Cosmic Inflation

The hot Big Bang is a triumph of extrapolation — run the expansion backwards and it correctly predicts the light-element abundances and the microwave background — yet its starting line is a scandal of fine-tuning. Three facts have no explanation within it. The universe is spatially flat when generic initial curvature should have grown catastrophically. It is uniform in temperature across patches of sky that, in the standard timeline, were never in causal contact. And it is empty of the magnetic monopoles that grand unified theories say should outweigh everything else. Each says the universe began in an extraordinarily special state — and a theory that must assume its own initial conditions has stopped explaining them. Cosmic inflation, proposed by Alan Guth in 1981 and made to work by Linde, Albrecht and Steinhardt a year later, dissolves all three at once with one mechanism: a brief burst of accelerated expansion, $a \propto e^{Ht}$ , driven by the potential energy of a slowly rolling scalar field. Stretch a tiny causal patch by $e^{60}$ and it becomes our entire observable universe — smooth because it was once in contact, flat because curvature was ironed out, and swept clean of relics. What lifts inflation from a fix to a theory is what it does for free: the quantum jitter of the inflaton, stretched to super-horizon size and frozen as classical density perturbations, seeds every galaxy and every hot and cold spot in the microwave sky — a nearly scale-invariant spectrum whose tilt $n_s \approx 0.965$ is a genuine, non-trivial confirmation.

Key takeaways

Inflation is a hypothesized epoch around $10^{-34}\ \mathrm{s}$ of quasi-exponential expansion, $a(t)\propto e^{Ht}$ , driven by the near-constant potential energy of a scalar field, the inflaton.
A single burst of at least $N\simeq 60$ e-folds solves three problems together: the horizon problem (why the CMB is isotropic), the flatness problem (why $\Omega\approx 1$ ), and the relic problem (dilution of monopoles).
Accelerated expansion ( $\ddot a>0$ ) requires an equation of state $w<-1/3$ ; a field with $\rho\approx V$ , $p\approx -V$ gives $w\approx -1$ , mimicking a cosmological constant.
Quantum fluctuations frozen at horizon crossing seed a nearly scale-invariant, adiabatic, Gaussian spectrum of curvature perturbations — the origin of the CMB anisotropies and large-scale structure.
Observations confirm the generic predictions: near-perfect flatness ( $|\Omega_0-1|\lesssim 0.005$ ), a red tilt $n_s=0.9649\pm 0.0042$ , and a tensor bound $r<0.06$ that already excludes the simplest $V\propto\phi^2$ model.
Inflation ends when slow roll breaks down; the inflaton oscillates and decays, reheating the universe and handing off to the standard hot Big Bang.
Key idea: A phase of accelerated expansion ( $\ddot a>0$ , requiring $w<-1/3$ ) driven by a slowly-rolling scalar field stretches a tiny causal patch to encompass the whole observable universe, flattening space and diluting relics, while quantum fluctuations frozen at horizon crossing seed a nearly scale-invariant spectrum of density perturbations.

Where this lives today

Inflation is now a precision science whose frontier is the hunt for primordial gravitational waves. Tensor modes leave a faint curl-like "B-mode" pattern in CMB polarization whose amplitude is a direct dial on the energy scale of inflation. The BICEP/Keck program at the South Pole has driven the bound from $r<0.1$ a decade ago to $r<0.036$ (2021), and next-generation experiments — the Simons Observatory, CMB-S4, and the LiteBIRD satellite — aim for $r\sim 10^{-3}$ . Every tightening of $r$ shears away another family of models: large-field potentials like $V\propto\phi^2$ are already gone, and the surviving favourites are flat "plateau" potentials such as Starobinsky's $R^2$ model and Higgs inflation, which predict $r$ of a few $\times10^{-3}$ . Meanwhile Planck has found no primordial non-Gaussianity, consistent with the simplest single-field models. The sharp open questions remain: what the inflaton is, how it connects to the Standard Model, and how reheating actually proceeds.

The physical picture

The object that makes every piece of inflation click into place is the comoving Hubble radius $1/(aH)$ — the size of the patch that can be in causal contact "now." In the ordinary hot Big Bang it grows without limit: early causal patches were microscopic, so the CMB's uniformity across the sky is a mystery. Inflation reverses the sign. During accelerated expansion $aH$ grows, so $1/(aH)$ shrinks, and a scale inside the horizon is pushed outside it — it "exits." Much later, in the decelerating radiation and matter eras, $1/(aH)$ grows again and the scale "re-enters." The whole observable universe is one patch that was in causal contact before inflation blew it up.

Figure

The comoving Hubble radius 1/(aH) plunges during inflation and grows afterward: because ä > 0 is equivalent to d/dt of 1/(aH) being negative, a comoving scale exits the horizon early and re-enters much later — solving the horizon problem needs N ≳ 60 e-folds

Where does the accelerated expansion come from? Picture the inflaton as a ball on a very shallow, nearly flat hill. While the slope is gentle the ball creeps along, so its height — the potential energy $V(\phi)$ — barely changes and dominates the energy budget, acting like a temporary cosmological constant that drives $a\propto e^{Ht}$ and, as a term $\propto a^{-2}$ , crushes any curvature toward $\Omega=1$ . When the hill steepens the ball speeds up, slow roll ends, and the field oscillates about the minimum, dumping its energy into a hot plasma. The same quantum jitter, stretched beyond the horizon, freezes into the density seeds of all structure. The payoff is the plot cosmologists use to judge models: the $n_s$ – $r$ plane. Every potential predicts a point here, and Planck plus BICEP/Keck have shrunk the allowed region enough to start killing models — the textbook $V\propto\phi^2$ is already outside it.

Figure

The n_s–r plane where inflation meets data: Planck plus BICEP/Keck give n_s = 0.965 and r < 0.06, ruling out the simplest chaotic model V ∝ φ² (n_s = 0.967, r = 0.13) while favouring flat plateau potentials like Starobinsky R² (n_s = 0.965, r = 0.004)

The core equations

Inflation is built from two equations of motion — one for the expansion of space, one for the field that drives it. We assemble them, extract the condition for acceleration, and end with the small parameters that turn the whole apparatus into two measurable numbers, $n_s$ and $r$ . The Friedmann equation ties the expansion rate to the energy content and the spatial curvature. It is the engine's tachometer: given $\rho$ and $k$ it returns $H$ .

H^2=\left(\frac{\dot a}{a}\right)^2=\frac{8\pi G}{3}\rho-\frac{kc^2}{a^2} \qquad (1)

where: $H$ — Hubble parameter, $\dot a/a$ ; $a$ — cosmic scale factor; $\rho$ — total energy density; $k$ — curvature constant ( $-1,0,+1$ ); $G$ — Newton's constant. Reading: the curvature term falls as $a^{-2}$ , so anything that grows $a$ enormously renders $k$ irrelevant. The acceleration equation is the second Friedmann equation, and it is where inflation lives: only a sufficiently negative pressure can make $\ddot a>0$ .

\frac{\ddot a}{a}=-\frac{4\pi G}{3}\left(\rho+\frac{3p}{c^2}\right) \qquad (2)

where: $\ddot a$ — acceleration of the scale factor; $p$ — pressure. Reading: gravity from ordinary matter and radiation ( $\rho,p>0$ ) always decelerates; accelerating requires $\rho+3p/c^2<0$ , i.e. $w<-1/3$ . A homogeneous scalar field supplies exactly that. Its energy density and pressure are a kinetic piece plus (or minus) the potential, so a field pinned high on a flat potential is almost pure vacuum energy.

\rho_\phi c^2=\tfrac12\dot\phi^2+V(\phi),\qquad p_\phi=\tfrac12\dot\phi^2-V(\phi),\qquad w=\frac{\tfrac12\dot\phi^2-V}{\tfrac12\dot\phi^2+V} \qquad (3)

where: $\tfrac12\dot\phi^2$ — kinetic energy density; $V(\phi)$ — potential energy density; $w$ — equation-of-state parameter. Reading: when $\dot\phi^2\ll V$ the pressure approaches $-\rho c^2$ , $w\to -1$ , and the field drives near-de Sitter expansion. The field obeys a Klein–Gordon equation in the expanding background, in which the expansion acts as friction — the $3H\dot\phi$ term — that lets the field creep rather than oscillate.

\ddot\phi+3H\dot\phi+V'(\phi)=0 \qquad (4)

where: $\phi$ — inflaton field; $V'=\mathrm{d}V/\mathrm{d}\phi$ — slope; $3H\dot\phi$ — Hubble friction. Reading: if $\ddot\phi$ is negligible, the field reaches "terminal velocity" $\dot\phi\approx -V'/3H$ — slow roll. The quality of slow roll is captured by two dimensionless parameters built from the shape of the potential — the slope and the curvature — both of which must be small for inflation to last.

\epsilon=\frac{m_\mathrm{Pl}^2}{16\pi}\left(\frac{V'}{V}\right)^2,\qquad \eta=\frac{m_\mathrm{Pl}^2}{8\pi}\frac{V''}{V} \qquad (5)

where: $\epsilon$ — first slow-roll parameter (slope); $\eta$ — second slow-roll parameter (curvature); $m_\mathrm{Pl}=\sqrt{\hbar c/G}$ — Planck mass. Reading: inflation proceeds while $\epsilon,|\eta|\ll 1$ and ends when $\epsilon\simeq 1$ ; these two numbers alone fix the observables. The total amount of inflation is measured in e-folds, the logarithmic growth of the scale factor — and it can be read directly off the potential.

N=\ln\frac{a_\mathrm{end}}{a_i}=\int_{t_i}^{t_\mathrm{end}}H\,\mathrm{d}t\approx\frac{8\pi}{m_\mathrm{Pl}^2}\int_{\phi_\mathrm{end}}^{\phi_i}\frac{V}{V'}\,\mathrm{d}\phi \qquad (6)

where: $N$ — number of e-folds; $\phi_i,\phi_\mathrm{end}$ — field values at the start and end of inflation. Reading: solving the horizon and flatness problems requires $N\gtrsim 60$ ; the exact number depends on the reheating temperature and the energy scale. Flatness is the cleanest consequence. The deviation of $\Omega$ from unity is set entirely by the comoving Hubble rate, which grows exponentially while $H$ is nearly constant.

|\Omega(t)-1|=\frac{|k|c^2}{a^2H^2}=\frac{|k|c^2}{(aH)^2} \qquad (7)

where: $\Omega=\rho/\rho_c$ — total density parameter; $aH$ — comoving Hubble rate; $1/(aH)$ — comoving Hubble radius. Reading: since $aH\propto e^{Ht}$ , $|\Omega-1|\propto e^{-2Ht}$ is driven exponentially to zero regardless of where it started. Finally, the fluctuations. The primordial curvature spectrum is nearly a power law whose tilt away from flat encodes the slow-roll parameters, and a companion tensor spectrum sets the tensor-to-scalar ratio.

\mathcal{P}_\mathcal{R}(k)=A_s\left(\frac{k}{k_*}\right)^{n_s-1},\qquad n_s-1=-6\epsilon+2\eta,\qquad r=16\epsilon \qquad (8)

where: $\mathcal{P}_\mathcal{R}$ — power spectrum of curvature perturbations; $A_s$ — scalar amplitude; $k_*$ — pivot scale; $n_s$ — scalar spectral index ( $n_s=1$ is exact scale invariance); $r$ — tensor-to-scalar ratio. Reading: a red tilt $n_s<1$ and a small $r$ are the fingerprints of a nearly flat potential.

Derivation AThe horizon problem and its inflationary resolution

The goal is to make the horizon problem quantitative and then show that a single sign flip cures it. In the standard hot Big Bang the microwave sky is divided into tens of thousands of patches whose past light cones never overlapped — yet they share a temperature to one part in $10^5$ . We estimate how many patches there are, show that $1/(aH)$ grows in any decelerating universe, and then show that accelerated expansion reverses that growth — so the whole sky descends from one causally connected region if inflation lasts about sixty e-folds.

Step 1. At recombination the comoving particle horizon subtends only about $1^\circ$ – $2^\circ$ on the sky, so the last-scattering surface breaks into of order $10^4$ – $10^5$ causally disconnected patches. Why: without a mechanism to correlate them, their near-identical temperatures are an unexplained conspiracy of initial data.

Step 2. In a decelerating universe the comoving Hubble radius grows, $\frac{\mathrm{d}}{\mathrm{d}t}\frac{1}{aH}=-\frac{\ddot a}{(\dot a)^2}>0$ when $\ddot a<0$ . Radiation ( $a\propto t^{1/2}$ ) and matter ( $a\propto t^{2/3}$ ) domination both satisfy this, so causal contact is a recent and expanding luxury — too late to homogenize the CMB. Why: this is the precise statement that "there was not enough time" for the sky to reach equilibrium.

Step 3. Accelerated expansion flips the sign: $\ddot a>0\ \Rightarrow\ \frac{\mathrm{d}}{\mathrm{d}t}\frac{1}{aH}<0$ , so the comoving Hubble radius shrinks. Scales that were sub-horizon before inflation are swept outside it, and the entire observable universe traces back to one pre-inflationary causal patch. Why: exit and later re-entry of scales is the whole geometric content of inflation.

Step 4. Demanding that the comoving scale entering the horizon today lay inside the patch that was causally connected before inflation gives the requirement below, with the exact figure set by the reheating temperature and inflationary scale. Why: sixty is not fundamental; it is the number of e-folds needed to place today's Hubble radius back inside the pre-inflationary horizon.

N_\mathrm{min}=\ln\!\left(\frac{a_\mathrm{end}H_\mathrm{end}}{a_0H_0}\right)\gtrsim 60 \qquad\text{(e-folds needed)}

Go deeper · Derivation B · 1 minSlow-roll dynamics and the spectral tilt

Derivation BSlow-roll dynamics and the spectral tilt

The goal is to reduce the coupled field-plus-Friedmann system to two algebraic relations and, from them, derive the observable spectral tilt as a direct measure of the flatness of the potential — assuming the kinetic energy is subdominant and the motion friction-dominated. We drop the two "fast" terms, define the slow-roll parameters that quantify how good that approximation is, and show that the amplitude a mode inherits at horizon crossing drifts slightly with scale — tilting the spectrum away from exact scale invariance by an amount fixed by $\epsilon$ and $\eta$ .

Step 1. Neglect acceleration and kinetic energy: drop $\ddot\phi$ in equation (4) and $\tfrac12\dot\phi^2$ in equation (3). The dynamics collapse to the boxed slow-roll pair below. Why: on a flat enough potential the field is overdamped, so its velocity is slaved to the local slope and its energy is almost pure $V$ .

Step 2. Define $\epsilon=\frac{m_\mathrm{Pl}^2}{16\pi}(V'/V)^2$ and $\eta=\frac{m_\mathrm{Pl}^2}{8\pi}(V''/V)$ . Both must be $\ll 1$ for the approximation of Step 1 to hold, and $\epsilon=1$ marks the moment acceleration ceases. Why: $\epsilon$ measures the slope and $\eta$ the curvature of the hill — the two ways the potential can fail to be flat.

Step 3. Each comoving mode $k$ has its amplitude fixed when it exits the horizon, $k=aH$ . Because $H$ and $\dot\phi$ evolve slowly, modes that cross at slightly different times inherit slightly different amplitudes, giving a spectrum that is nearly, but not exactly, independent of scale. Why: perfect scale invariance would require a perfectly static de Sitter background; the slow drift of $H$ is precisely what tilts the spectrum.

Step 4. Working the scale dependence through gives $n_s-1=-6\epsilon+2\eta$ , so a measured $n_s\neq 1$ is a direct readout of the potential's slope and curvature at the scale probed. Why: exact scale invariance ( $n_s=1$ ) needs $\epsilon=\eta=0$ ; Planck's $n_s=0.9649\pm0.0042$ is a nonzero, model-discriminating number.

3H\dot\phi\approx -V'(\phi),\qquad H^2\approx\frac{8\pi G}{3c^2}V(\phi) \qquad\text{(slow-roll equations)}

Go deeper · Derivation C · 1 minQuantum fluctuations and the amplitude of the primordial spectrum

Derivation CQuantum fluctuations and the amplitude of the primordial spectrum

The goal is to trace the observed lumpiness of the universe to the quantum vacuum, and in doing so pin down the amplitude $A_s$ that fixes the overall size of the inflaton potential. During inflation the near-de Sitter horizon endows any light field with an irreducible fluctuation of order the Hubble rate; stretched beyond the horizon it freezes into a classical shift in the local expansion history — a curvature perturbation. We turn that field fluctuation into a curvature perturbation, form its power spectrum, and evaluate it, working with the reduced Planck mass $M_\mathrm{Pl}=m_\mathrm{Pl}/\sqrt{8\pi}=2.4\times10^{18}\ \mathrm{GeV}$ .

Step 1. In a de Sitter background the vacuum fluctuation of a light scalar, evaluated as each mode leaves the horizon, is $\delta\phi\approx H/2\pi$ — the Gibbons–Hawking result. Why: the expanding horizon acts like a thermal bath at temperature $H/2\pi$ , imprinting a universal jitter on the field independent of the details of $V$ .

Step 2. A field displacement shifts the time the local patch spends inflating, so the frozen curvature perturbation is $\mathcal{R}=H\,\delta\phi/\dot\phi$ . Squaring and using $\dot\phi=-V'/3H$ gives the power spectrum $\mathcal{P}_\mathcal{R}=(H/\dot\phi)^2(H/2\pi)^2$ . Why: regions that inflate a little longer end up slightly denser — the "separate-universe" picture that turns a field jitter into a density seed.

Step 3. Substituting $H^2=V/3M_\mathrm{Pl}^2$ and the slow-roll parameter $\epsilon$ collapses the amplitude to the compact form boxed below, evaluated at horizon crossing of the pivot scale. Why: every model-dependent factor folds into $V$ and $\epsilon$ , so the amplitude directly constrains the height and steepness of the potential.

Step 4. Setting $\mathcal{P}_\mathcal{R}=A_s=2.1\times10^{-9}$ and using $r=16\epsilon$ fixes $V^{1/4}=[(3/2)\pi^2 A_s\,r]^{1/4}M_\mathrm{Pl}$ . Why: a detection of $r$ would immediately reveal the energy at which inflation occurred — near the grand-unified scale for $r\sim 0.01$ .

\mathcal{P}_\mathcal{R}=\frac{1}{24\pi^2}\frac{V}{M_\mathrm{Pl}^4\,\epsilon}=A_s\approx 2.1\times10^{-9} \qquad\text{(scalar amplitude)}

The amplitude formula $\mathcal{P}_\mathcal{R}=V/(24\pi^2 M_\mathrm{Pl}^4\epsilon)$ follows directly from the chain $\delta\phi=H/2\pi$ , $\mathcal{R}=H\delta\phi/\dot\phi$ , $\dot\phi=-V'/3H$ , $H^2=V/3M_\mathrm{Pl}^2$ , and yields $V^{1/4}\approx1.6\times10^{16}$ GeV at $r=0.06$ .

Limiting cases

The behaviour of an accelerating cosmology is governed by the equation of state $w=p/\rho c^2$ and the two slow-roll parameters; the regimes below bracket the physics.

Case	Result	Meaning
Pure de Sitter	$w=-1,\ H=\text{const},\ a\propto e^{Ht}$	The idealized limit: constant vacuum energy, exact exponential growth. Real inflation is only quasi-de Sitter — a tilt is needed to end it and to match $n_s<1$ .
Acceleration threshold	$\ddot a>0 \iff w<-\tfrac13 \iff \rho+3p/c^2<0$	The general condition for accelerated expansion; any fluid violating the strong energy condition inflates.
Slow-roll regime	$\epsilon\ll 1,\ \|\eta\|\ll 1$	Inflation proceeds while both parameters are small; the observables are $n_s-1=-6\epsilon+2\eta$ and $r=16\epsilon$ .
End of inflation	$\epsilon\simeq 1$	Slow roll breaks down and acceleration stops; the field enters the steep part of the potential.
Reheating	$\rho_\phi\propto a^{-3}$ (coherent oscillations)	The inflaton oscillates like pressureless matter, then decays into radiation, setting the initial temperature of the hot Big Bang.

Common trapInflation makes space expand faster than light and breaks causality

It does not, and the confusion is worth dismantling. Nothing moves through space faster than light during inflation; it is space itself that stretches, and the recession of two comoving points is not a local velocity that special relativity limits. The precise statement is geometric: the comoving Hubble radius $1/(aH)$ shrinks, so scales exit the horizon — no signal outruns light in any local frame. Two more traps: first, "exponential expansion" is only approximate — $H$ decreases slowly (quasi-de Sitter), and that very deviation produces the measured $n_s\neq 1$ ; a perfectly exponential universe would give the ruled-out $n_s=1$ . Second, "sixty e-folds" is no constant of nature — it is set by the reheating temperature and energy scale, and quoting it as fixed hides a real model dependence.

Worked examples

1 · Chaotic inflation with V = ½m²φ², worked to the observables (confidence builder)

The simplest chaotic model has $V=\tfrac12 m^2\phi^2$ . Using $N=60$ e-folds before the end of inflation, find the slow-roll parameters, the field value $\phi_i$ , and the predicted $n_s$ and $r$ — then compare with the data.

Step 1. Compute the slow-roll parameters from equation (5). With $V'=m^2\phi$ and $V''=m^2$ , $\epsilon=\frac{m_\mathrm{Pl}^2}{16\pi}(V'/V)^2=\frac{m_\mathrm{Pl}^2}{16\pi}(2/\phi)^2=\dfrac{m_\mathrm{Pl}^2}{4\pi\phi^2}$ , and identically $\eta=\frac{m_\mathrm{Pl}^2}{8\pi}(2/\phi^2)=\dfrac{m_\mathrm{Pl}^2}{4\pi\phi^2}$ . So for this potential $\boldsymbol{\epsilon=\eta}$ .

Step 2. Relate $N$ to $\phi$ with equation (6): $N=\frac{8\pi}{m_\mathrm{Pl}^2}\int_{\phi_\mathrm{end}}^{\phi_i}\frac{\phi}{2}\,\mathrm{d}\phi=\frac{2\pi}{m_\mathrm{Pl}^2}(\phi_i^2-\phi_\mathrm{end}^2)$ . Inflation ends at $\epsilon=1$ , i.e. $\phi_\mathrm{end}=m_\mathrm{Pl}/\sqrt{4\pi}\approx 0.28\,m_\mathrm{Pl}$ , which is negligible against $\phi_i$ . Hence $\phi_i^2\approx N m_\mathrm{Pl}^2/2\pi$ and, combining with Step 1, $\boldsymbol{\epsilon=1/(2N)}$ .

Step 3. Evaluate at $N=60$ . The field is super-Planckian, $\phi_i=\sqrt{N/2\pi}\,m_\mathrm{Pl}\approx \mathbf{3.1\,m_\mathrm{Pl}}$ (a hallmark of large-field models), and $\epsilon=\eta=1/120\approx 0.0083$ .

Step 4. Read off the observables from equation (8): $n_s-1=-6\epsilon+2\eta=-4\epsilon=-2/N=-2/60$ , so $n_s=1-2/N=\mathbf{0.967}$ , and $r=16\epsilon=8/N=8/60=\mathbf{0.13}$ . The tilt is a beautiful match to $n_s=0.9649$ , but the tensor ratio $r=0.13$ sits far above the bound $r<0.06$ . Conclusion: V ∝ φ² predicts n_s = 0.967 (good) but r = 0.13 (excluded) — the plot's red point, ruled out by BICEP/Keck.

2 · The energy scale of inflation from the tensor bound

Given the scalar amplitude $A_s=2.1\times10^{-9}$ and the upper bound $r<0.06$ , estimate the energy scale $V^{1/4}$ of inflation and comment on where it sits.

Step 1. From Derivation 3, $\mathcal{P}_\mathcal{R}=V/(24\pi^2 M_\mathrm{Pl}^4\epsilon)=A_s$ . Eliminate $\epsilon$ with $r=16\epsilon$ : $V=24\pi^2 M_\mathrm{Pl}^4(r/16)A_s=\tfrac32\pi^2 A_s\,r\,M_\mathrm{Pl}^4$ .

Step 2. Insert $r=0.06$ : $\tfrac32\pi^2(2.1\times10^{-9})(0.06)=1.9\times10^{-9}$ , so $V^{1/4}=(1.9\times10^{-9})^{1/4}M_\mathrm{Pl}=6.6\times10^{-3}\,M_\mathrm{Pl}$ .

Step 3. With $M_\mathrm{Pl}=2.4\times10^{18}\ \mathrm{GeV}$ , $V^{1/4}=\mathbf{1.6\times10^{16}\ GeV}$ — essentially the grand-unified scale, which is why a B-mode detection would probe physics far beyond the LHC. The bound r < 0.06 caps V¹ᐟ⁴ at ≈ 1.6×10¹⁶ GeV, the GUT scale.

Physics of the portrait

Inflation has a clear moment of birth: a young postdoc, worrying about magnetic monopoles, realized in December 1979 that a stretch of exponential expansion would erase them, flatten space, and homogenize the sky all at once.

Origins · 1 minAlan Guth and how this was discovered

Alan Guth

b. 1947 · the horizon, flatness, and monopole problems and the inflationary universe

Alan Harvey Guth was a particle theorist drifting between postdocs when, worrying about the overproduction of magnetic monopoles in grand unified theories, he wrote in his notebook the words "spectacular realization": a phase of exponential expansion would dilute the monopoles to invisibility, and — he soon saw — solve the flatness and horizon problems in the same stroke. His 1981 paper, "Inflationary universe: A possible solution to the horizon and flatness problems," gave the mechanism its name, though Guth was candid that his "old inflation," ending by bubble nucleation from a false vacuum, could not gracefully reheat. Within a year Andrei Linde, and independently Andreas Albrecht and Paul Steinhardt, replaced the first-order transition with a field slowly rolling down a smooth potential — "new inflation" — which reheats cleanly and became the template for everything since. The deepest twist was that the same quantum fluctuations Guth's expansion stretched would seed galaxies, worked out at the 1982 Nuffield workshop and later matched, digit for digit, by the microwave background. Guth, long at MIT, turned a puzzle about initial conditions into a predictive, tested theory of cosmic origins.

History

1980

— Starobinsky's R² gravity yields the first model of inflationary expansion

1981

— Guth frames the horizon, flatness, and monopole problems and proposes old inflation

1982

— Linde, and Albrecht and Steinhardt, introduce slow-roll new inflation

1983

— Linde proposes chaotic inflation from a simple V = ½m²φ² potential

2018

— Planck measures n_s = 0.965, ruling out the simplest V ∝ φ² models

Going further

Texts. Ryden, Introduction to Cosmology, 2nd ed., ch. 10 (The Inflationary Universe) · Dodelson, Modern Cosmology, 2nd ed., ch. 6 · Carroll & Ostlie, An Introduction to Modern Astrophysics, 2nd ed., §30.3 · Mo, van den Bosch & White, Galaxy Formation and Evolution, ch. 3 · Baumann, Cosmology (2022), Part III · Guth, Phys. Rev. D 23, 347 (1981). Key numerical checks: the acceleration condition $w<-1/3$ follows from equation (2); 60 e-folds suppress curvature by $e^{-2N}\sim e^{-120}\sim10^{-52}$ ; the slow-roll observables are $n_s-1=-6\epsilon+2\eta$ and $r=16\epsilon$ ; the $V\propto\phi^2$ model gives $n_s=1-2/N=0.967$ , $r=8/N=0.13$ , and $\phi_i\approx3.1\,m_\mathrm{Pl}$ at $N=60$ ; the scalar amplitude formula $\mathcal{P}_\mathcal{R}=V/(24\pi^2 M_\mathrm{Pl}^4\epsilon)$ yields $V^{1/4}\approx1.6\times10^{16}$ GeV at $r=0.06$ . Observational values ( $n_s=0.9649\pm0.0042$ , $A_s\approx2.1\times10^{-9}$ , $r<0.06$ tightening to $r<0.036$ , $\Omega_0=1.000\pm0.005$ ) are from Planck 2018 (A&A 641, A10) and BICEP/Keck 2021.

Prerequisites. Thermal History of the Universe · Equilibrium Thermodynamics of the Early Universe · Effective Number of Relativistic Degrees of Freedom. Leads to. Baryon Asymmetry and Baryogenesis · Big Bang Nucleosynthesis · Recombination and the Saha Equation · Entropy Conservation and Comoving Entropy.

Ask the tutor: Ask me to derive the slow-roll observables for any potential, walk through why $1/(aH)$ shrinking solves the horizon problem, or work out the energy scale of inflation from a value of $r$ .

Is inflation actually proven, or is it still a hypothesis?

It is a well-tested framework rather than a proven fact. Inflation was designed to solve the horizon, flatness, and monopole problems, so those successes are partly built in; its real credit comes from a prediction it did not have to get right — a nearly scale-invariant, adiabatic, Gaussian spectrum with a slight red tilt. That $n_s=0.965<1$ , together with the observed flatness and the acoustic-peak pattern, is exactly what generic slow-roll inflation forecast before the data arrived. What is not established is the specific model: the inflaton's identity is unknown, and alternatives (bouncing and string-gas cosmologies) still contend. A primordial B-mode detection would be the closest thing to a smoking gun.

What is the inflaton — is it a particle we know?

No known particle fits. The inflaton is a hypothesized scalar field whose potential energy drove inflation, and the only fundamental scalar in the Standard Model, the Higgs, has the wrong potential on its own (though "Higgs inflation," with a large non-minimal coupling to gravity, is a serious proposal). Model builders draw inflatons from grand unified theories, supersymmetry, or string-theory moduli, or simply posit a flat potential and ask what data require. Because the observables — $n_s$ , $r$ , the running — depend only on the shape of the potential near horizon crossing, we can constrain that shape tightly without knowing what the field "is." Pinning the inflaton to specific high-energy physics remains a central open problem.

How can regions expand apart faster than light without violating relativity?

Because the speed limit governs motion through space, not the stretching of space. During inflation two comoving points stay at fixed comoving coordinates; the physical distance between them grows because the scale factor grows, and that apparent recession can exceed $c$ without anything locally outrunning light. No information rides the stretching — you cannot send a signal on an expanding metric. Coordinate-free, the statement is that $1/(aH)$ shrinks, so scales cross outside the horizon while every local light cone still obeys special relativity — the same reason distant galaxies recede superluminally even today.

Why does everyone care so much that n_s is slightly less than 1?

Because that small deviation is the difference between a fit and a prediction. Exact scale invariance, $n_s=1$ (the "Harrison–Zel'dovich" spectrum), was an old guess with no dynamics behind it. Inflation instead predicts $n_s$ slightly below 1, because the Hubble rate falls slowly as the field rolls, and the shortfall is set by the slow-roll parameters through $n_s-1=-6\epsilon+2\eta$ . Measuring $n_s=0.9649\pm0.0042$ — five sigma from 1 — confirms the spectrum is tilted, and tilted the predicted way by a plausible amount. It is the cleanest quantitative success of the paradigm.

What are primordial gravitational waves, and why hunt for B-modes?

Inflation stretches not only the inflaton's fluctuations but the quantum fluctuations of the metric itself, producing a background of primordial gravitational waves — tensor perturbations. Their amplitude, encoded in $r=16\epsilon$ , scales with the fourth power of the inflationary energy, so measuring $r$ reads off the energy scale directly. These waves leave a distinctive curl-like ("B-mode") pattern in CMB polarization that density perturbations cannot produce at large angular scales, which is why B-modes are the prize target. A confirmed primordial B-mode signal would pin inflation near the grand-unified scale and be the first glimpse of quantum gravity in action.

How does inflation end, and where does all the matter come from?

Inflation ends when the field reaches the steep part of its potential and $\epsilon$ grows to 1: slow roll fails, acceleration stops, and the inflaton oscillates about the minimum of $V$ . Those coherent oscillations behave like pressureless matter, and as the field decays it pours its energy into a hot plasma — reheating, which sets the temperature at which the standard hot Big Bang begins. Crucially, inflation has diluted every pre-existing particle to nothing, so all matter, radiation, and entropy must be created here — which is why the observed baryon asymmetry cannot predate inflation and must be generated afterward by baryogenesis. (In many models the field's quantum jitter keeps some regions inflating forever — "eternal inflation" — spawning a multiverse whose status as prediction or embarrassment is unsettled.)