Parallel Application of Slitless Spectroscopy to Analyze Galaxy Evolution (PASSAGE): Survey Overview

TL;DR

PASSAGE is a JWST program that uses slitless spectroscopy to study galaxy evolution across many fields, providing unbiased, high-resolution near-infrared data that reveal faint and rare galaxies, advancing understanding of galaxy properties.

Contribution

This paper presents the PASSAGE survey, the largest JWST Cycle 1 program using slitless spectroscopy, offering new unbiased, high-resolution near-IR data for galaxy evolution studies.

Findings

Over ten thousand near-IR spectra of faint galaxies obtained.

Survey covers multiple fields to reduce cosmic variance.

Data reveal properties of low-mass emission-line galaxies.

Abstract

During the second half of Cycle 1 of the James Webb Space Telescope (JWST), we conducted the Parallel Application of Slitless Spectroscopy to Analyze Galaxy Evolution (PASSAGE) program. PASSAGE received the largest allocation of JWST observing time in Cycle 1, 591 hours of NIRISS observations to obtain direct near-IR imaging and slitless spectroscopy. About two thirds of these were ultimately executed, to observe 63 high-latitude fields in Pure Parallel mode. These have provided more than ten thousand near-infrared grism spectrograms of faint galaxies. PASSAGE brings unique advantages in studying galaxy evolution: A) Unbiased spectroscopic search, without prior photometric pre-selection. By including the most numerous galaxies, with low masses and strong emission lines, slitless spectroscopy is the indispensable complement to any pre-targeted spectroscopy; B) The combination of several dozen independent fields to overcome cosmic variance; C) Near-infrared spectral coverage, often spanning the full range from 1.0--2.3 $\mu$m, with minimal wavelength gaps, to measure multiple diagnostic rest-frame optical lines, minimizing sensitivity to dust reddening; D) JWST's unprecedented spatial resolution, in some cases using two orthogonal grism orientations, to overcome contamination due to blending of overlapping spectra; E) Discovery of rare bright objects especially for detailed JWST followup. PASSAGE data are public immediately, and our team plans to deliver fully-processed high-level data products. In this PASSAGE overview, we describe the survey and data quality, and present examples of these accomplishments in several areas of current interest in the evolution of emission-line galaxy properties, particularly at low masses.

Tables

Table 1. Table 1: List of observed PASSAGE fields with their target coordinates and exposure times.

Field	R.A. [deg]	Decl [deg]	Obs Date	F115W	F150W	F200W	F115W	F150W	F200W
				Direct	Direct	Direct	Grism	Grism	Grism
				[hr]	[hr]	[hr]	[hr]	[hr]	[hr]
Par001	170.0267050	6.5888780	2022-12-17	0.43	0.86	–	2.30*	4.60*	–
Par002	177.1453720	52.7788100	2022-12-28	0.86	0.43	–	3.45*	3.45*	–
Par003	150.4103960	2.4051780	2022-12-30	–	–	0.62	–	–	1.24*
Par004	186.7047850	21.6989890	2022-12-31	–	–	1.15	–	–	1.15
Par005	150.6536980	2.0855800	2023-01-07	0.43	0.43	–	2.29*	2.29*	–
Par006	150.6762160	2.1293050	2023-01-06	0.43	0.43	–	2.29*	2.29*	–
Par007	102.4303090	70.3818910	2023-01-08	–	–	0.89	–	–	0.89
Par008	170.1264770	6.7062310	2023-01-12	–	–	0.86	–	–	1.72*
Par009	183.1322960	27.4870880	2023-01-13	–	–	0.52	–	–	1.19*
Par010	205.5720260	9.3080670	2023-02-21	0.17	–	–	3.01	1.00	–
Par011	204.5789780	-19.8794780	2023-02-22	2.61	–	–	2.61	–	–
Par012	221.6227940	-23.2860630	2023-02-25	–	–	0.01	–	–	1.91
Par013	130.5298840	1.6983950	2023-03-19	–	0.52	–	0.67	0.52	–
Par014	224.3278890	24.7169780	2023-03-23	0.67	–	–	0.67	–	–
Par015	155.4060630	18.1657440	2023-04-08	10.52	–	–	18.04	–	–
Par016	152.6594150	-4.7508000	2023-04-14	–	0.29	0.14	1.19	0.29	0.29
Par017	150.3011090	1.8910820	2023-04-19	–	–	3.15	–	–	3.15
Par018	159.2102450	1.7162860	2023-04-20	–	0.67	–	1.10	1.10	–
Par019	130.3650700	48.6033460	2023-04-20	0.10	0.10	–	0.86	0.86	–
Par020	149.6229230	2.2251060	2023-04-23	–	0.13	–	4.87	1.05	–
Par021	161.1635170	33.9294610	2023-04-24	–	–	0.52	0.67	0.67	0.57
Par022	177.2619900	22.4367760	2023-04-28	–	–	3.15	–	–	3.15
Par023	150.0655290	2.5519210	2023-05-01	–	0.13	–	4.29	1.05	–
Par024	150.1315240	2.4996600	2023-05-02	–	0.13	–	4.87	1.05	–
Par025	150.2457730	1.9774740	2023-05-02	–	0.13	–	4.87	1.05	–
Par026	150.1227170	2.4398770	2023-05-06	–	0.13	–	4.87	1.05	–
Par027	172.9448010	-12.3606500	2023-05-09	0.29	0.29	0.29	2.58*	1.15	1.15
Par028	150.0874070	2.4185310	2023-05-20	1.79	1.43	1.43	21.26*	7.08*	5.37*
Par029	150.2522870	1.7285570	2023-05-22	–	0.13	–	4.87	1.05	–
Par030	216.0875670	24.2475070	2023-05-25	–	0.79	–	0.79	0.79	–
Par031	187.7406480	12.5610560	2023-05-31	–	–	0.41	–	–	6.49
Par032	180.4252840	-18.8087340	2023-06-07	–	1.00	–	1.00	1.00	–
Par033	176.4223950	20.3820560	2023-06-09	–	–	0.95	–	–	0.95
Par034	190.4031030	22.4995390	2023-06-21	1.24	1.24	1.24	6.20*	1.24	1.24
Par035	181.5579470	-8.6302920	2023-06-21	–	–	2.98	–	–	2.98
Par036	184.2851990	2.7440130	2023-06-21	–	–	0.20	–	–	2.03
Par037	180.3574960	2.3583210	2023-06-23	0.48	0.48	–	1.91	1.91	–
Par038	215.1343710	53.0534480	2023-06-25	–	0.13	–	4.87	4.87	–
Par039	182.9192700	-1.1318990	2023-06-29	–	–	0.67	–	–	3.82
Par040	181.3629010	-7.5363050	2023-07-01	–	–	0.95	0.95	0.95	0.95
Par041	191.4234500	3.5618430	2023-07-02	–	–	0.43	–	–	3.44
Par042	195.1846320	12.5341350	2023-07-04	–	0.57	0.57	2.58	1.00	1.00
Par043	213.2233650	-3.2187080	2023-07-25	–	–	0.14	–	–	0.14
Par044	265.7442370	67.0902050	2023-09-03	–	–	0.14	2.10	0.52	0.57
Par045	137.7250700	-4.4086120	2023-11-17	–	–	0.95	–	–	3.01
Par046	149.3468690	2.5384070	2023-12-03	–	–	0.20	–	–	1.62
Par047	149.4398780	2.1723810	2023-12-03	–	–	0.20	–	–	1.62
Par048	149.4029130	1.6714440	2023-12-03	–	–	0.20	–	–	1.62
Par049	149.7024380	2.0589470	2023-12-03	–	–	0.20	–	–	1.62
Par050	189.1800240	62.0654160	2023-12-22	–	2.04	2.04	12.02*	3.94	3.94
Par051	150.5036630	2.2633940	2024-01-02	0.79	–	–	7.09*	–	–
Par052	150.1552210	2.0418390	2024-01-03	–	2.04	1.36	12.02*	3.94	3.94
Par053	150.1030700	1.7931780	2024-01-04	–	–	0.20	–	–	1.62
Par054	214.5759310	52.2796830	2024-01-15	–	–	4.29	4.29	4.29	4.29
Par055	177.4168450	22.1796790	2024-01-16	–	0.79	–	0.79	0.79	–
Par056	169.9978930	6.5094450	2024-01-22	–	0.79	0.79	–	1.57	1.57
Par057	177.4628390	22.2952510	2024-01-24	–	–	0.67	0.91	0.91	0.67
Par058	177.5042440	22.6834370	2024-05-16	–	–	4.51	4.51	4.51	4.51
Par059	169.9723860	6.7628470	2024-05-19	–	1.43	–	2.86	1.43	–
Par060	157.5820920	5.5249110	2024-05-26	–	–	0.43	–	–	1.15
Par061	159.1929810	-2.4619510	2024-06-01	0.43	0.43	–	2.30*	2.30*	–
Par062	188.1922630	9.3749900	2024-06-20	–	–	0.20	–	–	2.23
Par063	149.9702590	-22.8246010	2023-12-09	–	–	0.14	–	–	0.14

Full text

PASSAGE brings unique advantages in studying galaxy evolution: A) Unbiased spectroscopic search, without prior photometric pre-selection. By including the most numerous galaxies, with low masses and strong emission lines, slitless spectroscopy is the indispensable complement to any pre-targeted spectroscopy; B) The combination of several dozen independent fields to overcome cosmic variance; C) Near-infrared spectral coverage, often spanning the full range from 1.0–2.3 $\mu$m, with minimal wavelength gaps, to measure multiple diagnostic rest-frame optical lines, minimizing sensitivity to dust reddening; D) JWST’s unprecedented spatial resolution, in some cases using two orthogonal grism orientations, to overcome contamination due to blending of overlapping spectra; E) Discovery of rare bright objects especially for detailed JWST followup. PASSAGE data are public immediately, and our team plans to deliver fully-processed high-level data products.

In this PASSAGE overview, we describe the survey and data quality, and present examples of these accomplishments in several areas of current interest in the evolution of emission-line galaxy properties, particularly at low masses.

JWST, Galaxies, Cosmic Evolution, Early Universe

1 Introduction

The value of unbiased spectroscopic selections of galaxies has long been recognized for understanding their evolution across cosmic time. For emission-line galaxies, this can crudely be done with ground-based imaging through narrow-band filters (Teplitz et al., 1999; Kashikawa et al., 2006; Ly et al., 2007; Kashikawa et al., 2011; Ly et al., 2012; Becker et al., 2018; Bunker et al., 1995). Full spectral surveys have been done using grism spectroscopy with HST/NICMOS, (McCarthy et al., 1999; Yan et al., 1999; Hicks et al., 2002; Backhaus et al., 2024), and then with HST/WFC3 (Atek et al., 2010; Malkan & WISP Team, 2013; Henry et al., 2013; Bagley et al., 2017; Henry et al., 2021; Battisti et al., 2024; Treu et al., 2015). More recently, JWST has also used grisms to obtain slitless spectroscopy of large samples of galaxies, e.g. (Estrada-Carpenter et al., 2024; He et al., 2024; Bradač et al., 2024; Rihtaršič et al., 2025; Watson et al., 2025).

Cosmic star formation likely peaked at $z\sim$ 2–3 (Ly et al., 2009; Madau & Dickinson, 2014). Most of our knowledge of galaxies at these epochs is restricted to those which are massive. We know less about the more representative galaxies with $\log(M_{\ast}/M_{\odot})\leq 10$, as samples are sparse and biased, with the stacking of low-SNR spectra often required (e.g., Henry et al., 2013; Sanders et al., 2018; Henry et al., 2021). Parallel NIRISS slitless spectroscopy is revolutionizing our understanding of galaxy evolution at high redshifts, especially in the important low-mass regime. By combining many fields with moderate depth, plus smaller numbers of deep fields, PASSAGE obtains coverage of statistically powerful numbers of galaxies spanning a very wide range of luminosities and stellar masses. We are measuring key optical emission line diagnostics, including [OII]$\lambda$3726,3729Å, [OIII]4959,5007Å, and hydrogen Balmer lines, across a wide range of redshift ($z=1-3.5$), and Paschen and [SIII]9532,9069 lines at lower redshifts, for of order ten thousand galaxies–by far the largest sample ever assembled.

In this paper we discuss our new PASSAGE observations in Section 2, the data reduction and sensitivity in Section 3, some examples of scientific results in Section 4, and future prospects in Section 5.

2 Observations

2.1 Slitless Spectroscopy with NIRISS

The Near Infrared Imager and Slitless Spectrograph (NIRISS) instrument on the James Webb Space Telescope (JWST) enables near-infrared slitless spectroscopy (0.8–2.3 µm) with two orthogonal medium resolution ($R\sim 150$ at 1.4 µm) grisms over a 133″$\times$133″ field of view (Willott et al., 2022; Doyon et al., 2023). NIRISS employs a set of specialized broad filters (PASSAGE uses F115W, F150W, and F200W), to selectively transmit specific wavelength ranges of infrared light, crucial for isolating key spectral features while blocking out-of-band radiation. The filters work in conjunction with the NIRISS grisms (GR150R and GR150C), enabling slitless spectroscopy by dispersing light across the detector, along either rows or columns, respectively. The combination of these filters and grisms helps minimize contamination from overlapping spectral orders and maximizes sensitivity in the target wavelength range, making them essential for capturing detailed spectrograms of faint astronomical sources. By using multiple filters in sequence, NIRISS can construct a comprehensive spectral profile of objects.

For the first Cycle of (JWST Cy1), a large program of pure parallel spectroscopy received an allocation of 591 hours. PASSAGE, JWST-1571 (PI: Malkan)–began in December 2022, when the pure parallel observing mode first became available for scheduling. Pure-parallel observations, as implied, are executed simultaneously with a separate instrument (NIRISS in our case) in parallel while a primary instrument is being used by a different guest observer program. JWST pure-parallel observations are subject to various constraints imposed by the prime observation. A collection of prime observations were assigned to our program where we were allowed to craft an observing strategy for the NIRISS pure-parallel observations – we refer to these as parallel opportunities.

Scheduling pure parallel slitless spectroscopy is challenging. Ideally, each wavelength setting requires 3 filter wheel motions (for a direct image and spectra with both orthogonal orientations of the dispersion axis). However, to eliminate any risk of losing telescope guiding, these changes can only be made when the primary observers are simultaneously reconfiguring their instrument, and not taking data. In many parallel opportunities, this restricts the number of wavelengths that can be covered, especially when the primary observer is mosaicking more than one sky region.

We individually crafted observations of each available pure parallel field to optimize the use of every parallel opportunity, down to 1.5 hours in length. These PASSAGE observations were adapted to any combination of primary exposure times and filter changes. Since the primary objective of the PASSAGE survey was to maximize the wavelength range of the spectroscopy, in opportunities with more allowed filter changes, our first priority was to extend wavelength coverage by adding spectroscopy through more broadband blocking filters (F115W, F150W, and F200W). The secondary objective when defining the observations was to reduce contamination due to overlapping spectra, and hence, our next priority was to add observations with the second GR150 grism (usually GR150C after GR150R), providing spectrograms at a second, orthogonal, orientation. We excluded potential fields at Galactic latitudes below 20 deg, to avoid spectrum overlaps in crowded fields.

Table 1 presents the PASSAGE Observing Log, detailing the sky coordinates of all NIRISS field centers observed to date, along with the exposure times for both direct imaging and spectroscopy in each waveband. While our exposure times were constrained by the scheduling of prime observations, we consistently assigned the longer integrations to spectroscopy, ensuring that the grism exposures have equal or greater duration than the imaging. The sky locations of these observations are illustrated in Figures 1 and 2.

Because the fainter emission lines of interest (Ly$\alpha$ and [OII]) generally fall in the shorter-wavelength spectral coverage, in all medium- and long-exposure parallel fields, we typically integrated 2–2.5 times longer in that F115W broadband filter (1.0–1.3$\mu$m) than in the longer-wavelength filters. Most of the shorter parallels fields (up to 4 hours total) were devoted entirely to spectroscopy through the F200W broadband filter (1.7–2.3$\mu$m), because this long-wavelength region was inaccessible to slitless spectroscopy with WFC3 on HST (Battisti et al., 2024). The detailed breakdown of exposure combinations for the survey is illustrated in the pie-chart of Figure 3. Four out of the 63 PASSAGE fields have multiple combinations orientations and filters, leading to a double counting in the wheel of Figure 3. For instance, a field that was observed with F115W in 2 PAs and with F150W and F200W with only 1 PA, contributes to both the purple and orange slices. Therefore, the number of fields in the wheels sum up to 67, rather than 63.

2.2 Emission Line Detection Sensitivity

In the fields covered by pure parallel visits longer than 3 hours, we usually obtained minimum total integration times of 4800 seconds in two grism filters. The main difference in the longer opportunities was the addition of longer integrations in the F115W filter (by factors of 2 to 3 times). Therefore our sensitivity in the F150W and F200W filters, as shown for some examples in Figure 4, was fairly uniform over most of our planned observations: for point sources, the line sensitivity was typically around $3-4\times 10^{-18}$ erg/sec/cm2 (5$\sigma$). This limiting line flux for H$\alpha$ is equivalent to a Star Formation Rate (SFR) $\sim$0.5 $M_{\odot}$/yr at $z\sim 2$ for a Chabrier IMF (Kennicutt, 1998). Throughout this paper assumes a flat lambda CDM cosmology with $H_{0}=70$ and $\Omega=0.3$.

We confirm the predictions of the exposure time calculator (ETC)111https://jwst-docs.stsci.edu; for a spatially extended galaxy (a Sérsic profile with major and minor axes of $1.0\arcsec\times 0.5\arcsec$, the limiting emission line flux in the central 4 pixels (a $0.13\arcsec\times 0.13\arcsec$ box) was 3 times higher.

These estimates are independently confirmed by published grism spectroscopy in a NIRISS field (Boyett et al., 2022). Although their integrations were somewhat shorter than what we typically obtained, the line flux sensitivities are consistent with our predictions, allowing detections of one or more rest-frame optical emission lines in more than 90% of their targeted galaxies.

In the shallow fields, we devoted the more limited time to the reddest grism (F200W), since its 1.8–2.3$\micron$ coverage has never been available for any slitless spectroscopy prior to to JWST (Colbert et al., 2024). The emphasis was on line pairs close in wavelength, viewed at higher redshifts: H$\beta$+[OIII] and H$\alpha$+[SII]$\lambda$6716,6731Å.

Contrary to a widely held perception, confusion due to spectral overlap is not generally very serious during the assembly of this sample of NIRISS slitless spectra. This is because the actual spectra of all objects in a typical high-latitude field cover only a small fraction of the detector area. The low spectral resolution results in spectra that are usually confined to only 6 x 60 = 360 pixels, which is only 1/10,000 of the total Teledyne H2RG detector area. Therefore, in most PASSAGE observations, only $\leq 15\%$ of the spectrograms suffer from serious overlap with a spectrum of another object. The second-order spectrograms are longer but about one hundred times fainter than the first order spectrograms. Thus, the only second order spectra that cause significant contamination tend to be from bright galactic stars, which is noted in our output files. Since we avoided fields of low galactic latitude, this contamination is very minor.

3 Data Reduction

Reduction of grism data is an important challenge not only for JWST, but also other ongoing and future missions (Euclid, Roman, Euclid Collaboration et al., 2023; Bagley et al., 2020). Getting the most out of NIRISS/WFSS observations requires extensive, careful data reduction efforts. For the preliminary reduction of our data, we use the Grism Redshift & Line analysis (Grizli222https://grizli.readthedocs.io, version 1.9.5; Brammer, 2023) software, which is a quantitative and comprehensive modeling and fitting tool for slitless spectroscopic observations. The details of our reduction will be fully described in a future publication, but we include a brief description here.

Grizli incorporates the default STScI data reduction pipeline to apply the initial detector-level corrections (performed by Detector1 stage of the STScI pipeline333https://jwst-pipeline.readthedocs.io) which includes a variety of steps such as linearity correction, persistence correction, flat-fielding, masking of bad pixels and cosmic rays, dark subtraction, and ramp fitting. Grizli further incorporates a custom algorithm for “snowball” masking444https://github.com/mpi-astronomy/snowblind (snowblind; Davies, 2024). This is followed by Grizli’s preprocessing step, which registers the WCS, performs astrometric alignment, subtracts sky background, and drizzles each exposure on a common reference frame to provide fully reduced individual exposures as well as mosaics for both the direct and dispersed images. All of our images are aligned to the Legacy Survey’s DR9 (Dey et al., 2019) astrometry. The next step is to identify all sources by running “Source Extraction and Photometry” (SEP555https://sep.readthedocs.io; Barbary, 2016; Bertin & Arnouts, 1996) on a detection image typically created by combining all filters where a direct image is available. The absolute flux calibration is performed on individual spectra using the calibration files provided as part of the NGDEEP calibration (Pirzkal et al., 2024; Ravindranath et al., 2023).

In order to model the overlapping spectrograms, which present one of the most challenging aspects of slitless spectroscopy, Grizli uses the segmented images of all detected sources (brighter than $m_{AB}$ ranging from 26 to 30, chosen depending on the depth of the direct imaging of the field), and models their associated spectrograms in each individual grism exposure. The pixels associated with an overlapping (contaminating) source are down-weighted when modeling spectral energy distributions (SED) of the primary source of interest and extracting the final grism spectra. Lastly, Grizli generates the 2D dispersed spectra by simultaneously modeling each grism exposure, allowing for independent noise and other features such as differences in morphological broadening per exposure. The final 1D spectra generated for each individual object are then modeled using a set of SED templates from Eazy666https://eazy-py.readthedocs.io to assist in redshift identification.

The 0th order contamination is of concern since it can masquerade as an emission line, as discussed below. The 2nd order is only prominent for bright stars and is therefore relatively rare. When modeling the contamination, Grizli flags contributions from both the 0th and 2nd orders, and these pixels are appropriately excluded when fitting the galaxy spectra.

For the brightest objects we measure, the systematic uncertainty in the flux scales becomes a significant additional source of error. Judging from the agreement of independent but overlapping spectra of the same object, our absolute fluxes may sometimes be systematically uncertain by up to 20%. The fluxing uncertainty is worse at the blue and red edges of each spectrum. Grizli does not accurately correct the fluxes where the grism transmission drops below 50%. In these small edge regions, we do not attempt to measure emission line fluxes because they are not trustworthy. Therefore we restrict our flux measurements to these reliably-fluxed spectral windows: F115W: 1.013 –- 1.283 $\micron$; F150W: 1.330 –- 1.671$\micron$: and F200W: 1.751 –- 2.240$\micron$ (see also Figure 4 and the graphical illustrations of the spectral windows in the one-dimensional plots of Figures 9–14). We warn other users of NIRISS that grism spectroscopy fluxes beyond these wavelength regions should not be used.

Experience with NIRISS grism spectra reduced by the Grizli software package showed that it picks up a significant number of artifacts (e.g., emission lines from other nearby sources, diffraction spikes, other image artifacts, modeling issues with the continuum, overlapping spectra or zeroth orders, etc.). These were incorrectly identified as real sources, which must be removed by (our) human intervention. Another problem is that when spectra of two objects partially overlap, Grizli’s attempt to automatically remove the contamination often leads to over-subtraction. This usually causes parts of the overlapping spectra to (artificially) drop below zero flux. Efforts to replace some of this labor intensive work with computer algorithms are ongoing.

4 PASSAGE Science Examples With Faint Emission-Line Galaxies

The resulting PASSAGE spectra are of great value for investigating many questions of strong current interest in research on galaxy evolution. Our predicted number of emission line-determined redshifts for the entire PASSAGE survey is shown in Figure 5. We now provide a few examples of relevant emission-line galaxy properties at intermediate redshifts that we are measuring with PASSAGE data.

4.1 Star Formation Main Sequence

According to equilibrium growth models (Lilly et al., 2013; Tacconi et al., 2020), galaxy evolution is controlled by cosmic accretion, merging, galactic gas depletion and star formation, and gas recycling into the circumgalactic medium. Models of these mechanisms aim to reproduce the observed scaling relations between galaxy stellar mass, star formation rate (SFR), metallicity, gas content, size, and structure (e.g., Noeske et al., 2007; Maiolino & Mannucci, 2019; Saintonge et al., 2011; van der Wel et al., 2014; Nedkova et al., 2021), along with their evolution over cosmic time. Observed relations appear to show large scatter, whose origin is debated. This can be modeled by short-term stochasticity, long-term differences in growth histories or gas fractions, or different scenarios for feedback from AGN and star formation (Lilly et al., 2013; Ma et al., 2016; De Rossi et al., 2017; Torrey et al., 2018, 2019; Davé et al., 2017, 2019). At low masses, it is unknown whether these relations change or break down completely (Atek et al., 2014a; Alavi et al., 2016; Amor´ın et al., 2017).

Unlike massive galaxies, low-mass ones often have high specific SFRs (normalized by stellar mass), with extremely strong high-ionization emission lines, elevated ionizing power, and perhaps low dust content. To test and constrain the stochastic star formation histories of hydrodynamical simulations (Shen et al., 2014; Dom´ınguez et al., 2015; Sparre et al., 2017; Boyett et al., 2024), we are deriving SFRs from the Balmer emission lines, which reflect the star formation activity on a few Myr timscale. The SFR is derived from the H$\alpha$ emission line, following the calibration of Kennicutt & Evans (2012). Where available, the Balmer line ratios(e.g., H$\alpha$/H$\beta$ allow us to estimate the effects of dust redenning. In our representative medium field, we measured the Balmer decrement in a sample of 122 galaxies. We derived a median value of E(B-V) ranging from $0.01$ to $0.02$, using the Cardelli et al. (1989) attenuation law or the SMC extinction law (Gordon et al., 2003), respectively.

Figure 6 presents the star-formation main sequence (SFMS) of galaxies at $z\sim 1-2$ in a representative medium-depth field. Stellar masses were estimated via SED fitting, using a combination of our PASSAGE broadband photometry and COSMOS-Web imaging, which covers this field (see Section 4.2). Out of the 63 PASSAGE fields, 18 have overlapping coverage with COSMOS-Web (Figure 2).

Photometric fluxes were obtained within 1″ apertures across the following bands: HST/ACS F814W; JWST/NIRISS F115W, F150W, F200W; JWST/NIRCam F115W, F150W, F277W, F444W; and JWST/MIRI F770W. We used the Bagpipes software(Carnall et al., 2018) to perform the SED (spectral energy distribution) fitting. We adopted a non-parametric star-formation history using the continuity prior of Leja et al. (2019), with five logarithmically-spaced age bins (fixing the most recent to cover 0–30 Myr). The metallicity was allowed to vary between 0.01 and 2.5 $\mathrm{Z}_{\sun}$, and the logarithm of the ionization parameter between $-3.5$ and $-2$. Nebular emission was included for populations as old as 20 Myr, and dust attenuation was parameterized following Cardelli et al. (1989), with $R_{V}=3.1$, and $A_{V}$ varying between zero and two magnitudes. Star formation rates, SFRHα, were computed using the Kennicutt & Evans (2012) calibration and are corrected for the small dust attenuation inferred from their Balmer decrements.

Our results in figure 6 show that firstly, at the higher redshifts of many PASSAGE galaxies ($z>1$), they tend to have higher specific star formation rates, resulting in higher H$\alpha$ equivalent widths. Secondly the observed SFMS increases towards lower masses and higher H$\alpha$ equivalent width, therefore PASSAGE is particularly sensitive to very low-mass galaxies which are undergoing intense bursts of star formation.

4.2 Dust Reddening

Dust attenuation traced by the Balmer decrement, H$\alpha$/H$\beta$, has been shown to correlate with various galaxy properties such as stellar mass, gas-phase metallicity, SFR (Garn & Best, 2010; Zahid et al., 2014; Shapley et al., 2022; Battisti et al., 2022) and more recently with velocity dispersion (Maheson et al., 2024). Among these properties, the primary dependency has been found to be with stellar mass (Maheson et al., 2024). Recently, Shapley et al. (2023) used JWST/NIRSpec data from the CEERS program to study the dust attenuation-stellar mass relation at higher redshifts $z=3-6$ (Roberts-Borsani et al., 2024). However, the positive correlation between dust attenuation and stellar mass has mostly been studied in galaxies with $M_{\ast}>10^{9}M_{\odot}$. Existing studies of dust attenuation in lower mass galaxies are primarily based on stacks of low-resolution, low-SNR spectra (e.g., Dom´ınguez et al. 2013; Battisti et al. 2022).

PASSAGE is obtaining hundreds (and ultimately a few thousand) of individual “gold standard” gas extinction measurements using the H$\alpha$/H$\beta$ ratio at $1.1\lesssim z\lesssim 2.4$, down to stellar masses of a few $10^{7}\ \text{M}_{\odot}$, $10\times$ less massive than current measurements (Reddy et al., 2015; Dom´ınguez et al., 2013; Battisti et al., 2022), as seen in Figure 7. To account statistically for blending of [NII] with H$\alpha$, we can rely on established calibrations for the [NII]/H$\alpha$ ratio as functions of stellar mass and redshift (Faisst et al., 2018), which are calibrated up to $z\sim 2.5$, and have been used previously for the purpose of measuring Balmer decrements from low spectral resolution grism data (e.g., Battisti et al., 2022). We note that the [NII] correction is not applied to the PASSAGE measurements in Figure 7 as most galaxies have low stellar masses, where [NII] contribution is $<10\%$ of H$\alpha$ flux (Faisst et al., 2018). PASSAGE detects large numbers of extreme emission line galaxies (EW [OIII] $>300\text{\AA }$). Their starlight continuum is extremely faint, suggesting stellar masses as low as log $M_{*}/M_{\odot}\sim 7$ that have been missed by most previous surveys. Figure 7 shows the correlation of gas extinction (from the Balmer decrement) with stellar mass. Most PASSAGE dwarf galaxies have negligible extinction, but a notable minority are substantially dusty. We note that the individual PASSAGE galaxies shown here are selected to have at least a $3\sigma$ detection in H$\beta$, which can bias the sample toward lower attenuation. This is evident in the figure, where many galaxies show negligible or zero attenuation. The Balmer line ratios used here are corrected for Balmer absorption lines from the atmosphere of stars, mainly A-type. We used an average stellar absorption corrections of $1.4\%$ and $11.8\%$ for H$\alpha$ and H$\beta$, respectively (Alavi et al. 2025 (submitted) Reddy et al., 2015).

For fields with ancillary photometric data, we can combine Balmer decrement with SED model fits to this photometry, to derive average dust attenuation curves (Reddy et al. 2015; Battisti et al. 2022). Such data are already available for 18 PASSAGE fields (29%) located within the COSMOS footprint, and more will be available in the future from follow-up programs with $HST$ (7 non-COSMOS fields; PI:Mehta, #17461), VLT/FORS2 (9 non-COSMOS fields; PI:Battisti, 114.27EF.001, 115.27WC.001/115.27WC.003), and Keck/LRIS (2 non-COSMOS fields with grism data in all three filters; PI:Nedkova, 64/2025A_N067). This is illustrated in Figure 8, which contrasts examples of an emission-line galaxy (top) with little if any extinction in either its starlight or gas, with another galaxy (bottom) having substantial continuum and emission-line reddenings. We will further constrain the optical shape of the dust extinction curve using higher-order Balmer lines (via stacking; eg., Reddy et al. 2020).

At lower redshifts ($z\leq 0.7$), our longest wavelength spectral coverage includes the Pa$\beta$ recombination line. We show the ratio of H$\alpha$ to Pa$\beta$ for individual galaxies in Figure 21. Not surprisingly, no galaxies have observed ratios comparable to the intrinsic ratio predicted by pure Case B recombination (top dotted line). Instead, as expected, many galaxies have significantly weaker H$\alpha$, which is in most cases consistent with the standard expectation of 1 magnitude of absorption at H$\alpha$ (Kennicutt, 1983). Interestingly, a significant fraction (4 out of 8) of the galaxies have H$\alpha$ even considerably weaker, indicative of 2 or more magnitudes of dust extinction.

4.3 Extreme Emission Line Galaxies

The slitless spectroscopy of PASSAGE is able to detect emission lines independently of the underlying continuum brightness. Thus in contrast to typical magnitude-limited spectral surveys, PASSAGE selects an emission-line flux-limited sample of galaxies, rather than a continuum-flux limited sample. PASSAGE can measure galaxies with emission lines of extremely high-equivalent widths, even when the continuum is very faint. This results in a large new sample of so called “extreme emission line galaxies”, which have line equivalent widths exceeding 200 \textÅ to more than 1000 \textÅ (Atek et al., 2014b, 2011). Such extreme but faint galaxies tend to be excluded from traditional spectroscopic surveys (Chen et al., 2024).

Figures 10, 12, 14, 16, and 18 show examples of PASSAGE extreme emission line galaxies (EELGs). They have H$\alpha$ emission lines with rest-frame equivalent widths of a few hundred Angstroms or more, indicative of extremely high specific star formation rates. This corresponds to extremely high specific star formation rates larger than $10^{-8}$ $yr^{-1}$. These imply that much of the star formation in these galaxies happened within a hundred million years of the epoch at which we observe them (Atek et al., 2014a). The low spectral resolution of the NIRISS grism blends in the [NII]6584/6548Å doublet with H$\alpha$. However we can remove the [NII] contamination with statistical corrections, which indicate that it is a small contributor in typical EELGs. The corresponding timescale to double their stellar masses may be only $\sim 10^{8}$ years (Atek et al., 2014a; Boyett et al., 2022). Furthermore, their intense bursts of star formation are often concentrated into extremely small volumes (Atek et al., 2011). It is not always easy to classify these highly disturbed dynamical systems into individual galaxies, or assembling galaxy fragments.

PASSAGE finds as many as a hundred such extreme H$\alpha$-emitters per field, out to redshifts of $z\sim 2.5$. For example Figure 6 shows that galaxies with EW of H$\alpha>200\text{\AA }$, the light blue and dark blue points dominate the PASSAGE sample of spectra in this single field. Figure 6 also illustrates that larger numbers of extreme emission line galaxies can be inferred from the detection in PASSAGE spectra of [OIII] line emission of very high equivalent width out to $z\sim 3.5$. And as illustrated by the example spectra in Figure 20, the strong UV-rest frame emission lines of quasars can be seen in PASSAGE at higher redshifts, out to $z>5$.

4.4 Gas-Phase Mass-Metallicity Relations

The mass-metallicity relation (MZR) is the correlation between galaxies’ stellar masses and their gas-phase metallicities. Constraining this relation over large ranges of stellar mass, and its evolution with redshift, provides crucial constraints on models of gas accretion, star formation, metal enrichment, and outflows, which are responsible for regulating galaxy growth. PASSAGE includes many low-mass galaxies, reaching largely unobserved ranges of stellar mass, gas fraction, and redshift (Figure 22, Sanders et al. 2018; Henry et al. 2021; Sanders et al. 2021; Wang et al. 2022; Curti et al. 2023; Revalski et al. 2024). At low stellar masses and high redshifts, outflows may preferentially remove metals (Henry et al., 2013), thus making the low mass MZR slope a key diagnostic of the role of gas recycling at different epochs, quantifying the role of feedback and of the stochastic processes in regulating metallicity. PASSAGE, due to its large sample size, is also measuring the scatter in the MZR for low-mass galaxies (see e.g., Torrey et al., 2019).

Prior to JWST, these regimes had not been constrained because of the low number and quality of spectra below 10${}^{9}M_{\odot}$, which required stacking. More recent results with NIRISS (Li et al., 2022; He et al., 2024) have shown the feasibility of metallicity measurements at lower masses, albeit with uncertain values and scatter. PASSAGE will address this by measuring gas metallicities of many individual low-mass galaxies—without requiring stacking—up to $z$=3.5.

For the several thousand galaxies in the redshift range $1.7\lesssim z\lesssim 3.5$ for which the multiwavelength coverage of the PASSAGE grism filters allows simultaneous detection of [OII], [OIII], and H$\beta$ lines, we are able to determine the MZR using strong line diagnostics such as the $R_{23}$ index (e.g., Pagel et al., 1979; Tremonti et al., 2004). In Figure 22, we show the position of 58 individual galaxies from a single deep PASSAGE field on the stellar mass–metallicity plane. These objects are selected to have S/N$\geq 10$ [O iii] detections. Metallicities are obtained using the Curti et al. (2017) strong line calibrations, which have been validated out to at least $z\sim 2.5$ (Revalski et al., 2024) and their stellar masses are derived as described in §4.1.

Due to the long-lasting interplay between the processes that drive the baryon cycle, the metallicity of galaxies has been also observed to anti-correlate with star formation rate at fixed stellar mass. This relation is commonly referred to as the “Fundamental Metallicity Relation” (FMR) (FMR; e.g. Ellison et al., 2008; Mannucci et al., 2010; Salim et al., 2015; Telford et al., 2016; Baker et al., 2023). PASSAGE is now establishing if high redshift galaxies follow the FMR, as at z$$\sim0 (Mannucci et al., 2010).

4.5 Absorption Line Measurements.

Stellar age and abundance measurements from absorption-line spectroscopy generally require rest-frame coverage from 3700Å to 5300Å. PASSAGE obtains this at $1.6<z<2$ for all medium and long visits, and at $2.0<z<3.3$ for long visits which include the F200W spectral band. In each field we typically find around a few spectra of $z=$1.6–2.5 quiescent galaxies which are bright enough ( $m_{F115W}\lesssim 24.5$) to have very high SNRF115W (ten or larger) per spectral pixel. In these cases, although emission lines are either weak or undetected, we can measure diagnostic stellar absorption features. If we restrict consideration to the strongest stellar absorptions, the Balmer and Ca II HK breaks–these can be measured even when the continuum is one to two magnitudes fainter, giving us an average of $\sim 5$ such spectra per field.

4.6 Feasibility of Spectroscopically Identifying and Measuring Ly$\alpha$ lines in High Redshift Galaxies

The total F115W grism exposures useful to search for faint Ly$\alpha$ lines and Lyman alpha forest breaks at $z\geq 8$, range from 8 ksec to more than 20 ksec in the deepest parallel visits. In the simulation shown in the left lower panel of Figure 23, a Ly$\alpha$ emission line is added to a Lyman break galaxy (LBG) spectrum. Even for these faint galaxies, the continuum break and the line are confidently detected, for rest-frame line equivalent widths as small as 50Å.

We have made a careful search of our deepest PASSAGE fields for strong isolated emission lines in the 1.0–1.3 $\mu$m range. This special “by-hand” analysis does not require that any object be initially detected in the broadband images. It has revealed 4 new Ly$\alpha$ emission lines at redshifts of z=7.7–9.5 (Runnholm et al., 2025). Their equivalent widths range from 40 to 100Å. Figure 24 shows a simulated grism spectrum for a Lyman break galaxy emitting stellar continuum only. Although there is no Ly$\alpha$ emission line in the simulation, since there are many observed continuum pixels above and below the Lyman break (at 1.12$\micron$ = 1216Å* 9.2), this sharp break is detected. And it is distinguishable from the gradual turn-downs seen in $z=2$ interloper galaxies, even if they lack any rest-frame optical emission lines. NIRISS slitless spectra have indeed detected Lyman breaks in 2 out of 3 faint $z\sim 8$ galaxies (Roberts-Borsani et al., 2022), with magnitudes and integration times similar to these simulations.

5 Summary and Legacy Impact of PASSAGE

Synergies with other programs and missions. The JWST/NIRISS capability for slitless near-infrared spectroscopy in a large number of parallel fields has opened many new prospects for better understanding galaxy evolution. Our preliminary analysis of the first PASSAGE observations demonstrates that the performance of this mode is as good, or even better, than was anticipated before launch. We show recent NIRISS examples of major improvements in sensitivity and extended long-wavelength coverage, to allow advances in the study of galaxies at higher redshifts and lower stellar masses than were previously attainable. We highlight progress in studying the star formation main sequence, the mass/metallicity correlation, and gas reddening. PASSAGE complements ongoing deep JWST spectroscopy of selected fields and targets, by measuring galaxy spectra without pre-selection in 63 independent fields, to overcome uncertainties due to cosmic variance.

The pure parallels provide a public treasure trove of spectroscopy. This multiplies their future value, allowing for JWST follow-up, as well as complementary data for the Euclid and Roman missions on how to optimize observing strategies. PASSAGE complements those missions by going much deeper in smaller areas of sky, thus conducting science that neither Euclid nor Roman missions can achieve.

The brighter (and rarer) objects discovered are particularly compelling for follow-up by JWST and other facilities. For example, PASSAGE is discovering galaxies bright enough to measure the key [O III] $\lambda$4363 line, to test or improve the calibration of current metallicities at $z\geq 2$ (Gburek et al. (2019); Sanders et al. (2020)). Even if the high-redshift strong-line metallicity calibration does need to be revised, this project provided the essential suite of lines on which it is based.

Legacy Value of PUBLIC fully processed datasets in the Archive. High Level Science Products (HLSPs) are used up to 10 times more than individual calibrated data, and the archive of slitless spectroscopy is sure to provide an unprecedented legacy of lasting scientific value. We will deliver fully processed data and HLSPs to MAST for the astronomy community.

6 Acknowledgments

We acknowledge support by NASA through grant JWST-GO-1571 (PI: Malkan). We are very grateful to staff members at Space Telescope Science Institute who worked hard to schedule pure parallel opportunities for this project, especially to our Program Coordinator Shelly Meyett. We thank the anonymous referee, for making extremely thorough readings and providing many comments and suggestions which have improved this paper. The JWST data presented in this article were obtained from the Mikulski Archive for Space Telescopes (MAST) at the Space Telescope Science Institute. The specific observations analyzed can be accessed via http://dx.doi.org/10.17909/6ca5-ba17 (catalog doi: http://dx.doi.org/10.17909/6ca5-ba17).

XW is supported by the National Natural Science Foundation of China (grant 12373009), the CAS Project for Young Scientists in Basic Research Grant No. YSBR-062, the Fundamental Research Funds for the Central Universities, the Xiaomi Young Talents Program, and the science research grant from the China Manned Space Project. BV, AA, and PW acknowledge support from the INAF Large Grant 2022 “Extragalactic Surveys with JWST” (PI Pentericci) and from the European Union – NextGenerationEU RFF M4C2 1.1 PRIN 2022 project 2022ZSL4BL INSIGHT. PW and BV acknowledge support from the INAF Mini Grant ‘1.05.24.07.01 RSN1: Spatially Resolved Near-IR Emission of Intermediate-Redshift Jellyfish Galaxies’ (PI Watson). AJB acknowledges funding from the “FirstGalaxies” Advanced Grant from the European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation program (Grant agreement No. 789056). HA is supported by CNES, focused on the JWST mission and the Programme National Cosmology and Galaxies (PNCG) of CNRS/INSU with INP and IN2P3, co-funded by CEA and CNES. MB acknowledges support from the ERC Grant FIRSTLIGHT and from the Slovenian national research agency ARRS through grants N1-0238 and P1-0188. MJH is fellow of the Knut & Alice Wallenberg Foundation.