This paper looks for small, repeating wiggles in the oldest light we can observe — the cosmic microwave background (CMB). Such wiggles would come from short-lived or periodic events in the very early Universe and would appear as oscillations in the primordial power spectrum, a simple way to describe how early density variations depend on scale. The authors compare two versions of the Planck data release (PR3, the 2018 legacy release, and PR4, called NPIPE) and use an “unbinned” analysis that keeps the finest detail in the data so it does not wash out rapid oscillations.

The team tested several concrete templates for what primordial oscillations might look like. These include sharp or step-like features that produce linear oscillations, resonant modulations that give logarithmic oscillations, localized bumps or dips in the inflationary potential, and signals from so-called “primordial standard clocks” (oscillations driven by massive fields). To be sensitive to fast oscillations they compared results across the two Planck products and used likelihoods that do not average neighbouring multipoles. They find a handful of frequencies in the dimensionless range ω ≈ 10–100 that improve the fit to the CMB by up to Δχ2 ≈ −10 to −15 compared with the simple, featureless model.

Despite these fit improvements, the authors emphasize that the evidence is not strong. A Bayesian model comparison — which penalises models for adding extra parameters that reduce predictability — does not favour the extended feature models. Adding three or four extra parameters produces an “Occam penalty,” meaning the modest fit gains do not outweigh the loss of simplicity. When the authors also account for the “look‑elsewhere” effect (the reduced significance that comes from searching many possible frequencies), the global statistical significance of any preferred frequency falls to at most about 2.6σ. They also note that some anomalies reported in earlier studies shrink when using the updated CamSpec likelihoods together with the PR4 maps.