We present two uniformly observed spectroscopic samples of Ly
The epoch of reionization is a key time in the evolution of the Universe, when the intergalactic medium (IGM) transitioned from being neutral to being very highly ionized. As we discuss below, it is likely that reionization is patchy and that the ionization begins with HII regions around the ionizing sources, which then expand and merge. However, our understanding is still quite limited. Intensive simulations are currently being carried out to model this evolution, but there are many issues that can only be treated in an empirical fashion. Perhaps most critically, we still do not know which objects were the source of the photons that reionized the IGM, nor the exact redshift range when reionization occurred. Answering these questions is a major goal of observational cosmology and fundamental for simulating galaxy evolution.
There has been enormous recent progress in identifying populations of high-redshift galaxies, in particular Lyα emitters (LAEs) at z 6. These samples have the potential to constrain when reionization occurred, since both the strength and shape of the Lyα line can be modified by the radiative damping wings of neutral hydrogen in the IGM (e.g., Haiman 2002;Haiman & Cen 2005;Mason & Gronke 2020). As the neutral hydrogen fraction in the IGM increases with increasing redshift, we expect the Lyα emission lines of LAEs to become narrower and less luminous, and only the red wings of the lines to be visible. At the highest redshifts, we will no longer see the Lyα line at all, except in the most highly ionized regions of the IGM.
This evolution of the LAE population is apparent in both a drop in the normalization of the LAE luminosity function (LF) toward higher redshifts and a decrease in the LAE line widths from z = 5 to z = 7, though both of these results appear to depend on the luminosities of the LAEs (Konno et al. 2014(Konno et al. , 2018;;Santos et al. 2016;Itoh et al. 2018;Taylor et al. 2020Taylor et al. , 2021;;Goto et al. 2021;Ning et al. 2022;Songaila et al. 2022).
A similar result is seen in UV continuum samples, where the fraction of objects with strong Lyα emission appears to decline beyond z = 6 (Stark et al. 2010;Pentericci et al. 2018;Hoag et al. 2019). However, both the significance of this result and the redshift where the drop occurs are controversial (see, e.g., Stark 2016;Fuller et al. 2020;Kusakabe et al. 2020).
Quasar spectra support the idea of patchiness in the reionization process (e.g., Becker et al. 2015), as do other lines of evidence, such as multipeaked LAE spectra (Hu et al. 2016;Songaila et al. 2018;Bosman et al. 2020;Meyer et al. 2021) and the distribution of LAEs. As an example of the latter effect, Larson et al. (2022) found a candidate LAE at z = 8.7 near a known source at a similar redshift (Zitrin et al. 2015). Furthermore, we are beginning to see LAEs well into the redshift range where we would expect the IGM to be substantially neutral (e.g., Oesch et al. 2015;Zitrin et al. 2015;Hashimoto et al. 2018;Hoag et al. 2018;Pentericci et al. 2018). Indeed, with the advent of JWST, Lyα emission has now been detected out to z = 10.60 (GN-z11; Bunker et al. 2023).
In combination, all of these observations can be matched to models of the evolution of LAEs and used to constrain the evolution of the IGM. Extensive simulations combining sophisticated radiative transfer codes with cosmological models, such as THESAN (e.g., Garaldi et al. 2022;Neyer et al. 2024) and SPHINX (e.g., Garel et al. 2021;Katz et al. 2023), are now being carried out for this purpose. Despite the beautiful results from these simulations, there are still many uncertainties. Perhaps most importantly, it is not easy to disentangle the effects of the IGM from the complexity of the galaxy's Lyα emission and the scattering at circumgalactic scales (Laursen et al. 2011;Jensen et al. 2013;Garel et al. 2015;Hassan & Gronke 2021;Guo et al. 2023).
In order to model the evolution of the IGM at very high redshifts, we need to separate the effects of the IGM from the intrinsic structure of the Lyα line emerging from the galaxy and its surrounding circumgalactic medium (CGM). The simplest approach is to assume that the emergent Lyα lines are varying more slowly with redshift than the IGM properties. However, even with such a crude assumption, we must have a large reference sample to understand how the significant variations in the emergent Lyα lines (e.g., Guo et al. 2023) combine with the IGM structure to match to the observations. More sophisticated modeling also requires detailed information on large samples of LAEs to constrain the underlying assumptions. A further complication is that the line properties appear to be a function of the Lyα luminosity (e.g., Santos et al. 2016;Taylor et al. 2020Taylor et al. , 2021;;Ning et al. 2022;Songaila et al. 2022), introducing a shape dependence in the evolution of the Lyα LF and a dependence of the velocity width on the Lyα luminosity. Thus, we also need to sample a wide range of Lyα luminosities.
In practice, the highest redshift where high-quality spectra of hundreds of LAEs can be straightforwardly obtained is z = 7, where giant optical imagers, such as Hyper Suprime-Cam (HSC) on the Subaru 8.2 m telescope, can be used to generate narrowband-selected samples (e.g., SILVERRUSH (Konno et al. 2018) and HEROES (Taylor et al. 2023); see also the LAGER survey, which uses DECam on the CTIO Blanco 4 m telescope (Wold et al. 2022)). These can then be followed up with efficient optical spectrographs to generate large spectroscopic samples. The lowest-background long-wavelength regions in the atmospheric emission are at 8160 Å and 9210 Å, which correspond to z = 5.7 and z = 6.6; ∼100 Å filters at these wavelengths are used to make the initial selections of narrowband excess objects (see, e.g., Shibuya et al. 2018;Ning et al. 2022;Kikuta et al. 2023;Taylor et al. 2023). Ideally, the LAE samples should be complete to a limiting observed Lyα luminosity and cover large contiguous areas (10s of deg 2 ) in fields with extensive ancillary observations. Our primary field is the 44 deg 2 HEROES field in the North Ecliptic Pole (NEP; Taylor et al. 2023), which has, or will have, deep coverage from the eROSITA X-ray and SPHEREx spectroscopy missions and contains the largest of the Euclid Deep Fields (20 deg 2 ). The goal of the present paper is to provide a publicly available atlas of uniform spectra for a very large sample of LAEs at z = 5.7 and z = 6.6 over a range of observed Lyα luminosities in this field together with a number of smaller areas. This sample can be used to characterize the LAE properties, to assess the evolution of the ionization structure, and to match to the models at these later stages in the reionization process. As an example, in Figure 1, we show the current spectroscopically confirmed LAE sample at z = 6.6 in HEROES, which we compare with a similarly sized area in the THESAN simulation of Lyα emission at high redshifts (Xu et al. 2023). In this simulation, the z = 6.6 LAEs are seen to occur in regions of the IGM that are more ionized (purple shading) and to be absent in regions that are more opaque.
We present our spectroscopic observations and discuss our observed Lyα line profiles in Section 2. In Section 3, we briefly summarize the various types of exotic LAEs seen in the sample. In Section 4, we make our Lyα line width measurements. In Section 5, we discuss the evolution of the velocity widths with redshift. Finally, in Section 6, we consider the constraints on the IGM evolution that may be derived from the observations. We use a H 0 = 70 km s -1 Mpc -1 , Ω M = 0.3, and Ω Λ = 0.7 cosmology throughout.
We have been obtaining follow-up spectroscopy of narrowband-selected samples (using the narrowband filters NB816 for z = 5.7 and NB921 for z = 6.6) in HEROES and other fields in the Subaru/HSC and Subaru/Suprime-Cam archives. These Subaru imaging observations provide very large samples (many thousands) of photometrically selected narrowband excess sources satisfying the LAE selection criteria: NB816 -i > 1.2 for z = 5.7, with the sources also undetected in the g and r bands, and NB921 -z > 1 for z = 6.6, with the sources also undetected in the g, r, and i bands. Thus, we have many available candidates for follow-up spectroscopy. Most of these lie in small areas (5-7 deg 2 ). HEROES (Figure 1) is the largest contiguous field (44 deg 2 ) in our spectroscopic survey.
We give the observed field for each object in Tables 1, 2, and 3, along with the R.A. and decl. of each object. The selection of the sources observed prior to 2010 is summarized in Hu et al. (2010). For sources observed after 2010, we give the narrowband and continuum magnitudes in Table 3, together with cutouts of the images when these were observed with HSC. The LAEs in the NEP (52 of the z = 6.6 sample and 34 of the z = 5.7 sample) were taken from the catalog of Taylor et al. (2023). The remaining sample comes from the SSA22, XMM-LLS, COSMOS, and GOODS-N fields, all of which were observed as part of the HSC-SSP initiative (Aihara et al. 2022).
We carried out our spectroscopic observations over a very extended period from 2002 to 2023 August, but the observations were all done in a uniform fashion with the same spectrograph configuration and wavelength range. We give the date of each observation in Table 3. Hu et al. (2010) described some of the earlier spectra, and spectroscopic observations for a number of the sources have appeared in the literature (e.g., Sobral et al. 2015;Jiang et al. 2017;Matthee et al. 2017;Shibuya et al. 2018;Songaila et al. 2018Songaila et al. , 2022)).
We used the Keck/DEIMOS spectrograph with the 830 line mm -1 grating and 1″ slits, which gives good red sensitivity and moderately high resolution. Because of ghosting in this configuration, we used three dithered 20 minutes observations. The total exposure time is 1 hr for most of the sources. Some LAEs have multiple observations and hence longer exposure times. The total exposure time for each source may be found in Table 3. Because the sources are sparse, there are usually only one or two LAEs in each DEIMOS mask. The data reduction is described in Cowie et al. (1996). We measure the instrumental resolution from the sky lines, giving an average of 85 km s -1 at z = 6.6 and 96 km s -1 at z = 5.7.
We summarize the LAE targets and the fields they are drawn from in Tables 1 (z = 5.7) and 2 (z = 6.6). We give the signalto-noise of the LAEs in Table 3.
On average, the LAEs at these high redshifts have a very generic (and well-known) spectral shape, with a sharp break at the blue side and a tail to the red side (see Figure 2). The small number of low-redshift contaminants of the narrowband candidate sources (∼5%) are easily recognized by the doublet nature of the [O II]3727 and [O III]5007 lines and were dropped from the sample. A larger number of the narrowband-selected sources (∼25%) were not confirmed by the spectroscopy and were also eliminated. The weak continuum also shows the break at the Lyα wavelength. We see very little change in the shape and width of the average Lyα line profile between z = 5.7 and z = 6.6. The shape is governed by the escape of Lyα from the galaxy and its circumgalactic gas, combined with the blue-side scattering from the neutral IGM. (See Guo et al. 2023 and references therein for an extensive recent discussion.) However, these average line profiles are, in reality, drawn from a fairly diverse set of individual line shapes and widths, as we illustrate in Figure 3, where we show two-dimensional (2D) spectra for a subset of our z = 5.7 LAEs. We see objects with a wide range of red velocity wings, including some spectacularly long-tailed objects, such as sources 4 and 69. We also see objects with blue emission relative to the Lyα peak, such as source 2, and lensed systems, such as source 114. This variety reflects the variation in the redshift of the galaxy with respect to the local IGM, the variation in the ionization surrounding the source, and the degree to which escaping from the galaxy and the CGM has moved the emission to the red.
As mentioned above, our large LAE spectroscopic samples allow us to find rare and interesting objects. In general, only one to a few of these have been found, emphasizing the need to continue to increase the numbers in our spectroscopically observed sample so we can find further examples and make statistical use of them. We describe three categories of such exotic LAEs here.
While double-peaked Lyα profiles resulting from the escape of the line from the galaxy and the CGM are seen in ∼30% of low-to moderate-redshift LAEs (e.g., Kulas et al. 2012), at high redshifts, we expect any blue wing to be removed by scattering in the IGM. However, we see a handful of LAEs with blue-side emission (e.g., Hu et al. 2016;Songaila et al. 2018;Bosman et al. 2020;Meyer et al. 2021 for published examples). We show a double-peaked LAE in Figure 4, which we compare with a single-peaked LAE of similar luminosity. >43.25 erg s -1 ) at z = 5.7 (blue; 55 objects) and z = 6.6 (red; 57 objects). The FWHM of the stacked spectra at the two redshifts are shown. The spectra are from the 830 line mm -1 grating on the Keck/DEIMOS spectrograph. The instrumental FWHM of 85 km s -1 at z = 6.6 and 96 km s -1 at z = 5.7 are not subtracted here. Double-peaked Lyα emission may be direct evidence for large ionized bubbles around these high-redshift LAEs, which may reveal reionization in progress. In this interpretation, we are seeing the Lyα profile emerging from the galaxy directly without scattering by the IGM. However, we must also consider the possibility that double-peaked Lyα profiles could result from processes other than ionized bubbles. Matthee et al. (2018) suggested that they could be the result of a merger, where both galaxies are producing Lyα lines redward of the IGM cutoff but with different velocities. This could then result in an observed double-peaked structure with different components in the LAE having significantly different redshifts.
As we noted in Section 2, we see a small number of LAEs with very extended red wings, in some cases to velocities beyond 2000 km s -1 . Yang et al. (2014) previously discussed extended red wings, but the object they considered only extends to 1000 km s -1 . The present sample contains much more extreme objects. The z = 5.7 sample contains 3 sources (4, 45, and 69) with tails extending beyond 1500 km s -1 . At z = 6.6, there are two such sources (35 and 43). Source 35 is also the double-peaked source NEPLA4.
There are also two active galactic nuclei (AGNs) in the z = 6.6 redshift range (Figure 5, right panel; see also Figure 11) and one in the z = 5.7 range. These sources are not included in the LAE samples.
In Figure 5 (left panel), we show the spectrum of one of the extended red wing LAEs (source 4 in the z = 5.7 sample). In addition to a narrow component, the spectrum shows a red wing extending to beyond 3000 km s -1 . It is difficult to reproduce such extreme line widths with models of the Lyα escape from star-forming galaxies. One possible alternative interpretation is that we are seeing a combination of narrow Lyα emission from the galaxy and broad Lyα emission from an AGN in the galaxy. These sources might be related to JWSTdetected AGNs at these redshifts, some of which also show narrow lines with underlying broad components, but in the restframe optical (Harikane et al. 2023;Kocevski et al. 2023;Greene et al. 2024).
Lensed LAEs are of particular interest, since the differential paths allow us to probe for the presence of small-scale structure in the circumgalactic and intergalactic gas. So far, we have only found one such object: an LAE cross at z = 5.749 in the SSA22 field (see Figure 6, upper panel). The components of the cross lie at separations from 2 5 to 5 2, and while the components with measured redshifts (A and B) lie at similar redshifts, their Lyα profiles are not identical. Thus, as we show in Figure 6 (lower panel), component B is slightly (but significantly) redward of component A by ∼35 km s -1 . These changes in the Lyα profiles may reflect changes in the Lyα scattering for scales of order 10 pc in the CGM to several hundred parsecs in the IGM. The variations in the Lyα profiles should provide constraints on the modeling of LAEs and on how the profiles change with viewing angle (Guo et al. 2023).
We will consider the double-peaked sources further in the discussion, but we postpone consideration of the extended red wing sources and the LAE cross to subsequent papers.
For each LAE, we first determine the redshift corresponding to the peak of the Lyα line. We then make a parameterized fit to determine the full width at half maximum (FWHM) in velocity. This gives an accessible measurement of the properties of the LAE and allows for a simple classification.
Following Shibuya et al. (2014) and Claeyssens et al. (2019), we fit the Lyα line with an asymmetric Gaussian profile,
where A is the amplitude (normalization), and Δv is the velocity relative to the peak of the Lyα profile. Here, a asym controls the asymmetry and d controls the line width. In terms of these free parameters, the width of the line (in km s -1 ) is
For the double-peaked sources, we fit only to the red wing of the LAE (right panel of Figure 4). As can be seen from the two examples shown in Figure 4, the asymmetric fit generally provides an extremely good representation of the line. At both redshifts, all of the asymmetry parameters are positive, corresponding to the wings being red. The median asymmetry parameter is 0.26 at z = 5.7 and 0.28 at z = 6.66, and the denominator in Equation (2) increases the FWHM by ∼10%, on average. The corresponding error in the FWHM is nearly fully dominated by the error in d, with only a few percent contribution from the a asym term. We tabulate the redshifts, Lyα luminosities, FWHM, and asymmetry parameters in Tables 1 and 2.
We next calculate the Lyα line fluxes and luminosities. Because of the difficulty of an absolute calibration for slit spectra, we calculate the Lyα line fluxes from the narrowband imaging. We assume that the narrowband flux is solely produced by the Lyα line. We do not correct for the continuum, which is extremely weak for these objects and not always measurable. The mean observed-frame Lyα equivalent width (EW) in the stacked normalized spectra is 600 Å at z = 6.6 and 410 Å at z = 5.7, while our selection chooses LAEs with observed-frame Lyα EWs of 130 Å or greater. Even for these lowest-EW objects, the continuum correction to the luminosities is <0.1 dex (Taylor et al. 2021). We use aperture-corrected 2″ diameter aperture magnitudes, with the correction (∼0.30 mag) determined via comparison with 4″ diameter apertures. We next divide the fluxes corresponding to the narrowband magnitudes by the filter transmission efficiencies at the spectroscopic redshifts to obtain our final Lyα line fluxes. We then determine the Lyα luminosities, L(Lyα), using the luminosity distances for our adopted cosmological geometry.
As a check, we compare our narrowband-derived L(Lyα) to our measured A × FWHM, which is roughly the spectral line strength. As can be seen from Figure 7, there is a near-linear relation between the two, which gives us confidence in our L (Lyα) measurements.
We plan on releasing an atlas of our LAE spectra on a regular basis as we update the sample. Our goal is to provide all of the information needed to underpin theoretical modeling. For each of the two redshifts, z = 5.7 and z = 6.6, we will release a binary FITS file containing the quantities given in Table 3. Table gives the first release as part of the present paper.
In Figure 8, we show our Lyα line widths versus L(Lyα) (blue circles: z = 5.7; red squares: z = 6.6). As we discuss below, the z = 5.7 sample FWHM distribution is relatively uniform over the observed luminosity range. In contrast, the z = 6.6 sample FWHM distribution is much narrower at lower luminosities (log L(Lyα) <43.25 erg s -1 ) than at high luminosities. At high luminosities, the two redshift samples have similar FWHM distributions.
In Figure 9, we compare the FWHM distributions of the full samples (black lines) with those at lower luminosities (shaded regions). At z = 5.7, the two distributions are similar, and a Kolmogorov-Smirnov test gives a 0.84 probability that they could be consistent. At z = 6.6, the two distributions are quite different, with the lower-luminosity LAEs being narrower. A Kolmogorov-Smirnov test gives only a 4.7 × 10 -5 probability that they could be drawn from the same distribution. The highluminosity LAEs at z = 6.6 are consistent with both the highand low-luminosity LAEs at z = 5.7. Comparing the highluminosity LAEs at the two redshifts gives a Kolmogorov-Smirnov probability of 0.51 (See also the stacked spectra of Figure 2).
These results suggest that, at z = 6.6, the high-luminosity LAEs may preferentially lie in more highly ionized regions than the lower-luminosity LAEs, while by z = 5.7, the IGM is much more uniformly ionized. We return to this point in the discussion (Section 6).
Our Lyα line widths versus L(Lyα) are mostly consistent with previously published values. In Figure 10, we compare our measurements with the measurements from Shibuya et al. (2018;green circles). We also compare with Ouchi et al. (2010) for fainter LAEs at z = 6.6 and with Mallery et al. (2012) for LAEs at z = 5.7. We do not compare with Ouchi et al. (2010) for LAEs at z = 5.7, as their errors are too large to be useful. We note that, while their z = 6.7 errors are also large, within these errors, their data are consistent with the present results. Finally, we compare with the stacked values of Ning et al. (2022;red diamonds), which are in very good agreement with our measurements at z = 6.6 but appear to be low at z = 5.7.
The differential evolution of the velocity widths with redshift discussed in Section 2 and shown in Figures 8 and 9 implies that the z = 5.7 LAEs must lie in a relatively uniformly ionized background, regardless of L(Lyα), while at z = 6.6, the highluminosity LAEs continue to lie in ionized regions, but the lower-luminosity sources show the effects of increasing IGM neutrality. Thus, at z = 6.6, it appears that the high-luminosity LAEs either are producing large ionized bubbles themselves, or they are marking overdense galaxy sites that are producing such bubbles, while the lower-luminosity LAEs are not. In order to avoid suppressing the red wings of the LAEs with the radiation-damping wings from surrounding neutral hydrogen in the IGM, we require a surrounding highly ionized region with a size R greater than about 1 pMpc (e.g., Mason & Gronke 2020 and references therein). This size does not depend on the LAE properties but only on the IGM neutrality.
The requirement that the observed LAEs lie in ionized bubbles has been used to estimate lower limits on the fractional ionized volume in the IGM (Malhotra & Rhoads 2006). Here, we use the present data to refine this calculation at z = 6.6.
We adopt the radius versus luminosity shape dependence from Yajima et al. (2018), whose normalization agrees closely with the present work. We then take the present results as an empirical normalization of the R-L(Lyα) relation. Using only the shape of the Yajima curve and normalizing to 1 pMpc at our transition log L(Lyα) = 43.25 erg s -1 , we obtain
where L 43 is L(Lyα) in units of 10 43 erg s -1 . Lyα sources fainter then log L(Lyα) = 43.25 erg s -1 have bubble sizes smaller than 1 pMpc and will suffer scattering in the red wing.
We can now use this calibration of R to estimate the fractional ionized volume marked by the luminous LAEs. The dependence of the ionized volume around a given LAE is ∝L(Lyα) 0.855 , so the ionized fraction is dominated by the more numerous lower-luminosity sources. In HEROES, there are 21 sources with log L(Lyα) above 43.5 erg s -1 , where the sample is substantially complete (Taylor et al. 2021 give 0.6 for the incompleteness at this luminosity). At log L(Lyα) above 43.4 erg s -1 , there are 32 sources in the field, as shown in Figure 11. The small covering area of these very luminous sources can be seen in Figure 11. The observed volume is 82,000 pMpc 3 , while the summed incompleteness corrected would only make the M1500 fainter. (The error bars for all the sources are shown in Figure 15.) In Figure 15, we compare the double-peaked LAEs with other LAEs in the HEROES and SSA22 fields. We can see that the double-peaked LAEs (larger symbols) are among the brighter NUV sources, with M1500 magnitudes in the range from -21 to -22.5.
Mason & Gronke (2020) computed models for the proximity radius at z = 6.6 and the corresponding peak velocity Figure 12. Two-dimensional spectral images of the four double-peaked sources in the z = 6.6 sample. The contour levels are 0.2, 0.36, 0.5, and 0.9 times the red peak value. We previously published two of these (COLA1 and NEPLA4), which we label. The dashed line marks the slit center, and the small blue vertical line the second peak. The 1D spectrum is also shown, together with the sky spectrum (red), both of which have arbitrary normalizations to fit on the plot.
Figure 13. Two-dimensional spectral images of four of the 15 double-peaked sources in the z = 5.7 sample. The contour levels are 0.2, 0.36, 0.5, and 0.9 times the red peak value. The dashed line marks the slit center, and the small blue vertical line the second peak. The 1D spectrum is also shown, together with the sky spectrum (red), both of which have arbitrary normalizations to fit on the plot. separation as a function of the NUV luminosity. In Figure 16, we show that three of our four z = 6.6 double-peaked LAEs lie beneath the Mason & Gronke (2020) reference model (black curve) for the peak velocity separation and are therefore inconsistent with this model (as Mason & Gronke 2020 pointed out for COLA1). However, the Mason & Gronke (2020) models show that reducing the local gas density by a factor of 2 (green curve) would permit the transmission of the blue peaks.
As Mason & Gronke (2020) point out, their reference model is highly optimistic, assuming an escape fraction of one and no clumping, so even the green curve may not be sufficient. For example, Meyer et al. (2021) argue via comparison of the peak velocity separations of such wide sources with those of local peak separations that the escape fractions should be considerably lower, with values of around 0.1-0.2 (Izotov et al. 2018).
Thus, multiple ionizing sources in the vicinity of the doublepeaked LAEs may be required in order to generate the large proximity radii necessary to see the blue peaks in these sources. Deeper observations in both the continuum and Lyα in the vicinity of the double-peaked LAEs are clearly needed in order to examine this possibility.
We present two uniformly observed large spectroscopic samples of LAEs (127 at z = 5.7 and 82 at z = 6.6) from our Keck/DEIMOS follow-up of narrowband-selected sources in the large, contiguous HEROES field (44 deg 2 ), plus several smaller fields from the Subaru/HSC and Subaru/Suprime-Cam archives. Our main results are as follows.
The generic spectral shape of the average Lyα line profile at these redshifts, with a sharp break at the blue side and a tail to the red side, belies the diversity of individual line shapes and widths that we see. This emphasizes the need for large spectroscopic samples, such as the present atlas, which can be combined with models of the IGM density and ionization structure to predict the LAE properties at higher redshifts.
Large spectroscopic samples, such as the present atlas, also reveal rare and interesting objects. We found small numbers of LAEs with double-peaked Lyα profiles; very extended red wings; broad-line AGNs; and a unique lensed LAE cross.
By comparing the Lyα line widths for the LAEs at the two redshifts, we found that the FWHM distribution for the z = 5.7 sample is relatively uniform over the observed luminosity range, while that for the z = 6.6 sample is not: it is much narrower at lower luminosities (log L(Lyα) <43.25 erg s -1 ) than at high luminosities. The z = 5.7 and z = 6.6 samples have similar FWHM distributions at high luminosities. This suggests that, at z = 6.6, the high-luminosity LAEs may preferentially lie in more highly ionized regions than the lower-luminosity LAEs. The ionized regions constitute at least 8% of the volume at z = 6.6, but this estimate is quite uncertain and depends strongly on the adopted Lyα LF. We also show the double-peaked sources in the low-redshift Green Pea galaxy sample of Yang et al. (2017;green diamonds) and the z = 6.9 double-peaked LAE of Meyer et al. (2021;purple triangle). Sources without double peaks are marked with small triangles at the bottom of the figure (red is z = 6.6, blue is z = 5.7). The Green Pea sources that do not satisfy our double peak definition are shown as green triangles. UV absolute magnitude at 1500 Å for the z = 6.6 double-peaked LAEs (red circles-HEROES; green square-COLA1; purple diamond-SSA22-333.68-0.572). The black curve shows the reference model of Mason & Gronke (2020), while the green curve shows the same model but with half the average density in the neighborhood of the source.
At z = 5.7, about 14% of the LAEs have double peaks, while at z = 6.6, this has dropped to about 6%. This can be compared with the 30% seen at lower redshifts. Comparing the observed z = 6.6 double peaks with the models of Mason & Gronke (2020) suggests that the LAEs themselves are not UV luminous enough to ionize the surrounding regions to the required level (see also Meyer et al. 2021). The double-peaked Lyα profiles may therefore mark sites with multiple ionizing sources.
In future work, we plan to migrate our ongoing spectroscopic follow-up from Keck/DEIMOS to the Keck/KCWI integral field spectrograph, which will provide spatially resolved spectra of the LAEs and also allow us to search for neighboring ionizing sources.
The Astrophysical Journal, 971:136 (15pp), 2024 August 20 Songaila et al.