Why do DESI and Dark Energy Survey hints favor evolving dark energy over specific modified-gravity models?

# Why DESI favors evolving dark energy over specific modified-gravity models

The hint toward evolving dark energy from DESI is real, but it comes with important caveats—and the data don't yet strongly prefer one theoretical explanation over another.

The observational signal

DESI's baryon acoustic oscillation (BAO) measurements, combined with CMB and supernova data, show a statistical preference for dark energy that changes over cosmic time (characterized by parameters $w_0$ and $w_a$ in the standard "CPL" parametrization) rather than Einstein's cosmological constant, where $w = -1$ always [AG-2024.07-1106, AG-2025.04-303]. This preference is driven primarily by two DESI samples at lower redshifts ($z \approx 0.5$ and $0.7$) [AG-2024.07-1106]. However, the datasets show real tensions: CMB, BAO, and supernovae don't perfectly agree, which weakens the robustness of the "dynamical dark energy" claim [AG-2025.04-303].

Why modified gravity doesn't fit as neatly

When researchers test specific modified-gravity theories against this data, they run into a problem: the data prefer solutions that cross the "phantom divide" ($w = -1$) without staying there [AG-2024.07-1106, AG-2024.07-448]. Simple scalar-field models and quintessence—the most natural modifications to gravity—struggle to match this behavior. Quintessence models, which respect basic energy conditions, produce only marginal improvements over a cosmological constant [AG-2025.05-293]. Even the most general Horndeski scalar-tensor theories, which can safely cross the phantom divide, require specific nonminimal coupling to gravity to match the DESI signal [AG-2024.07-448].

The deeper issue: the data aren't asking for a *specific* physical mechanism yet. The CPL parametrization (a simple 2-parameter model) captures the deviation from $\Lambda$CDM, but it may be too flexible—it can represent many behaviors that no actual field theory naturally produces [AG-2025.02-158]. When you fit a rolling scalar field with more realistic constraints, it doesn't fit DESI significantly better than $\Lambda$CDM [AG-2025.02-158].

A local-universe caveat

Critically, the signal for evolution in dark energy comes almost entirely from very low redshifts ($z < 0.1$)—our recent cosmic neighborhood [AG-2024.06-1160]. This raises an uncomfortable question: is the Universe's expansion history genuinely changing, or are we misunderstanding the local structure of space around us at scales of ~300 million light-years [AG-2024.06-1160]? Modified-gravity theories predict specific signatures in gravitational wave propagation that could, in principle, distinguish the two scenarios [AG-2025.09-201], but current data don't yet have the precision to test this.

Bottom line: DESI hints at evolving dark energy partly because the data show tensions that a constant $\Lambda$ doesn't fully resolve—and evolving models absorb that tension better statistically. But no single physical theory (quintessence, Horndeski, $f(R)$ gravity) yet emerges as the clear winner because the signal is weak, redshift-dependent, and possibly contaminated by local-universe systematics.