A new preprint from April 2026 (arXiv:2604.11879) reports results from an ALMA spectroscopic follow-up campaign targeting an 870μm-selected sample of dusty star-forming galaxies (DSFGs) in the GOODS-South field. Using targeted linescans, the team secured or confirmed spectroscopic redshifts for 20 additional sources, bringing the total with reliable spec-z to 52 out of 75 galaxies—a 69% completeness rate. For the brighter subset above 2.5 mJy, completeness reaches an impressive 97%. This represents the highest spectroscopic completeness achieved for an unbiased DSFG sample near the confusion limit of single-dish submillimeter telescopes.

Methodologically, the survey relies on ALMA's ability to detect redshifted emission lines such as CO transitions or [C II] without depending on near-infrared counterparts, which often fail for heavily dust-obscured systems. The sample size, while modest at 75 galaxies in a single deep field, is statistically powerful because of the near-total completeness at brighter fluxes. Limitations include potential cosmic variance from the GOODS-S patch, some 'tentative' spec-z assignments, and the fact that only 29% of the sample had complementary JWST/NIRSpec coverage—though nearly all of those with JWST data yielded successful redshifts regardless of flux or color.

The analysis reveals that every photometric redshift code tested shows outlier fractions exceeding 20%, with a systematic bias toward overestimating redshift. This inflates the apparent abundance of DSFGs at z>4. The authors conclude that no more than 10% of bright (f_870 ≳ 2 mJy) DSFGs lie above z=4, and fewer than 2% above z=5. This steep drop-off in the first 1.5 billion years after the Big Bang carries major implications.

Previous optical and near-infrared surveys have long been known to miss dust-obscured star formation, but this work flips the script at the highest redshifts: photo-z methods were actually overpredicting the dusty contribution beyond z=4. When synthesized with the landmark Madau & Dickinson (2014) cosmic star-formation history review (arXiv:1403.0007), which relied partly on UV luminosity functions corrected by IR templates, the new constraints suggest the high-redshift tail of the SFRD must be revised downward. Similarly, results from the Herschel-ATLAS and SCUBA-2 surveys (e.g., Dudzevičiūtė et al. 2020, arXiv:2006.04919) that modeled DSFG redshift distributions using less robust photo-z now appear biased high.

What much of the existing coverage missed is the interplay with JWST's emerging census of UV-bright galaxies at z>6. While JWST reveals surprisingly massive, star-forming systems early on, the ALMA data indicate the dustiest, most vigorously star-forming subset was not yet dominant—implying dust production via AGB stars or supernovae may have taken longer to ramp up than some chemical evolution models assume. This pattern echoes the 1990s discovery that submillimeter galaxies dominate the SFRD at z≈2–3; today's refinement shows a sharper cutoff at even earlier epochs.

Genuine analysis: These findings challenge hydrodynamical simulations (IllustrisTNG, EAGLE) that overproduce massive dusty galaxies at z>5 to match bright submm number counts. The systematic photo-z failure also warns against over-reliance on template-fitting codes for JWST-selected 'little red dots' or other dusty candidates. If optical surveys systematically underestimate total SFRD at moderate redshifts while overestimating the dusty high-z tail, our overall picture of galaxy assembly requires recalibration. Future ALMA 2.0 or ngVLA observations will be essential to push completeness even deeper and across wider fields, but this work already delivers a pivotal constraint that reshapes how we model the obscured universe.