Insight Editions · No. 228
The Hidden Reality
Parallel Universes and the Deep Laws of the Cosmos
Brian Greene
Brian Greene surveys nine kinds of multiverse, from quilted and inflationary to quantum, holographic and simulated, and asks which of them physics could ever test.
A study guide to The Hidden Reality by Brian Greene. Shared free for personal reading; not for sale and not endorsed by the authors or publishers. Please read and buy the original.
The Book on One Page
Brian Greene's The Hidden Reality (2011) asks whether our universe is the only one, and answers that the question is no longer idle. Five different branches of physics, pursued for their own reasons, keep producing parallel worlds as a by-product. The book walks through nine such multiverses, weighs whether any can count as science, and ends on its real theme: how far to trust mathematics as a guide to what exists.
The whole argument in eight moves
- Infinite space forces repetition. If space is infinite (consistent with all data), quantum mechanics and gravity allow only ≈1010122 distinguishable arrangements per cosmic horizon. Infinitely many horizons then repeat every arrangement infinitely often, you included: the Quilted Multiverse.
- Inflation never stops. The theory that explains the CMB's uniformity and its ripples is, in most versions, eternal: it keeps spawning bubble universes, each with possibly different field values (hence different particle physics), and each infinite from inside, so each contains a Quilted Multiverse.
- String theory supplies variety. String theory unifies gravity and quantum mechanics but requires extra dimensions with ≈10500 possible forms. That ended hopes of a unique prediction, but opened braneworlds (other universes millimetres away in a hidden dimension), cyclic brane collisions (universes parallel in time), and a landscape of vacua populated by eternal inflation.
- The cosmological constant needs explaining. Dark energy's density, ≈10−123 in Planck units, is ~120 orders below naive theory. Weinberg's anthropic argument predicted a small nonzero value, if there are more than 10124 universes with varied constants, which the string landscape could provide.
- Multiverses can be science, in principle. Science already accepts unobservable entities when the theory that implies them is well confirmed. A multiverse becomes predictive if its universes are distributed in a skewed or correlated way. The obstacles are computing those distributions, justifying typicality, and the unsolved measure problem of comparing infinities.
- Quantum mechanics may already describe many worlds. Take Schrödinger's equation seriously for everything, devices and brains included, and every outcome is realised in its own decohered branch (Everett). The probability problem (why some branches count more) is still open.
- Information is bounded by area. Black-hole thermodynamics shows a region's information capacity is set by its surface area in Planck units. Maldacena's AdS/CFT made this exact: string theory with gravity in a bulk equals a gravity-free quantum field theory on its boundary.
- Simulation and mathematics. If minds are substrate-independent, simulated universes are likely and Bostrom's argument suggests we may be simulated. The Ultimate Multiverse (every possible mathematics is real) is the only proposal Greene rejects as untestable by its nature. His bet: reality will prove fundamentally discrete.
Chapter map
| Ch | Title | One-line payload |
|---|---|---|
| P | Preface | Common experience deceives; the multiverse emerges from theories built for other reasons |
| 1 | The Bounds of Reality | Nine multiverses from five frontiers; none proven; all arise "unbidden" |
| 2 | Endless Doppelgängers | Infinite space + finite configurations (1010122) ⇒ infinite copies of you |
| 3 | Eternity and Infinity | Inflation explains the CMB, is usually eternal, and makes bubbles that are infinite inside |
| 4 | Unifying Nature's Laws | String theory: excellent on quantum gravity, indeterminate on predictions |
| 5 | Hovering Universes | Dualities, M-theory, branes; braneworlds and cyclic collisions; 10500 flux vacua |
| 6 | New Thinking About an Old Constant | Λ ≈ 10−123; Weinberg's anthropic bound; landscape + eternal inflation |
| 7 | Science and the Multiverse | When unobservable universes are legitimate; predictive distributions; the measure problem |
| 8 | The Many Worlds of Quantum Measurement | Everett's no-collapse quantum theory; decoherence; the probability puzzle |
| 9 | Black Holes and Holograms | Entropy = area; the holographic bound; Maldacena's bulk-boundary duality |
| 10 | Universes, Computers, and Mathematical Reality | Making universes; simulation argument; Ultimate Multiverse and computability |
| 11 | The Limits of Inquiry | Five questions; eight multiverses testable in principle; trust in mathematics |
How to read this edition. Each chapter opens with its payload, keeps every argument, number, person and thought experiment from the book, and closes with Going deeper (technical points from the book's notes), a Critical lens (contested claims and errors, flagged rather than silently corrected), a 2026 update (what fifteen years of data have done to the argument) and a So what box for practitioners.
Preface: Why Common Experience Misleads
Greene opens with a single premise that drives the whole book: everyday experience is a poor guide to the true nature of reality. Evolution equipped us for forests and savannas, not for quantum electrons or black-hole cosmology, yet our enlarged brains gave us instruments and mathematics, and those tools have already overturned our picture of space, time, matter and energy. The next upheaval may be the discovery that our universe is not the only universe.
The premise and the promise
By the turn of the twenty-first century, Greene says, it had become a foregone conclusion that common experience is deceptive when it comes to the deep structure of the world. That is no surprise: computing the quantum behaviour of electrons gave our ancestors no survival advantage, while a larger brain did. As our faculties grew, some of our species built equipment to extend the reach of the senses, and others mastered mathematics, a systematic method for detecting and expressing pattern. Together these let us peer behind appearances.
What we found has already demanded sweeping changes: space, time, matter and energy display a behavioural repertoire unlike anything we directly witness. The book explores the possibility that penetrating analyses of these discoveries now point to a multiverse, a reality in which our universe is one member of a larger collection.
How the book is built
- No preparatory chapters. Unlike The Elegant Universe and The Fabric of the Cosmos, Greene does not open with systematic tutorials on special relativity, general relativity and quantum mechanics. He introduces each idea "as needed", warning experienced readers which passages they may skip.
- Deeper end-of-chapter sections. Many chapters close with more technical material. Greene offers a short summary and permission to jump ahead, but urges readers to go as far as patience allows; the only prerequisite is "the will to persevere."
- Notes for specialists. The endnotes carry clarifications and extensions aimed at readers with formal training in mathematics or physics. This edition folds their substantive points into the relevant chapters.
- Metaphor, analogy and history. As in his earlier books, Greene relies on analogy interspersed with historical episodes, choosing among the field's twists and turns to build a path "peak by valley" from the everyday to the wholly unfamiliar.
- Readers and critics thanked: David Albert, Richard Easther, Simon Judes, Daniel Kabat, Juan Maldacena, Maulik Parikh, Marcus Poessel, Ken Vineberg and others.
- Colleagues whose conversations shaped the book: Raphael Bousso, Robert Brandenberger, Frederik Denef, Jacques Distler, Michael Douglas, Lawrence Krauss, Janna Levin, Andrei Linde, Seth Lloyd, Saul Perlmutter, Jürgen Schmidhuber, Steve Shenker, Paul Steinhardt, Andrew Strominger, Leonard Susskind, Max Tegmark, Henry Tye, Cumrun Vafa, David Wallace, Erick Weinberg and Shing-Tung Yau. The list reads as a roster of the book's protagonists: almost every multiverse in it has a champion here.
- Institutional base: Columbia University's Institute for Strings, Cosmology, and Astroparticle Physics (ISCAP), supported by the late Pentti Kouri and the US Department of Energy.
- Illustrator: Jason Severs drew the figures.
The book's method is its message: the multiverse is not a hypothesis anyone set out to prove; it is something that keeps falling out of theories built for other reasons. The preface frames everything that follows as an exercise in following the mathematics wherever it leads, while staying honest about what the data do and do not support.
The Bounds of Reality: On Parallel Worlds
Chapter 1 is the map of the territory. Greene argues that nearly every major frontier of fundamental physics (relativistic, quantum, cosmological, unified and computational) has independently led researchers to some version of parallel universes. He previews nine distinct multiverses, insists that none is established by data, and defends a single stance: these proposals deserve attention because they emerge unbidden from the mathematics of theories built to explain ordinary observations.
Two mirrors and an idea
Greene's boyhood room had two mirrors: one on the closet door aligned with one on the wall, producing an apparently endless corridor of reflections. Knowing that light travels at a finite speed, he understood that the images only seemed to move in unison; in his mind's eye each reflection lagged the one before. He would imagine "an irreverent me way down the line" who refused to fall into step, creating a new reality that informed all those that followed. Reflections have no minds of their own, but the daydream captures the theme of the book: worlds lying beyond the one we know, approached now as science rather than fantasy.
What "universe" now means
"Universe" once meant "all there is": the whole shebang, so a plurality of universes would be a contradiction in terms. Theoretical developments have quietly made the word context-dependent:
- Everything. Sometimes it still means absolutely all that exists.
- The accessible part. Sometimes it means only the portion of everything that someone like us could, in principle, have access to.
- One member of a collection. Sometimes it names a separate region, partly or fully, temporarily or permanently inaccessible, making our universe one member of a large, perhaps infinite, ensemble.
The wider canvas goes by many names, all synonymous: parallel worlds, parallel universes, multiple universes, alternate universes, metaverse, megaverse, multiverse. Greene refuses to fix a precise definition of where one universe ends and another begins. He borrows Justice Potter Stewart's line on obscenity, "I know it when I see it," and argues that labelling is only a question of language. What matters is whether regions of reality exist that make what we have long called the universe just one component of a far grander, mostly hidden reality.
Nine routes to many worlds
A striking fact, and part of Greene's reason for writing, is that relativistic physics, quantum physics, cosmological physics, unified physics and computational physics have each led to some variety of parallel universe. The nine variants differ in where the other universes are (vast distances away in space or time, millimetres away in an extra dimension, or in no meaningful location at all) and in what laws govern them (the same as ours, different but sharing a heritage, or wholly unlike anything known).
| Multiverse | Chapter | Source theory | What the other universes are |
|---|---|---|---|
| Quilted | 2 | Infinite space + finite configurations | Distant patches of one infinite space, some with copies of us |
| Inflationary | 3 | Eternal inflation | Bubble universes nucleating forever in an inflating sea |
| Brane | 5 | String/M-theory braneworlds | 3-D slabs floating in higher-dimensional bulk, possibly millimetres away |
| Cyclic | 5 | Colliding branes | Successive universes parallel in time, each born in a brane collision |
| Landscape | 6 | String landscape + inflation | Bubbles with different extra-dimensional shapes, hence different physics |
| Quantum | 8 | Many Worlds interpretation | Every outcome of every quantum process realised in its own branch |
| Holographic | 9 | Black-hole entropy, AdS/CFT | A boundary description that mirrors, exactly, the physics inside |
| Simulated | 10 | Computation | Universes running as software on superadvanced computers |
| Ultimate | 10 | Philosophical principle of plenitude | Every possible universe, every possible mathematics, realised |
The quantum seed (1950s)
The earliest scientific forays came in the 1950s from researchers puzzling over quantum mechanics. Quantum theory broke with classical mechanics by making predictions necessarily probabilistic: we can predict the odds of each outcome but generally not which will occur. A deeper and less publicised puzzle remains. Despite mountains of confirming data, no one has explained why only one of the many possible outcomes actually happens. More than a century after the quantum revolution began, there is no consensus on how a single definite experienced reality squares with the theory's mathematics.
The most startling proposal was among the first: maybe all outcomes happen, each in its own universe. If the calculation says a particle might be here or there, then in one universe it is here and in another it is there, each containing a copy of you who wrongly believes theirs is the only reality. Since quantum mechanics underlies everything from solar fusion to neural firing, the implication is sweeping: "there's no such thing as a road untraveled." After more than half a century, the Many Worlds approach remains controversial: some practitioners call it proved, others say its mathematical foundations do not hold together.
Greene notes how this idea fed a cultural current he loves: The Wizard of Oz, It's a Wonderful Life, the Star Trek episode "The City on the Edge of Forever," Borges's "The Garden of Forking Paths," Sliding Doors and Run Lola Run. Yet quantum mechanics is only one route, and not the first he treats.
The road ahead, chapter by chapter
- Chapter 2, infinite space. If space extends infinitely far (consistent with all observations and favoured by many cosmologists), then far out there must be copies of you and me living out alternate versions of our reality.
- Chapter 3, inflation. If the inflationary theory is right, the burst of super-fast expansion that made our region may not be unique; inflation may be spawning universe upon universe for all eternity, and each such universe is itself spatially infinite, containing infinitely many Chapter 2 worlds.
- Chapter 4, string theory. A brief primer and a status report on the leading attempt to unify all of nature's laws.
- Chapter 5, branes and cycles. The braneworld scenario makes our universe one of potentially many "slabs" in a higher-dimensional loaf, possibly with signatures at the Large Hadron Collider. Branes that slam together could trigger big bang-like beginnings over and over, giving universes parallel in time.
- Chapter 6, the landscape. The enormous number of possible shapes for string theory's extra dimensions, combined with inflation, suggests a vast collection of universes realising every possible form. This connects to one of the century's most surprising observations: space is filled with a uniform diffuse energy, perhaps Einstein's cosmological constant.
- Chapter 7, method. Can parallel universes be science? Greene previews his position that under specific conditions they fall unequivocally within science.
- Chapter 8, quantum measurement. The Many Worlds argument laid out step by step.
- Chapter 9, holography. Thirty years of black-hole research culminating in a string-theory result: what we experience may be a holographic projection of processes on a distant bounding surface. You can pinch yourself and the feeling is real, yet it mirrors a parallel process in a different reality.
- Chapter 10, artificial and ultimate. Can physics let us create universes? Could we be living in a simulation? And the most unrestrained proposal: every possible universe exists, which opens a closing inquiry into the role of mathematics in science.
The cosmic order: Greene's stance
No experiment or observation has established any version of the multiverse. Greene is not trying to convince you that we live in one; he says no one should be convinced of anything not supported by hard data. His claim is narrower and more interesting: every parallel-universe proposal worth taking seriously "emerges unbidden" from mathematics built to explain conventional data. Physicists are not standing with "multiverse nets" waiting to snare passing theories.
Greene states three goals. First, to lay out the chain of theoretical insights that leads, from many perspectives, to a multiverse. Second, to show how certain otherwise confounding observations become understandable inside a parallel-universe framework. Third, to describe the unresolved questions that so far keep that explanatory approach from being fully realised.
Marginalisation view: if we are one of countless universes, our place and importance shrink.
Greene's view: significance is not measured by relative abundance. What marks our arrival is the depth of our understanding, the capacity of analytical thought to bridge vast distances, to outer and inner space and perhaps beyond our universe, all from "our lonely vantage point in the inky black stillness of a cold and forbidding cosmos."
Framing to watch. Greene's "nine multiverses" are not independent hypotheses of equal standing. Several nest inside one another (Quilted inside Inflationary inside Landscape), two are interpretive (Quantum, Holographic) rather than claims about extra regions of space, and two are philosophical (Simulated, Ultimate). Counting them as nine can inflate the impression that physics "keeps finding" multiverses. Also note that the holographic "multiverse" is a duality: two descriptions of one reality, not two realities.
Architecture lens. Greene's taxonomy is a model of how to handle an over-loaded term: rather than legislate one definition of "universe", he classifies variants by two independent axes (location relative to us, and relation of their laws to ours). The same move clarifies loaded enterprise words like "platform", "agent" or "data product": separate the dimensions, then name the cells.
Endless Doppelgängers: The Quilted Multiverse
If space is infinite, then copies of you exist. Greene builds this conclusion from two premises that are each hard to deny: an infinite cosmos contains infinitely many causally separate regions, and quantum mechanics plus gravity allow each region only a finite number of distinguishable particle arrangements (about 1010122). Infinite trials over finite possibilities guarantee infinite repetition. To get there, the chapter rebuilds modern cosmology from Einstein's 1915 equations to the late-1990s discovery of dark energy, because whether space is finite or infinite is the hinge of the whole argument.
The question
Travel outward forever: does space go on indefinitely, stop abruptly, or bring you back home as Sir Francis Drake returned after circumnavigating the earth? Both a spatially infinite cosmos and a huge-but-finite one are compatible with all observations, and both have been studied vigorously. But one consequence of the infinite option has received scant attention. Somewhere in an infinite cosmos is a galaxy just like the Milky Way, a solar system just like ours, an earth, a house like yours, and someone who looks just like you reading this very sentence. There are infinitely many such copies: some reading along with you, some who have skipped ahead or gone for a snack, some you would not want to meet in a dark alley. And you won't meet them: they live so far away that light travelling since the big bang has not had time to cross the gap.
The father of the big bang
Solvay, Brussels, 1927. At the Hotel Metropole, during a week of debate on quantum theory, the Belgian Georges Lemaître (doctorate complete and ordained a Jesuit priest by 1923) approached Albert Einstein in his clerical collar. Lemaître had shown that general relativity implies the universe began as a tiny, astonishingly dense speck, later called the "primeval atom", which swelled into today's cosmos. Einstein had read the paper and found no error in the mathematics, yet replied: "Your mathematics is correct, but your physics is abominable."
Mount Wilson, 1933. Six years later Einstein heard a more detailed version and stood to call it "the most beautiful and satisfactory explanation of creation to which I have ever listened." Lemaître became known among scientists as the father of the big bang.
Lemaître was not the first. In 1921 the Russian mathematician and meteorologist Alexander Friedmann found solutions of Einstein's equations in which space stretches. Einstein first claimed Friedmann had made errors, was wrong, and retracted. But he "refused to be mathematics' pawn", preferring his intuition that the universe is eternal and, on the largest scales, fixed and unchanging.
Dates and quotes. Friedmann's expanding-universe paper appeared in Zeitschrift für Physik in 1922 (submitted that year), not 1921. Greene himself notes that the historian Helge Kragh questions the 1933 "most beautiful" quotation, which may be a garbled Newsweek report of Einstein speaking about cosmic rays. What is certain is that by 1933 Einstein had accepted a dynamic cosmos.
General relativity in brief
On 25 November 1915 Einstein sent the Annalen der Physik the culmination of a nearly ten-year odyssey: general relativity, a geometrical refashioning of gravity. He had begun around 1907, when most scientists thought Newton had long since settled gravity with the Universal Law of Gravity (late 1600s), still used by NASA for spacecraft trajectories and by astronomers for comets, stars and galaxies.
- Newton's law. F = Gm1m2/r2, where G is Newton's constant, the measured intrinsic strength of gravity.
- Newton's silence. In the Principia (1687) Newton admitted his law says nothing about how the sun reaches across 93 million miles of empty space to move the earth. He left the mechanism "to the consideration of the reader", and for two centuries readers simply read on. Einstein could not.
- Einstein's answer. There is something in empty space: space itself. Space (really spacetime) is gravity's medium.
The warped-table picture. A marble rolls straight across a flat metal table. If fire buckles the table, the marble follows the ruts. Empty space is the flat table; mass and energy warp spacetime, and the warps guide motion. The sun makes a bulge around itself, and the earth's orbit follows it. Three refinements: time curves as well as space; the table needs earth's gravity to hold the marble down whereas spacetime warps are gravity; and real space warps all around an object, not just underneath. Before Einstein, gravity was a mysterious force acting across space. After Einstein, gravity is a distortion of the environment caused by one body that guides the motion of others. You are anchored to the floor because you are trying to slide into an indentation in spacetime made by the earth.
For everyday gravity it is the curvature of time, not space, that dominates. A clock on the ground runs slow relative to one atop the Empire State Building by billionths of a second per year. General relativity says objects move toward regions where time runs more slowly: in a sense, every object "wants" to age as slowly as possible, which is why a dropped object falls.
The Einstein field equations (Rμν − ½ gμνR = 8πG Tμν) say precisely how matter and energy curve spacetime (with E = mc2 making mass and energy interchangeable) and how that curvature moves stars, planets, comets and light.
- 1915: Einstein explains the anomaly in Mercury's orbit; he told Adrian Fokker the thrill gave him heart palpitations for hours.
- 1919: Arthur Eddington's eclipse expedition finds starlight bent by the sun as predicted. New York Times: "LIGHTS ALL ASKEW IN THE HEAVENS, MEN OF SCIENCE MORE OR LESS AGOG." Einstein becomes Newton's heir apparent. (A later charge that Eddington culled inconvenient plates was examined by Daniel Kennefick (2007), who concluded the confirmation was reliable.)
- 1970s: Hydrogen maser clocks confirm earth's warping of spacetime to about 1 part in 15,000.
- 2003: Cassini-Huygens radio-wave timing near the sun confirms curved spacetime to about 1 part in 50,000.
- Today: GPS satellites must correct for the curvature they experience in orbit; without it, position fixes drift rapidly.
The universe and the teapot: the cosmological principle
In 1917 Einstein applied the theory to the whole cosmos. Theoretical physics, Greene says, is "the art of knowing what to ignore." Einstein's simplification was the cosmological principle: on the largest scales, the universe is uniform. A cup of tea is a chaos of H2O, polyphenol and tannin molecules up close, yet a uniform hazel to the eye. Einstein guessed the cosmos was the same, guided by the sense that no location is special, at a time when even the existence of other galaxies was unsettled.
- Observational support today. The principle holds, but only on scales of at least 100 million light-years, about 1,000 times the length of the Milky Way. Average galaxy counts, matter content and temperature in one such box are indistinguishable from those in another a billion light-years away: "if you've seen one 100-million-light-year chunk of the cosmos, you've pretty much seen them all."
- Why it matters. Describing every sand grain on a beach is hopeless; describing average weight per cubic metre is easy because the beach is uniform. Likewise, Einstein's ten interlocking field equations, a "Gordian knot", collapse to a single equation for a uniform universe: the Friedmann equation, (da/dt)2/a2 = (8πG/3)ρ − k/a2, where a(t) is the scale factor (the overall distance scale between galaxies), ρ the energy density and k = +1, 0, −1 for spherical, flat and hyperbolic space.
The unwelcome result: no static universe
The equation showed that the density of matter and energy cannot stay constant: it must grow or shrink. The physics is pedestrian. A baseball hit to centre field rises, slows and falls; it cannot hover like a blimp because gravity pulls one way and nothing pushes back (the blimp has buoyancy from helium lighter than air). The universe is more like the baseball than the blimp: with only attractive gravity, space must stretch or contract, never sit still. A box 100 million light-years on a side today will not be that size tomorrow. Einstein had founded scientific cosmology and was deeply distressed by where the math had taken him.
Taxing gravity: the third line
The popular story says Einstein "blinked" and mangled his equations. Greene says that is only partly true: his change was minimal and thoroughly sensible.
The general-relativity "tax form" has three lines:
- Geometry of spacetime: its warps and curves, the embodiment of gravity.
- Distribution of matter and energy: the source of the warps.
- Energy stitched into the fabric of space itself: an intrinsic property of spacetime, on an absolutely equal mathematical footing with the other two. Just as each cubic metre of water carries energy summarised by its temperature, each cubic metre of space carries energy summarised by this number.
In 1915 Einstein left line 3 blank, which is mathematically zero but psychologically "not relevant to me." Re-examining it, he saw no observational justification for zero.
Entering a positive value gives space a uniform positive energy, and (for reasons Chapter 3 explains) every region pushes away from every other: repulsive gravity. Tune the number precisely and the outward push exactly balances the inward pull of matter: a hovering blimp of a universe. Einstein called it the cosmological member or cosmological constant. He could not explain why space would carry exactly the right energy, but the augmented theory now matched the expected eternal cosmos.
A flaw Einstein was told about. Greene's notes add that Willem de Sitter pointed out the static solution is unstable: nudge it larger and it grows, smaller and it shrinks. Physicists distrust solutions that persist only under perfect, undisturbed conditions, so Einstein's misgivings predated the data.
The primeval atom vindicated
Having already wrestled the equations into a static universe and brushed off Friedmann, Einstein had little patience in 1927, faulting Lemaître for blindly following mathematics to an absurd conclusion. The setback was short-lived. In 1929, with what was then the world's largest telescope at Mount Wilson, Edwin Hubble showed that distant galaxies are all rushing away from the Milky Way. The universe is expanding (work that also leaned on Vesto Slipher, Harlow Shapley and Milton Humason). Einstein's motive for the cosmological constant evaporated, and the big bang model became the scientific story of creation. Friedmann got credit as the first to find expanding solutions; Lemaître for independently developing them into robust scenarios. Had Einstein not imposed his unjustified conviction of stasis, he might have predicted cosmic expansion more than a decade before it was observed. But the cosmological constant's story was not over.
Expansion needs no ongoing push. Like a ball already moving upward, galaxies carry on outward while attractive gravity slows them. The deeper question, what launched the expansion, is answered in Chapter 3 by a brief early burst of repulsive gravity.
The models and the data: what shape is space?
The big bang model is not one scenario but a handful, all expanding, differing in the overall shape of space and, decisively, in whether space is finite or infinite. The cosmological principle culls the shapes drastically (most bulge here and flatten there), but does not pick one. Two-dimensional analogues are mathematically exact guides.
| Shape (2-D analogue) | Curvature | Reflection in a mirror of that shape | Extent |
|---|---|---|---|
| Sphere (cue ball) | Constant positive | Bloats outward | Finite |
| Infinite tabletop | Constant zero (flat) | Undistorted | Infinite |
| Video-game screen (2-D torus) | Constant zero (flat) | Undistorted | Finite |
| Infinite Pringles chip (saddle) | Constant negative | Shrinks inward | Infinite |
- The torus explained. When Ms. Pac-Man leaves the left edge she reappears on the right; leave the bottom, reappear at the top. Roll the screen into a cylinder, bend the cylinder and tape the ends: you get the familiar doughnut. The doughnut looks curved only because it is drawn sitting in 3-D; intrinsically the torus is flat. Its "edges" are fictitious.
- Same local geometry, different global fate. On an infinite tabletop a fixed heading never brings you home; on the video-game screen you return without ever turning the wheel.
- Exotic extras. The list omits shapes like the binary tetrahedral space and Poincaré dodecahedral space, carved from these by "judicious slicing and paring" (technically, quotients of simply connected spaces by discrete isometry groups).
Central conclusion: the cosmological principle winnows the possible shapes, and some survivors are infinite while others are not. Expansion applies to all of them: an inflating balloon (positive), a uniformly stretched rubber sheet (flat), a stretched Pringles-shaped sheet (negative). Galaxies are glitter on the surface, separating as Hubble saw.
Weighing the universe: the critical density
General relativity reduces the question "which shape?" to one observable: the average density of matter and energy. A lot curves space back into a sphere; a little lets it flare into a saddle; "just the right amount", the critical density, gives zero curvature.
- Critical density today: about 2 × 10−29 g/cm3, Greene's figure, which he equates to about six hydrogen atoms per cubic metre, or one raindrop per earth-sized volume. (It was higher in the early universe.)
- Observed matter (stars plus inferred unseen mass): about 27% of critical, roughly two hydrogen atoms per m3, which alone would imply a negatively curved universe.
- Late-1990s surprise: a diffuse energy spread uniformly through space, contributing about 73% of critical.
- Total: 27% + 73% = 100% of critical, the value for zero spatial curvature.
- Age of the universe: about 13.7 billion years, for either finite or infinite space.
The uniform energy sounded like Einstein's retracted constant. Because no one yet knew whether it is truly constant or evolves, and because it emits no light, astronomers named it dark energy; "dark" also describes our ignorance of its origin and composition. Even so, Hubble Space Telescope and ground-based observations had reached consensus on its amount. Current data favour an ever-expanding universe that is flat: the 3-D version of either the infinite tabletop or the finite video-game screen.
Numbers that don't quite agree. The critical density is ρc = 3H02/8πG ≈ 1.9 × 10−29 h2 g/cm3. With today's h ≈ 0.68 that is about 9 × 10−30 g/cm3, which matches Greene's "six hydrogen atoms per m3". His 2 × 10−29 figure is the h = 1 value and corresponds to roughly twelve atoms. The two statements are off from each other by about a factor of two.
Finite or infinite? The observational stalemate
- Repeated images. In a finite universe, light could lap the cosmos and produce repeated images of the same galaxies, like the parallel mirrors of Chapter 1. Searches found none, which suggests that if space is finite, it is too large for light to have completed multiple laps. The larger a finite universe, the better it masquerades as infinite.
- Where the distinction doesn't matter. The age (13.7 billion years) follows from today's expansion rate and its history, whatever the extent.
- Where it matters: the beginning. A finite universe run backward really does shrink to an ever-smaller speck. An infinite universe was always infinite: shrinking it by a factor of 2 or of a million leaves infinity. Near time zero it becomes denser at every location, but its extent stays unending.
- What cosmologists believe. When pressed, most favour infinite space. Greene suspects two reasons: a historical neglect of the finite video-game shapes, which are harder to analyse, and the misconception that infinite versus huge-but-finite is merely academic. It is not academic: it changes the nature of reality.
Infinite space and the patchwork quilt
Step 1: infinitely many independent regions
Imelda's wardrobe. 500 dresses × 1,000 pairs of shoes = 500,000 outfits, about 1,400 years of daily changes. Live forever and she must wear every outfit infinitely many times.
Randy's decks. Fifty-two cards can be ordered in 52! ≈ 8 × 1067 ways. Shuffle more decks than that and some must match; shuffle infinitely many and every ordering repeats infinitely often.
Lesson: infinitely many occurrences over finitely many configurations guarantee infinite repetition.
Light's finite speed limits what we can see. The universe is 13.7 billion years old, but because space kept expanding while old light was in transit, the most distant sources whose light reaches us today are now about 41 billion light-years away, giving an observable universe roughly 82 billion light-years across. Beyond that is our cosmic horizon, like ships over the sea's horizon. Special relativity adds that no signal, influence or information travels faster than light, so regions beyond each other's horizons have never interacted and have evolved independently.
Picture space at one moment as a patchwork quilt of circular patches, each a cosmic horizon. Neighbouring patches overlap at their borders, but take every other row and every other column and the patches are wholly isolated. In 3-D the patches are spheres. Finite space means finitely many independent patches; infinite space means infinitely many.
Step 2: finitely many possibilities per region
The fly in the bedroom. You plead with a buzzing fly that a big bedroom offers "so many other places." "How many?" it asks. Classically, infinitely many: positions and speeds vary continuously. The fly objects: it cannot tell here from a billionth of a centimetre to the left, nor 0.25 from 0.249999999 km/h, so for any being with finite resolution, the variety is finite. If the smallest detectable step is 0.01 cm, each cubic centimetre offers 1003 = 1,000,000 locations and a bedroom about 100 trillion.
Better technology might seem to push resolution ever finer. Quantum mechanics says no, ever. The uncertainty principle (Werner Heisenberg) says sharpening the measurement of one property necessarily blurs its complement: position versus speed (momentum). The camera analogy: a fast shutter freezes the fly's location but tells you nothing about its speed; a slow shutter shows motion as a blur. Heisenberg's math makes the trade-off exact: (uncertainty in position) × (uncertainty in momentum) ≥ about Planck's constant. A sharper position measurement needs a higher-energy probe (bedside lamp shows legs and eyes; brief bursts of x-rays reveal wing muscles), and perfect resolution needs infinite energy.
Classical physics makes perfect resolution unattainable in practice; quantum physics makes it unattainable in principle. "Changes that are too small to be measured, even in principle, are not changes at all." Greene calls the residue quantum jitter. A finite region also caps position uncertainty at the region's size, which sets a floor on momentum resolution. Limited energy in a limited region gives finitely many distinct states.
Step 3: the gravitational cap on energy
How many arrangements fit in one patch (a sphere of radius 41 billion light-years)? As with Lego, more pieces mean more arrangements, but you cannot cram in pieces indefinitely. Too much energy in a region collapses it into a black hole, and adding more simply grows the event horizon to enclose more space. So a region of given size holds at most a fixed amount of matter and energy: for today's horizon, about 1056 grams. The size of the limit is not the point; the existence of a limit is.
Finite energy means finitely many particles (electrons, protons, neutrons, neutrinos, muons, photons and any undiscovered species), each with finitely many distinguishable positions and speeds. In proper quantum language: finitely many observably distinct quantum states per patch.
- Distinct configurations per cosmic horizon: about 1010122, a 1 followed by 10122 zeros.
- Derivation: a black hole filling our horizon has radius about 1028 cm ≈ 1061 Planck lengths; its entropy (Chapter 9) is its area in Planck units, about 10122; the number of states is eentropy ≈ 1010122.
- Nearest exact copy of our whole horizon: about 1010122 metres away.
- Nearest exact copy of a region a few light-years around the sun: about 1010100 metres away.
- Matter-energy cap for our horizon: about 1056 g.
Nothing but physics: why copies would be you
Greene declares his bias: a physical system is completely determined by the arrangement of its particles. Specify how the particles of earth, sun and galaxy are arranged and you have fully specified reality, minds included. Some believe life needs a nonphysical ingredient (spirit, soul, life force, chi); Greene remains open but has never encountered evidence for it. (He uses particles rather than fields only for convenience: the two languages are complementary, like describing an ocean wave by its water molecules.)
On that view, a patch whose particles duplicate ours looks and feels like ours in every way. There are many perfect copies of you, feeling exactly what you feel now, and no fact about which one is "really" you: all are physically, hence mentally, identical.
- Approximate copies are far more common. There is one way to duplicate a region exactly and many ways to almost duplicate it. Every decision is a particle arrangement: turn left or right, say yes or no. So every choice you made and every option you rejected is played out in some patch.
- Spectrum of variants. Some patches are barely distinguishable from ours; some realise your worst fears, some your wildest dreams; some are unrecognisable. Most patches contain no life at all.
- Growing horizons. Over time horizons grow and eventually overlap, so today's parallel universes merge. The conclusion survives: redraw the grid with bigger patches spaced farther apart, and infinite space has room for it. (With accelerating expansion, horizons stop growing beyond a limit, which only sharpens the separation.)
The Quilted Multiverse: at any moment an infinite space contains an infinite number of separate domains, our observable universe being one. Because particle arrangements must repeat, the reality in any given patch, ours included, is replicated an infinite number of times. (The idea was explored for a classical infinite universe by George Ellis and G. Brundrit, and in the quantum context by Jaume Garriga and Alexander Vilenkin.)
What to make of this? Three escape routes
Greene weighs the objection that such a bizarre conclusion proves an assumption wrong.
| Escape route | Possible? | Greene's verdict |
|---|---|---|
| Space beyond our horizon is empty | Yes | Requires theoretical contortions; the best cosmological theories point nowhere near it |
| Laws of physics change beyond the horizon | Yes | Chapter 3 shows laws can vary, but that variation does not undo the argument |
| Space is finite | Definitely | The most convincing way to upend the Quilted Multiverse: a smallish finite universe lacks room for many patches, let alone duplicates |
The stakes therefore rest on the finite-versus-infinite question, and here inflationary cosmology, developed by physicists pushing the big bang back to time zero, makes infinite space an almost inevitable conclusion while introducing an even more exotic multiverse.
Going deeper: the technical notes
- Friedmann's equation, read physically. In (da/dt)2/a2 = (8πG/3)ρ − k/a2, the left side is the squared expansion rate (the Hubble rate H2). Matter density dilutes as ρ ∝ a−3, radiation as a−4 (extra redshift loss), and a cosmological constant not at all. That one difference drives everything in Chapter 6: as a grows, Λ inevitably wins.
- Comoving coordinates. Cosmologists "paint" coordinates on galaxies and let the scale factor carry the expansion. Strictly, "a box 100 million light-years across grows" means the region containing a given set of N galaxies has a larger volume later.
- Density need not fall as space grows. The intuitive "expansion dilutes" assumes ordinary matter. For a cosmological constant, energy density stays fixed; in exotic cases density can even rise as space expands, because gravity can supply energy.
- Why GR forbids stasis. Einstein's static solution balanced Λ against matter, but de Sitter showed it is unstable: any perturbation runs away. Physicists distrust solutions that survive only in perfect conditions.
- The torus, honestly. The doughnut picture is an embedding artefact: a 2-D torus made by gluing a flat screen's edges is intrinsically flat. Its curved look comes from sitting in 3-D space.
- "Slicing and paring." The exotic finite spaces (binary tetrahedral, Poincaré dodecahedral) are quotients of simply connected spaces by discrete groups of isometries, which is why they share local curvature with the sphere or flat space but differ globally.
- Critical density is epoch-dependent. The 2 × 10−29 g/cm3 figure is today's; ρc = 3H2/8πG was far larger when H was larger.
- The horizon distance. Light emitted 13.7 billion years ago comes from a source now ≈41 billion light-years away (Greene's figure) because each segment of its path has stretched since the light crossed it; the observable universe is therefore ≈82 billion light-years across. In an accelerating universe horizons stop growing beyond a fixed size, so some regions are lost to us forever.
- The 1010122 estimate uses the Chapter 9 result that a black hole's entropy is its horizon area in Planck units: a horizon-filling black hole of radius ≈1061 Planck lengths has area ≈10122, and the number of states is the exponential of the entropy.
- Fields versus particles. Greene counts particle arrangements, but a field description gives the same count: particles are a field's quanta, like molecules in an ocean wave.
- Credit. Duplicate regions in an infinite classical universe were studied by George Ellis and G. Brundrit (1979); the quantum version by Jaume Garriga and Alexander Vilenkin (2001).
Critical lens on the Quilted argument.
- Infinity is assumed, not observed. Flatness to within measurement error is compatible with a finite torus or with a very large but finite sphere. The doppelgänger conclusion needs actual infinity.
- "Random distribution" across patches is an assumption. Greene invokes inflation's support for it; if initial conditions were correlated across space in some special way, the frequency estimates would change.
- Reductionism about persons is doing heavy lifting. That a particle-identical copy is "you" in every meaningful sense is a philosophical commitment, which Greene, to his credit, flags openly.
- Infinity breeds measure problems. When everything happens infinitely often, assigning relative frequencies is subtle, a problem Chapter 7 confronts directly.
2026 update: the numbers since 2011
- Composition (Planck 2018 + later): dark energy ≈ 68–69%, total matter ≈ 31% (ordinary ≈ 5%, dark ≈ 26%). Greene's 73/27 split reflects WMAP-era fits.
- Age: 13.79 ± 0.02 billion years (Planck 2018).
- Flatness: Planck plus baryon acoustic oscillations give Ωk = 0.001 ± 0.002: flat to about 0.2%.
- Observable radius: usually quoted as ≈ 46 billion light-years (diameter ≈ 93 Gly) with modern parameters; Greene's 41 Gly is an older estimate.
- Topology: searches for matched "circles in the sky" in Planck CMB data found none, ruling out a small torus or dodecahedral universe; any finite universe must be larger than our horizon.
- Dark energy may evolve: DESI's baryon acoustic oscillation results (DR1, 2024; DR2, March 2025), combined with CMB and supernova data, prefer a dark energy that weakens with time at about 2.8–4.2σ depending on supernova sample. If confirmed, Greene's "fixed or varying?" question tilts toward varying. Not yet a 5σ discovery.
- Hubble tension: local distance-ladder measurements (SH0ES, H0 ≈ 73 km/s/Mpc) disagree with the CMB-inferred ≈ 67–68, at about 5σ. JWST has confirmed the ladder's Cepheid photometry, so the cause is still open.
So what. The chapter's real lesson is a reasoning pattern: finite state space × unbounded trials ⇒ guaranteed repetition (the pigeonhole principle at cosmic scale). In engineering it is why hash collisions are certain at scale, why monitoring rules eventually fire falsely, and why any finite-capacity model evaluated over enough cases will meet "duplicates" of its training data. The quantum "fly" adds a second idea: distinctions below the resolution limit are not real distinctions, a useful discipline when deciding how finely to segment customers or risk classes.
# Pigeonhole at cosmic scale: how often must configurations repeat?
import math, matplotlib.pyplot as plt
print("Imelda:", 500*1000, "outfits ->", round(500*1000/365), "years")
print("Randy: 52! =", f"{math.factorial(52):.3e}")
# log10 of state count for a region of radius R (Planck lengths): area ~ R^2
R = [10**k for k in range(10, 62, 5)]
logN = [0.434 * r**2 for r in R] # log10(e^(R^2))
plt.loglog(R, logN, 'o-', color='#2563EB')
plt.xlabel('region radius (Planck lengths)'); plt.ylabel('log10(#states)')
plt.title('Holographic cap: states grow with AREA'); plt.show()
Eternity and Infinity: The Inflationary Multiverse
Inflationary cosmology was invented to explain why the cosmic microwave background is uniform to four decimal places; in most of its versions it never stops, and so it manufactures an endless supply of "bubble universes". Greene follows the story from Gamow's forgotten prediction of the relic radiation to Guth's repulsive-gravity burst, then shows why quantum jitters make inflation eternal, how bubbles can carry different values of Higgs-like fields (and so different physics under the same laws), and why each bubble, though finite from outside, is infinite from inside, which plants a Quilted Multiverse in every bubble.
Relics of a hot beginning
In the mid-1900s physicists realised that with the sun, the stars and every galaxy removed, space would still glow, not visibly, but in microwaves, in every direction. The glow's origin is the origin: it is a relic of the universe's creation. Its discovery was a triumph for the big bang, but in time it exposed one of the theory's deepest shortcomings and set up the next revolution after Friedmann and Lemaître: inflation.
- George Gamow: hulking six-foot-three Russian physicist, student of Friedmann in Leningrad. In 1932 he and his wife tried to defect by kayak across the Black Sea, stocked with chocolate and brandy; bad weather forced them back and Gamow talked his way out with a story about failed scientific experiments at sea. He later left "on dry land, with less chocolate."
- Ralph Alpher: Gamow's "phenomenally talented" graduate student; with Robert Herman he refined the prediction of relic radiation in the late 1940s.
- Hans Bethe: added to an Alpher-Gamow paper purely so the byline read "Alpher, Bethe, Gamow" (α, β, γ), a joke that made some colleagues take Gamow less seriously.
- Robert Dicke and Jim Peebles (Princeton): independently re-derived the relic radiation in the early 1960s; Dicke, a renowned experimentalist, set out to detect it with students David Wilkinson and Peter Roll.
- Arno Penzias and Robert Wilson (Bell Labs): found it by accident; Nobel Prize 1978.
The Gamow-Alpher picture (modernised)
- First three minutes: a stupendously hot, dense plasma expands and cools; the universe acts as a cosmic nuclear furnace, forging hydrogen, helium and traces of lithium.
- A few minutes later: temperature falls to about 108 K (≈ 10,000 × the sun's surface), too cool for further fusion. For eons, little happens but expansion and cooling.
- About 370,000 years: at about 3,000 K (half the sun's surface temperature), electrons and nuclei combine into neutral atoms. Before this, charged particles kept jostling photons, making the plasma opaque, like high beams in dense fog. Neutral atoms let photons slip through "like a hot knife through butter": the fog clears and the big bang's light is released.
- Since then: the photons keep travelling at light speed, but expansion lowers their frequency: violet to blue, green, yellow, red, infrared, microwave, and eventually radio. Calculations put them just above absolute zero: the cosmic microwave background (CMB).
Greene reread the late-1940s papers and calls them "marvels of theoretical physics" needing barely more than undergraduate physics. Yet they were ignored: physics was dominated by quantum and nuclear work, cosmology was not yet a quantitative science, and Gamow's playful style cost him credibility. No astronomer would search.
The most famous phone call in science. Bell Labs' Penzias and Wilson, under thirty miles from Princeton, could not silence a steady hiss in a radio antenna (built, coincidentally, to a Dicke design from the 1940s). A chain of conversation solved it: Peebles gave a talk at Johns Hopkins in February 1965; Carnegie radio astronomer Kenneth Turner told Bernard Burke of MIT, who was in touch with Penzias. Penzias and Wilson called Dicke, who confirmed they had tapped into the reverberation of the big bang. The two teams published side by side in the Astrophysical Journal: Princeton gave the cosmology; Bell Labs reported, in the most conservative language and with no mention of cosmology, a uniform microwave signal. Neither paper cited Gamow, Alpher or Herman, who spent years fighting for belated recognition.
A misplaced university. Greene places Gamow at Washington University in St. Louis in the 1940s. Gamow was in fact at George Washington University in Washington, D.C. (1934 to 1956), where the Alpher and Herman work was done (Alpher and Herman also worked at the Johns Hopkins Applied Physics Laboratory).
The uncanny uniformity of ancient photons
Cosmology is the one field that can witness its own history directly. Starlight seen by eye left its source years or millennia ago; telescopes catch light billions of years old. Because the universe is uniform on large scales, what happened there then is, on average, what happened here. "In looking up, we are looking back." The CMB photons are the oldest light we can ever see, since earlier photons were trapped in the fog; they show the universe as it was nearly 14 billion years ago.
- About 400 million CMB photons in every cubic metre of space.
- About 1% of the "snow" on an old analog TV tuned between stations is big bang radiation, sharing the airwaves with reruns of All in the Family and The Honeymooners.
- Temperature: 2.725 K in every direction, matching across the sky to better than four decimal places.
- Three hundred years of science took us from rudimentary telescopes and balls dropped from leaning towers to processes just after the universe's birth.
The cosmological principle predicts this uniformity, but it was an assumption adopted for mathematical convenience. The CMB confirms it and thereby shines a glaring spotlight on it: what mechanism made the cosmos uniform?
Faster than light: the horizon problem
A hot hand and a cold hand equalise when held; coffee cools to room temperature. Equalising temperatures requires contact, or at least an exchange of information; a thermos works by preventing it. Now take two points at opposite edges of the night sky, so far away that the first light from each is only now reaching us. Light from one has still not reached the other. They are beyond each other's horizons and have never interacted. Finding them at the same temperature to four decimal places is like finding two uncontacted civilisations speaking the same language with the same books in their libraries.
Greene confesses he was then "puzzled by the puzzle": if everything was once close together, how did regions get farther apart than light could cross? The answer is under-stressed in popular accounts: light's speed limit applies to motion through space. Galaxies don't have jet engines; they recede because space itself swells and drags them along. Relativity places no limit on how fast space can swell, so recession speeds can exceed light speed. In the early universe, standard big bang expansion drove regions apart faster than light, so they could never have equilibrated. This is the horizon problem.
Why "closer together in the past" does not help in the standard big bang. Expansion there decelerates, like a ball thrown upward that covers its first half-height in 2 s and the second in 4 s. At the halfway point in time, two regions are already more than halfway to their present separation: half the time to signal across more than half the distance. Going back in time makes communication harder, not easier. Inflation reverses this: during accelerating expansion, regions at the halfway time are less than halfway apart, so earlier times give proportionally more time for contact.
Broadening horizons: the inflationary fix
In 1979 Alan Guth (then at the Stanford Linear Accelerator Center) proposed inflationary cosmology; it was critically refined by Andrei Linde (Lebedev Physical Institute, Moscow) and by Paul Steinhardt and Andreas Albrecht (University of Pennsylvania). The outline: slow the early separation of regions so they have time for "cosmic handshakes" and reach one temperature, then insert a brief burst of enormously fast, ever-quickening expansion (inflationary expansion) that more than makes up for the slow start, flinging the regions to their present distant positions.
But physicists don't dictate expansion; Einstein's equations do. A Newton given a five-minute primer on relativity (one minute spent on baseball) would pound the table: gravity attracts, so outward expansion can only slow. He'd be wrong, and the fault would lie with the primer.
Pressure gravitates
In Newton's theory gravity comes only from mass. In Einstein's it comes from mass, energy and pressure. A squeezed bag of potato chips weighs fractionally more than an unsqueezed one, though its mass is unchanged. Positive pressure (air pushing out) adds to attractive gravity. Negative pressure (tension, like a stretched rubber band pulling in) does the opposite: it produces repulsive gravity. In the Friedmann acceleration equation the source is ρ + 3p: if pressure is sufficiently negative (p < −ρ/3), expansion accelerates.
A subtle point: uniform pressure, of either sign, exerts no push or pull at all (your eardrums pop only when pressure differs side to side). The repulsion comes from the gravity generated by the presence of uniform negative pressure. This also explains Einstein's cosmological constant: it fills space with uniform energy and uniform negative pressure, hence repulsive gravity. Einstein used it as a gentle, steady push for a static cosmos; inflation's pioneers wanted something different "as the finale of Mahler's Eighth differs from the drone of a tuning fork": a gargantuan, astoundingly short, thunderingly intense surge.
The remaining question, Newton's second objection: what could supply a fleeting, pervasive negative pressure? The pioneers' answer: quantum fields.
Quantum fields
- Newton described tangible bodies: stones, cannonballs, planets.
- Michael Faraday (early 1800s) introduced the field. A fridge magnet lifts a paper clip without touching it because it produces something that does touch it: a magnetic field, an invisible "mist" that fills space and does the magnet's bidding. Charges produce electric fields (the doorknob shock on a wool carpet). Faraday found changing electric fields generate magnetic fields and vice versa.
- James Clerk Maxwell (late 1800s) described fields as numbers assigned to each point of space: large magnetic values inside an MRI cavity, large electric values inside a thundercloud. His equations govern electromagnetic waves (phone, radio, Wi-Fi) and revealed that visible light is an electromagnetic wave.
- Quantum field theory (second half of the twentieth century) fused fields with quantum mechanics and now underlies our best theories of matter and force: strong and weak nuclear fields, electron, quark and neutrino fields, and one wholly hypothetical field, the inflaton (named, by the "-on" convention of electron, proton and muon, with the second "i" dropped).
Why a uniform field has negative pressure: the champagne argument
Uncorking Dom Pérignon. CO2 pushes the cork out (positive pressure). Doing that work drains energy from the champagne: the gas cools, and the vapour tendrils at the neck are condensation.
Uncorking a bottle filled with a uniform field. Each bit of new volume the cork exposes fills with field carrying the same energy per unit volume, so removing the cork increases the bottle's energy. That energy can only come from your muscles, which means you had to pull against the contents. A uniform field sucks inward: negative pressure. The same reasoning applies to empty space with a cosmological constant.
Guth's insight: a field with a uniform value fills space with uniform energy and uniform negative pressure, and therefore with repulsive gravity. Plugging in plausible early-universe values, the repulsion is many orders of magnitude stronger than Einstein's constant. The indispensable bonus is that a field can change while a constant cannot. As you change the field inside a microwave oven or MRI by flipping a switch, the inflaton can turn on for a burst and then turn off. Observations demand exactly this: blistering growth billions of years ago, then the statelier expansion astronomers measure. Repulsive gravity must be transient.
Cartman on the mountain: how inflation turns off
Picture Eric Cartman of South Park perched on a snowy peak: he holds potential energy, released as kinetic energy when he tumbles. Systems with pent-up energy release it whenever they can: "things fall." A field's energy as a function of its value is its potential energy curve.
- On the plateau: the inflaton, uniform across a region and high on a gently sloping plateau, fills space with large energy and negative pressure: inflation.
- Rolling down: as its value everywhere rolls to lower numbers, the energy and negative pressure dissipate and inflation ends.
- Reheating: the released energy is not lost; like steam condensing into droplets, it condenses into a uniform bath of particles, the raw material of stars and galaxies.
- Duration of the roll: about 10−35 s in typical versions.
- Expansion factor: about 1030, perhaps more.
- Pea-sized region → larger than the observable universe, in a time that the blink of an eye overestimates by a factor larger than a million billion billion billion.
- The region that became our observable universe was so small it easily reached a uniform temperature first. In inflation, uniformity is inevitable.
Eternal inflation
Inflation is a framework, not a specific theory: models differ in the number of inflaton fields, their potential curves and more. But they share consequences, and one, first fully realised by Alexander Vilenkin (Tufts) and developed by Linde and others, is the reason for this chapter: in many versions inflation is not a one-time event. Our region's history (rapid stretching, transition to slower expansion, particle production) happens over and over at far-flung locations, forever.
Two complications the single-Cartman picture hid:
- The field has a value at every point: picture numerous Cartman clones on closely packed identical peaks.
- Quantum jitters: the field's value fluctuates randomly, and the higher its energy, the bigger the jitters. During inflation they are large and dominant: random tremors, strong here, weak there, sometimes knocking a descending Cartman back up.
The terrain divides into domains, like US states: in some all the Cartmans have fallen, in others many remain perched. But here fields and Cartmans differ: a perched field drives fantastic expansion of its own region. Two processes compete:
| Process | Effect on high-energy (inflating) volume |
|---|---|
| Quantum jitters knock the field off its perch | Decrease |
| Inflationary expansion enlarges still-perched domains | Increase |
In cosmologically successful models, the increase wins: a field that falls off too quickly produces too little inflation to solve the horizon problem. So the inflating volume grows forever: once inflation starts, it never ends. Greene's analogy is a viral pandemic: to eradicate a virus you must destroy it faster than it reproduces, and inflation reproduces faster than it is eliminated.
Swiss cheese and the cosmos
Picture the universe as a block of Swiss cheese: the cheese is space where the inflaton is high and still inflating; the holes are regions like ours that have exited inflation and converted the field's energy into particles, later galaxies, stars and planets. Holes keep appearing as jitters knock the field down at random spots; the cheese keeps stretching. Result: an ever-expanding block riddled with an ever-growing number of holes. Each hole is a bubble universe (or pocket universe). "Bubble" sounds small, but each is as much a universe as ours: real, gigantic and dynamic.
Clever theorists can build non-eternal models by fiddling with the number of fields and their curves, but they are the exception. If inflation is right, and its physically relevant version is eternal, the Inflationary Multiverse is inevitable.
Changing perspectives: from shelved idea to serious contender
Guth falls asleep. In the 1980s Vilenkin excitedly visited Guth at MIT to explain eternal inflation. Midway, Guth's head drooped: he was asleep. Not necessarily a bad sign: Guth famously dozes in seminars (including Greene's) and wakes to ask the most insightful question. But the community was no keener, and Vilenkin shelved the idea. Then the evidence for inflation itself was thin; a multiverse built on it looked like "speculation piled upon speculation." Precision CMB data changed that.
The fingerprint: stretched quantum jitters
Inflation's pioneers realised it would not make the CMB perfectly smooth. As the inflaton rolled toward the end of inflation in our bubble, its value still jittered, a little higher here, a little lower there, like the waves in a sheet settling onto a mattress. Ordinarily these quantum variations are microscopic and irrelevant. But inflation stretches the microscopic to the macroscopic, as a tiny note on a deflated balloon becomes legible when the balloon is blown up. The energy differences become temperature variations of about a thousandth of a degree: 2.7245 K here, 2.7255 K there.
- 1948: Alpher, Bethe and Gamow; Alpher and Herman predict relic radiation of a few kelvin.
- 1965: Penzias and Wilson detect the CMB; Dicke, Peebles, Roll and Wilkinson interpret it.
- 1979 to 1982: Guth proposes inflation; Linde, Albrecht and Steinhardt create "new inflation"; the calculation of quantum ripples is developed by Mukhanov, Chibisov, Hawking, Starobinsky, Guth, Pi, Bardeen, Steinhardt and Turner.
- 1983 to 1986: Vilenkin, then Linde, develop eternal inflation.
- Early 1990s: COBE (over 1,000 researchers, led by John Mather and George Smoot) detects the variations; Nobel Prize 2006.
- 2000s: each more precise measurement (WMAP and others) confirms the predicted pattern.
Observers found the variations, about a thousandth of a degree, and their pattern across angular scales matches theory "spot-on": the temperature-variation curve rises to a dominant peak and then a series of smaller ones, with data points sitting on the theoretical line.
We should be grateful for the blotches. Slightly denser regions pulled in more matter, grew, pulled harder, and in a snowball effect over billions of years became galaxies and stars. Galaxies, stars, planets and life itself derive from microscopic quantum uncertainty amplified by inflationary expansion.
Going deeper: the technical notes
- Horizon problem, quantitatively. With deceleration, regions separate fast early and slowly later, so the ratio "time available / distance to cover" shrinks toward the past. Inflation inserts accelerated expansion: compared with the standard big bang, today's opposite-sky regions were much closer and separating more slowly at a given early moment, then were flung apart. Both statements ("inflation is faster" and "inflation is slower") are true, depending on whether you compare regions of fixed early separation or of fixed present separation.
- Pressure that collapses stars. Because pressure gravitates, a dying star that needs more outward pressure to resist contraction also generates more inward gravity. Past a threshold the remedy feeds the disease and collapse to a black hole is unavoidable.
- Chaotic inflation (Linde). A plateau is not required: even a simple parabolic potential inflates if the field starts high, which the hot, chaotic early universe could arrange. (Observations since have disfavoured the simple parabola; see the 2026 box.)
- Reheating. Inflation supercools space; as it ends, the inflaton oscillates about its minimum and dumps energy into particles, which "re-heats" the universe. The uniformity set during inflation makes reheating uniform too.
- Who found eternal inflation. Guth knew inflation could be eternal; Steinhardt wrote down a mathematical realisation in specific contexts; Vilenkin stated it in full generality; Linde developed it further.
- Jitters grow with expansion. The faster space expands, the more violently the inflaton jitters, a back-reaction that makes high-energy inflation self-perpetuating.
- Where inflation's energy comes from. As the inflating region grows, its total energy grows. The "muscles" pulling the cork are gravity's: gravitational energy can become arbitrarily negative, so gravity is "a bank with a bottomless credit line" (Chapter 10 returns to this).
- Reading the CMB plot. The temperature on the sky is expanded in spherical harmonics Yℓm; the plotted quantity is the size of each multipole's coefficient, with higher ℓ meaning smaller angles.
- Gravitational potential, not force, slows time. Inside a hollow at a star's centre you feel no force yet your clock still runs slow: what matters is the depth of the potential well.
- Bubble interiors are infinite: credit. The inside-outside result was articulated by Vilenkin and by Coleman and De Luccia.
- Why patches are random. The Quilted argument assumed particle arrangements vary randomly from patch to patch. Inflation supplies the reason: the inflaton's quantum jitters during conversion to particles randomise the arrangements, as they randomise the CMB spots.
Inflation's soft underbelly, as Greene concedes: the inflaton is hypothetical; its potential curve is posited, not observed; the field must somehow start high on its curve across a region. Agreement with data convinces many that inflation captures "a deep truth", and since many versions are eternal, theory plus data make an "indirect yet compelling case" for the Inflationary Multiverse. Critics, notably Paul Steinhardt (one of inflation's own architects), argue the reverse: because eternal inflation produces every possible outcome somewhere, it predicts nothing and so cannot be said to be confirmed by any particular observation (the "measure problem", Chapter 7).
Experiencing the Inflationary Multiverse
| Feature | Quilted Multiverse | Inflationary Multiverse |
|---|---|---|
| Separation between universes | No sharp divide: one continuous space | Sharply divided bubbles, separated by still-inflating space |
| Can they ever meet? | Horizons eventually overlap | No: intervening space recedes faster than light; no signal can cross |
| Laws | Same | Same fundamental laws |
| Observable physics | Same qualitative features; differences in detail | Can differ profoundly because fields settle differently |
Because every bubble results from the same process, all share the same fundamental laws. But like identical twins raised apart, identical laws can play out very differently in different environments.
The magnetic universe. Imagine a bubble with galaxies and planets but permeated by an unswitchable magnetic field thousands of times stronger than any MRI. Iron objects fly off along the field; particle, atomic and molecular properties shift; cells cannot function, so life as we know it cannot arise. Yet its fundamental laws are ours; the differences are purely environmental, and its scientists would eventually tease that out and find our equations.
The Higgs field: an environmental variable for mass
The Standard Model of particle physics posits that we are immersed in a Higgs field, proposed in the 1960s by Peter Higgs with Robert Brout, François Englert, Gerald Guralnik, Carl Hagen and Tom Kibble. Unlike a magnetic field it does not speed particles up, slow them down or bend their paths. Instead, fundamental particles acquire their mass by "rubbing" against this molasses-like field: the resistance you feel pushing an electron or quark is its mass. Remove the Higgs field and particles become massless; double it and they double in mass. (For composites like protons and neutrons, most mass comes from the energy of gluons binding the quarks, not from the Higgs.)
Most important, the Higgs field could have different values in different bubbles. Its potential curve typically has troughs away from zero. As two young bubbles cool, the Higgs in ours rolls into, say, the left trough, giving the particle properties we measure, while in the other it settles in the right trough. Same laws, different masses and forces. Theories with multiple Higgs fields (natural in string theory) have potential surfaces with many troughs and so many more kinds of bubbles.
Why small changes in particle properties are catastrophic:
- Heavier electron (a few times): electrons and protons merge into neutrons, so little hydrogen forms.
- Forces are carried by particles: heavier carriers move sluggishly, so the force has shorter range.
- Atoms depend on the electromagnetic and nuclear forces: modify them substantially and atoms fall apart or never form.
- So many bubbles are lethal: our universe may be "an island oasis in a gigantic but largely inhospitable cosmic archipelago."
Universes in a nutshell: inflation begets quilts
The two multiverses look unrelated (one needs infinite space, the other eternal inflation), but inflationary universes generate quilted ones, through time.
Relativity makes time fluid: observers moving relative to each other disagree on simultaneity; an hour near a black hole is an age far away. So what does "the universe is 14 billion years old" mean? It uses a particular clock. Every clock measures change in some system: hands on a dial, the sun's position, the decay of carbon-14 to nitrogen. On cosmic scales the best clock is the steadily falling average mass density: since it falls uniformly everywhere, calibrating watches to it (recalibrating after black-hole trips or near-light-speed jaunts) synchronises the whole bubble. In the earliest era, before matter, the clock is the inflaton's energy density, hence its value. Two locations are at the same moment when they have the same inflaton value, as two trees with the same number of rings are the same age.
Trixie and Norton. Trixie floats outside, in still-inflating space, watching a bubble form. Tracking the inflaton along a line across the bubble, she sees the low-value region grow ever larger with her time, but at any of her moments the bubble is finite.
Norton lives inside and keeps time by the inflaton value: he invites dinner guests to arrive "when the inflaton reads 60", and everyone arrives together. To draw "all of space at one moment" he paints by numbers, connecting all points with equal inflaton value. Each such curve extends indefinitely: inside, space at every moment is infinite.
"What appears as endless time to an outsider appears as endless space, at each moment of time, to an insider." In Hamlet's words, a bubble "could be bounded in a nutshell" and count itself "a king of infinite space." Each bubble is finite from outside and infinite from inside, so if the Inflationary Multiverse is real, we belong to a Quilted Multiverse too.
Greene first found the inflationary multiverse more plausible: it solves long-standing puzzles and matches data, while the quilted version seemed avoidable if space is finite. The inside-outside analysis removes the escape: "inflationary parallel universes beget quilted ones." The best theory for the best data suggests we live in one bubble of a vast inflationary multiverse, each bubble holding its own infinite quilt: "not only parallel universes but parallel parallel universes."
2026 update: inflation after Planck, BICEP/Keck and ACT
- Higgs boson found: ATLAS and CMS announced it on 4 July 2012 at ≈125 GeV; Englert and Higgs shared the 2013 Nobel Prize. Greene's "still up in the air" is settled; the Higgs field is real. The measured Higgs and top-quark masses even suggest our electroweak vacuum is metastable: not in the deepest trough, but with a lifetime vastly longer than the universe's age.
- Spectral tilt: Planck 2018 measured ns = 0.965 ± 0.004, a slight "red" tilt that slow-roll inflation predicted. ACT DR6 (2025), combined with DESI, pulls it up to about 0.974, creating mild tension with the simplest plateau models.
- Primordial gravitational waves not yet seen: BICEP/Keck (2021) bound the tensor-to-scalar ratio at r < 0.036, excluding simple m2φ2 and φ4 inflation and favouring plateau potentials (Starobinsky R2, Higgs inflation): Greene's "gentle plateau" picture survived. The BICEP2 claim of 2014 was retracted as galactic dust. Next: Simons Observatory (observing since 2024), LiteBIRD (planned 2030s).
- Peebles honoured: Jim Peebles won the 2019 Nobel Prize for theoretical cosmology.
- The critique sharpened: Ijjas, Steinhardt and Loeb (2017) argued inflation is "not testable" in its eternal form; 33 leading physicists, Guth and Linde among them, replied in defence. The debate over the measure problem is unresolved.
- CMB temperature: now quoted as 2.7255 K.
So what. Two transferable ideas. (1) Same rules, different equilibria: identical laws plus different settled "field values" yield radically different worlds. Organisations running one policy framework across business units see the same thing; the local parameters, not the rules, determine outcomes. (2) Clocks are conventions: "simultaneous" in a distributed system means "same value of some agreed counter" (Lamport clocks, event sequence numbers), exactly Norton's inflaton watch. Choose the counter and you choose what "the whole system at one moment" means.
Unifying Nature's Laws: On the Road to String Theory
Three of the coming multiverses grow out of string theory, so Greene pauses to give an honest status report on it. He shows why gravity and quantum mechanics are mutually incompatible in standard quantum field theory, how replacing point particles with vibrating strings in ten dimensions cures that, and why the theory still makes no definitive prediction: the equations allow an enormous variety of shapes for the extra dimensions, and the shape determines the particle physics. His scorecard rates the theory "excellent" on quantum gravity, unification, black-hole entropy, singularities and mathematics, and "indeterminate" on particle properties and experiment.
Einstein's unfinished dream
General relativity rebuilt space and time and fathered modern cosmology. Einstein wanted more: a unified field theory that stitches all of nature's forces into one coherent mathematical framework. History judges his decades on it as having "little lasting impact" (the dream was noble, the timing early), but others carried it forward, most ambitiously in string theory. Since Greene's The Elegant Universe and The Fabric of the Cosmos, the theory had faced a spate of public questioning, which he calls "completely reasonable": it has no definitive, testable prediction.
A brief history of unification
- 1860s: James Clerk Maxwell fuses electricity (current in a wire), magnetism (a child's magnet) and light (sunlight) into one electromagnetic trinity: currents make magnetic fields, moving magnets make currents, rippling fields are light. Einstein called it an archetypal contribution to human thought.
- 1919 to 1921: Theodor Kaluza unifies gravity and electromagnetism using a fourth spatial dimension; Einstein sits on the paper for two years, then champions it.
- 1926: Oskar Klein explains why the extra dimension is unseen: it is tiny and curled up.
- Mid-1920s to mid-1960s: physics is consumed by the atom; unification is of "passing interest."
- 17 April 1955: in Princeton Hospital, Einstein asks his secretary Helen Dukas for his pages of unified-field equations. He dies the next morning; his last scribblings shed no further light. Work on unification halts.
- Late 1960s to 1970s: quantum field theory captures the weak and strong nuclear forces. Sheldon Glashow, Steven Weinberg and Abdus Salam (Nobel) unify electromagnetism and the weak force; Glashow and Howard Georgi propose a grand unification of electromagnetic, weak and strong forces.
- Mid-1980s: superstring theory reconciles general relativity and quantum mechanics; thousands of journal pages follow.
- Mid-1990s: refinements beyond approximate methods unexpectedly thrust string theory into the multiverse story.
Einstein's programme failed for two further reasons: he downplayed quantum physics (believing the unified theory would supersede it), and he ignored two forces: the strong nuclear force (the glue of atomic nuclei) and the weak nuclear force (responsible for radioactive decay). Unification meant combining four forces, not two.
Unification turned from aesthetic preference into logical necessity. When quantum field theory was applied to gravity, calculations returned mathematical gibberish. "If the laws you have prove mutually incompatible, then clearly the laws you have are not the right laws."
Quantum fields redux
Quantum field theory adds two features to classical fields:
- Quantum jitter: a field's value at every point fluctuates randomly (as with the inflaton).
- Quanta: a field is composed of particles, as water is of H2O. A light bulb emits about 100 billion billion photons per second. Electrons are quanta of the electron field; quarks of the quark field. Physicists loosely picture particles as "knots" of field, but the mathematics treats them as points with no extent and no internal structure.
Not a single experimental result contradicts quantum field theory. Its crown jewel, quantum electrodynamics (QED), predicts the electron's magnetic properties in calculations taking decades, matching experiment to ten decimal places. By the late 1970s the weak and strong forces also fitted within it.
Why gravity breaks the machinery
Combine general relativity with quantum theory and compute, say, the chance of two electrons ricocheting given both their electric repulsion and gravitational attraction: the answer is infinity. A probability must lie between 0 and 1; infinity is not "very likely", it is meaningless, "like speaking of the thirteenth egg in an even dozen."
The San Francisco landlord. The three nongravitational fields are tenants in the house of spacetime: their quantum jitters are like raucous parties, a nuisance a generation of theorists learned to handle by the 1970s. Gravity's jitters are an earthquake: because the gravitational field is the fabric of spacetime, its fluctuations shake the whole structure, and the mathematical methods collapse.
The conflict was ignored because it matters only where things are both very massive and very small: the big bang and the centres of black holes, exactly where we most want answers.
- Planck mass: about 1019 proton masses, squeezed into about 10−99 cm3.
- Planck length: about 10−33 cm, where gravity and quantum mechanics collide.
- Gap to experiment: the Planck scale is some 100 billion billion (1020) times smaller than anything probed, and more than a million billion times beyond the reach of the most powerful accelerators.
- Picture it: magnify an atom to the size of the observable universe and the Planck length becomes the size of an average tree.
String theory in a nutshell
Particles are not dots but tiny vibrating filaments. Look inside an electron or a quark and you would find a string. The strings are identical; only their vibration patterns differ. An electron is lighter than a quark, so its string vibrates less energetically (E = mc2); its larger electric charge corresponds to subtler pattern differences. As different vibrations of a guitar string make different notes, different vibrations of a fundamental string make different particles. In fact the string doesn't just set the particle's properties: it is the particle.
Strings are about Planck-length sized. The Large Hadron Collider, smashing particles at a little more than 10 trillion times a proton's rest energy, resolves about 10−19 cm (a millionth of a billionth of a hair's width), still far too coarse. As Earth looks like a dot from Pluto, strings look like points to our best machines.
Three overarching points
Why string theory excites physicists:
- Particles are derived, not inserted. A quantum field theory must choose its fields: the Standard Model contains fifty-seven distinct quantum fields (electron, neutrino, photon, up, down, charm quark and more), an "ungainly assortment". In string theory each vibration pattern is a particle species, so enumerating the patterns would explain all properties of all particles, including any yet undiscovered. It seeks a complete description from the get-go.
- Gravity is built in. Pre-string research had established that a graviton must be massless, chargeless and spin-2 (spinning, roughly, twice as fast as a photon). John Schwarz, Joël Scherk and, independently, Tamiaki Yoneya found a string vibration with precisely those properties. When Schwarz and Michael Green showed in the mid-1980s that the theory is quantum-mechanically consistent, string theory became a quantum theory of gravity: its single most important achievement. Its extended strings smear out the violent short-distance jitters that killed point-particle attempts.
- It embraces its predecessors. Successful theories don't make old ones obsolete; they reduce to them in familiar regimes. Apply string mathematics to the large and massive (the sun) and out pop Einstein's equations; apply it to low-energy strings and it becomes quantum field theory. "Sufficiently revolutionary to transcend the barriers... Sufficiently conservative to allow the past three hundred years of discovery to snugly fit within its mathematics."
The dimensions of space
Early string theory suffered quantum anomalies, inconsistencies that let energy be spontaneously created or destroyed. Normally that kills a theory, and in the 1970s many thought strings should die. The few who persisted found the anomalies depend on the number of dimensions: with nine dimensions of space and one of time (ten in all), they vanish.
The "mathematical skinny": an equation in string theory contains a term (D − 10) × (Trouble), where D is the number of spacetime dimensions and Trouble produces physical nonsense such as violated energy conservation. Set D = 10 and the term is 0 × Trouble = 0. Greene admits he has never met anyone who can explain why the factor takes this form without the calculation itself: "That's how the math plays out. Really."
Hiding extra dimensions: Kaluza-Klein
Dimensions don't go missing like car keys or a sock, but they can be tiny and curled up:
- The tall straw. A straw as thin as usual but as tall as the Empire State Building, seen from across the Hudson River, looks like a one-dimensional line, though its surface has a second, circular dimension at every point.
- The salt-flat carpet. A carpet covering Utah's salt flats looks two-dimensional from a plane; up close each point carries a tiny loop of pile, a third, small dimension.
- The proposal. Space itself may have, at every point of the three big dimensions, extra dimensions curled up millions or billions of times smaller than an atom: everywhere, yet undetectable.
Kaluza's aria. Kaluza found that with a single stroke of the pen he could rewrite Einstein's equations for one extra spatial dimension. Inside the result were Einstein's gravity equations plus one more, which was Maxwell's equation for electromagnetism. His son recounted that the normally reserved Kaluza pounded his desk, leapt up and sang an aria from The Marriage of Figaro. Gravity ripples through the familiar three dimensions; electromagnetism ripples through the fourth. Klein supplied the reason we don't see it. Einstein, initially torn, held up publication for two years, then became a strong champion of hidden dimensions.
The programme stalled because it could not incorporate the detailed properties of matter particles such as electrons; by the mid-1940s it was largely abandoned. Thirty years later string theory required extra dimensions, and Kaluza-Klein provided a ready-made answer to "why haven't we seen them?": they're all around us, just too small. By the mid-1980s enthusiasts thought a complete theory of all matter and forces was a short time away.
Great expectations, sober assessment
The early pace was compared to the 1920s quantum gold rush. Some promised swift solutions to everything: merging gravity and quantum mechanics, unifying forces, explaining matter, fixing the number of dimensions, resolving black-hole singularities, explaining the origin of the universe. Seasoned researchers expected otherwise, and they were right: nearly three decades later, the theory is "but partway down the road", and with quantum gravity some 1020 times beyond experiment, "the road will be long."
String theory and the properties of particles
Why does the electron have its particular mass, and the up quark its particular charge? Had the electron been moderately heavier or lighter, or electric repulsion stronger or weaker, stellar nuclear processes would be disrupted: no stars, no sunlight, no life. The grand challenge is to calculate particle properties with pen, paper and computer and match the measured values.
- Quantum field theory cannot, permanently. Masses and charges are inputs that define the theory; a world with a different electron mass is just a different parameter.
- String theory, in principle, can. Strings vibrate into the curled-up dimensions too, and as the shape of a wind instrument dictates its air's vibrations, the size and shape of the extra dimensions dictate string vibrations, hence particle properties.
- The shapes: the equations require Calabi-Yau shapes (manifolds), after mathematicians Eugenio Calabi and Shing-Tung Yau, who studied them long before physics needed them.
The catch: there is no unique Calabi-Yau shape. Like musical instruments they come in a huge variety, each giving different particle properties, and nothing yet singles out "the one". This is the main obstacle to definitive prediction.
Greene's own catalogue. When he began in the mid-1980s, only a handful of Calabi-Yau shapes were known, so one could study each in search of real-world physics; his doctoral dissertation was an early step. As a postdoc working for Yau, the count had grown to a few thousand ("that's what graduate students are for"). It has since grown "more numerous than grains of sand on a beach. Every beach. Everywhere. By a long shot." Exhaustive analysis is out of the question.
So on particle properties string theory so far offers no improvement on quantum field theory. Greene's defence: the theory's claim to fame is quantum gravity, and achieving that alone would be monumental; no theory is required to explain particle masses. And Chapter 6 will suggest that hoping for a unique shape may be "plainly wrongheaded": as many forms of DNA give life's variety, many extra-dimensional shapes may give a multiverse's variety.
String theory and experiment
Probing a string's extended structure needs an accelerator a million billion times more powerful than the LHC: with known technology, galaxy-sized, using enough energy each second to power the world for a millennium. At accessible energies string theory impersonates quantum field theory. That is fine for most physics, since QFT takes measured properties as input and predicts other results superbly. But choosing fields in QFT corresponds to choosing the extra-dimensional shape in string theory, and the map from shape to particle properties is so tangled that working backwards from data is not yet possible.
The realistic strategy is circumstantial evidence: results explicable by other means but far more natural in string theory. Greene's analogy: you could theorise he types with his toes; fingers is the more natural hypothesis (and correct).
| Probe | Idea | What a signal would mean |
|---|---|---|
| Supersymmetry (LHC) | The "super" in superstring: every known particle has a heavier partner with the same force charges | Confirms a piece of string theory, though SUSY now lives in non-string models too; not a smoking gun |
| Short-range gravity | Extra dimensions dilute gravity like ink in a vat, explaining its weakness (your arm beats the whole earth lifting a coffee cup); below the extra-dimension size gravity should strengthen | No deviation found down to about a micron (Greene's figure) |
| Missing energy (LHC) | Collision debris (likely gravitons) escapes into extra dimensions, so less energy is recorded after than before | Strong evidence for extra dimensions |
| Mini black holes (LHC) | Stronger short-range gravity makes black holes easier to form; even two protons might make one, which would evaporate (Hawking) into a distinctive spray of particles | Unmistakable signature |
| Gravitational waves | A string stretched cosmic-sized (force above 1020 tons) vibrates and emits gravitational waves of highly distinctive shape | Clear observational signature |
| CMB fine structure | Inflation might stretch stringy imprints large enough for the Planck satellite | Depends on early-universe string behaviour |
The asymmetry of evidence. Positive results could be explained without strings, and negative results prove little: superpartners may be too heavy, dimensions too small, accelerators too weak, signals too faint. The best positive results would not prove string theory right, and negative results would not prove it wrong. Yet a collection of positive results would make non-string explanations as overwrought as "touch-typing toes".
Singularities and black holes
A singularity is a setting so extreme (huge mass, tiny size, enormous curvature, punctures or rips in spacetime) that relativity and quantum theory output the equivalent of a calculator's divide-by-zero error. A quantum theory of gravity should never break down. Here string theory has its most impressive record:
- Orbifold singularities (mid-1980s): Lance Dixon, Jeff Harvey, Cumrun Vafa and Edward Witten showed that certain punctures in space, fatal to Einstein's equations, are harmless to strings: a point particle can fall into a puncture, but a string can only bang into it, wrap it or stick to it.
- Flop singularities (1990s): Greene with Paul Aspinwall and David Morrison, and independently Witten, showed that even when a sphere in space is crushed to zero size string theory stays sound: a moving string sweeps around the shrinking sphere like a hula hoop around a soap bubble, forming a "string shield".
- More exotic cases: conifolds, orientifolds, enhançons and others are under full control; situations that would have left Einstein, Bohr, Heisenberg, Wheeler and Feynman saying "we just don't know."
- Still open: the big bang and black-hole singularities, the most severe and most relevant.
- Black-hole entropy: Jacob Bekenstein and Stephen Hawking (1970s) showed black holes carry a specific entropy (disorder), which should count the rearrangements of their "innards", but no one could identify the innards. Andrew Strominger and Cumrun Vafa (1996) built a string-theoretic black hole transparent enough to count its states, and the result matched Bekenstein-Hawking spot-on: the first firm quantum accounting of a black hole's disorder.
String theory and mathematics
Einstein mined the work of Gauss, Riemann and Lobachevsky to describe curved spacetime; string theory is repaying the debt by creating mathematics. In general relativity, geometry and physics are yin and yang: different spacetimes (Schwarzschild's for a black hole, Roy Kerr's 1963 solution for a spinning one) mean different physics. String theory breaks this one-to-one link: different shapes can give physically identical worlds.
The reason: geometry has always been built from points, but strings are loops, hence stringy geometry. A string on a cylindrical piece of space can slide like a point or wrap around it like a rubber band on a soda can. Fatten the cylinder and the wrapped string stretches while the unwrapped one does not; the two probe different features. String theorists found pairs of shapes that look completely different to unwrapped strings and completely different to wrapped strings, yet what the unwrapped strings see on one, the wrapped strings see on the other, making the full physics identical. This is mirror symmetry, and it turns overwhelmingly hard calculations on one shape into easy ones on its partner.
Counting spheres in a Calabi-Yau (the quintic). Mathematicians asked how many ways a sphere can be packed into the shape wrapping a portion of it a given number of times, like a lasso wound around a beer barrel, and found all but the simplest cases impenetrable. Mirror symmetry answered at once:
- Wrapping 5 times: 229,305,888,887,625
- Wrapping 10 times: 704,288,164,978,454,686,113,488,249,750
- Wrapping 20 times: 53,126,882,649,923,577,113,917,814,483,472,714,066,922,267,923,866,471,451,936,000,000
These numbers opened a whole new chapter of mathematical discovery (enumerative geometry, Gromov-Witten theory). Whatever happens physically, string theory is already a potent tool for exploring the mathematical universe.
The scorecard
| Goal | Required? | Status (Greene, 2011) |
|---|---|---|
| Unite gravity and quantum mechanics | Yes, the primary goal | Excellent: a wealth of calculations attest to the merger |
| Unify all forces | No | Excellent: every force's quanta are string vibrations |
| Incorporate past breakthroughs | No | Excellent: reduces to GR and QFT |
| Explain particle properties | No | Indeterminate: too many extra-dimensional shapes |
| Experimental confirmation | Yes, the only arbiter | Indeterminate: untested |
| Cure singularities | Yes | Excellent, but big bang and black holes outstanding |
| Black-hole entropy | Yes | Excellent: Bekenstein-Hawking reproduced |
| Mathematical contributions | No | Excellent |
A theory in progress with stunning achievements, untested on the most important scale. Greene stresses this is not peculiar to strings: any quantum gravity theory lives far beyond experiment. Through the mid-1990s research used approximate equations, good for gross features but too coarse for refined predictions. The breakthroughs that went beyond them opened new varieties of parallel worlds: the subject of Chapters 5 and 6.
Going deeper: the technical notes
- Electroweak and grand unification. Glashow, Salam and Weinberg's electroweak theory was confirmed by accelerator experiments in the late 1970s and early 1980s. Glashow and Georgi's grand unification added the strong force; its simplest version predicted proton decay at a rate experiments ruled out, but other versions with slower decay survive.
- Supersymmetry and supergravity. Supersymmetry, discovered inside string theory, is the last available symmetry of a non-trivial particle theory beyond rotations, translations and Lorentz/Poincaré symmetry. It relates particles of different spin, linking force carriers to matter. Supergravity adds gravity. Both emerge as the low-energy limit of string theory.
- "Every field has a particle" (precisely). Small fluctuations of a field around a local minimum of its potential behave as particles. Localising a particle at a point would need infinite energy, so "point particle" is itself an idealisation.
- Renormalization and its failure. The infinities of the three nongravitational forces were tamed by renormalization; for gravity it fails. The modern view treats every theory as valid only above some length scale: with a built-in cutoff, no infinities arise, but the theory is silent below the cutoff. A full quantum gravity would lift the cutoff.
- Where the Planck scale comes from. A particle's quantum wavelength shrinks as its mass grows, while its black-hole size grows. Where the two meet, both quantum mechanics and gravity matter: that mass and size are the Planck mass and Planck length. (Holography, Chapter 9, suggests the conflict can surface even at low density in very large volumes.)
- Spin. Experiments show particles carry an intrinsic, immutable angular momentum in whole multiples of ħ/2; "spinning like a top" is a mental image, not a mechanism.
- How strings tame gravity. Point particles probe arbitrarily short distances, where gravitational jitters become arbitrarily violent. A string cannot probe below its own size, which caps the jitters' energy and keeps the mathematics finite.
- Massless moduli. String theory's freedom in particle properties shows up as numbers that can vary at no energy cost, which means massless particles ("moduli"); too many would conflict with data. Fluxes (Chapter 5) give these fields potentials and masses.
- How string theory could be falsified anyway. It requires unitarity (probabilities sum to 1), local Lorentz invariance, and technical properties of scattering such as analyticity and crossing symmetry. Evidence violating these (say at the LHC) would be hard to reconcile with strings.
- A black hole's "centre" is a time, not a place. Inside the horizon, radial direction and time swap roles: falling inward is moving into the future, so the singularity is a last moment rather than a location.
- Who found mirror symmetry. Hints came from Lance Dixon and independently Wolfgang Lerche, Nicholas Warner and Cumrun Vafa. Greene and Ronen Plesser built the first concrete mirror pairs (and named them) and showed hard calculations on one equal easy ones on the other; Philip Candelas, Xenia de la Ossa, Paul Green and Linda Parkes used this to compute the sphere counts quoted above. Yau and Nadis's The Shape of Inner Space (2010) tells the history.
Critical lens.
- The scorecard is self-graded. "Excellent" on required goals rests on internal consistency checks, not experiment; critics such as Lee Smolin (The Trouble with Physics, 2006) and Peter Woit (Not Even Wrong, 2006) argued that a framework with no unique vacuum and no prediction should not be rated so highly. Greene lists their kind of objection but doesn't engage them by name.
- "Explain particle properties: not required." True for a quantum gravity theory, but explaining the Standard Model was a central selling point in the 1980s; demoting it after the fact is a shifting of goalposts.
- The micron claim. The definitive torsion-balance tests of the inverse-square law (Eöt-Wash) reached about 50 μm around 2007; below that, Casimir-type experiments constrain only much stronger deviations. "No deviation down to a micron" overstates the sensitivity.
- Strominger-Vafa covered special black holes (extremal, supersymmetric); entropy counting for realistic Schwarzschild black holes is still not done microscopically.
2026 update: fifteen years of LHC and observation
- Supersymmetry: LHC Runs 1 to 3 (to 13.6 TeV) have found no superpartners; gluinos are excluded below about 2.3 TeV and squarks of the first generations below about 1.8 to 2 TeV in simple models. "Natural" weak-scale SUSY is under heavy pressure; SUSY at higher scales remains viable.
- Extra dimensions and mini black holes: no missing-energy or black-hole signal; large-extra-dimension models are pushed to fundamental gravity scales above about 9 to 11 TeV. Torsion-balance limits on extra dimensions now reach about 40 μm (Eöt-Wash, 2020).
- Gravitational waves: LIGO's first detection, GW150914 (14 September 2015), earned the 2017 Nobel Prize. Pulsar-timing arrays (NANOGrav and others, June 2023) found a nanohertz background; cosmic strings are one candidate source, though merging supermassive black holes are favoured. Cosmic-string tension is bounded at roughly Gμ ≲ 10−7 (Planck) and tighter from pulsar timing.
- CMB: Planck (final release 2018) found no stringy imprint.
- QED precision: the electron's magnetic moment was measured to 0.13 parts per trillion (Northwestern, 2023), still in agreement with QED; the Fermilab muon g-2 final result (2025) matches the Standard Model when lattice-QCD hadronic inputs are used.
- Mirror symmetry has been proved mathematically for these enumerative predictions (Givental; Lian, Liu and Yau, 1996 to 1997); string-inspired mathematics remains vigorous.
- The swampland programme (Vafa and others, from 2005; the de Sitter conjecture, 2018) asks which effective theories cannot arise from string theory, the first serious attempt to extract general predictions without picking a shape.
So what. Greene's scorecard separates required goals from nice-to-have goals and grades each separately: a template for evaluating any ambitious platform (an agentic-AI architecture, a core-banking replacement). Two further lessons transfer directly. (1) Reduction to the incumbent: a successor that reproduces the old system's results in the old regime (as strings reproduce GR and QFT) is far easier to trust and adopt. (2) The underdetermination trap: when a framework admits too many configurations (Calabi-Yau shapes, or a model zoo with countless hyperparameters), it can fit anything and predict nothing until something selects the configuration.
Hovering Universes in Nearby Dimensions: The Brane and Cyclic Multiverses
When string theorists finally went beyond their approximations in the mid-1990s, the theory changed shape. The five string theories turned out to be one theory (M-theory) seen through different dualities; space had ten dimensions, not nine; and the theory contained branes of every dimension, not just strings. That opened two new multiverses. In the Brane Multiverse our universe is a three-brane slab, with other universes possibly millimetres away in an extra dimension, invisible because only gravity can cross the gap. In the Cyclic Multiverse colliding branes reset each other every trillion years or so, giving universes parallel in time. The same advances multiplied the possible extra-dimensional shapes to about 10500.
The ladder problem and the art of approximation
Six pages for a falling ladder. Late one night at Cornell, hungry and in a hurry, Greene "spiced up" the honours freshman final by tweaking a standard problem (a ladder sliding down a wall) so that the ladder's density varied along its length. Writing the solutions during the exam, he found the half-page problem now took six pages ("I write big. But you get the point."). Textbook problems are specially designed to be solvable; change one assumption and they become as hard as the real world.
Almost nothing in nature can be solved exactly. The theorist's job is to decide which complications to discard. Earth's orbit needs the sun; the moon helps but raises complexity sharply (Charles-Eugène Delaunay filled two 900-page volumes on the sun-earth-moon problem); all the planets together are overwhelming. "The art of physics lies in deciding what to ignore." But approximation carries peril: a raindrop doesn't change a boulder's weight, yet on a boulder teetering at a cliff edge it can start an avalanche. In the mid-1990s string theorists found their raindrop.
Beyond approximations: perturbation theory
Every established theory has central equations (Newton's, Maxwell's, Schrödinger's, Einstein's) that are solvable exactly only in the simplest cases. String theory was worse: even its central equations were known only approximately, and solving them required further simplification: "approximations of approximations."
Ralph's lottery. Ralph plays two weekly draws, each a 1-in-a-billion chance, and claims 2 in a billion. Alice: "That's close." Winning twice doesn't make him win more, so subtract the chance of winning both (10−18): 0.000000001999999999. For ten draws, the first pass gives ten in a billion; the second pass corrects for double wins (∝ 10−18), the third for triple wins (∝ 10−27), and so on, each correction far smaller than the last. This is a perturbative approach.
Particle physics works the same way. Two particles colliding at the LHC: first pass, they exchange one force-carrying "bullet" (e.g. A photon); second pass, two; then three. Each extra exchange brings a factor called the coupling constant, the likelihood that one particle fires a bullet and the other absorbs it.
| Force / particle | Coupling constant | Perturbation theory? |
|---|---|---|
| Electromagnetic (electrons, photons) | ≈ 0.0073 (≈ 1/137) | Excellent: 0.00732 ≈ 0.0000533; 0.00733 ≈ 0.000000389 |
| Weak nuclear (neutrinos) | ≈ 10−6 | Excellent |
| Strong nuclear (quarks at the LHC) | somewhat less than 1 | Marginal |
| String coupling | Unknown | Assumed small, for convenience |
String theory has its own string coupling constant, the chance one string bumps another. With no experimental value, theorists assumed it small, partly like the drunk searching for keys under the lamppost because that's where the light is, and partly, more charitably, because previous successful theories had small couplings, so the light had often been in the right place. If the coupling were large, each correction would be bigger than the last, the series would never be justified, and "years of research would collapse." Even with moderate coupling, approximations might miss a raindrop.
Duality: five theories become one
In the 1980s there were five string theories: Type I, Type IIA, Type IIB, Heterotic-O and Heterotic-E. Each was studied only at small coupling; beyond that, theorists "shrugged and threw up their hands."
Spring 1995: Edward Witten (building on Joe Polchinski, Michael Duff, Paul Townsend, Chris Hull, John Schwarz, Ashoke Sen and others) showed that dialling one string theory's coupling ever larger turns it steadily into another string theory with ever smaller coupling. Type I at strong coupling is Heterotic-O at weak coupling. The five theories are not different after all. Two mathematical descriptions of one physics is a duality, and it doubles the calculational arsenal: a calculation impossible in one language is easy in its dual.
Greene's image is Aude Oliva's (MIT) hybrid picture: Albert Einstein close up, Marilyn Monroe from afar, ambiguous in between. Seeing only the extremes you would swear there were two pictures; scanning the intermediate distances shows one portrait. Witten argued, and others filled in, that all five theories are linked by a network of dualities; their union is M-theory.
Branes
As a student Greene asked what many later asked him: why only strings? Why not 2-D sheets or 3-D blobs? The 1980s answer was that higher-dimensional constituents gave inconsistencies such as negative probabilities, while strings, being highly symmetric, cancelled them. Strings seemed to be in a class of their own. "Or so it seemed."
The new non-perturbative methods showed that rationale was an artefact of the approximations. A small army of theorists found that objects of many dimensions lurk in string theory's mathematical shadows:
- One-branes: strings themselves, now just one entry on a long list.
- Two-branes: membranes, like Frisbees or flying carpets (one meaning of the "M" in M-theory).
- Three-branes, four-branes … nine-branes: all able to vibrate and wiggle like strings.
Space has ten dimensions, not nine: spacetime has eleven. The (D − 10) × Trouble argument was itself perturbative. Witten showed the string coupling controls the size of a hidden tenth spatial dimension; assuming the coupling small had made the dimension small, so small the mathematics itself couldn't see it. Greene recalls "dazed and wide-eyed looks everywhere" at the 1995 conference at the University of Southern California where Witten announced this: the opening of the Second Superstring Revolution (the first being Green and Schwarz, 1984).
Branes and parallel worlds: the braneworld
A string's length is set by its energy. Known particles correspond to minuscule strings, but pump in energy and a string stretches: an advanced civilisation could make strings any size, and cosmic expansion could stretch strings wrapped around space. Branes can likewise be big. A long string is like an overhead cable to the horizon; a big two-brane like an endless tablecloth or flag. An enormous, perhaps infinite, three-brane would fill our space like water in a fish tank, so we should think of it as the substrate of space itself. Space would be a thing, and we would move in and through a three-brane: the braneworld scenario.
String theory has more than three spatial dimensions, so there is room for more than one three-brane. Evolution prepared us to picture objects within 3-D space, not two coexisting space-filling 3-D entities, so drop a dimension: think of our universe, with the Orion, Horsehead and Crab nebulae, the Milky Way, the Andromeda, Sombrero and Whirlpool galaxies, all on one giant, extraordinarily thin slice of bread. A second universe is another slice, shifted slightly along the extra dimensions. Add more slices, a cosmic loaf, or branes oriented any which way, of any dimension.
All share the same fundamental laws (one theory, string/M-theory), but as with inflationary bubbles, environment matters: fields on each brane, even the brane's number of dimensions, change its features. Some may host galaxies and self-aware beings who think their slice is everything. Our universe would be one member of the Brane Multiverse.
Sticky branes: why we can't see our neighbours
The rules of confinement (Joe Polchinski's insight):
- Strings come in two shapes: loops and snippets (open strings).
- Snippet endpoints can only live on branes. Pulling an endpoint off is as impossible as making π smaller or √2 bigger, or removing the north pole from a bar magnet. Snippets glide freely within a brane but cannot leave.
- Ordinary matter and the nongravitational forces are snippets. You can jump, throw a baseball from first to second, send sound from radio to ear, but you cannot jump off the brane. Photons are snippets too, so light crosses from Andromeda to the Wilson Observatory but never across the gap to a brane a millimetre away.
- Gravitons are loops. Spin-2 is twice the snippet spin, so a graviton is like two snippets joined end to end. Loops have no endpoints, so gravity alone can leave and re-enter our brane: our only probe of the beyond.
The experimental bonus
Before branes, extra dimensions were assumed Planck-sized (10−33 cm). If only feeble gravity can feel them, they could be much larger and still unseen.
- About 10−4 cm across (roughly a billion billion billion times Planck size): short-range gravity experiments could see gravitons that haven't yet leaked off the brane, so gravity would appear stronger at short distance. The more gravitons leak, the weaker gravity looks at long range: an explanation of why gravity is so weak.
- About 10−18 cm across: the LHC could knock debris (gravitons) off our brane, producing missing energy.
- Mini black holes need gravity to strengthen at short range: another braneworld by-product.
These experiments are therefore tests of whether we live on a brane, and a positive result would be indirect evidence for other universes: if we live on one brane, the mathematics gives no reason to think ours is the only one.
Time, cycles and the multiverse
In the Quilted, Inflationary and Brane multiverses, the other universes are "out there" in space (far away; beyond our bubble; or a short hop through another dimension). Braneworlds also allow a multiverse in time.
"What would it mean for all of space to move ten feet to the right?" is a brain-teaser, but in a braneworld it is mundane: branes can move, and our three-brane could be gliding through the higher-dimensional bulk. If it hits a neighbouring brane, like two cymbals crashing, the relative motion's energy becomes a fiery rush of particles and radiation that obliterates all structure.
To Paul Steinhardt, Neil Turok, Burt Ovrut and Justin Khoury, that sounded like conditions just after the big bang. A three-brane full of hot plasma expands and cools, particles clump, stars and galaxies form. Some nicknamed it the "big splat", but the branes don't stick: they bounce apart. Gravity slows them, they reach maximum separation, fall back, collide, and reset everything again.
- Cycle length: about one trillion years (Steinhardt and co-workers' estimate) from birth through evolution to death.
- Our universe: merely the latest in a temporal series, earlier members of which may have held intelligent life and culture, now erased, as ours will be.
The long history of cyclic cosmologies
- Hindu tradition: one of the oldest cosmologies, with cycles within cycles spanning millions to trillions of years in some interpretations.
- Heraclitus and Cicero: universes consumed by fire and reborn from the embers. Christianity's unique genesis later dominated, but cyclic ideas never vanished.
- Alexander Friedmann (1923 popular book): some of his solutions suggest an oscillating universe: expand, reach maximum size, contract to a "point", perhaps re-expand.
- Einstein (1931): investigated oscillating models after abandoning stasis.
- Richard Tolman (Caltech, 1931 to 1934): the most thorough mathematical studies, launching a tradition that has bubbled in and out of the mainstream ever since. (Dicke's 1960s CMB thinking also began from an oscillating universe.)
The appeal: cycles sidestep the beginning. Each cycle has a physical cause, the end of the previous one, and the series as a whole has none. The steady state theory also avoided a beginning (new matter created continually as space expands) but was killed by evidence of a hot, chaotic early phase. Cyclic models seem to have it both ways: a big-bang-like start every cycle, but no ultimate beginning.
Zanstra's objection (1950s, implicit in Tolman): the Second Law of Thermodynamics says entropy increases, as kitchens, laundry bins and playrooms prove by nightfall. Tolman had shown that more entropy means a stronger rebound and a longer cycle. Run time backward and earlier cycles had less entropy, hence shorter durations; far enough back they shrink to nothing. The cycles would have had a beginning.
Steinhardt and Turok's reply: in the brane model what cycles is the separation between the branes, not the size of space. The branes keep expanding through every cycle. Total entropy rises, but entropy density falls; by each cycle's end it is diluted nearly to zero, a full reset. No beginning to time is needed.
A falsifiable difference from inflation
Inflation's violent stretching should produce primordial gravitational waves that imprint the CMB; a brane collision, lacking the stretch, should produce waves far too weak to leave a lasting signal. So detection of primordial gravitational waves would be strong evidence against the Cyclic Multiverse; continued non-detection would challenge many inflationary models and make the cyclic framework more attractive. The community knows the idea well but views it "almost as widely, with much skepticism."
In flux: 10500 ways to dress the extra dimensions
The non-perturbative revolution was supposed to pick out one shape for the extra dimensions. It did the opposite: it found vast new troves of candidates.
The key is flux. Electrons have electric fields, magnets magnetic fields, and branes have brane fields. Faraday measured a field's strength by the density of its field lines through a surface, which he called flux. Raphael Bousso, Polchinski, Steven Giddings, Shamit Kachru and others realised that a full description of the extra dimensions needs shape, size and the brane fluxes threading them. Earlier work (including Greene's in the 1980s and early 1990s) studied "naked" Calabi-Yau shapes. But their many open regions, like the inside of a beach ball, a doughnut's hole, or a blown-glass sculpture, can be wrapped by branes and threaded by flux: "dressed-up" Calabi-Yaus.
The counting argument:
- Flux is quantised: like photons and electrons, it comes in whole units (1, 2, 3 …), never 1.2 or 6.4.
- Practical cap about 10: larger flux distorts the shape and breaks reliable methods, so researchers keep flux numbers about 10 or less.
- Multiply per region: one open region gives 10 dressings; two give 10 × 10 = 100; three give 103.
- Some Calabi-Yaus have about 500 open regions, giving about 10500 forms for the extra dimensions: "far more outfits than there are particles in the observable universe" (about 1080).
Three reactions followed:
- Despair: without a way to choose the shape and its flux outfit, string theory loses predictive power. "Some string theorists lost heart."
- Patience: a missing principle that fixes the extra dimensions may yet be found, soon or far off.
- Radicalism: decades of failure may be telling us something. Perhaps all the shapes and fluxes are real, each the extra-dimensional part of its own universe, and perhaps that is exactly what's needed to explain the cosmological constant (Chapter 6).
Going deeper: the technical notes
- The four central equations. Classical mechanics: F = ma. Electromagnetism: dF = 0, d*F = *J. Quantum mechanics: iħ ∂ψ/∂t = Hψ. General relativity: Rμν − ½gμνR = 8πG Tμν. Each is simple to state and hard to solve in realistic cases.
- The electromagnetic coupling is the fine-structure constant, α = e2/ħc ≈ 1/137 ≈ 0.0073 at everyday energies.
- The duality web in detail. Type I at strong coupling is Heterotic-O at weak coupling (and vice versa); Type IIB at strong coupling is Type IIB again at weak coupling (self-dual); Type IIA and Heterotic-E at strong coupling grow an eleventh dimension and become M-theory.
- Why strings are special (perturbatively). A moving string sweeps out a 2-D surface, so its dynamics is a two-dimensional quantum field theory with infinite-dimensional conformal symmetry. That symmetry removes the negative-probability states that the time direction would otherwise create, and preserving it quantum mechanically is what fixes the number of dimensions. Higher-dimensional objects lack this symmetry, which is why they looked inconsistent until non-perturbative methods arrived.
- Who found branes. Groundwork and key results came from Michael Duff, Paul Howe, Takeo Inami, Kelley Stelle, Eric Bergshoeff, Ergin Sezgin, Paul Townsend, Chris Hull, Chris Pope, John Schwarz, Ashoke Sen, Andrew Strominger, Curtis Callan, Joe Polchinski, Petr Hořava, Jin Dai, Robert Leigh, Hermann Nicolai and Bernard de Wit, among others.
- Time-based versus space-based multiverses. The Inflationary Multiverse also involves time (each bubble's wall is its big bang), but its analysis is spatial; the Cyclic Multiverse is temporal from the outset.
- Dark energy's job in the cyclic model. In each cycle's last phase, accelerated expansion (today's dark energy) dilutes the entropy density, preparing the reset; the model thus explains today's acceleration as part of the cycle.
- Friedmann's oscillating universe appeared in his 1923 popular book The World as Space and Time.
- Why fluxes are capped. Large flux not only spoils the mathematics but tends to make the Calabi-Yau grow large, which would make the extra dimensions visible, so realistic constructions keep fluxes modest.
Critical lens.
- Credit for braneworlds. The large-extra-dimension proposal that makes branes testable is chiefly due to Arkani-Hamed, Dimopoulos and Dvali (1998), and the warped braneworld to Lisa Randall and Raman Sundrum (1999); neither is named, which makes the experimental programme look more purely "string theoretic" than it is.
- "Most experimentally accessible" depends on lucky parameters. Nothing in string theory requires extra dimensions large enough for the LHC; detectability is hope, not prediction.
- The cyclic model's problems are not only social. Critics raised technical issues: what happens to the extra dimension when the branes touch (a singular moment the model needed string theory to resolve, which it hadn't), and whether the post-collision fluctuations really yield the observed nearly scale-invariant spectrum. Greene mentions only "skepticism."
- 10500 is a heuristic. It assumes independent flux choices with a cap of 10 and does not check that each dressed shape is a stable, consistent vacuum. Later estimates range from 10272,000 (Taylor and Wang, 2015, for F-theory geometries) to the swampland suspicion that few or none are fully consistent de Sitter vacua.
2026 update
- No braneworld signal at the LHC: no missing-energy excess, no mini black holes, no Kaluza-Klein graviton resonances through Run 3.
- Primordial gravitational waves still unseen: r < 0.036 (BICEP/Keck 2021). Greene's test is live and, so far, non-detection is the outcome, which gives the cyclic model some relief, though plateau inflation models also predict small r.
- The cyclic idea evolved: Anna Ijjas and Paul Steinhardt (2019 onward) proposed a non-brane cyclic model in which a slow contraction phase smooths the universe and a classical bounce, with dark energy decaying, sets up each new cycle; entropy is diluted by net growth of the universe each cycle. Roger Penrose's conformal cyclic cosmology is a separate cyclic proposal; claimed "Hawking points" in the CMB are not supported by independent analyses.
- Swampland pressure: the 2018 de Sitter conjecture challenges whether the flux constructions (KKLT, 2003) truly give stable vacua with positive vacuum energy; the question is still argued in 2026.
So what. (1) Perturbation is a design pattern and a risk. Like Ralph and Alice, most enterprise estimates (credit loss, operational risk, AI eval scores) are "first pass plus small corrections", valid only when the coupling (interaction between factors) is small. In a stressed regime the corrections grow and the series breaks: precisely the 2008 lesson about correlated defaults. (2) Duality as an engineering move: when one representation makes a problem intractable, find an equivalent representation where it is easy (Fourier transforms, primal-dual optimisation, rewriting a graph query as a join). (3) The flux explosion is the combinatorics of configuration: 10 settings across 500 independent switches is 10500 states; it is why exhaustive testing of configurable platforms is impossible and why you need a selection principle.
# Ralph and Alice: perturbation series for "win at least once" in n draws
import matplotlib.pyplot as plt
p, n = 1e-9, 10
exact = 1 - (1-p)**n
terms, s = [], 0.0
from math import comb
for k in range(1, 5): # inclusion-exclusion passes
s += (-1)**(k+1) * comb(n, k) * p**k
terms.append(abs(s-exact))
print(f"exact={exact:.18e}")
plt.semilogy(range(1,5), [t+1e-40 for t in terms], 'o-', color='#DB2777')
plt.xlabel('number of passes'); plt.ylabel('error'); plt.show()
New Thinking About an Old Constant: The Landscape Multiverse
The 1998 discovery that cosmic expansion is speeding up revealed a cosmological constant of about 10−123 in natural (Planck) units: tiny, but not zero. Quantum theory predicts a value some 120 orders of magnitude larger, and no symmetry has been found that cancels it. Greene shows how Steven Weinberg's 1987 anthropic argument predicted a small nonzero value, but only if there exist more than 10124 universes with different constants, and how string theory's 10500 extra-dimensional forms, populated by eternal inflation and quantum tunnelling, could supply them. This Landscape Multiverse is the first in the book invoked to explain a measurement in our own universe, which forces the question: is this science?
A tiny number with huge consequences
The difference between 0 and 0.000…000138 (122 zeros after the decimal point, then 138) hardly seems to matter. Yet it may force a radical shift in how we picture reality. The number, first measured in 1998 by two teams studying exploding stars in distant galaxies, appears to be the entry on line three of Einstein's tax form: the cosmological constant, the density of dark energy suffusing space. Decades of theory had convinced nearly everyone it was zero. Not quite everyone: a contentious line of thought had suggested a nonzero value might be found, on one supposition. We live in one of many universes. Many universes.
The return of the cosmological constant
Einstein introduced the constant in 1917 to balance gravity; after Hubble he dropped it. The story that he called it his "greatest blunder" comes from George Gamow, whose penchant for "playful hyperbole" makes some historians doubt it. Certain is that Einstein later wrote that had "Hubble's expansion been discovered at the time of the creation of the general theory of relativity, the cosmological constant would never have been introduced." Hindsight blurred his earlier clarity. In 1917, writing to Willem de Sitter, he had been more nuanced:
"The general theory of relativity allows the inclusion of the cosmological constant in the field equations. One day, our actual knowledge... Will probably have come far enough for us to be able to decide empirically the question of whether or not the cosmological constant vanishes. Conviction is a good motive, but a bad judge." Albert Einstein to Willem de Sitter, 1917
Eight decades later the Supernova Cosmology Project (led by Saul Perlmutter) and the High-Z Supernova Search Team (led by Brian Schmidt) did exactly that, by studying spectral lines from distant stars. To the shock of many, the constant does not vanish.
Cosmic destiny: what the teams set out to measure
Neither team was hunting the cosmological constant. They wanted the rate at which expansion is slowing, which decides the universe's fate:
- Big slowdown: expansion halts and reverses: contraction, a big crunch (or a cyclic bounce).
- Small slowdown: like a ball thrown faster than escape velocity, space expands forever.
The plan: measure how fast space was expanding at various past times by finding, for galaxies at many distances, how fast they were receding when they emitted the light we see. That needs two things: distance and speed then.
Distance and brightness
- Parallax. Close each eye in turn and a book's corner jumps; the angle shift gives distance (your brain does this automatically, which is why you see in 3-D, and why 3-D movies work). For stars, use Earth's positions six months apart. In 1838 Friedrich Bessel won the race, measuring 61 Cygni's parallax at 0.000084 degrees, about 10 light-years. Satellites now reach a few thousand light-years.
- Brightness. Light spreads as it travels, so compare apparent brightness with intrinsic brightness to get distance. The hitch: is a star dim because it's far, or because it's feeble? You need standard candles of known intrinsic brightness.
- Type Ia supernovae. A white dwarf siphons material from a companion (often a red giant) until it reaches about 1.4 solar masses, can no longer support itself, collapses and explodes, briefly outshining the 100 billion or so other stars in its galaxy. Because every such explosion follows the same physics at the same mass, they reach similar peak brightness: ideal standard candles, visible across billions of light-years.
- Catching rare events. A typical galaxy has one once every few hundred years. Like epidemiologists studying rare conditions in large populations, the teams used wide-field detectors watching thousands of galaxies at once, found dozens of Type Ia supernovae, and followed them with conventional telescopes.
Whose distance is it? The static map
Which distance do these measurements give: then-to-then, now-to-then, or now-to-now? Greene's tool is the unchanging map with a changing legend.
The expanding-map method. On a US map, New York to Los Angeles = 39 cm, Los Angeles to Austin = 19 cm, Austin to New York = 24 cm; at 1 cm = 100 km these are about 3,900, 1,900 and 2,400 km. If Earth's surface swelled uniformly, the map stays valid: just update the legend (1 cm = 200 km at noon, 300 km at 2 pm, 400 km at 4 pm). Galaxies, like cities on a swelling Earth, don't move under their own power; space swells. So keep one static cosmic map and update its legend, the scale factor, which grows over time.
Applied to a supernova in the "Noa Galaxy": the light's dilution measures the area of the sphere it has spread over by now, hence its radius, the light's whole trajectory converted at today's legend, including all the stretching of earlier segments after the light passed. So brightness gives the distance now between us and the galaxy's current location.
Colours of cosmology: redshift as a ruler of size
A neon sign glows red because excited electrons in neon atoms drop to lower energies and emit photons of characteristic colours; blue "neon" is usually mercury, gold usually helium. Every element has a unique fingerprint of wavelengths. During the 1868 eclipse, Pierre Janssen and Joseph Norman Lockyer found a line no known substance produced: a new element, helium, the only element discovered in the sun before on Earth.
Light from distant sources shows familiar fingerprints shifted to longer wavelengths (3%, 12%, 21% longer): redshift. Vesto Slipher and Edwin Hubble traced it to expansion. On the static map the wave's crests are evenly spaced, but converting with a legend that has grown means the real wavelength grows, like thread stitched into spandex.
Redshift measures size, not speed directly: if light is stretched 21%, the universe is 21% larger now than when the light left. Each supernova is a dated pencil mark on a child's wall; enough marks reveal growth spurts. Dating the marks needs care, because light covers distance partly by its own motion and partly by space stretching under it, like a traveller on an airport moving walkway.
Cosmic acceleration
After checking, rechecking and checking again, both teams announced: for the last 7 billion years the expansion of space has not been slowing down. It has been speeding up. Until about 7 billion years ago the scale factor's growth slowed as expected; then the curve turned upward.
The island universe of the far future. In 100 billion years, every galaxy outside our Local Group (a gravitationally bound cluster of about a dozen galaxies) will have crossed our horizon. Unless they inherit records, future astronomers will theorise about an island universe of galaxies "numbering no more than students in a backwoods school, floating in a static sea of darkness." We live in a privileged age: "Insights the universe giveth, accelerated expansion will taketh away."
The cosmological constant's value
A ball that speeds up after being thrown upward is being pushed. The supernova data required something pushing outward: the job description of repulsive gravity from a cosmological constant, returned "not through the bad judgment of conviction... But through the raw power of data."
- Dark-energy density: just under 10−29 g/cm3 (as mass, via m = E/c2).
- Why acceleration began late: ordinary matter dilutes as space grows; the constant does not, since every cubic metre of space contributes the same push. More space between objects means more push. Matter's pull won for 7 billion years; then the constant took over.
- In Planck units (Planck mass ≈ 10−5 g per Planck volume ≈ 10−99 cm3): Λ ≈ 1.38 × 10−123. (Asking for "1015 picograms of potatoes" is odd; natural units are kilograms, and for vacuum energy they are Planck units.)
- Confirmation: the case has grown stronger, and CMB measurements dovetail with the supernova results.
Alternatives exist: an exotic quantum field masquerading for a while as a constant (as the inflaton did), or modified gravity that weakens faster with distance. But the simplest, most convincing explanation is a nonzero cosmological constant. To many researchers it is the single most surprising observational result of their lifetimes.
Explaining zero: the worst prediction in physics
Greene's first reaction: "It just can't be." Most theorists had concluded the constant was zero, first from "greatest blunder" lore, then from a powerful argument.
Quantum jitters carry energy, and they cannot be removed from empty space. Since the cosmological constant is energy uniformly filling space, quantum field jitters generate one. Einstein introduced the constant abstractly; quantum physics makes it inevitable: had Einstein not invented it, someone would have.
How big should it be? The calculation in three steps:
- Naive answer: infinite. A field in a box can jitter in infinitely many wave shapes, each shorter (and more energetic) than the last.
- Cut off at the Planck length. Jitters shorter than 10−33 cm need quantum gravity; quarantine them. The answer becomes finite: about 1094 g/cm3, more than all the stars of all known galaxies crushed into a thimble, i.e. 1 Planck mass per Planck volume (Λ ≈ 1).
- Compare with observation: at that value everything from galaxies to atoms would be ripped apart. Theory exceeds the observational bound by more than a hundred orders of magnitude (observed Λ ≈ 10−123).
The hope of symmetry
Archie's homework. Add 163 + 263 + … + 1063 to (−1)63 + (−2)63 + … + (−10)63. Archie grinds through numbers with more than five dozen digits. Edith: "Use symmetry, Archie." Each term cancels its negative partner, like children of equal weight on a seesaw: the answer is 0, no arithmetic needed.
Physicists hoped an unidentified symmetry would similarly cancel the vacuum energy, the only strategy available, and so concluded Λ = 0. Supersymmetry gives a concrete version: every field's jitter is matched by its partner's (selectron for electron, squark for quark, sneutrino for neutrino) with equal size and opposite sign. But exact cancellation needs equal masses, and since superpartners haven't been seen they must be much heavier. Broken symmetry, unbalanced seesaw, huge result again. Many other cancellation proposals also failed, yet most researchers read this as incomplete understanding, not as a sign that zero was wrong.
Weinberg's heresy and the anthropic principle
One who challenged the orthodoxy was Nobel laureate Steven Weinberg. In 1987, more than a decade before the supernova data, he derived a small but nonzero cosmological constant (Cambridge's George Efstathiou was another early, convincing advocate). His tool was perhaps the most polarising idea in decades, revered by some and vilified by others: the anthropic principle.
From Copernicus to Carter
Copernicus began a series of nested demotions: we are not at the centre of the solar system, the galaxy, or the universe, and not even made of the dark ingredients that dominate the cosmos's mass. This is the Copernican principle: everything points to humans not occupying a privileged position. At a conference in Kraków marking Copernicus's 500th birthday (1973), Australian physicist Brandon Carter (echoing Alfred Russel Wallace, Abraham Zelmanov and Robert Dicke) noted one arena where we are indispensable: our own observations. So we must correct for selection bias.
| Biased study | What it wrongly concludes | Correction |
|---|---|---|
| Counting trout only in the Sahara | "Why are trout extinct?" | Sample hospitable habitats too |
| Surveying subscribers of Can't Live Without Opera | "Why is opera interest surging?" | Survey a representative population |
| Interviewing refugees who survived a brutal trek | "This group is uniquely hardy" | Fewer than 1% survived; only the strongest are in the sample |
| Explaining why Earth is 93 million miles from the sun (Kepler's quest) | Seeks a fundamental law | Gravity allows any orbit; we can only exist where liquid water is possible |
The shoe shop. You think a shop stocks one shoe size and marvel that it's yours: "lucky coincidence or deeper explanation?" Learn the shop stocks every size and the question evaporates. With many planets at many distances, it is no surprise one is life-friendly, and of course we live on it. Earth's distance is a contingent historical fact of the gas cloud that formed the solar system; seeking a deeper reason is "a fool's errand."
Carter's name for this accounting, the anthropic principle, is unfortunate (it applies to any observers, humans or otherwise). No one objected to it for things inside the universe. The controversy came when Carter applied it to the universe itself: perhaps the electron mass (0.00054 of the proton's), the electromagnetic coupling (0.0073) and the cosmological constant (1.38 × 10−123) have no fundamental explanation, because they vary across a multiverse and we necessarily find ourselves where they permit life. Asking why they have their values would then be the wrong kind of question.
The argument fails without many universes. A unique universe still invites "lucky coincidence or deeper explanation?" Only a shop stocked with many sizes makes the fit unremarkable. Anthropic reasoning makes the mysterious mundane only inside a robust multiverse.
The three assumptions of anthropic explanation:
- Our universe is part of a multiverse. Once anathema; previous chapters give it serious standing.
- The constants vary over a broad range across universes. Expected in the Inflationary and Brane multiverses, and examined further below.
- Most variations would prevent life as we know it. The subject of Weinberg's calculation.
Life, galaxies and nature's numbers
Many constants are finely balanced: stronger gravity and stars burn out before life evolves; weaker and galaxies don't hold together; stronger electromagnetism and hydrogen nuclei repel too strongly to fuse and power stars. For the cosmological constant, Weinberg could not model life, so he used a clever proxy: galaxy formation. Galaxies grow by a snowball effect: a clump denser than its surroundings pulls in more matter, grows, pulls harder. Too large a cosmological constant pushes young, fragile clumps apart before they can grow.
- Weinberg's bound: galaxies cannot form if Λ exceeds a few hundred times the current matter density (a few protons per cubic metre), i.e. About 10−121 in Planck units.
- Negative Λ is constrained even more tightly: it adds to gravity and collapses the universe before stars ignite.
- Weinberg's sharper prediction: expect Λ near the upper limit, not far below it, since a much smaller value would need a non-anthropic explanation, the very kind physics had failed to find.
- 1998: observed Λ ≈ 10−123, within a factor of about 100 of the bound. "Prophetic."
How many universes does the argument need?
Casting Pulp Friction. Producer Harvey W. Einstein wants a lead between one and two metres tall, and whatever height he picks, someone must match. At centimetre precision there are 100 possible heights, so to be safe you call a few hundred actors; at millimetre precision, 1,000 heights and a few thousand actors. (You don't remind him that quantum uncertainty means he doesn't need every height; you remember what happened to the talking fly.)
Apply this to Λ. Take values between 0 and 1 in Planck units (negative values collapse; values above 1 strain our mathematics). Today's measurement precision is about 10−124, so there are 10124 distinguishable values. Allowing for duplicates, a safe multiverse needs perhaps 10130 universes. For scale: cells in your body ≈ 1013; seconds since the big bang ≈ 1018; photons in the observable universe ≈ 1088. Weinberg's explanation works only if we inhabit a multiverse whose universes realise some 10124 distinct cosmological constants.
From vice to virtue: the string landscape supplies the universes
String theory's ≈10500 flux-dressed Calabi-Yau shapes dwarf 10124: subtract 10124 a billion times and you barely dent 10500. Most important, Λ varies among them:
- Fluxes carry energy, as magnetic flux does, and the amount depends on the shape's geometric details.
- The extra dimensions sit at every point of ordinary space, like carpet pile on a backing; soak the fibres and the whole carpet is uniformly heavier. So the energy of the dressed shape fills our three dimensions uniformly: it contributes to the cosmological constant.
- Raphael Bousso and Joe Polchinski (2000) argued the 10500 values are spread fairly uniformly across a broad range, so many lie very close to the observed value.
- Finding one explicitly is another matter: at one per second for a billion years, a computer would examine only 1032 shapes.
The very feature that dismayed string theorists, no unique universe, becomes a virtue: "string theory makes the least plausible part of Weinberg's explanation suddenly seem plausible."
The final step: eternal inflation populates the landscape
Possible universes are not actual ones. The gap is filled, as Leonard Susskind emphasised most forcefully (inspired by Shamit Kachru, Renata Kallosh, Andrei Linde and Sandip Trivedi, "KKLT", 2003), by eternal inflation. The three-sentence summary: the Inflationary Multiverse produces an endless supply of bubbles; melded with string theory, eternal inflation scatters the 10500 forms of the extra dimensions across the bubbles, one per bubble, realising all possibilities; we live in a bubble whose extra dimensions yield a life-friendly Λ and the physics we observe.
The string landscape
Greene updates the Chapter 3 mountain metaphor in three ways:
- Altitude = total vacuum energy from all sources (inflaton plus every field's jitters), rather than the inflaton's alone.
- The valley floor is not at sea level: a valley's altitude is that bubble's cosmological constant. Our mystery becomes "why is our valley so close to, but not exactly at, sea level?"
- Valleys = shapes: the terrain has a lush assortment of valleys, ledges and outcroppings, one for each of the ~10500 forms. Strictly it is a 500-dimensional terrain (one axis per flux), impossible to draw, but connected. This is the string landscape (Susskind's term).
Quantum tunnelling between valleys
Quantum tunnelling turns classical "can't penetrate" into "small but nonzero probability." An electron facing ten feet of steel can, thanks to its jitters, occasionally materialise on the other side, at a random moment. It really happens: the sun shines only because hydrogen nuclei tunnel through their electric repulsion to fuse.
Sidney Coleman with Frank De Luccia (1980), independently of string theory, scaled tunnelling up to a whole universe: a universe whose field sits in a higher trough, separated by a bump from a lower one, can tunnel through. Not all at once: a seed event creates a small, randomly located bubble of the lower-energy value, which then grows, like Vonnegut's ice-nine.
How the Landscape Multiverse forms:
- Start high: a universe whose extra dimensions sit in a high valley has a large Λ and inflates rapidly.
- Tunnel: at a random place and time, a tiny region's extra dimensions transmute (shape, size and fluxes change) to the form of a lower valley. Nothing "moves"; the form changes.
- Grow, but more slowly: the new bubble has a smaller Λ, so it expands more slowly than its faster-inflating parent.
- Repeat: tunnelling happens again in parent and child, giving bubbles within bubbles within bubbles, "like a sequence of Matryoshka dolls, each painted by a different artist."
- Exhaust the possibilities: the relentless series of tunnelling events realises every form of the extra dimensions in some bubble. This is the Landscape Multiverse, and somewhere in it are bubbles with Λ ≈ 10−123. We live in one.
The rest of physics?
The cosmological constant draws special attention because the theory-observation chasm is so vast. But what of three neutrino species, the electron's mass, the weak force's strength? Surveys of the landscape show these also vary from place to place, at least in current understanding. So calculating them might be as misguided as deriving Earth's distance from the sun. For this to be credible, though, we need at minimum one bubble with the right Λ and the right particles and forces. That is the goal of string model building: hunting the landscape for shapes, sizes, fluxes and branes that reproduce our world, a craft that combines rigour with "an artistic sensibility." No one has yet found an exact match, but with 10500 options the consensus is that our universe has a home somewhere in the landscape.
Is this science?
The chapter turns a logical corner. Earlier multiverses broadened our picture of what might exist. The Landscape Multiverse is invoked to explain observations we make here, using worlds we may never visit, see, test or influence. Chapter 7 takes up the question directly.
Going deeper: the technical notes
- Two extra dimmings. Redshift lowers each photon's energy, and it also stretches the arrival rate of photons. A supernova's measured brightness therefore falls by the area of the sphere and by two factors of (1 + z), which astronomers must correct for. This gives the luminosity distance.
- The light-travel distance formula. In a flat expanding universe, ds2 = c2dt2 − a2(t)dx2. Light has ds = 0, so the comoving (map) distance is ∫c dt/a(t), and the physical distance today is a(tnow) ∫c dt/a(t): every segment of the path is multiplied by how much it has stretched since the light crossed it. Without expansion this reduces to c(tnow − temitted).
- "Then" equals "now" location. Since galaxies ride the expansion rather than move through space, a galaxy is at the same comoving location now as when it emitted the light, so Greene's second and third candidate distances agree when properly read.
- The exact value. Dark-energy density ≈ 7.12 × 10−30 g/cm3. Converting: × 4.6 × 104 Planck masses per gram × (1.62 × 10−33 cm per Planck length)3 gives 1.38 × 10−123 Planck units.
- Quintessence. A field rolling down a gentle, long potential gives mild, prolonged acceleration, unlike inflation's brief intense burst. This is the main alternative to a true constant, and the one DESI's 2024 to 2025 hints point toward.
- Why the vacuum sum diverges. Each jitter's energy is inversely proportional to its wavelength, so summing over ever-shorter wavelengths gives infinity.
- Why supersymmetry cancels it. Supersymmetry pairs bosons (integer spin, described by commuting variables) with fermions (half-integer spin, anticommuting variables); the anticommutation produces the relative minus sign that makes paired vacuum energies cancel, but only if partner masses are equal.
- "Fine-tuned for life" is debated. Some argue the life-permitting range is wider than often claimed, especially when several constants vary together (see Barrow and Tipler, The Anthropic Cosmological Principle; Barrow, The Constants of Nature; Davies, The Cosmic Jackpot; Stenger, Has Science Found God?).
- Infinite universes are not enough. A multiverse explains Λ only if (a) the laws produce a multiverse, (b) it contains far more than 10124 universes, and (c) Λ actually varies among them. Infinitely many universes with the same few Λ values would not help.
- KKLT and after. Kachru, Kallosh, Linde and Trivedi were the first to show fully that flux-stabilised Calabi-Yau shapes can give small positive Λ; with Juan Maldacena and Liam McAllister they then showed how to embed inflation in string theory (KKLMMT, 2003).
- The landscape's dimensions. The terrain lives in a roughly 500-dimensional space (one axis per flux), plus all sizes and shapes; valleys are local minima of the extra-dimensional potential energy. Tunnelling uphill is possible but far less likely than downhill.
Critical lens.
- A missing laureate. The High-Z team's acceleration paper was led by Adam Riess (1998), who shared the 2011 Nobel Prize with Perlmutter and Schmidt. Greene names only the team leaders.
- Weinberg's "prediction" is softer than it looks. He gave an upper bound; the claim that Λ should be near the bound depends on assuming a flat "prior" over Λ across the multiverse and on which observers to count (galaxies? Galaxies of what mass?). Later analyses (Martel, Shapiro and Weinberg, 1998; Bousso's "causal patch" measure, 2007) got values within roughly an order of magnitude of observation, but with choices that critics call tunable.
- Varying only Λ is a convenient slice. If other constants vary simultaneously, life-friendly universes with very different Λ may exist (Aguirre, 2001, showed universes with different primordial fluctuation amplitudes tolerate much larger Λ). The anthropic window then widens and the "prediction" weakens.
- "Uniform distribution" of Λ across the landscape is an assumption from Bousso-Polchinski toy models, not a theorem about the full set of vacua.
- The existence of the de Sitter vacua themselves (KKLT) has been contested since 2018; if positive-Λ stable vacua don't exist in string theory, the Landscape Multiverse loses its foundation.
- Rhetorical framing: "the consensus is that our universe has a home somewhere in the landscape" is a statement of hope, since no explicit construction exists.
2026 update: dark energy after DESI
- Nobel: 2011 Physics Prize to Saul Perlmutter, Brian Schmidt and Adam Riess.
- Dark energy may not be a constant. DESI baryon acoustic oscillation data (DR1 April 2024; DR2 March 2025), combined with CMB and supernova samples, prefer an equation of state that was phantom-like (w < −1) in the past and is weakening today, at 2.8σ to 4.2σ. If it holds, the "constant" is a slowly evolving field, rewriting Greene's framing (though not the size problem: the energy density is still ~10−123).
- New surveys: Euclid (launched July 2023) and the Vera C. Rubin Observatory (first images June 2025, LSST under way) will test this at higher precision; the Nancy Grace Roman Space Telescope is due later in the decade.
- Swampland: the 2018 de Sitter conjecture (Obied, Ooguri, Spodyneiko and Vafa) argued string theory may admit no stable de Sitter vacua, pointing instead to rolling quintessence, a picture DESI's hints would fit. Defenders of KKLT-type constructions dispute it. The landscape's foundations are now an active controversy rather than a settled backdrop.
- Weinberg died in July 2021; his anthropic bound remains the most cited success of anthropic reasoning.
So what. Two lessons travel well beyond cosmology. (1) Selection bias is the default, not the exception. A model validated only on the loans that were approved, the customers who stayed, or the AI pilots that "succeeded" is the refugee interview: survivors only. Before seeking a deep cause for a pattern, ask whether the act of observing filtered the sample. (2) Distinguish laws from contingencies. Some parameters (like Earth's orbital distance) are historical accidents, not consequences of principle. In architecture reviews, separating "this must be so" constraints from "this happened to be so" decisions prevents teams from defending accidents as principles.
Science and the Multiverse: On Inference, Explanation, and Prediction
Can a theory that invokes unobservable universes be science? Greene answers yes in principle, not yet in practice. Science already accepts unobservable ingredients (probability waves, black-hole interiors, regions beyond our horizon) when the theory that implies them earns its keep with confirmed predictions. A multiverse can even be predictive, if its universes are distributed in a sharply skewed or tightly correlated way. The obstacles are real: we can't yet compute those distributions, anthropic arguments rest on our being typical, and comparing infinite collections of universes (the measure problem) has no agreed solution and may even undermine inflation's own successes. The payoff, if it works, is a clean sorting of which mysteries are real and which are artefacts of assuming one universe.
Passions at the battlefront
- David Gross (Nobel 2004): fights the Landscape Multiverse quoting Churchill (29 October 1941): "Never give in... Never, never, never, never... Never give in."
- Paul Steinhardt (Princeton; co-architect of modern inflation): reaches for an unfavourable comparison to religion.
- Martin Rees (Astronomer Royal): sees the multiverse as the natural next step in our grasp of all there is.
- Leonard Susskind: those who ignore it are averting their eyes from a vision they find overwhelming.
- Greene: in a quarter century of string theory he has never seen passions run so high or language turn so sharp. Many see this as "a battleground for the very soul of science."
The soul of science
The questions apply to every multiverse theory: Is it legitimate to invoke domains inaccessible even in principle? Is a multiverse testable or falsifiable? Does it give explanatory power we'd otherwise lack? Detractors say no. Proponents say the connection to observation may be more indirect, less explicit, and dependent on fortune, but in respectable proposals it is not absent.
Textbook summaries say science finds regularities, explains them by laws, and tests the laws by prediction. Greene adds that real science is messier: asking the right questions matters as much as answering them, and today's questions are shaped by yesterday's insights. So any development, multiverses included, should be judged also by its effect on which questions we ask. That prospect "invigorates some and infuriates others."
Accessible multiverses: collisions
The multiverse is not monolithic (five varieties so far, four more to come), and some allow contact. In the Brane Multiverse, closed string loops can travel between branes. In the Inflationary Multiverse, bubbles separate because inflating space lies between them, but expansion is cumulative: two bubbles that form really close together have so little space between them that their own growth outpaces their separation, and they collide.
Groups including Jaume Garriga, Alan Guth and Alexander Vilenkin; Ben Freivogel, Matthew Kleban, Alberto Nicolis and Kris Sigurdson; and Anthony Aguirre and Matthew Johnson showed that while some collisions would wreck a bubble's interior, gentler "fender-benders" could leave shock waves that modify the CMB's hot and cold pattern. Detecting such a fingerprint would be direct evidence of other universes. But what if no such signal is ever found?
Science and the inaccessible I: is it legitimate in principle?
Every theory has an architecture: its ingredients and their laws, which also fix the questions it can ask. Physics has grown steadily more comfortable with the unseen:
| Theory | Core ingredient | Direct access? | Why we believe it |
|---|---|---|---|
| Newton | Positions and velocities of rocks, balls, moon, sun | Yes, tangible | Countless confirmed predictions |
| Maxwell | Oscillating electric and magnetic fields | No; eyes see light, not fields | Instruments plus confirmed predictions |
| Special and general relativity | Spacetime, then curved spacetime | No; "I've never grasped spacetime the way I grasp my chair" | Extraordinarily accurate predictions |
| Quantum mechanics | Schrödinger's probability wave (mid-1920s) | Never: the theory itself makes it unobservable | Predictions of where particles turn up |
| GR: black-hole interiors | Region behind the event horizon | Never, from outside, by the laws themselves | Full consensus it is real |
| GR: beyond the horizon | Galaxies more than ≈20 billion light-years away in an accelerating universe | Never: like a kayaker against a current faster than she can paddle | Same theory that describes what we see |
A theory's success justifies its architecture after the fact, even when that architecture is inaccessible. A galaxy that slips over our horizon does not drop out of reality. So confidence in a theory does not require every feature to be verifiable; a robust and varied set of confirmed predictions is enough. If a theory with strong support and a tight mathematical structure with no room for cherry-picking implies other universes, you must take them on board. Imagine string theory confirmed by stringy accelerator signals, extra dimensions, CMB features and cosmic strings, and shown to "absolutely, positively, incontrovertibly" generate the Landscape Multiverse: "the time for 'giving in' had arrived." In that context, failing to accept a multiverse would itself be nonscientific prejudice.
Greene declares himself a realist (mathematical theories give direct insight into reality), not an instrumentalist (theories merely predict what instruments register); philosophers of other orientations would weigh these points differently.
Science and the inaccessible II: where we stand in practice
| Multiverse | Underlying theory | Status of that theory | Verdict |
|---|---|---|---|
| Quilted | General relativity + infinite space | GR is confirmed but allows, not requires, infinite space; eternal inflation would require it but is unproven | Tentative |
| Inflationary | Eternal inflation | Data support inflation, but say nothing about whether it is eternal; some models give one universe | Tentative |
| Brane, Cyclic, Landscape | String theory | No testable predictions yet; unclear whether branes or the many shapes will survive refinement | Speculative |
So none of the multiverses can yet ride on an independently confirmed theory. Their prediction of other universes must be confronted head-on.
Predictions in a multiverse I: can inaccessible universes still help predict?
To many, a multiverse is an admission of failure: if the constants vary, calculating the value of the electromagnetic force is as meaningless as asking a pianist to pick out the note. Greene, who spent decades hoping string theory would compute every constant, empathises. But variation need not kill prediction. It comes down to statistics.
Predicting dogs. Dogs range from Chihuahuas under two pounds to Old English mastiffs over two hundred. Know your neighbourhood's breed mix, weight spreads, and how often each breed is walked, and you can predict the most likely next dog. If 80% are 60-lb Labradors and 20% assorted 30-lb breeds, bet on 55 to 65 lb. If 95% are 62-lb Labs, bet with confidence. The next dog may be a fluffy shih tzu, but probably not.
Three ways a multiverse can be predictive:
- Invariant features. If every universe shares a particular set of particles with fixed masses and charges, failing to find them falsifies the whole theory, multiverse and all.
- Skewed distributions. If most universes have Λ close to ours, our value is "nothing special", no more surprising than meeting a 62-lb Lab on a stroll.
- Tight correlations. If every universe with Λ in our range also has, say, a particle 5,000 times the proton's mass (beyond 20th-century accelerators, within reach of 21st-century ones), finding or failing to find it tests the theory.
The essential feature: a predictive multiverse is not a grab-bag; its physical properties are distributed in a sharply skewed or highly correlated pattern.
Predictions in a multiverse II: where we stand in practice
Dog counts can be read off a licensing authority's records; multiverses have no census bureau. We must compute the distribution from how the multiverse is populated. In the Landscape Multiverse the engines are inflation and quantum tunnelling; tunnelling is random in time and place, but its rates depend on landscape features (peak and valley heights), so in principle one could compute how often each kind of universe arises. "The rub is that so far no one has been able to do so." The landscape is too vast and rugged; studies are rudimentary. Proponents therefore add one more ingredient: anthropic selection.
Predictions in a multiverse III: anthropic reasoning
Most universes will be lifeless, and we can't find ourselves in them. If a multiverse had a unique life-friendly universe, we'd compute its properties and compare: a clean test. More plausibly there are many, so Weinberg, Linde, Vilenkin, Efstathiou and others propose counting observers rather than universes: some universes host the odd "cactus in a harsh desert", others teem with life. Observer census data predict what a typical observer, you, should see.
Weinberg, Martel and Shapiro (1997). For a multiverse in which only Λ varies, they used galaxies as the proxy for life:
- Negative Λ: collapse before galaxies, so discard.
- Λ above ≈200 × critical density (≈ 10−27 g/cm3, ≈ 10−121 Planck units): galaxies blocked, so discard.
- Within the window: near the upper limit Λ acts like a strong wind blowing dust piles apart, so few galaxies form; near zero, many form. So most observers see small Λ, few see Λ near the limit.
- Result: the typical observer sees Λ somewhat larger than measured but in the same ballpark; roughly 1 in 10 to 1 in 20 observers see a value like ours (~10−123).
A mismatch of 120 orders of magnitude becomes about as surprising as meeting a shih tzu in a neighbourhood of Labs.
For: proponents cite Weinberg-Martel-Shapiro as a success of anthropic reasoning.
Against: Max Tegmark and Martin Rees noted that if the size of primordial quantum jitters (the seeds of galaxies) also varies, bigger seeds can withstand bigger Λ, as bigger piles of leaves survive a brisker breeze. Then a typical observer sees a larger Λ, and the fraction seeing ours drops, potentially sharply.
Greene's verdict: the debate is premature. These are exploratory first passes; results are highly sensitive to assumptions, so real predictions need the actual variation of properties derived from theory, not assumed. Not yet achieved.
Predictions in a multiverse IV: the three hurdles
What a predictive multiverse requires:
- Know what varies, and how it is distributed. That needs the population mechanism (e.g. Bubble nucleation in the landscape). With luck the distribution, over all universes or over life-bearing ones, is sharply skewed.
- Justify typicality. Anthropic predictions assume we are average observers, Vilenkin's principle of mediocrity. If our observations match the average, confidence grows in both us-as-typical and the theory; if not, either the theory is wrong or we are atypical (even in a 99% Labrador neighbourhood you can meet a Doberman), and telling these apart may be hard. Better understanding of how intelligent life arises is needed; Weinberg's assumption that observers ∝ galaxies is reasonable (life needs a warm planet, a star, a galaxy) but tentative.
- Divide up infinity: the measure problem. Perhaps the last hurdle standing.
Dividing up infinity
Three Labs and one dachshund: Labs are 3× as likely. Likewise 300 and 100, or 3 million and 1 million. But infinitely many of each? Is three times infinity bigger than infinity? In the Inflationary Multiverse this is unavoidable: bubbles form forever, and each bubble contains an infinite Quilted Multiverse.
Monty's offer on Let's Make a Deal. You've won infinitely many envelopes: $1, $2, $3, …. Monty offers to double each. Envelope by envelope, the deal is obviously good. Collection to collection, you'd go from all whole numbers to only even numbers, apparently losing the odd amounts. "Compared envelope by envelope, the deal looks good. Compared collection to collection, the deal looks bad."
| Comparison method | Whole numbers vs even numbers | Conclusion |
|---|---|---|
| Subset | Evens are half the wholes | More wholes |
| Pairing n ↔ 2n (Monty doubles: same envelopes) | Every whole matched to exactly one even | Equal |
| Pairing n ↔ 4n (Monty quadruples) | Every whole matched to a multiple of 4, leaving 2, 6, 10, … unmatched | More evens |
There is no absolute answer: the size comparison depends on how you compare. So unless physics picks out a precise basis for comparison, we cannot say which kinds of universes, galaxies or observers predominate, and therefore cannot say what a typical observer should see. Various proposals (analogues of the pairings) exist; they give different answers, and none is derived and agreed. This is the measure problem: "Solving the measure problem is imperative."
A further contrarian concern: does eternal inflation undermine inflation?
Inflation's case rests on three successes: uniform temperature, flat space (expansion irons out curvature) and CMB ripples from stretched quantum jitters. The ripple predictions are probabilistic, but we can't rerun the big bang, so we read "99% probable and observed" as strong support, meaning 99% of universes produced by the same physics would look like ours.
The contrarian: in an infinite multiverse, 99% of infinity is infinity and 1% of infinity is infinity. Every outcome, even one with a 0.0000000001% quantum probability, happens in infinitely many universes. Without a measure, the very predictions that built confidence in inflation are compromised.
The optimist: the spectacular CMB agreement must reflect a deep truth: a measure exists in which anomalous universes have tiny measure. Deriving it is an immense challenge, but most researchers believe the data mean we will one day succeed.
Mysteries and multiverses: explanatory power
Traditional physics predictions are exquisitely precise: Mercury's perihelion, the energy of a uranium nucleus splitting into barium and krypton, the electron's magnetic moment, calculated as 2.0023193043628 and measured as 2.0023193043622, agreement better than 1 part in 10 billion. Multiverse predictions will likely say only that a quantity is "highly likely" to lie in some range, unless we are lucky enough to find extreme skew (say 99.9999% of observers see our Λ) or astonishing correlations (electrons exist only where Λ = 10−123). For some physicists that loss of precision is an unacceptable price.
Greene changed his mind. He long held that position, but now accepts that some features of nature may lie beyond precise prediction. A theory's equations churning out the electron mass (0.000000000000000000000091095 Planck masses), the top quark mass (0.0000000000000000632) or Λ = 10−123 "strains even the optimist's optimism," and string theory seems no closer to calculating any of them than when he started. A well-developed multiverse gives a sharp vetting of which single-universe mysteries persist and which are artefacts: if Λ varies finely across universes, its value goes from mysterious to prosaic, "explained away" like the shoe shop that stocks every size.
Newton's apple, the cautionary tale. Imagine Newton deciding that apples fall up in some universes and down in others, so a falling apple merely tells us which universe we're in; or that up-falling apples have already left for deep space. Absurd, since there was never reason for such thinking, but the point is serious: multiverse reasoning could sap the drive to solve mysteries that have ordinary explanations, tempting us to quit when "harder work and deeper thinking" was needed. Hence a serious multiverse proposal must be strongly motivated by theory and must specify its universes precisely. Yet turning away from a multiverse because it could lead to a blind alley is equally dangerous: "we might well be turning a blind eye to reality."
Going deeper: the technical notes
- Collisions happen at birth. Bubble walls are where the inflaton is high, which inside each bubble means early times. So collisions occur near each universe's big bang, which is why their ripples can affect the CMB.
- Where probability waves live. For n particles the wave lives not in ordinary 3-D space but in configuration space, with three axes per particle (3n in all). This is one sense in which probability waves "slither outside the arena of everyday reality."
- A shared feature: negative curvature. In the Inflationary Multiverse, every bubble's equal-inflaton slices are saddle-shaped, so all bubbles have negative spatial curvature, though inflation can make it unmeasurably small. Freivogel, Kleban, Rodríguez Martínez and Susskind (2006) pointed out a test: measuring small positive curvature at about 1 part in 105 would count strongly against landscape tunnelling. (Current limits, |Ωk| ≲ 0.002, are not yet that sensitive.)
- What Weinberg-Martel-Shapiro left out. They didn't commit to a specific multiverse, so they weighted universes only by galaxy count, assuming every Λ in the narrow life-friendly range (0 to about 10−120) is equally likely a priori, like assuming 59.99997-lb and 59.99999-lb dogs are equally common. A complete calculation needs the actual distribution from a definite multiverse theory.
- Typicality depends on the yardstick. A "typical American family" changes if you count cars, love of opera or political engagement. A "typical observer" likewise depends on whether we weight by population, galaxies, brain-hours or something else.
- Levels of infinity. Cantor showed infinities come in sizes: the whole numbers have cardinality ℵ0, the reals a larger one, and the set of subsets of reals larger still. All of Greene's examples are ℵ0, equal in Cantor's sense, which is precisely why cardinality is useless for physics. The standard strategy is to compare finite subsets and take a limit; the problem is choosing a physically justified cutoff that gives stable answers.
- Inflation and monopoles. Grand unified theories predict copious magnetic monopoles in the early universe; none are seen. Inflation's dilution explains their absence, another of its successes.
- How deep is the measure problem? Some see it as a technical detail; Steinhardt thinks solving it would require a different theory altogether; Greene (with a small but growing group) suspects it reaches "a deep problem at the very root of physics" that may require overhauling foundations.
- Small negative Λ. If negative but tiny enough, the universe's collapse could come late enough for galaxies; Greene glosses over this edge case.
Critical lens on the chapter.
- The analogy to black-hole interiors is weaker than it looks. We believe in black-hole interiors because GR is independently confirmed and the interior is a continuous extension of verified regions. Most multiverses rest on unconfirmed extensions (eternal inflation, string vacua), which Greene concedes in practice but which makes the "in principle" argument less reassuring.
- Falsifiability is only partly addressed. Greene's footnote admits the worry that with 10500 universes some universe fits any data, but the skewed/correlated rescue is hypothetical; no actual multiverse model has produced such a distribution.
- Typicality can be gamed. Choosing the "reference class" of observers (all observers? Galaxies? Brains? Computations?) changes predictions, and the Boltzmann brain problem (random fluctuations producing isolated observers who would outnumber evolved ones under some measures) is a known pathology that the chapter does not mention.
- The precision trade-off is philosophically heavy. Accepting "explained away" as explanation shifts what counts as success in physics; critics such as George Ellis and Joseph Silk ("Defend the integrity of physics", Nature, 2014) argue this risks lowering the bar of empirical testing.
2026 update
- Bubble-collision searches: Feeney, Johnson, Mortlock and Peiris (2011) searched WMAP data with a dedicated algorithm and found no evidence of collisions; Planck-era analyses also found none. Absence constrains only the expected number of observable collisions; it does not refute the Inflationary Multiverse.
- "Non-empirical confirmation": philosopher Richard Dawid (String Theory and the Scientific Method, 2013) argued theories can gain credibility from "no alternatives" and "unexpected explanatory coherence" arguments; a 2015 Munich conference ("Why Trust a Theory?") aired the dispute between Dawid's view and critics such as Ellis, Silk and Carlo Rovelli. It remains unresolved.
- The measure problem is still open. Leading proposals (causal-patch and fat-geodesic measures, scale-factor cutoff, watcher measures) still disagree, and each must avoid Boltzmann-brain domination and the "youngness paradox".
- Electron magnetic moment: 2023 Northwestern measurement g/2 = 1.001 159 652 180 59(13), the most precise measurement of any property of an elementary particle, still in agreement with QED given the measured fine-structure constant (though two atomic-physics determinations of α differ at the 5σ level).
So what. The chapter is a model for governing models whose internals cannot be inspected. An enterprise ML model, an LLM's latent features, a counterparty's internal risk process: you trust the architecture because its observable predictions keep coming true across a varied set of tests, never merely because it is elegant. And Monty's envelopes carry a practical warning for metrics: ratios over unbounded or ill-defined populations depend on how you count ("conversion rate per what?"). Fix the measure (the denominator) before comparing outcomes, or the comparison is meaningless.
The Many Worlds of Quantum Measurement: The Quantum Multiverse
Quantum mechanics predicts probabilities superbly, but its central equation never makes a probability wave "collapse" to one outcome. Bohr's Copenhagen approach simply suspends Schrödinger's equation during measurement, without saying what counts as a measurement. Hugh Everett III took the equation at its word: apply it to everything, measuring devices and brains included, and every possible outcome is realised, each in its own decohered branch. Greene argues this is the most conservative formulation of quantum theory, not the most extravagant. Its deepest open problem is probability: if every outcome happens, what does it mean that one is "more likely"? Answers range from Everett's subjective ignorance to the Oxford decision-theory programme, and no consensus exists.
A different kind of multiverse
The previous multiverses await their verdict. This one has already been judged by many physicists; they just haven't all reached the same verdict. The split comes down to the unresolved problem of getting from quantum probabilities to the definite reality of experience.
Quantum reality and Everett's story
- 1925 to 1926: Heisenberg's matrix mechanics (1925), Schrödinger's equation (1926), Max Born's probability interpretation (1926; Nobel almost three decades later, 1954).
- 1954: Princeton graduate student Hugh Everett III sees a "gaping hole" Bohr danced around.
- January 1956: Everett gives his draft thesis to adviser John Wheeler, who is impressed.
- May 1956: Wheeler visits Bohr in Copenhagen; the reception is icy.
- March 1957: under Wheeler's pressure (Wheeler revered Bohr), Everett submits a trimmed thesis, cutting criticism of Bohr and presenting his work as an extension of orthodoxy; he needed the PhD for a Defense Department job (he later shaped nuclear-weapons policy for the Eisenhower and Kennedy administrations).
- April 1957: Princeton accepts it. July 1957: published in Reviews of Modern Physics, and ignored.
- c. 1967 to 1970: Bryce DeWitt, inspired by his student Neill Graham, revives it; his 1970 Physics Today article candidly describes his "shock" at Everett's enormous "multiworld", igniting a debate still running.
The quantum break with classical physics
Classical physics is deterministic: tell me how things are now and I'll tell you any past or future, chaos and complexity aside. In the microworld this fails. Put one electron in each of 100 identically prepared boxes, measure all positions after 10 minutes, and you get 100 different answers. The regularity appears only statistically: if one batch finds 27% near the lower left corner, 48% upper right and 25% in the middle, the next batch will show nearly the same split. Quantum mechanics predicts probabilities, not outcomes. In the more than eighty years since, not a single verifiable experiment or observation has contradicted those predictions.
The puzzle of alternatives: the double slit
A lucky accident at Bell Labs, April 1925. Clinton Davisson and Lester Germer were firing electrons at nickel when a glass tube exploded and tarnished the sample. Rather than replace it, they heated it clean, which fused many small nickel crystals into a few large ones. Electrons now bounced from a few concentrated spots, and the data were unlike anything seen before. By 1927 it was clear they had shown electrons behave as waves; a Nobel Prize followed within a decade (Davisson, 1937). Their setup is the ancestor of the double-slit experiment.
- One slit open (left or right): electrons pile up in a single bright band behind that slit.
- Both slits open: not the sum of the two, but alternating bright and dark bands. Dark bands appear where either slit alone would put many electrons.
- The paradox: each electron passes through one slit, yet the presence of the other slit changes where it can land, as if adding a second door to an office building made your office unreachable. Each electron somehow "knows" about both slits.
Probability waves
The light-dark pattern "screams waves." Two pebbles in a pond make overlapping ripples: crest on crest reinforces, trough on crest cancels, giving an interference pattern. In 1926 Max Born combined the two clues (waves and statistics): the wave associated with a particle is a probability wave. Where its values are large, the particle is likely to be found; where they vanish, never. The electron's wave passes through both slits; the two fragments interfere; the resulting high and low values set where electrons land, building up the bands one dot at a time. Photons, neutrinos, muons, quarks: every fundamental particle is described by a probability wave.
A simplification to note. Greene says probability is set by the wave's "height" or "size". Precisely, the Born rule says probability is proportional to the squared magnitude of the wave (|ψ|2), and the wave's values are complex numbers. Interference arises because amplitudes add before squaring. The simplification does no harm to the argument but matters for the probability debate later in the chapter.
Not so fast: three questions
Question 1. Why does Newton work for big things? Their probability waves are extraordinarily narrow: near-100% probability at the peak, and the peaks follow Newton's trajectories. Over billions of years you'd essentially never see a macroscopic deviation. Small things have spread-out waves, hence quantum strangeness.
Question 2. Can we see the probability waves? No. In the Copenhagen interpretation (Bohr and colleagues), looking (measuring position) makes the wave collapse: it surges to 100% at one spot and drops to 0 everywhere else; look away and it spreads again. Like Lucille, who claims to be blonde until someone looks, then turns redhead, it sounds like snake oil, but it is testable: if the wave is 2.874 times bigger here than there, you find the particle here 2.874 times as often. Predictions like this have been "enormously successful."
Question 3, the bear. What happens during measurement? Schrödinger's equation cannot produce collapse, the sudden spiking of a wave at one point "like a lone parishioner standing in a megachurch while everyone else kneels." Bohr's remedy: use Schrödinger's equation except when you look, then declare collapse.
Bohr's prescription is ungainly, arbitrary, mathematically baseless, and not even clear. What counts as "looking"? A human? "A sidelong glance from a mouse", as Einstein asked? A computer probe, a bacterium, a virus? Bohr drew a line between small things (Schrödinger applies) and big things (it doesn't), but never said where. Experiments keep confirming Schrödinger's equation for ever-larger collections, "like floodwaters rising from your basement into your living room and threatening your attic." You, your apparatus and your brain are made of particles that obey Schrödinger's equation, so a measurement is just one particle collection meeting another. If the equation never stops, waves never collapse. Then how does one definite outcome arise?
Linearity and its discontents
Schrödinger's equation is linear: the whole is the sum of its parts. Split a wave into pieces, evolve each separately, add the results, and you get the same answer as evolving the whole. This divide-and-conquer property underlies most quantum calculations (it's how we got the double-slit pattern). It is also the root of the measurement problem.
Electrons in a tabletop Manhattan.
- One spike: the electron's wave is spiked at Thirty-fourth Street and Broadway (Macy's). A well-made device, its trillion trillion atoms engineered for the job, reads "Thirty-fourth and Broadway". Likewise for the Hayden Planetarium (81st and Central Park West) or Bill Clinton's office (125th near Lenox).
- Two spikes: the wave is spiked at Strawberry Fields (John Lennon's memorial) and Grant's Tomb ("the electron's in one of its dark moods"). Linearity says: combine the two single-spike results. Naively, the display shows both names flashing at once, like a computer about to crash, and since photons from the display entangle with your retina and neurons, your mind registers both locations at once.
- Many spikes: a smooth wave is a union of spikes at every location, so the device would register every location and your mind would be profoundly puzzled.
No working device or person ever experiences this. Bohr: "Hold the Dramamine": measurement suspends Schrödinger, collapses all but one spike (chosen with probability set by its height), "dizziness done."
For Everett, and later DeWitt, that cost was too high: Schrödinger's equation describes all particles; why would it fail for the particular arrangements we call devices and experimenters?
Many Worlds
Everett's insight: the confusion comes from combining the results sloppily. Measuring a single spike at Strawberry Fields gives a device and a mind registering Strawberry Fields; a spike at Grant's Tomb gives a device and a mind registering Grant's Tomb. Combine them carefully and literally: you get a device and a mind registering Strawberry Fields, and a device and a mind registering Grant's Tomb. Not one confused mind, but two clear ones. In broad strokes, you, the device and everything else "split" into two realities, each experiencing one sharp result. Five spikes give five worlds; a further seven-spike measurement splits each into seven more; a smooth wave gives a vast number. Everything quantum mechanics deems possible (nonzero probability) is realised in its own world: the Quantum Multiverse, the sixth multiverse.
A tale of two tales
Greene admits "split" is a sledgehammer word that Everett and DeWitt used, and that he now has to repair the damage. He insists that Many Worlds is not extravagant speculation but, in some ways, the most conservative framework for quantum physics. The key is that physicists always tell two stories:
| Story 1: the mathematics | Story 2: the experience | |
|---|---|---|
| Newton | x is the ball's position, v its velocity | Identical: every symbol has a transparent physical meaning |
| Many Worlds | Schrödinger's equation always governs the probability wave. Period. No splitting, no worlds, no multiverse appear in the equations | When the evolved wave has several well-separated peaks, each peak describes a device with a definite reading and an observer who has seen it: each peak is a world |
- The scale of the task. You and your device contain about 1027 particles; about 1024 in the display rearrange, "like performers in a crisply choreographed halftime show", to spell the reading. Greene's undergraduates struggle to solve Schrödinger's equation for one particle.
- Picturing it. A proper probability wave for 1027 particles needs three axes per particle, about a billion billion billion axes. Greene draws each axis as a bundle of axes and gives the many-particle wave a glowing outline. A peaked many-particle wave means each particle sits, almost certainly, where it belongs (hand particles in the hand, device particles in the device), not all at one spot.
- The measurement in story 1: the joint wave of electron, device and you evolves from one peak into two peaks: one corresponding to "Strawberry Fields" in display and brain, one to "Grant's Tomb". Nothing distinguishes the peaks except the reading; neither is more real.
- Include everything: add the lab, the earth, the sun; they just "come along for the ride." Each peak is what we'd call a bona fide universe.
There is no cloning machine and no splitting machine, only "a probability-wave-evolution machine driven by the lean mathematical law of quantum mechanics." When the wave takes a multi-peaked shape, the story-2 retelling finds in each peak a sentient being, certain they saw one definite result. Interview them all: exact replicas except for the result each reports. Everett, in a "note added in proof" slipped in apparently without Wheeler's notice, made his stance unmistakable: "From the viewpoint of the theory, all ... Are 'actual,' none any more 'real' than the rest."
When is an alternative a universe? Decoherence
A single electron with a two-spiked wave is one world with an electron of ambiguous position, not two worlds. What changes on measurement? The quick answer is that with no observer there is no story 2 to tell. The deeper answer is decoherence.
- Double slit, no detectors: the two wave fragments intermingle and interfere; calling them separate universes makes no sense, since their mixing is the interference pattern.
- Put a meter behind each slit: now the two paths differ in billions and billions of particles in the meter displays. The two branches' waves become so different they no longer overlap, so they cannot interfere. The striped pattern disappears, replaced by the plain sum of the one-slit results. The waves have decohered.
- Consequence: decohered branches evolve independently and become mutually invisible; each "thinks" the others have vanished. Within each universe it is as if the wave collapsed.
Bohr: big things lie outside Schrödinger's law; measurement collapses all outcomes but one, by fiat. One outcome is real.
Many Worlds + decoherence: big things obey the same law, but their many particles make branches stop interfering. Each branch sees an apparent collapse. All outcomes are real. The one point of agreement with Bohr: big and small things do behave differently, though not for Bohr's reason.
A probable problem
The short version for readers stopping here: probability normally expresses ignorance of which outcome will happen. Many Worlds says all outcomes happen. How can some be "likely" and others "unlikely" when all take place? This is cutting-edge research, and opinions vary widely.
Objection 1: too baroque (answered)
Critics call it too extravagant to be true. Proponents reply that complexity should be judged by a theory's assumptions, not its implications. Many Worlds assumes one equation, always in force: hard to beat for economy. Copenhagen adds a vague rule for switching the equation off and an even vaguer collapse to replace it. A rich reality is no more a mark against Everett than the diversity of life is against Darwin's simple mechanism. Occam's razor isn't sharp enough to cut Many Worlds.
Objection 2: probability (the real problem)
A physicist must tell a third story: why we trust the theory. For quantum mechanics: when it predicts one outcome 9.62 times as often as another, that is what experiments find; had they not, confidence would fall (with caution: even a fair coin tossed 1,000 times has odd runs). A theory must be vulnerable to data in this way.
Many Worlds seems to undercut this. If a coin lands heads in one world and tails in another, with a Brian Greene in each, where is the 50:50? There is no chance anywhere: waves evolve deterministically, "no dice are thrown; no roulette wheels are spun." Copenhagen brought probability in through collapse; Many Worlds abolishes collapse and with it the traditional entry point for probability.
The tempting fix, "only one of the copies is really me", smuggles Copenhagen back in through the back door. Greene had that reaction himself years ago, but it erases the reason for Many Worlds.
Everett's answer: subjective ignorance
The Zaxtarian cloning offer. Aliens from Zaxtar will clone you overnight. One copy wakes to a normal life plus any wish granted; the other is taken to a torture chamber on Zaxtar forever. Most refuse: each clone really is you, so a you will be tortured. They raise the stakes: 1,000,001 copies, a million with wishes, one tortured. You waver.
- Wrong reason to waver: "the odds are it won't be the real me." Each copy is you; a tortured you is 100% certain.
- Right reason: after the cloning, each copy, eyes still shut under a warm duvet, can ask "what are the odds that I am Zaxtar-bound?" Each knows only one in a million and one is, so each reckons the odds low; a million are proved right.
Probability enters with no dice at all, through each copy's subjective ignorance of which outcome it will witness.
Everett called his theory "objectively deterministic" with probability "reappearing at the subjective level," and thought it reconciled Einstein (no dice in fundamental physics) with Bohr (probabilities explain observations). Greene's image: Einstein soaring above the many worlds, watching Schrödinger's equation dictate everything, concluding that God doesn't play dice, while Bohr, inside one world, uses probabilities with stupendous precision. (Greene notes this non-chancy view argues for dropping "probability wave" for the technical term "wavefunction".)
The branch-counting failure
Those who study Everett's thesis agree his intent was clear but the execution unconvincing. The natural rule, "probability ∝ number of yous who see the outcome", works when spikes are equal (one you per outcome, 50:50 as quantum theory says). But if the wave is 100 times larger at Strawberry Fields than at Grant's Tomb, quantum mechanics predicts 100:1, while there is still one you per outcome: 50:50, the wrong answer. The number of yous follows the number of spikes; quantum probabilities follow their heights. Everett and others argued that low-amplitude worlds should get less weight, but is the Strawberry Fields world "a hundred times as genuine" as the other? That sits uneasily with all worlds being equally real.
The Oxford programme
David Deutsch, Simon Saunders, David Wallace and Hilary Greaves ask: how should a rational gambler who believes Many Worlds bet on quantum experiments? Their mathematical answer: exactly as Bohr would. Averaged over all the inhabitants who claim to be you (a notion that would have sent Bohr "into a tizzy"), the numbers physicists call quantum probabilities are the right weights for decisions. The theory is deterministic, but you should treat the numbers as if they were probabilities. "Some are convinced that this completes Everett's program. Some are not."
Probability is slippery even classically: 1,200 rolls of a die should give "about 200 threes", but the count will likely deviate, so "it is highly probable that a sixth are threes" defines probability using probability. Add "you" referring to many people, and it is no wonder the debate continues. Greene expects clarity eventually, "but not yet, and perhaps not for some time."
Predictions and understanding
Practically, there's a "quantum algorithm": evolve the wave with Schrödinger's equation, read probabilities off the wave, compare with data. It works whatever happens "really." Some say that's all physics is. Greene offers three rebuttals:
- Coherence. A theory must be mathematically complete. Copenhagen "retreats into mathematical silence" at the moment of observation; Many Worlds tries to fill the gap.
- Different predictions. Copenhagen collapses the two-peaked wave of you-plus-device to one peak. If one could make macroscopically distinct branches interfere, that would prove collapse didn't happen. Decoherence makes this extraordinarily hard, but in principle they predict different things, so calling them "interpretations" is "an abuse of language."
- Understanding beats prediction. A black box that predicted every experiment would not end physics. "There's a difference between making predictions and understanding them." Physics' raison d'être is insight into why.
Greene does not expect consensus in his lifetime on whether quantum mechanics describes one universe, many, or something else.
Going deeper: the technical notes
- Amplitudes, not probabilities. Schrödinger's wave takes complex values; probability is the squared magnitude, ψ*ψ. Interference requires values that can cancel, which positive probabilities could not do.
- Configuration space. For n particles in m spatial dimensions the wavefunction is a complex function on an mn-dimensional space. "Sharply peaked" means concentrated in a small ball there; a big object's wavefunction can be peaked (each particle nearly definitely located) while its particles are in different places. Newtonian motion emerges for the centre of mass.
- Finitely many worlds? Wave values vary continuously, so "infinitely many worlds" is common loose talk. But the wave lives in a Hilbert space; if that space is finite-dimensional (believed true for finite volume and energy), measurements have finitely many distinct outcomes, and finitely many worlds per finite sequence of measurements.
- Wheeler's warning to Everett: "get the bugs out of the words, not the formalism," and beware "the difficulty of expressing in everyday words the goings-on in a mathematical scheme that is about as far removed as it could be from the everyday description." In 1959 Everett went to Copenhagen to make his case at Bohr's blackboard; a hoped-for two-week showdown amounted to a few unproductive conversations. No minds changed.
- The frequency-operator result. For n copies of a system, as n → ∞, the joint wavefunction approaches an eigenstate of the frequency operator with eigenvalue |ck|2: the fraction of outcome k matches the Born rule, except in branches whose total amplitude-norm goes to zero. Found independently by Graham, DeWitt, Hartle, Farhi-Goldstone-Gutmann, Deutsch, Coleman, Albert and Greene himself. The catch: for a 98:2 spin state, the vast majority of branches (counted equally) show near 50:50 statistics, and only low-count, high-norm branches show 98:2. Calling zero-norm branches "nonexistent" is exactly the unresolved step. Greene reads the result as strong motivation for, and a consistency check on, the Born rule rather than a derivation.
- Vaidman's timing proposal. Probability enters in the window after the measurement but before you read the result. Critics say this is too late (science must predict) and fragile (read the result instantly and probability vanishes). And Greene's twist: if the only difference between "me" and "the other Brian" is the screen reading, I couldn't have seen the other outcome without being someone else, so my ignorance may not support probabilistic reasoning at all.
- The alternatives (Greene's notes). GRW (Ghirardi, Rimini, Weber): modify Schrödinger's equation so collapse happens spontaneously, negligibly for single particles, rapidly for big objects. De Broglie-Bohm: particles always have definite positions guided by the wave via a second equation, v = (ħ/m) Im(∇ψ/ψ), derived from probability conservation; it is non-local (each particle's velocity depends instantly on all others) and hard to make relativistic. Greene admits he'd like Many Worlds to be right, but stays open.
- An in-principle test. Start with an electron spin-up along z; measure spin along x; then reverse the entire evolution (apparatus included; the equations are time-reversible) and measure z again. Many Worlds predicts spin-up with 100% certainty; a coherent collapse theory predicts 50%. Practically impossible, but it shows the approaches are empirically distinct.
Critical lens.
- The alternatives get short shrift. Greene frames the choice as Copenhagen versus Everett. But other serious, mathematically complete options exist: Bohmian (pilot-wave) mechanics (particles with definite positions guided by the wave; one world), objective-collapse theories such as GRW and Diósi-Penrose (Schrödinger's equation modified so large superpositions collapse spontaneously; testable), and epistemic or relational views (QBism, Rovelli's relational QM). Each dissolves the measurement problem differently, and several make different predictions. They are mentioned only in the notes, if at all.
- The preferred-basis problem. Decoherence explains why branches stop interfering in a given basis, but why the world splits along position-like definite outcomes rather than other decompositions needs further argument (Zurek's "einselection" is the usual answer). The chapter treats this as solved.
- "Most conservative" is contestable. Economy of postulates is one virtue; ontological extravagance and the unsolved probability problem are costs. Reasonable physicists weigh them differently.
- Born rule simplification. "Wave height" rather than squared amplitude hides why branch counting fails so badly: the mismatch is between counting and |ψ|2.
2026 update: quantum foundations since 2011
- Bell tests closed: three loophole-free Bell tests (2015) (Delft, Vienna, NIST) confirmed quantum non-locality; Aspect, Clauser and Zeilinger won the 2022 Nobel Prize. Every interpretation must accommodate this; Many Worlds does so without non-local collapse.
- Superpositions keep growing: matter-wave interference with molecules of about 2,000 atoms (≈25,000 amu) (Vienna, 2019); a 16-microgram sapphire crystal placed in a "cat state" of motion (ETH Zürich, 2023). Bohr's line keeps retreating, as Greene predicted.
- Collapse models squeezed: underground radiation searches at Gran Sasso (2020) ruled out the simplest parameter-free Diósi-Penrose gravity-collapse model; GRW parameters are increasingly constrained. They are not yet excluded.
- Born rule in Many Worlds: Sebens and Carroll (2014 to 2018) derived it from self-locating uncertainty (Greene's Zaxtar intuition, made rigorous); Wallace's The Emergent Multiverse (2012) is the definitive statement of the Oxford programme. Critics remain unconvinced; the issue is still live.
- Extended Wigner's-friend arguments: Frauchiger and Renner (2018) and follow-ups ("Local Friendliness" experiments, 2020) show that quantum theory, single outcomes for observers, and locality can't all hold, sharpening exactly the choice Greene describes.
- Quantum computing: Deutsch has long argued quantum computers draw on many worlds; Google's Willow chip (December 2024) showed error rates falling as codes grow (below threshold), and Google's announcement explicitly invoked the multiverse. Most physicists stress that quantum computation's success is interpretation-neutral.
So what. Greene's two-stories discipline (separate what the formal model says from how we narrate it to users) is directly useful in AI governance. A model's math (logits, embeddings) is story 1; "the model believes / decided / hallucinated" is story 2. Many confusions in AI risk discussions are story-2 words smuggled into story 1, just as "splitting" misleads about Everett. And the measurement problem has a systems analogue: a component that behaves one way "unless observed" (logging changes timing, monitoring alters load) is a design smell. Prefer architectures where observation is part of the same model, not an exception to it.
Black Holes and Holograms: The Holographic Multiverse
The strangest multiverse in the book is not a set of distant regions but a second, complete description of reality living on a distant boundary. Greene follows a thirty-year chain: Wheeler's view that information is fundamental; Bekenstein's claim that black holes carry entropy; Hawking's proof that they radiate and that their entropy equals their horizon area in Planck units; the deduction that the information any region can hold is bounded by its surface area, not its volume; and Maldacena's 1998 discovery that string theory in a ten-dimensional bulk is exactly equivalent to a gravity-free quantum field theory on its four-dimensional boundary. That holographic duality already helps calculate real quark-gluon plasma properties, and may serve as the very definition of string theory.
Plato's cave turned inside out
Plato likened our perceptions to shadows on a cave wall, faint hints of a richer reality. Two millennia later, the cave may be more than metaphor, and inverted: reality may take place on a distant boundary surface, with everything we experience in three dimensions a projection of it, like a holographic movie. The holographic principle says that if we understood the laws on that surface and how they link to experience here, we would grasp all there is. The route runs through general relativity, black holes, thermodynamics, quantum mechanics and string theory, and the thread tying them is information.
Information
Lunch with Wheeler, Princeton, 1998. Greene asked John Wheeler (mentor of Everett, Richard Feynman, Kip Thorne and Jacob Bekenstein) what would dominate physics in the coming decades. Wheeler dropped his head; the silence was so long Greene wondered if he'd forgotten the question. Then he looked up and said one word: "Information."
Physics traditionally studies things: planets, atoms, fields, and the forces among them. Wheeler argued that things are secondary carriers of something more fundamental: information (where a particle is, which way it spins, what its charge is). He did not call matter illusory; it is the material realisation, as a skyscraper realises the information in an architect's blueprints. The universe then becomes an information processor: it takes information about now and produces information about the next now. Whether or not that stance comes to dominate, puzzles about information in black holes, driven largely by Gerard 't Hooft and Leonard Susskind, have forced a major shift.
Black holes
- 1916: within a year of general relativity, Karl Schwarzschild finds the first exact solution of Einstein's equations, the spacetime around a spherical mass, while calculating artillery trajectories on the Russian front. Einstein had only approximate solutions; impressed, he presents the work to the Prussian Academy.
- The startling implication: cram enough mass into a small enough ball and spacetime curves so steeply that anything, light included, is trapped: early names were "dark stars" and, since time halts at the edge, "frozen stars." Einstein accepted the maths for stars and planets but scoffed at the extreme case.
- 1960s: Wheeler, "nearly as adept at marketing as at physics", popularises the name "black holes."
- Since then: overwhelming evidence that black holes are real and plentiful; most large galaxies have a giant one at the centre.
- Sun as a black hole: squeezed to a ball "about three kilometres across" (Greene's figure).
- Earth as a black hole: squeezed to about a centimetre across.
- Milky Way's central black hole: about three million solar masses (Greene's figure).
- LHC: might produce microscopic black holes if extra dimensions strengthen short-range gravity (Chapter 4).
Two numbers to correct. The sun's Schwarzschild radius is about 3 km, so its black-hole diameter would be about 6 km, not 3 km across. The Milky Way's central black hole, Sagittarius A*, is about 4.3 million solar masses (Genzel and Ghez, Nobel 2020); "three million" was an older estimate already superseded when the book appeared.
The Second Law, entropy and lions
"Less is more": "executive summary," "just the facts," "TMI." Meeting a lion on the savanna, you don't want the trajectory of every reflected photon, only whether it's crouched and coming at you. Some physics needs every detail: the LHC's detectors along its seventeen-mile tunnel record so much that a year's data would fill a stack of DVDs about fifty times the height of the Empire State Building. But thermodynamics, and its modern form statistical mechanics, deliberately sacrifices detail for aggregate understanding. A steam engine's push depends only on how many H2O molecules hit the piston and how fast on average, not on which molecule hit where.
Entropy, defined
Felix and Oscar. Felix cries burglary. Oscar opens his own room: clothes, pizza boxes and crushed beer cans everywhere: "It looks just like it always does." Felix: "Ransack a pigsty and you get a pigsty. But look at my room": vitamins not lined up by size, Shakespeare out of alphabetical order, black socks in the blue bin. Toss Oscar's 26 shirts and 42 beer cans around and nobody can tell; disturb Felix's room and anyone can.
Boltzmann's definition (1870s): entropy counts the rearrangements of a system's constituents that leave its macroscopic appearance unchanged (strictly, the logarithm of that number). Many rearrangements: high entropy (steam, Oscar's room). Few: low entropy (ice, Felix's room). The Second Law: total entropy increases, simply because high-entropy configurations vastly outnumber low ones. Baking bread fills the house with aroma because spread-out arrangements outnumber "all in one corner" by trillions. The law is statistical: your milk could swirl into a floating Santa Claus figurine, "but don't hold your breath."
Black holes threaten the Second Law
In the early 1970s Wheeler noticed a loophole: throw any disordered system into a black hole and its entropy seems to vanish. His student Jacob Bekenstein disagreed: perhaps entropy is transferred, not destroyed, just as energy thrown in is not destroyed but makes the black hole heavier. The experts "shot Bekenstein down":
- Ultimate order. Infalling mess is crushed to the centre ("the ultimate in orderly trash compaction"); outside it, a black hole is just empty spacetime to the horizon.
- No-hair theorems. Two black holes with the same mass, charge and spin are identical, like the bald performers of Blue Man Group: nothing to rearrange, so no entropy.
- Entropy implies temperature, and temperature implies radiation, yet nothing escapes a black hole. So black holes can't have entropy.
One result was on Bekenstein's side. In 1971 Stephen Hawking proved the area theorem: in any process, the total horizon area of black holes never decreases. Few laws in physics require a quantity to increase; the parallel with entropy was obvious, but most dismissed it as coincidence. Until Hawking's next calculation.
Hawking radiation
Hawking combined quantum fields with the fixed curved spacetime of a black hole: a partial union (spacetime itself not jittering), but enough.
- Pair production: in empty space, quantum jitters briefly create particle-antiparticle pairs (e.g. electron and positron) that quickly annihilate. Energy conservation means one member carries negative energy, tolerated only because the uncertainty principle hides its sign if it lives briefly enough.
- Near a horizon: if a pair forms close enough, one partner falls in and the other escapes. The black hole's extreme warping makes a particle that has negative energy as judged from outside have positive energy inside, so the infalling partner finds "a safe haven" and no annihilation is needed.
- Result: the escaping particles form Hawking radiation; the infalling negative-energy partners reduce the black hole's mass. Black holes glow like burning coal and slowly shrink. They are not completely black.
Hawking temperature is inversely proportional to mass (T = ħc3/8πGMk). Escaping radiation is "tired" after climbing out of the gravitational well.
- Galactic-centre black hole: less than a trillionth of a degree above absolute zero.
- Solar-mass black hole: less than a millionth of a degree, far colder than the 2.7 K cosmic background.
- Hot enough to barbecue dinner: mass about one ten-thousandth of Earth's.
- Entropy: from temperature and thermodynamics, Hawking derived the entropy, and it was proportional to horizon area, exactly as Bekenstein proposed. By the end of 1974 the Second Law was law again, with black holes as active contributors to growing disorder.
Entropy as hidden information
Entropy is also the gap between what you know (macroscopic features) and what you don't (the microscopic arrangement): the hidden information.
Oscar's thousand silver dollars (poker winnings scattered on the floor), gathered in a heap of heads and tails:
- Count the arrangements: 2 coins have 4 (HH, HT, TH, TT); 3 have 8; 1,000 have 21000 ≈ 10301. A handful are distinctive (all heads, 999 heads); the rest look the same.
- Take the logarithm: entropy = the exponent = 1,000.
- Count the information: information is measured by the number of yes-no questions it answers: "Is coin 1 heads? Is coin 2?" That's 1,000 bits (a bit, binary digit, answers one yes-no question).
- They match: the number of arrangements is the number of possible answer sequences, so entropy = hidden information = number of yes-no questions the microscopic details can answer.
Entropy, hidden information and black holes
Hawking's recipe: tile the event horizon with a grid of cells one Planck length (10−33 cm) on a side (each 10−66 cm2). The entropy is the number of cells: horizon area in Planck units. It is as if each cell carries one bit, a 0 or 1 answering one yes-no question about the black hole's microscopic make-up.
So general relativity and the no-hair theorems, ignoring quantum mechanics, miss a vast amount of information: many black holes share mass, charge and spin yet differ microscopically. Bekenstein and Hawking told us how much information is hidden but not what it is or what questions it answers.
The deepest puzzle: why area? The information in the Library of Congress depends on the space inside for shelves and microfiche; the information in your head on the volume of your brain. Physicists had assumed one bit per Planck cube (10−99 cm3), so storage should scale with volume. For a black hole it scales with the number of Planck squares on its surface. This was the first hint of holography.
Where is a black hole's information? Two perspectives
Einstein's equivalence principle: in free fall you are weightless, and inside a windowless capsule you cannot distinguish free fall from floating in empty space. So falling through a horizon, you notice nothing: no surface, no signposts, no flashing lights, until the centre's devastation. Then how can the horizon store information?
Your view (falling in): you float through the horizon uneventfully, carrying all your information (body, brain, laptop), then (less happily) hit the singularity.
Distant observer's view: she measures cold Hawking radiation (say 10−13 K, for a galactic-centre-size hole), knows it's cold only because it's tired, and infers that closer to the horizon it is ever hotter. She sees you bombarded by increasingly intense radiation until only charred remains are left, your information absorbed and scrambled in a sizzling layer just above the horizon. To her, the horizon is a real place populated by real things carrying the bits.
Susskind's answer: both are right (black hole complementarity). The two accounts can never be confronted: you can't climb out, and by the time she has decoded your demise from the radiation it is too late to jump in and catch you before the singularity. "Physics has a built-in fail-safe against paradoxes."
Beyond black holes: the holographic bound
Take any region (the Library of Congress, all of Google's computers, the CIA's archives) enclosed by an imaginary sphere, with ordinary density far from black-hole conditions. What is the most information it can hold?
The argument in four steps:
- Stuff it: add memory chips, hard drives, books, Kindles, and raw matter (whose molecules also carry information).
- Hit the threshold: at some point one more grain of sand turns the region into a black hole whose horizon coincides with the sphere. Add more and the black hole grows, overspilling the region.
- Invoke the Second Law: entropy rose throughout, so everything you packed in held less hidden information than the black hole.
- Invoke Bekenstein-Hawking: the black hole's information is its horizon area in Planck units, i.e. The area of the sphere.
So the information any region of space can contain, in any objects of any design, is always less than the area of its bounding surface in Planck units. This holds whether or not a black hole is present.
- Five terabyte hard drives fit in a sphere of radius 50 cm, whose surface holds about 1070 Planck cells: capacity about 1070 bits, about a billion trillion trillion trillion trillion terabytes. "No one in Silicon Valley cares much about these theoretical constraints."
The practical limit is irrelevant; the principle is not. If all the information describing physics inside a region can be fully encoded on its boundary, then, as Susskind and 't Hooft argued, perhaps the boundary is where the fundamental processes actually happen, and our 3-D reality is a holographic projection of 2-D physics. Processes on a distant surface, "much like a puppeteer pulls strings," would be fully linked to Greene's fingers typing at his desk. Holographic Parallel Universes: "as connected as me and my shadow."
The fine print
Each step is theory, untested by experiment: do black holes really have entropy and temperature with the predicted values? Is a region's capacity really set by its boundary? Is one bit per Planck area really the limit? Greene thinks yes because the conclusions fit a "coherent, consistent, and carefully constructed theoretical edifice," but admits a future advance could "lay to waste the holographic idea." And the argument concerns only the quantity of information. How is it stored on the boundary? What form does it take? Is there a dictionary between boundary and interior? 't Hooft doubted string theory could help: "Nature is much more crazy at the Planck scale than even string theorists could have imagined." Less than a decade later, string theory proved 't Hooft wrong by proving him right.
String theory and holography: Maldacena
The Maldacena Macarena (Santa Barbara, Strings 1998). At the conference banquet, physicists performed a song-and-dance version of the Macarena honouring Juan Maldacena ("Black holes used to be a great mystery; / Now we use D-branes to compute D-entropy"). Greene, one of the few whose talk wasn't about Maldacena's result, opened his talk the next morning by doing the Macarena moves on stage, bunny-hop and quarter turn included, to laughter. More than a decade later, many would agree no string theory work since has matched its impact.
Maldacena studied string theory in a universe with a different shape from ours but easier to analyse: one with a boundary, an impenetrable surface enclosing the interior. He argued convincingly that everything in the interior reflects laws and processes on the boundary: the first mathematical example of Holographic Parallel Universes.
How he did it: one stack of branes, two descriptions
Maldacena's argument:
- Set-up: a tight stack of three-branes, open strings (snippets) stuck to the branes, closed strings (loops) free in the bulk.
- Perspective A, intrinsic: look at low-energy strings. Low energy means low mass, so gravity is negligible, and open and closed strings decouple. The open strings on the branes are described by a quantum field theory (technically, a conformally invariant supersymmetric quantum gauge field theory) in four spacetime dimensions, with no gravity.
- Perspective B, extrinsic: treat the stack as a massive object that warps spacetime; enough branes form a black brane. Low-energy physics now includes any closed string near the black brane's horizon, since gravitational redshift makes it look slow and low-energy. These strings move in ten spacetime dimensions and include gravity.
- Equate them: both perspectives describe the same stack; the free bulk strings match, so the rest must match too: a gravity-free point-particle quantum field theory in 4 dimensions = string theory with gravity in a 10-dimensional curved region (anti-de Sitter five-space times a five-sphere, AdS5 × S5), whose boundary is the brane stack. Greene can think of no two things in the real world more dissimilar. "This is holography come to life."
Within months, Edward Witten and, separately, Steven Gubser, Igor Klebanov and Alexander Polyakov supplied a precise dictionary translating any boundary process into its bulk image and vice versa. Like an ordinary hologram, whose etched swirls look nothing like the 3-D image until a laser reconstructs it, the boundary theory bears no obvious resemblance to the bulk string theory; shown both, a physicist would think them unrelated. The dictionary is the laser, and information appears scrambled in translation.
Witten's black hole. A black hole in the bulk translates on the boundary into a bath of hot particles, nothing gravitational at all, as the Wizard of Oz's frightening face was projected by an ordinary man. Because the boundary theory is ordinary quantum mechanics that never destroys information, this equivalence gave conclusive evidence that information falling into a black hole is ultimately returned to the outside world: quantum mechanics needs no modification. (Susskind's The Black Hole Wars tells that story.) Unlike Plato's diminished shadows, Maldacena's flattened world tells the full story, and may be "the primary narrator."
Parallel universes or parallel mathematics? Testing with quark soup
Maldacena's result remains conjectural (no airtight proof) but has passed many stringent mathematical tests and entered mainstream thought. One reason proof is hard is also why it's powerful: when one side's coupling is small, the other's is large, so perturbative checks on both sides at once are impossible, but hard strong-coupling calculations on one side become easy weak-coupling calculations on the other.
The perfect fluid at RHIC. At Brookhaven's Relativistic Heavy Ion Collider, gold nuclei collide just shy of light speed, creating matter more than 200,000 times hotter than the sun's core, hot enough to melt protons and neutrons into a quark-gluon plasma, the state of matter just after the big bang. Its theory, quantum chromodynamics, has a large coupling, so perturbative calculations fail; measured shear viscosity (the drag each fluid layer exerts on its neighbours) came out far smaller than they predicted.
Reverse the holographic picture: put our quarks and gluons on the boundary; translate into strings in the bulk, where the coupling is small. Pavel Kovtun, Dam Son and Andrei Starinets did the calculation, and their result came "impressively close to the experimental data."
The boundary theory isn't our full universe (it lacks gravity, which hardly matters for light particles at RHIC), so here string theory is a calculational tool, not a theory of everything. Conservatively, "a potent string-based mathematical trick"; less conservatively, the higher-dimensional string description might be physically real. Either way Greene, "not a fan of hyperbole", calls it among the most exciting advances in decades.
Coda: the future of string theory
The chain from Wheeler's "information" through entropy, the Second Law, black-hole horizons and the holographic bound has produced a new unifying idea, which Greene expects to be "a beacon for physicists well into the twenty-first century." Within string theory, Maldacena's result does more:
- A definition at last. For nearly thirty years string theory lacked a full mathematical definition. The boundary quantum field theory is among the best understood in physics and contains no gravity (bringing GR directly into QFT is "like setting a campfire in a gunpowder factory"). Since it generates string theory, gravity included, it can serve as the mathematical definition of string theory, at least in AdS spacetimes.
- Two sides of one coin. String theory doesn't merely reduce to QFT in limits; string theory and quantum field theory may be the same theory in different languages, a translation complicated enough to take more than forty years to find.
- Holography as foundation. "The holographic parallel universes may thus be more than a potential outgrowth of fundamental laws; they may be part of the very definition of the fundamental laws." The open challenge is extending this to universes of any shape, including ours.
Going deeper: the technical notes
- Einstein's anti-black-hole calculation was mathematically correct but rested on assumptions (in effect, excluding implosion) too restrictive to reveal black holes; GR itself admits them.
- The Second Law has a ceiling. Entropy increases until it reaches the maximum the system allows, as with steam uniformly filling a vat.
- The laws of black-hole mechanics. In 1972 James Bardeen, Brandon Carter and Stephen Hawking found black-hole equations identical in form to thermodynamics, with horizon area ↔ entropy and surface gravity ↔ temperature. Bekenstein's idea required that the parallel be literal, hence a real temperature.
- Why the infalling partner's energy flips sign. A particle's energy relates to the rate of time for its field's oscillations. At the horizon, distant time freezes; inside, time and the radial direction swap roles, so "positive energy" means motion toward the singularity. The partner with negative energy as judged from outside has positive energy inside, which is why it can exist there indefinitely.
- Hawking's precise formulas. Temperature T = ħc3/(8πGMkB); entropy S = kBA/(4ℓP2), one quarter of the area in Planck units (in natural-log units). An evaporating black hole's lifetime scales as M3: about 1067 years for a solar mass.
- Entropy in bits. With entropy measured as a base-2 logarithm, entropy is the number of yes-no questions the microstate answers. The 1,000-coin example (entropy 1,000 bits) is exact; Hawking's "one bit per Planck cell" is approximate to factors of 4 and ln 2.
- What we still don't know. Bekenstein-Hawking gives the quantity of hidden information, not its content: which yes-no questions, about which constituents. String theory's Strominger-Vafa counting (Chapter 4) identified the constituents for special black holes.
- The bound beyond black holes. If adding matter to a region can at most create a black hole filling it, and entropy must rise throughout, then no region can hold more information than its boundary area in Planck units. This is 't Hooft and Susskind's holographic principle (1993 to 1995); Bousso's covariant entropy bound (1999) made it precise for general spacetimes.
- Complementarity's fine print. The distant observer needs time to decode the infaller's fate from Hawking radiation (about half the black hole's lifetime, the Page time); by then, jumping in cannot catch the infaller before the singularity. This "no-cloning" consistency check is what AMPS later challenged (2012).
- AdS5 × S5 in words. Anti-de Sitter space is a maximally symmetric spacetime with negative cosmological constant, shaped like a saddle stretched to a boundary at infinity that light reaches in finite time. The boundary theory is N = 4 supersymmetric Yang-Mills with gauge group SU(N); large N and strong coupling on the boundary correspond to classical gravity in the bulk.
- Information is preserved. Because the boundary theory is ordinary unitary quantum mechanics, Witten's identification of a bulk black hole with a hot boundary plasma implies no information is lost. In 2004 Hawking conceded his famous bet with John Preskill on this point.
- A definition, more than a model. Since the boundary QFT is well defined, it can serve as a non-perturbative definition of quantum gravity in AdS: physicists now often say "quantum gravity in AdS is a CFT."
Critical lens.
- "Parallel universe" is a stretch here. A duality is two descriptions of one reality, not two realities. Calling the boundary a parallel universe is evocative but conceptually different from every other multiverse in the book.
- Our universe is not anti-de Sitter. AdS has negative vacuum energy and a timelike boundary; ours has positive vacuum energy (Chapter 6) and no such boundary. Holography for de Sitter or flat space remains unproven, so applying "you are a projection of a distant surface" to us is extrapolation.
- The bit-per-Planck-cell picture is a simplification: the Bekenstein-Hawking entropy is S = A/4 in Planck units (natural-log units), so roughly one bit per four Planck areas, divided by ln 2.
- Complementarity was later challenged. In 2012 Almheiri, Marolf, Polchinski and Sully (AMPS) argued that complementarity plus unitarity plus smooth horizons are inconsistent, suggesting a "firewall" at the horizon: exactly the scenario Greene describes as impossible to confront.
- RHIC success is qualitative. The holographic viscosity ratio η/s = 1/4π comes from a theory (N = 4 super-Yang-Mills) that differs from QCD; agreement is suggestive, not a derivation of QCD.
2026 update: black holes and holography since 2011
- Seeing black holes: the Event Horizon Telescope imaged M87* (April 2019) and Sagittarius A* (May 2022). LIGO's 2015 merger detection opened gravitational-wave astronomy; analysis of the loud event GW250114 (published 2025) confirmed Hawking's area theorem with high significance, the first sharp observational test of the area law.
- Hawking radiation analogues: Jeff Steinhauer's sonic black holes in Bose-Einstein condensates (2016, 2019) showed thermal, entangled analogue Hawking radiation. Real astrophysical Hawking radiation remains undetectable.
- The information paradox, round two: after the 2012 firewall argument, 2019 papers (Penington; Almheiri, Engelhardt, Marolf and Maxfield; then "replica wormhole" analyses) computed the Page curve, showing via "islands" inside the black hole how information escapes in Hawking radiation, at least in AdS-based models. Most researchers now regard information as preserved; the mechanism is still debated.
- Entanglement builds space: the Ryu-Takayanagi formula (2006) relates boundary entanglement to bulk areas; Van Raamsdonk (2010) and Maldacena and Susskind's "ER = EPR" (2013) propose that spacetime connectivity is quantum entanglement, extending Wheeler's "information" programme.
- Quantum simulation: a 2022 "wormhole teleportation" experiment on Google's Sycamore processor was widely publicised; later analyses argued the small model did not faithfully capture holographic dynamics. Holography-inspired quantum simulation is an active but early field.
- Losses: Jacob Bekenstein died in 2015, Stephen Hawking in 2018.
- Heavy-ion physics: RHIC and the LHC confirm the quark-gluon plasma is a near-perfect fluid, with η/s close to the holographic value 1/4π.
So what. Three ideas transfer. (1) Entropy = hidden information = yes-no questions: a crisp way to quantify uncertainty in any system, from feature stores to fraud signals; it is the basis of information gain in decision trees and of cross-entropy loss in LLM training. (2) Capacity limited by surface, not volume: in architecture, a service's real capacity is often bounded by its interfaces (API throughput, network edge), not its internal resources. (3) Duality as a computational strategy: the RHIC result shows a hard problem solved by translating it into an equivalent but easier representation. That is the move behind kernel methods, embeddings and reformulating an optimisation as its dual.
Universes, Computers, and Mathematical Reality: The Simulated and Ultimate Multiverses
Until now the mathematics drove us to multiverses; here Greene asks what happens if we take the wheel. Could we create a real universe? Inflation shrinks the recipe to ten grams in a 10−26 cm seed, but the result would be invisible to us and probably needs a white hole to start. Could we create simulated universes with conscious inhabitants? Nothing in physics forbids it, and Bostrom's argument then suggests we are probably simulated ourselves. Finally, Nozick's and Tegmark's Ultimate Multiverse says every possible mathematics is a real universe. Greene stays sceptical of that last step, but shows how simulations could generate its "computable wing", and bets that reality will prove fundamentally discrete.
Seizing the wheel
The credibility of earlier theories varies (quantum mechanics is established fact, inflation has observational support, string theory is thoroughly speculative), but the pattern holds: hand the wheel to the mathematics and you are driven to parallel worlds. Now Greene asks whether we could create them, with the same intent you feel choosing a movie or a meal (setting aside free will). "You" means you or a distant descendant with arbitrarily advanced technology, still bound by physics. He warns that this is "territory even more speculative" than anything so far.
To create a universe
- Mass within our cosmic horizon: about 1055 g ("10 billion billion billion billion billion billion grams"). If it differed much, the CMB splotches would be the wrong angular size.
- Big bang recipe: gather more than 1055 g and squeeze it to an infinitesimal speck, then somehow ignite a bang that expands space itself. The big bang theory takes the raw material as an unexplained given.
- Inflationary recipe: a seed about 10−26 cm across holding inflaton energy equivalent to less than ten grams: "a lump you could put in your wallet."
- The catch: that seed's density is about 1067 times nuclear density; it "would weigh less than a handful of popcorn," but needs compression trillions and trillions of times beyond our capability.
Gravity: matter's sugar daddy
Inflation creates matter from almost nothing because gravity is an inexhaustible energy source. A high-valued inflaton drives expansion; the bigger region is also filled with inflaton at the same energy per volume, so total energy grows. The energy comes from gravity, which, "much like the U.S. Treasury, has no fear of debt". Jump off a cliff and your kinetic energy grows while gravitational energy goes more negative by the same amount; climbing back out of the rabbit hole is how you repay the debt. The only independent energy budget inflation needs is the tiny seed.
The good news and the bad news
Good news (Steven Blau, Eduardo Guendelman and Alan Guth): a lab-made seed won't swallow us. Inside: positive energy, negative pressure. Outside: zero pressure, which is larger, so the surroundings squeeze the seed like water on a diver's eardrums. Instead, like air bulging between the fingers of hands clasping a balloon, the seed sprouts a new expanding region of space, which beyond a critical size severs its umbilical cord and becomes an independent inflating universe. From our side, all we'd see is a black hole, so we could never confirm we'd succeeded. "Physics protects us, but the price for safety is total separation from our handiwork."
Bad news (Edward Farhi and Alan Guth): the newborn universe needs a strong kick-start that only a white hole (a hypothetical object that spews matter out, its conditions so extreme known maths breaks down) can supply. "No one anticipates generating white holes in the laboratory. Ever."
- Possible escapes. Farhi, Guth and Jemal Guven: create the seed by quantum tunnelling, avoiding the white hole, but with a probability so small it would essentially never happen. Nobuyuki Sakai, Ken-ichi Nakao, Hideki Ishihara and Makoto Kobayashi: a magnetic monopole (a particle with only a north or only a south pole) might catalyse inflation, but after nearly forty years of searching none has been found (ironically, inflation's dilution of monopoles is one explanation of why).
- Verdict: the door to creating universes is open, but only barely, and depends on hypothetical ingredients.
- Would we bother? Andrei Linde, with his flair for mock drama, says the allure of playing god would be irresistible. Greene doubts it: by then, other spectacular pursuits whose results we could actually experience would make an invisible creation less interesting. For real universes we can see or interact with, no known strategy is compatible with physics. Virtual universes are another matter.
The stuff of thought
Greene's fever dream. During a feverish flu he hallucinated being in a sparse hotel room, a hallucination within a hallucination, where he was certain days and weeks passed; dragged back to the primary hallucination, he'd find hardly any time had passed. Knowing he'd be swallowed whole again, he resisted drifting back. When the fever eased he'd surface to ordinary life. Lesson: "Our grip on reality is more tenuous than day-to-day life can lead us to believe."
This is epistemology's classic worry, the brain in a vat, popularised by The Matrix, The Thirteenth Floor and Vanilla Sky. You can't know for sure: stimulate the brain exactly as eating pizza or skydiving does and the experience is identical; "experience is dictated by brain processes, not by what activates those processes." Push further: dispense with biology entirely; might you be software on a hyper-advanced supercomputer? The danger is skeptical collapse: if simulated, why trust simulated neurobiology texts? Greene "drops anchor": assume for now that everything real is real, and estimate.
Brains versus computers
- Retina: 100 million neurons, smaller than a dime, a few sheets of paper thick. Hans Moravec: matching it takes ≈109 operations per second.
- Whole brain: scale up by volume (≈100,000×): ≈1014 ops/s (Moravec); synapse counts and firing rates suggest up to ≈1017 ops/s.
- 2010-era machines: Greene's computer ≈109 ops/s; top supercomputers ≈1015 ("a statistic that no doubt will quickly date this book"). At the high brain estimate: a hundred million laptops or a hundred supercomputers ≈ one brain.
- Earth-sized computer with present technology: 1033 to 1042 ops/s.
- All human thought ever: 1017 ops/s × 100-year life ≈ 1024 per person × ≈100 billion people since "Lucy" ("my archaeology friends tell me I should say 'Ardi'") ≈ 1035 operations: under two minutes on an earth-sized computer at the conservative rate.
- Quantum computing: a laptop-sized quantum computer might reproduce all human thought since the dawn of our species in a tiny fraction of a second.
Futurists call the coming surpassing of brainpower a singularity: human-technology boundaries blurring, thinking machines, minds uploaded in silico with backups. Greene calls it possibly hyperbolic: raw power is not in dispute; turning it into mind is. That question runs from Plato and Aristotle, Hobbes and Descartes, Hume and Kant, Kierkegaard and Nietzsche, James and Freud, Wittgenstein and Turing.
| Theory of mind | Core claim | Does a perfect copy think and feel? |
|---|---|---|
| Dualism | An essential nonphysical component is essential to mind | No (for both copies below) |
| Physicalism | Every subjective experience corresponds to a unique brain state | Yes, if copied molecule by molecule |
| Functionalism | What matters is processes and relationships, not substrate | Yes, even in silicon, if the connections and processes are preserved |
Artificial sentience requires functionalism: consciousness is the sensation of a particular kind of information processing, whether in "a three-pound biological mass" or in circuits. It could be wrong: perhaps sentience needs wet, wrinkled matter, or actual molecules, or is nonphysical after all.
The Blue Brain Project (IBM with EPFL, Lausanne) models brain tissue on Blue Gene, a more powerful successor to Deep Blue, which beat Garry Kasparov in 1997. From tens of thousands of experiments on a pinhead-sized rat neocortical column, it built a 3-D simulation of ≈10,000 neurons and ≈10 million connections, responding to stimuli with encouraging fidelity. A human brain has ≈100 billion neurons. Leader Henry Markram anticipated a full human-brain model before 2020 and speculated it might speak and feel, though its aim was medical, a tool for treating mental illness.
How would we know a computer is sentient? Greene can't verify his wife's sentience, nor she his: consciousness is private. But circumstantial evidence makes solipsism absurd, and conversing with, consoling and cajoling computers may one day make "the simplest explanation" that they are conscious. Greene adopts functionalism to see where it leads.
Simulated universes
Some would put sentient software in robot bodies (C-3PO, Data); Greene's interest is simulated worlds with simulated, self-aware inhabitants, like The Sims or Second Life with real minds. Fidelity would improve iteration by iteration. The Simulator decides whether inhabitants know; those who guess might be led away by simulated technicians in white coats. Most would find the idea too silly to consider.
Simulations can cut corners: render only what lies within the cosmic horizon; render stars only during simulated clear nights when someone might look up; render cells, molecules and atoms only when simulated scientists study them, by tracking inhabitants' intentions and anticipating observation. A cheaper replica with detail on demand would suffice. Such worlds would realise Wheeler's primacy of information: "generate circuits that carry the right information and you've generated parallel realities that are as real to their inhabitants as this one is to us." This is the eighth multiverse, the Simulated Multiverse.
Are you living in a simulation?
The idea goes back to the 1960s: Konrad Zuse and Edward Fredkin (the universe as "cosmic Fortran", as Greene's IBM summer boss, the late John Cocke, liked to say). Greene thought it absurd, until he met Nick Bostrom's argument (also hinted at by Moravec).
Bostrom's simulation argument:
- Real universes are hard to create and invisible if created; simulated universes are easy and we already build them. The open question is only how realistic they become.
- If realistic sentient simulations are possible, our descendants will run immense numbers of them, often, "kick back and fire up the create-a-universe software."
- Then simulated minds vastly outnumber biological ones.
- So, by brute statistics, we are probably simulated, perhaps by future historians curious about life on twenty-first-century earth.
The escape routes (Bostrom's own): (a) sentience cannot be simulated; (b) civilisations destroy themselves before reaching that capability; (c) mature civilisations choose not to run such simulations (moral reasons, or more interesting pursuits). A further loophole: if real universes are infinite (e.g. The Quilted Multiverse), comparing infinite real and simulated populations runs into the measure problem.
Does this collapse into total scepticism? Partly: will the sun rise tomorrow? "Maybe, as long as whoever is running the simulation doesn't pull the plug." Are memories reliable? The keyboard operator may edit them. But Bostrom notes one thing we'd still know about the root reality: it permits realistic simulations. The scepticism aligns with that knowledge rather than undermining it, so the "anchor" was useful but not necessary: logic alone can't ensure we're not simulated. Greene's tongue-in-cheek advice: make your mark; "whoever is running the simulation is bound to get tired of wallflowers."
Seeing beyond a simulation
The Simulator might reveal herself: a "ding-ding" in the shower and a floating window with a smiling programmer, or a planet-wide booming voice. Absent that, clues might come from imperfections:
Emergent strategy: stitch together the best knowledge for each domain: quantum field theory for proton collisions, Newton for batted balls, biochemistry, physiology and psychology for a mother watching her child's first steps, political theory, history and economics for leaders. The challenge is keeping the patchwork consistent where domains meet; tweaks might appear to inhabitants as sudden, baffling, causeless changes, and inconsistencies could build until the simulation crashes.
Ultra-reductionist strategy: one set of fundamental equations plus initial conditions, evolved forward. No meshing problems, but if the laws involve continuous numbers, a finite computer can only approximate them (tracking 0.9 to 1 through 0.95, 0.958, 0.9583, 0.95831… requires infinitely many digits). Round-off errors accumulate; inhabitants would see anomalies, failed predictions, and once-converged measurements drifting.
The iconoclast's demonstration. After exhausting explanations for mounting anomalies, a simulated physicist suggests: we are in a simulation. "That's nutty." "Really? Take a look." She shows a simulation she built using the very same deep laws, and its simulated scientists are puzzling over exactly the same strange data.
Greene concedes a careful Simulator could reset the program and erase memories of anomalies, and that glitches are far more plausibly our own scientific failings. But one serious conclusion stands: if we ever create apparently sentient simulations, the principle of mediocrity says we should not assume we are the first; there would likely be a "swarming ocean" of simulations, and since most sentient beings would then be simulated, we probably are too.
The Library of Babel: the Ultimate Multiverse
Nozick's challenge. In his first college semester Greene took an introductory philosophy course from the late Robert Nozick, then completing Philosophical Explanations (1981). Physics overturned intuition about the fast, heavy and tiny; philosophy overturned everyday experience. Greene told Nozick he wanted to work on quantum gravity and unified theories to find "eternal truths." Nozick: "Let's say you find the unified theory. Wouldn't you still be left asking why that particular theory... Was the correct theory of the universe?" Greene replied we might have to accept some things as given: exactly where Nozick wanted him.
Nozick's principle of fecundity defangs the question: don't single out any theory. Imagine every possible universe exists: beyond Many Worlds branches or inflationary bubbles or string vacua (each still leaves "why quantum mechanics? Why inflation? Why strings?"), but any universe whatsoever, "made of the usual atomic species, but a universe made solely of melted mozzarella would serve just as well." Since every proposed multiverse consists of possible universes, all are contained in this one: the Ultimate Multiverse, the ninth and largest possible.
- Why our laws? Anthropics: we live where life is possible; nearby life-friendly universes (tiny parameter changes) have people asking the same question.
- Existence is no distinction: "There are no sterile laws. All sets of laws describe real universes."
- Leibniz answered: "Why is there something rather than nothing?" troubled Greene at ten or eleven; he would pace his room with a hand hovering behind his head, trying to see it as a way of grasping absence. Nozick notes the Ultimate Multiverse contains a universe of absolutely nothing (not empty space; nothing). No imbalance needs explaining; "a nothing universe is nothing to get exercised about."
- Borges's library: a theorist reads it as every consistent mathematical structure realised physically: "La Biblioteca de Babel" written in mathematics. A few books hold general relativity and quantum mechanics; most hold equations nobody has written down. "Every book in the Library of Mathematical Babel is a real universe."
Is mathematics discovered or invented?
Discovered: maths fits physics uncannily. Paul Dirac (1931) found an "extraneous" solution to his electron equation describing a positively charged electron; Carl Anderson (1932) discovered the positron in cosmic rays: the first antimatter, born in a notebook.
Invented: evolution shaped us to spot patterns for finding the next meal; mathematics is the language of pattern, a human tool. Aliens might reach the same results because of shared needs (counting, trading, surveying), not because maths transcends us.
Greene vacillates. Mid-proof, when pieces "suddenly coalesce," it feels like discovery, "a grand vista hidden by the morning fog." Surveying maths objectively, it looks like literature written in a precise language. He once imagined aliens saying, "Oh, math. Yeah, we tried that for a while... Ultimately it was a dead end. Here, let us show you how it really works." but admits he can't say how they'd finish, or what wouldn't count as maths under a broad definition.
The Ultimate Multiverse is unequivocal: all maths is real. Newton's equations with only solid billiard balls; an empty universe of 666 spatial dimensions under higher-dimensional Einstein equations: each is a universe.
Multiverse rationalisation: Greene's objection
Every earlier multiverse was a by-product. Quantum mechanics was built for the microworld, inflation for cosmic observations, string theory to reconcile quantum mechanics and gravity. That they spawn multiverses (perhaps solving measurement, time's beginning, gravity's weakness, dark energy, heavy-ion data) is secondary. The Ultimate Multiverse, by contrast, does one job: deleting "why these laws?" from the to-do list, by assuming a multiverse. "Cooked up specifically to address one issue," it lacks independent rationale.
The counter-view: maths is reality
Structural realists suggest physicists have a false dichotomy between maths and physics. Greene's bridge: Samuel Johnson kicked a stone to refute Bishop Berkeley ("I refute it thus"). Put that kick inside a high-fidelity simulation: the computer is just mathematical rules evolving arrangements of bits, so studying the maths you would see the kick, the rebound, the thought and the words. Then remove the computer: it is an "inessential intermediate step." "Dr. Johnson is in the mathematics." Everything you experience would be "the experience of mathematics. Reality is how math feels." Greene has not taken that leap.
Tegmark's Mathematical Universe Hypothesis (MUH): Max Tegmark (MIT) argues:
- The deepest description of reality should be free of human-dependent concepts ("human contagion").
- Mathematics (operations on abstract sets yielding relations, like 1 + 2 = 3) is exactly that language.
- Any feature distinguishing a universe from its mathematics would be either mathematical (then absorbable into the maths) or non-mathematical (then human-tainted, so not fundamental).
- So nothing distinguishes them: mathematical existence = physical existence. "There's no switch that turns math 'on'." Applied to all maths, this yields the Ultimate Multiverse.
Greene's scepticism: he wants a process (fluctuating inflaton, colliding branes, landscape tunnelling, Schrödinger evolution) that could generate a multiverse. In the Inflationary and Landscape multiverses the same equations hold everywhere; only environments differ. What process within one set of laws could produce different laws? "Like the number five desperately trying to be six." Except simulations: the hardware obeys ordinary physics, but each simulated world runs on whatever equations the programmer chooses.
Simulating Babel: the computable wing
- Continuous laws can only be approximated. A batted ball classically passes through infinitely many points with infinitely many speeds; computers track it on a fine grid. Quantum theory adds discreteness (finitely many distinguishable outcomes in a region) but still uses continuous wave values.
- Computable functions can be evaluated exactly by a finite set of discrete instructions, perhaps after many cycles, with no originality needed: "just a matter of grinding out the result."
- Noncomputable functions never finish. Greene's example is the simulated chef who cooks for exactly those who don't cook for themselves: who feeds the chef? (a version of Russell's barber paradox). The computer hangs. So only computable universes survive in a Simulated Multiverse.
The Simulated Multiverse can generate part of the Ultimate Multiverse with an actual process: armies of future users, temperamentally like Second Life enthusiasts, simulating ever-different equations, working through the Library of Babel's computable wing. Jürgen Schmidhuber (extending Zuse) showed it is easier to program all computable universes than one: simulating one baseball game needs a mountain of specific data; a short master program that systematically enumerates every possible game contains them all. "I'd happily pay the programmer by the hour" (but not the computer time). Tegmark adds that a complete simulation is itself a mathematical structure, so if all maths is real, no simulation need actually run.
The cost: generating only the computable wing revives Nozick's question: why should computable laws hog the spotlight? Greene's speculation: computable maths evades Gödel's ghost. Kurt Gödel's incompleteness theorem shows some mathematical systems contain true but unprovable statements; physicists have wondered whether physics might be forever incomplete. Computable universes lie squarely within calculation, so Gödel, Turing and Church limits wouldn't bite. Maybe that is what singles them out.
Is our universe computable exactly, rather than approximately? If so, the continuum must go: space and time would be fundamentally discrete. Space has been probed only to a billionth of a billionth of a metre; the one-bit-per-Planck-area result of Chapter 9 points toward discreteness. Greene's guess: whether or not sentient simulations ever exist, the world will prove fundamentally discrete.
The roots of reality
In a Simulated Multiverse there's no ambiguity about which universe is real: the root, whose computers' crash would bring everything down, however many simulations-within-simulations run. But simulated scientists with autonomy would discover the code of their world and call it natural law, though their laws need only be "good enough" to produce sentient inhabitants. If many law-sets qualify and we are typical, science's mission of revealing fundamental truths about the root reality would be undermined.
Greene is not losing sleep: until a sentient simulation takes his breath away, he won't seriously believe he's in one. And he suspects the novelty of unaware simulated worlds would fade ("there's just so much reality TV you can watch"). The lasting applications would connect simulated and real worlds: simulated beings migrating into reality, biological people joining simulations, until "the distinction between real and simulated beings might become anachronistic." Then the Simulated Multiverse would become part of what we mean by reality.
Going deeper: the technical notes
- The big bang takes its raw material as given. In the standard model the universe's mass-energy was always present, only compressed; inflation is the first theory to manufacture most of it.
- Negative gravitational energy, again. Falling particles gain kinetic energy as gravitational energy becomes more negative; inflation draws on the same bottomless account. In a closed universe the total energy (matter plus gravity) can be exactly zero, which is why some call the universe the "ultimate free lunch."
- The seed numbers. The inflationary seed's size (≈10−26 cm) and mass-energy (≲10 g) come from requiring that its own inflaton energy drive inflation inside a region about the size of the inflationary horizon; its density, 1067 times nuclear density, is set by the inflationary energy scale.
- Monopoles twice over. Inflation is credited with diluting grand-unification monopoles; Sakai and collaborators' proposal would use a monopole to seed new inflation.
- The Bostrom loophole from infinity. If real universes are infinite in number (a Quilted Multiverse), both real and simulated populations are infinite, and "most observers are simulated" becomes a measure-problem statement with no definite meaning yet.
- Finite states, continuous mathematics. Quantum mechanics allows only finitely many distinguishable outcomes in a finite region, yet its formalism uses continuous wave values. A computable universe would need the formalism itself to be discrete, not merely the outcomes.
- Borges's library actually contains every string of characters, meaningful or not; Greene's mathematical version keeps only consistent strings.
- Computability is about the whole history. Tegmark notes that a complete simulation is itself a mathematical structure, so if all maths is real, no computer need run. And computability should be judged on the relations defining the entire spacetime history, not only on step-by-step evolution in time.
- Moravec's arithmetic. Retina ≈109 ops/s × ≈105 (brain/retina volume) ≈ 1014 ops/s; synapse-based estimates (≈1014 to 1015 synapses × up to ≈100 Hz) push toward 1017. Brain power consumption is about 20 watts, a figure that, unlike ops/s, today's computers are nowhere near matching for equivalent work.
Critical lens.
- Bostrom's argument is a trilemma, not a probability. Greene presents the "probably simulated" horn prominently; Bostrom's actual conclusion is that at least one of three propositions is true. Critics (e.g. Sean Carroll) add that the reasoning risks self-undermining: if most simulations are cheaper, coarser simulations, typical observers would find themselves in low-fidelity worlds that can't run detailed simulations of their own.
- Moravec-style ops/second estimates equate synaptic events with digital operations, a contested assumption; energy, connectivity and biophysical detail may matter more than raw speed.
- Blue Brain did not deliver a human brain by 2020. The project produced detailed rodent cortical models; the EU's Human Brain Project (2013 to 2023) ended without a whole-brain simulation, amid public criticism of its goals in 2014.
- The "noncomputable chef" is a self-reference paradox, not a noncomputable function in Turing's sense (the halting problem is the canonical example); the analogy conveys flavour, not the technical idea.
- Gödel and physics: incompleteness concerns provability in formal systems; a computable universe can still contain undecidable questions about its own long-term behaviour (as the halting problem shows). "Free of Gödel's ghost" overstates it.
- Greene's "process" criterion is a personal methodological preference; Tegmark would answer that demanding a generating process smuggles in a pre-mathematical notion of time.
2026 update
- Compute caught up with brains: Frontier (2022) broke the exaflop (1018 ops/s) barrier; El Capitan (2024) reached about 1.7 exaflops; Europe's JUPITER became the first European exascale system (2025). Single machines now exceed Greene's high brain estimate. Frontier AI training runs use about 1025 to 1026 operations, comparable to his estimate of a lifetime of human thought (1024).
- Machines that talk: large language models now converse fluently, which is exactly the circumstantial evidence Greene says might make machine consciousness "the simplest explanation." Expert opinion remains sharply divided; a 2023 multi-author report (Butlin, Long and colleagues) assessed current systems against scientific theories of consciousness and concluded none was a strong candidate, while finding no obvious barrier to building one.
- Connectomes: the complete adult fruit-fly brain wiring diagram (FlyWire, 2024, ≈140,000 neurons, ≈50 million synapses) has been mapped and used for whole-brain simulation; the MICrONS project (2025) mapped a cubic millimetre of mouse visual cortex. A human connectome remains far off.
- Testing for simulation: Beane, Davoudi and Savage (2012) proposed that a lattice-based simulation would leave signatures in the highest-energy cosmic rays; no such effect is seen. Fermilab's Holometer (2015) found no holographic spacetime jitter at its sensitivity, and gamma-ray-burst timing (Fermi, from GRB 090510) rules out the simplest Planck-scale discreteness that would make light speed depend linearly on energy. Discreteness, if real, must be subtle.
- Magnetic monopoles: still undiscovered; the LHC's MoEDAL experiment continues to set limits.
- Quantum computing: Google's Willow (2024) and others have demonstrated error correction below threshold, but a laptop that replays "all human thought" remains science fiction.
So what. Greene's simulation discussion maps closely onto enterprise digital twins and AI agents. Render on demand (simulate detail only where someone is looking) is how game engines, digital twins and LLM agent sandboxes control cost, and it is also how inconsistencies creep in at boundaries, Greene's "emergent strategy" problem. The emergent versus ultra-reductionist split is the familiar choice between composing specialised models (fast, inconsistent at interfaces) and one foundational model (consistent, expensive, approximation errors that accumulate). Schmidhuber's insight that "generate everything" can be cheaper to specify than "generate this one thing" is the logic of synthetic-data generation and of search over program spaces.
The Limits of Inquiry: Multiverses and the Future
Greene cannot answer "universe or multiverse?", so he closes with five questions physicists will keep asking. Is the multiverse the capstone of the Copernican demotions? Can multiverse theories be tested, and are these ones testable yet? How would a multiverse change what counts as explanation, turning initial conditions, constants and even laws from deep mysteries into local facts? And how seriously should we take mathematics? His answers: multiverses keep arising unbidden from well-tested theories; eight of the nine are open to judgment while the full Ultimate Multiverse is not; and the single thread through the book, hence its singular title, is mathematics' power to reveal hidden truths.
Newton's pebble and the virtue of surprise
Newton found that a few equations govern motion on earth and in the heavens, but never thought them eternal truths: "I seem to have been only a boy playing on the seashore... Whilst the great ocean of truth lay before me all undiscovered." Greene is glad Newton's laws have limits; had they held at every scale, "the cosmic flavor was vanilla through and through." Beyond everyday conditions, common experience has proved untrustworthy, and big changes to our picture are rightly resisted until the evidence reaches critical abundance: "Reality will wait." Science assumes patterns exist at all scales, not that familiar patterns recur at all scales. "The surprise would have been to find no surprises."
Physics enjoys local uncertainty (no generation knows how history will judge its work: diversion, passing fascination, stepping-stone or lasting insight) but global stability: new theories incorporate and extend old ones rather than erase them, so the story stays coherent.
The nine multiverses
| Multiverse | Core idea | Status (Greene) | What could sink or support it |
|---|---|---|---|
| Quilted | Infinite space forces conditions, and copies of you, to repeat | Tentative | Sunk by evidence that space is finite |
| Inflationary | Eternal inflation produces endless bubble universes | Tentative, with inflation supported | Supported by bubble-collision patterns in the CMB; weakened if inflation needs contorted potentials |
| Brane | Our universe is a three-brane in a higher-dimensional bulk with other branes | Speculative | Supported by SUSY, missing energy or mini black holes at colliders |
| Cyclic | Brane collisions give big bangs repeatedly: universes parallel in time | Speculative | Distinguished from inflation by primordial gravitational waves (detection would count against it) |
| Landscape | Inflation + 10500 extra-dimensional forms give bubbles with different physics and Λ | Speculative | Bubble collisions; landscape maps and their statistics; a string inconsistency would kill it |
| Quantum | Every outcome of a quantum probability wave is realised (Many Worlds) | Controversial | Killed by any failure of linearity; in-principle interference tests between branches |
| Holographic | Our universe is exactly mirrored by physics on a distant boundary | Strong theoretical support | Growing contact with RHIC data; physical versus mathematical reading still open |
| Simulated | Advanced computers could host universes with sentient beings | Depends on substrate-independent sentience | Settled if we ever build or enter a convincing sentient simulation |
| Ultimate | Every possible universe (every mathematics) exists | Not falsifiable in full form | Only its computable subset has a generating process (simulation) |
Question 1: Is the Copernican pattern fundamental?
Experience suggests we are the hub: sun rising and setting, constellations wheeling, the lead role we each play in our inner lives. Science has steadily corrected this. Earth is not central among neighbours, the sun not central in the galaxy, the Milky Way not central among galaxies, and even protons, neutrons and electrons are not the main ingredient of the cosmos. Once felt as assaults on human worth, these corrections we now value as enlightenment.
The multiverse may be the capstone Copernican correction: our universe itself may be one among many. What lifts it above idle speculation is the book's refrain: scientists were not hunting for ways to complete Copernicus. They built theories from data, followed them, and kept running into multiverses: "They're harder to avoid than they are to find." "The more we understand, the less central we appear."
Question 2: Can multiverse theories be tested?
Multiverses differ from earlier demotions by invoking domains possibly forever beyond examination, threatening "just so" stories ("the way things just happen to be here"). Greene restates his four routes to testability:
How a multiverse theory can make falsifiable predictions:
- Universal features: properties shared by all universes; fail to find them here and the theory dies.
- Correlations: e.g. Every universe containing electrons, or containing all known particles, also contains a particular undiscovered species; look for it.
- Distributions: if most universes share our value of, say, Λ, confidence grows.
- Anthropic typicality: among life-friendly universes, are we typical? If we "stick out like a sore thumb," the proposal loses force (though unlikely outcomes do happen).
Requirement for all four: the demographics of actual universes, derived from the cosmological process that creates them, beyond the list of possible ones.
Question 3: Can we test the multiverses we've met?
Physicists' hunches guide research "behind the scenes," but proposals are judged by one standard only: explaining or predicting data. That is science's singular beauty: wide creative latitude, then a final reckoning. But some theories have outrun our ability to test them; string theory has been the "poster child," and multiverses are "an even more sprawling example." None yet meets the criteria. The landscape is mapped only roughly, "like ancient seafarers"; tunnelling is understood in principle but not quantitatively in string theory (Greene's own group has done "initial reconnaissance").
Where Greene draws the line. What counts as "respectable science": only what Earth-bound humans can test now? In a hundred years? Two hundred? Or any path radiating from confirmed concepts, even into permanently hidden domains? It's a matter of taste, and Greene's is expansive: bounding thought by "where we are, when we are, and who we are" is parochial. But he excludes ideas that cannot be confronted by observation because of their inherent nature, not human limitations. Only the full Ultimate Multiverse fails: whatever we observe, it "will nod and embrace our result." The other eight each come from well-motivated reasoning and each is open to judgment.
Question 4: How does a multiverse change scientific explanation?
Science explains details (elliptical orbits, the blue sky, transparent water, a solid desk) and big pictures (hundreds of billions of stars per galaxy, hundreds of billions of galaxies, dark energy everywhere; a century ago the universe was thought static and to be just the Milky Way). Sometimes it makes us re-examine science itself.
The three ingredients of any physical description:
- Equations for the laws (Newton's, Maxwell's, Schrödinger's).
- Constants appearing in them (strengths of gravity and electromagnetism, particle masses).
- Initial conditions (a ball hit from home plate at a given speed and angle; an electron with 50% odds at Grant's Tomb and 50% at Strawberry Fields).
Classical physics then predicts outcomes; quantum physics predicts probabilities. For the whole universe each ingredient invites a "why?", and multiverses change which "why"s make sense.
- Quilted: laws the same, initial conditions vary, so asking why ours were what they were is "single-universe mentality in a multiverse setting"; ask instead whether universes like ours exist and abound. Then initial conditions need no more explanation than a New York shoe store stocking your size.
- Inflationary, Brane, Cyclic, Landscape: constants vary (electron mass, whether there is an electron, electromagnetism's strength or existence, Λ). Asking the equations to pick our values is as meaningless as asking for the word with which Shakespeare wrote *Macbeth*.
- Simulated and Ultimate: even the laws vary; we experience ours "because these are among the laws compatible with our existence."
Collectively, multiverses can make prosaic the three things that are most mysterious in a single universe: the initial conditions, the constants and even the laws.
Question 5: Should we believe mathematics?
"Our mistake is not that we take our theories too seriously, but that we do not take them seriously enough. It is always hard to realize that these numbers and equations we play with at our desks have something to do with the real world." Steven Weinberg, on the Alpher-Herman-Gamow microwave background prediction
That prediction followed straight from general relativity and basic cosmology, yet was discovered theoretically twice, a dozen years apart, then found by serendipity. But Weinberg's remark needs care: not every equation on a theorist's desk is reality; deciding which maths to take seriously is "as much art as it is science."
Why the title is singular. The nine multiverses suggest "a panorama of hidden realities," but Greene titles the book The Hidden Reality to name the one theme beneath them all: "the capacity of mathematics to reveal secreted truths about the workings of the world."
Einstein, Everett and taking maths seriously:
- Late 1800s: Maxwell's equations give light's speed, ≈300,000 km/s, but relative to what? Physicists invent the aether as a rest frame.
- 1905: Einstein takes Maxwell more seriously: if the equations need no rest frame, there is none; light's speed is 300,000 km/s relative to anything. Special relativity follows. Everyone had the maths; Einstein embraced it fully.
- 1915 to 1919: mastering mathematics most physicists barely knew, Einstein completes general relativity; told of the eclipse confirmation, he says that otherwise he'd have felt "sorry for the dear Lord, since the theory is correct."
- But: Einstein didn't take his equations seriously enough to accept black holes or expansion; Friedmann, Lemaître and Schwarzschild did. In his last twenty or so years he went the other way, perhaps "blinded" by a thicket of unified-field equations. Even Einstein misjudged which maths to trust.
- 1926 / 1957: Schrödinger's equation was thought to apply only to the small; Everett echoed Einstein's charge, take the math seriously, applied it to everything, and found the Quantum Multiverse.
Every multiverse rests on the belief that mathematics is stitched into reality's fabric, the Ultimate Multiverse most extremely (maths is reality). Whether that takes our theories too seriously or not seriously enough, "only time will tell."
If the maths proves right, Einstein's question (could the universe have been otherwise?) gets a definite answer: yes. Our universe is not the only one possible; its properties could have been, and elsewhere are, different. Seeking a fundamental reason for everything would be pointless; "statistical likelihood or plain happenstance" would enter our understanding of a vast cosmos. Greene doesn't know if that's how things will turn out, "no one does," but "it's only through fearless engagement that we can learn our own limits."
Critical lens on the conclusion.
- The "harder to avoid than to find" claim has selection bias. Theories that don't produce multiverses (most of physics) aren't counted; and the nine are not independent (several share eternal inflation or string theory), so "nine routes" overstates the convergence.
- The Copernican framing is double-edged. Anthropic reasoning, which several multiverses depend on, reintroduces observers as central to explanation, arguably an anti-Copernican move Greene doesn't address.
- "Explained away" is not universally accepted as explanation. Critics (Ellis, Silk, Steinhardt, Hossenfelder) argue that relabelling mysteries as environmental accidents can halt inquiry, exactly the "Newton's apple" danger Greene raised in Chapter 7.
- The expansive-taste line is fuzzy. "Testable in principle, given enough time and technology" excludes almost nothing except the full Ultimate Multiverse; a sharper criterion would weigh how likely a test ever is, not just whether one is conceivable.
2026 update: the scorecard fifteen years on
- Quilted: Planck confirms flatness to ≈0.2% and finds no small-topology signal; infinite space remains allowed, not proven.
- Inflationary: inflation's predictions (spectral tilt, Gaussian fluctuations) hold; no bubble collisions found; r < 0.036; the measure problem is unsolved.
- Brane, Landscape: no SUSY, no extra dimensions, no mini black holes at the LHC through Run 3; the swampland challenge to de Sitter vacua is unresolved; DESI's hint of evolving dark energy (2024 to 2025) would favour rolling fields over a landscape-picked constant.
- Cyclic: primordial gravitational waves still undetected, so the inflation-versus-cyclic test remains open.
- Quantum: loophole-free Bell tests (Nobel 2022), larger superpositions, and tightening limits on collapse models; Many Worlds remains one of several live interpretations.
- Holographic: the Page curve and "island" calculations (2019 onward) strengthened holography's role in the information paradox; a holographic description of our de Sitter-like universe is still missing.
- Simulated: exascale computing and fluent AI systems have made the simulation question feel less remote; machine sentience remains unestablished.
- Ultimate: no change: still untestable in its full form.
So what. Greene's five questions form a reusable review template for any bold architecture or AI strategy: (1) Does it follow from evidence, or from wanting a particular answer? (2) What would it predict that we could check? (3) Can we check it now, and if not, what would make it checkable? (4) Does it change which questions are worth asking (e.g. Treating some variance as environmental rather than a defect)? (5) Are we taking our models too seriously, or not seriously enough? The Einstein history shows the same mind can err both ways.
Master Takeaways
Strip away the analogies and The Hidden Reality makes a small number of strong claims about physics, and a larger number about how science should reason when its best theories outrun its instruments. The first six below are the ones to remember.
The ideas that matter most
- Multiverses are by-products, not goals. Infinite space, eternal inflation, string theory, quantum mechanics and black-hole physics were each built to explain ordinary data; each independently yields parallel worlds. That convergence, not any single argument, is why the idea deserves attention.
- Finite states plus infinite space means infinite repetition. Quantum uncertainty and the black-hole limit allow ≈1010122 distinguishable configurations per horizon; in infinite space every configuration, including you, recurs infinitely often.
- Inflation is probably eternal, and its bubbles are infinite inside. Expansion outpaces quantum decay of the inflaton, so bubbles form forever; the outsider's endless time is the insider's endless space, so inflation creates Quilted Multiverses.
- The cosmological constant is the multiverse's strongest empirical hook. Λ ≈ 10−123, about 120 orders below naive theory and within ~100 of Weinberg's anthropic bound (1987, before the 1998 discovery). The explanation works only with more than 10124 varied universes, which string theory's ≈10500 vacua plus eternal inflation could supply.
- Unobservable is not unscientific; untestable is. Science accepts inaccessible entities (probability waves, black-hole interiors, regions beyond the horizon) when the theory implying them is independently confirmed. A multiverse can predict through invariant features, correlations, skewed distributions or anthropic typicality. Only the full Ultimate Multiverse is untestable by its nature.
- Information is bounded by area, and reality may be holographic. A region's information capacity is set by its boundary area in Planck units; Maldacena showed string theory with gravity in a bulk equals a gravity-free quantum field theory on its boundary, a duality that already computes real quark-gluon-plasma properties.
- The measure problem is the deepest obstacle. In infinite multiverses every outcome happens infinitely often, and comparing infinities depends on the method. Until physics supplies a measure, multiverse predictions, and possibly inflation's own successes, rest on unproven footing.
- Many Worlds is the conservative reading of quantum mechanics, with one hole. Applying Schrödinger's equation to everything, with decoherence separating branches, removes Copenhagen's ill-defined collapse. The unresolved cost is probability: branch counting gives the wrong answer, and decision-theoretic or self-locating derivations of the Born rule are not universally accepted.
- String theory's scorecard is split. Excellent on quantum gravity, unification, singularities, black-hole entropy and mathematics; indeterminate on particle properties and experiment, because the extra dimensions have too many possible forms.
- Dualities are physics' most powerful translation tool. The five string theories are one (M-theory); mirror Calabi-Yaus make hard sums easy; AdS/CFT turns strong coupling into weak. The same reality can have radically different, equally complete descriptions.
- Same laws, different environments. In the Inflationary, Brane and Landscape multiverses the fundamental equations are identical everywhere; masses, forces and Λ differ because fields settle differently, as a magnet-permeated universe would look lawless yet obey our laws.
- Multiverses change which questions are meaningful. Initial conditions (Quilted), constants (Inflationary, Brane, Cyclic, Landscape) and even laws (Simulated, Ultimate) can become environmental facts, like Earth's distance from the sun. Asking for their "fundamental" values may be like asking for the word with which Shakespeare wrote *Macbeth*.
- Anthropic reasoning is selection-bias correction, and only works with many universes. The shoe shop explains your fit only if it stocks many sizes; a unique universe still begs the question.
- Selection effects can masquerade as laws. Kepler's search for a reason behind Earth's 93-million-mile orbit was a "fool's errand"; some of today's constants may be the same kind of accident.
- Black holes are thermodynamic objects. They have entropy (area/4) and temperature (∝ 1/M), radiate, and preserve the Second Law; their information is stored, from outside, at the horizon.
- Complementarity: incompatible stories can both be true if they can never be compared. The infaller crosses the horizon unharmed; the outside observer sees her burned and scrambled; no observer can collect both accounts.
- Inflation creates matter from almost nothing. Gravity's energy can go arbitrarily negative, so a ≲10-gram, 10−26 cm inflaton seed could in principle become a universe, invisible to its makers, who would see only a black hole.
- If sentient simulations are possible, we are probably not the first to build them. Bostrom's trilemma: simulations are impossible, or civilisations don't survive or don't choose to run them, or most minds are simulated.
- Computability may be a selection principle. If physics is exactly computable, space and time must be discrete; Greene bets they are.
- The singular theme: trust in mathematics. Einstein took Maxwell's equations seriously (special relativity) but not his own (black holes, expansion); Everett took Schrödinger's seriously. Deciding which equations to believe is "as much art as science."
Frameworks and mental models
| Name | What it says | Use it when |
|---|---|---|
| Pigeonhole repetition (Ch 2) | Finite states × unbounded trials ⇒ guaranteed repeats | Reasoning about collisions, duplicates, false positives at scale |
| Resolution limit (Ch 2) | Differences below measurable resolution are not real differences | Deciding how finely to segment or model anything |
| Three-line tax form (Ch 2) | GR has geometry, matter, and vacuum energy; leaving a line blank is a choice | Auditing a model's implicit zero assumptions |
| Static map, changing legend (Ch 6) | Keep coordinates fixed; put growth in a scale factor | Thinking about expanding systems or normalised metrics |
| Pressure gravitates (Ch 3) | Negative pressure produces repulsive gravity | Understanding inflation and dark energy |
| Expansion versus decay (Ch 3) | If growth outpaces removal, the process never ends | Epidemics, technical debt, self-reinforcing loops |
| Required versus nice-to-have scorecard (Ch 4) | Grade goals separately by necessity | Evaluating platforms and research programmes |
| Reduction to predecessor (Ch 4) | Good new theories reproduce old ones in old regimes | Migrating systems; validating replacements |
| Perturbation series (Ch 5) | First pass plus small corrections, valid only for small coupling | Estimates under stress; knowing when they break |
| Duality (Ch 5, 9) | Two descriptions of one reality; hard in one is easy in the other | Reformulating intractable problems |
| Selection bias / anthropics (Ch 6) | Observers see only what permits observers | Survivorship bias in data, pilots and backtests |
| Predictive multiverse criteria (Ch 7) | Invariants, correlations, skew, typicality | Making any "many possible worlds" model testable |
| Measure problem (Ch 7) | Ratios over infinite sets depend on how you count | Defining denominators before comparing |
| Two (three) stories (Ch 8) | Separate the maths, its experiential reading, and why we trust it | Explaining AI or quantitative models without smuggled language |
| Decoherence (Ch 8) | Interacting with many particles makes branches stop interfering | Why quantum effects vanish at scale |
| Entropy = hidden information (Ch 9) | Entropy counts yes-no questions the microstate answers | Uncertainty, compression, information gain |
| Holographic bound (Ch 9) | Capacity scales with boundary area, not volume | Interface-limited systems |
| Simulation trilemma (Ch 10) | Impossible, or not done, or we're likely simulated | Reasoning about observers and reference classes |
| Five closing questions (Ch 11) | Copernican fit, testability in principle and practice, explanation, trust in maths | Reviewing any bold theory or strategy |
Critical lens on the whole book.
- Taxonomy inflates convergence. The nine multiverses are not independent: Quilted, Inflationary and Landscape share eternal inflation; Brane, Cyclic and Landscape share string theory; Holographic is a duality, not a plurality; Simulated and Ultimate are philosophical. "Physics keeps finding multiverses" is truer of two or three engines than of nine.
- String-theory optimism. Greene is a leading string theorist, and the book's assessments ("excellent" on required goals; "the consensus is that our universe has a home in the landscape") lean favourable. Fifteen years later the LHC has found no supersymmetry or extra dimensions, and the swampland debate questions whether the landscape's de Sitter vacua exist.
- Alternatives under-represented. Loop quantum gravity, asymptotic safety and other quantum-gravity programmes get a passing mention; Bohmian mechanics and collapse models appear only in notes; non-inflationary cosmologies other than the cyclic model are absent.
- Errors flagged in this edition: Friedmann's paper date (1922, not 1921); Gamow's university (George Washington, not Washington University); the sun's black-hole size (≈6 km across, not 3); Sagittarius A* mass (≈4 million, not 3 million suns); the critical-density figure versus "six hydrogen atoms"; short-range gravity tests (≈50 μm, not 1 μm); Adam Riess omitted from the acceleration discovery.
- Out of date by 2026: the Higgs is found (2012); dark energy may evolve (DESI 2024 to 2025); the cosmic budget is 69/31, not 73/27; the Page curve and "islands" reshaped the information paradox; exascale computers and fluent AI have overtaken the book's computing benchmarks.
- What the book does superbly: it separates in-principle from in-practice testability, states its own biases (reductionism, realism, expansive taste), and treats critics fairly. Its methodological chapters (7 and 11) remain among the clearest popular treatments of the multiverse-as-science debate.
Who's who
| Person | Role | Why they matter |
|---|---|---|
| Albert Einstein | General relativity (1915); cosmological constant (1917) | Built the framework, then distrusted its predictions of expansion and black holes |
| Alexander Friedmann | Expanding solutions (1922) | First to find GR's dynamic universe; also oscillating models (1923) |
| Georges Lemaître | "Primeval atom" (1927) | Father of the big bang |
| Edwin Hubble | Galaxy recession (1929) | Proved expansion; killed the static universe |
| George Gamow, Ralph Alpher, Robert Herman | Relic radiation (late 1940s) | Predicted the CMB; ignored for years |
| Robert Dicke, Jim Peebles | Princeton CMB theory and search (1960s) | Re-derived the CMB; interpreted the discovery |
| Arno Penzias, Robert Wilson | Bell Labs (1965) | Found the CMB by accident; Nobel 1978 |
| Alan Guth | Inflation (1979) | Repulsive gravity from a uniform field |
| Andrei Linde | New and chaotic inflation; eternal inflation | Developed inflation and the multiverse it implies |
| Paul Steinhardt | New inflation; cyclic model | Co-architect of inflation, later its sharpest critic |
| Alexander Vilenkin | Eternal inflation; mediocrity principle | Showed inflation generically never ends |
| John Mather, George Smoot | COBE (1990s) | Detected CMB ripples; Nobel 2006 |
| Peter Higgs and colleagues | Higgs mechanism (1960s) | Mass from a field whose value can differ between bubbles |
| Theodor Kaluza, Oskar Klein | Extra dimensions (1919 to 1926) | Unified gravity and electromagnetism with a hidden dimension |
| John Schwarz, Michael Green, Joël Scherk, Tamiaki Yoneya | Early string theory | Found the graviton; proved consistency (1984) |
| Edward Witten | M-theory (1995); AdS/CFT dictionary | Unified the five string theories; key holography results |
| Joe Polchinski | D-branes; Bousso-Polchinski | Branes as string endpoints; flux landscape |
| Raphael Bousso | Landscape; entropy bounds | Λ distribution across vacua |
| Kachru, Kallosh, Linde, Trivedi | KKLT (2003) | First string vacua with small positive Λ |
| Leonard Susskind | Landscape; holography; complementarity | Chief advocate of the anthropic landscape |
| Steven Weinberg | Anthropic bound on Λ (1987) | Predicted small nonzero Λ before it was seen |
| Brandon Carter | Anthropic principle (1973) | Formalised observer selection |
| Saul Perlmutter, Brian Schmidt (and Adam Riess) | Supernova teams (1998) | Discovered cosmic acceleration; Nobel 2011 |
| Max Tegmark, Martin Rees | Anthropics critique; Mathematical Universe | Varying seeds weaken Weinberg's result; Tegmark champions the Ultimate Multiverse |
| Hugh Everett III | Many Worlds (1957) | Took Schrödinger's equation seriously for everything |
| John Wheeler | Everett's and Bekenstein's adviser | "Information"; named black holes |
| Bryce DeWitt | Revived Everett (1970) | Popularised the "multiworld" |
| Deutsch, Saunders, Wallace, Greaves | Oxford Everettians | Decision-theoretic route to quantum probability |
| Jacob Bekenstein | Black-hole entropy (1972) | Proposed entropy ∝ horizon area |
| Stephen Hawking | Area theorem (1971); radiation (1974) | Proved black holes glow; fixed entropy |
| Gerard 't Hooft | Holographic principle (1993) | Information on boundaries |
| Juan Maldacena | AdS/CFT (1997 to 1998) | Exact holographic duality |
| Kovtun, Son, Starinets | Holographic viscosity | Connected strings to RHIC data |
| Robert Nozick | Principle of fecundity | Every possible universe exists |
| Nick Bostrom | Simulation argument (2003) | Most minds may be simulated |
| Jürgen Schmidhuber, Konrad Zuse, Edward Fredkin | Digital physics | Programs that generate all computable universes |
Connect it
- AI evaluation and the measure problem. "Model X is better 70% of the time" depends entirely on the sampling measure over prompts; fix the reference distribution before claiming superiority, exactly Greene's warning about infinite universes.
- Selection bias in banking data. Credit models trained on approved loans, and fraud models trained on detected fraud, are anthropic samples. Reject inference and counterfactual evaluation are the shoe-shop correction.
- Architecture through dualities. Event-sourced versus state-based designs, and batch versus stream, are often dual descriptions of the same system; choose the representation in which the hard operation is easy.
- Explainability as "two stories." Regulators need the experiential story; model risk needs the mathematical one. Keep them distinct and avoid importing narrative words ("the model decided") into the formal account.
- Capacity at the boundary. Like the holographic bound, real throughput in distributed platforms is usually limited by interfaces (APIs, queues, network edges), not internal compute.
About this book
The Hidden Reality: Parallel Universes and the Deep Laws of the Cosmos by Brian Greene, 2011. 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