physics2u
PhysicsMathsChemistryBiology
Tier
⌕ Search ⌘K
Derivation

Particle Horizon, Horizon and Flatness Problems

Statement

For a spatially flat, single-fluid FLRW universe with constant equation of state \(w\), the comoving particle horizon is \(\chi_p(a)=\dfrac{2c}{H_0(1+3w)}\,a^{(1+3w)/2}\), finite only when \(w>-\tfrac13\); simultaneously the curvature term obeys \(\Omega-1=\dfrac{kc^2}{a^2H^2}\), whose magnitude grows as \(|\Omega-1|\propto a^{1+3w}\). In a decelerating (\(1+3w>0\)) hot big-bang, the finite horizon means causally disconnected patches share the same CMB temperature (the horizon problem) and \(\Omega=1\) is a repulsive fixed point requiring extreme early-time fine-tuning (the flatness problem).

Why it matters

The identical exponent \((1+3w)\) governing both the comoving Hubble radius and the deviation \(|\Omega-1|\) shows that the two "problems" of standard cosmology are one problem wearing two costumes: for ordinary matter and radiation the comoving Hubble radius grows, so scales enter the horizon over time and any initial flatness is amplified away.

Reversing the sign of \(1+3w\) — a fluid with \(w<-\tfrac13\) that violates the strong energy condition — makes the comoving Hubble radius shrink. That single algebraic observation is the seed of cosmic inflation, which is why this derivation is the pivot between hot-big-bang kinematics and the inflationary paradigm.

Assumptions
Homogeneous, isotropic FLRW metric.Without the Robertson–Walker form there is no single scale factor \(a(t)\); the notion of one particle horizon and one \(\Omega\) dissolves into direction-dependent quantities.
A single fluid with constant equation of state \(w=p/(\rho c^2)\).If \(w\) varies (radiation→matter→dark-energy transitions), the clean power law \(a^{(1+3w)/2}\) is replaced by piecewise integrals; the qualitative conclusions survive but the exponents must be tracked era by era.
General relativity with the Friedmann equations as the dynamical law.Drop GR and the relation \(\Omega-1=kc^2/(a^2H^2)\) — which is just the Friedmann constraint rewritten — no longer holds, so the flatness argument evaporates.
The universe extends to \(a\to 0\) with the same fluid (no earlier phase).If a pre-hot-big-bang epoch (e.g. inflation) precedes \(a=0\)-extrapolation, the lower limit of the horizon integral is set by that epoch, not by a true singularity, and the horizon can be arbitrarily large — precisely the escape route.
Standard energy conditions: \(w>-\tfrac13\) so that \(\ddot a<0\).If violated the horizon integral's convergence and the sign of \(d|\Omega-1|/dt\) both flip, converting the "problem" into the "solution."
Derivation
1
\[ \chi_p(t)=\int_0^{t}\frac{c\,dt'}{a(t')} \]
Comoving particle horizon: the maximal comoving distance a light signal (null geodesic \(ds^2=0\Rightarrow c\,dt=a\,d\chi\)) can have travelled since \(t=0\). A
2
\[ \chi_p(a)=\int_0^{a}\frac{c\,da'}{a'\,\dot a'}=\int_0^{a}\frac{c\,da'}{a'^2 H(a')} \]
Change variable \(dt'=da'/\dot a'\) and use \(H\equiv\dot a/a\Rightarrow\dot a'=a'H\). Purely kinematic. A
3
\[ H(a)=H_0\,a^{-\tfrac{3}{2}(1+w)} \]
Flat single fluid: continuity \(\dot\rho+3H(\rho+p/c^2)=0\) with \(p=w\rho c^2\) gives \(\rho\propto a^{-3(1+w)}\); Friedmann \(H^2=\tfrac{8\pi G}{3}\rho\) then fixes the power. Normalised to \(H_0\) at \(a=1\). B
4
\[ \chi_p(a)=\frac{c}{H_0}\int_0^{a}a'^{\,-2}\,a'^{\tfrac{3}{2}(1+w)}\,da'=\frac{c}{H_0}\int_0^{a}a'^{\tfrac{3w-1}{2}}\,da' \]
Substitute Step 3 into Step 2 and collect the exponent: \(-2+\tfrac32(1+w)=\tfrac{3w-1}{2}\). Symbols only. B
5
\[ \int_0^{a}a'^{\tfrac{3w-1}{2}}\,da'\ \text{converges}\iff \frac{3w-1}{2}>-1\iff w>-\tfrac13 \]
The lower endpoint \(a'\to0\) is integrable only if the power exceeds \(-1\). This is the crux: for radiation (\(w=\tfrac13\)) and matter (\(w=0\)) the horizon is finite. C
6
\[ \boxed{\ \chi_p(a)=\frac{2c}{H_0\,(1+3w)}\,a^{\tfrac{1+3w}{2}}\ }\qquad(w>-\tfrac13) \]
Evaluate the power-law integral: \(\int_0^a a'^{(3w-1)/2}da'=\dfrac{a^{(3w+1)/2}}{(3w+1)/2}\). Physical horizon is \(d_p=a\chi_p\). B
7
\[ R_H(a)\equiv\frac{c}{aH}=\frac{c}{H_0}\,a^{\tfrac{1+3w}{2}} \]
Comoving Hubble radius from Step 3. It carries the same exponent as \(\chi_p\): horizon growth and Hubble-radius growth are the same phenomenon for constant \(w\). B
8
\[ H^2=\frac{8\pi G}{3}\rho-\frac{kc^2}{a^2}\ \Longrightarrow\ 1-\frac{\rho}{\rho_c}=-\frac{kc^2}{a^2H^2},\quad \rho_c\equiv\frac{3H^2}{8\pi G} \]
Divide the full (non-flat) Friedmann equation by \(H^2\) and insert the critical density. Pure algebra. A
9
\[ \boxed{\ \Omega-1=\frac{kc^2}{a^2H^2}\ },\qquad \Omega\equiv\frac{\rho}{\rho_c} \]
Rearrange Step 8. \(k=+1,0,-1\) fixes the sign; \(\Omega=1\) exactly iff \(k=0\). A
10
\[ |\Omega-1|=\frac{|k|c^2}{a^2H^2}=\frac{|k|}{c}\,R_H^2\ \propto\ a^{\,1+3w} \]
Use Step 7: \(a^2H^2=(c/R_H)^2\cdot\dots\); equivalently \(a^2H^2=H_0^2\,a^{-(1+3w)}\). For any decelerating fluid \(1+3w>0\), so \(|\Omega-1|\) grows monotonically — \(\Omega=1\) is a repeller. C
Result
\[ \chi_p(a)=\frac{2c}{H_0(1+3w)}\,a^{\tfrac{1+3w}{2}}\quad(w>-\tfrac13),\qquad |\Omega-1|=\frac{|k|c^2}{a^2H^2}\propto a^{\,1+3w} \]

Reading. In a hot big bang the age of the universe is finite and light has only had time to cross a finite comoving distance, so the sky is tiled by many patches that have never exchanged a signal — yet the CMB shows them at one temperature to \(1\) part in \(10^5\) (horizon problem). The same exponent makes \(|\Omega-1|\) grow with \(a\): to land near \(\Omega_0\approx1\) today the universe must have begun with \(|\Omega-1|\) tuned to \(\sim10^{-60}\) at the Planck epoch (flatness problem).

Units check. \(c/H_0\) has units \((\text{m s}^{-1})/(\text{s}^{-1})=\text{m}\); with \(a\) dimensionless, \(\chi_p\) is a length. For \(\Omega-1\): with \(a\) dimensionless and \(k\) carrying \(\text{m}^{-2}\), \(kc^2/(a^2H^2)=(\text{m}^{-2})(\text{m}^2\text{s}^{-2})/(\text{s}^{-2})\) is dimensionless, as a density ratio must be.

Limiting cases
  • Radiation, \(w=\tfrac13\): \(\chi_p=\dfrac{c}{H_0}a^{2}\); physical horizon \(d_p=a\chi_p=c/H=2ct\).
  • Matter, \(w=0\): \(\chi_p=\dfrac{2c}{H_0}a^{1/2}\); physical horizon \(d_p=2c/H=3ct\).
  • de Sitter / inflation, \(w=-1\): exponent \(1+3w=-2<0\); the integral from \(a=0\) diverges — no particle horizon in the naive sense, and \(R_H\) and \(|\Omega-1|\) both shrink \(\propto a^{-2}\). This is the escape.
  • Curvature scale \(w=-\tfrac13\) (coasting): exponent vanishes; \(R_H\) and \(|\Omega-1|\) are constant — the marginal case separating growth from decay.
  • Present epoch: with dark energy (\(w\approx-1\)) now dominant, \(R_H\) has recently begun to shrink again, so \(\Omega\) is being driven back toward \(1\) today.
Breaks when
  • \(w\le-\tfrac13\) (accelerated expansion). The horizon integral no longer converges at \(a\to0\) and \(|\Omega-1|\) decreases rather than grows; the boxed \(\chi_p\) formula and the "fine-tuning" conclusion both fail — deliberately, since this is the inflationary regime.
  • Multi-component universe near an era transition. The single power law \(a^{(1+3w)/2}\) is invalid; one must integrate \(\chi_p=\int c\,da/(a^2H)\) with \(H^2=H_0^2\sum_i\Omega_{i,0}a^{-3(1+w_i)}\) numerically, and \(|\Omega-1|\) tracks the dominant component's exponent piecewise.
  • Near the Planck epoch \(a\to0\). Classical GR and the FLRW continuity relation are unreliable at \(t\lesssim t_{\rm Pl}\); the extrapolated \(10^{-60}\) tuning is a statement about the classical theory's boundary condition, not a measured quantity.
  • Anisotropic or inhomogeneous cosmologies (Bianchi, LTB). A single \(a(t)\) and a single \(\Omega\) do not exist; shear terms alter both the horizon structure and the curvature evolution.
Failure modes
  • Confusing the particle horizon with the Hubble radius. They coincide up to a factor for constant \(w\), but in a general (varying-\(w\)) universe \(\chi_p\) is a cumulative integral while \(R_H=c/(aH)\) is instantaneous; scales can be inside \(R_H\) yet outside \(\chi_p\) and vice versa.
  • Confusing the particle horizon with the (event) horizon or with the sound horizon. The CMB acoustic peak is set by the sound horizon \(\sim c_s/\sqrt3\times\) travel time; the horizon problem uses the causal (particle) horizon. Students conflate the \(\sim1^\circ\) sound scale with the \(\sim2^\circ\) causal scale.
  • Writing \(\Omega-1\propto a\) universally. That holds only in matter domination; in radiation domination \(|\Omega-1|\propto a^2\). Forgetting the era-dependence gives the wrong Planck-time estimate by many orders of magnitude.
  • Claiming inflation "creates" causal contact retroactively. Inflation does not make distant patches touch; it enlarges the pre-inflationary causally-connected region and stretches it beyond today's horizon, so patches that were in contact are now widely separated.
  • Treating \(\Omega=1\) fine-tuning as a probability statement without a measure. "Why start so close to 1?" presumes a natural measure on initial conditions; asserting the tuning is "unlikely" hides a measure-theoretic assumption.
Discussion

The unifying content of this derivation is a single exponent. The comoving Hubble radius \(R_H\propto a^{(1+3w)/2}\) is the fundamental object: the comoving particle horizon integrates it, and the curvature deviation \(|\Omega-1|\propto R_H^2\) squares it. Whenever gravity is attractive (\(1+3w>0\), i.e. \(\rho c^2+3p>0\), the strong energy condition), \(R_H\) grows, so comoving scales continually enter the horizon and any initial spatial flatness is progressively spoiled. The horizon and flatness problems are therefore not two coincidences but two faces of the fact that decelerating expansion has a growing comoving horizon.

The resolution inverts the sign. A transient epoch with \(w<-\tfrac13\) makes \(R_H\) shrink in comoving units, so a small causally-connected patch inflates to encompass the entire observable universe (solving the horizon problem) while \(|\Omega-1|\) is driven exponentially toward zero (solving the flatness problem). The required duration is set by demanding that today's comoving Hubble radius fit inside the pre-inflationary one, which for GUT-scale inflation is about \(N\simeq60\) e-folds: \(a_{\rm end}/a_{\rm start}=e^{60}\).

Historically these puzzles (Dicke and Peebles; Guth 1981) were not inconsistencies — a hot big bang with finely tuned initial data reproduces all observations — but statements about naturalness. Their force is that a \(10^{-60}\) coincidence and \(\sim10^4\) independent patches at one temperature demand explanation, and inflation supplies a dynamical mechanism that renders both generic rather than tuned.

A subtler modern view frames flatness through the phase-space of solutions. In the standard FLRW system \(\Omega=1\) (the \(k=0\) trajectory) is a fixed point that is a repeller for \(1+3w>0\) and an attractor for \(1+3w<0\); inflation works precisely by turning the repeller into an attractor. There remains genuine debate about whether "fine-tuning" is well-posed without a measure on the space of initial data (the Gibbons–Hawking–Stewart measure and its divergences are the technical heart of this), so the flatness problem is best stated as: within the classical GR phase flow, generic decelerating histories do not land near \(\Omega=1\) today, whereas an inflationary attractor makes them do so.

Common misconceptions. (i) "The universe is bigger than the horizon, so information travelled faster than light" — no; the horizon is finite because time is finite, not because expansion is superluminal in any local frame. (ii) "Inflation solves flatness by adding energy to flatten space" — rather, rapid expansion dilutes the curvature term \(kc^2/a^2\) relative to \(H^2\). (iii) "A finite horizon means a finite universe" — the particle horizon is an observational boundary, not a spatial edge; space can be infinite.

Worked examples
1
Angular size of the causal horizon at recombination and the number of disconnected patches on the CMB sky.
Matter-dominated approximation, \(z_{\rm rec}\approx1100\Rightarrow a_{\rm rec}=1/1101\). B
2
\[ \frac{\chi_p(a_{\rm rec})}{\chi_p(a_0)}=\left(\frac{a_{\rm rec}}{a_0}\right)^{1/2}=\left(\frac{1}{1101}\right)^{1/2}\approx0.0301 \]
Ratio of the boxed matter-era horizon \(\chi_p\propto a^{1/2}\); the comoving distance to last scattering is \(\approx\chi_p(a_0)\). B
3
\[ \theta\approx\frac{\chi_p(a_{\rm rec})}{\chi_p(a_0)}\approx0.030\ \text{rad}\approx1.7^\circ \]
Small-angle: the causal horizon at recombination, seen from here, subtends this angle. A
4
\[ N\approx\frac{4\pi}{\pi\theta^2}=\frac{4}{\theta^2}=\frac{4}{(0.030)^2}\approx4.4\times10^{3} \]
Sky solid angle \(4\pi\) divided by a patch of angular radius \(\theta\). B
\[ \theta\approx1.7^\circ,\qquad N\sim10^{4}\ \text{causally disconnected patches} \]

Reading. Of order ten thousand independent regions all sit at \(T=2.725\,\text{K}\) to \(1\) part in \(10^5\), with no time to have equilibrated — the horizon problem quantified.

Units check. \(\theta\) dimensionless (radians); \(N\) dimensionless. Consistent.

1
Required flatness at the Planck epoch given today's bound \(|\Omega_0-1|<0.01\).
Scale back with \(|\Omega-1|\propto a^{1+3w}\), radiation-dominated (\(w=\tfrac13\)) to good approximation over most of cosmic history: \(|\Omega-1|\propto a^{2}\). C
2
\[ \frac{a_{\rm Pl}}{a_0}=\frac{T_0}{T_{\rm Pl}}=\frac{2.35\times10^{-4}\,\text{eV}}{1.22\times10^{28}\,\text{eV}}\approx1.9\times10^{-32} \]
Temperature scales as \(a^{-1}\); \(T_0=2.725\,\text{K}=2.35\times10^{-4}\,\text{eV}\), \(T_{\rm Pl}=1.22\times10^{19}\,\text{GeV}\). B
3
\[ |\Omega_{\rm Pl}-1|\approx|\Omega_0-1|\left(\frac{a_{\rm Pl}}{a_0}\right)^{2}<10^{-2}\times(1.9\times10^{-32})^2 \]
Apply the radiation-era scaling \(a^2\) across the whole span (order-of-magnitude; the matter/DE eras change the coefficient, not the exponent's dominance). C
4
\[ |\Omega_{\rm Pl}-1|\lesssim10^{-2}\times3.6\times10^{-64}\approx4\times10^{-66} \]
Arithmetic. Even allowing a few orders of magnitude for era mixing, the result sits near \(10^{-60}\). A
\[ |\Omega_{\rm Pl}-1|\lesssim10^{-60} \]

Reading. The density at the Planck time had to equal the critical density to sixty decimal places for the universe to look as flat as it does now — the flatness fine-tuning made explicit.

Units check. Every factor is a dimensionless ratio (temperatures cancel, \(\Omega\) dimensionless); the result is dimensionless.

Problems
  1. Show that for a flat matter-dominated universe the physical particle horizon equals \(d_p=3ct\), and evaluate it numerically at \(t=t_0=13.8\,\text{Gyr}\) treating the universe as matter-only.
    Solution From the boxed result with \(w=0\), \(\chi_p=\tfrac{2c}{H_0}a^{1/2}\), so \(d_p=a\chi_p=\tfrac{2c}{H_0}a^{3/2}=\tfrac{2c}{H}\) using \(H=H_0a^{-3/2}\). In matter domination \(a\propto t^{2/3}\Rightarrow H=\tfrac{2}{3t}\), hence \(d_p=2c/H=3ct\). Numerically \(d_p=3(3.0\times10^8\,\text{m/s})(13.8\times10^9\times3.156\times10^7\,\text{s})=3.9\times10^{26}\,\text{m}\approx13\,\text{Gly}\). (The true \(\Lambda\)CDM value \(\approx46\,\text{Gly}\) is larger because dark energy and the radiation era modify the integral.)
  2. Derive the convergence condition \(w>-\tfrac13\) for the particle horizon directly from the integral \(\int_0^a a'^{(3w-1)/2}da'\), and state what happens exactly at \(w=-\tfrac13\).
    Solution The integrand is \(a'^{p}\) with \(p=(3w-1)/2\). \(\int_0^a a'^p\,da'\) converges at the lower limit iff \(p>-1\), i.e. \((3w-1)/2>-1\Rightarrow3w-1>-2\Rightarrow w>-\tfrac13\). At \(w=-\tfrac13\), \(p=-1\) and \(\int_0^a a'^{-1}da'=\ln a'\big|_0^a\) diverges logarithmically — the marginal case with no finite particle horizon. For \(w<-\tfrac13\) the integral diverges as a power, again giving no (finite) particle horizon.
  3. Using \(|\Omega-1|\propto a^{1+3w}\), compute the ratio of \(|\Omega-1|\) at matter–radiation equality (\(z_{\rm eq}\approx3400\)) to its value today, assuming matter domination throughout that interval.
    Solution Matter domination: \(w=0\Rightarrow|\Omega-1|\propto a\). Thus \(\dfrac{|\Omega_{\rm eq}-1|}{|\Omega_0-1|}=\dfrac{a_{\rm eq}}{a_0}=\dfrac{1}{1+z_{\rm eq}}=\dfrac{1}{3401}\approx2.9\times10^{-4}\). So at equality the universe was flatter by a factor \(\sim3400\) than today — deviations grow as the universe evolves toward the present, consistent with \(\Omega=1\) being a repeller.
  4. Estimate the number of e-folds of inflation needed to solve the horizon problem if the comoving Hubble radius at the start of inflation must be at least as large as today's, \(R_H(t_i)\ge R_H(t_0)\). Take \(H_i\) constant during inflation, \(a_i=a_{\rm end}e^{-N}\), and reheating directly into radiation domination from GUT scale \(T_{\rm reh}\sim10^{15}\,\text{GeV}\) to today.
    Solution Comoving Hubble radius \(R_H=c/(aH)\). During inflation \(H\approx H_i\) const, so \(R_H(t_i)/R_H(t_{\rm end})=a_{\rm end}/a_i=e^{N}\). After inflation (radiation) \(R_H\propto a\), growing by the factor \(a_0/a_{\rm end}=T_{\rm reh}/T_0\sim(10^{15}\,\text{GeV})/(2.35\times10^{-13}\,\text{GeV})\approx4\times10^{27}\). Requiring \(R_H(t_i)\ge R_H(t_0)\): the pre-inflation shrink \(e^{N}\) must beat the post-inflation growth, \(e^{N}\gtrsim T_{\rm reh}/T_0\approx4\times10^{27}\Rightarrow N\gtrsim\ln(4\times10^{27})\approx64\). Hence the canonical \(N\gtrsim60\) e-folds.
  5. For a flat radiation-dominated universe show explicitly that \(|\Omega-1|\propto a^2\), starting from the Friedmann-derived scalings, and hence estimate \(|\Omega-1|\) at nucleosynthesis (\(T_{\rm BBN}\approx1\,\text{MeV}\)) given \(|\Omega_0-1|<0.01\).
    Solution Radiation: \(\rho\propto a^{-4}\Rightarrow H\propto a^{-2}\), so \(a^2H^2\propto a^2\cdot a^{-4}=a^{-2}\), giving \(|\Omega-1|=kc^2/(a^2H^2)\propto a^{2}\) (equivalently \(1+3w=2\) with \(w=\tfrac13\)). Scale factor ratio \(a_{\rm BBN}/a_0=T_0/T_{\rm BBN}=(2.35\times10^{-4}\,\text{eV})/(10^{6}\,\text{eV})=2.35\times10^{-10}\). Then \(|\Omega_{\rm BBN}-1|\approx|\Omega_0-1|(a_{\rm BBN}/a_0)^2<10^{-2}\times(2.35\times10^{-10})^2\approx5\times10^{-22}\). Allowing for the matter/dark-energy eras (which contribute additional factors) the standard textbook figure is \(\sim10^{-16}\)–\(10^{-18}\); either way the flatness at BBN had to be extraordinary.