), i.e. Λ ≈ N−6 in Planck units, with N ≈ 1020.
- N ≈ 1020. Dirac (1937): integer powers of N recur in dimensionless ratios involving gravity. Electric-to-gravitational force between electron and proton in hydrogen: about 1040 ≈ N2. Age of the universe: about N3 Planck times.
- Planck time tP = √(ħG/c5) ≈ 5.4 × 10−44 s; Planck length lP = ctP ≈ 1.6 × 10−35 m.
- Planck mass ≈ 2.1 × 10−5 g; Planck energy ≈ 2.0 × 109 J.
- Planck temperature (k = 1): given as 2.5 × 1032 K.
- Dirac thought G must fall as the universe ages (as 1/age2); measurements rule out that rate. Robert Dicke (1961), refined by Brandon Carter: observers living mid-way through a main-sequence star's life will find an age near N3. If Λ ≈ N−6 can be derived, the coincidence dissolves.
| What crosses into the next aeon | What does not |
|---|---|
| Conformal geometry of the crossover 3-surface (overall spatial geometry, open, flat or closed) | The metric scale: sizes, durations, temperatures as absolute numbers |
| Electromagnetic radiation (Maxwell fields, conformally invariant) | Rest-mass, which must have faded to zero |
| Gravitational-wave information, via K and the normal derivative of C | Weyl curvature itself (zero at the join) |
| Massless Yang–Mills fields | Information swallowed by black holes, destroyed at their singularities |
| Possibly altered constants such as N (Wheeler's idea) | Entropy accumulated in black holes, devalued by the reset |
Going deeper: the cosmological-constant coincidence
Two puzzles about Λ are usually run together. The first is its size: naive vacuum-energy estimates overshoot by about 10120, the worst prediction in physics. The second is its timing: Λ's repulsion is only now comparable with the attraction of matter, which was enormously larger in the past and will be negligible in the future. Penrose treats the second as no deeper than older puzzles. If Λ really equals about N−6 in Planck units, and if observers like us inevitably live when the universe's age is about N3 Planck times (the Dicke–Carter argument from stellar lifetimes), then the matter density at our epoch scales as roughly 1/age2 ≈ N−6, the same order as Λ. The "coincidence" would then be a consequence of where in time observers can exist, provided someone can derive Λ ≈ N−6. Penrose admits these are "clearly speculative matters".
3.3 Earlier pre-Big-Bang proposals
Friedmann's oscillating universe. The closed (K > 0, Λ = 0) Friedmann model's radius traces a cycloid, the curve of a point on a rolling hoop; continued, it gives an endless series of expansions and crunches, which briefly interested Einstein in 1930. Each bounce is a singularity, and the model has no room for entropy to grow.
Tolman's improvements (1934). Richard Chace Tolman gave the matter an internal degree of freedom so entropy could rise: successive cycles grow longer and larger. Still FLRW, so no gravitational clumping and only a mild entropy increase, but "one of the surprisingly few serious attempts" to put the Second Law into cosmology.
Tolman's radiation models and the sign flip. Near the Big Bang, matter is better modelled as hot incoherent radiation than dust. Tolman built radiation analogues of all six Friedmann models. For K > 0 the radius is a semicircle, whose analytic continuation is a full circle, so the radius would turn imaginary: no straightforward bounce. The behaviour of the conformal factor at the singularity also differs:
Friedmann dust
- Near 𝒰, ω = Ω−1 behaves like the square of conformal time.
- Smooth continuation keeps ω positive: no sign change.
- Not what CCC needs.
Tolman radiation
- ω varies linearly with conformal time.
- Smoothness forces ω (and Ω) to change sign across 𝒰.
- Matches CCC: after the Ω ↦ −Ω−1 switch, the new aeon starts as a Tolman radiation universe, the realistic early state (ignoring inflation).
Varying constants and anthropic reasoning. John Wheeler suggested that dimensionless constants might change at each bounce, since physics has been abandoned there anyway. Fine-tuning examples: ice floating on water (dismissible as parochial), the neutron slightly heavier than the proton (without it, no variety of stable nuclei and no chemistry), and Fred Hoyle's predicted carbon energy level, confirmed by William Fowler (Fowler shared the 1982 Nobel with Chandrasekhar; "strangely, Hoyle was passed over"). Brandon Carter coined the anthropic principle. Penrose is unsure of his own position but thinks it is over-used to prop up implausible theories. In CCC, N might change between aeons.
- Lee Smolin, Life of the Cosmos (1997): black-hole interiors bounce into baby universes, constants vary slightly, giving cosmic natural selection for black-hole production; makes testable statistical predictions about black holes and neutron stars. Its collapse-to-expansion is unlike CCC's smooth join, and its link to the Second Law is "obscure".
- Gabriele Veneziano (string pre-big-bang), about seven years before CCC: shares conformal rescalings and places the exponential "inflationary" phase before the bang; but rests on string-theory culture, hard to compare.
- Paul Steinhardt and Neil Turok (ekpyrotic/cyclic): aeons separated by colliding D-branes, with crossover after only some multiple of 1012 years, when astrophysical black holes still exist.
- Quantum bounces: most replace the singularity in simplified models and still leave singularities; the best is loop quantum cosmology (Ashtekar and Bojowald).
- Common failing: none addresses the suppression of gravitational degrees of freedom at the Big Bang; most stay inside FLRW. Einstein hoped irregularities would avoid singularities; the singularity theorems of the late 1960s ended that hope, and a clumpy crunch cannot match a smooth bang.
How CCC compares with the main alternatives
| Scheme | Before our Big Bang | Explains low gravitational entropy? | Status in 2026 |
|---|---|---|---|
| Standard inflation | Nothing, or an eternally inflating parent | No: assumes a smooth enough patch to start | Mainstream; no primordial gravitational waves yet (r < 0.036) |
| Eternal inflation / multiverse | Endless bubble nucleation | Appeals to anthropics | Popular, hard to test |
| Friedmann / Tolman oscillation | Previous crunch | No: entropy would accumulate | Abandoned |
| Loop quantum bounce (Ashtekar, Bojowald) | Collapsing quantum phase | Not without fine-tuning the collapse | Active research |
| Ekpyrotic / cyclic (Steinhardt, Turok) | Brane collision | Partly, via slow contraction smoothing | Active; tensions with string theory |
| Smolin's cosmic natural selection | Parent black hole | No clear link | Minority view |
| Veneziano pre-big-bang | String-theory phase | Not addressed | Largely dormant |
| CCC (Penrose) | Remote future of previous aeon | Yes, by construction: C = 0 forced at the join | Minority; CMB claims disputed |
3.4 Squaring the Second Law
The conundrum. The very early universe (conformally stretched) and the very remote future (conformally squashed) look alarmingly similar, differing by an enormous change of scale, and scale changes do not alter phase-space volume. If anything the early state has more freedom (dark matter; Tod even allows finite Weyl curvature). Yet entropy rises enormously between them through gravitational clumping. How can a cycle repeat if each aeon ends with vastly more entropy than it began?
- CMB: about 109 (1 000 000 000).
- Today, mostly black holes: about 1021 (1 000 000 000 000 000 000 000), estimated from the Milky Way's 4 × 106 M☉ hole via Bekenstein–Hawking.
- These will grow hugely before they evaporate.
Black holes are Second-Law citizens. Their entropy and temperature follow from Bekenstein's 1972 thought experiments (lowering objects into holes) using only the Second Law, quantum mechanics and relativity, up to the exact coefficient; Hawking's temperature TBH = κ/M (κ = 1/4π in Planck units) then follows thermodynamically. A stationary black hole needs very few parameters, so a vast phase-space volume corresponds to each, hence huge entropy. Evaporation takes about 2 × 1067(M/M☉)3 years; quantum gravity matters only at the Planck scale, leaving about a Planck mass that should last about a Planck time. Stable remnants cause QFT difficulties; CCC needs nothing massive to persist, so the "pop" is acceptable.
Where does the information go? Inside the hole everything meets a singularity where "physics comes to an end", or continues into something utterly foreign. Penrose finds it very hard to believe that a million-solar-mass hole's information floods out at the tiny pop, and doubts the alternative that it leaks out in entanglements in the Hawking radiation, which would seem to require copying quantum information, violating basic principles.
- 1974: Hawking's original derivation splits an incoming wave between what escapes and what falls in; assuming the infalling part is lost is what makes the outgoing radiation thermal.
- For years: Hawking champions information loss.
- 2004, GR17, Dublin: Hawking reverses, concedes the bet he and Kip Thorne made with John Preskill, and argues information is retrieved. Penrose: Hawking "should have stuck to his guns".
- The orthodoxy's motive: faith in unitary evolution (U), the reversible Schrödinger dynamics that cannot destroy information.
Why Penrose distrusts the escape routes. The two popular ways to save unitarity both trouble him. If information emerges only at the final pop, then the enormous information content of a hole of millions of solar masses must flood out of a Planck-sized region in a Planck time, which he finds implausible. If instead it leaks out gradually, encoded in correlations in the Hawking radiation, then the same information seems to exist twice, once falling toward the singularity and once in the radiation outside, which collides with the quantum no-cloning principle. Hawking's own 1974 derivation, read with its conformal diagram, makes the sharing explicit: part of each incoming wave escapes, part falls in and is lost, and that loss is exactly what makes the radiation thermal. Penrose therefore regards information loss not as a paradox to be dissolved but as evidence that the true theory of quantum gravity is not unitary.
U and R. Quantum mechanics alternates two procedures: U, continuous, deterministic, time-symmetric Schrödinger evolution of the state ψ; and R, the reduction or collapse on measurement, where ψ jumps at random to one of the alternatives ψj with probability Pj. Schrödinger reportedly said that if "all this damned quantum jumping were really here to stay" he would be sorry he got involved. Penrose sides with Schrödinger, Einstein and Dirac that quantum mechanics is provisional: R is a real departure from unitarity that occurs when gravity becomes involved, and black-hole information loss is part of the case.
The resolution: phase space shrinks. Information loss means loss of degrees of freedom, so the phase space itself becomes smaller, a completely new kind of dynamics in which the arena shrinks as the state moves.
Entropy counting outside the holes goes on normally; the Second Law is never violated locally. But once holes evaporate, the degrees of freedom they swallowed are gone, so phase-space volumes "count for far less", as in a country experiencing currency devaluation. Because of the logarithm, this amounts to subtracting a huge constant from the universe's entropy, exactly like dropping the galaxy's external factor W in §1.3. Black holes are the entropy sinks that let each aeon begin afresh with gravity unexcited. A rough measure of the reset: the Bekenstein–Hawking entropy the holes would have reached without evaporating.
An accountant keeps the books of an aeon in "phase-space currency". Every process outside black holes is audited and found honest: entropy never falls. But each black hole is a sealed vault: deposits go in, and when the vault finally evaporates, the radiation that comes out is a thin thermal trickle carrying none of the account details. At the end of the aeon the accountant must restate the currency: all those vaults' contents have ceased to exist as possibilities, so the unit of account shrinks. Nothing was stolen and no rule was broken, yet the new aeon's opening balance is tiny. That restatement is Penrose's resolution, and its legitimacy depends entirely on whether the vaults truly destroy their contents.
Going deeper: the entropy ledger of one aeon
Rough numbers make the scale of the reset concrete. At last scattering, the entropy of our observable region was dominated by the CMB: about 1089 (in units of k), roughly 109 photons for each of about 1080 baryons. Today, supermassive black holes dominate, with estimates for the observable universe of order 10104 (Egan and Lineweaver's 2010 census of supermassive black holes gives about 3 × 10104). If all the matter ended in one hole, the ceiling would be about 10124. CCC's claim is that by the crossover, black-hole evaporation and information loss have subtracted almost all of this, leaving a starting state for the next aeon comparable to our own CMB-dominated start, with gravitational degrees of freedom again unexcited. The ledger never shows a decrease within an aeon; it is the definition of the zero point that is reset between aeons.
3.5 CCC and quantum gravity
The reset, restated. After all holes evaporate, the entropy zero must be re-set; the allowed states at the next big bang are greatly restricted, enforcing a Weyl curvature hypothesis and providing the potential for clumping in the new aeon.
Is there a cosmological entropy? Many cosmologists assign Bekenstein–Hawking entropy to the cosmological event horizon of a Λ > 0 universe: SΛ = AΛ/4 with limiting area AΛ = 12π/Λ, so SΛ = 3π/Λ, and a temperature TΛ = (1/2π)√(Λ/3). With the observed Λ: TΛ ≈ 10−30 K and SΛ ≈ 3 × 10122, far above the ~10115 reachable by forming and evaporating black holes in the observable universe. Penrose argues against treating it as real.
- Which region? Not the whole universe (if infinite, today's entropy could already exceed SΛ). Within one horizon, the material inside an ultimate particle horizon is about (3/2)3 ≈ 3.4 times today's; one hole from it would have entropy about (3/2)6 ≈ 11.4 × 10124 ≈ 10125 ≫ 3 × 10122, a gross violation, though such a hole would stay colder than TΛ and never evaporate.
- Many smaller holes: about 1081 baryons' worth (3.4 × 1080, times about 3 for dark matter) split into 100 holes of 1079 protons, each reaching about 10119 and evaporating: total about 10121, which he compares with "3 × 10120".
- No discernible degrees of freedom: a constant Λ adds nothing to phase space; matching ℐ+ to 𝒰− leaves no room for them.
- A dark-energy field would be strange: positive energy yet repulsive, and it would grossly violate the weak energy condition if allowed to vary.
- No information loss: unlike a black hole there is no singularity; global time slices show nothing is lost, only hidden from one observer.
- Observer-dependence: like the Unruh temperature felt by Rindler (uniformly accelerating) observers in flat space, TΛ is felt by accelerating observers, not free-fallers; as Λ → 0 the horizon becomes a Rindler horizon with infinite entropy, absurd for Minkowski space.
- The derivation rests on elegant analytic continuation valid only for exactly symmetric space-times like de Sitter.
In item 2, Penrose compares his 10121 total against 3 × 10120 and says it "seems still to violate the Second Law, but not by so very much". The bound he set up is SΛ ≈ 3 × 10122, and 10121 is below that, so on his own numbers the many-hole case does not violate the bound. The single-giant-hole case (10125) still does. He calls the figures "perhaps too rough for a definitive conclusion" anyway.
The Unruh analogy in more detail. In flat Minkowski space, an observer with constant acceleration a follows a hyperbola and has a Rindler horizon: a null plane behind which events can never send signals to her. Quantum field theory says she perceives the vacuum as a thermal bath at the Unruh temperature, proportional to a (about 4 × 10−21 K per m/s2), while an inertial observer in the same region sees nothing. Close to a large black hole's horizon, observers hovering at fixed height are in exactly this situation: they feel the local Hawking temperature, while free-fallers do not. Penrose's point is that de Sitter's TΛ belongs to the same family, felt by accelerated observers, not by those moving freely with the cosmic expansion. A temperature that depends on the observer's motion, he argues, cannot be paired with an objective entropy counted in the universe's phase space.
Vacuum energy. QFT's vacuum seethes with virtual particles; relativistic symmetry says its energy tensor is Tv = λg, just like a cosmological constant, with λ = Λ/8π. Calculations give λ = ∞ ("the most honest... But also the most wrong"), λ = 0 (favoured while Λ = 0 fitted the data), or λ ≈ tP−2, while observation gives Λ ≈ 10−120 tP−2. "Something is clearly seriously wrong." Either way, SΛ and TΛ do not affect CCC, and Penrose ignores them.
Planck-scale menagerie. Theorists expect classical space-time to fail when curvature radii approach lP (about 10−20 of a proton's classical radius). Proposals for what replaces it: Wheeler's quantum foam of wormholes; loops, spin foams, lattices, causal sets, polyhedra; non-commutative geometry; strings and membranes in extra dimensions; or space-time emerging from something more primitive (Machian theories, twistor theory). "There is no agreement whatsoever."
CCC's answer is conservative. Quantum gravity takes over when Weyl curvature radii reach the Planck scale, not Ricci/Einstein curvature. At the Big Bang, C = 0 (infinite Weyl radius) while Einstein curvature diverges, so the bang is essentially classical, and the next aeon is fully determined by the previous aeon's remote future through deterministic equations. Black-hole singularities, with wild Weyl curvature, are where exotic, information-destroying, time-asymmetric quantum gravity lives. This explains why the two ends of time look so different, a puzzle Penrose had long found baffling.
Standard view
- Quantum gravity governs the Big Bang.
- CMB fluctuations start as quantum fluctuations of the inflaton.
- Both kinds of singularity are treated by the same unknown theory.
CCC
- The Big Bang is classical and smooth; only black-hole singularities need quantum gravity.
- Fluctuations are imprints from the previous aeon (§3.6).
- The two ends of time are different in kind, matching the Second Law.
| Term | Meaning in Part 3 |
|---|---|
| Aeon | One universe history from big bang 𝒰− to future infinity ℐ+. |
| Crossover 3-surface | ℐ+ of one aeon identified with 𝒰− of the next. |
| Conformal factor Ω | The scale function in ĝ = Ω2g; goes to zero at ℐ+, switches to its reciprocal across the join. |
| Phantom field ϖ | Scaling freedom before the join; becomes real proto-dark matter after it. |
| K-field | Rescaled gravitational radiation field; finite at ℐ+, forces C = 0 there. |
| Mass fading | CCC's requirement that all rest-masses decay to zero over eternity. |
| Information loss | Destruction of degrees of freedom at black-hole singularities; the entropy reset. |
| SΛ, TΛ | Proposed cosmological-horizon entropy 3π/Λ and temperature; rejected by Penrose. |
Objections to CCC
- "Infinite cold cannot be infinite heat."
- "Massive particles survive for ever."
- "Entropy would pile up aeon after aeon."
- "Quantum gravity must govern the Big Bang."
- "Horizon entropy 3π/Λ dwarfs black holes."
Penrose's replies in the book
- Conformal rescaling makes both finite and identical; the physics is scale-blind.
- Rest-mass slowly fades; charge cannot, so no charged massless particles are needed.
- Black-hole information loss shrinks phase space; the zero of entropy is reset.
- Only Weyl curvature at the Planck scale demands it; C = 0 at the bang.
- No discernible degrees of freedom, observer-dependent, no information loss; ignore it.
3.6 Observational implications
Seeing through the Big Bang. The extreme organisation of the Big Bang, implied by the Second Law, is what allows conformal extension and deterministic evolution from the previous aeon, so we might see through to it. The overall spatial geometry (open, flat, closed) must match across aeons, but that gives nothing new. The real hope is in subtle CMB irregularities left by the previous aeon's last activity, assuming it resembled ours.
- Super-horizon correlations. Without inflation, the last-scattering surface 𝒲 sits so close to 𝒰− that only points within about 2° could ever have been in causal contact, yet CMB correlations extend to about 60°. Inflation stretches the conformal gap between 𝒰− and 𝒲; CCC puts the correlations' origin in the previous aeon.
- Scale invariance. Harrison and Zel'dovich (1970) proposed scale-invariant initial fluctuations before inflation existed; inflation explains them by self-similar exponential expansion. CCC's prior aeon ends in a self-similar exponential expansion too. - In effect CCC (like Veneziano, 1998) moves the "inflationary" phase to before the Big Bang.
Open questions for testing. Wheeler's varying constants (is N the same each aeon?) would complicate matters but could become testable. CCC needs: Λ truly constant; Hawking evaporation real and complete for every hole, its energy reaching the crossover and potentially leaving an imprint in our CMB (remarkable if so, since Hawking radiation in our aeon is hopelessly unobservable); and mass decay, for which no rate is predicted. If all masses decay in proportion, it would mimic a slowly weakening G; the 1998 limit is under about 1.6 × 10−12 per year, "small beer" against the 10100-year horizon.
The key prediction: circles. Close encounters between supermassive black holes in the previous aeon (deflections, captures, mergers) emit enormous, effectively instantaneous bursts of gravitational waves, carrying away a considerable fraction of the pair's mass. The burst spreads as a thin spherical shell along the future light cone of the encounter e. Because ∇K = 0 is conformally invariant, K stays finite at the crossover, where C = 0 but its normal derivative is not. Two effects follow: the crossover geometry acquires Cotton–York conformal curvature (so our aeon is not exactly FLRW at the bang), and the reborn ϖ dark matter gets a "kick" in the direction of the wave.
Why circles. Our view of the CMB is the intersection of our past light cone with the last-scattering surface, which sits only about 1% of an aeon's conformal height above the crossover, so roughly a sphere on 𝒰−. Each burst's light cone meets 𝒰− in another sphere. Two spheres meet in a circle. The kick is toward or away from us uniformly around it, so each encounter adds a circle uniformly warmer or colder than average.
Rain falls on a still pond; each drop sends out a ring, and soon the overlapping rings look random. When the rain stops (the last black holes evaporate), a photograph shows only a confused pattern, yet the right statistical analysis could reconstruct the impacts and prove the pattern came from discrete drops. The Epilogue returns to Tom watching exactly this.
- May 2008, Princeton: Penrose asks David Spergel whether anyone had seen such circles. "No", then "but then nobody has ever looked!"
- Amir Hajian (Spergel's postdoc) analyses WMAP data: 171 radii from about 1° to 60° in steps of about 0.4°, circles centred on 196 608 points; histograms of average circle temperature against Gaussian expectations.
- Spikes appear but are spurious, caused by circles crossing known hot or cold regions near the Milky Way; masking the galactic plane removes them.
- Many spike circles had radii over 30°, which CCC forbids (below).
- Remaining: an apparent excess of cold circles at radii 7° to 15°.
How the circle search worked in practice. For each trial radius, the analysis drew circles of that radius around each of 196 608 sky points (an even pixelisation of the sphere), averaged the CMB temperature around each circle, and plotted the histogram of those averages. For purely random, Gaussian fluctuations the histogram has a predictable bell shape. CCC circles would add excess tails: too many circles uniformly warmer or colder than chance allows. The difficulty is that the real sky is not purely Gaussian: the Milky Way, point sources and scanning patterns leave structure that a circle can cross. Masking the galactic plane removed the early spikes, and the remaining deviations were then tested for dependence on circularity, which is the logic behind the twist experiment.
The 60° cut-off. If the previous aeon resembles ours, galactic black-hole encounters began no earlier than its "present", two-thirds of the way up. Geometry then caps circle radii at 30° and correlations at 60°. Penrose notes that observed CMB correlations do fall away near 60°, unexplained by inflation as far as he knows, "perhaps" support for CCC.
The geometry of the 30° limit. In a strict conformal diagram, each aeon is drawn with a fixed height, and we sit about two-thirds of the way up ours. Our past light cone, traced back to the crossover, therefore spans a sphere whose radius on the crossover is about two-thirds of an aeon's height. A black-hole encounter that happened two-thirds of the way up the previous aeon sends out a burst whose light cone has only about one-third of an aeon's height left to spread before it reaches the crossover. Two spheres, one large and one at most half its radius, intersect in a circle that subtends no more than about 30° as seen from us. Earlier encounters would give larger circles, but in an aeon like ours supermassive black holes did not exist that early. Hence the predicted cap on circle radii and the related 60° limit on correlations.
The twist test. To check that the non-Gaussian signal really depends on circularity, Penrose had the analysis rerun with an area-preserving twist of the celestial sphere (θ′ = θ, φ′ = φ + 3aπθ2 − 2aθ3), turning circles into ellipses, in three versions: no twist, small twist, large twist. He expected the effect to weaken with twist. In autumn 2008 Hajian found, systematically across radii 8.4° to 12.4° (12 histograms), that the small twist enhanced the effect and the large twist erased it. Penrose's explanation: Weyl curvature from large mass inhomogeneities in our aeon (voids, for instance) distorts circles into ellipses, so a small twist sometimes matches the real distortion. He calls the situation "distinctly tantalizing".
If a gravitational-wave burst in one aeon leaves a circle in the next aeon's sky, then in principle a sufficiently advanced civilisation, near the end of its aeon, could write something into the next one by arranging the collisions of supermassive black holes: a pattern of circles as a message. Penrose does not propose this in the book, but his own account of why the ℐ+/𝒰− identification within one aeon would create causal paradoxes depends on exactly this possibility: information really does cross. CCC is therefore not only a cosmology but a claim that causal influence flows from aeon to aeon, which is what makes it testable at all.
After the book, Penrose and collaborators (Daniel An, Krzysztof Meissner, Paweł Nurowski) shifted from circles to Hawking points. The idea: a supermassive black hole in the previous aeon, evaporating over 10100-odd years, releases its entire mass-energy as Hawking radiation. Conformally compressed onto the crossover, that energy lands in a tiny patch, which after expansion should appear in our CMB as a warm spot about 4° across with a characteristic radial temperature profile. Unlike circles, Hawking points would be a direct observation of Hawking radiation, made in a previous universe.
- High-resolution CMB maps (Simons Observatory, now observing; the planned LiteBIRD satellite for polarisation) can sharpen searches for circles and Hawking points with better foreground control.
- Gravitational-wave background measurements (pulsar timing arrays found a nanohertz background in 2023) constrain supermassive black-hole encounter rates in our own aeon, the analogue of CCC's source.
- Any detection of primordial B-mode polarisation from inflationary gravitational waves would favour inflation; CCC predicts no standard inflationary tensor signal.
- A confirmed evolving dark energy would remove CCC's constant-Λ foundation.
- Dark energy that evolves (not a true constant Λ), since CCC needs a spacelike ℐ+ in the remote future.
- A stable massive charged particle that cannot lose mass, or a direct demonstration that rest-mass is exactly conserved over cosmological times.
- Proof that black-hole evaporation is unitary, removing the phase-space reset that keeps entropy from accumulating.
- Detection of primordial gravitational waves with the amplitude and spectrum inflation predicts, since CCC replaces inflation.
- High-precision CMB data showing no circle or Hawking-point excess, once the signal's predicted strength is pinned down. Penrose's side argues the predicted strength is still uncertain, which also makes this test soft.
- Mass fading is unobserved and unspecified. CCC requires every particle's rest-mass to decay to zero, with no predicted rate; Penrose admits no evidence and no clear test. Lunar laser ranging now limits any fall in G to well under 10−13 per year, though that barely touches a 10100-year process.
- Information loss is a minority view. CCC's entropy accounting depends on black holes destroying information, which most physicists reject. Since 2019 the "island" and replica-wormhole calculations (Penington; Almheiri, Engelhardt, Marolf, Maxfield) reproduce the unitary Page curve for evaporating black holes in semiclassical gravity, strengthening the opposite view.
- Arithmetic slip in the SΛ comparison (10121 vs 3 × 10120 rather than 3 × 10122).
- The Planck temperature is given as 2.5 × 1032 K; with his own Planck energy (2.0 × 109 J) and k = 1.38 × 10−23 J/K the value is about 1.4 × 1032 K.
- Dark matter is said to "comprise some 70% of ordinary matter"; it is about five times ordinary (baryonic) matter.
- The 60° claim: the lack of large-angle CMB correlations is a known anomaly but of debated significance (look-elsewhere effects, cosmic variance), so it is weak support.
- The 2008 twist result was preliminary, unpublished at the time, and explained post hoc.
- The circles debate. Gurzadyan and Penrose (2010) reported concentric low-variance circles in WMAP and BOOMERanG data at about 6σ. Wehus and Eriksen, Moss, Scott and Zibin, and Hajian (all 2010), and later DeAbreu et al. (2015), found no significant excess over standard ΛCDM simulations; the disagreement traced to how simulations were built. Gurzadyan and Penrose replied with a simulation-free "sky-twist" method (2013) and claimed Planck confirmation (2015).
- Hawking points. An, Meissner, Nurowski and Penrose (2018 to 2020) proposed a new signature: warm spots up to about 4° across, the conformally compressed final Hawking radiation of supermassive black holes from the previous aeon. Jow and Scott (2020) found them consistent with standard inflationary skies once the look-elsewhere effect is included.
- 2024 reanalysis. Bodnia et al. (JCAP 2024), re-examining Planck and WMAP, found no low-variance circles and no statistically significant Hawking points under their modelling, noting that absence of evidence does not disprove CCC.
- Erebons. Penrose and collaborators later named the decaying dark-matter particles of CCC "erebons" (Planck-mass scale), whose decays might show in gravitational-wave detectors; Penrose (2017) even suggested CCC signals in LIGO's correlated noise, a claim not taken up by the LIGO collaboration.
- Λ under pressure. CCC needs a true cosmological constant; DESI's 2024 to 2025 hints of weakening dark energy, if confirmed, would undercut the spacelike-ℐ+ premise or force a reformulation.
- The Higgs (anticipated here as the mass mechanism) was found at the LHC in 2012, supporting the picture of massless particles above the electroweak scale.
CCC is a lesson in architectural invariants. Penrose identifies what the system's components actually depend on (null-cone structure, not scale) and designs the join between "releases" around that invariant, then pays for the leftovers (rogue massive particles) with an explicit, unverified requirement (mass fading). In enterprise migrations the same discipline applies: name the invariant the interface truly needs, and list the non-conforming residue honestly as a risk rather than burying it.
Penrose argues geometrically first and algebraically second. Each step of the book is a picture (a phase-space sketch, a null cone, a conformal diagram, a pair of intersecting spheres) that carries the logic, with the equations deferred to an appendix. The pictures are not decoration: CCC was found by looking at a conformal diagram of a Λ > 0 universe and noticing that its future edge has the same shape as its past edge. Readers who learn to read those diagrams can follow, and check, nearly every argument in the book.
Tom Watches the Rain
Tom calls CCC "the craziest idea I ever heard of" and strides off toward his aunt's car. Then he stops by a pond beside the mill where the drizzle is now light enough to show individual raindrops, each making its own spreading ring, and he "couldn't help himself wondering". The image closes the loop with §3.6: discrete impacts leave circles that statistics might recover.
The Mathematics of the Crossover, Condensed
Penrose's Appendices A and B, for experts, write CCC in 2-spinor notation (Penrose and Rindler 1984 and 1986). The point of each step is summarised below; Penrose stresses that parts are "incomplete and provisional", but that well-defined classical equations do propagate one aeon into the next.
Appendix A: conformal rescaling, spinors, Maxwell and Einstein theory
- A1, 2-spinors. Tangent vectors are pairs of spinor indices (unprimed and primed, swapped by complex conjugation). Maxwell's field becomes a symmetric spinor φAB, with equation ∇A′BφAB = 2πJAA′. Einstein's equations: Eab = ½Rgab − Rab = 8πGTab + Λgab.
- A2, massless free fields. One equation, ∇AA′φAB...L = 0, is the "Schrödinger equation" for a massless particle of any spin n/2: neutrinos (n = 1), photons (n = 2), gravitons (n = 4).
- A3 to A5, curvature. The Riemann tensor splits into the Weyl spinor ΨABCD (free gravity), the trace-free Ricci spinor ΦABA′B′ and the scalar curvature; the Bianchi identities make gravity's source equation ∇AA′ΨABCD = ∇B′(BΦCD)A′B′. Massless sources have a trace-free energy tensor.
- A6, conformal rescaling. Massless free-field equations are conformally invariant if the field scales as φ̂ = Ω−1φ; Maxwell with sources stays invariant with Ĵ = Ω−4J.
- A7, Yang–Mills. Same as Maxwell with internal indices (U(2), SU(3)), so conformally invariant until the Higgs adds mass.
- A8, scalar fields. The massless scalar wave equation (□ + R/6)φ = 0 is invariant with weight Ω−3 rather than Ω−1; energy tensors of massless fields scale consistently.
- A9, Weyl scalings. The geometric Weyl spinor Ψ and the radiative field ψ must scale differently (ψ̂ = Ω−1ψ), giving the relation Ψ = Ωψ used in §3.2 and forcing Weyl curvature to vanish at ℐ+.
Why spinors. Penrose works in 2-spinor notation because it exposes conformal behaviour transparently. Every massless field of any spin becomes a totally symmetric spinor with one index type, and every such field obeys the same first-order equation. Under a conformal rescaling, the field picks up a single power of Ω (its conformal weight), and the equation is preserved exactly. In tensor language the same facts are true but scattered across different-looking equations for different spins. The one wrinkle, and the source of CCC's subtleties, is gravity: the geometric Weyl curvature and the radiative graviton field must carry different weights, so they cannot both be finite at ℐ+. CCC resolves the conflict by letting the geometric curvature vanish there while the radiation field stays finite.
Appendix B: equations at the crossover
- B1, three metrics. ĝ (Einstein's metric before the crossover, sources massless), ǧ (Einstein's metric after), and a smooth bridging metric g across the crossover 3-surface, with ĝ = Ω2g and ǧ = ω2g.
- B2, the ϖ-equation. Requiring scalar curvature R = 4Λ makes the scaling field obey (□ + R/6)ϖ = (2/3)Λϖ3, a Calabi equation: conformally invariant, self-coupled.
- B3, the phantom field. Ω itself is a solution; it carries no degrees of freedom, only "keeps track" of Einstein's metric. Before the crossover, Einstein's equations become Tab = Tab[Ω].
- B4 to B5, normals and horizons. Friedrich's results make a smooth spacelike ℐ+ nearly automatic with Λ > 0. A by-product: any cosmological event horizon's limiting cross-section area is exactly 12π/Λ.
- B6, the reciprocal proposal. ω = −Ω−1, encoded in the smooth 1-form Π = dΩ/(Ω2 − 1), which is blind to Ω ↦ Ω−1; the R = 4Λ normalisation fixes the factor to exactly −1.
- B7 to B8, dynamics across. Massless fields (Maxwell, massless Yang–Mills, the Dirac–Weyl equation ∇AA′ωA = 0) continue across; a conformally invariant operator (Eastwood and Rice) forces the crossover to be umbilic (shear-free normals).
- B9, keeping G positive. Naive continuation gives a negative effective gravitational constant; switching to the ǧ interpretation restores it, turning the phantom field into a real field (proto-dark matter).
- B10, fixing the freedom. Two conditions per point of the crossover are needed to pin Ω uniquely; Penrose lists candidates (third-order vanishing of 3ΠaΠa − Λ is his preference) without fully settling.
- B11, matter after the bang. The post-bang energy tensor must acquire a small trace: rest-mass appears unavoidably, perhaps the Higgs mechanism; best delayed as long as possible.
- B12, gravitational radiation at the crossover. Weyl curvature vanishes there, but its normal derivative measures the previous aeon's gravitons. It feeds the Cotton–York tensor (intrinsic conformal curvature of the crossover) and kicks the new dark matter, the mechanism behind the predicted circles.
Master Takeaways
Penrose's book is two things at once: the clearest account anywhere of why the Big Bang's low entropy is the deepest fact behind the arrow of time, and a daring, minority cosmological proposal built to explain it. The first is close to consensus physics; the second stands or falls on assumptions most physicists reject and on CMB evidence that has not held up.
- The Second Law is not a consequence of the laws of motion: those are time-reversible, so the law needs an extra boundary condition, a Big Bang of extraordinarily low entropy.
- The Big Bang's specialness lies in gravity: matter and radiation were in thermal equilibrium, but gravitational degrees of freedom were unexcited, with Weyl curvature essentially zero.
- The odds are about 1 in 1010124: the phase volume of a black-hole-riddled start dwarfs that of our smooth one; this is the number any theory of the beginning must explain.
- Inflation cannot explain it, because its dynamics are time-symmetric: run a lumpy collapse backwards and inflation never switches on.
- Massless particles cannot keep time: a clock needs mass (ν = mc2/h), so a universe of radiation has no physical scale, only conformal structure.
- CCC identifies each aeon's infinite future with the next aeon's Big Bang, joined smoothly by a conformal rescaling, and in doing so forces Weyl curvature to vanish at every bang.
- Entropy is k log V, and its robustness comes from the stupendous ratios between coarse-graining volumes, not from any perfectly objective definition.
- Run statistics backwards and you get nonsense: reconstructing any past from the present silently imports the low-entropy past hypothesis.
- Life runs on low entropy, not energy: the Earth returns all the Sun's energy, but as roughly twenty times as many photons.
- For gravitating systems, uniform means low entropy and black holes are the maximum; supermassive holes now hold most of the universe's entropy (about 1021 per baryon versus 109 in the CMB).
- Black-hole formation is generic (Penrose's 1965 trapped-surface theorem); cosmic censorship and the BKL picture remain the great open problems of classical relativity.
- A positive Λ makes future infinity spacelike, matching the spacelike character of a Big Bang; CCC requires a true cosmological constant.
- CCC requires all rest-mass to fade to zero over eternity, a prediction with no evidence and no specified rate.
- Dark matter is reborn at each crossover from the conformal "phantom field", and both dark matter and dark energy are structurally required.
- CCC's entropy books balance only if black holes destroy information, shrinking phase space between aeons, a view most physicists now reject after the Page-curve calculations of 2019 onwards.
- Penrose rejects horizon entropy 3π/Λ as observer-dependent and devoid of discernible degrees of freedom.
- Quantum gravity matters only where Weyl curvature reaches the Planck scale, so the Big Bang is classical and black-hole singularities are where the strange physics lives.
- The two ends of time are different in kind: smooth, conformal beginnings; chaotic, information-destroying black-hole endings.
- CCC predicts concentric CMB circles from supermassive black-hole encounters in the previous aeon, and later Hawking points; independent analyses find no significant signal.
- Penrose names his alternatives' common failing: bounce, brane, string and baby-universe schemes do not address the suppression of gravitational degrees of freedom.
- Dirac's large number N ≈ 1020 may tie Λ ≈ N−6 to other constants, and might differ between aeons.
- Quantum mechanics is provisional for Penrose: state reduction R is a real, gravity-linked departure from unitary evolution U.
- Phase-space volume is scale-invariant: a conformal rescaling cannot hide entropy, so any cyclic cosmology must say where the accumulated entropy goes.
- The Sun's gift is organisation: a hot spot in a cold sky; a uniformly bright sky would make its energy useless.
- Mass is a clock and charge is a count: mass is non-additive and unexplained, which is why Penrose thinks it, unlike charge, can fade.
- Rival bounce cosmologies founder on clumping: a crunch obeying the Second Law is full of black holes and cannot match a smooth bang.
- CCC is falsifiable: its premises (constant Λ, non-unitary black holes, no inflationary tensor modes) and its CMB signatures all face current experiments.
Then and now: the book's premises in 2010 and 2026
| Premise | In the book (2010) | In 2026 |
|---|---|---|
| Age of the universe | ~1.37 × 1010 years | 13.787 ± 0.020 × 109 years (Planck 2018) |
| Dark energy | Constant Λ "most plausible" | ΛCDM still fits, but DESI hints at evolution (2.8 to 4.2σ) |
| Dark matter | "Some 70%" of matter | About 85% of matter; still undetected directly |
| Higgs boson | Anticipated at the LHC | Discovered 2012; Nobel 2013 |
| Black holes | Strong indirect evidence | Gravitational waves since 2015; EHT images 2019 and 2022 |
| Information paradox | Hawking had recanted (2004) | Island and Page-curve results (2019+) favour unitarity |
| Inflation | Near-universal acceptance | Still dominant; r < 0.036 and no B-mode detection yet |
| CCC circles | Tantalising preliminary WMAP hints | Not confirmed by independent analyses; Hawking points disputed |
| Penrose | Rouse Ball Professor emeritus | Nobel Prize in Physics 2020 |
| Neutron-star maximum | ~1.5 M☉ | At least ~2.1 M☉ observed |
Key equations in one place
| Equation | Meaning |
|---|---|
| S = k log V | Boltzmann entropy of a coarse-graining region of volume V |
| E = mc2, E = hν ⇒ ν = mc2/h | Every massive particle is a clock; massless particles cannot tell time |
| g ↦ Ω2g | Conformal rescaling: angles and null cones kept, scale changed |
| E = 8πT + Λg (Planck units) | Einstein's field equations with cosmological constant |
| SBH = A/4 = 4πM2 | Bekenstein–Hawking entropy (non-rotating hole) |
| tevap ≈ 2 × 1067(M/M☉)3 yr | Hawking evaporation time |
| Ĉ = ΩK̂ | Weyl curvature vanishes where Ω does, at ℐ+ |
| Π = dΩ/(Ω2 − 1) | Scale data blind to Ω ↦ Ω−1, smooth across the crossover |
| (□ + R/6)ϖ = (2/3)Λϖ3 | The ϖ-equation; its solutions are the allowed rescalings |
| Λ ≈ N−6, N ≈ 1020 | Penrose's numerological guess for the cosmological constant |
| AΛ = 12π/Λ, SΛ = 3π/Λ | Horizon area and the disputed horizon entropy |
| 1 in 1010124 | Chance of a Big Bang as special as ours |
Frameworks and mental models
| Name | What it says | Use it when |
|---|---|---|
| Boltzmann entropy | S = k log V of the coarse-graining region | Quantifying how special a macrostate is |
| Coarse graining | Lump microstates that look the same | Defining entropy; judging what "macroscopic" means |
| Liouville's theorem | Hamiltonian flow preserves phase volume | Seeing why entropy rise is about region sizes, not compression |
| Past hypothesis | The universe began in a tiny region ℬ | Explaining any time asymmetry |
| Gravitational entropy inversion | For gravity, clumped is high entropy | Reasoning about stars, galaxies, black holes |
| Weyl vs Ricci curvature | Free field vs source; astigmatic vs magnifying | Deciding what "smooth" means for a space-time |
| Weyl curvature hypothesis | C = 0 at initial singularities | Characterising a low-entropy beginning |
| Conformal diagram | Ω → 0 squashes infinity, Ω → ∞ stretches singularities | Seeing global causal structure |
| Mass as clock | ν = mc2/h | Asking whether time has physical meaning in a regime |
| Trapped surface | Local sign of irreversible collapse | Predicting black holes without symmetry |
| Bekenstein–Hawking entropy | S = A/4 (Planck units) | Entropy budgets involving black holes |
| Information-loss reset | Phase space shrinks as holes evaporate | Following CCC's entropy accounting |
| U and R | Unitary evolution vs state reduction | Discussing measurement and Penrose's gravity-collapse idea |
| Two-sphere intersection | Our past cone and a burst's future cone meet in a circle | Interpreting CCC's CMB predictions |
- Part 1 and Part 2 are strong. The case that the Big Bang's low entropy is gravitational, and that inflation does not explain it, is widely respected even by people who reject CCC, though inflationists reply that inflation never claimed to explain the past hypothesis, only the observed flatness and fluctuation spectrum.
- CCC rests on three heterodox pillars: universal mass decay, non-unitary information loss in black holes, and a pure cosmological constant. Each is a minority or untested position; the second is now more contested than in 2010.
- The observational case has not survived scrutiny. The circles of 2010 and the Hawking points of 2018 to 2020 were not confirmed by independent groups (Wehus and Eriksen; Moss, Scott and Zibin; Jow and Scott; Bodnia et al. 2024), largely because of how statistical significance and simulations were handled.
- Several numbers are dated or slipped: dark matter fraction, neutron-star maximum mass, the Planck temperature, the SΛ comparison and the bath-entropy figure; flagged in the chapters.
- What is missing: a predicted rate for mass decay, a fully settled crossover prescription (Appendix B10 leaves options open), and engagement with the holographic and AdS/CFT arguments for unitarity.
- Credit where due: CCC makes specific, falsifiable predictions about the CMB, which many rival cosmologies of the early universe do not.
Who's who
| Person | Role in the book |
|---|---|
| Ludwig Boltzmann | Entropy as log of phase volume (1875) |
| James Clerk Maxwell | Electromagnetic field equations; the demon |
| Albert Einstein | General relativity; Λ (1917); E = mc2 |
| Hermann Minkowski | 4D space-time geometry (1908) |
| Max Planck | Black-body spectrum; E = hν (1900) |
| Alexander Friedmann, Georges Lemaître | Expanding-universe models; Big Bang proposal (1922 to 1927) |
| Edwin Hubble, Vesto Slipher | Galaxy red shifts and expansion |
| Bondi, Gold, Hoyle | Steady-state model (1948) |
| Dennis Sciama | Penrose's mentor; switched to the Big Bang |
| Penzias and Wilson; Dicke; Gamow | Discovery and prediction of the CMB |
| Smoot and Mather | COBE; 2006 Nobel |
| Oppenheimer and Snyder | First collapse model (1939) |
| Chandrasekhar | White-dwarf mass limit (1931) |
| John Wheeler | Prompted Penrose's collapse work; varying constants; quantum foam |
| Lifshitz, Khalatnikov, Belinski; Misner | BKL singularities; mixmaster |
| Stephen Hawking | Black-hole radiation (1974); reversed on information loss (2004) |
| Jacob Bekenstein | Black-hole entropy (1972) |
| Paul Tod | Conformal-boundary form of the Weyl curvature hypothesis |
| Helmut Friedrich | Smooth conformal infinity with Λ > 0 |
| R. C. Tolman | Entropy-growing cycles; radiation-filled models (1934) |
| Lee Smolin; Veneziano; Steinhardt and Turok; Ashtekar and Bojowald | Rival pre-Big-Bang schemes |
| Paul Dirac; Brandon Carter | Large-number hypothesis; anthropic principle |
| David Spergel, Amir Hajian | First CCC circle search in WMAP data (2008) |
Connect it
- AI and generative models. Diffusion models generate images by learning to run a noising process backwards; they work only because training pins one end of the process to structured data, a direct echo of Penrose's past hypothesis: reversal without a low-entropy anchor yields noise.
- Data and model risk. Penrose's critique of the CCC-circle statistics, and its rebuttals, is a case study in the look-elsewhere effect and simulation dependence: any validation team testing for "anomalies" in transaction or model data faces the same trap.
- Enterprise architecture. CCC's join works because both sides depend only on an invariant interface (conformal structure); the unresolved residue (rogue massive particles) is handled by an explicit requirement. Good migration design names its invariants and its leftover risks the same way.
- Monitoring and drift. Penrose's "coarse graining" is the analyst's choice of features and buckets; his robustness argument explains why well-chosen aggregate metrics are stable, and his spin-echo cases explain why a population can look unchanged by every dashboard metric while carrying hidden structure that a better probe would reveal.
- Model governance. Penrose's habit of listing what his theory requires (mass decay, information loss, constant Λ) as explicit premises is the discipline regulators ask for in model documentation: state assumptions so they can be monitored and, if broken, trigger review.
- Banking and capacity. "Entropy as log of possibilities" maps onto information-theoretic measures used in fraud and anomaly detection, where surprise is the log of improbability; Penrose's warning that entropy depends on the chosen coarse graining applies to every feature-engineering choice.
About this book
Cycles of Time: An Extraordinary New View of the Universe by Roger Penrose, 2010. This is a chapter-by-chapter study of the book, with summaries, maps, tables and diagrams; short quotations keep their page references. It is shared free for personal study, is not for sale, and is not endorsed by the author or publisher. Please read and buy the original.
Insight Editions · rajeshrmahapatra.com/insight-editions