A cosmic collision in the making: JWST/NIRISS reveals a merging and maturing protocluster core at z 3 around a red quasar

We report the discovery of The Step, a dense protocluster at z = 3 revealed by JWST/NIRISS Wide-Field Slitless Spectroscopy (WFSS) of the field around the luminous quasar and starburst SDSSJ165202.64+172852.3. The protocluster contains at least 18 member galaxies and exhibits signs for two substructures in the 3D position-velocity space – notably a distinctive step-like distribution, from which the nickname derives. We propose that the Step may be undergoing a merger of two haloes, which could also explain the unusually large velocity span (700 − 850km/s) between galaxies along the densest line of sight. This sightline contains four/seven galaxies within projected distances of 20/70 kpc from the quasar. Within a 320 × 230 kpc2 region, we find 10 galaxies (including the quasar), consistent with a remarkably dense protocluster core. We derive a galaxy overdensity of a factor > 100 in this core and > 15 in the full field (1 Mpc2), making it one of the densest structures known at this redshift, and a very massive halo of 13.1 < log Mhalo /M⊙ < 13.8, which suggests the structure could potentially evolve into a massive, Coma-like cluster at z = 0. Compared to the field, the star-forming galaxies in the Step appear to show slightly elevated levels of star formation activity. Yet, we find two quiescent galaxies (log M⋆ /M⊙ ∼ 10.2, 10.4 ± 0.8) with strong continuum breaks and no evidence for ongoing star formation. Compared to other protoclusters at z = 3, the Step structure exhibits broadly similar, or slightly suppressed star-forming activity for its halo mass. We also identify two new AGN candidates based on optical line ratios and find tentative evidence for an enhanced AGN fraction (10 − 20%) with respect to the field. The protocluster constitutes one of the emerging handful of maturing structures at 2 ⩽ z ⩽ 4 that host both star-forming and quenched galaxies. This redshift regime may thus be a key transitional epoch for cluster studies, where overdensities transition from being the most active sites of star formation in the early Universe towards shaping the massive passive ellipticals seen in the cores of today's clusters.

Improvements to AI Systems:

1. 3D Overdensity Detection & Substructure Classification

• Improve clustering algorithms to identify non-Gaussian, step-like distributions in position-velocity space (e.g., using topological data analysis or change-point detection) rather than assuming smooth spherical overdensities.

• The improved AI can automatically flag merging halo candidates from spectroscopic surveys, distinguishing between single virialized structures and two-component mergers based on velocity offsets and spatial alignment.

2. Halo Mass Estimation from Sparse Galaxy Samples

• Enhance Bayesian inference models to incorporate overdensity factors, velocity dispersions, and substructure signatures simultaneously, with priors from simulations of massive halos at z=3.

• The improved AI can predict halo mass ranges (e.g., log M halo = 13.1–13.8) for protoclusters with only 10–20 confirmed members, reducing reliance on large-N statistics.

3. Star-Formation Quenching Prediction in Dense Environments

• Train a classifier on multi-wavelength SEDs (JWST/NIRISS, photometric breaks) to separate quiescent vs. star-forming galaxies in overdensities, using continuum break strength and lack of emission lines as key features.

• The improved AI can identify early quenched galaxies in protoclusters (like the two found here) and predict their future evolution into passive ellipticals, even when their stellar masses are low (10 10 M sun).

4. AGN Fraction Enhancement Detection

• Develop a pipeline that combines optical line-ratio diagnostics (e.g., BPT diagrams) with spatial proximity to overdensity cores to estimate AGN fraction with confidence intervals.

• The improved AI can automatically flag AGN candidates in dense regions and statistically compare their fraction (10–20%) against field samples, correcting for selection biases from slitless spectroscopy.

5. Merger-Driven Velocity Span Modeling

• Implement a dynamical model that fits observed velocity distributions (e.g., 700–850 km/s) as a superposition of two halo components, using Monte Carlo Markov chains to infer merger geometry and line-of-sight projection effects.

• The improved AI can distinguish between a single massive halo with high velocity dispersion and a true merger, providing probabilities for each scenario from 3D spatial-velocity data.

6. Transitional Epoch Classification

• Build a redshift-dependent classifier that categorizes structures (2 ≤ z ≤ 4) into active star-forming protoclusters, transitional, or pre-virialized based on quenched fraction, AGN fraction, and star-formation rate relative to field.

• The improved AI can automatically map the evolutionary stage of any overdensity and predict its z=0 descendant mass (e.g., Coma-like cluster) using scaling relations calibrated on this and similar discoveries.

What the Improved AI System Can Do:

• Given a spectroscopic survey (e.g., JWST/NIRISS WFSS) of a high-z field, it can autonomously:

• Detect and characterize non-trivial substructures (e.g., steps, filaments) in 3D.

• Estimate halo mass and merger probability with uncertainty.

• Identify quenched galaxies and AGN candidates without manual SED fitting.

• Predict the structure’s evolutionary fate (e.g., massive cluster vs. group) and compare its star-formation activity to cosmic averages.

• Flag rare, maturing structures at z=2–4 that are key for studying the transition from starburst to quiescent cluster cores.