NICE: NOEMA Forming Cluster Survey

• NICE is a dual-facility survey that systematically identifies high-redshift protocluster cores using uniform IRAC/Herschel selection and mm-wave spectroscopic confirmation.
• The survey uses precise CO and [CI] line detections from NOEMA and ALMA to robustly measure molecular gas, star formation rates, and dark matter halo properties.
• Results indicate enhanced star formation and substantial gas reservoirs in compact protocluster cores, supporting sustained growth of early massive cluster galaxies.

Searching arXiv for the NICE survey papers and related records. Massive protoclusters at $z\sim1.5$–4 occupy the peak of the cosmic star formation history and are therefore central to the study of how massive galaxies in present-day clusters assemble. The Noema formIng Cluster survEy (NICE) is a NOEMA Large Program with a complementary ALMA program designed to build a homogeneous, statistically selected, spectroscopically confirmed sample of forming cluster cores and compact groups at high redshift, and to characterize their star formation, molecular gas, baryons, and dark matter through CO lines, millimeter continuum, and uniform ancillary data (Zhou et al., 14 Jul 2025). Across the survey literature, NICE is defined by a uniform candidate selection based on overdensities of red IRAC sources spatially coincident with red Herschel colors, followed by mm-wave spectroscopic confirmation and multi-method halo-mass inference (Sillassen et al., 2024).

1. Survey definition, scope, and scientific rationale

NICE is described as a two-facility millimeter spectroscopic program designed to build a homogeneous, statistically powerful census of massive groups and proto-clusters in the early Universe and to characterize their baryons and dark matter (Sillassen et al., 2024). The Large Programs comprise 159 hr with NOEMA and 40 hr with ALMA. One survey description states that 48 overdensities were selected across five extragalactic legacy fields for NOEMA Band 1 spectroscopy, with an ALMA Cycle 8 program complementing southern targets (Zhou et al., 2023). A later survey paper states that the program targets 69 massive galaxy group candidates at $z>2$ in six deep fields—COSMOS, Lockman Hole, Elais-N1, Boötes, XMM-LSS, and ECDFS—over $46\,\mathrm{deg}^2$ (Sillassen et al., 2024). This suggests that NICE evolved from an initial operational sample to a broader multi-field program.

The primary scientific motivation is the lack of a statistically and homogeneously selected and spectroscopically confirmed sample of massive protoclusters at $2\lesssim z\lesssim4$, where cold gas accretion and concentrated star formation are expected to be efficient (Zhou et al., 2023). Earlier proto-cluster work is described as heterogeneous in tracer choice, often based on radio galaxies, narrow-band emission-line overdensities, or Ly$\alpha$ tomography, and therefore prone to differences in selection function and evolutionary stage. NICE was designed to mitigate those limitations by adopting a single color-plus-overdensity selection and a common spectroscopic confirmation strategy (Zhou et al., 14 Jul 2025).

Within COSMOS, NICE-COSMOS consists of eight confirmed protoclusters at $1.5$r\leq1\,R_{\rm vir}$, with virial radii of 141–429 pkpc and 13–31 members per structure (Zhou et al., 14 Jul 2025). The COSMOS pilot paper reports that these eight systems are confirmed at $1.65\leq z\leq3.61$ and have best halo-mass estimates of $\log(M_{\rm h}/{\rm M_\odot})=12.8$–13.7 with uncertainty of 0.3 dex (Sillassen et al., 2024). The Lockman Hole discovery paper adds an even higher-redshift compact group, LH-SBC3 at $z=3.95$, confirmed via CO(4–3) and CI in four massive galaxies (Zhou et al., 2023).

2. Candidate selection and spectroscopic confirmation

NICE candidate selection combines an IRAC overdensity criterion with Herschel/SPIRE color cuts. The IRAC selection requires

$z>2$0

with a surface-density overdensity threshold $z>2$1, where

$z>2$2

and $z>2$3 or 10 nearest neighbors (Zhou et al., 14 Jul 2025). The SPIRE “350 $z>2$4m peaker” cuts are

$z>2$5

in the survey overview, with one COSMOS pilot description listing $z>2$6 (Sillassen et al., 2024). Both formulations appear in the NICE literature. The stated motivation is that these criteria select overdense, intensively star-forming massive environments at $z>2$7–4 (Zhou et al., 14 Jul 2025).

This dual selection is meant to identify dusty, star-forming structures while reducing projection contamination relative to submillimeter overdensities alone (Zhou et al., 2023). At the same time, the survey literature explicitly notes selection effects: the IRAC/SPIRE-based targeting emphasizes intensively star-forming protoclusters and may over-represent systems in active growth phases (Zhou et al., 14 Jul 2025). Severe blending can also suppress the apparent IRAC overdensity in compact cores, as discussed for HPC1001, which satisfied the red Herschel color cuts but was not an IRAC overdensity because of source blending at IRAC resolution (Sillassen et al., 2024).

Spectroscopic confirmation is carried out with NOEMA and ALMA through CO and [CI] line detections, with Subaru/FMOS H$z>2$8 used for the $z>2$9 COSMOS system where the ALMA tunings did not cover strong mm lines (Sillassen et al., 2024). NOEMA observations at 3 mm were designed with two frequency setups per target, covering CO(3–2) and CO(4–3) over $46\,\mathrm{deg}^2$0, with typical line sensitivity of 0.13 mJy beam$46\,\mathrm{deg}^2$1 over 500 km s$46\,\mathrm{deg}^2$2 at $46\,\mathrm{deg}^2$3 GHz, continuum sensitivity of $46\,\mathrm{deg}^2$4Jy beam$46\,\mathrm{deg}^2$5, and $46\,\mathrm{deg}^2$6 synthesized beams (Sillassen et al., 2024). ALMA Bands 4 and 5 used four tunings spanning 135–183 GHz, targeting CO(3–2), CO(4–3), CO(5–4), and [CI]$46\,\mathrm{deg}^2$7–$46\,\mathrm{deg}^2$8 depending on redshift, with line sensitivity of 0.13 mJy beam$46\,\mathrm{deg}^2$9 over 500 km s$2\lesssim z\lesssim4$0 at $2\lesssim z\lesssim4$1 GHz, continuum $2\lesssim z\lesssim4$2Jy beam$2\lesssim z\lesssim4$3, and typical resolution $2\lesssim z\lesssim4$4 (Sillassen et al., 2024).

A unified NOEMA+ALMA line-search pipeline extracted spectra directly in the uv domain with GILDAS uvfit, adopted a $2\lesssim z\lesssim4$5 power law for the continuum, and searched for lines with an integrated-S/N algorithm (Sillassen et al., 2024). In the COSMOS pilot sample, 22 significant lines—CO(3–2), CO(4–3), CO(5–4), and CI—were detected in 20 galaxies, with typical $2\lesssim z\lesssim4$6 over the full band (Sillassen et al., 2024). Photometric-redshift PDFs from COSMOS2020 were used to disambiguate single-line identifications, and 58% (95%) of spectroscopic solutions lie within the $2\lesssim z\lesssim4$7 ($2\lesssim z\lesssim4$8) $2\lesssim z\lesssim4$9 PDF (Sillassen et al., 2024).

3. Observational datasets and measurement framework

NICE combines mm-wave spectroscopy with broad ancillary datasets. The survey overview lists COSMOS2020 photometry and LePhare-based stellar masses and SFRs, VLA 3 GHz, the super-deblended FIR catalog, Herschel/SPIRE, and JWST/COSMOS-Web imaging for two structures (Zhou et al., 14 Jul 2025). In crowded cores, integrated FIR/sub-mm fluxes are measured with a dedicated super-deblending procedure: one prior per group is placed at the SCUBA-2 850 $\alpha$0m peak, sources within $\alpha$1 are excluded from the ancillary prior list, PSF fitting is performed on Herschel/PACS and SPIRE, SCUBA-2, and MeerKAT images, and $\alpha$2m totals are built from the sum of member fluxes (Sillassen et al., 2024). The integrated SEDs are then fit with STARDUST using the Magdis et al. (2012) dust templates at the group spectroscopic redshift, treating $\alpha$3 points as detections and the rest as $\alpha$4 upper limits (Sillassen et al., 2024).

For member-galaxy star formation rates, the primary estimator is VLA 3 GHz when $\alpha$5, using the infrared–radio correlation $\alpha$6 from Delvecchio et al. (2021) and a spectral index $\alpha$7; the conversion to SFR follows Kennicutt (1998), adjusted via $\alpha$8 inferred from radio via IRRC (Zhou et al., 14 Jul 2025). For 3 GHz $\alpha$9, SFR is derived from $1.5Zhou et al., 14 Jul 2025). For the integrated quantity $1.5SED fits (Zhou et al., 14 Jul 2025).

The survey provides explicit cross-checks between SFR estimators for radio detections with $1.5Zhou et al., 14 Jul 2025). A radio-loud galaxy, COSMOS-SBC6 ID 839791, uses $1.5AGN radio contamination because its radio luminosity exceeds the AGN-dominance threshold from Wang 2024 (Zhou et al., 14 Jul 2025).

Stellar masses are taken from COSMOS2020 using LePhare, with uncertainties as reported in COSMOS2020 and not re-derived in the NICE analysis (Zhou et al., 14 Jul 2025). Molecular gas masses are preferentially estimated from Rayleigh–Jeans dust continuum at observed 2–3 mm, corresponding to rest-frame $1.5Zhou et al., 14 Jul 2025). A CO-based cross-check converts CO(3–2) or CO(4–3) to CO(1–0) using the ASPECS main-sequence CO SLED at $1.5Zhou et al., 14 Jul 2025). The two gas-mass methods agree within uncertainties, with

$r\leq1\,R_{\rm vir}$0

a median log ratio $r\leq1\,R_{\rm vir}$1 dex, and scatter $r\leq1\,R_{\rm vir}$2 dex; typical $r\leq1\,R_{\rm vir}$3 values are 3.52–4.78 $r\leq1\,R_{\rm vir}$4 (Zhou et al., 14 Jul 2025).

The main derived quantities are defined explicitly:

$r\leq1\,R_{\rm vir}$5

Main-sequence and starburstiness are parameterized using Schreiber et al. (2015), with

$r\leq1\,R_{\rm vir}$6

and

$r\leq1\,R_{\rm vir}$7

where starbursts are defined by $r\leq1\,R_{\rm vir}$8 (Zhou et al., 14 Jul 2025).

4. Halo masses, structural inference, and the COSMOS confirmed sample

NICE introduces six halo-mass estimators and adopts a consensus mass based on three of them (Sillassen et al., 2024). These methods include a BCG stellar-to-halo mass estimate, two total-stellar-mass to halo-mass conversions, an overdensity-plus-galaxy-bias estimate, and two NFW-based fits to the radial stellar mass density (Sillassen et al., 2024). The adopted cosmology in the COSMOS pilot is flat with $r\leq1\,R_{\rm vir}$9, $1.65\leq z\leq3.61$0, and $1.65\leq z\leq3.61$1 (Sillassen et al., 2024). The NICE census paper uses $1.65\leq z\leq3.61$2, $1.65\leq z\leq3.61$3, and $1.65\leq z\leq3.61$4, with a Chabrier (2003) IMF (Zhou et al., 14 Jul 2025).

The overdensity-plus-bias method defines the projected overdensity as

$1.65\leq z\leq3.61$5

and uses

$1.65\leq z\leq3.61$6

where $1.65\leq z\leq3.61$7 is the mean matter density at $1.65\leq z\leq3.61$8 and $1.65\leq z\leq3.61$9 is the linear bias calibrated from Tinker et al. (2010); because $\log(M_{\rm h}/{\rm M_\odot})=12.8$0 depends on $\log(M_{\rm h}/{\rm M_\odot})=12.8$1, an iterative solution is adopted (Sillassen et al., 2024). The NFW-based methods fit the projected stellar mass surface density $\log(M_{\rm h}/{\rm M_\odot})=12.8$2 in annuli from $\log(M_{\rm h}/{\rm M_\odot})=12.8$3 pkpc to $\log(M_{\rm h}/{\rm M_\odot})=12.8$4 pMpc, centered on the FIR-peak-weighted barycenter of spectroscopically confirmed dusty members (Sillassen et al., 2024).

A central result is that the radial stellar mass density of all eight COSMOS structures is consistent with an NFW profile, supporting the interpretation that they are collapsed structures hosted by a single dark matter halo (Sillassen et al., 2024). The profile is written as

$\log(M_{\rm h}/{\rm M_\odot})=12.8$5

Compared to Ludlow et al. (2016), the best-fit concentrations show a scatter of $\log(M_{\rm h}/{\rm M_\odot})=12.8$6 dex and a mean offset of $\log(M_{\rm h}/{\rm M_\odot})=12.8$7 dex, implying higher concentration on average (Sillassen et al., 2024). This suggests early assembly and relatively compact halos for several systems.

Across the halo-mass estimators $\log(M_{\rm h}/{\rm M_\odot})=12.8$8, $\log(M_{\rm h}/{\rm M_\odot})=12.8$9, and $z=3.95$0, the inter-method scatter is 0.2–0.3 dex with mean offsets $z=3.95$1 dex. NICE therefore adopts the median of these three as the “best” halo mass with a conservative $z=3.95$2 dex uncertainty (Sillassen et al., 2024). The eight COSMOS systems span $z=3.95$3–13.7 with $z=3.95$4–430 pkpc; examples include HPC1001 $z=3.95$5, COS-SBCX1 $z=3.95$6, and COS-SBC4 $z=3.95$7 (Sillassen et al., 2024).

The COSMOS pilot further derives baryonic accretion rates using the Goerdt et al. (2010) scaling,

$z=3.95$8

yielding $z=3.95$9–$z>2$00 for the COSMOS groups (Sillassen et al., 2024). A simple two-regime model relates SFR and BAR through the theoretical cold-stream threshold $z>2$01, and the combined NICE+literature sample yields the empirical relation

$z>2$02

with a scatter of $z>2$03 dex (Sillassen et al., 2024). This provides an explicit observational link between gas supply and star formation in dense environments.

5. Star formation and cold gas properties in protocluster members

The main NICE-COSMOS census reports a steep increase in star formation rates per halo mass with redshift in intensively star-forming protoclusters (Zhou et al., 14 Jul 2025). Specifically, NICE-COSMOS, together with other NICE prototypes, lies approximately 1–2 dex above field values at $z>2$04 when compared to field evolution proportional to $z>2$05 (Zhou et al., 14 Jul 2025). The inferred scaling for (proto)clusters is consistent with

$z>2$06

extending trends observed at lower redshift to $z>2$07, with the data showing an approximately 1 dex increase from $z>2$08 to $z>2$09 (Zhou et al., 14 Jul 2025).

NICE argues that this enhancement is not driven by a higher starburst fraction. Member galaxies generally follow the star-forming main sequence, and the survey reports a low starburst fraction indicated by the absence of a systematic elevation in the $z>2$10 distribution relative to the field (Zhou et al., 14 Jul 2025). Instead, the dominant effect is the concentration of massive, gas-rich star-forming galaxies in protocluster cores (Zhou et al., 14 Jul 2025). The radial mass segregation is strong: the median stellar mass within $z>2$11 is approximately $z>2$12, compared with approximately $z>2$13 in the outskirts, a difference of more than 1 dex (Zhou et al., 14 Jul 2025). Among 31 massive galaxies with $z>2$14, 21 lie within $z>2$15, and all but one are detected in CO (Zhou et al., 14 Jul 2025).

Detected protocluster galaxies have median gas properties

$z>2$16

(Zhou et al., 14 Jul 2025). The most massive protocluster galaxies show enhanced gas reservoirs relative to field galaxies. For $z>2$17, also described as the most massive protocluster galaxies with $z>2$18, $z>2$19 is typically approximately $z>2$20 the field level at $z>2$21, declining to approximately $z>2$22 at $z>2$23, after normalization to A3COSMOS field scaling relations that account for $z>2$24, $z>2$25, and $z>2$26 (Zhou et al., 14 Jul 2025). These galaxies remain on the main sequence rather than in the starburst regime, implying sustained growth via large gas reservoirs rather than brief starburst episodes (Zhou et al., 14 Jul 2025).

Less massive members are detected mainly when they are strongly starbursting, and their $z>2$27 and $z>2$28 are comparable to field galaxies at the same $z>2$29, $z>2$30, and $z>2$31 (Zhou et al., 14 Jul 2025). The survey sensitivity is relevant here: the $z>2$32 continuum sensitivity corresponds to $z>2$33 at $z>2$34, which biases detections toward massive members and lower-mass starbursts (Zhou et al., 14 Jul 2025). The mean radiation field $z>2$35 derived from integrated protocluster FIR SEDs follows main-sequence evolution, with some systems slightly lower, consistent with moderately longer gas depletion times inferred for members (Zhou et al., 14 Jul 2025).

A notable clarification in the NICE census concerns a notation issue: the abstract threshold “$z>2$36” is explicitly stated to refer to galaxy stellar mass and should be read as $z>2$37 (Zhou et al., 14 Jul 2025). This addresses a potential misconception, since halo masses of the protoclusters themselves are much larger.

6. Early discoveries, environmental interpretation, and limitations

The first published NICE discovery, LH-SBC3 in the Lockman Hole at $z>2$38, established the feasibility of the survey strategy for compact, IR-luminous structures (Zhou et al., 2023). Four compact members within approximately 180 pkpc were confirmed in CO(4–3), two also in CI, with an average redshift $z>2$39 and velocity offsets of $z>2$40, $z>2$41, $z>2$42, and $z>2$43 km s$z>2$44 (Zhou et al., 2023). The core has an estimated halo mass of $z>2$45 and total SFR of $z>2$46 (Zhou et al., 2023). One member hosts a radio-loud AGN with

$z>2$47

above the common radio-loud threshold (Zhou et al., 2023).

Using $z>2$48 to convert CO(4–3) to CO(1–0) and adopting a ULIRG-like $z>2$49, the paper derives molecular gas masses of approximately $z>2$50, $z>2$51, $z>2$52, and $z>2$53 for the four confirmed members (Zhou et al., 2023). Corresponding depletion times are approximately 0.10, 0.022, 0.037, and 0.025 Gyr, identifying the system as a compact starbursting core (Zhou et al., 2023). The halo-integrated quantity $z>2$54 is approximately $z>2$55–$z>2$56, about 2 dex above clusters at $z>2$57–2 (Zhou et al., 2023).

The later NICE-COSMOS census indicates that such extreme systems should not be taken as representative of the entire survey population. In COSMOS, the member galaxies generally remain on the field main sequence, and the core enhancement in $z>2$58 arises primarily from the concentration of massive, gas-rich main-sequence galaxies rather than from a global excess of starbursts (Zhou et al., 14 Jul 2025). This establishes an environmental picture in which the dominant mechanism is not universally short-lived burst activity, but sustained star formation supported by substantial gas reservoirs in the most massive core galaxies (Zhou et al., 14 Jul 2025). A plausible implication is that NICE samples multiple evolutionary phases, from compact starbursting cores at $z>2$59 to gas-rich, main-sequence-dominated protocluster cores at $z>2$60–3.

The survey literature is explicit about limitations. Candidate selection favors dusty, intensely star-forming systems and may bias the sample toward active phases (Zhou et al., 14 Jul 2025). Membership at low stellar mass is based on photometric redshifts and is therefore subject to contamination; the high-mass end is more secure because the most massive and high-SFR members are often CO-detected and spectroscopically confirmed (Zhou et al., 14 Jul 2025). Small-number statistics remain important: eight protoclusters in COSMOS are described as substantial but still limited (Zhou et al., 14 Jul 2025). Gas masses also carry systematic uncertainties because the dust-continuum method depends on Rayleigh–Jeans calibration and dust properties, while the CO-based method depends on excitation corrections and $z>2$61–metallicity scaling; the two methods agree within $z>2$62 dex, but absolute $z>2$63 uncertainties remain (Zhou et al., 14 Jul 2025). Halo masses inherit uncertainties from the use of multiple proxies, and the survey notes that underestimation of $z>2$64 at high redshift could bias $z>2$65 high (Zhou et al., 14 Jul 2025).

7. Evolutionary significance and place in the literature

NICE places its results in the context of previous studies of cluster and protocluster evolution. The steep evolution of $z>2$66, extending to $z>2$67 and consistent with $z>2$68, aligns with extrapolations of lower-redshift trends discussed by Webb, Alberts, and others (Zhou et al., 14 Jul 2025). The lack of a global starburst excess is presented as consistent with several H$z>2$69-emitter studies of young protoclusters, including Spiderweb-like systems (Zhou et al., 14 Jul 2025). At the same time, NICE differs from some extreme $z>2$70 starbursting cores such as SPT2349-56 and DRC: by $z>2$71, the most massive NICE core galaxies are gas-richer than field counterparts, while at $z>2$72 the enhancement is milder, approximately $z>2$73 (Zhou et al., 14 Jul 2025). This suggests environmental effects and selection may evolve with redshift.

The COSMOS pilot compares the confirmed structures with simulations and concludes that all eight are consistent with being progenitors of present-day clusters with $z>2$74 (Sillassen et al., 2024). Their most massive member galaxies have stellar masses consistent with brightest-cluster-galaxy progenitors in the TNG300 simulation (Sillassen et al., 2024). Combined with the NFW-like stellar-mass profiles and the concentration of massive, gas-rich galaxies toward the core, this supports the interpretation that NICE is observing collapsed group-scale halos in late stages of assembly, in which proto-BCGs are rapidly growing (Sillassen et al., 2024).

The NICE census further argues that the concentration of gas-rich, massive galaxies with $z>2$75 in protocluster cores at $z>2$76–3, together with $z>2$77 Gyr, suggests sustained in-situ growth and early assembly of central massive ellipticals in present-day clusters (Zhou et al., 14 Jul 2025). The elevated $z>2$78 is interpreted as an early, efficient build-up phase in cluster cores, consistent with a “downsizing” scenario (Zhou et al., 14 Jul 2025). This suggests that, in the environments selected by NICE, the early formation of massive cluster galaxies is driven by persistent gas-rich star formation within dense cores rather than exclusively by brief starburst episodes or immediate quenching.

Recommended future work in the NICE literature includes increasing spectroscopic completeness at lower stellar mass, obtaining more uniform $z>2$79 constraints through low-$z>2$80 CO and metallicity measurements, conducting deeper mm-continuum observations below $z>2$81 at $z>2$82–3, extending mosaics beyond $z>2$83 to map environmental gradients and inflows, and comparing NICE systematically with differently selected protocluster samples and simulations of gas accretion and quenching timescales (Zhou et al., 14 Jul 2025). These directions follow directly from the survey’s central premise: a homogeneous, CO/continuum-anchored census can turn high-redshift cluster formation from a collection of heterogeneous case studies into a comparative empirical framework.