THE IMPACT OF JWST BROADBAND FILTER CHOICE ON PHOTOMETRIC REDSHIFT ESTIMATION

Our first galaxy sample contains 2422 galaxies with secure spectroscopic redshifts (z_spec) from the ESO public compilation in the Chandra Deep Field South (CDFS; Cristiani et al. 2000; Croom et al. 2001; Bunker et al. 2003; Dickinson et al. 2004; Stanway et al. 2004a, 2004b; Strolger et al. 2004; Szokoly et al. 2004; van der Wel et al. 2004; Doherty et al. 2005; Le Fèvre et al. 2005; Mignoli et al. 2005; Ravikumar et al. 2007; Popesso et al. 2009; Balestra et al. 2010; Silverman et al. 2010; Kurk et al. 2013; Vanzella et al. 2014) and from Morris et al. (2015). In order to obtain multiwavelength photometry for these galaxies, we cross-matched this sample with the public Cosmic Assembly Near-infrared Deep Extragalactic Legacy Survey (CANDELS; Grogin et al. 2011; Koekemoer et al. 2011) catalog for the Great Observatory Origins Deep Survey South field (GOODS-S) obtained by Guo et al. (2013). This catalog contains photometry in 17 broadbands, from 0.37 through 8.0 μm, obtained with ground- and space-based telescopes, including HST and Spitzer (Ashby et al. 2015). This wavelength sampling allows for a good-quality SED fitting of the galaxy stellar emission and a proper redshift recovery in the vast majority of cases, using standard galaxy spectral templates and photometric redshift codes.

Our aim here is to assess the impact of choosing different JWST filter sets on the derived photometric redshifts, rather than the performance of different photometric redshift codes and/or SED template libraries. Therefore, we explicitly excluded from our analysis any galaxy from the parent spectroscopic sample for which we could not correctly recover the redshift photometrically using all the CANDELS photometry up to 4.5 μm. We did not use the two IRAC band data at longer wavelengths because their signal-to-noise ratio (S/N) is considerably lower than for the other bands and they could be contaminated by polycyclic aromatic hydrocarbon emission at low z. The excluded galaxies are only ∼8% of all the galaxies with publicly available z_spec in the CDFS.

To obtain photometric redshift estimates for all the galaxies in the parent spectroscopic sample, we ran the public code LePhare (Arnouts et al. 1999; Ilbert et al. 2006) using templates from Bruzual & Charlot (2003, hereafter BC03) with solar metallicity, a range of exponentially declining star formation histories with different characteristic times τ from 0.01 to 10 Gyr, and ages from 0.01 to 10 Gyr. We allowed for emission lines in LePhare and applied the Calzetti et al. reddening law (Calzetti et al. 2000) with A_V extinction values from 0 to 4.

We considered the subsample of 2422 objects with resulting photometric redshifts within 2σ (σ = 0.13) of the spectroscopic redshifts (Figure 1), which constitute 92% of the parent spectroscopic sample, and adopted it as our sample 1 for all further analysis. The sample 1 redshift distribution is shown in Figure 2. This distribution peaks around z ∼ 1 and has a high-z tail up to z = 6, although the statistics are rather poor at z > 4. So, our main purpose with sample 1 is to study the percentage of low-redshift galaxies that can leak to high z when different filter sets, and thus different wavelength coverages, are available.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Our second sample consists of 1375 galaxies at z = 4–7 taken from the CANDELS GOODS-S public catalog. For this sample, we considered as fiducial redshifts the consensus photometric redshifts that have been compiled by Dahlen et al. (2013). The CANDELS consensus photometric redshifts result from the comparison of 11 independent SED-fitting runs. We considered only CANDELS galaxies with z > 4 here, as lower redshifts are well represented in our sample 1.

As we did for the sample with spectroscopic redshifts (sample 1), we obtained our own photometric redshifts by running LePhare on the CANDELS 15-band photometry and using the same SED templates and parameter values stated before. In this case, our own photometric redshifts agree within 2σ (σ = 0.10) with the CANDELS consensus redshifts for 1375 (95%) of the GOODS-S galaxies with CANDELS redshifts z_CANDELS = 4–7 (Figure 3), consistently with the typical level of dispersion present in the individual SED-fitting CANDELS runs. We adopted these 1375 galaxies as our sample 2. The resulting redshift distribution is shown in Figure 2.

Zoom In Zoom Out Reset image size

We built our third sample with mock galaxies at fixed redshifts z = 7–10 (for an analysis of more redshifts, see Appendix A), simulating their SEDs with two different methods. First, we considered the BC03 synthetic galaxy templates with four metallicities (1, 0.4, 0.2, and 0.02 Z_⊙), exponentially declining star formation histories, and five different ages, depending on redshift. We applied dust extinction by using the Calzetti et al. reddening law with A_V = 0–1 and the effect of absorption by the intergalactic medium (IGM; Madau 1995).

The BC03 models do not include emission lines, so we manually added the Hα, Hβ, [O iii] λ5007 and [O ii] λ3727 emission lines to those templates for which the age is equal to or lower than the characterisitic τ of the exponentially declining star formation history, as follows. We started from the observed mean value of the Hα+[N ii]+[S ii] equivalent width (EW), which is 422–423 Å at z = 3.8–5 (Smit et al. 2015). Then, we assumed line ratios from Anders & Fritze-v. Alvensleben (2003) in order to derive the EWs of Hα alone at different metallicities. Fumagalli et al. (2012) obtained a relation between Hα EW and redshift, namely, EW(Hα) $\propto {(1+z)}^{1.8}$, which has been confirmed to be consistent with observations up to z ∼ 5 (Stark et al. 2013).¹⁴ Following this relation, we derived EW(Hα) for all considered redshifts. We also derived EWs for Hβ assuming case B recombination and the corresponding EWs for [O iii] and [O ii] by using the line ratios tabulated in Anders & Fritze-v. Alvensleben (2003). We explicitly decided not to include the Lyα line because the importance of its emission in galaxies at $z\gt 7$ is still unclear (e.g., Stark et al. 2013; Konno et al. 2014; Pentericci et al. 2014). Finally, we normalized every model at 29 AB mag at 1.5 μm (corresponding to the NIRCam F150W filter pivot wavelength).

The complete list of adopted parameter values is summarized in Table 1, while some SED examples at z = 7 are shown in Figure 4. Our sample contains a total of 864 galaxies, corresponding to 240 BC03-based SED models at each redshift 7 and 8, and 192 SED models at each redshift 9 and 10. For each model, we considered independently three possible S/N values (S/N = 3, 5, and 10) for the flux corresponding to [F150W] = 29 AB mag. The S/N values of the photometry in all other bands have been scaled, as we explain in Section 2.2.1.

Zoom In Zoom Out Reset image size

Second, we simulated further z = 7–10 galaxy SEDs with the population synthesis code Yggdrasil (Zackrisson et al. 2011), which has been especially conceived to model high-redshift galaxies. We used a set of parameter values similar to the ones used for the BC03 templates, except that the star formation histories are somewhat different, consisting of a step function for which the duration of the constant star formation period is set as a free parameter. And, in addition to the normal stellar templates, we also adopted a template with a metal-free, Population III single stellar population (Schaerer 2002). In all cases, we considered the two possible extreme gas covering factors, namely, f_cov = 0 and 1. The complete list of adopted parameter values is given in Table 2. We show four SED template examples for z = 7 galaxies in Figure 5.

Zoom In Zoom Out Reset image size

Table 2.  Parameter Values Used to Create Yggdrasil SED Models for Simulated Galaxies at z = 7–10

Parameter	Values
Metallicity (Z⊙)	1, 0.4, 0.2, 0.02,a 0a
Constant SFH (Myr)	10, 30, 100
AV	0, 0.5,1
Age (Gyr)	0.01, 0.05, 0.2, 0.4, 0.6b
fcov	0, 1
z	7, 8, 9, 10
S/Nc	3, 5, 10

Notes.

^aFor this metallicity we considered only ages $t\lt$ 0.2 Gyr. ^bWe considered this age only up to redshift z = 8. ^cFor the flux corresponding to the normalization magnitude, i.e., [F150W] = 29 mag.

Download table as:  ASCIITypeset image

The Yggdrasil models include the effects of nebular line and continuum emission, so we did not need to add any extra line manually as we did for the BC03 templates. We did include the effect of dust reddening and IGM absorption in the same way as for the BC03 models. We removed the Lyα emission line from the Yggdrasil models because its presence and importance are debated at $z\gt 7$, as we explained before. Once again, we normalized every SED template to [F150W] = 29 mag at 1.5 μm. Our Yggdrasil simulated sample comprises a total of 1260 galaxies, corresponding to 342 SED models at each redshift 7 and 8 and 288 SED models at each redshift 9 and 10. The complete redshift distribution of our galaxy sample 3, simulated with both the BC03 and Yggdrasil templates, is shown in Figure 2.

Based on the best-fit SEDs obtained running LePhare on the CANDELS photometry for the galaxies in samples 1 and 2 and the simulated SEDs for our mock galaxies in sample 3, we measured the photometry expected at the eight NIRCam broadbands and the two MIRI broadbands F560W and F770W for all our galaxies (see Table 3 for a summary of the different JWST filters considered in this work). Our procedure to interpolate/extrapolate the JWST photometry is similar for our three galaxy samples. The main difference is that in samples 1 and 2, we simply adopted S/N values for the photometry scaled from the CANDELS photometry, while for sample 3 we tested three different S/N values in a controlled manner, in order to assess the effect of S/N on our results.

Table 3.  List of JWST Filters Considered in This Work, Their Pivot Wavelengths, and Their Bandwidths

Filter Name	λ (μm)	Bandwidth (μm)a
NIRCam/F070W	0.70	0.18
NIRCam/F090W	0.90	0.23
NIRCam/F115W	1.15	0.30
NIRCam/F150W	1.50	0.38
NIRCam/F200W	1.99	0.53
NIRCam/F277W	2.77	0.82
NIRCam/F356W	3.56	0.99
NIRCam/F444W	4.41	1.29
MIRI/F560W	5.60	1.41
MIRI/F770W	7.70	2.30

Note.

^aThis bandwidth corresponds to the entire wavelength range in which the filter transmission is $\gt 1 \%$.

Download table as:  ASCIITypeset image

As a first step, for every galaxy in our real galaxy samples (samples 1 and 2), we derived NIRCam broadband fluxes by convolving the best-fit SED template with the transmission curve of each JWST filter (Meyer et al. 2004). To each flux, we assigned an error bar corresponding to an S/N equal to that of the closest filter in the CANDELS catalog. We adopted as NIRCam flux a random value (selected from a uniform distribution) within the corresponding error bar.

Note that, in practice, NIRCam will reach the depth of the HST CANDELS photometry in very short integration times, so the S/N achieved for CANDELS galaxies in typically deep NIRCam surveys will be higher than what we consider here. However, our assumed S/Ns are already sufficiently high in most cases: virtually all galaxies in sample 1 and more than 90% of galaxies in sample 2 have ${\rm{S}}/{\rm{N}}\geqslant 5$ in their derived NIRCam F150W photometry. Thus, our results derived here for these real galaxy samples are mostly independent of S/N, as we explain below (the low-S/N effect will be more clearly manifested in our sample 3, where we explicitly set three different S/N values for each template, including a low value of S/N = 3).

As a second step, we measured the photometry expected at the MIRI bands for our sample 1 and 2 galaxies, which results from the extrapolation of the best-fit template redward of the CANDELS IRAC 4.5 μm coverage. We convolved the best SED model with the two MIRI bands' transmission curves (Bouchet et al. 2015; Glasse et al. 2015). To each MIRI band flux, we assigned an error bar corresponding to an S/N scaled using as reference the S/N in the NIRCam F150W band: we assumed that an MIRI F560W or F770W flux corresponding to a magnitude 28 (AB) would have the same S/N as a NIRCam F150W flux corresponding to 29 mag (whose S/N was, in turn, scaled assuming that the "measured" NIRCam F150W flux S/N for each galaxy was the same as the real CANDELS HST F160W flux S/N). Fixing the S/N in the NIRCam and MIRI photometry implicitly assumes specific integration times with the JWST instruments. In practice, this assumption does not matter, as our results will be equally valid for any galaxies with similar S/N values, independently of the MIRI fluxes at which these S/Ns are reached.¹⁵ Finally, we randomized every flux within its own error bar.

For sample 3 with mock galaxies, we convolved every simulated SED template with the transmission curve of each NIRCam filter. In our sample every SED template is considered three times, in order to apply three fixed, reference S/N values (10, 5, and 3) to the NIRCam F150W fluxes. We assumed the same integration times for all the other NIRCam bands and took into account the differences in filter sensitivities to assign S/N in each band. A different strategy that assumes the same observational depth for all NIRCam bands is presented in Appendix B. We considered every flux below the 2σ level as a nondetection. The 2σ detection limits in each band for each fixed reference S/N value at F150W are listed in Table 4. Note that, as a conservative approach, we adopted 3σ flux upper limits (rather than 2σ) in all cases of nondetections for running the z_phot code, as is explained below.

Finally, we obtained 10 random realizations of each NIRCam band flux within the corresponding error bar. Note that, in contrast, for samples 1 and 2 we considered a single realization of each flux, simply because in the case of real galaxies the NIRCam fluxes are well constrained by interpolation of the CANDELS fluxes, so their error bars are typically smaller than in sample 3.

In a similar way, we derived fluxes in the MIRI bands by convolving all SED templates in the sample with the two MIRI filter transmission curves. Then, we scaled each MIRI flux S/N, such that a mag_AB = 28 in MIRI has the same S/N as a NIRCam magnitude [F150W] = 29. This is a realistic assumption of how the NIRCam and MIRI data will typically be matched. As for the NIRCam bands, we randomized every MIRI flux 10 times within its own error bar.

Figures 12–15 show the distributions of the recovered photometric redshifts for each input redshift (z = 7–10) for the sample 3 galaxies simulated with the BC03 templates, obtained by running LePhare with different JWST filter combinations. Since in this case the input redshifts are four fixed values, it is clearer to show the output z_phot distributions for each input redshift rather than ${z}_{\mathrm{phot}}-{z}_{\mathrm{phot}}$ plots, which would contain four vertical columns each.

Zoom In Zoom Out Reset image size

As before, in each plot we quote the mean value of the normalized redshift difference distribution $\delta z=({z}_{\mathrm{phot}}-{z}_{\mathrm{fiduc}.})/(1+{z}_{\mathrm{fiduc}.})$, taking into account all galaxies. However, for sample 3 galaxies, we adapted our method for the rms calculation. As can be seen in Figures 12–15, the sample 3 redshift histograms are characterized by quite a narrow distribution around the main peak and the presence of some catastrophic failures producing secondary peaks at much lower redshifts. These catastrophic redshift failures would significantly bias the computed σ of $| {z}_{\mathrm{phot}}-{z}_{\mathrm{fiduc}.}| /(1+{z}_{\mathrm{fiduc}.})$ if they were considered in the statistics. So, for sample 3 galaxies (both based on the BC03 and Yggdrasil models) we explicitly excluded all galaxies with $| {z}_{\mathrm{phot}}-{z}_{\mathrm{fiduc}.}| /(1+{z}_{\mathrm{fiduc}.})\gt 0.15$ for the σ computation, and we considered those excluded galaxies to be the outliers of our sample. We provide the statistics on the outlier fractions for the BC03 SED galaxies in Table 7.

For S/N = 10, the eight NIRCam broadbands alone allowed us to derive accurate redshift estimates, with mean values for the redshift distributions close to 0 and $\sigma \leqslant 0.011$ at all redshifts (Figure 12). The percentage of outliers quoted in Table 7 is very small (0.1%). Adding MIRI photometry to the NIRCam photometry only slightly enhances the already excellent quality of the output redshifts, resulting in $\sigma \lt 0.01$ at all redshifts. These results indicate that the incorporation of the MIRI photometry into the NIRCam photometry has in practice little effect on deriving photometric redshifts for z = 7–10 galaxies with BC03-like SEDs when the NIRCam S/N is very high.

The MIRI effect on the photometric redshift estimation is more evident at S/N = 5 (Figure 13). The NIRCam data alone are sufficient for a reasonably good redshift recovery (σ ≈ 0.01–0.03), but the addition of MIRI data improves the photometric redshift quality, taking the outlier percentage to $\lt 1 \%$ (when F560W photometry is included). Inspection of Figure 4 suggests that this is a k-correction effect: galaxies are brighter in the MIRI bands either because they are old or because of the presence of emission lines. Indeed, both effects are important. In Figure 14 we show separately our results only for galaxies with emission lines: it can be seen that the effect of incorporating the MIRI photometry is similar to the general case with S/N = 5. Note that, on the one hand, the presence of emission lines helps in the redshift identification, but on the other hand having prominent emission lines in contiguous broadbands (Figure 16) makes the observed photometry mimic a less featured SED, resulting in a more difficult redshift recovery.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

In the case of S/N = 3 at F150W (Figure 15), assuming identical integration times in all NIRCam filters implies that some galaxies are undetected at the shortest-wavelength NIRCam bands. In this case, having complementary MIRI photometry becomes very important, even with the assumption that we made here, i.e., that the MIRI depth will be 1 mag shallower than the NIRCam depth at 1.5 μm. This assumption is reasonable, as achieving the same depth as in NIRCam will require too long exposure times.¹⁶

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The result of this poorer-quality photometry is that the output redshift distributions are broader and secondary peaks at lower redshifts become more significant (Figure 15). The incorporation of MIRI photometry has a significant effect in reducing the overall outlier fraction: it goes down from 10.1% with NIRCam data only to 3.0% after adding the two MIRI bands (Table 7). Note that, among the two MIRI filters, F560W appears to have a more important effect in improving the z_phot determination of the BC03 SED sources at z = 7–10.

As a summary, we conclude that considering MIRI photometry of 1 mag shallower depth than the NIRCam photometry keeps the fraction of z_phot outliers low for the $z\geqslant 7$ galaxies that have a moderate or low S/N in the NIRCam images. This effect is evident for different SED types (with or without emission lines).

Figures 17–20 show the distributions of the recovered photometric redshifts for the sample 3 galaxies simulated with the Yggdrasil templates at each fixed redshift, obtained by running LePhare with different JWST filter combinations. The outlier percentages for each analyzed S/N value are listed in Table 8. Yggdrasil templates are more complex than the BC03 ones, given that Yggdrasil automatically incorporates nebular emission lines and continuum emission when galaxy ages are young and the gas covering factor is ${f}_{\mathrm{cov}}\gt 0$, which may be the case of many high-z galaxies.

Zoom In Zoom Out Reset image size

When considering objects with S/N = 10 at F150W (Figure 17), we see that the NIRCam data alone provide a good photometric redshift recovery (σ = 0.01–0.02). The incorporation of the MIRI bands has little effect on the resulting σ values in most cases, and the fraction of catastrophic outliers is only slightly reduced (see Table 8).

As in the case of BC03 SED galaxies, the impact of incorporating MIRI photometry is more evident at S/N = 5 in F150W (Figure 18). The overall percentage of catastrophic outliers reduces from 20.4% to 12.1% by adding MIRI photometry in this case (Table 8).

Zoom In Zoom Out Reset image size

If we consider only those galaxies with emission lines in their SEDs (i.e., young and with gas covering factor f_cov = 1), also at S/N = 5, we see two interesting effects (Figure 19 and Table 8). First, the degeneracies in redshift space are more important than for the general galaxy sample with S/N = 5. This is manifested in a much larger outlier percentage, namely, ∼49% when considering the NIRCam data alone. Second, the effect of incorporating MIRI photometry becomes more noteworthy: including both MIRI bands reduces the percentage of outliers to ∼17%. These effects are produced by two factors: on the one hand, the nebular line emission boosts the MIRI fluxes and, thus, the S/N of the MIRI data; on the other hand, the multiple emission lines in the Yggdrasil templates with nebular emission result in similarly bright observed fluxes in contiguous filters, masking the continuum SED shape and making the SED fitting more difficult (as explained in Section 3.3.1), so having a longer baseline is necessary for a proper redshift recovery.

Zoom In Zoom Out Reset image size

For the lowest-quality photometry (S/N = 3; Figure 20), the percentage of catastrophic outliers in the entire sample is 44.4% using only NIRCam photometry. The advantage of incorporating MIRI data, which also have lower S/N, is less obvious in this case, as the percentage of z_phot outliers still remains high even with the two MIRI bands (36.7%). A significant fraction of galaxies are detected in only a few (NIRCam and MIRI) bands, so the photometry provides poor constraints for the SED fitting, making degeneracies in redshift space become evident.

Zoom In Zoom Out Reset image size

Indeed, as is the case for the BC03 SED galaxies, the catastrophic outliers correspond mainly to sources with detections in four or fewer JWST bands (open regions in the histograms of Figure 20). However, in the case of the Yggdrasil SED galaxies, the secondary peaks appearing in the output redshift distributions are much more pronounced than for the BC03 galaxies. These secondary peaks are mainly produced by two kinds of templates that are not represented in our BC03 SED sample: (1) young galaxies with nebular emission, where the combination of multiple emission lines and low S/N produces an enhancement in the number of catastrophic outliers; and (2) young galaxies with f_cov = 0 (i.e., no emission lines) and the highest extinction adopted to build our mock sample (i.e., A_V = 1). These galaxies have no analogs either in the BC03-based sample, as we explicitly incorporated the main emission lines in all galaxies with ages lower than the characteristic τ in the star formation history.

As a summary, we conclude that the benefit of considering MIRI photometry along with the NIRCam data for Yggdrasil SED galaxies is mainly obvious for NIRCam sources with moderate S/N values, and particularly helpful in the case of SEDs with nebular emission. At low S/N the advantage is less obvious. This is in contrast to our conclusion for the BC03 SED galaxies, indicating that at low S/N the importance of having MIRI data depends significantly on the galaxy spectral type.

In this section we show the MIRI magnitudes and NIRCam/MIRI colors of our simulated galaxies in sample 3, as this can be a useful reference when planning JWST observations. In particular, we analyze how the MIRI detections vary with the galaxy SED type. Given our chosen normalization [F150W] = 29 mag, all these galaxies will easily be observed in NIRCam deep surveys (as it takes only ∼30 minutes of integration per pointing to reach [F150W] = 29 mag, 3σ with NIRCam, according to the public JWST/NIRCam exposure time calculator¹⁷ ). So our aim here is to discuss the fractions of such galaxies that will be detected in typically deep MIRI observations. Of course, these same results only need to be rescaled to remain valid for other NIRCam magnitudes.

Figures 21–24 show the expected MIRI magnitude distributions for our sample 3 galaxies generated with the BC03 and Yggdrasil models. As the NIRCam F150W magnitude for all these galaxies has been fixed to [F150W] = 29 mag, there is a direct correspondence between the MIRI F560W magnitudes and the F150W–F560W colors, which are also shown in each plot. For the BC03 galaxies (Figures 21 and 22), we analyzed separately the cases with and without emission lines, which here correspond to galaxies with ages lower and higher than their characteristic SFH decay parameter τ, respectively. For the Yggdrasil galaxies (Figures 23 and 24), we analyzed three cases: galaxies with ages larger that their constant star formation period, which are galaxies in passive evolution; and galaxies that are still forming stars with two different gas covering factors, f_cov = 0 (implying no line or nebular continuum emission) and 1 (with nebular emission). The median values of all distributions are provided in Tables 9 and 10.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The magnitude distributions in Figure 21 show that the vast majority of BC03 galaxies with [F150W] = 29 mag at z = 7, 8, and 10, and more than half of those at z = 9, will be detected by an MIRI survey with a detection limit [F560W] = 28 mag. This result is valid for galaxies with or without emission lines. All the latter are cases in which a 4000 Å break is at least partly developed, so the SEDs are brighter at the MIRI wavelengths than at 1.5 μm, particularly at z = 7 (see Figure 4). This k-correction effect causes the median [F560W] value at this redshift to be brighter for galaxies without emission lines, as those with emission lines correspond to younger galaxies with fainter continua that partly cancel the emission-line flux enhancement (Table 9). At z = 8 and 10, instead, the combination of the continuum k-correction and the nebular emission results in a net flux boosting in the case of emission-line galaxies.

For F770W (Figure 22), the flux enhancement due to the emission-line presence is only evident at z = 9 and 10, as the Hα emission line enters the F770W wavelength window at these redshifts (Table 10). Most young galaxies at z = 7 and 8, instead, remain below our reference [F770W] = 28 mag detection limit, as their continua are blue and no emission line is present in this filter to help with the detection. So, the F770W filter will only allow us to detect evolved galaxies at $z\geqslant 7$ or young line emitters at z = 9 and 10 to an [F770W] = 28 mag cut.

The vast majority of the Yggdrasil SED galaxies with [F150W] = 29 mag that finished forming stars at $z\geqslant 7$ will have mag $\lt 28$ in both MIRI bands (Figures 23 and 24, and Tables 9 and 10). This is the result of the same k-correction effect discussed above.

Zoom In Zoom Out Reset image size

For the Yggdrasil star-forming galaxies, there is a clear enhancement of the MIRI fluxes produced by the nebular lines (compare the cases with f_cov = 0 and 1 in Figures 23 and 24 and the corresponding tables). In this case, the flux enhacement is more obvious than for the BC03 templates because here we are comparing only young galaxies with and without lines (segregated by their different gas covering factors), so all their continua are similarly blue. In both cases, however, the median MIRI magnitude values are $\gt 28$ mag. This implies that an MIRI galaxy survey to our reference depth will only detect a minority of star-forming galaxies with [F150W] = 29 mag at $z\geqslant 7$ if their SEDs are similar to the Yggdrasil templates. Note that these galaxies are considerably brighter in F560W if they are at z = 7 or 8, while slightly brighter in F770W if at z = 9 or 10.

Note that, conversely, our color analysis shows that detecting more than one-half of red MIRI galaxies at $z\geqslant 7$ will require NIRCam data at least 3 mag deeper. This is the case both for "passive galaxies" (i.e., galaxies that finished forming stars) and for galaxies that are red simply because of the presence of prominent emission lines.

So, as a summary, "old" galaxies at $z\geqslant 7$ will easily be detected with MIRI in reasonably deep surveys, independently of their galaxy type. For young galaxies, the situation is more complex, as it depends on to what extent line emission can compensate for the very faint continua that are characteristic in these sources. F560W will provide more detections than F770W if maps of the same depth are obtained in both passbands.