Mapping dusty galaxy growth at z > 5 with FRESCO: detection of Hα in submm galaxy HDF850.1 and the surrounding overdense structures

ABSTRACT

We report the detection of a 13|$\sigma$| H|$\alpha$| emission line from HDF850.1 at |$z=5.188\pm 0.001$| using the FRESCO (First Reionization Era Spectroscopically Complete Observations) NIRCam F444W grism observations. Detection of H|$\alpha$| in HDF850.1 is noteworthy, given its high far-infrared (IR) luminosity, substantial dust obscuration, and the historical challenges in deriving its redshift. HDF850.1 shows a clear detection in the F444W imaging data, distributed between a northern and southern component, mirroring that seen in [C ii] from the Plateau de Bure Interferometer. Modelling the spectral energy distribution of each component separately, we find that the northern component has a higher mass, star formation rate (SFR), and dust extinction than the southern component. The observed H|$\alpha$| emission appears to arise entirely from the less-obscured southern component and shows a similar |$\Delta$|v|$\sim$| + 130 |${\rm{km\,s}}^{-1}$| velocity offset to that seen for [C ii] relative to the source systemic redshift. Leveraging H|$\alpha$|-derived redshifts from FRESCO observations, we find that HDF850.1 is forming in one of the richest environments identified to date at |$z\gt 5$|⁠, with 100 |$z=5.17$|–5.20 galaxies distributed across 13 smaller structures and a |$\sim$|(15 cMpc)|$^3$| volume. Based on the evolution of analogous structures in cosmological simulations, the |$z=5.17$|–5.20 structures seem likely to collapse into a single |$\gt 10^{14}$||${\rm M}_{\odot }$| cluster by |$z\sim 0$|⁠. Comparing galaxy properties forming within this overdensity with those outside, we find the masses, SFRs, and |$UV$| luminosities inside the overdensity to be clearly higher. The prominence of H|$\alpha$| line emission from HDF850.1 and other known highly obscured |$z\gt 5$| galaxies illustrates the potential of NIRCam-grism programs to map both the early build-up of IR-luminous galaxies and overdense structures.

1 INTRODUCTION

One of the most important frontiers in extragalactic astronomy has been understanding the formation and evolution of massive galaxies in the early universe. Despite the many significant insights that have been gained into both the build-up of the star formation rates (SFRs) and stellar masses of |$UV$|-bright galaxies from both space and ground-based telescopes (e.g. Madau & Dickinson 2014; Davidzon et al. 2017; Bouwens et al. 2021; Stefanon et al. 2021; Harikane et al. 2022), it is essential we achieve an equally complete census of star formation from obscured galaxies given how significantly obscured star formation contributes to galaxies with masses |$\gt 10^{10}$||${\rm M}_{\odot }$| (e.g. Reddy et al. 2006, 2008; Whitaker et al. 2017).

In addition to the great strides made by the Atacama Large Millimeter/submillimeter Array (ALMA) in this area (e.g. Casey et al. 2021; Smail et al. 2021; Bouwens et al. 2022; Dayal et al. 2022), JWST is further revolutionizing this science area thanks to its sensitive near-infrared (IR) photometric and spectroscopic capabilities to 5|$\mu$|m and beyond, sampling bright Balmer lines like H|$\alpha$| to |$z\sim 7$| (e.g. Helton et al. 2024; Oesch et al. 2023) and Paschen series line like Pa|$\alpha$| and Pa|$\beta$| mostly at intermediate redshifts (e.g. Reddy et al. 2023) and in at least one case at |$z\sim 7$| (Álvarez-Márquez et al. 2023). Through the identification of bright line emission from e.g. H|$\alpha$| (e.g. the candidate |$z=5.58$| dusty galaxy shown in Oesch et al. 2023), one can assemble substantial spectroscopic samples of far-IR luminous, dusty star-forming galaxies in the early universe with JWST. Particularly useful for this endeavor are NIRCam grism observations that allow for a probe of H|$\alpha$| emission over the full morphology of sources over a |$\sim$|9 arcmin|$^2$| NIRCam field, facilitating redshift determinations even when the escape of H|$\alpha$| occurs over just an isolated region.

In this paper, we report on the successful detection of H|$\alpha$| emission from HDF850.1, one of the first submillimetre galaxies to be identified in the high-redshift universe, leveraging new grism observations from the First Reionization Era Spectroscopically Complete Observations (FRESCO: Oesch et al. 2023) program. HDF850.1 was initially identified using sensitive submillimetre observations over the Hubble Deep Field (HDF) North with the SCUBA (Submillimetre Common-User Bolometer Array STScI) camera (Hughes et al. 1998), but no optical counterpart could be identified in the deepest Hubble Space Telescope (HST) images available at the time. Based on the available information, HDF 850.1 appeared to have a redshift z > 2, implying that the SFR of this single source could rival the total star formation activity of all coeval galaxies in the HDF combined. This discovery triggered a >10-yr effort to improve the location of the source from the coarse position provided by the 17 arcsec full width at half-maximum (FWHM) SCUBA beam, to search for possible counterparts, and to pin down its redshift (Downes et al. 1999; Richards 1999; Dunlop et al. 2004; Wagg et al. 2007; Cowie et al. 2009).

A redshift for HDF 850.1 was finally obtained through a millimetre line scan obtained at the Institut de radioastronomie millimétrique (IRAM) Plateau de Bure Interferometer (PdBI) by Walter et al. (2012) who detected multiple lines in the rest-frame submillimetre ([C ii] as well as multiple CO lines), thus unambigously pinpointing the redshift to z = 5.183. This study also revealed that the system is part of a fairly significant galaxy overdensity at these redshifts. Higher resolution imaging of the [C ii] line emission from HDF 850.1 (Neri et al. 2014) provided evidence that HDF 850.1 is a merging system.

In addition to enabling the detection of the H|$\alpha$| line from HDF850.1, the new NIRCam imaging observations from FRESCO allow us to measure the continuum flux from HDF850.1 in the rest-optical while facilitating a modelling of the overall spectral energy distribution (SED) of both components of the source. Moreover, the NIRCam grism observations allow for the identification of a large number of H|$\alpha$| emitters over a |$\sim$|62 arcmin|$^2$| FRESCO area in the GOODS North field, allowing us to map out the |$z=5.17$|–5.20 overdensity much more extensively than had been possible in earlier work by Walter et al. (2012), Arrabal Haro et al. (2018), and Calvi et al. (2021).

We organize the paper as follows. In Section 2, we provide a summary of the observational data we utilize for this analysis and describe our procedures for constructing multiwavelength catalogues. In Section 3, we describe our discovery of a detection of H|$\alpha$| from HDF850.1, the photometry we derive for the two components of HDF850.1 and inferences based on SED modelling, and then present information on both the structure of the |$z=5.17$|–5.20 overdensity in which HDF850.1 resides, and on the characteristics of the large number of other member galaxies. In Section 4, we include a short summary of our primary results and provide a brief look to the future. For consistency with previous work, we express all quantities using the so-called concordance cosmology with |$\Omega _m = 0.3$|⁠, |$\Omega _{\Lambda } = 0.7$|⁠, and |$H_0 = 70$||${\rm {km\, s}}^{-1}\, {\rm {Mpc}}^{-1}$|⁠. Stellar masses and SFRs are presented assuming a Chabrier (2003) initial mass function (IMF). All magnitudes are presented using the AB magnitude system (Oke & Gunn 1983).

2 DATA

2.1 FRESCO NIRCam grism and imaging data

In this analysis, we make use of NIRCam data obtained by the JWST FRESCO survey (GO-1895; see Oesch et al. 2023 for details). FRESCO covered both the GOODS-South and the GOODS-North fields with |$\sim$|62 arcmin|$^2$| of NIRCam/grism spectroscopy in the F444W filter. This coverage is achieved with two 2|$\times$|4 mosaics of NIRCam/grism observations with significant column overlap in order to maximize the wavelength coverage over the field. Only the GRISMR was used due to overhead costs. The maximal wavelength coverage is from 3.8 to 5.0 |$\mu$|m, which is achieved over 73 per cent of the full mosaic. The exposure times per pointing are 7043 s, resulting in an average 5|$\sigma$| line sensitivity of |$\sim 2\times 10^{-18}$| erg s|$^{-1}$| cm|$^{-2}$| at a resolution of R|$\sim$| 1600. The grism data of GOODS-N used here were obtained in 2023 February.

The NIRCam/grism observations are reduced using the publicly available grizli code¹ (see also Brammer et al., in preparation). Specifically, we start from the rate files that are obtained from the Mikulski Archive for Space Telescopes archive. These are then aligned to a Gaia-matched reference frame. The direct images of a given visit are used to align the associated grism exposures. Before combination of the long-wavelength data, a custom bad-pixel map is applied, which masks residual bad pixels.

Following Kashino et al. (2023), we use a median filtering technique to remove the continuum spectra for each row of the individual grism exposures. The filter uses a 12 pixel central gap, which avoids self-subtraction in case of strong emission lines. After the first pass filtering, pixels with significant line flux are identified and masked, before running the median filtering again.

For each source of interest from the photometric catalogue (see the next section), we derive an extraction kernel based on the image morphology in the F444W band and the segmentation map. This kernel is used to perform an optimal extraction of 1D spectra from the aligned and combined grism exposures. We use slightly modified sensitivity curves and spectral traces based on the v4 grism configuration files.²

In addition to the grism observations, FRESCO obtained imaging in the two short-wavelength filters F182M and F210M, as well as direct imaging in the long-wavelength filter F444W. The 5|$\sigma$| depths of these images are, respectively, 28.3, 28.1, and 28.2 mag, as measured in circular apertures of 0.32 arcsec diameter.

We note that deeper NIRCam imaging data are available over |$\sim$|50 per cent of the FRESCO mosaics thanks to observations from the JWST Advanced Deep Extragalactic Survey (JADES) team (Robertson 2022). However, these data are not yet publicly available for inclusion in this analysis.

2.2 FRESCO multiwavelength catalogues

In addition to the new JWST NIRCam data, we also make use of all ancillary HST imaging available in the GOODS-North field. Being a key extragalactic field for more than a decade, GOODS North has been targeted by all the main telescope facilities through a large number of programs. A complete listing of HST programs can be found on the Hubble Legacy Field (HLF) release page³ (see also Illingworth et al. 2016; Whitaker et al. 2019). Most importantly, the field was covered with ACS imaging from GOODS (Giavalisco et al. 2004), with ACS and WFC3/IR imaging by the CANDELS survey (Grogin et al. 2011; Koekemoer et al. 2011), as well as WFC3/IR grism imaging by the AGHAST survey (Weiner & AGHAST Team 2014).

Here, we use a re-reduction of all available HST ACS and WFC3/IR data in the archive in filters that cover the FRESCO pointings, which we drizzled to a common pixel frame of 40 mas pixel⁻¹ as the JWST NIRCam data. The 5|$\sigma$| depth (in the same 0.32 arcsec diameter apertures) for the ancillary data are well matched to the FRESCO NIRCam imaging: they range from |$\sim$|28.6 mag for the ACS data (F435W, F606W, F775W, F814W, and F850LP) to |$\sim$|28.2 mag for the WFC3/IR data (F105W, F125W, and F160W), with the exception of F140W that reaches to |$\sim 27.4$| mag.

In total, we derive photometry in 12 bands for the GOODS-North data set. The F444W images are used as a detection image, and we run SExtractor (Bertin & Arnouts 1996) in dual-image mode to measure matched-aperture photometry. The images are all point spread function (PSF) matched to the F444W detection image, and fluxes are corrected to total using the default SExtractorauto parameters in addition to a small correction for remaining flux outside the auto aperture based on WebbPSF’s F444W curve of growth (Perrin et al. 2014).

2.3 Spectroscopic sample of H|$\alpha$| emitters

We briefly summarize the construction of the H|$\alpha$| catalogue at |$z\approx 4.9-6.6$| over the GOODS-North FRESCO field and refer readers to Brammer et al. (in preparation) and Naidu et al. (in preparation) for more detailed description of the overall methods and catalogue validation. Line extractions from the grism data are based on eazy photometric redshifts (⁠|$z_{\rm {phot}}$|⁠). For each object, we search for lines in a window around |$z_{\rm {phot}}\pm 0.05\times (1+z_{\rm {phot}})$|⁠, corresponding to the mean 1|$\sigma$| width of the redshift-likelihood distribution function of all H|$\alpha$| emitters in our sample (after 3|$\sigma$| clipping). The photometric redshifts are derived from the PSF-matched photometry described above (Section 2.2). Additionally, we also use the default grizli photometric redshifts catalogue that uses a stack of the FRESCO imaging (F182M + F210M + F444W) as a detection image, and makes two different choices in performing the photometry (e.g. with different source deblending parameters). We use both sets of photometric redshifts to marginalize over such choices. In practice, the best-fitting redshifts at |$z_{\rm {phot}}\gt 4.5$| have a small median difference of |$|\Delta _{\rm {z}}|\lt 0.1$| and in most cases result in the same line candidates, but a fraction (⁠|$\approx 25~{{\ \rm per\ cent}}$|⁠) have |$|\Delta _{\rm {z}}|\gt 0.3$|⁠.

Every line extracted with S/N |$\gt 5$| is visually inspected to verify that its morphology is consistent with the source and also to look for false positives (e.g. from broad polycyclic aromatic hydrocarbon features in the foreground that slip through the median filtering). Physical properties (e.g. stellar masses) are derived by jointly fitting the photometry and line fluxes with a non-parametric star formation history (SFH) and Chabrier IMF using the prospector SED-fitting code (Leja et al. 2017). We refer readers to Naidu et al. (in preparation) for further details about the modelling choices. To assess the robustness of H|$\alpha$|-emitters identified as part of our selection, we compare our catalogue to those from the independent analyses made in Sun et al. (2024) and also Covelo-Paz et al. (2024). Over the same redshift range as considered here and only considering detections with |$\text{SNR}_{\text{H}_\alpha }\gt 5$|⁠, Sun et al. (2024) find 100 sources where we find 75, conducting a search over those areas of FRESCO that overlap with the JADES photometric data set in GOODS North. Those numbers include six multicomponent sources whose clumps are counted as different sources in one of the two catalogues. We investigated the 25 H|$\alpha$| emitters from the Sun et al. (2024) catalogue not found in our own catalogues. Among these 25, 14 are in the general FRESCO catalogue but were either associated with different lines than H|$\alpha$| (three sources) or which were flagged due to heavily contaminated spectra from other nearby galaxies (11 sources). As such, we can conclude that most of the differences that occur between our catalogues result from the differing robustness criteria we apply. Comparing our H|$\alpha$| selection with Covelo-Paz et al. (2024), we find |$\sim$|93 per cent overlap with their selection, providing further evidence for the reliability of our H|$\alpha$| catalogue.

Regarding completeness, we assume the results of Covelo-Paz et al. (2024) largely apply to our own selection given the >90 per cent overlap between our catalogues and similar line identification procedures. Covelo-Paz et al. (2024) estimate the total completeness for their selection by multiplying together the completeness function determined from four separate steps (1) photometric detection, (2) extraction, (3) line detection, and (4) visual inspection (see section 3 of Covelo-Paz et al. 2024 for details). The completeness of H|$\alpha$|-emitter selections over FRESCO is >70 per cent complete for H|$\alpha$| fluxes in excess of |$3\times 10^{-18}$||${\rm{erg\,s}}^{-1}\, {\rm{cm}}^{-2}$| and detection magnitudes of |$\lt 27$| mag. As 60 per cent of our catalogue satisfy both of those conditions, our catalogue should include the majority (>70 per cent) of H|$\alpha$| emitting galaxies to 27 mag.

3 RESULTS

3.1 Identification of H|$\alpha$| emission from HDF850.1

One of the first sources we examined after creating catalogues of line emitting sources over the GOODS-North FRESCO field was the well-known dusty star-forming galaxy HDF850.1 with a spectroscopic redshift |$z=5.1853$| from the 157.74 |$\mu$|m [C ii] line (Walter et al. 2012; Neri et al. 2014). As we discussed in Introduction, that source evaded a direct redshift determination for >10 yr following its initial discovery (e.g. Wagg et al. 2007), and a redshift only came in 2012 thanks to a dedicated spectral scan for [C ii] and various CO lines with PdBI (Walter et al. 2012).

The JWST F182M, F210M, and F444W NIRCam imaging we have available for HDF850.1 from FRESCO is shown in the centre and right panels of Fig. 1 together with the blueshifted and redshifted high spatial resolution [C ii] contours from PdBI. Also shown is the imaging data of HDF850.1 at 0.8 |$\mu$|m from F814W, again demonstrating that HDF850.1 radiates essentially no flux at |$\lt 1\mu$|m. From this imaging data, it is clear that while both components are detected in the NIRCam F444W data, only the southern component shows a clear detection in the F182M and F210M imaging data.

We made use of the two prominent components of HDF850.1 in the F444W imaging data to construct segmentation maps, shown in the central and right panels of Fig. 1 as red and blue dashed lines. We then used grizli to extract spectra of each component. In Fig. 2, we show both a 2D and 1D spectral extraction for the southern component to HDF850.1. In the lower panel to Fig. 2, we present a collapsed 1D spectrum which not only reveals a |$13\sigma$| detection of the H|$\alpha$| line, but shows the detection of [N ii]|$_{6583}$| at |$4 \sigma$| from HDF850.1. No significant detection of [N ii]|$_{6548}$| is apparent in the FRESCO grism spectra, but this is not surprising given the fact that [N ii]|$_{6548}$| is |$\approx 3\times$| fainter than [N ii]|$_{6583}$| (e.g. Dojčinović, Kovačević-Dojčinović & Popović 2023).

For H|$\alpha$|⁠, we measure a total flux of |$(6.4\pm 0.5) \times 10^{-18}$||${\rm{erg\,s}}^{-1}\, {\rm{cm}}^{-2}$| for the southern component to HDF850.1 and derive a redshift of 5.188|$\pm$|0.001. The [N ii]/H|$\alpha$| ratio we measure for the southern component to HDF850.1 is |$0.4 \pm 0.1$|⁠. Such a high [N ii]/H|$\alpha$| ratio is commonly found for galaxies at higher masses with solar metallicities (e.g. Shapley et al. 2015). Shocks could also be a contributing factor to the high [N ii]/H|$\alpha$| ratio we find (e.g. Kewley et al. 2013; Freeman et al. 2019) – which would not be especially surprising given the apparent merging activity in HDF850.1 based on its two component structure (Neri et al. 2014).

The redshift measurement we derived from the H|$\alpha$| line implies a velocity offset of 133|$\pm$|34 |${\rm{km\,s}}^{-1}$| relative the systemic redshift measurement |$z=5.1853$| derived by Neri et al. (2014). Neri et al. (2014) find a velocity offset of 130 |${\rm{km\,s}}^{-1}$| for the [C ii] line from the southern component to HDF850.1, which is almost exactly the same velocity offset in [C ii] for the southern component as we find here. A detailed comparison of the line profiles and intensity of our new H|$\alpha$| detection is shown in Fig. 3 for the southern component to HDF850.1, both in terms of the raw extraction (grey histogram) and Gaussian fits to the lines (yellow lines). Fig. 3 also shows the line profiles for the northern and southern components to HDF850.1 as found by Neri et al. (2014). To further emphasize this similarity, a vertical dotted line is shown at the velocity offset for the southern component found by Neri et al. (2014). From Fig. 4, it is furthermore clear that H|$\alpha$| emission we detect (white contours) appears to mostly originate from the southern component of HDF850.1, being offset from both the [C ii] and continuum IR emission lying to the north (blue and orange contours, respectively).

Additionally, the H|$\alpha$| line we extract from the southern component of HDF850.1 appears to be very broad overall. Ignoring for the moment the impact spatial extension has on the width of the lines, the best-fitting FWHM we find for the southern component is of 7.6|$\times 10^2$||${\rm{km\,s}}^{-1}$|⁠. To account for the impact that source morphology has in broadening the line, we make use of grizli to forward model the source based on its direct image morphology using the same dispersion direction as in the observations. Based on this forward modelling, we find that the non-zero size of HDF850.1 contributes 6.2|$\times 10^2$||${\rm{km\,s}}^{-1}$| to the measured FWHMs. Subtracting this contribution in quadrature from the southern component, we derive an FWHM of (4.4 |$\pm$| 0.9) |$\times$| 10|$^2$||${\rm{km\,s}}^{-1}$|⁠. The width of this line is therefore consistent with what Neri et al. (2014) derive for [C ii] from the southern component. Given similar velocity offsets for both lines (130 |${\rm{km\,s}}^{-1}$| for [C ii] versus (1.3|$\pm$|0.3) |$\times$||$10^2$||${\rm{km\,s}}^{-1}$| for H|$\alpha$|⁠), it seems clear that interstellar medium (ISM) material producing both lines show a very similar kinematic structure.

In contrast to the southern component, the northern component of HDF850.1 does not show a clear, localized detection of H|$\alpha$| line emission. As a result and due to the proximity of the two components and dispersion direction (shown in Fig. 1), line emission from the southern component partially extends into the same spatial region where H|$\alpha$| line flux measurements need to be made for the northern component. As a result of these challenges, we only report an upper limit to the H|$\alpha$| flux for the northern component of |$f_{\text{H}\alpha } \lt 2.1 \times 10^{-18}$||${\rm{erg\,s}}^{-1}\, {\rm{cm}}^{-2}$| (3|$\sigma$|⁠). We derived this upper limit by comparing the observations with the spatial profile expected for H|$\alpha$| emission from the northern component assuming a similar spatial distribution to the continuum light in the direct image. By computing the 2D least-squares residuals between the expected and observed line morphologies, we concluded that there is no significant H|$\alpha$| line flux emanating from the northern component, and any line flux evident in the segmentation map for the northern component is consistent with contamination from the southern component. The flux measurements for the two components are provided in Table 1.

3.2 Impact of dust obscuration on the H|$\alpha$| line emission

Thanks to our new measurements of the H|$\alpha$| fluxes for both components of HDF850.1, we can compute an observed SFR for each component. By comparing this SFR to the SFR implied by the respective [C ii] luminosities, we can attempt to estimate the approximate dust obscuration of each component. We use the conversion factor from Kennicutt & Evans (2012):⁴

The SFR we estimate from the observed H|$\alpha$| flux for the southern component is 6.0|$\pm$|0.5 (1.5/|$\mu$|⁠) M|$_{\odot }$| yr⁻¹, while for the northern component the SFR we estimate is <1.8 (1.7/|$\mu$|⁠) M|$_{\odot }$| yr⁻¹. In specifying the SFR for each source, we have divided the result by the fiducial magnification factors for the two components derived in Neri et al. (2014) based on the isothermal model they constructed for a nearby |$z=1.224$| elliptical galaxy, i.e. 1.7 for the northern component and 1.5 for the southern component. Our quoted results in Table 1 include the factors (1.7/|$\mu$|⁠) and (1.5/|$\mu$|⁠) for clarity and to allow for a scaling of the results in case of updated lensing magnification factors.

We can estimate the approximate dust extinction for each component by relying on multiple total SFR tracers, we use the [C ii] luminosities for HDF850.1 from Neri et al. (2014) as well as the inferred SFRs and IR luminosity from the bagpipes and magphys modelling. The ALPINE program (Béthermin et al. 2020; Faisst et al. 2020; Le Fèvre et al. 2020) provided the following calibration of the |$L_{{\rm [CII]}}$|-SFR relation (Schaerer et al. 2020):

We infer [C ii] SFRs of (1.6 |$\pm$| 0.5)|$\times 10^2$||$(1.7/\mu)$| M|$_{\odot }$| yr⁻¹ and (0.9 |$\pm$| 0.3)|$\times 10^2$||$(1.5/\mu)$| M|$_{\odot }$| yr⁻¹ for the northern and southern components to HDF850.1.

Dividing the apparent H|$\alpha$| SFRs by the [C ii] SFRs, we infer SFR|$_{\rm H\alpha }$|/SFR|$_{\rm [CII]}$| ratios of <0.01 and 0.07|$\pm$|0.03, respectively. However, given the uncertain reliability of this tracer due to the infamous [C ii] deficit (Díaz-Santos et al. 2013). We also consider the total IR luminosity following the method in Kennicutt & Evans (2012):

If we rely on the total IR luminosities we infer for the two components from our magphys SED modelling (Section 3.3.2), we infer SFRs of |$412_{-210}^{+449}$||$(1.7/\mu)$| and |$354_{-276}^{+457}$||$(1.5/\mu)$| M|$_{\odot }$| yr⁻¹. This leads to SFR|$_{\rm H\alpha }$|/SFR|$_{\rm TIR}$| ratios of <0.005 and |$0.02_{-0.02}^{+0.01}$|⁠, respectively. Alternatively, if we rely on those from our bagpipes SED modelling (Section 3.3.2), we obtain |$119_{-49}^{+31}$||$(1.7/\mu)$| and |$90_{-25}^{+22}$||$(1.5/\mu)$| M|$_{\odot }$| yr⁻¹. This leads to SFR|$_{\rm H\alpha }$|/SFR|$_{\rm TIR}$| ratios of <0.02 and 0.07|$\pm$|0.02, respectively, for the two components. Similar SFR fractions are also obtained utilizing the SFRs we infer from bagpipes SED modelling (Section 3.3.2) and by making use of the inferred |$A_V$| (⁠|$\sim$|3.5–4.7 mag) from the SED modelling (Table 1).

Collectively, these results demonstrate that even with a clear detection of H|$\alpha$| from HDF850.1 dust attenuation has a substantial impact on the observed line emission.

3.3 UV+optical + far-IR SED model of HDF850.1

3.3.1 Photometry on the individual components of HDF850.1

Because of the very extended wings to the profile from a bright nearby elliptical galaxy, obtaining accurate flux measurements for HDF850.1 can be challenging to obtain using simple aperture photometry and therefore we elected to measure the flux for HDF850.1 by modelling the light with analytic Sérsic profiles using galfit (Peng et al. 2002). This provides a very effective way of coping with contamination from the bright elliptical spilling over onto our aperture measurements.

We begin by using galfit to model the light in the NIRCam F444W band where both the northern and southern components of HDF850.1 are clearly detected, while also fitting to light in the nearby elliptical galaxy. We model the light in the southern component of HDF850.1 with a single Sérsic profile and the light from the northern component with 1, 2, and 3 profiles (fixing the centres of each component to be cospatial). To account for the uncertainties associated with the extended wings of the nearby elliptical galaxy and the star-forming arc westward to HDF850.1, we mask HDF850.1 and fit the light profile of the two prominent foreground galaxies with Sérsic profiles. We then fix the light profiles of the two foreground galaxies, unmask the HDF850.1 region, and fit the two components of HDF850.1 with Sérsic profiles. To estimate uncertainties on the flux of HDF850.1, we repeat these fits 12 (⁠|$=2\times 2\times 3$|⁠) different times, alternatively modelling the flux of the foreground lensing galaxy (to the east) with one and two Sérsic cocentric profiles, the star-forming lensed arc (to the west) with one and two Sérsic profiles, the northern component with one, two, and three Sérsic profiles, and the southern component with just one Sérsic profile. We take the flux to be equal to the mean derived from these fits. For each galaxy and fit, we fix the center to that inferred from the F444W NIRCam imaging data.

We present the flux measurements we obtained for the two components of HDF850.1 in Table 2, the values represent the mean of the fits in each band and the errors are taken as their standard deviation. We measure an F444W-band magnitude of 24.2|$\pm$|0.1 mag for the northern component to HDF850.1 and 25.7|$\pm$|0.2 mag for the southern component. We remark that the F444W flux we measure for the southern component is |$\sim 4\times$| what we would expect accounting for the line emission from H|$\alpha$| alone, suggesting that stellar continuum from HDF850.1 contributes substantially to the F444W flux that we measure for the southern component. For the northern component, the contribution from the stellar continuum seems to dominate the F444W flux given the absence of a clear line detection at its location.

3.3.2 SED modelling of UV + optical and far-IR light from HDF850.1

The JWST NIRcam observations presented here provide us with the first direct constraints on the stellar content of HDF850.1. We complement the photometry listed in Table 2 with the flux limits in Herschel bands, and the 1 mm flux density continuum reported for the two components of the system in Neri et al. (2014). We adopt the upper limits in Herschel bands from Walter et al. (2012) and perform our own analysis of the flux in the SPIRE bands using the mophongo software (e.g. Labbé et al. 2006, 2010). Based on an attempted deblending of the low-resolution data at 250 |$\mu$|m with mophongo, we measure no significant flux from HDF850.1, and hence adopt the |$2 \sigma$| upper limit from Liu et al. (2018) of 3.14 mJy. While there is a |$\sim 5\sigma$| detection in the Herschel data at 350 and 500 |$\mu$|m, respectively, the low resolution makes it impossible to know the fraction of the flux that should be ascribed to each component of HDF850.1 or to a neighbour. Therefore, we use these values with asymmetric error bars such that the lower bound goes to 0, that is to say |$11.6^{+2.2}_{-11.6}$| mJy and |$11.3^{+2.8}_{-11.3}$|⁠, respectively. We note that the uncertainties we derive are consistent with that found in both Elbaz et al. (2011) and Liu et al. (2018). Additionally, as we are fitting the two components individually, we make the assumption that each component is responsible for 50 per cent of the flux in those bands and accordingly adjust the flux densities. We also make use of the 1 mm dust detections from Neri et al. (2014), namely 2.2 and 2.4 mJy for the northern and southern components, respectively, along with their 20 per cent uncertainties.

We model the SED of the northern and southern components of HDF850.1 using two different SED-fitting codes. This was done to better assess the large uncertainties expected from the small number of detections associated with this source. We make use of bagpipes (Carnall et al. 2018) as well as the high-redshift implementation of magphys (da Cunha, Charlot & Elbaz 2008; da Cunha et al. 2015). They both use the stellar populations based on Bruzual & Charlot’s (2003) spectral libraries both assume energy balance between energy absorbed by dust in the rest-frame optical range and re-emitted in the far-IR. We fix the redshift at the H|$\alpha$| value and assume a delayed exponential function as the SFH. The best-fitting models are shown in Fig. 5. We infer the best fit of the free parameters and their uncertainties from the distribution of the posterior probabilities, interpolated at the 16 per cent, 50 per cent, and 84 per cent levels. Our best-fitting stellar masses, SFRs, far-IR luminosities, and dust attenuation for the two components of HDF850.1 are provided in Table 1.

The NIRCam observations sample the stellar continuum around the Balmer break, which lies at 2.1|$\mu$|m. Our fits suggest that both components of HDF850.1 are significantly dust reddened. The northern component of HDF850.1 appears slightly more massive and star forming, with a similar specific SFR, compared to the southern component. The precise flux measurements made possible thanks to our new NIRCam photometry pins down the stellar component of the fit; this however, combined with the energy balance assumption in the code, and the large uncertainties associated with the Herschel photometry of HDF850.1 as a whole, leads to the sizeable uncertainties in the dust continuum SED of HDF850.1. A better sampling of the IR continuum emission separately for the two components of HDF850.1 will be necessary to improve our characterization of the overall SED.

3.4 Extended galaxy structures surrounding HDF850.1

3.4.1 Redshift overdensity at |$z\sim 5.2$| and comparison with earlier studies

Given the substantial clustering of other star-forming galaxies expected around massive galaxies like HDF850.1 and earlier results showing a significant overdensity of galaxies at |$z\sim 5.2$| (Walter et al. 2012; Arrabal Haro et al. 2018; Calvi et al. 2021), it is logical to make use of the substantial number of sources with spectroscopic redshifts in the GOODS North field from FRESCO to investigate this matter more extensively.

Using the techniques described in Section 3.4.2, we constructed catalogues of H|$\alpha$| emitters over GOODS North FRESCO field. Fig. 6 shows the number of sources as a function of redshift, and it is clear there is a huge overdensity of sources at |$z=5.16$|–5.20, with 100 sources found in that narrow-redshift interval. Walter et al. (2012) and Calvi et al. (2021) had both previously reported the same overdensity of galaxies at |$z\sim 5.2$|⁠, adding 13 and 6 spectroscopic members, respectively. Analysis of the narrow-band SHARDS observations indicate 44 additional sources whose redshifts are consistent with lying in this overdensity (Arrabal Haro et al. 2018), but the redshift measurements from SHARDS are much less precise, i.e. |$\Delta z\sim 0.07$|⁠, and so much more difficult to tie to distinct individual structures identified here.

Nineteen of the 100 H|$\alpha$| emitters that we identified in the redshift range |$z=5.16$| to 5.20 were previously flagged as probable members of the |$z\sim 5.2$| overdensity by Walter et al. (2012), Arrabal Haro et al. (2018), and Calvi et al. (2021). These earlier studies appear to have been most efficient at identifying those member galaxies in the |$z\sim 5.2$| overdensity with the highest H|$\alpha$| fluxes, likely as a result of these same sources showing more prominent Ly|$\alpha$| emission lines for redshift determinations. Three out of the seven brightest H|$\alpha$| emitters (i.e. 43 per cent) were identified as part of earlier spectroscopic studies, as well as five out of the 14 brightest H|$\alpha$| emitters (i.e. 36 per cent). This compares with just 19 out of 100 sources (i.e. 19 per cent) that appeared in these earlier compilations.

Comparing our redshift measurements to those from Walter et al. (2012) and Calvi et al. (2021), the redshift measurements we derive are |$\Delta z = -0.009\pm 0.011$| lower in the mean than those in the literature. This is consistent with what we would expect for Ly|$\alpha$| velocity offsets of 436|$\pm$|582 |${\rm{km\,s}}^{-1}$|⁠, in the general range of what has been found in many studies at |$z\sim 3$|–8 (e.g. Erb et al. 2014; Tang et al. 2023). Somewhat unexpectedly, only 6 of the 19 sources in the spectroscopic redshift catalogues of Walter et al. (2012) and Calvi et al. (2021) with coverage from FRESCO show up in our own catalogues of H|$\alpha$| emitters. Given the high completeness levels expected for our H|$\alpha$|-emitter searches, this suggests that previous spectroscopic samples were dominated by sources with high Ly|$\alpha$| escape fractions.

There are three sources over the GOODS North field discussed by Walter et al. (2012) and Calvi et al. (2021) which have spectroscopic redshifts suggesting they are part of the prominent |$z=5.17$|–5.20 overdensity discussed here, but which lie outside the |$\sim$|62 arcmin|$^2$| FRESCO mosaic. These include a quasi-stellar object (QSO) at |$z\sim 5.18$| (Barger et al. 2002) and two other galaxies SHARDS20008777 and a source at 12:36:00.0, 62:12:26.1. Nineteen of the plausible 50 sources with redshifts in the range |$z=5.12$|–5.27 from Arrabal Haro et al. (2018) lie outside the FRESCO mosaic. Of the 31 sources from Arrabal Haro et al. (2018) which lie in the redshift range |$z=5.12$|–5.27 and lie within the FRESCO mosaic, 17 are members of our spectroscopic sample of H|$\alpha$| emitters in the redshift range |$z=5.16$| to 5.20.

With 100 spectroscopic members of the overdensity at |$z=5.17$|–5.20, this overdensity is particularly extreme. |$\sim$|28 per cent of the total number of |$z=5.0$|–6.0 H|$\alpha$| emitters found over the HDF-North field lie in a |$\Delta z\sim 0.03$| interval. This exceeds even the 45 galaxies present in |$z\sim 5.4$| overdensity identified over the GOODS-South field (Helton et al. 2024) and also the 20 galaxy and 24 galaxy overdensities identified at |$z\sim 6.19$| and |$\sim 6.33$|⁠, respectively, over the J0100 + 2802 field by Kashino et al. (2023).

Given the much larger number of star-forming galaxies in the immediate vicinity of HDF850.1 than even present over the J0100 + 2802 field, this may suggest the halo masses associated with this overdensity may be even more extreme than the bright |$z\sim 6.33$| QSO J0100 + 2802 that the EIGER program targeted. In particular, the |$z=5.17$|–5.20 overdensity extends over the entire GOODS-North FRESCO field, i.e. a 18 cMpc |$\times$| 18 cMpc area and 15 cMpc along the line of sight, which translates to a 4.8|$\times 10^3$| cMpc|$^3$| comoving volume. Assuming a similar efficiency for line emission across the redshift interval |$z=5.0$| to 6.0, this translates to an amplitude of |$\delta _g + 1 = 8\pm 1$| for the identified |$z=5.17$|–5.20 overdensity.

Interestingly enough, the present overdensity is very similar in comoving size (17 cMpc |$\times$| 20 cMpc |$\times$| 26 cMpc and 17 cMpc |$\times$| 20 cMpc |$\times$| 36 cMpc), volume (6.7|$\times 10^3$| cMpc|$^3$| and 1.2|$\times 10^4$| cMpc|$^3$|⁠), and |$\delta _g$| overdensity factors (4.8|$\pm$|1.8 and 9.5|$\pm$|2.0) to the overdensities at |$z=3.065$| and 3.095, respectively, identified over the SSA22 field by Steidel et al. (1998) and characterized in more detail by Topping et al. (2018).

3.4.2 Extended galaxy structures around HDF850.1

To gain insight into the structures that make up the overdensity at |$z=5.17$|–5.20 and overdensities at |$z=5.20$|–5.32, we constructed a list of candidate extended structures within the FRESCO GOODS-North volume based on our H|$\alpha$|-emitting catalogues. For this, we made use of the dbscan density-based clustering algorithm – which is part of the scikit-learn (Pedregosa et al. 2011) to identify the structures. We set the minimum number of sources to define an overdensity to 3 and the linking length for the algorithm to 3.4 cMpc, which corresponds to 0.5|$\times$| the |$1\sigma$| width (⁠|$\Delta z\sim 0.025$|⁠) of the |$z=5.17$|–5.20 structure along the line of sight. As a result of the impact of a |$\sigma _{v}\sim 700$||${\rm{km\,s}}^{-1}$| line-of-sight velocity for structures whose components sum to a mass |$\gt 10^{13}$||${\rm M}_{\odot }$|⁠, we expect structures to appear |$\sim$|7 cMpc longer along the line of sight. Accordingly, we divide the line-of-sight distances between sources by a factor of 2. For completeness, we also include several additional highly obscured galaxies identified by Xiao et al. (2024) to those identified using procedures from Section 2.3. These include GN10, which was previously identified as part of earlier work (Daddi et al. 2009; Riechers et al. 2020).

Using a linking length of 3.4 cMpc, we identify nine structures from |$z=5.15$| to 5.32 using dbscan. Not surprisingly, the most prominent structure lies at |$z=5.16$|–5.20 and includes all H|$\alpha$|-emitting sources between |$z=5.16$| and 5.20. We also identify eight smaller structures at z |$\sim$| 5.23, 5.27, and 5.30. Given the number of sources in the z |$=$| 5.17–5.20 overdensity, we can also look at possible substructure within this overdensity. Again we do so using the dbscan density-based clustering algorithm but with a linking length that reduced by a factor of 2. With a linking length of 1.7 cMpc, we find 13 small structures embedded within the larger structure. The number of structures recovered with dbscan is sensitive to the linking length. Lengths of 1.0, 1.5, 2.0, and 2.5 cMpc result in the identification of 10, 13, 7, and 5 structures, respectively. Linking lengths smaller than 1.5 cMpc results in a decrease in the number of structures identified.

To help illustrate the 3D spatial distribution of galaxies and extended structures within this |$z=5.17$|–5.20 overdensity, we have included Fig. 7, which shows six smaller structures we have identified at |$z=5.170$|–5.185 (left panel) and seven smaller structures we have identified at |$z=5.185$|–5.200 (centre panel). The rightmost panel shows the redshift distribution of sources in structures 1A–1F (red histogram) and structures 1G–1M (black histogram), pointing to the presence of a possible bimodality in the |$z=5.17$|–5.20 overdensity. Additionally, the structures located in the lower redshift mode lie predominantly to the south and western parts of the FRESCO field, while structures in the higher redshift mode, lie predominantly to the north and east.

We have included a full list of the H|$\alpha$| emitters in the |$z=5.17$|–5.20 overdensity in Table B1 from Appendix  B. Galaxies which lie in some nearby overdensities at |$z\sim 5.22$|⁠, |$\sim 5.27$|⁠, and |$\sim 5.3$| over the GOODS North FRESCO field are also presented in this table. In Appendix  A, we present the spatial distribution in the plane of the sky of these slightly higher redshift overdensities. The spatial extent of the structures in these overdensities range from 1 to 4–5 cMpc in the plane of the sky, much less substantial than we see in the |$z=5.17$|–5.20 overdensity that contains HDF850.1, which is the focus of this manuscript.

We include in Table 3 the central coordinates, mean redshift, nominal velocity dispersion, approximate radius, overdensity factor relative to the mean volume density of H|$\alpha$| emitters in the GOODS North FRESCO field, and number of member galaxies of these extended structures. The centre of each structure is provided both in right ascension and declination and offset from the center of the FRESCO GOODS North field. The median |$1\sigma$| dispersion in redshift for the extended structures we identified is 0.0050, while the median velocity dispersion we find is 240|$_{-21}^{+5}$||${\rm{km\,s}}^{-1}$|⁠. If instead of considering structure 1 as a whole and instead consider the smaller structures that make it up, the median |$1\sigma$| dispersion in redshift for structures is 0.0030, while the median velocity dispersion is 143|$_{-13}^{+24}$||${\rm{km\,s}}^{-1}$|⁠. This is 17 per cent larger and 30 per cent smaller, respectively, than the velocity dispersions Kashino et al. (2023) find, i.e. 212 and 197 |${\rm{km\,s}}^{-1}$|⁠, for the overdensities in the J0100 + 2802 field at |$z\sim 6.19$| and |$\sim 6.33$|⁠, respectively. Not surprisingly, the dispersion in redshift and line-of-sight velocity is sensitive to whether we consider substantial overdensities as single structures or as an ensemble of smaller structures.

Following Long et al. (2020) and Calvi et al. (2021), we also use the stellar masses of the member galaxies to estimate the integrated halo masses of the extended structures that host the galaxies in the extended structures. We do so by computing the total stellar mass we estimate from the member galaxies and then supposing an integrated star formation efficiency of 5 per cent based on detailed studies comparing the galaxy stellar mass functions to the halo mass functions (Behroozi & Silk 2018), but we note that detailed studies (McLeod et al. 2021; Stefanon et al. 2021) of the ratio of the stellar mass to halo mass density from galaxy stellar mass functions at |$z\sim 0$|–10 integrated down to 10|$^8$||${\rm M}_{\odot }$| give factors between 0.5 per cent to 2 per cent. We make use of the masses from Naidu et al. (in preparation) SED fits with prospector – except for the sources in the optically faint samples of Xiao et al. (2024). The derived integrated halo masses range from 10|$^{10.6}$| and 10|$^{12.6}$||${\rm M}_{\odot }$|⁠. Not surprisingly, HDF850.1 and GN10 lie in the halos that belong to the structures with the highest integrated halo masses. Assuming a survey area of |$\sim$|62 arcmin|$^2$| and |$\Delta z \sim 1$| volume (⁠|$\sim 2\times$||$10^{5}$| cMpc|$^3$|⁠), we estimate that the GOODS North FRESCO volume should contain approximately one |$10^{12.2}$||${\rm M}_{\odot }$| halo. For this calculation, we made use of the public halo mass function calculator HMFcalc by Murray, Power & Robotham (2013) adopting a Planck Collaboration VI (2020) cosmology.

The radius of each extended structure |$R_{\rm structure}$| is computed using the standard formula from Heisler, Tremaine & Bahcall (1985) used in computing dynamical masses of galaxy clusters based on a discrete number of galaxies with measured positions and line-of-sight velocities:

where N is the number of galaxies in an extended structure and |$R_{\perp ij}$| is the projected distance in the plane of the sky between galaxy i and galaxy j.

We also estimate overdensities for the extended structure we identified as part of Table 3 by comparing the volume density of H|$\alpha$| emitters within the estimated radius and a |$|\Delta z| \lt 2\sigma _z$| redshift width of an extended structure to that found between |$z=5.0$| to 6.0 for the GOODS-North FRESCO field as a whole. We compute overdensity factors between 3 and 238, far in excess of the linear regime. Based on the volume density of H|$\alpha$| emitters we find in the GOODS North FRESCO volume, i.e. 1.6|$\times 10^{-3}$| cMpc|$^{-3}$|⁠, we can estimate a nominal bias for this population assuming abundance matching and find an average bias factor of |$\sim$|4.7 using the Trenti & Stiavelli (2008) ‘cosmic variance’ calculator. Adjusting for this bias factor, this is suggestive of overdensities in the matter distribution ranging from 1 to 50, which is in excess of |$\approx$|1 for the linear regime but less than |$200\times$| overdensities expected for completely collapsed structures.

It is interesting to ask whether we would expect protoclusters to be present over the FRESCO fields and indeed within the GOODS North FRESCO field. Given that clusters generally have a mass of at least 10|$^{14}$||${\rm M}_{\odot }$| (e.g. Kravtsov & Borgani 2012) at |$z\sim 0$|⁠, we can use abundance matching to estimate whether the FRESCO volume contains any such objects. Using the halo mass functions from HMFcalc, we estimate the FRESCO GOODS-North volume to include four such clusters with mass 10|$^{14.0}$||${\rm M}_{\odot }$| and one cluster reaching a mass of 10|$^{14.4}$||${\rm M}_{\odot }$|⁠.

3.4.3 Comparison of structure sizes to the expected sizes of protoclusters

We can compare the size of the extended structures we find to that expected from the detailed study of protocluster growth (Chiang et al. 2017) making use of several semi-analytic galaxy formation models (Guo et al. 2013; Henriques et al. 2015) applied to the Millennium simulations (Springel et al. 2005). Chiang et al. (2017) present both the comoving sizes of collapsed protoclusters at various redshifts and the sizes of cosmic volumes that will collapse into |$\gt 10^{14}$||${\rm M}_{\odot }$| galaxy clusters by |$z\sim 0$|⁠. Of relevance, Chiang et al. (2017) find protoclusters to have a comoving radius of |$\sim$|9 cMpc at |$z\sim 5.2$|⁠, implying overdense structures to extend over a comoving distance of |$\sim$|18 cMpc, very similar to the spatial extent of the overdensities we have identified over the GOODS North FRESCO field (Fig. 7). Chiang et al. (2017) suggest that the entire |$z=5.17$|–5.20 structure here will likely collapse into a single |$\gt 10^{14}$||${\rm M}_{\odot }$| galaxy cluster by |$z\sim 0$|⁠.

We estimate that the halo masses of the most massive collapsed halo in those extended structures would be 10|$^{12.0}$| and 10|$^{12.2}$||${\rm M}_{\odot }$|⁠, respectively, at |$z\sim 5.2$|⁠. Assuming that average overdensity factor in the collapsed halos is 200, the size of the halos would be 0.3 and 0.4 cMpc at |$z\sim 5.2$|⁠, i.e. 8 to 13 arcsec in the plane of the sky. In computing a radius of the identified structures, the above equation gives greater weight to galaxy pairs that have smaller separations. The fact that many of these extended structures have computed sizes not especially larger than 0.4 cMpc suggests that at least a few of the member galaxies in the extended structures are part of the same haloes.

3.4.4 Characteristics of galaxies in the |$z=5.17$|–5.20 overdensity versus those outside

To put the extreme properties of HDF850.1 in context, we use the SED-fitting code prospector (Leja et al. 2017; Johnson et al. 2021) to derive galaxy properties from our extensive spectroscopically confirmed sample of galaxies in FRESCO GOODS-North. Our prospector fits use FSPS with the MIST stellar models. We do not fit for redshift and set it to the derived |$z_{\text{grism}}$| value infered by our pipeline using grizli. Along with the photometry, fluxes from emission lines identified in the grism 1D spectra are also used as an input. We make use of prospector’s more flexible non-parametric SFH with eight bins evenly separated in lookback time with a ‘continuity’ prior for smoother SFH. The output catalogue contains all sources detected in the detection image that were visually inspected by at least one member of the FRESCO team along with the inferred properties from SED-fitting such as the retained stellar mass |$M_\star$|⁠, UV slope |$\beta$|⁠, star formation rates, SFR or age at 50 per cent of the total stellar mass.

To highlight the distinct properties galaxies may have within the overdensity or its structures and outside of it, we compare two samples. One contains all the galaxies in the |$z = 5.17$|–5.20 overdensity that have been identified as being part of a structure (‘in’ with 40 sources), while the second sample contains all the remaining galaxies from |$z = 4.9$| to 5.5 not identified in such structures (‘out’ with 86) and also excluding the redshift bin |$z = 5.17$|–5.20. We also apply a magnitude cut of 27 AB mag in F182M. While this will limit our inferences to galaxies intrinsically brighter than |$-20$| mag, this choice was made to alleviate the uncertainties on the inferred parameters from faint sources – particularly for stellar masses – and avoid a bias towards strong H|$\alpha$| emitters only. We report the output of the prospector fits for these two samples with kernel density estimate (KDE) plots as shown in Fig. 8. We also compare SFRs inferred from H|$\alpha$| using equation (1). We use the H|$\alpha$| flux corrected for dust attenuation with the empirical relation linking it to |$\beta$| slopes from Shivaei et al. (2020). Finally, we use a two-sample Kolmogorov–Smirnov (KS) test to comment on the dissimilarities in our selection of sources.

Based on the sizes of our two samples, we require a critical D-value of 0.3311 to demonstrate at 99.5 per cent confidence (⁠|$\alpha = 0.005$|⁠) that the distributions of the properties for sources inside and outside overdensities are different. While we do not find a clear differences (at 99.5 per cent confidence) in the distributions for |$\beta$| slopes and ages at 50 per cent of the SFH, the D- and p-values for the masses and ages show clear differences between the two populations of galaxies at this redshift. Our results suggest that the galaxies in the overdensity are more evolved, more massive and with higher SFRs than the galaxies outside of these structures. The median stellar mass |$M_\star$| in the overdensity is |$\sim 2.5 \times 10^{9} {\rm M}_{\odot }$| versus |$\sim 8 \times 10^{8} {\rm M}_{\odot }$| for the galaxies outside, a factor of |$\sim$|3 difference. Earlier work by Steidel et al. (2005) had shown that the masses and ages of galaxies in a similarly prominent (i.e. |$\delta _g \sim 7$|⁠) overdensity at |$z=2.3$| were twice as high as those outside the overdensity in line with our results.

More evidence from FRESCO for the most extreme sources being found in overdense environments can be found in the Xiao et al. (2024) H|$\alpha$| selection of highly obscured sources over the GOODS North field. In particular, the five |$z=4.9$|–6.6 sources with the highest estimated stellar masses and SFRs all lie within the overdensities at |$z\sim 5.16$|–5.20 and |$\sim 5.30$|⁠. We show one of these sources in Fig. 9, a neighbour to HDF850.1 which was given source ID S3. That source has a redshift |$z=5.179$| derived from the detection of both H|$\alpha$| (⁠|$\gt 20\sigma$|⁠) and the [S ii]|$_{6719,6730}$| doublet (⁠|$\sim 9\sigma$|⁠: Xiao et al. 2024). Remarkably, this source shows even more extreme stellar masses and SFRs than HDF850.1 (Xiao et al. 2024). Xiao et al. (2024) present a much larger sample of obscured, H|$\alpha$| emitters at |$z\gt 5$|⁠, leveraging the FRESCO NIRCam grism data to derive spectroscopic redshifts. The existence of large numbers of obscured galaxies with detectable H|$\alpha$| emission may arise because of prevalent merging activity amongst the brightest far-IR sources (e.g. Clements et al. 1996; Tacconi et al. 2008) and the possibility that the merging activity could create less obscured sightlines by which H|$\alpha$| photons could escape from galaxies (e.g. Le Reste et al. 2024). One of the components in a merging system might also be subject to substantially less dust obscuration, which would also enhance the detectability of IR luminous sources.

4 SUMMARY

In this paper, we present the detection of a 13|$\sigma$||$\mathrm{ H}\alpha$| line for HDF850.1 in the NIRCam F444W grism observations over the GOODS North field from FRESCO (Oesch et al. 2023), recovering a redshift of 5.188|$\pm$|0.001. Detection of H|$\alpha$| in HDF850.1 is particularly noteworthy, given how obscured star formation from the source is. HDF850.1 was one of the first submm galaxies to be identified with SCUBA (Hughes et al. 1998) and evaded efforts to pin down its redshift for >10 yr until the PdBI secured the detection of [C ii] and various CO lines from a spectral scan (Walter et al. 2012).

In addition, HDF850.1 is clearly detected in the F444W imaging observations available over the source with NIRCam, with the emission segregated into a distinct northern and southern component. These two distinct components were also evident in the earlier observations of [C ii] from HDF850.1 with PdBI (Neri et al. 2014). Modelling the SEDs of the two components of HDF850.1 based on the available HST + NIRCam F182M, F210M, and F444W imaging observations, we find a much higher SFR, stellar mass, and dust obscuration for the northern component than the southern component.

The majority of the H|$\alpha$| emission for HDF850.1 appears consistent with originating from the southern component, not only due to the spatial localization of the emission but also due to its showing a very similar |$\Delta v$||$\sim$|130 |${\rm{km\,s}}^{-1}$| velocity offset to that seen in the southern component from [C ii] emission (Neri et al. 2014). Comparison of the SFR inferred from the observed H|$\alpha$| emission with that seen from [C ii] is suggestive of the H|$\alpha$| emission from the southern component being 93|$\pm$|3 per cent attenuated and the northern component >98 per cent attenuated.

Leveraging redshift determinations possible from the FRESCO NIRCam grism observations in F444W over the GOODS North field, we note the existence of 100 galaxies in total in the |$z=5.17-5.20$| interval, indicative of a huge 8|$\times$| overdensity of galaxies in that redshift interval relative to the |$z=5.0-6.0$| population of H|$\alpha$| emitters we see over GOODS North. Earlier work by Walter et al. (2012), Arrabal Haro et al. (2018), and Calvi et al. (2021) had previously demonstrated the existence of a substantial overdensity around HDF850.1, but with a much smaller number of spectroscopically confirmed sources than we find as part of this study.

Taking advantage of both the spatial and redshift information we have of galaxies in the |$z=5.17$|–5.20 and slightly higher redshift (⁠|$z=5.22$|–5.31) overdensities, we have used the dbscan density-based clustering algorithm to identify 9 more extended structures using a linking length of 3.4 cMpc and decomposed the largest structure at |$z=5.17$|–5.20 into 13 smaller structures using a smaller linking length (1.7 cMpc). For each structure, we have computed radii, masses, velocity dispersions. The median |$1\sigma$| dispersion found the more extended structures is 0.0050 in redshift and 240|$_{-21}^{+5}$||${\rm{km\,s}}^{-1}$| in |$v_{\mathrm{ los}}$|⁠. If we consider smaller structures within the z |$=$| 5.17–5.20 overdensity, the 1|$\sigma$| dispersion in redshift and in |$v_{\mathrm{ los}}$| becomes 0.0030 and 143|$_{-13}^{+24}$||${\rm{km\,s}}^{-1}$|⁠. This is 17 per cent larger and 30 per cent smaller, respectively, than what Kashino et al. (2023) find for the extended structures in the J0100 + 2802 field at |$z\sim 6.19$| and |$\sim 6.33$|⁠.

Interestingly, the comoving physical size of the extended structures we find around HDF850.1 extend over a 18 cMpc |$\times$| 18 cMpc |$\times$| 15 cMpc volume, very similar to the |$\sim$|18 cMpc physical size, Chiang et al. (2017) expect for protoclusters based on an analysis of various semi-analytic galaxy formation results building on the Millennium simulation (Springel et al. 2005: Section 3.4.3). The link to the |$z=5.17$|–5.20 overdensity being the progenitor to a |$\gt 10^{14}$||${\rm M}_{\odot }$||$z\sim 0$| galaxy cluster is further strengthened by noting that within the FRESCO GOODS North volume we would expect four |$\gt 10^{14}$||${\rm M}_{\odot }$| clusters and one |$\gt 10^{14.4}$||${\rm M}_{\odot }$| cluster to form by |$z\sim 0$|⁠.

Additionally, we have made a systematic comparison of galaxy masses, SFRs, |$UV$| luminosities, ages, and apparent dust extinctions inside the |$z=5.17$|–5.20 overdensity to that outside the |$z=5.17$|–5.20 overdensity and other overdensities in the field. We find strong evidence (⁠|$\gt 3\sigma$|⁠) that galaxies inside the overdensities have higher stellar masses, SFRs, and |$UV$| luminosities than those outside these overdensities. This conclusion is strengthened by the inclusion of optically faint, massive, high-SFR dust-obscured galaxies at |$z=5.0$|–6.0, almost all of which lie inside some overdensity within the GOODS North.

In the future, it should be possible to significantly expand the number of dusty star-forming galaxies identified in the |$z\gt 5$| universe thanks to ongoing NIRCam grism observations from EIGER (Kashino et al. 2023), ASPIRE (Wang et al. 2023), a new NIRCam grism program from MIRI GTO team over GOODS South (program ID 4549), as well as deeper NIRCam grism observations approved over a deep NIRCam parallel field (program ID 4540) and various HFF clusters (program IDs 2883, 3516, and 3538). Of course, to obtain the strongest constraints, a much wider area NIRCam program will be essential, and so it is exciting that in cycle 3, there will be the COSMOS-3D program obtaining grism data over a 1200 arcmin|$^2$| area over the COSMOS field (Kakiichi et al. 2024). This program will be especially useful for further explorations such as the present one leveraging far-IR observations available from the ex-MORA program over the COSMOS field (Casey et al. 2021) and CHAMPS program (2023.1.00180.L, PI: Faisst) covering |$\gtrsim$|50 per cent of the NIRCam grism mosaic.

ACKNOWLEDGEMENTS

We are grateful to Roberto Neri and collaborators for providing us with spatially resolved information on both the dust-continuum and [C ii] line emission from their high spatial resolution PdBI observations. This project was made possible in part by the Leiden University Fund/Bouwens Astrophysics Fund. RJB acknowledges support from NWO grants 600.065.140.11N211 (vrij competitie) and TOP grant TOP1.16.057. The Cosmic Dawn Center (DAWN) is funded by the Danish National Research Foundation under grant no. 140. Cloud-based data processing and file storage for this work is provided by the AWS Cloud Credits for Research program. Support for this work was provided by NASA through grant JWST-GO-01895 awarded by the Space Telescope Science Institute, which is operated by the Association of Universities for Research in Astronomy, Inc., under NASA contract NAS 5–26555. RPN acknowledges funding from JWST programs GO-1933 and GO-2279. Support for this work was provided by NASA through the NASA Hubble Fellowship grant HST-HF2-51515.001-A awarded by the Space Telescope Science Institute, which is operated by the Association of Universities for Research in Astronomy, Incorporated, under NASA contract NAS5-26555. MS acknowledges support from the CIDEGENT/2021/059 grant, from project PID2019-109592GB-I00/AEI/10.13039/501100011033 from the Spanish Ministerio de Ciencia e Innovación – Agencia Estatal de Investigación. This study forms part of the Astrophysics and High Energy Physics programme and was supported by MCIN with funding from European Union NextGenerationEU (PRTR-C17.I1) and by Generalitat Valenciana under the project n. ASFAE/2022/025. RAM acknowledges support from the ERC Advanced Grant 740246 (Cosmic_Gas) and the Swiss National Science Foundation through project grant 200020_207349.

This work is based on observations made with the NASA/ESA/CSA JWST. The data were obtained from the Mikulski Archive for Space Telescopes at the Space Telescope Science Institute, which is operated by the Association of Universities for Research in Astronomy, Inc., under NASA contract NAS 5–03127 for JWST. These observations are associated with program no. 1895.

This paper made use of several publicly available software packages. We are indebted to the respective authors for their work: ipython (Pérez & Granger 2007), matplotlib (Hunter 2007), numpy (Oliphant 2006), scipy (Virtanen et al. 2020), jupyter (Kluyver et al. 2016), astropy (Astropy Collaboration 2013, 2018), grizli (v1.7.11; Brammer 2018; Brammer et al. 2022), eazy (Brammer, van Dokkum & Coppi 2008), and SExtractor (Bertin & Arnouts 1996).

DATA AVAILABILITY

All data used here are from the JWST Cycle 1 (GO 1895) and publicly available from the Barbara A. Mikulski Archive for Space Telescopes (MAST: https://mast.stsci.edu). Data were reduced with the publicly-available code grizli (https://grizli.readthedocs.io). The reduced imaging data is available at https://s3.amazonaws.com/grizli-v2/JwstMosaics/v7/index.html or through MAST: https://archive.stsci.edu/hlsp/fresco.

Footnotes

REFERENCES

APPENDIX A: EXTENDED GALAXY STRUCTURES AT z|$\sim$|5.22–5.31

For completeness and as an illustration of the extended galaxy structures that exist at redshift just above the prominent |$z=5.17$|–5.20 overdensity. Similar to Fig. 7 from the main text, we show in Fig. A1 the spatial distribution of galaxies in the galaxy overdensities at the slightly higher redshifts of |$z\sim 5.23$|⁠, |$\sim 5.27$|⁠, and |$\sim 5.30$| for context. As in the main text, we have segregated sources in the overdensities into different structures using the same approach as described in Section 3.4.2.

While in some cases these overdensities consist of just a single group of |$\sim$|3–10 galaxies lying within 1 cMpc of each other (e.g. structures 9 and 15), in other cases the overdensities extend over larger (⁠|$\sim$|3–5 cMpc) scales. However, even for the most extended structures in these overdensities, galaxies in these overdensities are much less spatially extended than in the prominent |$z=5.17$|–5.20 overdensity that is the focus of this study.

 

APPENDIX B: H|$\alpha$| EMITTERS IDENTIFIED IN THE z|$\sim$|5.15–5.31 OVERDENSITY OVER THE GOODS NORTH FRESCO FIELD

In this appendix, we provide a full catalogue of the star-forming galaxies identified as part of the prominent |$z=5.15$|–5.31 overdensity over the GOODS North FRESCO field. Coordinates for individual sources in the overdensity, together with their redshift determinations, flux measurements in the F210M band, H|$\alpha$| fluxes, stellar masses, SFRs, |$UV$|-continuum slopes, and V-band dust attenuations are provided in Table B1–B5. Also indicated are whether sources were previously reported as part of the Walter et al. (2012), Arrabal Haro et al. (2018), Riechers et al. (2020), Calvi et al. (2021), or Matthee et al. (2024) studies. Given the |$\Delta z \sim 0.07$| uncertainties in the photometric redshifts derived as part of Arrabal Haro et al. (2018), sources in Arrabal Haro et al. (2018) are considered to be likely members of the overdensity at |$z\sim 5.2$| if their photometry redshifts are |$z\gt 5.13$| and |$z\lt 5.27$|⁠. A machine readable format of this table is available as Supporting Information.