Negative Neutrino Mass: Are DESI DR2 and the CMB Breaking the Standard Cosmological Model?

The DESI DR2 and CMB Neutrino Anomaly

1. Baseline Oscillation Constraints

   For decades, the standard cosmological model (ΛCDM) has relied on terrestrial particle physics experiments to set absolute physical boundaries on the neutrino sector. Solar, atmospheric, and reactor neutrino experiments, such as Super-Kamiokande and KamLAND, have conclusively demonstrated that neutrinos undergo flavor oscillations, necessitating non-zero rest masses. In the normal hierarchy, the sum of the three active neutrino masses, Σm_ν, is strictly bounded below by 0.06 eV, while the inverted hierarchy demands a minimum of 0.10 eV. These oscillation floors serve as hard priors in traditional cosmological parameter estimation, ensuring that the background Friedmann equations and the resulting thermal history reflect a physically viable universe.

2. The 2025–2026 DESI DR2 Dataset

   The landscape of precision cosmology shifted dramatically with the unprecedented volume and precision of the 2025–2026 DESI DR2 release. By mapping millions of galaxies and quasars, DESI DR2 provided an exquisitely precise measurement of the Baryon Acoustic Oscillation (BAO) scale across a vast range of redshifts. When these late-time geometric constraints are combined with the high-redshift primary anisotropies measured by the Planck satellite and the Atacama Cosmology Telescope (ACT), the combined likelihoods tightly constrain the matter density Ω_m and the expansion rate H₀. However, the high-fidelity redshift-space distortion data in DR2 also tightly restricts the amplitude of structure growth at late times, placing immense pressure on the standard model's parameter volume.

3. The Unphysical Signal at −0.16 eV

   When the hard physical prior of Σm_ν ≥ 0.06 eV is relaxed to allow for mathematical continuity in the likelihood surface, the joint DESI DR2 + CMB posterior distribution does not merely peak near zero; it descends deep into the negative regime. The data show a clear statistical preference for an effective negative neutrino mass sum of Σm_ν,eff ≈ −0.16 eV, bounded at approximately a 3σ confidence level below the normal hierarchy floor. This indicates that the observed universe exhibits less free-streaming suppression—or equivalently, more late-time structural clustering—than a universe with massless neutrinos would allow. The emergence of this signal has sparked intense debate over whether standard Boltzmann codes are accurately capturing the true underlying physics.

The τ–As–Σmν Degeneracy and Parameter Space

1. Primordial Amplitude and Optical Depth

   To understand the root of this negative mass preference, one must dissect the profound parameter degeneracies governing the cosmic microwave background. The primary CMB temperature and polarization power spectra are uniquely sensitive to the combination of the primordial scalar amplitude, A_s, and the reionization optical depth, τ. Because free electrons during reionization scatter CMB photons, they suppress the primary temperature anisotropies by a factor of e⁻²τ at scales smaller than the horizon at reionization.

   C_ℓ^TT = ∫ d(ln k) P_ℛ(k) |Δ_Tℓ(k)|² ∝ A_s e⁻²τ

   As a result, a lower measured value of τ directly forces a corresponding reduction in A_s to maintain the observed height of the acoustic peaks. This strict mathematical coupling means that any systematic downward shift in the measurement of τ inherently propagates into a suppressed initial condition for the growth of cosmic structure.

2. Late-Time Structure and Neutrino Compensation

   With a reduced primordial amplitude A_s, the universe begins with less initial power. However, the late-time clustering amplitude measured by DESI DR2 BAO and redshift-space distortions remains robustly high. In the standard model, massive neutrinos suppress the growth of structure below their free-streaming scale. If the initial power A_s is low, adding massive neutrinos will suppress structure further, drastically failing to match the late-time DESI observations. To compensate for the suppressed A_s, the fitting algorithms are forced to remove the neutrino suppression entirely. When zero mass is not enough to bridge the gap between a low A_s and the high observed late-time clustering, the likelihood surface naturally spills over into negative mass territory to artificially enhance structure growth.

3. Breaking the Degeneracy

   As demonstrated by Sailer et al., relying solely on primary CMB anisotropies leaves this degeneracy largely intact, leading to severely skewed posteriors when combined with aggressive late-time constraints. CMB lensing provides a theoretical pathway to break the τ–As–Σm_ν degeneracy, as lensing probes the integrated matter distribution and is directly sensitive to A_s independent of τ. However, if the lensing amplitude itself contains unmodeled systematic biases or points toward an internally inconsistent clustering amplitude, the combined likelihood will continue to pull Σm_ν into unphysical negative spaces to resolve the tension between the early-universe initial conditions and the late-universe matter distribution.

Phenomenological Prescription for Negative Mass

1. Boltzmann Code Limitations

   Standard cosmological Boltzmann solvers, such as CAMB and CLASS, are strictly programmed to integrate the phase-space distributions of massive particles. Because a negative mass squared implies an imaginary momentum or energy in the standard relativistic dispersion relation, these codes traditionally crash or truncate when sampling algorithms attempt to cross the m = 0 threshold. As highlighted by Elbers et al., this truncation creates an artificial boundary prior. When the true maximum of the likelihood surface lies in the negative regime, truncating the space forces the posterior to pile up artificially against the m = 0 boundary, masking the severity of the cosmological tension.

2. The sgn(m)√|m| Effective-Energy Prescription

   To safely navigate the unphysical regime without breaking the background Friedmann integration, theorists employ the sgn(m)√|m| effective-energy prescription. By smoothly continuing the energy density of the neutrino fluid across the zero-mass boundary, the modified background Lagrangian permits an analytical extension of the parameter space. The effective mass squared is analytically continued such that the background energy density responds inversely to negative mass values.

   ρ_ν(a) ∝ ∫ y² dy √( y² + a² sgn(Σm_ν) |Σm_ν|² / T_ν² )

   By mapping the mass parameter via sgn(m)√|m|, the Boltzmann solvers treat a negative mass as a fluid that actively enhances, rather than suppresses, the effective clustering potential. This mathematical trick does not propose that neutrinos literally possess negative mass, but rather provides a continuous, derivable likelihood surface that faithfully reports the statistical demands of the DESI DR2 data.

3. Prior Volume and Likelihood Masking

   The critical impact of this prescription was thoroughly analyzed by Green & Meyers, who demonstrated that restricting the parameter space to Σm_ν ≥ 0.06 eV heavily distorts the marginalized posteriors of other correlated cosmological parameters, such as the dark energy equation of state and the Hubble constant. By allowing the MCMC chains to explore the negative mass space via the effective-energy prescription, researchers uncovered that the −0.16 eV preference is not a subtle statistical fluctuation, but a deep, structural pull from the data. The data actively rejects the physical region, indicating that if the model is forced into the physical regime, the fit to the BAO and CMB observables degrades significantly.

Free-Streaming Suppression of Structure Growth

1. The Free-Streaming Scale

   Massive neutrinos decouple from the primordial plasma while still highly relativistic, retaining large thermal velocities. This high velocity prevents them from collapsing into dark matter halos below a characteristic free-streaming wavevector, k_fs. Modes larger than this wavevector (smaller k) evolve as if neutrinos were cold dark matter, while modes smaller than the free-streaming scale (larger k) experience a deficit in the gravitational potential.

   k_fs = √( (3/2) Ω_m ℋ² / v_th² )

   This scale dictates the exact point in the matter power spectrum where the suppression of structure begins. In a universe with physically massive neutrinos, the power spectrum is strictly damped at high wavenumbers relative to a massless neutrino cosmology.

2. Modifying the Growth Factor

   The underlying physics of this suppression can be derived from the Euler-Lagrange formalism applied to the perturbed cosmological fluid. As established by Loverde & Weiner, the evolution of cold dark matter and baryon density perturbations (δ_cb) in the presence of massive neutrinos requires modifying the background source term in the second-order differential growth equation. The fraction of energy density in neutrinos, f_ν = Ω_ν / Ω_m, effectively removes a portion of the clustering matter from the gravitational source term.

   δ̈_cb + ℋ δ̇_cb − (3/2) ℋ² Ω_m (1 − f_ν) δ_cb = 0

   Because f_ν is strictly positive in a physical universe, the driving term (1 − f_ν) is strictly less than one, mathematically guaranteeing a retarded growth rate for δ_cb at sub-horizon scales.

3. Reversing the Suppression

   When the MCMC sampler utilizes the sgn(m)√|m| prescription to explore negative mass sums, it effectively injects a negative f_ν into the growth equations. This transforms the (1 − f_ν) term into a value strictly greater than one, supercharging the gravitational source term. The result is an artificial, unphysical enhancement of the matter power spectrum at high wavenumbers. The DESI DR2 and CMB joint analysis actively seeks out this enhancement to reconcile the low primordial amplitude (driven by lower τ) with the high clustering amplitude observed in the local universe. This mathematical reversal of free-streaming suppression is the exact mechanism by which the anomalous −0.16 eV signal manifests in the data.

Implications: New Physics or Parameter-Space Artifact?

Conclusion