Carving out the Multifield Cosmological Collider Landscape

Technical Summary: Carving out the Multifield Cosmological Collider Landscape

Problem Statement
The paper addresses the challenge of connecting theoretical predictions of multifield inflation to cosmological observations, specifically within the framework of the "cosmological collider." Inflation embedded in ultraviolet (UV) completions (e.g., string theory) naturally involves a plethora of scalar fields (moduli, Kaluza–Klein towers, axion-like states). While the decay of these fields into inflaton fluctuations produces distinctive signatures in the soft limits of cosmological correlators, interpreting these signals is complicated by the vast parameter space of masses and mixings.

Standard lore suggests that heavy fields (principal series, $\Delta = 3/2 \pm i\mu$) produce suppressed oscillatory signals, while light fields (complementary series, $\Delta = 3/2 \pm \nu$) produce enhanced power-law signals. The local shape ($S \sim k_s/k_\ell$) corresponds to the unitarity boundary ($\Delta_- = 0$). A central open question is whether enhanced, almost-local signals are rare, fine-tuned occurrences or generic predictions when the number of fields ($N_f$) is large, as suggested by the string landscape ($N_f \sim \mathcal{O}(100)$).

Methodology
The author employs the Effective Field Theory (EFT) of inflationary fluctuations to model the interaction between the Goldstone boson of broken time translations ($\pi_c$) and $N_f$ additional massive scalar fields ($\sigma_I$).

1. Lagrangian Construction: They construct the most general quadratic Lagrangian including boost-breaking mixings. The fields are collected in $\chi_A = (\pi_c, \sigma_I)$. The Lagrangian includes kinetic mixing ($\rho_I \dot{\pi}_c \sigma_I$), antisymmetric mixing ($\Omega_{IJ} \dot{\sigma}_I \sigma_J$), and mass mixing ($M^2_{IJ} \sigma_I \sigma_J$).
2. Scaling Dimensions: By solving the coupled linear equations of motion with a power-law ansatz $\chi_A \sim (-\tau)^\Delta$, they derive a characteristic polynomial $\det M(\Delta) = 0$. The roots $\Delta$ determine the late-time scaling dimensions, which dictate the soft limit behavior of the bispectrum.
3. Statistical Approach: Instead of scanning specific points, the author adopts a random-matrix approach. They scan the parameter space (masses $m_I$, mixings $\rho_I$, and twists $\Omega_{IJ}$) using various statistical priors (uniform, normal distributions) for $N_f$ ranging from 2 to $\sim 100$.
4. Constraints: They enforce subluminality conditions ($N_{AB} \succeq 0$) and require stability (no super-horizon instabilities, i.e., $\text{Re}(\Delta) \geq 0$).

Key Contributions and Results

• Small $N_f$ Regime ($N_f = \mathcal{O}(1)$):
  For a small number of fields, the author confirms that enhanced power-law signals (where the real part of the scaling dimension $r \equiv \text{Re}(\Delta) \ll 1$) are highly fine-tuned.

  • For $N_f=2$, requiring $r \leq 1$ occupies $\sim 30\%$ of the parameter space, but requiring $r \ll 1$ (approaching the local shape) drops this to $\mathcal{O}(1\%)$.
  • Oscillatory signals ($\text{Im}(\Delta) \neq 0$) are even rarer ($< 0.1\%$).
  • The author identifies a parity effect: odd $N_f$ forces the antisymmetric mixing matrix $\Omega_{IJ}$ to be singular, creating a null direction that systematically enhances the probability of oscillatory signals compared to even $N_f$.

• Large $N_f$ Regime ($N_f = \mathcal{O}(100)$):
  The picture inverts dramatically as $N_f$ increases.

  • Generic Enhancement: For almost any distribution of bare masses and mixings (except for highly localized heavy mass spectra), the median of the dominant scaling dimension, $r_{\text{dom}} = \min_a \text{Re}(\Delta_a)$, decays as a clean power law with $N_f$.
  • Unitarity Boundary: Mixings generically drive the scaling dimensions toward the unitarity boundary ($\Delta_- = 0$), producing almost-local bispectrum shapes without fine-tuning of masses or couplings.
  • Mechanism: This behavior is a direct consequence of extreme value statistics and random-matrix universality. Since $r_{\text{dom}}$ is the minimum of $\mathcal{O}(N_f)$ roots, the Fisher–Tippett–Gnedenko theorem dictates that the minimum converges to zero as $N_f$ increases, provided the root density near $r=0$ is non-zero.
  • Oscillations: Unlike the real part, the imaginary part (oscillations) does not become generic; oscillatory signals remain rare even at large $N_f$.

• Bispectrum Shapes:
  Numerical computations of the full bispectrum shape confirm that for parameters yielding $r_{\text{dom}} \ll 1$, the squeezed limit is enhanced and approaches the local shape. The competing mode (at $\Delta \approx 3 - r_{\text{dom}}$) remains far away, ensuring the signal is not spoiled.

Significance and Claims
The paper claims to fundamentally alter the interpretation of cosmological collider signals in multifield scenarios.

1. Shift from Fine-Tuning to Generality: While enhanced signals were previously considered fine-tuned signatures for small numbers of fields, the author argues they are a generic prediction for inflation scenarios with a large number of fields ($N_f \sim 100$), as suggested by concrete UV embeddings.
2. Model Independence: The result relies on extreme value statistics and random-matrix universality rather than specific model details, reducing model-dependent bias in data interpretation.
3. Observational Implications: The author suggests that almost-local collider signals are a robust target for observational searches. Because these signals are free from equilateral contamination from inflaton self-interactions, they can be efficiently probed via scale-dependent bias, making constraints on $f_{\text{NL}}^{\text{loc}}$ a natural discriminator for new physics involving many species during inflation.

The author concludes that the standard distinction between heavy and light field signatures is radically altered by the number of active fields and their linear mixings, with mixings universally driving scaling dimensions toward the unitarity boundary in the large-$N_f$ limit.