MAMMOTH-Subaru. III. Lyα Halo Identified by Stacking ∼3300 Lyα Emitters at z = 2.2–2.3

In this paper, we present a Lyα halo (LAH) identified by stacking ∼3300 Lyα emitters (LAEs) at z = 2.2–2.3. We carry out imaging observations and data reduction with Subaru/Hyper Suprime-Cam. Our total survey area is ∼12 deg2 and the imaging depths are 25.5–27.0 mag. Using the imaging data, we select 1240 and 2101 LAE candidates at z = 2.2 and 2.3, respectively. We carry out spectroscopic observations of our LAE candidates and data reduction with Magellan/IMACS to estimate the contamination rate of our LAE candidates. We find that the contamination rate of our sample is low (8%). We stack our LAE candidates with a median stacking method to identify the LAH at z = 2. We show that our LAH is detected until ∼100 kpc at the 2σ significance level and likely extended to ∼200 kpc at a surface brightness level of ∼10−20 erg s−1 cm−2 arcsec−2. Compared to those of previous studies, our LAH is brighter at radii of ∼25–100 kpc, which is not likely caused by the contamination in our sample but by the different redshifts, fields, and selection methods instead. To investigate how central galaxies affect surrounding LAHs, we divide our LAEs into subsamples based on the Lyα luminosity (L Lyα ), rest-frame Lyα equivalent width (EW0), and UV magnitude (M uv). We stack the subsamples and find that higher L Lyα , smaller EW0, and brighter M uv cause more extended halos. Our results suggest that more massive LAEs generally have more extended LAHs.

Until recently, LAHs are only identified at radii smaller than ∼100 kpc due to limited depths (e.g., Momose et al. 2014;Xue et al. 2017;Arrigoni Battaia et al. 2019;Cai et al. 2019;Wu et al. 2020).Kakuma et al. (2021) and Kikuchihara et al. (2022) found that LAHs are possibly extended to ∼200 kpc, although the radius bin sizes are not small enough to investigate the transition from central galaxy to CGM at the scale of several kiloparsecs to several hundred kiloparsecs.By stacking 15 z ∼ 2 QSOs, Arrigoni Battaia et al. (2016) reported a 2.3σ detection of an LAH in the radius bin of 50-500 kpc.On the other hand, Lujan Niemeyer et al. (2022) used relatively small radius bin sizes and identified an LAH extended to 320 kpc at z = 1.9-3.5, although there were only two bins at radii 100 kpc (80−160 and 160−320 kpc).
After the identification of LAHs, previous studies investigated how central galaxies affect LAHs.Momose et al. (2016) found that galaxies with a lower Lyα luminosity (L Lyα ), smaller rest-frame Lyα equivalent width (EW 0 ), and brighter UV magnitude (M uv ) have more extended halos.In contrast, Xue et al. (2017) and Lujan Niemeyer et al. (2022) found that galaxies with higher L Lyα show more extended halos.The reason for the contradictory results is still not clear.
In this paper, we present an LAH identified by stacking ∼3300 LAEs at z = 2.This paper is structured as follows.We introduce our imaging observations and data reduction in Section 2. The sample selection is presented in Section 3. Section 4 shows our spectroscopic observations, data reduction, and contamination rate estimation.Our results and discussion are presented in Section 5. Finally we summarize this paper in Section 6.

Imaging Observations and Data Reduction
We carry out narrowband (NB) and broadband (BB) imaging observations with Subaru/Hyper Suprime-Cam (HSC; Furusawa et al. 2018;Kawanomoto et al. 2018;Komiyama et al. 2018;Miyazaki et al. 2018).The NB filters we use are NB387 (λ c = 3863 Å, FWHM = 55 Å) and NB400 (λ c = 4003 Å, FWHM = 92 Å), and the BB is g (λ c = 4754 Å, FWHM = 1395 Å).The central wavelengths of the NB387 and NB400 filters are chosen to detect redshifted Lyα emission at z = 2.2 and 2.3, respectively.More details of the observations are described in Liang et al. (2021) and Y. Liang et al. (2023, in preparation).In brief, we observe four fields (J0210, J0222, J0924, and J1419) with NB387 and four other fields (J0240, J0755, J1133, and J1349) with NB400 between 2018 January and 2020 March.The detailed field selection is described in Cai et al. (2016), Liang et al. (2021), Z. Cai et al. (2023, in preparation), and Y. Liang et al. (2023, in preparation).In short, our fields except J0240 and J0755 are selected to contain coherently strong Lyα absorption systems (CoSLAs), while the J0240 and J0755 fields are selected to contain grouping QSOs.It should be noted that our eight fields are expected to host galaxy overdensities at z ∼ 2. The onsource exposure times are ∼2.8-3.8,2.5-3.0, and 0.2-0.8hr for the NB387, NB400, and g filters, respectively, depending on the weather conditions.The seeing sizes are 0 7-1 3 depending on the fields, with the smallest seeing observed in J1419 and the largest seeing observed in J0210.The total survey area is ∼12 deg 2 .
The imaging data are reduced with the HSC pipeline dubbed hscPipe (Aihara et al. 2018;Bosch et al. 2018).The NB387 and g images of the four NB387 fields were reduced by Liang et al. (2021).We reduce the NB400 and g images of the four NB400 fields in the same manner as that in Liang et al. (2021).
In brief, we first carry out bias subtraction, dark subtraction, flat-field calibration, and stacking of individual exposures.We then use the imaging data from the Panoramic Survey Telescope and Rapid Response System 1 (Pan-STARRS1; Schlafly et al. 2012;Tonry et al. 2012;Magnier et al. 2013) survey to calibrate the astrometry and photometry with sky subtraction.Flux calibration of NB400 is carried out in the same manner as that adopted for NB387 (Liang et al. 2021).Because the g − NB400 color may be affected by the 4000 Å break of local stars and galaxies, for the flux calibration we use extended objects with NB400 magnitudes of ∼24 mag that are most likely high-z galaxies.The uncertainty of flux calibration is 0.1 mag for both NB387 and NB400.We estimate the detection limits by fitting a Gaussian function to the histograms of the sky aperture fluxes.The fluxes are measured in apertures with a fixed diameter, which are randomly put across sky regions with no detected sources.The 5σ detection limits of the final imaging products are 24.3-25.0, 25.5-25.8, and 26.2-27.0mag for NB387, NB400, and g, respectively.The above detection limits are measured in a 1 7 diameter aperture, except for the J0210 field, for which they are measured in a 2 5 diameter aperture because the seeing in J0210 is worse than that in other fields.The details of the fields are summarized in Table 1.
We carry out source detection and photometry with SExtractor (Bertin & Arnouts 1996) to build our source catalogs.The source catalogs of the four NB387 fields were obtained by Liang et al. (2021), and we make the source catalogs of the four NB400 fields in the same manner.First, we convolve the NB400 and g images with proper Gaussian kernels to match the point-spread functions (PSFs) of the NB and BB filters.Then, we use the NB400 images as the reference images to detect sources in the dual-image mode of SExtractor.The detection criterion is 15 adjacent pixels above the 1.2σ limit.To reduce the effect of different depths across the field, we use the sky background rms map as the weight image.The mesh size used for sky background estimation is 128 pixels.We do not use regions with low S/N or bad pixels such as field edges and bright-star vicinities.

Sample Selection
Using the reduced images and source catalogs, we select LAEs by an excess of the NB -BB color in a manner similar to that in previous studies (e.g., Konno et al. 2016;Liang et al. 2021).Our NB387 LAE selection is based on the LAE catalogs in Liang et al. (2021), but we apply more strict criteria mainly to reduce spurious faint sources (see also Ma et al. 2023).Our selection criteria for NB387 LAEs are where the subscripts "ap" and "tot" mean the aperture and total magnitudes, respectively.We use the "AUTO" magnitude in SExtractor as the total magnitude.The superscripts "3σ" and "5σ" represent the 3σ and 5σ limits, respectively.If the g magnitude of a source is fainter than the 2σ limit, we use the 2σ limit instead.The (g − NB387) 3σ is the 3σ color error, which is calculated by NB387 , where f err,g and f err,NB387 are the 1σ errors in g and NB387, respectively.The NB -BB color limit (>0.3) corresponds to a rest-frame Lyα equivalent width of >20 Å.After applying the above criteria, we carry out visual inspection to remove spurious sources such as cosmic rays and satellite trails.After the visual inspection, the final number of our NB387 LAE candidates is 1240.
We select NB400 LAEs in the same manner as we do NB387 LAEs.As shown in Figure 1, our selection criteria for NB400 LAEs are where the meaning of the symbols is the same as that in Equation (1).After visual inspection, the final number of our NB400 LAE candidates is 2101.

Spectroscopic Observations and Data Reduction
To estimate the contamination rate of our NB387 and NB400 LAE candidates, we carry out spectroscopic observations with Magellan/IMACS on 2022 September 29 and 30.Magellan/ IMACS is set in the multislit spectroscopy mode with the f/2 camera, which provides a field of view of 12″ in circular radius.We use the 400 line mm -1 grism combined with a filter to cover the wavelengths between ∼3600 and 5700 Å.The spectral resolution is ∼7 Å using this setup with a slit width of 1 2 and a slit length of 8 0. We observe one pointing in each of the J0210 and J0222 fields for NB387 LAEs, and two pointings in the J0240 field for NB400 LAEs.The on-source exposure times are 7500 and 6000 s for NB387 and NB400 LAEs, respectively.
We reduce the spectroscopic data using the official pipeline named COSMOS.We carry out bias subtraction, flat calibration, wavelength calibration, sky subtraction, two-dimensional spectrum extraction, and stacking of individual exposures.We use dome flat frames for the flat calibration, and helium-mercury lamp spectra for the wavelength calibration.We do not conduct flux calibration yet, because it is not necessary for the purpose of contamination rate estimation.

Contamination Rate Calculation
After the data reduction, we obtain 120 and 151 spectra for NB387 and NB400 LAEs, respectively.Among the total 271 spectra, there are 120 spectra with a detected emission line at the expected wavelengths of Lyα emission.The detection fraction is not high because the real on-source exposure time is shorter than expected.Another reason is that we also observe low-priority faint LAE candidates as the number of available slits on the slit mask is much greater than that of our highpriority bright targets.Among the 120 spectra with detections, 22 spectra are confirmed LAEs at z = 2 with 2 emission lines, and two spectra are foreground objects with 2 emission lines.The 22 LAEs with 2 emission lines are typically detected in Lyα, C IV, and/or He II.The two foreground objects with 2 emission lines include a C IV + C III emitter at z = 1.711 and a C IV + He II + C III emitter at z = 1.793.Because our spectral resolution (∼7 Å) is not high enough to resolve the doublet of a foreground [O II] emitter, we only use the spectra with 2 emission lines when estimating the contamination rate.The contamination rate of our LAE candidates is thus 2/(22 + 2) ≈ 8%.The NB magnitudes of the 24 objects with 2 emission lines are 21.3-25.3mag.The spectroscopic properties and statistics of these objects will be presented in a future work.
Our estimated contamination rate (∼8%) is likely an overestimation for the following reasons.First, most LAEs at z ∼ 2 should be detected only in a single line on our spectra, and candidates with 2 emission lines on our spectra are more likely contaminated by low-z sources.Second, the LAEs targeted with spectroscopy are not representative of our total LAE sample, because we mainly select bright (NB  25 mag) LAE candidates for spectroscopy due to the limited sensitivity.However, bright LAE candidates generally have a higher contamination rate than faint candidates (e.g., Davis et al. 2023).Third, [O II] emitters at z < 0.1 are expected to be the largest contaminants for our z = 2 LAEs, and the contamination rate contributed by these [O II] emitters in the literature is low (e.g., <3% in Mentuch Cooper et al. 2023).There is a sequence of LAE candidates with black arrows in the upper right corner.This is because if the g magnitude of a source is fainter than the 2σ limit, we use the 2σ limit instead.

LAE Stacking
After obtaining our LAE catalogs, we stack the NB387 and NB400 LAE candidates to detect the faint and extended LAH at z = 2. First, we match the PSFs of all filters (NB387, NB400, and g) in all the eight fields to the same FWHM (∼1 3) by convolution with proper Gaussian kernels.Then we make cutout NB (NB387 or NB400) and BB (g) images with a size of 84″ × 84″ for each LAE.During this process, we remove six NB387 LAEs because they are too close to the field edges and we cannot make their cutout images with the 84″ × 84″ size.We assume a flat UV continuum slope and subtract the BB images from the NB images to obtain the Lyα images.We stack the Lyα images with a median stacking method for the NB387 and NB400 LAEs separately.Because the expected redshifts of the NB387 and NB400 LAEs are very close (Δz ∼ 0.1), we also stack the NB387 and NB400 LAEs together and this is referred to as the all sample hereafter.The number of LAEs used for stacking is 1234, 2101, and 3335 for the NB387 LAEs, the NB400 LAEs, and all LAEs, respectively.It should be noted that the 3335 all LAEs contain 117 Lyα blobs (LABs), whose details are described in M. Li et al. (2023, in preparation) and Zhang et al. (2023).We do not remove these LABs during the stacking, because there are no evidences showing that extended Lyα emission of LABs is generally distinct (Zhang et al. 2020).After stacking, we globally subtract the median value (∼ −1.24 × 10 −18 erg s −1 cm −2 arcsec −2 ) measured in a radius of 34″-42″ (∼280-350 kpc) from the image to remove sky oversubtraction.
We estimate the uncertainty of our stacking results by the following method.First, we randomly choose sky regions and make cutout sky images (continuum-subtracted) with the same number as our LAE samples.A sky region refers to a region with no detected objects within a radius of 20 pixels (3 4) from the center.This is because when we visually inspect LAEs in Section 3, we only remove LAE candidates contaminated by close (separation 3″) non-LAE objects.As a result, it is natural that far (separation 3″) contaminants exist around our LAEs.Then we stack the sky images in the same manner as for our LAEs.We repeat the above procedure 100 times and obtain 100 stacked sky images.Finally we plot histograms of the surface brightness using the stacked sky images and fit a Gaussian function to measure the 1σ uncertainties at different radii.The surface brightness distributions of the sky at different radii are shown in Figure 8 in the appendix.We find that the 1σ uncertainties at radii of 75 kpc are between 4 × 10 −20 and 5 × 10 −20 erg s −1 cm −2 arcsec −2 .A sky residual of ∼ −1.25 ± 0.01 × 10 −18 (−1.24 × 10 −18 ) erg s −1 cm −2 arcsec −2 is found at radii of 75 kpc (between ∼280 and 350 kpc).It should be noted that this sky residual is consistent with the one measured and subtracted from the stacked Lyα image.We find that the sky residual increases after continuum subtraction, as the NB images have negative sky residuals and the BB images have positive sky residuals.
Figure 2 shows our stacked Lyα images and Figure 3 shows the Lyα surface brightness profiles.Clearly, the Lyα emission is detected until ∼100 kpc at the 2σ significance level (surface brightness of ∼10 −19 erg s −1 cm −2 arcsec −2 ) and likely extended to ∼200 kpc at a surface brightness level of ∼10 −20 erg s −1 cm −2 arcsec −2 , much more extended than the PSF and UV continuum.The PSF is obtained by stacking 533 point sources with NB magnitudes of 18−22 mag.All of the three profiles (NB387, NB400, and all) are consistent within the 1σ error without any scaling.We find that the UV continuum is slightly brighter than the PSF at radii of 20 kpc, which suggests that the UV continuum is likely resolved.Similarly, Kikuta et al. (2023) report resolved "UV halos" that are possibly caused by satellite galaxies.
We compare our stacking result with those of previous LAH studies at z = 2-3 as shown in Figure 4. Using a method similar to the one in this study, Momose et al. (2014, hereafter M14) stacked 3556 LAEs at z = 2.2 with Subaru/Suprime-Cam and identified an LAH extended to ∼80 kpc.Our result is consistent with M14 after scaling, although the uncertainty of M14 is relatively large at radii larger than 30 kpc.We also compare our results with those of Lujan Niemeyer et al. (2022, hereafter N22).N22 stacked 968 spectroscopically confirmed LAEs at z = 1.9-3.5 with the HETDEX data, and identified Lyα emission extended to 320 kpc.Our result is consistent with N22 at radii smaller than ∼25 kpc, but is brighter at radii of ∼25-100 kpc.Further at radii of ∼100-200 kpc, there are no clear differences between the result of N22 and our result beyond the error bars.The N22 halos seem to be more extended than the one in this study at radii larger than 200 kpc, although the last bin size of N22 is large (160−320 kpc).By  this comparison, the most notable difference is at radii of ∼25-100 kpc.
The reason this difference is not clear.The difference is not likely caused by the contamination in our LAEs, because the contamination is typically foreground point sources that make the surface brightness profile more compact, which cannot explain our more extended profile compared to that of N22.Even if the foreground sources such as the [O II] emitters are extended, the contribution from the contamination is expected to be small (8%) because the contamination rate of our LAE candidates is low (8%; see Section 4) and we use a median stacking method.The difference is not likely caused by active galactic nucleus (AGN) contamination either.Although our LAEs may contain AGNs, the AGN fraction should be low (<5%) for the majority (91%) of our LAEs (L Lyα < 10 43 erg s −1 ; Zhang et al. 2021), and the AGN contamination should have negligible effect on our median-stacked LAH.Additionally, it is still not clear if the Lyα surface brightness profiles of galaxies with AGNs are generally distinct from those of galaxies without AGNs.
There are three possible causes for the different results.The first possibility is the redshift difference.N22 stacked LAEs with a larger redshift range (z = 1.9-3.5)than ours (z = 2.2-2.3) to increase their sample size, and a possible evolution of LAHs from z = 3 to z = 2 may cause the difference.Indeed, it is shown that LAHs at z = 2 are 0.4 dex fainter than those at z = 3 at radii smaller than ∼100 kpc (Cai et al. 2019).However, this evolution from z = 3 to z = 2 does not greatly change the profile shape (slope), and the redshift evolution alone may not explain the difference between the results of N22 and our results.The second possibility is the field difference.Our fields are either selected based on a high H I density along background QSO sightlines, or selected to contain multiple QSOs (see Cai et al. 2016;Liang et al. 2021;Z. Cai et al. 2023, in preparation, for details).As a result, our special field selection may cause the different LAHs.The third possibility is the different LAE selection methods.LAEs in N22 were selected with integral-field spectrographs and the integration window widths of Lyα emission were 5-18 Å in the observed frame (smaller than our NB widths of 55 and 92 Å).
Note that there are several other stacking studies at z = 2-3 such as Wisotzki et al. (2018) and Kikuchihara et al. (2022).Because results from these studies are similar to those from M14 and N22, and their surface brightness limits are not as deep as those of N22, we only compare our results to those of M14 and N22.

Connection between Central Galaxies and LAHs
To investigate how central galaxies affect the surrounding LAHs, we divide our LAEs into subsamples based on Lyα luminosity (L Lyα ), rest-frame Lyα equivalent width (EW 0 ), and UV magnitude (M uv ).If the properties (L Lyα , EW 0 , and M uv ) of an LAE are smaller than the median values, we assign this LAE to the low sample.In contrast, we assign an LAE to the high sample if its properties are larger than the median values.Each subsample thus contains one-half of the all sample.The median properties of the subsamples are summarized in Table 2. Histograms of the L Lyα , EW 0 , and M uv of our LAEs are shown in Figure 5. EW 0 is calculated in the same manner as that in the appendix in Shibuya et al. (2018).If the g magnitude of a source is fainter than the 2σ detection limit, we use the 2σ limit instead.
We stack the subsamples separately using the same method in Section 5.1.Figure 6 shows the stacking results of our subsamples.The subsamples show clear differences at radii of ∼10-40 kpc.We find that higher L Lyα , smaller EW 0 , and brighter M uv cause more extended LAHs.Because the three properties are correlated and more massive LAEs generally have higher L Lyα , smaller EW 0 , and brighter M uv (see Muv_low and Muv_high subsamples in Table 2), our results suggest that more massive LAEs generally have more extended halos.Similar results are shown in Zhang et al. (2020).Interestingly, recent simulations also show that more massive galaxies should be characterized by brighter LAHs (e.g., Byrohl et al. 2021).
It should be noted that the typical (median) errors of Lyα, EW 0 , and M uv of our LAEs are 4.0 × 10 41 erg s −1 , 6.9 Å, and 0.05 mag, respectively.We find that 343, 328, 209, 245, 103, and 144 LAEs may be placed in the opposite subsample due to the errors in the L_low, L_high, EW_low, EW_high, Muv_low, and Muv_high subsamples, respectively.After removing these uncertain LAEs, we stack the new subsamples separately.We find that the changes as compared to Figure 6 are negligible, and that the differences of the surface brightness profiles between the new subsamples remain clear.Our results suggest  that the different surface brightness profiles between the are not caused by the errors of Lyα, EW 0 , and M uv .
Because LAEs with higher L Lyα generally have more extended halos, we investigate whether the halo difference between the result of N22 and our result is caused by luminosity differences.The median Lyα luminosities of L_low, L_high, and N22 are 42.2, 42.7, and 42.8 erg s −1 , respectively.Because a higher L Lyα corresponds to a more extended profile, the profile of N22 is expected to be more extended than those of L_low and L_high.However, we find that the profile of N22 is clearly more compact at radii of ∼25-100 kpc, as shown in Figure 7.This comparison suggests that our more extended LAH compared to N22 cannot be explained by the luminosity difference, but instead by other reasons such as the two possibilities we discussed in Section 5.1.
Previous studies have also investigated the connection between central galaxies and LAHs as we briefly introduced in Section 1. Momose et al. (2016) also found that galaxies with smaller EW 0 and brighter M uv have more extended halos.However, galaxies with higher L Lyα show less extended halos in Momose et al. (2016), which is the opposite of the result in this study.Xue et al. (2017) found that galaxies with higher L Lyα and brighter M uv have more extended halos, but the halo correlation with EW 0 is not clear.It should be noted that the halos in Momose et al. (2016) and Xue et al. (2017) are only identified at radii smaller 80 kpc and have larger uncertainties than the ones in this study.Similarly, N22 showed that the LAH becomes more extended for a higher L Lyα until a radius (∼250 kpc) farther than our results, possibly because the detection limit in N22 is deeper (1.3 × 10 −20 erg s −1 cm −2 arcsec −2 ).However, the relations with EW 0 and M uv were not investigated in N22.In summary, the relation between central galaxy properties (L Lyα , EW 0 , and M uv ) and LAHs in this study is consistent with most previous studies.

Summary
In this study, we identify the LAH by stacking ∼3300 LAEs at z = 2.2-2.3.Our results are summarized below.1.We carry out imaging observations and data reduction with Subaru/HSC targeting eight fields that are expected to host galaxy overdensities at z ∼ 2. The total survey area is ∼12 deg 2 and the imaging depths are 25.5-27.0mag.Using the imaging data, we select 1240 and 2101 LAE candidates at z = 2.2 and 2.3, respectively.2. We carry out spectroscopic observations of our LAE candidates and data reduction with Magellan/IMACS to estimate the contamination rate of our LAE candidates.
We find that the contamination rate of our sample is low (8%).3. We stack our LAE candidates with a median stacking method to identify the LAH at z = 2.We find that the LAH is detected until ∼100 kpc at the 2σ significance level (surface brightness of ∼10 −19 erg s −1 cm −2 arcsec −2 ) and likely extended to ∼200 kpc at a surface brightness level of ∼10 −20 erg s −1 cm −2 arcsec −2 .4. Comparing to previous studies, our LAH is consistent with M14 after scaling, but is clearly brighter than that of N22 at radii of ∼25-100 kpc.The halo difference is not likely caused by the contamination in our sample, but by the different redshifts, fields, and selection methods instead. 5. We divide our LAEs into subsamples based on Lyα luminosity (L Lyα ), rest-frame Lyα equivalent width (EW 0 ), and UV magnitude (M uv ).We stack the subsamples and find clear differences between them at radii of ∼10-40 kpc.Our result shows that higher L Lyα , smaller EW 0 , and brighter M uv cause more extended halos, which suggests that more massive LAEs generally have more extended LAHs.

Figure 1 .
Figure 1.Color-magnitude diagram of NB400 LAE candidates (red) and all detected sources in the source catalog (black).The "all detected sources" group includes contamination such as foreground objects, cosmic rays, and satellite trails.The dashed lines represent our selection criteria.The curved dashed line corresponds to the 3σ color error (g − NB) 3σ .This figure shows the LAE selection in the J1349 field as an example, and we use aperture magnitudes for g and NB400.There is a sequence of LAE candidates with black arrows in the upper right corner.This is because if the g magnitude of a source is fainter than the 2σ limit, we use the 2σ limit instead.

Figure 3 .
Figure 3. Lyα surface brightness profiles of NB387 (blue), NB400 (red), UV continuum (yellow), and all (black) LAEs.The horizontal error bars indicate the ranges of radius bins, while the vertical error bars and shaded regions are the 1σ uncertainties of surface brightness obtained by our simulation.The PSF (dotted line) is obtained by stacking 533 point sources with NB magnitudes of 18−22 mag.We scale the PSF and UV continuum to the all profile at the radius of ∼0 kpc for comparison, while the NB387, NB400, and all profiles are not scaled.

Figure 4 .
Figure 4. Lyα surface brightness profiles of Momose et al. (2014; magenta) and Lujan Niemeyer et al. (2022; cyan) compared with this study's (black).We scale the profiles of M14 and N22 to this study at the radius of ≈0 kpc for comparison.

Figure 5 .
Figure 5. Histograms of Lyα luminosity (left), rest-frame Lyα equivalent width (middle), and UV magnitude (right) of our LAEs.The vertical dashed lines indicate the median values.In the middle panel, all LAEs with EW 0 > 240 Å are counted in the 240-260 Å bin for clarity.

Figure 6 .
Figure6.Lyα surface brightness profiles of our subsamples (blue for low and red for high) divided based on L Lyα (left), EW 0 (middle), and M uv (right).All of the profiles are scaled to the all sample at the radius of ∼0 kpc for comparison, as the profiles have different surface brightness at ∼0 kpc.Because we do not find clear differences between the subsamples beyond 1σ errors at radii of 40 kpc, we only show the profiles until 60 kpc for clarity.

Figure 7 .
Figure 7. Lyα surface brightness profiles of our subsamples (blue for low and red for high) and the all sample (black) compared with those of N22 (cyan).All of the profiles are scaled to the all sample at the radius of ∼0 kpc for comparison.