A lack of LAEs within 5Mpc of a luminous quasar in an overdensity at z=6.9: potential evidence of quasar negative feedback at protocluster scales

We observed the area around VIK J2348–3054 with DECam housed on the 4m Blanco telescope at the CTIO (PROPID: 2021B-0905). The observing run was split into two parts: the first took place over four half-nights from the UT 2021-08-30 to the UT 2021-09-03 where the moon did not rise during the observations, while the second occurred over two half-nights on UT 2021-10-20 and 2021-10-21 during full-moon. All observations were done under acceptable observing conditions and all frames were used. From the combined seven half-nights of observations we obtained integrations of 5.8, 8.6, and 14.2 hours in the i-band ($\lambda_{c}=7840$ Å; $\Delta\lambda$ = 1470 Å), z-band ($\lambda_{c}=9260$ Å; $\Delta\lambda=1520$ Å), and NB964 narrowband ($\lambda_{c}=9640$ Å; $\Delta\lambda$ = 94 Å), respectively. The filter transmission curves are shown in Figure 1.

This particular choice of filters allow us to search for LAEs within a redshift range of 6.89 ¡ z ¡ 6.97, making it well suited for a search around VIK J2348–3054 at z = 6.9018 (Venemans et al., 2016). Given that the Ly-$\alpha$ line tends to be redshifted from the galaxy’s systemic velocity by $\sim$ 200 km s^-1 (Verhamme et al., 2018), we can detect galaxies with peculiar velocities ranging from $\sim$ -700 km s^-1 to $\sim$ 2250 km s^-1 from the quasar’s systemic velocity. Sources around high-z quasars and in protocluster environments can have peculiar velocities within 1000 $\pm$ km s^-1 (Overzier et al., 2009). This means that we should detect all sources associated with the protocluster environment with redshifted peculiar velocities, but may miss some sources with extreme blue shifted peculiar velocities.

DECam has 64 individual CCDs with numerous gaps between them. To account for this and obtain the best possible data quality, we implemented a dithering pattern during the observations, randomly shifting the pointing position within a 120 arcsecond box for each set of exposures. The pointing was also offset globally by 5 arcminutes in order to ensure that the quasar was not centered on a chip gap¹¹1The full observation script-generator is available at https://github.com/TrystanScottLambert/DECam_Photometry/blob/main/scriptmaker.py.

Each band was stacked using the DECam community pipeline (Valdes et al., 2014). The DECam Community Pipeline provides the following operations: linearity correction, bias subtraction, dome, photometric, and dark sky flat fielding, bad pixel masking, masking of saturation effects and cosmic rays, an astrometric solution, removal of background patterns including stray light, pupil reflection, and fringing, and projection to a standard tangent plane. For multiple, dithered exposures a co-add is produced which includes transient masking. Three methods for masking cosmic rays and other transients were used in this work: the “maskcosmics” algorithm, from the Dark Energy Survey, using the traditional comparison to the PSF; an algorithm that identifies “long” cosmic rays based on the ratio of boundary to interior pixels from the IRAF segmentation and cataloging package; individual exposures are compared to a median by image differencing detection; and finally, statistical outlier rejection may be used when making the final stack.

After being stacked the images were astronomically aligned to the Gaia DR2 catalog (Gaia Collaboration et al., 2018) and finally trimmed to cover the same area in all bands.

We masked several obvious CCD artifacts, which included stellar haloes and image “ringing”, i.e., clusters of two or three unrealistically bright and/or negative pixels, by eye on a case by case basis. We also masked noisy edges by creating a polygon region within the borders of the combined images. The final usable field of view (FoV) is 2.87 deg². At the redshift of VIK J2348–3054 this area is equivalent to 1034 pMpc²

We use the Lyman Alpha Galaxies in the Epoch of Reionization (LAGER) CDFS field observations from Hu et al. (2019), for comparison since it does not contain a known quasar or any other feature that would make it unusual at the redshift of VIK J2348–3054. This field was observed with the same telescope, instrument, and filters that we used to search for LAEs around VIK J2348–3054. The CDFS images have several CCD artifacts as well as large areas of low S/N, mainly within the chip gaps due to insufficient dithering. We manually masked CCD artifacts and automatically masked the chip gap regions in each image using the corresponding weights image. The effective area available for source detection after masking was $\sim 1.56$ deg², or $\sim 570$ pMpc² at the redshift of VIK J2348–3054.

We photometrically calibrate the CDFS images with the same method used for our DECam data.

While the narrowband depths are similar between the observations around VIK J2348–3054 and CDFS, the broadbands in the latter are about 2 magnitudes deeper (see Hu et al., 2019). To create a comparison sample with the same photometric properties, it is necessary to degrade the CDFS photometry to mimic our depths. To do this we create a source catalog using SExtractor in the same way that we did for our data and then assign new errors to every detected source based on the fitted exponential functions from the depth calculations for our data (discussed Sect. 2.1.3) resulting in a degraded magnitude error, $\sigma_{\rm degraded}$. To replicate the expected scatter due to a lower depth image, we randomly assign a new source magnitude which is drawn from a normal distribution with a mean equal to the magnitude determined in the CDFS image and a standard deviation equal to $\sqrt{\sigma_{\rm degraded}^{2}-\sigma_{\rm CDFS}^{2}}$, where $\sigma_{\rm CDFS}$ is the original magnitude error determined by SExtractor in the CDFS image.

A typical LAE has very strong Ly-$\alpha$ emission and a faint continuum (Malhotra & Rhoads, 2002; Ouchi et al., 2020). We plot a synthetic LAE spectrum at the redshift of VIK J2348–3054, along with the filter curves in Figure 1. The standard IGM extinction from Madau (1995) was assumed. The line was made by assuming a Gaussian profile and adopting a 200 km s^-1 FWHM value and an equivalent width of 50 Å. The luminosity is the L^∗ and was derived using the $\alpha=-1.5$ luminosity function, reported in Matthee et al. (2015). As expected, the strong Ly-$\alpha$ emission is contained within the NB964 filter, with practically no flux in the i-band, while the z-band is dominated by the source continuum. Given this, we adopt a similar selection criteria to that used by Bañados et al. (2013) and Mazzucchelli et al. (2017a), namely:

Note that when written like this, the requirement (Eq. 5) deviates at faint magnitudes from a strict lower limit in the $f_{\rm NB}/f_{z}$ flux ratio and in practice ends up requiring a detection in the z-band. This makes the color selection significantly more robust against interlopers.

We also require that the Ly-$\alpha$ emission within the narrowband is bright with respect to the continuum (Eq. 6). Figure 3 shows the estimated $z$-NB964 color over a range of redshifts for LAEs with different equivalent widths. For all of them, the $z$-NB964 color is above 1.3. The criterion in Eq. (5) implies that a source with a color of 1.3 requires a corresponding error of $<0.52$, allowing us to capture LAEs within the photometric scatter without incurring in significant contamination (see section 4.1).

Finally, we require a significant identification of the continuum break short of Ly-$\alpha$ (Eq. 7). Figure 1 shows clearly that the expected continuum emission on the blue side of the Ly-$\alpha$ line is very weak due to the IGM absorption in combination with the Lyman break, whilst the continuum on the red side should still be detected in the z-band. Our chosen criterion of i-z¿1 enforces this and limits the numbers of interlopers (see section 4.1). We also require SNR($i$) $<2$ (Eq. 8).

To further justify our selection we take the synthetic LAE spectrum from Figure 1, and derive the expected colors using the python package synphot³³3Version 1.2.1 in the redshift range $5<z<7.2$. This color-color track is shown in Figure 4 for equivalent widths of 10Å and 50Å. The Figure shows that the LAEs meet the $i-z$ criteria at the targeted redshift for all EWs of Lyman alpha considered in Figure 3. At the same time we tested two types of interlopers using the spectral energy distributions provided in Polletta et al. (2007): M82, and QSO2. These represent a star forming galaxy and a type-2 quasar respectively. We plot their color-color tracks from $0<z<1.3$ in green and blue and include no dust extinction and E(B-V) = 1. Neither of these two low-z interlopers enter our color selection (indicated as the top right green box in Figure 4). This justifies our color selections, demonstrating that these choices are optimal to select Ly-$\alpha$ emitters around the quasar redshift.

The color-color plot of all the detected sources in our field is shown in Figure 5. The grey points are the detected sources, the red dashed lines indicated the color selection criteria, and the red points highlight the LAE candidates which meet the selection criteria. The green box highlights the color selection for LAEs.

We find 39 sources sources in the VIK J2348–3054 field which meet our selection criteria: 38 LAE candidates as well as the quasar itself. The 20 pix $\times$ 20 pix (10.8” $\times$ 10.8”) postage stamp cut outs of all the candidates are presented in Figure 11 in the appendix. We also indicate the magnitudes in the respective bands as well as their S/N. While all candidates formally meet our selection criteria (see Sect. 2.3), we do note that three candidates (LAE-25, LAE-30, and LAE-38) seem to have obvious i-band detections despite SExtractor assigning them very low S/N values. By contrast, only two candidates were identified in the CDFS images after the photometry degradation (see Sect. 2.3).

Our candidate properties are presented in Table 2 in the appendix, including their positions, separation from VIK J2348–3054, i-band, z-band, and NB964 magnitudes, Ly-$\alpha$ luminosities, and star formation rates (see sections 3.2 and 3.3). VIK2348–3054 also met the selection criteria and is therefore also presented in Table 1. The z magnitude determined for the quasar is nearly 0.8 mag fainter than the one reported in Venemans et al. (2013), upon discovery, who measured a Gunn z filter magnitude of $22.96\pm 0.18$ on the 3.8m New Technology Telescope (NTT). The slight difference between the two is easily explainable by the difference in filter sets between DECam and NTT; the z-band in DECam is much redder than in EFOCS2 (because of the lower efficiency of the CCD at long wavelengths). This is also notable in Table 3 of Mazzucchelli et al. (2017b) where the PS1 z-band magnitudes differ from the DECam z-band photometry by substantial amounts due to the filter curve. Additionally, our DECam z value is also expected given the relationship between previous quasars’ DECam z magnitudes and their J band magnitudes (see Table in 3 in Mazzucchelli et al., 2017a) which have comparable redshifts and luminosities to VIKJ2348-3054.

The on-sky distribution of the LAE candidates around the quasar can be seen in Figure 6.

We follow the same method used by Hu et al. (2019) to determine the Ly-$\alpha$ luminosities of our candidates. Specifically, we use the following equation to estimate the flux density from the Ly-$\alpha$ emission line ($f_{\lambda,\text{line}}$),

where $T_{x}(\lambda)$ is the transmission function of filter $x$ (in our case $x$ is either NB964 or the z-band), $\bar{f_{\nu}}$ is the observed flux density, and $k_{x}$ is a constant defined as:

where $\bar{\lambda_{x}}$ is the central wavelength of filter $x$ and $c$ is the speed of light. For simplicity, we assume that the Ly-$\alpha$ emission line ($f_{\lambda,\text{line}}$) is a $\delta$-function, centred on the NB964 filter, and the continuum flux ($f_{\lambda,\text{cont}}$) is a power law spectrum with a slope of -2 and attenuated by the IGM using the model from Madau (1995). Making these assumptions allows us to express Equation (10) as:

where $f_{\alpha}$ is the Ly-$\alpha$ flux, $C$ is a constant, and $\tau$ is the average IGM absorption optical depth at the redshift of the quasar. When we substitute the narrowband and broadband values into Equation (12) we get two equations:

where $\lambda_{\alpha}$ is the Ly-$\alpha$ emission wavelength. We can solve Equations 13 and 14 simultaneously for $C$ and $f_{\alpha}$, using the latter to determine the Ly-$\alpha$ luminosity ($L_{\text{Ly}\alpha}$).

We estimate the star formation rates (SFRs) by following the same procedure used by Mazzucchelli et al. (2017a). Using the derived Ly-$\alpha$ luminosities ($L_{\text{Ly}\alpha}$), we can determine H$\alpha$ luminosities ($L_{\text{H}\alpha}$) by assuming case-B recombination and exploiting the relationship presented in Osterbrock (1989): $L_{\text{Ly}\alpha}=8.7\times L_{\text{H}\alpha}$. SFRs can be calculated using the relationship between SFR and $L_{\text{H}\alpha}$ presented in Kennicutt & Evans (2012):

However, it is worth noting that these star formation rate estimates are highly uncertain given that we do not know the intrinsic Ly-$\alpha$ luminosity due to star formation and the observed Ly-$\alpha$ luminosity is most likely a fraction of the intrinsic total.

Considering the number of candidates (section 3.1) and usable area of each field (section 2), we find the surface density of LAEs around VIK J2348–3054 to be $13.2\pm 2.2$ deg^-2, whilst in CDFS we only find a surface density of $1.3\pm 0.9$ deg^-2 to the same depth (see Sect. 2.3 for details). If we define $\delta=1$ as a density consistent with a field density, then this quasar is sitting within an overdense region with an overdensity factor of $\delta=10^{+5.3}_{-5.9}$, where the uncertainty corresponds to the 68$\%$ confidence interval. We note that given this highly skewed distribution, the probability of the field not being overdense (i.e, the probability that $\delta\leq 1$) is 3$\times 10^{-6}$. To further explore the overdensity factor and make sure the result is not driven by the CDFS density, we constructed a Ly-$\alpha$ luminosity function (see the values used in Table 1) and compared it to the combined CDFS and COSMOS fields from Hu et al. (2019) and the combined LAGER 4-field luminosity function from Wold et al. (2022) in Figure 8. We define the luminosity function as:

where $V_{\rm eff}$ is the effective volume of our survey–$1.8\times 10^{6}$ cMpc³–determined by taking into account the survey area and the redshift range probed, and $\Delta L$ is the bin-width = 0.1 $\log L$. We also include a detection completeness factor—$f_{\rm comp}$—as described in Hu et al. (2019) and Wold et al. (2022). This factor is calculated by inserting, and then recovering, false sources in the narrowband. The $f_{\rm comp}$ factor is shown in Figure 8 as a function of magnitude. Specifically, we use the best-fit error function shown in the Figure to estimate the luminosity function. To further constrain which equivalent widths we are sensitive to we also investigated what range of equivalent widhts would pass our selection criteria for a given Ly-$\alpha$ luminosity. In particular: we assumed zero flux in the i-band; for each bin in Figure 8 we calculated the percentage of synthetic LAEs that would pass our selection criteria, with equivalent widths ranging from 1 - 200 $\AA$; we then divide the luminosity function by this completeness correction factor, which results in the blue points in Figure 8. Our limit-assumption correction suggests an even larger overdensity. However, the uncommon shape of the LF observed after correction suggests that there is likely a more complex distribution of EWs, and the correct LF sits in between the black and blue dots.

While the overdensity is not clear in the two faintest bins, this may be due to incompleteness from our selection function. Further observations, both deeper and in other bands (as has been done for the LAGER survey, e.g., Hu et al. 2019) are likely required to determine the source of this. To estimate $\delta$, we fit the amplitude of the luminosity functions from Hu et al. (2019) and Wold et al. (2022) to our data, without considering the two faintest bins. We find that the candidates around VIK J2348–3054 have a higher space density. The luminosity functions from Hu et al. (2019) and Wold et al. (2022) are in tight agreement (as is evident in Figure 8), and both imply an overdensity of $\delta=4\pm 0.4$. These luminosity functions were corrected for selection incompleteness whereas our data has not been corrected in the same way. This has the consequence of the measured $\delta=4\pm 0.4$ value representing a lower limit, which again, indicates that the quasar lives in an overdense environment.

Interestingly, we are only able to find this overdensity due to the large size of our field, as there appears to be a distinct lack of sources in the vicinity of the quasar (see Figure 6). Figure 9 shows the LAE candidates’ radial distribution from the quasar. The blue shaded region represents the average CDFS surface density of $1.3\pm 0.9$ deg², with an uncertainty determined by considering the Poisson single-sided, $1\sigma$ upper limit for $n=2$ (Gehrels, 1986). Radial bins with no detections have been given an upper limit of 1.846 objects which is again the $1\sigma$ upper limit for a single-sided Poisson distribution. Figure 9 also shows the maximum radius that objects are expected to collapse into a cluster at $z=0$, 75 cMpc (9.5 pMpc) expected at this redshift (Overzier et al., 2009) and is also shown as the outer ring in Figure 6. Figure 9 shows that the field is overdense throughout all the area covered by our images, except within the inner  5pMpc from the quasar. We discuss this further in the next section.

Figure 6 shows the on-sky distribution of LAE candidates. The inner shaded region represents the quasar proximity zone—the theoretical region around a quasar where its UV radiation would have ionized a pure hydrogen ISM (Fan et al., 2006). Mazzucchelli et al. (2017b) determined the radius of the proximity zone of VIK J2348–3054 to be 2.64 pMpc. The Figure also shows ring with the largest radius that Overzier et al. (2009) found to contain overdensities. It is worth noting that there is a small observational artifact in the form of striations, to the south of the quasar, which overlaps the proximity zone. This is caused by a relatively strong main reflection pattern in DECam, possibly by moonlight. However, the area affected is very small and this effect had no impact on LAE candidate selection.

What is striking about the on-sky distribution is the discernible lack of candidates towards the center of the image, around the quasar position (also see Figure 9). In fact, the nearest candidate to the quasar is $\sim$ 0.27^∘ away, i.e., 5.15 pMpc. This distance is indicated by the dashed ring in Figure 6. If we take the area outside of the maximum radius at which objects are expected to collapse into clusters at $z=0$, i.e, the region beyond the outer ring in Figure 6, where we do not expect the quasar to have much, if any, influence, and use it to estimate the average field density, then the Poisson probability of finding the inner region void of LAEs is 1.2$\%$. However, when coupled with the fact that this underdense region is centered on the already rare luminous high-z quasar, we can conclude it is highly unlikely that this is a chance occurrence, but instead shows a real dearth of LAEs. Furthermore, we would naively expect the quasar to be close to the center of overdensity, where the environment should be more concentrated than in the outskirts, making this underdensity even more striking. In the following sections we discuss potential explanations for this observed lack of LAEs in the vicinity of VIK 2348-3054.

The varying results throughout the literature have been difficult to constrain and many explanations have been given for the scatter. In Figure 10 we summarize the results of previous studies that targeted 20 quasars in the redshift range $z=4.87$ to $z=7.08$, that searched for either LAEs or LBGs in their vicinity. Specifically, we show the search area and the value of $\delta$ determined by each study.

Of the 20 quasars only six were searched for neighboring LAEs. And of these six observations, none found overdensities. Furthermore there were no multiple LAE searches done on any of these quasars, meaning that only six LAE searches were done in total (Bañados et al., 2013; Mazzucchelli et al., 2017a; Kikuta et al., 2017; Goto et al., 2017; Ota et al., 2018). On the other hand, 16 of the quasars had LBG surveys done, and some quasars were studied multiple times by different groups. In total there were 24 LBG searches done, of which 16 (67$\%$) found overdensities, whilst 8 (33$\%$) did not. The fact that many LAE searches don’t detect overdensities, whilst at the same time, most LBG searches do, suggests that quasars may be suppressing star-formation in their environments. As we mentioned previously, the younger, lower mass LAEs would be more affected than the older, higher mass LBGs. Therefore, we would expect to find less overdensities when looking for LAEs. Alternatively, narrowband imaging might not have enough sensitivity if the equivalent widths are low.

Several quasars have had both LAEs and LBGs searches done. VIKJ1030–0524 was investigated for LBGs by Morselli et al. (2014) and Balmaverde et al. (2017). Later, Mignoli et al. (2020) performed spectroscopy using MUSE at VLT and spectroscopically identified the LBGs within their FoV, and found two LAE within 5 pMpc of the quasar. Mignoli et al. (2020) themselves note that the low number of detected LAEs close to the quasar may be hinting at negative feedback; confirming an overdensity in general but lacking in LAEs.

Ota et al. (2018) searched for both LBGs and LAEs around VIK J0305–3150 and found, at low masses, a slight overdensity of LBGs and an underdensity of LAEs. Later, Champagne et al. (2023) reported a massive overdensity of LBGs very near the same quasar. Whilst narrowband searches for LAEs around the quasar have resulted in non-detections, spectroscopic observations, in particular with MUSE on the Very Large Telescope, identified a LAE very near to the quasar (Farina et al., 2017). Recently, James Webb Space Telescope (JWST) slitless spectroscopy has identified 10 [OIII] emitting galaxies, which trace a filamentary structure around this quasar (Wang et al., 2023). These results highlight the importance that spectroscopic follow up can play in determining overdensities.

Utsumi et al. (2010) and Goto et al. (2017) both observed CFHQS J2329–0301 for LBGs and LAEs respectively, with Subaru/Suprime-Cam. Once again, the quasar had an overdensity of LBGs and an underdensity of LAEs.

Bañados et al. (2013) and Mazzucchelli et al. (2017a) searched for LAEs around ULAS J0203+0012 and PSO J215.1512–16.0417 respectively, and also performed Lyman-break selections around their respective targets. They found that the number of LBG candidates was consistent with blank fields. It is worth noting that both used the FOcal Reducer/low dispersion Spectrograph 2 (FORS2) on the Very Large Telescope (VLT) which has a FoV of 37 arcmin² and they did not use an optimum filter set up, which may have influenced results.

Recently, Champagne et al. (2023) performed an LBG search around the target of our study for LAEs—VIK J2348–3054—using the Advanced Camera for Surveys (ACS) onboard the Hubble Space Telescope (HST). Whilst they claim no LBG enhancement, they still find 10 LBG candidates within 0.3 pMpc, well within the quasar proximity zone (5 secure detections and 5 marginal detections). On the other hand, we do not find any LAEs within a much larger area around VIK J2348–3054. Once again, the lack of LAE and the detection of LBGs is expected, further hinting at star formation suppression.

There is also evidence at lower redshifts ($z\sim 4$) which suggest a similar effect might be occurring. Specifically, García-Vergara et al. (2017) used a custom set of narrowband filters which were optimized for identifying LBGs around six quasars; the QSO-LBG cross-correlation function showed an overdensity. Later, García-Vergara et al. (2019) performed a similar work, but for LAEs around 17 quasars and again the QSO-LAE cross-correlation function found an overdensity. However, the LBG overdensity signal was stronger than the LAE one by a factor of 10 (see the bottom panel of Figure 6 in García-Vergara et al., 2022).

The notion of LAE suppression due to the quasar seems to explain a lot of the results throughout the literature. However, at the same time, our study is the only one which has found an overdensity of LAEs around a high-redshift quasar ($z>5$) using the narrowband search technique. And is the largest area ever searched for either LBGs or LAEs (see the solid purple star in Figure 10). Moreover, had we observed VIK J2348–3054 with any other observational set up, on any of the other telescopes that were used in previous observations, we would not have detected an overdensity. This would seem to suggest that the lack of overdensities of LAEs in previous studies might have been due to not probing a large enough area, and wider FoV studies would possibly have revealed overdense regions.