ALMA-CRISTAL: ISM Mapping at Cosmic Dawn

• ALMA-CRISTAL Program is a survey designed to spatially resolve [CII] emission and far-infrared dust continuum in typical main-sequence star-forming galaxies at z=4–6.
• It combines ALMA’s multi-configuration observations with HST and JWST imaging to achieve kiloparsec-scale mapping of gas kinematics, outflows, and stellar structure.
• Key insights include the discovery of extended [CII] disks, turbulent yet moderately rotation-supported systems, and the calibration of early galaxy assembly and ISM phase balance.

ALMA-CRISTAL—“[CII] Resolved ISM in STar-forming galaxies with ALMA”—is an ALMA Large Program (ID 2021.1.00280.L) designed to spatially resolve the [CII] $158\,\mu\mathrm{m}$ line and far-infrared dust continuum in typical main-sequence star-forming galaxies at $4HST and JWST. The program targets the interval when the Universe was $\sim 1$ Gyr old and reionization had just ended, with the explicit aim of building a kiloparsec-scale census of ISM structure, dust-obscured star formation, kinematics, outflows, and circumgalactic gas in representative massive galaxies rather than compact starburst-selected systems (Herrera-Camus et al., 9 May 2025, Lee et al., 15 Jul 2025).

1. Program definition and scientific rationale

CRISTAL is motivated by the need to move beyond small, heterogeneous, and often starburst-biased high-redshift samples. Its central tracer, the [CII] line at rest frequency $\nu_{\rm rest}=1900.537$ GHz, is a dominant coolant of the multiphase ISM and can arise in photodissociation regions around young star clusters as well as in diffuse neutral and ionized gas. In the CRISTAL redshift range, the line is shifted into ALMA Band 7 and remains bright enough to trace gas kinematics and morphology to and beyond effective radii (Ikeda et al., 2024, Lee et al., 15 Jul 2025).

Program papers organize the science around four primary goals: kinematics, outflows, spatial distribution of the ISM and star formation, and ISM conditions. The survey was designed to couple ALMA Band 7 spectroscopy and continuum imaging with HST/WFC3 rest-UV morphology and, in later work, JWST/NIRCam and JWST/NIRSpec observations. This combination enables direct comparison of stellar light, unobscured star formation, dust-embedded star formation, and cool gas geometry on $\sim$kpc scales (Herrera-Camus et al., 9 May 2025, Faisst et al., 17 Oct 2025).

A defining feature of CRISTAL is its emphasis on typical main-sequence galaxies. The program explicitly contrasts with earlier ALMA work dominated by massive dusty star-forming galaxies and quasars. This shift in target class is essential because the representative galaxy population at $z\sim 5$ governs most stellar-mass build-up, yet is difficult to characterize in the far-infrared and submillimeter regime. CRISTAL therefore functions both as a resolved ISM survey and as a calibration dataset for interpreting JWST-era studies of early galaxy assembly (Mitsuhashi et al., 2023, Herrera-Camus et al., 9 May 2025).

2. Survey design, target selection, and sample growth

The initial targeted sample consisted of 19 main-sequence galaxies drawn from ALPINE. Selection required robust spectroscopic redshifts at $4\le z\le 6$, [CII] detections in ALPINE, stellar masses $\log(M_\star/M_\odot)\gtrsim 9.5$, HST imaging, and specific star formation rates within a factor of three of the star-forming main sequence. Several papers also note the practical role of HST/WFC3 UV coverage in ensuring that stellar and UV morphologies could be compared directly to ALMA maps (Herrera-Camus et al., 9 May 2025, Mitsuhashi et al., 2023).

As the program developed, the working sample expanded through deeper CRISTAL detections, archival ALMA data, and pilot targets. Different program papers report different accounting schemes tied to their scientific scope: the resolved SED study spans 25 fields containing 36 galaxies, the outflow analysis refers to 37 galaxies including serendips and resolved multiples, and the survey overview reports 39 main-sequence galaxies after including newly detected galaxies in CRISTAL fields, archival data, and pilot study targets (Li et al., 2024, Birkin et al., 24 Apr 2025, Herrera-Camus et al., 9 May 2025).

The survey overview gives the broadest census: 39 main-sequence galaxies with median $z\approx 5.1$, median stellar mass $M_\star \approx 10^{10.1}\,M_\odot$, and median $4signal-to-noise ratio and spatial resolution, while the dedicated [CII]-size study reports structural parameters for $4Herrera-Camus et al., 9 May 2025, Lee et al., 15 Jul 2025, Ikeda et al., 2024).

This sample construction matters methodologically. The survey was designed to be representative of the massive main-sequence population at $4COSMOS and GOODS-S fields with rich ancillary data (Lee et al., 15 Jul 2025, Herrera-Camus et al., 9 May 2025).

3. Observational strategy and analysis framework

CRISTAL’s ALMA setup was optimized for both uv-coverage and spatial dynamic range. Targets were observed in at least two 12 m-array configurations: a compact configuration sensitive to large scales and an extended configuration sensitive to sub-kpc structure. Typical final [CII] cubes have synthesized beams of $4Ikeda et al., 2024, Herrera-Camus et al., 9 May 2025).

A recurrent methodological choice is direct visibility fitting. In the [CII]-size analysis, emission is modeled with UVMULTIFIT as a 2D exponential disk with parameters for position, total line flux, major-axis FWHM, axis ratio $4

$\sim 1$0

with

$\sim 1$1

This uv-plane approach is used to avoid CLEAN-mask choices, weighting trade-offs, and PSF deconvolution uncertainties. Rest-UV sizes are measured with GALFIT assuming $\sim 1$2, and FIR sizes are fit as circular exponentials in the uv-plane (Ikeda et al., 2024).

The kinematic papers adopt a separate forward-modeling framework. DysmalPy constructs full 3D model cubes including projection, beam convolution, and spectral broadening, and fits 1D velocity and dispersion profiles extracted along the kinematic major axis. The baryonic model comprises an exponential disk and, where needed, a compact de Vaucouleurs bulge; the dark matter halo follows an NFW profile. Pressure support is included through

$\sim 1$3

and inclinations are inferred from axis ratios using an intrinsic disk thickness $\sim 1$4 (Lee et al., 15 Jul 2025).

CRISTAL also introduced resolved UV-to-FIR SED modeling at $\sim 1$5–1 kpc resolution. The 14-galaxy study uses PSF-matched HST, JWST, and ALMA data with magphys, while the JWST PRIMER study of four galaxies uses BAGPIPES pixel-by-pixel fits to NIRCam photometry. These analyses quantify $\sim 1$6, $\sim 1$7, $\sim 1$8, $\sim 1$9, and related maps, and show explicitly that ALMA continuum data reduce uncertainties in $\nu_{\rm rest}=1900.537$0, $\nu_{\rm rest}=1900.537$1, and SFR by breaking age–dust degeneracies (Li et al., 2024, Lines et al., 2024).

4. Spatial distribution of gas, dust, and star formation

The most systematic morphological result is that [CII] emission is more extended than both stellar UV light and, to a lesser degree, FIR dust emission. The dedicated size census finds $\nu_{\rm rest}=1900.537$2 in the range $\nu_{\rm rest}=1900.537$3–$\nu_{\rm rest}=1900.537$4 kpc, with average $\nu_{\rm rest}=1900.537$5 kpc. The mean size ratios are $\nu_{\rm rest}=1900.537$6 and $\nu_{\rm rest}=1900.537$7. Median [CII]-to-UV peak offsets are only $\nu_{\rm rest}=1900.537$8 arcsec, implying co-located centers but more extended [CII] outskirts (Ikeda et al., 2024).

A key negative result is that the achieved data do not require a bright secondary [CII] halo component. Residual visibilities show no statistically significant excess beyond an exponential disk, and stacked short-baseline residuals remain below $\nu_{\rm rest}=1900.537$9. In the stacked fit of single galaxies, a one-component model gives $\sim$0 kpc, while a two-component “galaxy + halo” fit yields a very faint, poorly constrained extended component with flux fraction $\sim$1 and $\sim$2 kpc, with reduced $\sim$3 indistinguishable from the single-disk fit (Ikeda et al., 2024).

The dust-continuum program reaches a complementary conclusion. In 26 galaxies, 19 are individually detected in rest-frame $\sim$4 continuum, with $\sim$5–12.4. Dust effective radii are on average $\sim$6 kpc and are $\sim$7 times larger than the rest-UV. The median infrared surface density is $\sim$8, about an order of magnitude below compact DSFG starbursts, supporting a picture of disk-wide rather than nuclear star formation in typical $\sim$9–6 galaxies (Mitsuhashi et al., 2023).

Resolved SED work further shows that CRISTAL galaxies obey a resolved star-forming main sequence at $z\sim 5$0. Combining all 14 galaxies yields a best-fit slope $z\sim 5$1 with median scatter $z\sim 5$2 dex, and the summed stellar masses and SFRs from spatial bins agree with integrated SED fits, implying no systematic outshining bias in these data. The same study finds that less obscured star formation is located where $z\sim 5$3 and $z\sim 5$4 are highest, typically in central regions, and explicitly speculates that feedback may be driving gas and dust out of those regions (Li et al., 2024).

The program also ties [CII] extent to ISM phase balance. The [CII]-size paper reports a negative correlation between $z\sim 5$5 and $z\sim 5$6 with Spearman $z\sim 5$7 and $z\sim 5$8, a positive correlation between $z\sim 5$9 and $4\le z\le 6$0 with $4\le z\le 6$1 and $4\le z\le 6$2, and a possible negative correlation between $4\le z\le 6$3 and $4\le z\le 6$4. The interpretation advanced there is that inner [CII] primarily traces PDRs directly associated with star formation, while the extra spatial extent likely includes a substantial diffuse neutral atomic-gas contribution (Ikeda et al., 2024).

Resolved ISM-condition work on HZ10 extends this logic to component scales. Multi-band ALMA SED fitting gives a global $4\le z\le 6$5 K, while the more UV-obscured HZ10-W component has $4\le z\le 6$6 K and a lower [CII]/FIR ratio than HZ10-C and the Bridge. HZ10-C follows the local UV-starburst IRX–$4\le z\le 6$7 relation, whereas HZ10-W and the Bridge are better described by dust-screen models with holes in front of a hard UV radiation field (Villanueva et al., 2024).

5. Kinematics, interactions, and circumgalactic structures

CRISTAL’s resolved kinematic analysis provides the first systematic kpc-scale characterization of typical massive main-sequence galaxies at $4\le z\le 6$8. In the 32-galaxy kinematic sample, $4\le z\le 6$9 are classified as Best Disks, $\log(M_\star/M_\odot)\gtrsim 9.5$0 as Disks, and $\log(M_\star/M_\odot)\gtrsim 9.5$1 as Non-Disks, giving a total disk fraction of $\log(M_\star/M_\odot)\gtrsim 9.5$2. The disks have median intrinsic velocity dispersion $\log(M_\star/M_\odot)\gtrsim 9.5$3, typical $\log(M_\star/M_\odot)\gtrsim 9.5$4–$\log(M_\star/M_\odot)\gtrsim 9.5$5, and median $\log(M_\star/M_\odot)\gtrsim 9.5$6, indicating highly turbulent, moderately rotation-supported systems. Roughly half the disks have $\log(M_\star/M_\odot)\gtrsim 9.5$7, and the median enclosed dark-matter fraction is $\log(M_\star/M_\odot)\gtrsim 9.5$8 (Lee et al., 15 Jul 2025).

These disk properties are interpreted within a marginally Toomre-stable, gas-rich framework. Using CRISTAL medians $\log(M_\star/M_\odot)\gtrsim 9.5$9–0.4 and $z\approx 5.1$0, the kinematic paper infers typical $z\approx 5.1$1, with the sample spanning $z\approx 5.1$2. The elevated dispersions are described as consistent with dominant gravity driving, and the weak or absent correlation of $z\approx 5.1$3 with $z\approx 5.1$4 argues against feedback alone as the dominant driver of turbulence across the disk sample (Lee et al., 15 Jul 2025).

A recurrent lesson of the program is that apparently simple systems often resolve into interacting or multi-component structures at higher angular resolution. HZ10 (CRISTAL-22) at $z\approx 5.1$5, studied with ALMA and JWST/NIRSpec, separates into at least three components—HZ10-C, HZ10-E, and HZ10-W—with projected separations of $z\approx 5.1$6 and 4 kpc from the central component. The [CII] morphology and kinematics closely resemble the broad [OIII] $z\approx 5.1$7 component, leading to dynamical scenarios ranging from a disturbed rotating disk plus companions to triple or quadruple merger configurations (Telikova et al., 2024).

CRISTAL-05/HZ3 at $z\approx 5.1$8 provides a second representative example. High-resolution [CII] maps show that a source previously treated as single is instead a close interacting pair, C05-NW and C05-SE, separated by $z\approx 5.1$9 arcsec ($M_\star \approx 10^{10.1}\,M_\odot$0 kpc), and embedded in extended carbon-enriched gas. The global [CII] radius is $M_\star \approx 10^{10.1}\,M_\odot$1, significant [CII] emission extends to $M_\star \approx 10^{10.1}\,M_\odot$2 kpc, and the dominant component is fit by a rotation-supported, baryon-dominated DysmalPy model with only a minor dark-matter contribution within the radial range probed. Spatially resolved [CII]/FIR ratios in the outskirts are consistent with merger-driven shocks or turbulence enhancing [CII] emission (Posses et al., 2024).

CRISTAL also resolves extreme circumgalactic structures. In the J1000+0234 system at $M_\star \approx 10^{10.1}\,M_\odot$3, the program identified an elongated [CII]-emitting plume extending $M_\star \approx 10^{10.1}\,M_\odot$4 kpc from the central DSFG, with no clear counterpart at other wavelengths. Its centroid velocity increases from $M_\star \approx 10^{10.1}\,M_\odot$5 to $M_\star \approx 10^{10.1}\,M_\odot$6 relative to systemic along the plume, while the linewidth narrows from $M_\star \approx 10^{10.1}\,M_\odot$7–600 to $M_\star \approx 10^{10.1}\,M_\odot$8. The paper evaluates four possible origins—conical outflow, cold accretion stream, ram-pressure stripping, and gravitational interactions—and explicitly disfavors ram-pressure stripping while leaving the others viable (Solimano et al., 2024).

Outflows are present in the program, but not as a dominant explanation for all extended [CII]. A stacking analysis of 15 disk-like galaxies, after excluding systems with disturbed kinematics or obvious interactions, finds only weak evidence for a broad [CII] component in the composite spectrum, and that evidence is largely driven by CRISTAL-02. Interpreting the modest residuals at $M_\star \approx 10^{10.1}\,M_\odot$9 as an outflow yields $4Birkin et al., 24 Apr 2025).

6. Program significance, limitations, and evolving legacy

At the program level, CRISTAL has established a quantitative picture of the cold ISM in typical galaxies during the first billion years. Its core empirical synthesis is that [CII] in these systems is usually well described by a single extended exponential disk rather than by a statistically required bright halo, that [CII] is systematically larger than rest-UV and moderately larger than FIR continuum, and that a substantial fraction of the emission likely arises outside the main star-forming radius in diffuse neutral gas. In parallel, the kinematic program shows that many $4Ikeda et al., 2024, Lee et al., 15 Jul 2025).

The survey also demonstrates the value of strict multi-wavelength co-registration. The ALPINE-CRISTAL-JWST program combines CRISTAL ALMA maps with JWST/NIRSpec IFU and JWST/HST imaging for 18 representative galaxies at $4NII], [SII], stellar continuum, [CII], and dust on $4Faisst et al., 17 Oct 2025).

Several limitations are emphasized across the literature. Angular resolution remains finite, with many analyses operating at $4uncertainty in $4Ikeda et al., 2024, Mitsuhashi et al., 2023, Herrera-Camus et al., 9 May 2025).

Future work follows directly from these limitations. Program papers explicitly call for 3D analyses of Ly$4anchor of a wider observational program linking ALMA, HST, JWST, and IFU spectroscopy into a resolved benchmark for early galaxy assembly (Herrera-Camus et al., 9 May 2025, Faisst et al., 17 Oct 2025).