Six DESI DR2 distance pairs give a weak bound on past acceleration in a flat Gaussian distance model

We test whether public DESI DR2 distances reject a flat expansion history that never accelerates. The test uses six transverse and radial distance pairs, linear contrasts, and the released Gaussian covariance. It requires no sound horizon calibration. We allow any positive, differentiable expansion rate in the point-distance model. The largest violation is 2.6725 standard errors at redshift 0.934. Correction for all six bins gives p = 0.02238. A bound that does not require independence between bins gives p = 0.02259. Test inversion gives a simultaneous 95% upper confidence limit of -0.01378 for the minimum deceleration parameter between redshift zero and 0.934. The evidence is modest. Omitting this pair gives p = 0.625 for the remaining five pairs. Increasing all standard errors by 12% removes rejection at the 5% level. Setting the negative within-bin covariance to zero would overstate the evidence. A radial-only control gives p = 0.493. We also give finite constraints for sampled flat histories with no acceleration and independently verify their numerical fit. The result assumes spatial flatness and Gaussian point distances at the effective redshifts. It does not locate acceleration in redshift or measure it today. The package preserves code, unchanged source data, provenance, results, figures, and reproduction instructions.

DESI DR2 · baryon acoustic oscillations · cosmic acceleration · deceleration parameter · Gaussian likelihood · simultaneous confidence bounds · covariance · reproducible research