Planck reanalysis finds a modest hint of inflation-era particle signatures

Searching for particles through cosmic patterns

The “cosmological collider” is not a machine. It is a research program that asks whether particles present during inflation could be identified through patterns in primordial fluctuations. Those fluctuations left imprints in the cosmic microwave background, or CMB, which the Planck satellite measured.

The target is primordial non-Gaussianity: departures from Gaussian statistics in the early fluctuations. This study examines the bispectrum, which describes correlations among three Fourier modes. Different arrangements of those modes form different triangle shapes, and different inflationary processes predict different patterns across those triangles.

In cosmological collider models, massive particles can produce oscillations when one mode has a much longer wavelength than the other two—the squeezed limit. Particle spin can also introduce characteristic angular dependence. These patterns could, in principle, help distinguish particle properties rather than simply indicate that primordial fluctuations were non-Gaussian.

The work, presented in an arXiv preprint revised on November 5, 2026, reports a search strategy and tentative statistical hints, not an established particle discovery.

Separating overlapping signatures

A central difficulty is that a possible massive-particle contribution can resemble signals generated by interactions of the inflaton, the field driving inflation. A fit can therefore constrain a mixture of contributions rather than isolate the distinctive particle signature.

The researchers construct four simplified collider-template families: two for spin-zero particles with different interactions, one for spin-one particles and one for spin-two particles. Their theoretical setup assumes scale invariance and weak coupling, and focuses on lower spins and lower-derivative interactions.

They then “orthogonalize” each template. This means subtracting components that overlap with the standard equilateral and orthogonal templates describing the single-field contributions considered in their model. The remaining pattern is designed to isolate information that those two standard shapes cannot reproduce.

This is not the removal of measured particles or foregrounds from a sky map. It is a change to the mathematical patterns being fitted. Because the overlapping standard shapes have low significance in Planck data, removing those components can increase the signal-to-noise ratio of the remaining collider pattern.

What the Planck search found

The team applies the Modal analysis pipeline to Planck’s official SMICA map combining CMB temperature and polarization. Modal represents complicated bispectra using sums of separable functions, making a broad parameter search computationally manageable.

For each template family, the search varies a dimensionless mass parameter from 1 to 6 and a sound-speed ratio from 0.01 to 100. The latter compares the propagation speeds associated with the inflaton and the additional particle field; it is not an absolute speed measurement.

The strongest result comes from the orthogonalized Scalar-II template. Its best fit has mass parameter 1.85 and sound-speed ratio 0.09. Its raw significance is 3.29 sigma, falling to 2.35 sigma after the authors’ look-elsewhere adjustment—approximately the 2.4 sigma highlighted in the abstract.

That adjustment matters because scanning many parameter combinations gives random fluctuations more opportunities to produce an apparently interesting match. The corrected result remains insufficient to establish primordial non-Gaussianity or a new particle.

The best orthogonalized spin-one and spin-two results reach adjusted significances of 2.14 and 1.98 sigma, respectively. The best-fit significance increases after orthogonalization in all four template families, although the size of the improvement varies.

Why the interpretation remains limited

The templates are simplified representations, built by taking the squeezed-limit particle signature and extending it to other triangle configurations. They do not exhaust every possible inflationary interaction or particle model.

The separation is also approximate in observational terms. Templates made orthogonal in primordial momentum space can retain correlations of order 10 percent when evaluated through the CMB estimator. The authors argue that this difference does not materially affect the reported statistical significances.

Numerical accuracy restricts the search as well. Larger mass parameters produce increasingly oscillatory patterns that the Modal expansion represents less accurately. The researchers report agreement with the independent CMB-BEST pipeline where the expansions remain accurate, providing a methodological cross-check rather than confirmation from a separate sky dataset.

The immediate contribution is therefore a way to organize and compare searches for distinctive primordial patterns. Better CMB measurements and large-scale-structure observations could test these templates further. Whether the modest Planck preferences reflect inflationary physics or statistical fluctuations remains unresolved.