The Horizon Problem

The hot Big Bang, built on the FLRW metric and the Friedmann equations, is one of the most successful physical theories ever written down: it predicts the expansion of the universe, the abundances of the light elements, and the near-perfect blackbody spectrum of the cosmic microwave background. Yet the same model harbours a puzzle not about dynamics but about initial conditions. The CMB photons arriving from opposite directions on the sky last scattered at recombination, redshift $z\simeq 1090$ , on a thin spherical shell — the last-scattering surface — and they carry one temperature, $T_0=2.725\,\mathrm{K}$ , matched across the whole sphere to within fluctuations of order $10^{-5}$ . The question is not what the temperature is. It is why it agrees at all. The heart of the matter is a single sign. For any fluid that gravitates normally — matter or radiation, with $w>-1/3$ — the expansion decelerates and the comoving Hubble radius $1/(aH)$ grows with time. The largest comoving distance light could have travelled since the Big Bang, the particle horizon $\chi_{\rm ph}=c\!\int\! \mathrm{d}t/a$ , is therefore finite, and at recombination it subtends only about $1.7^\circ$ on today's sky. Two CMB points farther apart than that lay outside each other's horizons when they emitted — no signal of any kind could have passed between them. The sky is a mosaic of $\sim 10^4$ causally disconnected patches painted, impossibly, the same colour. This is not a contradiction of general relativity; it is a demand for an outrageously fine-tuned initial state. The standard cure is inflation — a finite early epoch with $w\approx -1$ and $\ddot a>0$ , during which $1/(aH)$ shrinks, stretching a single thermalized patch to engulf everything we now see so that the CMB's uniformity is explained rather than assumed. This article makes that argument quantitative.

Key takeaways

In the hot Big Bang the particle horizon at recombination subtends only $\theta_{\rm hor}\approx 1.7^\circ$ on the sky, yet the CMB is isotropic to one part in $10^{5}$ over the entire celestial sphere.
Regions of the last-scattering surface separated by more than $\theta_{\rm hor}$ were never in causal contact, so no physical process could have equilibrated their temperatures — and yet they match.
The root cause is the sign of $w$ : for ordinary matter and radiation ( $w>-1/3$ ) the comoving Hubble radius $1/(aH)\propto a^{(1+3w)/2}$ grows, so the largest causally connected comoving scale was tiny in the early universe.
Counting patches gives $N\sim 4/\theta_{\rm hor}^2\sim 10^{4}$ independent regions on the CMB sky, each of which must have had its temperature set by hand in the standard model.
The horizon problem is a fine-tuning of initial conditions, not a failure of GR; the standard resolution is cosmic inflation, an early epoch of accelerated expansion ( $\ddot a>0$ , $w<-1/3$ ) that reverses the sign and shrinks $1/(aH)$ .
It travels with the flatness problem and the relic-monopole problem — collectively the three motivations Guth used to propose inflation in 1981.
Key idea: for any fluid with $w>-1/3$ the comoving Hubble radius $1/(aH)$ grows, so the particle horizon at recombination is small and the CMB is many causally disconnected patches at one temperature; reversing the sign — making $1/(aH)$ shrink, i.e. $\ddot a>0$ — is exactly what inflation does to solve the problem.

The core insight

For any cosmological fluid with $w>-1/3$ the comoving Hubble radius $1/(aH)$ grows with time, so the particle horizon at recombination is small and the CMB consists of many causally disconnected patches that nevertheless share one temperature; reversing the sign — making $1/(aH)$ shrink, i.e. $\ddot a>0$ — is exactly what inflation does to solve the problem.

Where this lives today

The horizon problem is now the central design constraint of early-universe cosmology. Precision CMB experiments — WMAP, Planck, and ground-based ACT and SPT — measure the angular power spectrum whose first acoustic peak sits at multipole $\ell\approx 220$ (angular scale $\sim1^\circ$ ), the sound horizon at recombination and a close cousin of the causal horizon. Crucially, they detect temperature correlations on scales larger than that horizon — super-horizon correlations the hot Big Bang forbids but inflation predicts, because those scales were in causal contact before they exited the horizon during inflation. The same shrinking comoving Hubble radius also freezes quantum fluctuations into a nearly scale-invariant spectrum with spectral index $n_s\approx 0.965$ , which Planck now pins to better than one percent. The frontier is the hunt for primordial gravitational waves: the tensor-to-scalar ratio $r$ , sought through CMB B-mode polarization by BICEP/Keck, the Simons Observatory, LiteBIRD, and CMB-S4, would fix the inflationary energy scale and directly probe the epoch that made the sky causally whole. Meanwhile the horizon problem remains the litmus test for every alternative — bouncing, string-gas, ekpyrotic, and varying-speed-of-light scenarios — each of which must supply its own mechanism for connecting the CMB.

The physical picture

Think of the CMB sky as a mosaic. Each tile is a region that has had time, since the Big Bang, for a light signal to cross it — a causally connected patch. The arithmetic below fixes the tile at a couple of degrees across, so the full $4\pi$ sphere is thousands of tiles. The scandal is that every tile is the same colour to one part in $10^5$ , though no signal could ever have passed between most of them. The cleanest way to see the impossibility is a conformal spacetime diagram: draw the past light cone of each CMB photon back to the Big Bang and watch the cones of two opposite points fail to overlap.

Why is the horizon so small? Because gravity decelerates ordinary expansion, and a decelerating universe has a comoving Hubble radius that grows. Run the film backwards and $1/(aH)$ collapses toward zero at the Big Bang, so the comoving distance light could cover early on is tiny — there was not enough conformal time to build a large horizon. Inflation flips the sign: a phase of acceleration makes $1/(aH)$ shrink going forward, so a microscopic causal patch inflates to contain the whole observable universe. The single most useful figure in the subject plots $1/(aH)$ against the scale factor and shows a fixed comoving scale exiting the horizon during inflation and re-entering long afterward.

Figure

The comoving Hubble radius 1/(aH) ∝ a^((1+3w)/2) versus scale factor. During inflation w ≈ −1 so it shrinks (∝ a⁻¹); afterwards radiation w = 1/3 gives ∝ a and matter w = 0 gives ∝ a^(1/2), so it grows. A fixed comoving scale (dashed) exits the horizon during inflation and re-enters much later — the same scale that is super-horizon at recombination. Solving the horizon problem needs the inflationary drop to span at least N_e ≈ 60 e-folds.

The core equations

Everything begins with the geometry. On large scales the universe is described by the spatially homogeneous, isotropic FLRW line element, whose single dynamical unknown is the scale factor $a(t)$ . Light propagation, distances, and horizons are all read off from this metric.

\mathrm{d}s^2 = -c^2\,\mathrm{d}t^2 + a^2(t)\left[\frac{\mathrm{d}r^2}{1-kr^2} + r^2\,\mathrm{d}\Omega^2\right] \qquad (1)

where: $a(t)$ — scale factor, normalised $a(t_0)=1$ today; $k$ — spatial curvature constant, $k=0$ for the observed nearly flat universe; $r$ — comoving radial coordinate; $\mathrm{d}\Omega^2=\mathrm{d}\theta^2+\sin^2\theta\,\mathrm{d}\phi^2$ — the unit two-sphere metric. Reading: proper distances stretch as $a(t)$ while comoving coordinates stay fixed, so a photon's coordinate speed is $c/a$ . The causal reach at any epoch is the comoving particle horizon: the total comoving distance a photon can have covered since the Big Bang. It is a cumulative integral over the whole past, and its convergence is the crux of the whole problem.

\chi_{\rm ph}(t) = c\int_0^{t}\frac{\mathrm{d}t'}{a(t')} \qquad (2)

where: $\chi_{\rm ph}$ — comoving distance light has travelled since $t'=0$ ; $c$ — speed of light; $t$ — cosmic time at the epoch of interest. Reading: because $a\to 0$ at early times, the integrand can diverge — but for $w>-1/3$ it does not, so the horizon is finite and, at recombination, small. The diagnostic that decides the sign of everything is the comoving Hubble radius, $1/(aH)$ : the comoving scale that is just crossing the horizon now. Its dependence on the scale factor follows from combining energy conservation with the Friedmann equation for a single fluid.

\frac{1}{aH} \propto a^{\frac{1+3w}{2}} \qquad (3)

where: $H=\dot a/a$ — Hubble parameter; $w$ — equation-of-state parameter, $p=w\rho c^2$ ; $1/(aH)$ — comoving Hubble radius. Reading: the exponent flips sign at $w=-1/3$ . For $w>-1/3$ (matter, radiation) it grows; for $w<-1/3$ (a cosmological constant, inflation) it shrinks. That sign flip is the entire story. The dynamical meaning of $w=-1/3$ is written cleanly in the second Friedmann (acceleration) equation: pressure gravitates, and sufficiently negative pressure accelerates the expansion.

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

where: $\ddot a$ — second time derivative of the scale factor; $G$ — Newton's constant; $\rho,\,p$ — energy density and pressure. Reading: $\ddot a>0$ requires $\rho+3p/c^2<0$ , i.e. $w<-1/3$ — exactly the condition under which $1/(aH)$ shrinks. Acceleration and a shrinking comoving Hubble radius are the same statement. Translating the small horizon into an observable means dividing the proper horizon size at recombination by the angular-diameter distance to the last-scattering surface, then tiling the sphere with patches of that angular radius.

\theta_{\rm hor} \approx \frac{d_{\rm hor}(t_{\rm rec})}{d_A(t_{\rm rec})} \sim 1.7^\circ,\qquad N \sim \frac{4}{\theta_{\rm hor}^2}\sim 10^{4} \qquad (5)

where: $d_{\rm hor}$ — proper particle-horizon size at recombination; $d_A$ — angular-diameter distance to the last-scattering surface; $\theta_{\rm hor}$ — angle the horizon subtends on today's sky (in radians for $N$ ); $N$ — number of causally independent patches. Reading: a degree-scale horizon slices the sky into thousands of regions that never communicated. Against that tiny causal patch stand the numbers the CMB actually shows: a single monopole temperature and fluctuations smaller than it by five orders of magnitude, over the whole sky.

1+z_{\rm rec}\simeq 1090,\quad T_{\rm rec}=T_0(1+z_{\rm rec})\simeq 3000\,\mathrm{K},\quad \frac{\Delta T}{T}\sim 10^{-5} \qquad (6)

where: $z_{\rm rec}$ — redshift of recombination / last scattering; $T_{\rm rec}$ — photon temperature there; $T_0=2.725\,\mathrm{K}$ — present CMB monopole; $\Delta T/T$ — rms fluctuation with the dipole removed. Reading: $\sim 10^4$ independent patches agree to $10^{-5}$ — the fact demanding explanation.

Derivation AWhy the particle horizon is finite and small at recombination

The goal is to show that the comoving particle horizon integral $\chi_{\rm ph}=c\!\int_0^t \mathrm{d}t'/a$ converges at the Big Bang for radiation and matter, that it evaluates to a quantity of the order of the comoving Hubble radius, and that at recombination this comes out far smaller than the comoving distance to the last-scattering surface — which is precisely why opposite CMB points fall outside each other's horizons. We insert the single-fluid scale-factor solution, do the integral, and read off the size.

Step 1. Solve the Friedmann equation for a single fluid. Energy conservation gives $\rho\propto a^{-3(1+w)}$ , and $H^2\propto\rho$ then integrates to $a\propto t^{2/[3(1+w)]}$ , so $a\propto t^{1/2}$ for radiation ( $w=1/3$ ) and $a\propto t^{2/3}$ for matter ( $w=0$ ). Why: the horizon integral needs $a(t')$ explicitly, and these are the two eras that set it.

Step 2. Do the radiation integral. With $a=A\,t^{1/2}$ , $\chi_{\rm ph}=c\!\int_0^t \mathrm{d}t'/(A\,t'^{1/2}) = 2c\,t^{1/2}/A = 2ct/a(t)$ . Since $H=1/(2t)$ gives $aH=a/(2t)$ , this is exactly $\chi_{\rm ph}=c/(aH)$ . Why: the comoving horizon equals the comoving Hubble radius in radiation domination — so the Hubble radius is a faithful stand-in for causal contact. (In matter domination the same steps give $\chi_{\rm ph}=3ct/a=2c/(aH)$ ; both are order $1/(aH)$ .)

Step 3. Notice that convergence is the whole point. The integrand $1/a\propto t'^{-1/2}$ is integrable at $t'=0$ , so $\chi_{\rm ph}$ is finite — there was only a limited amount of conformal time before recombination, and hence only a limited causal reach. Why: a divergent integral would mean an infinite horizon and no problem at all; finiteness is what makes the CMB a mosaic of strangers.

Step 4. Compare with the distance to last scattering. The comoving radius of the last-scattering surface seen today, $\chi_{\rm rec\to 0}$ , is far larger than $\chi_{\rm ph}(t_{\rm rec})$ , so two points on opposite sides of that surface lie well outside each other's horizons: $\chi_{\rm ph}(t_{\rm rec})\ll \chi_{\rm rec\to 0}$ . Why: this inequality is the mathematical form of "they never communicated".

Step 5. Convert to an angle. Dividing the proper horizon $d_{\rm hor}=a_{\rm rec}\chi_{\rm ph}$ by the angular-diameter distance $d_A$ gives $\theta_{\rm hor}\sim 1.7^\circ$ — a patch a couple of degrees across on the present sky. Why: this degree-scale number is the quantitative statement of the horizon problem, and it is directly checkable against the CMB. *The results $\chi_{\rm ph}=c/(aH)$ (radiation) and $2c/(aH)$ (matter) follow analytically from the single-fluid scale-factor solutions, and the horizon angle $\theta_{\rm hor}\approx 1.7^\circ$ follows from $\theta\approx\sqrt{a_{\rm rec}}=(1+z_{\rm rec})^{-1/2}$ with $z_{\rm rec}=1090$ .*

Go deeper · Derivation B · 1 minThe comoving Hubble radius and the sign of w

Derivation BThe comoving Hubble radius and the sign of w

The goal is to derive the scaling $1/(aH)\propto a^{(1+3w)/2}$ from first principles and show that its sign — whether the comoving Hubble radius grows or shrinks — is fixed entirely by the equation of state, through the same combination $\rho+3p/c^2$ that appears in the acceleration equation. This one relation is the pivot on which the whole problem and its cure turn, so it is worth building from conservation and the Friedmann equation.

Step 1. Start from the flat Friedmann equation $H^2=(8\pi G/3)\rho$ and the continuity equation $\dot\rho+3H(\rho+p/c^2)=0$ , which for constant $w$ integrates to $\rho\propto a^{-3(1+w)}$ . Why: these two equations contain all the background dynamics; everything else is algebra.

Step 2. Substitute the density scaling into the Friedmann equation: $H^2\propto a^{-3(1+w)}$ , so $H\propto a^{-3(1+w)/2}$ . Why: this expresses the expansion rate purely in terms of the scale factor and the equation of state.

Step 3. Multiply by $a$ and invert. $aH\propto a^{1-3(1+w)/2}=a^{-(1+3w)/2}$ , hence $1/(aH)\propto a^{(1+3w)/2}$ . Why: the comoving Hubble radius is the physically meaningful causal scale, and now its entire behaviour is encoded in the exponent $(1+3w)/2$ .

Step 4. Read off the sign. For $w>-1/3$ the exponent is positive and $1/(aH)$ grows (radiation $\propto a$ , matter $\propto a^{1/2}$ ); for $w<-1/3$ it is negative and $1/(aH)$ shrinks; at $w=-1/3$ it is constant. This is precisely the boundary $\ddot a=0$ from the acceleration equation, since $\ddot a\propto -(\rho+3p/c^2)=-\rho(1+3w)$ . Why: the same critical value $w=-1/3$ governs both whether the universe accelerates and whether the comoving Hubble radius shrinks — they are one phenomenon. *The exponents evaluate to $(1+3w)/2 = 1,\,1/2,\,-1$ for $w=1/3,\,0,\,-1$ respectively, and the sign change at $w=-1/3$ coincides exactly with $\ddot a=0$ .*

Go deeper · Derivation C · 1 minHow many e-folds of inflation are needed

Derivation CHow many e-folds of inflation are needed

The goal is to turn "inflation stretches one causal patch to cover the sky" into a number — the minimum amount of exponential expansion, measured in e-folds $N_e=\ln(a_f/a_i)$ , needed so that the comoving Hubble radius at the start of inflation was at least as large as the comoving scale of today's observable universe. This is the condition an inflationary model must satisfy to actually solve the horizon problem, and it fixes the famous benchmark of about sixty e-folds.

Step 1. State the condition. For the observable universe to have originated inside one pre-inflationary Hubble patch, the comoving Hubble radius at the onset of inflation must exceed the present one: $(a_iH_i)^{-1}\gtrsim(a_0H_0)^{-1}$ . Why: any comoving scale we observe today must have been inside the horizon before inflation, i.e. below the comoving Hubble radius then.

Step 2. Use the inflationary scaling. With $w\approx-1$ , $H\approx$ constant and $1/(aH)\propto a^{-1}$ , so during inflation the comoving Hubble radius simply falls as $e^{-N_e}$ while $a$ grows as $e^{N_e}$ . Why: near-de Sitter expansion is what does the shrinking, converting a large starting radius into a small one exponentially fast.

Step 3. Estimate the required drop. The comoving Hubble radius would otherwise have grown, from the end of inflation to today, by roughly the factor $T_{\rm reh}/T_0$ (the expansion since reheating). Inflation must therefore supply at least that much shrinkage: $N_e\gtrsim\ln(T_{\rm reh}/T_0)$ . Why: this balances the post-inflationary growth of $1/(aH)$ against the inflationary shrinkage.

Step 4. Put in GUT-scale numbers. Taking $T_{\rm reh}\sim 10^{15}\,\mathrm{GeV}\sim 10^{28}\,\mathrm{K}$ and $T_0\simeq 2.7\,\mathrm{K}$ gives $N_e\gtrsim\ln(10^{28}/2.7)\approx 64$ , i.e. the canonical $N_e\gtrsim 60$ . Why: this is why "about sixty e-folds" is quoted everywhere — it is set by the enormous hierarchy between the inflationary energy scale and today's CMB temperature. *Explicitly, $\ln(1.16\times10^{28}/2.725)\approx 63.6$ , confirming the canonical benchmark $N_e\gtrsim 60$ .*

Limiting cases

Every case is governed by one exponent, $(1+3w)/2$ , in the comoving Hubble radius $1/(aH)\propto a^{(1+3w)/2}$ . The equation of state $w$ decides whether the causal scale grows (problem) or shrinks (cure).

Case	Result	Meaning
Radiation, $w=1/3$	$1/(aH)\propto a$	Comoving Hubble radius grows linearly; $\chi_{\rm ph}=c/(aH)$ finite and small — the regime that creates the problem in the early universe.
Matter, $w=0$	$1/(aH)\propto a^{1/2}$	Still growing; $\chi_{\rm ph}=2c/(aH)$ . The horizon keeps expanding, but most of the integral was already fixed during radiation domination.
Critical, $w=-1/3$	$1/(aH)=\text{const}$	The boundary: $\ddot a=0$ . For $w>-1/3$ the radius grows (problem); for $w<-1/3$ it shrinks (cure).
Inflation / $\Lambda$ , $w=-1$	$1/(aH)\propto a^{-1}$	Comoving Hubble radius shrinks; $\ddot a>0$ . This is the sign reversal that solves the horizon problem.

Common trapConfusing the particle horizon with the Hubble radius

These two lengths are numerically comparable in a single-fluid universe, so they are constantly conflated — but they are conceptually different, and the horizon problem lives in the gap between them. The particle horizon $\chi_{\rm ph}=\int_0^t c\,\mathrm{d}t'/a$ is a cumulative integral over all past time; it measures what could ever have reached us and it never decreases, even during inflation. The comoving Hubble radius $1/(aH)$ is an instantaneous quantity, the scale crossing the horizon right now, and it can shrink when $\ddot a>0$ . The causal question — were two points ever in contact? — is answered by the particle horizon. But it is the comoving Hubble radius whose growth is the clean diagnostic of the problem and whose reversal by inflation is the cure. Saying "inflation shrinks the horizon" is loose: inflation shrinks the comoving Hubble radius, letting scales exit the horizon, while the particle horizon integral keeps growing.

Worked examples

1 · How many causal patches tile the CMB sky (fully worked)

Estimate the angular size of the particle horizon at recombination and the number of causally disconnected patches on the CMB sky, using $z_{\rm rec}=1090$ and a flat, matter-dominated approximation for the conformal-time geometry.

Step 1. Relate the horizon angle to conformal time. In a flat universe the comoving horizon at recombination is the conformal time there, $\eta_{\rm rec}$ , and the comoving distance to the last-scattering surface is $\eta_0-\eta_{\rm rec}\approx\eta_0$ . In matter domination $\eta\propto a^{1/2}$ , so $\eta_{\rm rec}/\eta_0=(a_{\rm rec}/a_0)^{1/2}=(1+z_{\rm rec})^{-1/2}$ .

Step 2. Put in the redshift. $\theta_{\rm hor}\approx\eta_{\rm rec}/\eta_0=(1+z_{\rm rec})^{-1/2}=(1091)^{-1/2}=\mathbf{0.0303\ rad}=\mathbf{1.73^\circ}$ . The causal horizon at last scattering is under two degrees across on today's sky.

Step 3. Tile the sphere. A patch of angular radius $\theta_{\rm hor}$ subtends solid angle $\Omega_{\rm patch}\approx\pi\theta_{\rm hor}^2$ , so the number covering $4\pi$ steradians is $N=4\pi/\Omega_{\rm patch}=4/\theta_{\rm hor}^2$ .

Step 4. Evaluate. $N=4/(0.0303)^2\approx\mathbf{4.4\times10^{3}}$ . A more careful treatment — the horizon grows through the matter era, and one may count three-dimensional volumes — pushes this to the often-quoted $\sim10^{4}$ – $10^{5}$ . Either way: thousands of regions, all at one temperature to $10^{-5}$ , with no causal way to have agreed. That is the horizon problem in one number.

2 · The number of e-folds inflation must supply

Estimate the minimum number of e-folds of inflation needed to solve the horizon problem for a GUT-scale model with reheating temperature $T_{\rm reh}\sim 10^{15}\,\mathrm{GeV}$ .

Step 1. The condition is that inflation shrink the comoving Hubble radius by at least the factor by which it has grown since reheating, $T_{\rm reh}/T_0$ , so $N_e\gtrsim\ln(T_{\rm reh}/T_0)$ .

Step 2. Convert the reheating scale to a temperature: $T_{\rm reh}\sim 10^{15}\,\mathrm{GeV}$ . Using $1\,\mathrm{GeV}\leftrightarrow 1.16\times10^{13}\,\mathrm{K}$ , this is $T_{\rm reh}\sim 1.16\times10^{28}\,\mathrm{K}$ .

Step 3. Evaluate the logarithm with $T_0=2.725\,\mathrm{K}$ : $N_e\gtrsim\ln(1.16\times10^{28}/2.725)=\ln(4.3\times10^{27})=\mathbf{63.6}$ .

Step 4. Round to the benchmark. This is the origin of the ubiquitous $N_e\gtrsim 60$ : a lower inflationary scale needs somewhat fewer e-folds, a higher one more, but $\sim 60$ is canonical for GUT-scale inflation. The same $\sim 60$ e-folds also flatten the universe and dilute relic monopoles — one mechanism, three cures.

Physics of the portrait

The horizon problem is inseparable from the man who made it a household phrase and, in the same 1981 paper, proposed its cure. Alan Guth did not invent the observation that the CMB is too uniform — Rindler had formalized cosmological horizons in 1956, and Misner had spent the late 1960s trying to erase the puzzle dynamically — but Guth fused the horizon, flatness, and monopole puzzles into a single argument and showed that one idea, accelerated expansion, dissolves all three.

Origins · 1 minAlan Guth and how this was discovered

Alan Guth

b. 1947 · cosmic inflation and the horizon problem

Alan Harvey Guth, a particle theorist by training, was working on grand-unified-theory monopoles at Cornell and SLAC in 1979–1980 when he noticed that the same supercooled false-vacuum state that overproduced monopoles would also drive a phase of exponential expansion. In his January 1981 paper Inflationary universe: a possible solution to the horizon and flatness problems, he showed that a brief early epoch with $\ddot a>0$ stretches a single causally connected, thermalized patch to encompass the entire observable universe — turning the CMB's near-perfect uniformity from an unexplained initial condition into a dynamical consequence, while simultaneously flattening the spatial curvature and diluting unwanted relics. Guth's original "old inflation" ended too violently to be viable (the graceful-exit problem) and was superseded by the slow-roll "new inflation" of Linde and of Albrecht and Steinhardt in 1982; but the framing was decisive. The horizon problem as we teach it — the small particle horizon at recombination, the $\sim10^4$ disconnected patches, the shrinking comoving Hubble radius — is essentially Guth's synthesis. COBE's 1992 measurement of $\Delta T/T\sim10^{-5}$ confirmed both the scandalous uniformity and the small fluctuations inflation predicts on top of it.

History

1956

— Rindler formalizes particle and event horizons in cosmology, isolating the causality puzzle

1969

— Misner's Mixmaster program tries to explain CMB uniformity by dynamical mixing, and largely fails

1981

— Guth proposes inflation, resolving the horizon, flatness, and monopole problems together

1982

— Linde and Albrecht–Steinhardt develop slow-roll new inflation, fixing the graceful-exit flaw

1992

— COBE measures ΔT/T ∼ 10⁻⁵, confirming both the uniformity and the predicted fluctuations

Going further

Texts. Misner, Thorne & Wheeler, Gravitation (1973), §27 · Wald, General Relativity (1984), Ch. 5 · Carroll, Spacetime and Geometry (2004), §8.4 · Hartle, Gravity (2003), Ch. 18 · Schutz, A First Course in General Relativity (2009), Ch. 12 · Dodelson, Modern Cosmology (2003), Ch. 6 · Mukhanov, Physical Foundations of Cosmology (2005), Ch. 1–2 · Guth, Phys. Rev. D 23, 347 (1981); The Inflationary Universe (1997) · Baumann, TASI Lectures on Inflation (2009). The key results are: $\chi_{\rm ph}=c/(aH)$ for radiation ( $a\propto t^{1/2}$ ) and $2c/(aH)$ for matter ( $a\propto t^{2/3}$ ) — note that $2c/(aH)$ is the correct matter result, not radiation; the horizon angle $\theta_{\rm hor}\approx(1+z_{\rm rec})^{-1/2}\approx1.7^\circ$ ; the patch count $N\sim4/\theta_{\rm hor}^2\sim10^{3}$ – $10^{5}$ ; and the e-fold requirement $N_e\gtrsim\ln(T_{\rm reh}/T_0)\approx60$ . Standard Planck-era values used throughout: $z_{\rm rec}\simeq1090$ , $T_{\rm rec}\simeq3000\,\mathrm{K}$ , $T_0=2.725\,\mathrm{K}$ , $\Delta T/T\sim10^{-5}$ , $n_s\approx0.965$ .

Builds on: The FLRW Metric · The Scale Factor and Hubble Parameter · Comoving and Proper Coordinates · The Friedmann Equations · The Radiation-Dominated Era. Leads to: The Cosmological Constant and Dark Energy · The Dark-Energy-Dominated Era · The Cosmological Principle.

Ask the tutor: Want me to walk through the conformal diagram, or show how the same $\sim60$ e-folds also solve the flatness problem?

Isn't the CMB uniformity just explained by the whole universe starting from one Big Bang?

No — and this is the subtle point. A common Big Bang origin does not imply causal contact. The Big Bang happened everywhere at once, not at a single point, and immediately afterward different regions were already separated by more than light could subsequently cross. The finiteness of the particle horizon $\chi_{\rm ph}=c\!\int_0^t\mathrm{d}t'/a$ is exactly the statement that "same origin in time" does not mean "able to have exchanged signals": two regions can both date from $t=0$ and still never have shared a past light cone. That is why a common Big Bang does not, by itself, explain the shared temperature.

Why does the particle horizon integral converge instead of blowing up at the Big Bang?

Because the integrand $1/a(t')$ grows only mildly as $t'\to0$ . In radiation domination $a\propto t'^{1/2}$ , so $1/a\propto t'^{-1/2}$ , and the integral of $t'^{-1/2}$ is finite at the origin (it gives $2t'^{1/2}$ ). Physically, the universe expanded fast enough at early times that a photon simply could not accumulate an infinite comoving distance. Had the early expansion instead been dominated by something with $w<-1/3$ — as during inflation — the integral would diverge and the horizon would be effectively unbounded. So the convergence is not a technicality; it is the mathematical signature of the problem, and inflation cures it precisely by adding an early phase that makes the causal reach enormous.

What is the difference between the particle horizon and the Hubble radius, in one sentence each?

The particle horizon is the total comoving distance light has travelled since the Big Bang, an accumulated integral $\int_0^t c\,\mathrm{d}t'/a$ that never decreases and answers "what could ever have reached me?". The Hubble radius $c/H$ (comoving: $1/(aH)$ ) is an instantaneous length, the scale crossing the horizon right now, and it can shrink when the expansion accelerates. In a single-fluid decelerating universe they are the same order of magnitude, which is why they are so often used interchangeably; but only the comoving Hubble radius shrinks during inflation, and keeping the two ideas distinct is the difference between understanding the cure and merely reciting it.

How can inflation connect regions if the particle horizon never actually shrinks?

The trick is that inflation makes the particle horizon enormous, not small. By inserting an early epoch with $w<-1/3$ , the integral $\int\mathrm{d}t'/a$ picks up a huge contribution while $a$ was tiny, so the true particle horizon at recombination is vastly larger than the naive hot-Big-Bang estimate — large enough to encompass the whole last-scattering surface. Equivalently, every scale we observe was inside the comoving Hubble radius before inflation, then exited as that radius shrank, then re-entered later. Both say the same thing: the observable universe descends from a single patch that was in thermal equilibrium before inflation stretched it beyond the horizon.

Does the horizon problem prove that inflation happened?

No. It proves only that the hot Big Bang, with ordinary matter and radiation, cannot explain the CMB's uniformity without extreme fine-tuning of initial data — a strong motivation, not a proof. Inflation is the leading solution because it also explains the flatness problem, dilutes monopoles, and predicts a nearly scale-invariant spectrum matching the CMB in detail ( $n_s\approx0.965$ ). But alternatives exist — bouncing, ekpyrotic, and varying-speed-of-light models — each engineered to make the CMB causally connected. What the horizon problem does is set the bar: any serious theory of the early universe must say how the sky came to be uniform, and inflation clears that bar most economically.

Why exactly sixty e-folds, and not thirty or a hundred?

The number is set by a ratio of energy scales. Inflation must shrink the comoving Hubble radius by at least as much as it has grown since inflation ended, and that post-inflationary growth is roughly the factor by which the universe has cooled, $T_{\rm reh}/T_0$ . For GUT-scale inflation, $T_{\rm reh}\sim10^{15}\,\mathrm{GeV}\sim10^{28}\,\mathrm{K}$ against today's $2.7\,\mathrm{K}$ , giving $N_e\gtrsim\ln(10^{28}/2.7)\approx64$ . Lower the inflationary scale and you need fewer e-folds (down to $\sim40$ ); raise it and you need more. So "sixty" is not a fundamental constant but a logarithm of the hierarchy between the inflationary scale and today's CMB temperature — which is why it is quoted as $\gtrsim60$ .