Cosmic Inflation: Theory and Evidence

The horizon problem: how did distant regions share similar conditions?

Run an ordinary, decelerating hot Big Bang history backward, without an earlier inflationary stage. Widely separated regions that emitted the CMB then lack enough shared causal history for signals to establish their strikingly similar temperatures. This is the horizon problem. It concerns conditions at the time the radiation last scattered, even though light from both regions reaches us today.[1]

A sufficiently long inflationary stage allows the region we observe to originate within a much smaller region with an earlier opportunity for causal contact. Expansion can carry that shared history to enormous scales. Whether suitable initial conditions arise is a further question; stretching alone does not explain every property of the starting state.[2]

The flatness problem: why is spatial curvature so small?

“Flat” describes spatial geometry, not a universe shaped like a sheet. The curvature contribution relative to expansion obeys the following relation in a homogeneous cosmological model, where Ω_tot is total density divided by critical density.[3]

During radiation domination this departure grows approximately as a²; during matter domination, as a. Inflation reverses the trend because aH grows. For nearly constant H, the departure falls approximately as a⁻². Sufficient inflation therefore suppresses curvature within the observable region. It does not prove that all of space is exactly flat, infinite, or topologically simple.[3]

The scale factor, a(t), describes how distances between ideal observers following the cosmic expansion change. Inflation means ä > 0: the growth accelerates. The Hubble parameter, H = ȧ/a, need not increase; in many models it decreases slowly while a grows rapidly.[4]

A slowly evolving scalar field

A common model uses a hypothetical inflaton, a scalar field with a potential energy V(φ). For a homogeneous field with the standard, canonical kinetic term, its energy density and pressure are:[5]

When potential energy dominates, pressure is approximately the negative of energy density. This can sustain inflation. In slow-roll models the field changes gradually enough to maintain that state for many expansion times. “Rolling” describes evolution through possible field values; the field is not a small object sliding down a physical hill.[5]

Counting expansion in e-folds

One e-fold multiplies the scale factor by e, about 2.72. The count is N = ln(a_end/a_start); sixty e-folds increase lengths by roughly 10²⁶. Frequently quoted estimates of around 50–60 e-folds refer to the interval between observable scales leaving the Hubble radius and inflation ending. The value depends on the inflationary energy scale and the subsequent thermal history, especially reheating. It is not a measurement of inflation’s total duration.[6]

Reheating connects inflation to the hot universe

Inflation must end in our region. In many models the field begins oscillating, loses energy to other fields, and produces particles. Their interactions eventually establish a hot thermal plasma. This process is called reheating; an early, rapid stage of particle production is sometimes called preheating.[7]

The details depend on couplings and the inflation model. Reheating is therefore both a necessary connection to the later hot Big Bang and an important uncertainty when translating observations into early-universe predictions. It does not imply that physicists have measured the temperature or particle content of this transition.[7]

Expansion can also dilute pre-existing relics, including hypothetical magnetic monopoles. That argument applies to relics present before or sufficiently early during inflation; new production afterward requires its own accounting.[1]

A quantum field is not perfectly featureless even in its vacuum state. In the standard inflationary calculation, quantum fluctuations in the field and geometry evolve as their wavelengths are stretched. They generate primordial curvature perturbations: small differences that later influence the distribution and motion of matter.[8]

For the usual single-field, adiabatic slow-roll conditions, a suitable curvature perturbation becomes approximately constant on scales much larger than the Hubble radius. After inflation, its subsequent evolution helps produce the patterns measured in the CMB and in cosmic structure. Additional fields or departures from these conditions can change that behavior.[8]

Comoving distances factor out the overall expansion. The comoving Hubble radius, (aH)⁻¹ in units with c = 1, shrinks during inflation and grows during radiation and matter domination. It is a useful scale for following fluctuations. The particle horizon, which depends on the entire prior expansion history, is a different quantity.[9]

This is a proposed physical origin for the initial patterns. Measuring their later effects tests the prediction; it does not amount to directly watching quantum fluctuations during inflation.

Inflation earns its place in cosmology because many simple models produce a pattern of initial conditions that works well when evolved through the later universe. CMB data support predominantly adiabatic perturbations, a spectrum close to scale invariant, and only small departures from Gaussian statistics. Agreement constrains models without uniquely selecting an inflaton or potential.[10]

A joint analysis revised in June 2026 combines Planck, the Atacama Cosmology Telescope, the South Pole Telescope, and BICEP/Keck. It finds n_s = 0.9682 ± 0.0032 (68% interval), a slightly “red” spectrum with more relative power on larger scales. Additional datasets shift the estimate.[12]

Gaussianity is a statistical test

A Gaussian random field has a simple statistical structure: its average and correlations between pairs of points determine its higher-order correlations. Researchers look for additional correlations—primordial non-Gaussianity—that could distinguish different fields and interactions. Planck’s tested non-Gaussian templates are consistent with no detected primordial signal. This constrains particular possibilities; it does not eliminate every multi-field model or establish perfect Gaussianity.[13]

CMB polarization can be separated into E-mode and B-mode patterns. Primordial gravitational waves could create B-modes, but B-modes also arise when gravitational lensing remaps E-mode polarization. Polarized emission from Galactic dust and synchrotron radiation adds foreground signals. Observing B-mode power alone therefore does not identify inflationary gravitational waves.[14]

The tensor-to-scalar ratio, r, compares primordial tensor power with scalar power at a reference scale. The same analysis reports r < 0.034 at 95% confidence. This upper limit is consistent with the earlier BICEP/Keck result; it does not constitute a tensor detection.[12]

Under standard general-relativistic, vacuum-fluctuation assumptions, tensor amplitude is related to the expansion rate and energy scale during inflation. A credible primordial detection would therefore offer a powerful new measurement. The conversion depends on the model; a null result constrains models predicting detectable tensors without ruling out inflation as a whole.[10]

Even a primordial gravitational-wave signal would require further interpretation. Its spectrum and correlations would help distinguish its origin, since inflation is not the only proposed early-universe source.[16]

Choosing a field, its interactions, and its gravitational couplings produces a specific inflation model. Models can share important predictions while differing in their underlying physics. A match to one observable is therefore only part of the comparison.

Slow roll describes a regime of evolution, rather than a mutually exclusive family alongside these examples. Several constructions can use slow roll for part of their history.

Does inflation imply a multiverse?

Some models permit eternal inflation: inflation ends in certain regions while continuing elsewhere. Quantum fluctuations or slow decay of an inflating state can help sustain that behavior. Turning the resulting picture into probabilities for observations introduces difficult questions about how regions should be counted. Eternal inflation is not a separately established observation, and evidence for ordinary inflationary predictions would not by itself demonstrate a multiverse.[19]

Initial conditions and a physical mechanism

An inflationary model must explain when inflation starts and which initial conditions allow it to proceed. Numerical studies find that success depends on the potential and the initial field configuration; some regions can collapse while others inflate. There is no single result saying that every arbitrary initial state becomes the smooth universe we observe.[20]

Other questions concern the identity of the inflaton, the stability of its potential against quantum corrections, and its connection to known particles. Reproducing an expansion history is not yet a complete microscopic explanation of its cause.

Could another early history produce similar clues?

Bouncing and ekpyrotic cosmologies explore histories with a contracting phase before expansion. Some constructions generate nearly scale-invariant perturbations, making it misleading to say that only inflation can reproduce that broad feature. Their challenges include controlling instabilities, unwanted anisotropy, and the transition through a bounce. Each proposal must be assessed through its detailed predictions and physical consistency.[16]

The useful comparison is between explicit models and the full body of evidence. An alternative must explain the successes already achieved, while inflationary models must keep making predictions precise enough to test.

Measure polarization more precisely

The Simons Observatory is pursuing measurements of CMB polarization and lensing from Chile. Its combination of telescope scales and observing frequencies supports both the primordial B-mode search and the characterization of foregrounds and lensing. Greater sensitivity is most useful when those additional contributions are controlled.[21]

LiteBIRD is a planned space mission aimed at large-scale CMB polarization across many frequencies. JAXA reports that it entered Phase A in June 2026 and targets launch in Japanese fiscal year 2036. This is a development plan, with the observing program still ahead.[22]

Look beyond a single number

Improved limits on tensors, departures from Gaussian statistics, and the scale dependence of fluctuations can test different aspects of early physics. A stronger case emerges when a model correctly connects several observables, rather than fitting just one.

Inflation remains a productive framework because it links questions about the earliest conditions to observations we can refine. Its unfinished task is to turn that broad success into a more specific account of what happened.

Cosmology and the Universe’s Large-Scale Structure

1. Cosmic Inflation: Theory and Evidence · You are here
2. The Cosmic Web: Filaments, Voids, and Superclusters
3. The Cosmic Microwave Background’s Detailed Structure
4. Baryon Acoustic Oscillations
5. Redshift Surveys and Mapping the Universe
6. Gravitational Lensing: A Natural Cosmic Telescope
7. Measuring the Hubble Constant: The Tension
8. Dark Energy Surveys
9. Anisotropies and Inhomogeneities
10. Current Debates and Outstanding Questions