Studying [CII] Emission in Low-mass Galaxies at 𝑧 ∼ 7

We report on a [ CII ] 158 𝜇 m search using the Atacama Large Millimeter/submillimeter Array (ALMA) on three lensed, confirmed Ly 𝛼 emitting galaxies at 𝑧 ∼ 7. Our targets are ultra-violet (UV) faint systems with stellar masses on the order of 𝑀 ∗ ∼ 10 9 𝑀 ⊙ . We detect a single [CII] line emission (4 𝜎 ) from the brightest ( 𝐿 ∼ 2 . 4 × 10 10 𝐿 ⊙ ) galaxy in our sample, MACS0454-1251. We determine a systemic redshift ( 𝑧 [ CII ] = 6 . 3151 ± 0 . 0005) for MACS0454-1251 and measure a Ly 𝛼 velocity offset of Δ 𝑣 ≈ 300 ± 70km s − 1 . The remaining two galaxies we detect no [CII] but provide 3 𝜎 upper limits on their [CII] line luminosities which we use to investigate the 𝐿 [CII] − SFR relation. Overall our single [CII] detection shows agreement with the relation for local dwarf galaxies. Our [CII] deficient galaxies could potentially be exhibiting low metallicities ( 𝑍 < 𝑍 ⊙ ). Another possible explanation for weaker [CII] emission could be strong feedback from star formation disrupting molecular clouds. We do not detect continuum emission in any of the sources, placing upper limits on their dust masses. Assuming a single dust temperature of 𝑇 𝑑 = 35K dust masses ( 𝑀 dust ) range from < 4 . 8 × 10 7 𝑀 ⊙ to 2 . 3 × 10 8 𝑀 ⊙ . Collectively, our results suggest faint reionization era sources could be metal poor and/or could have strong feedback suppressing [CII] emission


INTRODUCTION
In the past decade, the Atacama Large Millimeter/submillimeter Array (ALMA) observations of metal fine structure lines such as the [CII] 158m line have opened up studies into the epoch of reionization (EoR;  > 6) by providing an unobscured view of galaxies.With the advent of the James Webb Space Telescope (JWST), which can also provide a similar a view of non-resonant optical lines (e.g., H), ALMA observations still provide a complementary view for far infrared (FIR) emission lines.FIR lines hold immense value for EoR studies because they are not affected by dust extinction, in contrast to rest-optical lines accessible with JWST.There have been numerous [CII] detections in UV-bright, high- ( > 6) galaxies (e.g., Willott et al. 2015;Carniani et al. 2017;Laporte et al. 2017;Smit et al. 2018a;Matthee et al. 2019;Harikane et al. 2019;Bakx et al. 2020).However, there are considerably fewer recorded [CII] detections for characteristically faint,  <  * ∼7 (where  * is the characteristic luminosity) EoR galaxies (e.g., Schaerer et al. 2015;Watson et al. 2015;Knudsen et al. 2016;Bradač et al. 2017;Fujimoto et al. 2021;Laporte et al. 2021).Faint galaxies are much more numerous than bright ( <  * ∼7 ) galaxies (e.g.Bouwens et al. 2022;Bolan et al. 2022) and, as such, can be a key driver of reionization.However, this connection has been heavily debated (e.g.Finkelstein et al. 2019;Naidu et al. 2020;Robertson 2022;Endsley et al. 2023).In order to resolve this question, we need to study the physical properties of these fainter primordial systems; this step is key to understanding their role in cosmic reionization.
Here we use [CII] observations to study  ∼ 7 galaxies.The [CII] line is of particular interest because it predominantly traces the dense neutral gas in photodissociation regions (PDRs, Wolfire et al. 2022) associated with molecular clouds, and the diffuse neutral gas (Wolfire et al. 2003;Hollenbach & Tielens 1999).It is the most luminous line in the FIR band (∼ 0.1 − 1% of the FIR luminosity, Baier-Soto et al. 2022) and one the strongest emission lines of starforming galaxies at FIR/radio wavelengths (Carilli & Walter 2013;Stacey et al. 1991Stacey et al. , 2010)).Additionally, the [CII] line can be used to trace the systemic redshift, and therefore velocity, of the host galaxy because it is optically thin and not affected by dust extinction.When paired with detected Ly emission, we can estimate interstellar medium (ISM) properties of the galaxy under some assumptions.For instance, we can estimate the amount of neutral gas assuming the offset stems from Ly resonant scattering in ISM gas (Mason et al. 2018). 1 Ly photons travelling through neutral ISM scatter more frequently and emerge with a larger Δ compared to photons traveling through less neutral ISM (Yang et al. 2016(Yang et al. , 2017;;Guaita et al. 2017).Previous observations of  > 5 galaxies have typically measured Δ ≲ 500kms −1 (Cassata et al. 2020;Carniani et al. 2018a;Pentericci et al. 2016;Bunker et al. 2023;Prieto-Lyon et al. 2023) with the largest Δ ≈ 1000kms −1 recorded by Baier-Soto et al. (2022).Cassata et al. (2020) find that galaxies (4.4 <  < 6) with smaller Δ have larger Ly rest-frame equivalent widths and  esc (Ly).However, for intrinsically faint systems ( UV ≳ −20.5) at  > 6, a limited sample size restricts our ability to make robust conclusions (see Endsley et al. 2022 and references therein).
As mentioned previously, majority of EoR studies investigating [CII] have targeted UV-bright galaxies, which traditionally have higher SFRs.These larger and brighter galaxies are inherently easier to observe but do not exemplify galaxies at this epoch.Lensing, in our case by foreground galaxy clusters, serves as an excellent way to increase the apparent brightness and size of these fainter systems.In particular, the flux of such objects is increased by a magnification factor  and resolution improved by √ .In this paper we report on a [CII] study carried out with ALMA that targeted a set of lensed, sub- * galaxies with SFRs<20 ⊙ yr −1 .The lensed sizes of our galaxies are compact and smaller than the beam sizes of observations.All objects in our sample are spectroscopically confirmed EoR galaxies via detected Ly emission.We acknowledge a sample of galaxies all with detected Ly emission could potentially bias our results.However, recent results by Bolan et al. in prep find no significant difference in stellar UV properties (e.g, stellar mass, star formation rate, specific star formation rate, UV magnitude, -slope, age) between Ly emitters and non-emitters at 5 <  < 8.2 in the similar stellar mass range to the one studied here.
The paper is organized as follows.Section 2 explains various ob- 1 We remind that outflows might also generate comparable velocity offsets.servations used to compile our sample and the specific ALMA data reduction performed on our data.In Section 3 we discuss the measurements and derived properties from the ALMA data as well as the spectral energy density (SED) fitting performed for galaxy property estimates.In Section 4 we analyze and discuss our results with literature findings.In Section 5 we summarize our conclusions.Throughout this paper we assume a ΛCDM concordance cosmology with Ω  = 0.27, Ω Λ = 0.73 and Hubble constant  0 = 73km s −1 Mpc −1 .
Coordinates are given for the epoch J2000.0,magnitudes are in the AB system, and we use the Chabrier (2003) initial mass function (IMF).

OBSERVATIONS AND DATA REDUCTION
The lensing galaxy clusters of our sample have been extensively studied in past works.All three clusters have imaging from the Hubble Space Telescope (HST).
Values have been corrected for lensing using  best .
Values derived from SED fitting described in Section 3.
ALMA observations (Proposal ID: 2019.1.00003.S) were carried out between March of 2020 and April 2021 using ALMA Band 6 with 43 12-m antennae in array configurations C43 − 4 and C43 − 5.The precipitable water vapor (PWV) ranged from 0.5mm to 2.1mm.The spectral setup consisted of one spectral window centered on the expected observed frequency of [CII] 158m estimated from the Ly redshift.The remaining two spectral windows were used for continuum measurements.The on-source time varied from 20 to 35 minutes per target.
The data was reduced and calibrated with the Common Astronomy Software Applications (CASA) package, version 6.1.1.15following standard procedures.We reimaged the data with the CASA task TCLEAN, adopting Briggs weighting with ROBUST=2.0(effectively natural weighting) and adding UVTAPER=0 ′′ .6.We adopt the taper = 0 ′′ .6,given the typical [CII] effective radius of ∼ 2kpc at  ∼ 4 − 6 (Fujimoto et al. 2020) which corresponds to 0 ′′ .3.The angular resolution for each object is equal to their respective beam sizes which increased globally by a factor of ∼ 4 when the taper was applied.The RMS also increased when the taper was applied by a factor of ∼ 1.56 (MACS0454-1251), ∼ 1.36 (MACS2129-1412), and ∼ 1.13 (RXJ1347-018) respectively.The final beam sizes and RMS values are quoted in the captions of the contour figures.All three targets were unresolved in their final data products and were spectrally binned into 25-km s −1 velocity channels.To ensure positional accuracy, we astrometrically calibrated HST reference images to GAIA DR2 when comparing [CII] to UV emission.Continuum maps were made with all spectral windows.
The remaining two galaxies, MACS2129-1412 and RXJ1347-018, yielded non-detections for [CII] emission (see Fig 3).Because their intrinsic [CII] line widths are unknown, we estimate their  [CII] (3) upper limits by integrating over the same width of channels ( 9 channels, 225km s −1 ) as was done in MACS0454-1251.In the absence of known systemic redshifts, we use Ly redshifts to predict expected [CII] emitting frequency of each object2 .The RMS uncertainty was calculated using the 1 intensity value extracted from a 0. ′′ 5 aperture centered to the HST imaging centroid of the target.We list the calculated  [CII] (3) upper limits in Table 2.
We constructed a continuum map for MACS0454-1251 from all four spectral windows (SPWs), masking out the channels expected to contain [CII] emission.As seen in Fig 2, the [CII] line of MACS0454-1254 appears to be extended over roughly nine channels which corresponds to a width of 225km s −1 .All nine channels were masked when producing the final continuum map from which we derived an upper limit on  FIR .We assume a wavelength range of 8 − 1000m for  FIR calculation.For the remaining non-detection galaxies, we constructed continuum maps in the same manner excluding nine channels centered on the central frequency.To estimate  FIR , we used a grey-body spectral energy distribution model (Casey 2012), assuming a spectral index of  = 1.5.In the absence of multi-band observations, we assumed a uniform dust temperature ( d = 35K) across all three objects and estimated 3 FIR which is recorded in table 2. In similar studies where dust continuum was not detected (Fujimoto et al. 2022;Fudamoto et al. 2023;Witstok et al. 2022), assumed dust temperatures varied  d = 40 − 60K.Other studies on mostly massive galaxies (Fujimoto et al. 2021;Bakx et al. 2021;Algera et al. 2024;Witstok et al. 2022;Sommovigo et al. 2022) also have a sizeable range on  d = 32 − 59K with considerable error at low and high ends (note that employing single band techniques to estimate dust temperature was shown to overestimate temperatures by ∼ 15K).Because  FIR and  dust are directly related to the assumed   , assuming a lower  d sets a lower limit on  FIR and a conservative upper limit on  dust .
Because we do not detect continuum in all three galaxies, we provide SFR FIR and  dust limits in Table 2.We estimated SFR FIR using eq.3 from Kennicutt (1998b) and converted from Salpeter IMF to Chabrier IMF.We assumed a dust mass absorption coefficient of  =  0 (/ 0 )  d , where  0 = 0.232m 2 kg −1 at  0 = 250m (Draine 2003;Bianchi 2013).
To estimate age and stellar mass ( stellar ) of the sample, we use Bayesian Analysis of Galaxies for Physical Inference and Parameter EStimation (BAGPIPES, Carnall et al. 2018).BAGPIPES fits physical parameters using the MultiNest sampling algorithm (Feroz & Hobson 2008;Feroz et al. 2009).We use the default set of stellar population templates from Bruzal and Charlot (Bruzual & Charlot 2003, BC03).The SED fitting is done assuming the Kroupa ( 2001) IMF which we convert to Chabrier (2003), a metallicity of 0.02 ⊙ , the Calzetti dust law (Calzetti et al. 2000), and a constant star formation history.We allow dust extinction to range from   = 0 − 3 magnitudes.The  stellar values reflected in Table 1 have been converted to Chabrier IMF via conversion factor 0.923.We take the general prescription for the fitting from Strait et al. (2020).See Bolan et al. in prep.for more information on the SED fitting.

Velocity and Spatial Offset
Observational studies have used [CII] emission as a tracer of systemic velocities (e.g Pentericci et al. 2016;Matthee et al. 2019).Using the peak [CII] emission from the extracted spectrum of MACS0454-1251, we find Δ Ly−[CII] ≈ 320 ± 70km s −1 .This is shown in Fig 2 where the velocity axis is centered on the [CII] emission and the grey dashed line represents Ly.The magnitude of the offset falls within literature ranges quoted for  = 2 ∼ 3 (Erb et al. 2014) and high- galaxies (e.g., Cassata et al. 2020;Endsley et al. 2022).
The spectrum shows a clear redshift of Ly emission.Given the resonant nature of Ly, revealing the direct cause of the offset is both a complex and active place of research.Redshifted Ly could indicate scattered emission from outflowing (or expanding) gas from the galaxy.Outflows may originate from strong star formation feedback which could reduce the covering fraction of neutral gas in the ISM and boost Ly escape (Jones et al. 2013;Trainor et al. 2015;Leethochawalit et al. 2016).Ly could also be redshifted by neutral hydrogen inside a galaxy's ISM, where the emerging Δ would be a proxy for the column density of neutral hydrogen (Yang et al. 2016(Yang et al. , 2017;;Guaita et al. 2017).Additionally, the model put forth by Mason et al. (2018) used Δ as a way to measure the intergalactic medium (IGM) neutral fraction.A more neutral IGM will scatter Ly photons more causing Ly to emerge at a higher Δ relative to systemic.
It is not uncommon for  > 5 galaxies to have [CII] emission 21 h 29 m 24.5 s 24.4 s 24.3 s 24.2 s 24.1 s 24.tracing the systemic redshift of the galaxy but spatially offset from the UV component (Maiolino et al. 2015;Willott et al. 2015;Capak et al. 2015;Carniani et al. 2017Carniani et al. , 2018b,a;,a;Jones et al. 2017;Matthee et al. 2019;Fujimoto et al. 2022).In fact, spatial offsets between Ly and UV have also been found at  > 5 (Hoag et al. 2019a;Lemaux et al. 2021).Carniani et al. (2018a) shows that most of the [CII] spatial offsets are indeed physically motivated but further observations are needed to understand the mechanisms causing the offsets.Recently, Fujimoto et al. (2022) was able to determine the necessity of past outflow activity in a galaxy at  ∼ 8.5 based on dual observations with ALMA and JWST.We roughly estimate the [CII]-UV spatial offset in MACS0454-1251 using the brightest pixels found in the [CII] moment0 map and the centroid of the HST rest-UV image.The lensed spatial offset is ∼ 0. ′′ 5 which is greater than the ALMA astrometric accuracy of ∼ 0. ′′ 08 3 .Taking into account the magnification, we estimate the lens-corrected offset to be ∼ 1.4kpc 4 .The offset could be physically associated with intrinsic ISM properties (e.g.different distribution in the ionized vs neutral gas phase).A possible explanation for the spatial offset could also be the ejection of material by galactic outflows or galaxy mergers (e.g., Maiolino et al. 2015;Vallini et al. 2015;Pallottini et al. 2017;Katz et al. 2017;Gallerani et al. 2018;Kohandel et al. 2019).

𝐿 [CII] − SFR Relation
In Fig 4 we show the  [CII] − SFR UV relation for our galaxies (stars) alongside available  > 6 observations from the literature.The reported SFR UV values for our objects can be found in Table 1 along with SFR SED values that were derived through SED fitting described in Section 3. SFR UV is calculated assuming no dust attenuation and therefore should be considered a lower limit as it does not include obscured star formation.We also recognize that in the event of a more top heavy IMF (i.e., the median of the mass-to-light ratio of stars being born decreases), SFR UV would be an overestimate assuming constant dust properties.
Our single [CII] detection, MACS0454-1251, falls within the 1 scatter of the De Looze et al. ( 2014) relation for dwarf (∼ 0.2 dex below).We note, however, MACS0454-1251 is the most UV luminous galaxy ( = 0.74 * ) in our sample.The detection is also consistent with the C1 model put forth by the Vallini et al. (2015) simulations which corresponds to a galaxy with constant solar metallicity.If we compare this to our non-detections, the upper limit set by RXJ1347-018 also is consistent with the De Looze et al. ( 2014) relation; having a scatter of ∼ 0.3 dex from the average.The upper limit set by MACS2129-1412 is not consistent with the De Looze et al. ( 2014) relation, falling ∼ 0.6 dex from the dwarf galaxies.The one displayed Vallini et al. (2015) model it could possibly support is the C005 model which represents a galaxy with a constant metallicity of  = 0.05 ⊙ .
The analysis presented here focuses on the SFR UV component, but SFR SED as well as upper limits on SFR FIR are provided for completeness.The SFR UV /SFR FIR limits are not very constraining for our galaxies given we only have upper limits on SFR FIR making us unable to exclude any model considered in this work.We do note the SFR FIR limit for MACS2129-1412 shows SFR FIR < SFR UV .For the  dust / * , all three of our galaxies showed limits log  dust /log  * < ∼ 0.8.Our limits are consistent within 1 of the overall values reported in Sommovigo et al. (2022) (log  dust /log  * ≈ 0.74 − 0.81) who studied 13 higher-mass galaxies at  ∼ 7. Our limits are also consistent with the dust build up reported in Pozzi et al. (2021).
We also advise caution when drawing comparisons between our objects and the dwarf relation in De Looze et al. (2014).As shown in Fig 4, majority of dwarf galaxies (gray diamonds) used to make the best-fit relation (gray dotted line) are much lower luminosity systems than those reported in this paper as well as literature.Above SFR ∼ 1M ⊙ yr −1 the dwarf relation is less constrained, being driven by only a handful of dwarf galaxies.
The solar and sub-solar metallicity models by Vallini et al. (2015) provide a unique view on the metal content of our galaxies.The possibility of a sub-solar metallicities is not surprising for our three low-mass galaxies and aligns with recent work (Curti et al. 2023;Nakajima et al. 2023).Although Curti et al. (2023) probed a moderately lower mass range than our sample, they found their  > 6 galaxies exhibiting sub-solar metallicities ( <  ⊙ ), with a scaling relation drawn near ∼ 0.1 ⊙ .A more comparable study by Nakajima et al. (2023) reported  <  ⊙ for a sub-sample of galaxies ranging from  = 6 − 10 with a somewhat larger range of masses.While we do not have the observations to determine the true metallicities of our sample, a low metallicity could explain the absence of [CII] in our non-detections and would not contradict their respective upper limits.Another possible reason for the lack of detected [CII] emission could be negative feedback disrupting molecular clouds (MCs, Vallini et al. 2015).For example, [CII] emission predominately originates from PDRs.Negative feedback disrupting MCs would reduce the PDR layer at the edge of the MC where most of the [CII] emission comes from.Additionally, we know stellar feedback is efficient in galaxy centers, places where there is very active star formation.In fact, studies done at lower- have supported a connection between feedback efficiency and SFR surface density (e.g., Hayward & Hopkins 2016;Heckman et al. 2011).Since high- galaxies are more compact than local analogs, this would also support negative feedback suppressing [CII] emission at the systemic redshift of the galaxy.

CONCLUSIONS
We present new ALMA observations investigating [CII] 158m line emission in three lensed galaxies at  > 6.We find a 4 [CII] detection in our most luminous galaxy ( ∼ 0.7 * ) MACS0454-1251 and calculate the systemic redshift to be  [CII] = 6.3151 ± 0.0005.We measure a Δ = 320 ± 20km s −1 and calculate  [CII] = 1.5 +0.5 −0.4 × 10 8  ⊙ .For the remaining two galaxies we do not detect [CII] emission but provide 3 upper limits for their  [CII] .Our main findings are: (i) MACS0454-1251 exhibits both a velocity and a spatial offset between between ALMA [CII] and rest-frame HST UV emission.The spatial offset is larger than the the ALMA astrometric uncertainty which could mean the offset is physically motivated by intrinsic ISM properties.
(ii) Feedback is a very important process in both the emission of [CII] and Ly.On one hand, strong feedback can suppress [CII] emission by disrupting MCs (Vallini et al. 2015).On the other hand strong feedback from star formation can drive outflows which redshifts the Ly line.A possibility that puts together the spatial offset and the [CII] deficit is that feedback destroys the emitting MCs at the center, allowing only displaced MCs to contribute to the [CII] emission.
(iii  2014) for  > 6 galaxies.That being said, the upper limits set by our RXJ1347-018 and MACS2129-1412, would argue an inconclusive result, especially with the 3 upper limit of MACS2129-1412 still falling well below De Looze et al. (2014).In general, more observations of low-mass, UV faint galaxies are needed in order to break this degeneracy.
(iv) Low metallicity is a possible justification for fainter galaxies falling below the De Looze et al. ( 2014)  [CII] − SFR relation (Vallini et al. 2013;Ferrara et al. 2019).Based on recent work (Curti et al. 2023), we would expect lower mass, ( stellar ⪅ 10 9.5 ) galaxies at  > 6 to exhibit sub-solar metallicites.A possible scenario for [CII] deficient systems with very low metalicity is powerful feedback; involving the destruction of star forming sites occurring during bursty evolutionary phases in relatively chemically unevolved systems (Vallini et al. 2015;Ferrara et al. 2019).

Figure 1 .
Figure 1.MACS0454-1251 velocity-integrated [CII] line intensity overlaid on a HST/WFC3 F160W image.The contours are shown in red and are spaced linearly by intervals of 1 which range from 2, 3, and 4 (RMS = 87.1mJy/beam).The beam size ( 0 ′′ .88× 0 ′′ .82) is given in the bottom right with a 2 ′′ × 2 ′′ zoom-in shown in the bottom left.The galaxy is located at an RA and DEC of 04:54:08.56,−03:00:14.82(UV centroid) with blue cross hairs marking the target in the center.The direction of shear is shown by the green line.

Figure 2 .
Figure 2. Extracted spectrum showing the detected [CII] emission.Flux (left axis) shown as a function of frequency (bottom axis) and velocity (top axis).Red dashed line represents the best fit Gaussian and grey dashed line represents Lyman- redshift with boxed 1 uncertainty.
Our single [CII] detection in MACS0454-1251 falls within the 1 scatter of the De Looze et al. (2014)  [CII] − SFR relations.As such, it would support with the applicability of the  [CII] − SFR relation put forth by De Looze et al. (

Table 1 .
Treu et al. 2015)12)and Supernova Survey with Hubble (CLASH,Postman et al. 2012)program observed MACS2129 and RXJ1347.The two CLASH clusters were also spectroscopically observed with The Grism Lens-Amplified Survey from Space (GLASS ,Treu et al. 2015)program.MACS0454 was im-Properties of observed galaxies.
All limits are 3