SPAM: Survey Overview and JWST/NIRCam Multi-Band Photometry Catalog

="2D

Introduction

Early Universe ($z>4$) discovery with JWST has been driven by the legacy extragalactic deep fields, such as the Cosmic Evolution Early Release Science Survey (CEERS; [29]), the Public Release IMaging for Extragalactic Research (PRIMER), the Ultradeep NIRSpec and NIRCam ObserVations before the Epoch of Reionization (UNCOVER; [3]), and the JWST Advanced Deep Extragalactic Survey (JADES; [81]), among others. Deep near-IR photometry from NIRCam [?] in these fields, combined with extensive HST/ACS and WFC3 photometry from large surveys such as the Great Observatories Origins Deep Survey (GOODS; [80]), the Cosmic Assembly Near-infrared Deep Extragalactic Legacy Survey (CANDELS; [27, ?]), and the All-Wavelength Extended Groth Strip International Survey (AEGIS; [26]), provides photometric coverage from the rest-UV through the rest-optical at $4<z<9$. This coverage enables, among other discovery spaces, searches for early luminous galaxies, probes of star-formation-history timescales, and identification of the earliest forming active galactic nuclei (AGN).

However, many extragalactic fields largely use NIRCam broad-band (BB) filters and often lack the photometric coverage necessary to de-blend emission lines and resolve the shape of the rest-UV to rest-optical continuum at $2<z<9$. One possible avenue to mitigate this is to break up the spectral energy distribution (SED) by adding NIRCam medium-band (MB) filters, which provide higher wavelength resolution when combined with the BB filters. This technique has a long history in imaging predating JWST, including ground-based surveys such as the NEWFIRM Medium-Band Survey (NMBS; [94]), which showed that better-sampled SEDs yield improved photometric redshifts [93], and the Survey for High-z Absorption Red and Dead Sources (SHARDS; [1]) which used 24 MBs to provide an intermediate spectral resolution (R$\sim$50) over the GOODS-N field. The FourStar Galaxy Evolution Survey (ZFOURGE; [95]) similarly provided MB coverage across a wide ($\sim400 \, \rm{arcmin}^2$) field of view with ground-based IR imaging and demonstrated improvement for SED sampling at $1<z<3$.

Such uses of MBs have proved crucial in the JADES field [81], where the JWST Extragalactic Medium-band Survey (JEMS; [14]) covers $\sim40\,\rm{arcmin}^2$ across 5 MBs. In the lensed Canadian NIRISS Unbiased Cluster Survey (CANUCS; [82]), 12 MBs cover $\sim 30 \,\rm{arcmin}^2$, where the addition of MBs has been used to rigorously identify high-z galaxies [84]. MBs have also been utilized in UNCOVER [83], where 11 MBs cover $\sim 30 \,\rm{arcmin}^2$. Such surveys have also shown improvements in SED sampling from other telescopes, such as Spitzer (i.e., the F2 Extragalactic NearIR K-Split Survey (FENIKS) [96]). However, these surveys consistently probe small ($\leq 40 \,\rm{arcmin}^2$) areas of sky and/or rely on coverage from lower-resolution telescopes.

The CEERS survey covers $\sim100 \,\rm{arcmin}^2$ with JWST Cycle 1 coverage including one MB and six BB JWST/NIRCam photometric filters: F115W, F150W, F200W, F277W, F356W, F410M, and F444W [29]. This limited-resolution coverage introduces degeneracies, especially when emission lines are photometrically blended, which can produce spuriously high-redshift solutions in galaxies with emission lines in the BB filters [35, 30]. NIRSpec observations effectively break degeneracies between source types, but they require careful targeting of individual sources, are time-expensive, and are incomplete over large survey areas. Such spectroscopic observations can constrain a galaxy’s properties, such as redshift or star formation history (SFH), but we can recover them only for a small subset of imaged galaxies. Improved SED sampling, enabled by the increased resolution of MBs, enables this science on a larger scale within a single imaging footprint.

Adding MBs is therefore essential to expand this discovery space in the early universe within legacy fields. In this work, we present the Star Formation from the Addition of Medium Bands (SPAM, GO#8559, PIs: Davis & Larson) survey, which adds ten new filters, nine MB filters and one BB filter, to the CEERS legacy footprint at depths comparable to the legacy CEERS coverage. The SPAM program is presented here alongside photometry from a complementary program in the same JWST cycle over the CEERS footprint: Medium-band Imaging with NIRCam to Explore ReVolutionary Astrophysics (MINERVA, GO#7814, PIs: Muzzin, Marchesini, & Suess; [62]). Together, the programs add all remaining BB and MB filters from NIRCam over a total of $\sim 100 \, \rm{arcmin}^2$. The CEERS footprint now represents the legacy extragalactic field with the most complete SED sampling over the largest sky area, enabling a wide variety of science.

In this paper, we present the SPAM MB survey, provide reduced imaging mosaics for all MB and BB filters in the CEERS field, which includes a complete photometric catalog, alongside photometric redshifts for every source, and describe the program’s key science cases. In § 2, we describe the survey, present the imaging data-reduction process, and produce a photometric catalog inclusive of the SPAM + MINERVA dataset, which we make publicly available here to enable community science. In § 3, we present the photometric redshift catalog which accompanies the photometric catalog and compare it to $\sim4500$ sources in CEERS with spectroscopic redshifts to assess improvement. In § 4, we show early results to demonstrate the improvement the SPAM + MINERVA dataset brings to (1) quantifying overdensities, (2) identifying emission-line galaxies, and (3) improving star-formation histories (SFHs). We summarize our conclusions in § 5.

Data

The SPAM photometric dataset follows the CEERS (ERS-#1345; PI: Finkelstein; [29]) survey’s legacy, adding nine new MB filters and one BB photometric filter: F070W, F140M, F162M, F182M, F210M, F300M, F335M, F360M, F430M, and F480M. This unprecedented coverage, combined with the two additional MB filters from MINERVA [62]: F250M and F460M, gives the CEERS legacy footprint coverage across all broad-band and medium-band photometric filters available on NIRCam. MINERVA also adds additional depth to four blue MB filters: F140M, F162M, F182M, and F210M.

We obtained these data with JWST/NIRCam between 24-26 June 2026 at a position angle of PA=131${}^{\circ}$ to align with the NIRCam pointings of the CEERS program observed in 2022. We choose to use the INTRAMODULEBOX 4 dither pattern to cover all detector gaps and provide the most uniform depth across our survey field, ensuring sensitivity to even the faintest emission lines. We use the FULL subarray with the MEDIUM8 readout pattern (more groups per fixed exposure time, less susceptible to cosmic rays), with 8 groups, 1 integration, and 4 exposures per specification for the F480M/F430M/F360M + F070W/F140M/F162M filters. We reduced our groups to 7 for the F335M + F210M filters and to 6 for the F300M + F182M filters, but kept 1 integration and 4 exposures to take advantage of JWST’s increased sensitivity at these wavelengths and use telescope time more efficiently. All data from the SPAM survey are available in MAST at .

Our exposure times are those needed to detect a significant MB excess, with a formal signal-to-noise of ${\rm SNR}>3.0$ for the MB/BB flux ratio. The red filter(s) drive the depth, while the blue filters are matched to each red filter based on the sensitivity curves in order to achieve the most uniform depth. We reach this in 0.9 hours (3349.9s) of integration per pointing in F480M/F430M/F360M (+F070W/F140M/F162M), 0.8 hours (2920.4s) in F335M (+F210M), and 0.7 hours (2490.9s) in F300M (+F182M) due to the increased sensitivity of JWST at those wavelengths. At this depth, we are sensitive to galaxies with observed-frame EW_{\hbox{\sc [O\,iii]}} down to $\sim 400$Å in the red MB filters.

Data Reduction

The SPAM photometric dataset is processed alongside complementary medium-band data from the MINERVA survey [62], legacy CEERS broad-band coverage [29], an additional two filters from GO#2344 (PI: E. Bañados), and all other public NIRCam imaging falling within the CEERS footprint as of July 17, 2026 (See Appendix § 6). We ensure a uniform reduction across the imaging filters presented. Table tbl:filters lists all BB and MB filters that cover the CEERS legacy footprint and were reduced for this survey. Figure 1 shows a full 20-filter NIRCam color image illustrating the CEERS footprint we use for our mosaics and catalogs. This is inclusive of all public NIRCam imaging in these filters as of the date of publication.

We reduce all available NIRCam imaging data through a custom pipeline that largely follows [32] and [18]. Briefly, we processed the raw NIRCam images through the JWST Calibration Pipeline version 1.20.2 [67], with several custom modifications, including subtracting $1/f$ noise and sky background. We use the Calibration Reference Data System (CRDS)¹ pmap 1481, which corresponds to NIRCam instrument mapping imap 0342. We subtract $1/f$ noise through iterative source masking and amp-row median subtraction, as described in [32].

We remove wisps — scattered light features present in several NIRCam short wavelength (SW) detectors — by rescaling and subtracting the templates provided by STScI.² After stage2 of the pipeline, we visually inspect all exposures for residual unsubtracted wisps or other imaging artifacts (e.g., “claws,” “ginko leaves,” “dragon’s breath”); we then manually mask and remove these artifacts from the combined mosaics.

Astrometric calibration is conducted via the JWST/HST alignment tool [JHAT; 17] using as an absolute reference a photometric catalog generated from the CEERS DR1 NIRCam imaging [32, 75], itself aligned to the DESI Legacy Imaging Surveys [92] DR9 catalog as described in [32].

We create the final mosaics in Stage 3 of the pipeline with a pixel size of 0\farcs03/pixel on the same pixel grid as the CEERS DR1 imaging. We perform background subtraction on the final mosaics following the iterative procedure described in [32], which yields a highly uniform background without subtracting significant flux in the wings of bright sources.

We show the SPAM data depths in red in Figure 2, with the legacy CANDELS HST depths [27, ?] in gray and the legacy CEERS v1.0 depths in blue [29]. Our catalog has improved depths for the CEERS NIRCam filters compared to the public CEERS imaging [32], as shown in black in Figure 2. We discuss how the SPAM catalog improves depths over the initial public CEERS reductions in Appendix § 6. We list the depths of all the JWST/NIRCam filters in our SPAM catalog in Table tbl:filters. Fully reduced image mosaics can be found on the SPAM website³.

Photometry Catalog

The full photometric catalog presented here can be accessed from the SPAM website⁴. This catalog contains 174,465 sources across the legacy CEERS footprint, plus additional coverage from other public NIRCam data in the vicinity (See Appendix § 6). This catalog provides photometry across 28 filters: 7 HST/ACS and WFC3 filters from the CANDELS data [27, ?], and other public archival HST data, reprocessed to the latest calibrations, plus 20 JWST/NIRCam filters: seven from the CEERS survey [29], inclusive of additional public data in the field, two from GO#2234, two from MINERVA [62], and 10 from SPAM, all in nJy. Table tbl:filters lists all JWST/NIRCam filters and depths included in the photometry catalog.

The photometric catalog was created as part of the Uniform Near-Infrared CatalOgs from Robust imaging (UNICORN) project. UNICORN aims to measure highly accurate photometric colors, total fluxes, and photometric redshifts. This full process will be described in S. Finkelstein et al. (in prep), though it is based predominantly on the process described in [75]. The README associated with this release includes all details, but we provide a summary here.

We match the PSF of all ACS bands and all NIRCam bands bluer than F277W to that of F277W. For bands with larger PSFs (the redder NIRCam bands, and all WFC3 bands), we calculate correction factors by convolving the F277W image to the larger PSF and measuring the flux ratio in the native image to that in the convolved image for each source. By applying this correction factor to fluxes measured in images with larger PSFs, we correct for the missing flux (assuming the morphology does not change significantly).

We performed photometry with Source Extractor (v2.25.0, [60]) in two-image mode, using a weighted sum of the native-resolution F277W and F356W as the detection image. For the detection image, we used a 1/sqrt(WHT) map with map_rms weighting, so Poisson noise (which is in the ERR maps) did not affect the detections. For the measurement image, we used the relevant ERR map.

We ran both hot- and cold-mode catalogs. The hot-mode catalog has DETECT_THRESH=1.75 and DETECT_MINAREA=5, derived iteratively via visual inspection to minimize spurious sources while maximizing completeness; this has been validated to work reasonably well across a wide range of JWST imaging depths. We ran a cold-mode catalog with DETECT_THRESH=3.5 and DETECT_MINAREA=40. For both catalogs, we used a top-hat convolution kernel with a width of 4 pixels and a kernel size of 5x5 pixels, and set DEBLEND_NTHRESH=32. For the hot catalog, we used DEBLEND_MINCONT 0.0001, while for the cold catalog we used DEBLEND_MINCONT 0.005.

Empirical flux uncertainties were derived by placing random non-overlapping apertures with diameters from 0.1-1.5" in empty regions of the image (avoiding the segmentation map), fitting a four-parameter function to the Normalized Median Absolute Deviation (NMAD) of the fluxes measured in each aperture as a function of the aperture area. We calculate the flux error for each object from this fit for the area in a given object’s aperture.

We measure our fiducial fluxes in small Kron apertures and apply two aperture corrections. The first is based on the ratio of the default larger MAG_AUTO Kron aperture to our small aperture. We then calculate an additional aperture correction to measure flux missed on larger scales. In the NIRCam bands, we ran source injection simulations, injecting small Sersic profiles [7] with Galfit into the F277W and F356W images. We then create a detection image from F277W+F356W, and measure photometry in F277W. Comparing input and output fluxes, we find the input fluxes are $\sim$5-20% higher than the output, with this discrepancy slightly increasing for fainter magnitudes. We fit a linear function to this offset and apply it to our fluxes. This is typically a factor of 1.02 at $m=24$, rising to 1.10 at $m=28$ (we cap this value at 1.20, since we cannot reliably measure recovered fluxes near the imaging depth in a given field).

We did the same for the hot- and cold-mode detection parameters. To combine the catalogs, we first included all cold-mode objects. We then included hot-mode objects whose central pixel did not lie in a cold-mode segmentation map, after dilating the cold map with a 5x5 kernel. Finally, we created a separate catalog using only F444W as the detection image in hot mode, and then added any sources from this catalog that were not already in the hot$+$cold catalog. This helps ensure we still include faint, red sources in our catalog.

Our released photometric catalog includes fiducial fluxes and flux errors in all filters, as well as fluxes and errors measured in a number of circular-diameter apertures. Further information is available on the team website.

Photometric Redshift Measurements

Photometric redshifts are the backbone of many science goals enabled by the SPAM data (e.g., see § 4). We estimated photometric redshifts ($z_{phot}$) with LAZY⁵ a Julia adaptation of the EAZY software [78]. The methodology is similar to that in CEERS Key Paper I [16], including new templates from [15], and is described further in S. Finkelstein et al. (in prep). Our fiducial photometric redshifts use 23 filters in the catalog (as described in § 2.2), from both HST (ACS) and JWST, and were fit over $z=0-35$. We provide a photometric redshift for all sources in the SPAM photometric catalog under the ZA column (the redshift where the goodness-of-fit is maximized), with upper and lower 68th-percentile errors in the L68 and U68 columns. 95th percentile errors on the $z_{phot}$ are in the L95 and U95 columns. We also provide full $z_{phot}$ probability distribution functions and the best-fitting templates. We also provide full $z_{phot}$ catalogs with the fiducial filter set, including (separately) an extreme emission line galaxy (EELG) and little red dot (LRD) template, as well as our fiducial template set with only broad bands, plus WFC3 filters. The results of these fits are available on the team website.

Spectroscopic Redshift Catalog

The spectroscopic redshift ($z_{spec}$) measurements come from several input catalogs. The first is from the CEERS spectroscopic redshift catalog (Arrabal Haro et al., in prep), which uses CEERS NIRSpec data. We also use public JWST spectroscopic redshifts from the Dawn JWST Archive⁶ (DJA; [25, 24]) for recent spectroscopic redshifts from public NIRSpec data in EGS. We also measure spectroscopic redshifts ourselves on the public NIRSpec data in EGS and add these measurements to the list. Additionally, we include products from CANDELS [27, ?], specifically the public CANDELS HST+Spitzer catalog [23]. Finally, we include the ground-based redshifts from Keck/MOSFIRE: from the MOSDEF survey [22], other MOSFIRE programs [21, 20], and the Keck/DEIMOS observations in EGS [19]. We have spectroscopic redshifts for 7,004 sources in the SPAM survey. These values are in the SPAM_Specz_Catalog file under the ZSPEC column found on the SPAM website⁷.

Each input catalog includes a grade for the reliability of its redshift measurement, and we use those grades to determine a vetting value for the SPAM catalog. We vet spectroscopic redshifts by assigning a value of 2 if the $z_{spec}$ is believed, 1 if it is likely but uncertain, 0 if it is not believed, and $-1$ if it is not vetted. For example, those sources in the DJA that have a grade of 3 (flagged by DJA as robust) are assigned a $z_{spec}$ vetted value of 2, but those that have a DJA grade of 0 (flagged by DJA as a data quality issue) or 1 (flagged by DJA as having no features) are assigned a vetted value of 0. We have confident (grade $= 2$) spectroscopic redshifts for 4,557 sources in the SPAM catalog, and these are what we use to test our photometric redshifts (Figure 3). Our grades are in the SPAM_Specz_Catalog file under the ZSPEC_VETTED column.

Improvements to Photo-z’s with Medium Bands

We compare photometric redshifts to our catalog of confident (zspec_vetted = 2) spectroscopic redshifts in Figure 3. The left panel of Figure 3 compares the photometric redshifts before (blue) and after (red) the addition of SPAM and MINERVA MBs. We include several lines that trace possible sources of degeneracy in SED template fitting. The figure shows that adding medium bands improves photometric redshift accuracy across a broad redshift range. The blue line in Figure 3 traces  mistaken for , and the green line traces  mistaken for . We find  mistaken for  to be a rare source of confusion. However, several sources, especially at $4<z<7$, have  mistaken for  (along the green line). This is often tied to AGN, especially LRDs, where  line strengths are much greater than , as [63] noted.

The yellow line in Figure 3 traces cases where the Balmer break has been mistaken for the  break, and the purple line traces cases where the Lyman break has been mistaken for the Balmer break. We find that the Balmer break is mistaken for the  break in rare cases, and that several interlopers at $z<2$ are mistaken for higher-redshift sources due to this confusion. The opposite case, where the  break is mistaken for the Balmer break, is our most significant source of confusion for SED fitting before adding the medium bands, as the largest outlier fraction occurs at these redshifts.

The middle panel of Figure 3 shows the $\Delta z$ / ($1+z_{spec}$) for both the CEERS (blue) and SPAM (red) catalogs. Here $\Delta z$ = $z_{phot} - z_{spec}$. This highlights the improvements in photometric redshift fitting after including the MBs (red). To understand this more statistically, we calculate the NMAD for our standard deviation calculations, as it is more outlier-resistant:

Here, the normalization value is $X = 0.5 \times \sqrt{2}$ or 0.67449, which makes NMAD equivalent to a standard deviation for a Gaussian distribution. This is shown as the solid line in the right panel of Figure 3 for our photometric redshift fits both with (blue) and without (red) the addition of the MBs.

We also calculate the outlier fraction, defined as the fraction of points where $|\Delta z| > 0.2 \times (1 + z_{spec})$, often considered catastrophic failures of our photometric redshift fitting. These fractions are shown as dotted lines in the right panel of Figure 3 both with (red) and without (blue) the new MBs. The outlier fractions are much higher around $z=3$ before the addition of the MBs (blue), which are often the points along the orange line in the left panel of Figure 3 where the Balmer break has been mistaken for the  break, or along the purple line where the  break has been confused for the Balmer break. Outlier fractions increase at higher redshift due to low-number statistics.

Improved Selection for Distant Galaxies

SPAM adds the bluest NIRCam BB filter to the CEERS legacy dataset: F070W. This filter probes  break dropouts from $5<z<12$ as it lies between the HST F606W and F814W filters. SPAM achieves fainter depths than CANDELS HST F814W imaging and comparable depths to the F606W. This further constrains high-z candidates by increasing the confidence in  break dropout galaxies. This coverage, in combination with the addition of the MBs, gives the most complete look to date at the distant ($z>5$) Universe in CEERS.

Early CEERS work included searching for the furthest galaxies [39], with notable early candidates at $z\sim12$ [40] and $z\sim16$ [35]. However, follow-up spectroscopy confirmed the former at $z=11.4$, but revealed that the $z\sim16$ candidate was a $z\sim4.9$ reddened extreme emission-line galaxy [31]. We present the SEDs of the candidate $z=16$ source in Figure 4 using only the BBs (blue) and after adding the SPAM + MINERVA coverage (red), where we recover the correct photometric redshift at $z=4.86$ (black). The WB only fits presented here are fit through a methodology identical to § 3 with the MBs removed. Using only BBs produced an SED for this source with photometrically blended lines that mimicked those of a high-z source with a  break at $z\sim16$ (blue). However, the MBs localize the photometric impact of the strong emission lines at certain wavelengths, enabling identification of the continuum between the strong emission-line features and accurate photometric redshift measurement (black).

Science Cases

Rare sources are typically identified via photometric surveys and then followed up via expensive spectroscopic observations. Compared to these expensive targeted observations, SPAM offers an efficient way to enable science over a large sample of sources in only a few (10) NIRCam pointings with $\sim2-3$ orders of magnitude more sources than available from all NIRSpec surveys in CEERS combined. SPAM is also a non-proprietary program. This ensures the community can use the added SPAM filters, enabling a wide range of discoveries when combined with the already prolific CEERS dataset and the similarly non-proprietary MINERVA program. Here, we demonstrate how adding MBs improves the three key science cases of the SPAM Survey: detecting overdensities in the CEERS field, identifying emission-line galaxies, and constraining star formation histories.

Quantifying Overdensities Powering Reionization

The history of cosmic reionization is poorly constrained, both in which objects drove it and how fast the process occurred. Current theory favors “late reionization" [e.g., 49], where UV emission from galaxies accounts for the ionizing photons needed to match the constraints. This model requires an ionizing photon escape fraction of 20% for galaxies across all luminosities; if this holds, then the brightest galaxies ($>$0.01L$^{\star}$) naturally dominate reionization. Other simulations predict that the ionizing photon escape fraction depends on galaxy halo mass, with a higher escape fraction from galaxies in the lowest-mass halos (log M$_h$/M$_{\odot}<$9). In this scenario, the faintest galaxies ($<$0.01L$^{\star}$) instead dominate reionization [50], and the high abundance of faint galaxies at very early times [47, 40] drives an earlier start to reionization and a smoother time evolution [38]. In either scenario, over-dense regions of the universe ionize first due to the cumulative contribution of galaxies’ ionizing radiation output [49]. If large, early reionized patches can be robustly identified and characterized at $z > 8$, they would support an earlier reionization start.

While the confirmed galaxies in these epochs are bright, they are unlikely to dominate the local ionizing budget alone. This is not surprising, however, as overdensities should be traced not only by these bright galaxies but also by an overabundance of faint galaxies. Detecting an abundance in the fainter galaxy population would not only confirm these overdensities, but also provide high-value targets for JWST spectroscopy to probe faint  from these systems and map the full extent of these ionized regions.

SPAM enables an efficient method to detect and quantify these additional faint sources, as galaxies at this redshift are known to exhibit very strong  +  emission (combined rest-frame EW $\geq$ 800Å; [34, 74, 51]). SPAM covers the  +  emission line complex at $z=7.3-8.0$ and $z=8.3-8.9$, respectively, through the F430M and F480M filters. This encompasses two known overdensities at $z=7.8$ [51] and $z=8.7$ [21, 52, 2]. SPAM selects members by detecting excess emission in the MB imaging relative to the CEERS BB image, as the emission lines pass through the MB filters. This tightens constraints on the redshift probability, $P(z)$, as shown in Figure 5, where the $P(z)$ before (blue) is much broader than after adding the medium bands (red). This lets us identify robust overdensity candidates, as those with tightly constrained $P(z)$ distributions around the overdensity redshifts. Preliminary tests on the SPAM catalog identify 25 candidate galaxies at $z=7.8$ and 13 candidate galaxies at $z=8.7$. More detailed analysis on this can be found in Larson et al. (in prep).

Emission-Line Galaxies at $2 < z < 9$

NIRSpec detections have confirmed strong  +  emission in $4<z<9$ galaxies [51, 44, 45, 77, 63], first identified in NIRCam photometry. In some cases, equivalent widths (EWs) $>2000$ Å have been reported [34, 36, 74], dwarfing low-z emission line strengths. These high EWs indicate high ionization parameters and consequently high ionizing production rates ($\xi_{ion}$, [51, 37, 52, 48, 77]). The strength of these lines, combined with the sensitivity of JWST’s NIRCam, has enabled studies of  at $4<z<7$ and  +  at $4<z<9$ from broadband photometry alone [34, 74, 76].

However, the confidence with which these emission lines can be disentangled from the broadband photometry depends directly on the availability of MB photometric filters [34, 43, 42]. Emission lines are most reliably detected when they are captured by a MB filter, while  and  +  are most confidently de-blended from the photometry when a medium-band filter probes the continuum between the lines [34]. However, the legacy CEERS observing design included only one MB filter, F410M.

Consequently, high-confidence photometric  and  +  studies in CEERS at $z>4$ are limited to a narrow redshift range. Figure 6 shows how we can now deblend these emission lines, with legacy CEERS coverage in gray, SPAM in red, and MINERVA in blue. Solid lines indicate which emission lines the new filters probe at $1<z<15$. SPAM provides coverage of F300M, F335M, and F360M, which, in combination with existing CEERS F410M coverage, samples the continuum between  and  +  at $4<z<6.7$ with no gap in coverage. The 430M and 480M filters further break up the 444W band to isonalte  emission at $6.7<z<9$, which enables emission-line deblending and improved $\rm{z_{phot}}$ measurements at the highest redshift available from NIRCam photometry alone. Anticipating that probing the continuum between emission lines would also improve detection of emission-line galaxies at lower redshifts, SPAM includes two bluer filters, F210M to probe the wavelength range between the F200W and F277W broadband filters and deblend  and  +  at $2<z<3.5$. Similarly, SPAM provides F182M to probe the region between the F150W and F200W filters, deblending these emission lines at $1.5<z<2$. We also include the 140M and 162M filters to similarly break apart the wavelengths covered by the F150W filter.

The two sources in Figure 7 show the emission-line isolation now possible photometrically with SPAM + MINERVA, enabling new emission-line science across expanded populations derived photometrically. The source in the top panel, SPAM 33593, is an EELG from [64] whose  and  +  lines were previously blended in the BB filters. In the legacy CEERs photometry (blue), the source was placed at $\rm{z}_{phot} = 1.04^{+4.10}_{-0.02}$, with wide error reflecting the degenerate nature of the SED. The  line is captured by the single MB from the legacy CEERS coverage, F410M, but falls at a redshift where the line boosts the neighboring BB filters (F356W and F444W) while the  emission boosts F277W. This creates a false continuum shape and the SED fitting finds a $\rm{z}\sim4$ EELG and a $\rm{z}\sim1$ galaxy with a stellar bump boosting the BBs degenerate. The improved MB coverage (red) successfully isolates the continuum emission between the two lines, and our new SED places the galaxy at $\rm{z}_{phot} = 4.94^{+.02}_{-0.02}$, increasing the accuracy of the photometric redshift through isolation of the  and  +  emission lines without a need for follow-up spectroscopy.

The bottom panel of Figure 7 shows a Little Red Dot (LRD) identified in [68]. The WB-only fit places the LRD at $\rm{z}_{phot} = 5.58^{+0.08}_{-0.06}$, and our fit from the SPAM+MINERVA coverage places the source at $\rm{z}_{phot} = 5.98^{+0.02}_{-0.02}$, in agreement with $\rm{z_{spec}} = 5.95$ from the Skyfire survey [97]. While the z recovery from the BB only fit performs well, the strength of the  and  emission lines are difficult to recover, blended with the reddened rest-optical continuum. Without the emission-line isolation enabled by MBs, the rest-optical continuum is under-constrained. The complete MB photometry enables recovery of the strengths of these lines and continuum in a system with sub-extreme emission lines, demonstrating the expanded line recovery enabled by MBs.

SFHs of Massive Galaxies at $z > 4$

Early JWST results have revealed an apparent overabundance of high-redshift, massive galaxies at $z>4$ [13, 85, 79, 56, 87], challenging galaxy formation models to create adequate amounts of stellar mass in a short enough timescale to reproduce the observed sources [53, 12, 11], motivating theories requiring highly efficient star formation in the early Universe [86, 89, 90]. Furthermore, observations have confirmed galaxies that have ceased their star formation entirely, hosting an older-than-expected stellar population [33, 41, 88, 10, 9] at early times. Together, this implies a Universe in which galaxies cannot only form extreme levels of stellar mass in a short time but also rapidly quench star formation on a galactic scale. To properly constrain galaxy formation histories in the early Universe, medium bands are necessary to isolate line emission from the underlying continuum flux. SPAM provides coverage of F300M, F335M, and F360M to deblend  and  +  from the photometry at $4<z<5.5$ and F140M, F162M, F182M, and F210M to probe the UV continuum at $2<z<12$.

Both  and UV fluxes not only constrain photometric redshift-derived distances, but also directly probe star formation histories (SFHs).  emission from galaxies requires massive, short-lived O- and B-type stars, directly probing SF on recent ($\sim 10$ Myr) timescales [46]. The medium bands isolate this line emission from the broadband continuum and enable 2D emission-line studies, building SEDs from pixels as in [55], and directly constraining SF from .

UV continuum emission traces sustained SF populations over $\sim 100$ Myr timescales. Studies of SFHs in JWST have begun to uncover evidence for burstiness in early stellar populations [54, 91], and MB coverage from SPAM + MINERVA further quantifies population-level statistics for bursty star formation by isolating  and UV fluxes.

In the more evolved Universe, the four blue MB filters cover the Balmer break region at $3<z<5$, one of the most information-rich regions of the earliest-known massive, quiescent galaxies. The shape of the Balmer break can be effectively determined using multiple MB filters [8], adding necessary constraints on the age and dust content of these galaxies and breaking the degeneracy between Balmer breaks and dust reddening. The SPAM F070W photometric filter extends this coverage to the Balmer break at $0.5<z<1.25$, which, combined with existing F090W coverage, completes Balmer-break detection over cosmic history in CEERS.

In Figure 8, we show an example star-formation history (SFH) of a galaxy using only CEERS BB photometry (blue) versus that recovered with the addition of the SPAM+MINERVA MBs (red). The MB information significantly improves the galaxy’s SFH by capturing emission lines and capturing the exact location (improving redshifts) and strength of the Balmer break, leading to much stronger constraints on galaxy stellar populations.

Summary

We present a photometric catalog from the SPAM and MINERVA surveys which, combined, leave the CEERS legacy footprint with coverage from all BB and MB NIRCam photometric filters. The only other fields with such complete filter coverage are CANUCS and JADES, with coverage over $\sim30 \,\rm{arcmin^2}$ and $\sim10 \,\rm{arcmin^2}$, respectively, compared to $\sim 100 \,\rm{arcmin^2}$ presented here, a greater area than all surveys predating Cycle 5 combined. This photometric catalog includes both the legacy CEERS coverage and the SPAM + MINERVA coverage in a uniform reduction, along with photometric redshifts. The catalog is publicly available and will be maintained in the most up-to-date version on the SPAM website⁸.

The SPAM coverage improves photometric redshifts across all redshifts, and we include photometric redshifts with the photometry catalog. We compare these photometric redshifts to $\sim4500$ spectroscopic redshifts available in the EGS field. The SPAM-inclusive photometric redshifts improve on the legacy CEERS BB coverage, especially in cases of chronic degeneracy tied to unresolved emission lines and confused SED breaks.

This coverage not only makes the CEERS field an excellent test bed for photometric science, but also provides an equivalent to very-low-resolution spectroscopy that can recover the continuum shape and photometrically isolate  and  + . This will enable a wide range of science goals. Here we highlight three key improvements:

• Overdensities in CEERS: Improved photometric redshifts enable identification of resolved overdensities in CEERS, partly due to photometric deblending of strong rest-optical emission lines.

• Emission Line Galaxy Identification: Emission lines, especially strong rest-optical lines, can now be isolated photometrically for all sources in the CEERS legacy footprint. This improves identification of these strong lines and will help characterize many source types, from EELGs to LRDs, in future work.

• Star Formation Histories: Better Balmer-break fitting and emission-line detection with the additional MB data enable more well-constrained SFHs, as they break degeneracies between continuum and emission-line contributions.

Acknowledgments

We thank the MINERVA team for maintaining open communication, which enabled this unprecedented dataset. The 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 JWST Cycle 4 GO programs 8559 and 7814. Support for the program JWST-GO-8559 was provided by NASA through a grant from the Space Telescope Science Institute (STScI), which is operated by the Association of Universities for Research in Astronomy (AURA), Incorporated, under NASA contract NAS5-26555. The material is based upon work supported by NASA under award number 80GSFC24M0006.

RLL and LYAY appreciate support from a Giacconi Fellowship at the Space Telescope Science Institute, which is operated by the Association of Universities for Research in Astronomy, Inc., under NASA contracts NAS 5-26555 and NAS5-03127. KD and RLL acknowledge support from NASA JWST-GO-8559. KD acknowledges support from an NSF Graduate Research Fellowship award number 2040433. KD and JRT acknowledge support from NASA JWST-GO-8410. KD acknowledges support from Los Alamos National Laboratory. This work was released under LA-UR-26-26360. LYAY acknowledges support from the Taiwanese Ministry of Education (MoE) Yushan Fellowship (MOE-115-YSFMS-0002-005-P1).

Improved Depths over Legacy CEERS

Our SPAM catalog includes data from all public JWST/NIRCam imaging near the legacy CEERS footprint. We reduced all data using the same pipeline and process, then mosaicked them to create a CEERS tile that encompasses a $0.370 \times 0.123$ degree rectangle around the original CEERS NIRCam footprint (which is the same as the SPAM footprint). Figure 9 shows this extended coverage, with the complete F200W coverage on the left and the F444W coverage on the right.

We calculate the 5$\sigma$ point-source depths reported in Figure 2 and Table tbl:filters using the measured noise function in the field. We use an aperture diameter of 0.2 arcsec and correct to total flux based on the flux enclosed for a given band’s PSF. The improved depths over the legacy CEERS data we report are largely due to the increased coverage from additional imaging over the field, but pipeline improvements do play a small part (See Figure 10, Left). These pipeline improvements include new flats, which now use all Cycle 1-3 data and most Cycle 4 data. New darks have been added since the original CEERS data release, and the readnoise and superbiases have also been improved. We have also tuned the 1/f correction using a Gaussian process model.

Most of the depth improvements we report come from additional data over the field in each filter from other public NIRCam programs. Coverage is not uniform across the field in any filter, as shown in Figure 9. This means we calculate a median depth over the full footprint, but the minimum and maximum depths span a broad range. We show this by plotting the area vs. depth in Figure 10. The center panel shows depth vs. area for all public BB data in the CEERS footprint (See Figure 9) available at the time of publication. Comparing the center panel to the left panel shows improved depths relative to the original CEERS DR1 [32] data due to the addition of ancillary data in the field. The right panel of Figure 10 shows depth vs. area for all MB and one narrow-band (NB) filter in the mosaics from both SPAM and MINERVA, as well as a few ancillary pointings in the CEERS footprint. Annotations indicated the 5$\sigma$ depths at 50% area.