The universe rang piercingly in its first moments of life. During inflation, when space expanded faster than ever after, energies reached levels no earthly accelerator can match. If exotic particles existed then, they would have set up oscillations in the primordial plasma—much like a stone thrown into water creates ripples. These ‘ripples’ got imprinted in the cosmic microwave background and the distribution of matter, becoming a unique key to ultra-high-energy physics. Alan Guth’s idea of inflation evolved into a tight theory, and the collider concept put forward by Juan Maldacena and collaborators showed that collisions of the inflaton field with other fields leave characteristic oscillatory traces in three-point correlations of density. To detect such oscillations in large-scale structure is to probe a ‘collider’ that operated at temperatures of $10^{29}$ K.
The main difficulty lay in recreating these signals in computer simulations. Standard methods expand the bispectrum in a set of preselected functions, but for models with rapidly varying squeezed behavior they break down. The team devised an elegant workaround: they divided the three-dimensional wavenumber space into cells and stitched the values inside each cell with cubic polynomials. To avoid infrared divergences, they used the reduced bispectrum—the ratio of the three-point correlation to a combination of power spectra. This yielded better than percent-level accuracy in reproducing the kernel of the model bispectrum. Then, on a supercomputer, they ran an ensemble of ten cubes of side 3 Gpc/h, filled with $1024^3$ dark matter particles. The evolution was followed with the Pkdgrav3 code down to the present epoch (z=0). Halos were identified with the Rockstar algorithm; the bias was computed as the ratio of the halo–matter cross-spectrum, suppressing cosmic variance. The whole work rested on standard cosmology, with Hubble's law and the Friedmann equations governing the expansion dynamics.
For the first time in simulation history, pronounced non-monotonic oscillations of the bias appeared. For the baseline model with frequency ω=1.5, the wave is clearly visible in the scale range k≈$10^{-2}$ h/Mpc. The positions of peaks and troughs depend dramatically on mass: increasing the halo mass tenfold—from $10^{14}$ to $10^{15}$ M☉/h—shifts the characteristic scale $k_c$ from 0.015 to 0.0068 h/Mpc, almost precisely following the change in the Lagrangian radius R∝$M^{1/3}$. The amplitude also grows with mass. The peak-background split theory, incorporated into a two-parameter model (b_φ and b_σ), describes all measurements beautifully. For samples where halos were selected only by mass, b_φ≈$δ_c$($b_{1-1}$) and b_σ=1, as predicted by the universality of the mass function. But when halos were split by concentration (using the $V_{max}$ parameter), the phase of oscillations shifted: denser halos yield oscillations on noticeably smaller scales—an effect reaching a factor of two for objects of $10^{14}$ M☉/h. The reason is that selection by secondary properties breaks universality, and b_φ, b_σ deviate from their standard values. Interestingly, as the frequency ω increases, the signal in the bias is rapidly suppressed: while in the primordial bispectrum the oscillations have constant amplitude, in the halo bias they get averaged within the filter window and at ω≈8 drop to several percent. This means the observationally accessible frequency range is limited.
The discovery of oscillating bias offers a powerful new test of inflationary models. Unlike the classic local non-Gaussianity (local fNL), whose signal on large scales is monotonic and can be mimicked by instrumental systematics, here we have a non-monotonic, periodic structure with predictable frequency and phase. Moreover, the phase depends on mass and assembly, and systematics would be hard-pressed to fake a consistent behavior across multiple scales and different sample types. Thus, collider signals may provide an even cleaner window into primordial physics than the benchmark local type. A practical takeaway: the calibrations of the b_φ parameter previously obtained for local PNG work just as well for collider models, meaning existing data analysis pipelines (DESI, Rubin, Euclid) can easily be extended.
The next natural step is to develop methods to extract the oscillations from real catalogs of galaxies and clusters. This will require carefully modeling systematic effects: photometric bias, critical differences by redshift, correlations with dust maps. A separate challenge is to understand whether the signal shape can simultaneously constrain the amplitude fNL and the selection parameters b_φ, b_σ, or whether they must be fixed from additional statistics. Theorists also plan to study how the collider signal will manifest in millimeter-wave surveys of galaxy clusters via the Sunyaev–Zel'dovich effect.
Cosmology (early Universe, origin of structure), astrophysics (formation and evolution of halos), particle physics (properties of ultraheavy particles).
Building predictions for DESI, Rubin, and Euclid surveys; testing self-calibration of the bias parameters; searching for the collider signal in SPT and Planck data.
Addresses problems: the nature of the inflaton field, the energy scale of Grand Unification, the existence of supersymmetric particles and axions, and dark energy—via the connection to the expansion.
This section is addressed to the author of the paper and their colleagues. Neighbouring works were found by meaning-based search across our archive, not by word overlap. We do not review: we show the field around the work.
What the work does
This paper introduces a new method for modeling the oscillating halo bias, reproducing the non-monotonic oscillations in the bispectrum predicted by cosmological collider models. Using an ensemble of N-body simulations, the authors for the first time detected the dependence of these oscillations on halo mass and showed their agreement with the peak-background split analytical model.
Where its strength is
The strength of the work lies in demonstrating that signals from ultra-high-energy physics can be extracted from the large-scale structure through careful processing of numerical simulations. This paves the way for directly testing inflationary models against galaxy survey data.
Where it could go next
Worth trying
Why this may matter
If the oscillating bias can be robustly extracted from observational data, it would provide a direct window into physics at energies unattainable in the laboratory and would allow us to distinguish between competing inflationary models.
Nearby works in our archive
The archive is crowded around this work: 10 reviews nearby, closest at 0.70. This ground is well walked — and there is plenty to compare against.
as written by the authors arXiv:2607.24939v1
The initial conditions of our Universe contain a wealth of information about the particle physics of very high energies. One such class of signatures, called cosmological colliders, generates oscillations in the three-point correlations (or bispectra) of the primordial density field, and these imprint scale-dependent oscillations in the halo bias. We develop a new method for simulating cosmological collider models that foregoes traditional template-based basis-decomposition methods and can reproduce the scale-dependent signals of the input template at percent-level accuracy. Using this method, we produce simulations for one class of collider models and present the first measurements of oscillating halo bias in simulations. The amplitude and phase of the oscillations show a clear dependence on halo mass, with a factor of ten shift in halo mass causing a factor of two shift in the location of the oscillations. Increasing the frequency of the primordial bispectra model suppresses the signal in the halo bias, as the oscillations average down over the window function of the halo. The phase of the signal is also sensitive to assembly bias. In all cases, the scale-dependent halo bias can be accurately modeled using a simple peak background-split theory. The oscillations and their mass/selection-dependent phase offsets are a unique signature that is not easily mimicked by known observational systematics and is therefore a more robust target. Our simulations and underlying initial conditions code are both made publicly available.
The cosmological collider isn't science fiction: it's the Universe itself at $10^{-32}$ seconds old. The temperature then reached $10^{29}$ kelvin, and collisions were 10 trillion times more energetic than at the Large Hadron Collider.
In Carl Sagan's 'Contact,' an extraterrestrial intelligence encodes information in a pulsar — here the Universe itself has left a 'message' in the distribution of galaxies, which we are only beginning to decode.