Polycyclic Aromatic Hydrocarbon and CO (2-1) Emission at $50{-}150$ pc Scales in 70 Nearby Galaxies

We use JWST MIRI and NIRCam imaging of 19 galaxies from the PHANGS-JWST Cycle 1 Treasury (GO 2107, PI: J. Lee; Lee et al., 2023). These data cover the F335M, F770W, and F1130W bands, each of which captures a strong PAH feature (e.g., Tielens, 2008). The F335M filter captures the prominent 3.3 $\mu$m feature, attributed to C–H stretching modes of small PAHs, while the F770W filter captures the 7.7 $\mu$m PAH feature, attributed to C–C stretching modes of larger PAHs, and the F1130W filter captures the 11.3 $\mu$m feature, attributed to C–H out-of-plane bending modes of larger PAHs (Tielens, 2008).

We also analyze 51 galaxies from a new Cycle 2 Treasury (GO 3707, PI: A. Leroy), which we describe in Appendix A. These new observations include the F335M and F770W filters. From the 54 available Cycle 2 galaxies, we remove NGC 3344 due to the lack of CO (2-1) observations, NGC 5530 because the observations are still to be executed, and NGC 1068 because the background level in the JWST images remain too uncertain for the current analysis. In total we analyze 70 galaxies. The distribution of gas-phase metallicities of these galaxies has median $12+\log\mathrm{O/H}=8.49$ dex and robust standard deviation of 0.09 dex.

Reduction for both data sets was done using the PHANGS-JWST pipeline (pjpipe³³3https://pjpipe.readthedocs.io/en/latest/), following Williams et al. (2024). pjpipe is a wrapper on jwst⁴⁴4https://jwst-pipeline.readthedocs.io/en/latest/index.html, tailored to produce mosaics of images with extended emission.

The Rayleigh-Jeans tail of the stellar continuum contributes to F335M and F770W (more so for F335M). This contribution must be subtracted from the total surface brightnesses to obtain emission from PAHs. We calculate stellar-continuum-subtracted F770W surface brightness $I_{\rm F770W}^{\rm PAH}$ by subtracting F200W (Cycle 1) or F300M (Cycle 2) times a scaling factor (one factor per band) following Sutter et al. (2024). In our sample, the F770W stellar continuum correction (F770W${}_{\star}=0.22\times$F300M, or $0.13\times$F200W for Cycle 1) tends to be largest in the gas-poor, high stellar surface density inner regions of galaxies, including bars and bulges, which have high stellar-to-gas ratios. The coverage of F300M and F200W is smaller than that of F770W, and so for pixels without F300M or F200W data we subtract median(F770W_⋆/F770W)$\times$F770W, where the median (which is 2–10%, in agreement with Whitcomb et al., 2023b) is computed separately for each galaxy.

For F335M data, the stellar continuum subtraction is critical to any estimate of PAH emission (Sandstrom et al., 2023b). We use the continuum subtraction method defined by H. Koziol et al. (in preparation) based on Sandstrom et al. (2023b), where the F300M and F360M (for Cycle 1) are used to estimate continuum. We denote the continuum-subtracted F335M surface brightness as $I_{\rm F335M}^{\rm PAH}$. A similar effort using the F300M is underway for Cycle 2 but not yet ready, and so our analysis of $I_{\rm F335M}^{\rm PAH}$ focused only on the Cycle 1 targets.

The F770W filter also captures emission from hot, very small dust grains, but this has a small effect (in the 19 PHANGS-JWST Cycle 1 galaxies, 7.7 $\mu$m PAH emission is found to be about $5\times$ brighter than the underlying continuum; Baron et al., 2024a) and we do not subtract any underlying dust continuum. In both data sets, we also masked out a few bright stars and background galaxies with elliptical apertures that cover all of the emission from these sources on top of the F770W images. We do not use any data within these apertures.

We use PHANGS–ALMA CO (2-1) observations, which are described in Leroy et al. (2021). All maps include short- and zero-spacing data and are expected to achieve full flux recovery. The native resolution is $\approx 1.0\arcsec$, varying slightly from galaxy to galaxy. We use a noisy but uniform and high-completeness version of the CO (2-1) maps to ensure that $>90\,\%$ of the CO flux enters our analysis and that our mean trends will be unbiased by signal-to-noise based clipping of the CO data (see Leroy et al., 2023b). These “flat” maps are produced following Neumann et al. (2023) using a procedure analogous to spectral stacking (Schruba et al., 2011). For each line of sight, we integrate over a fixed-width velocity window around the local mean velocity defined via either low-resolution CO emission, H$\alpha$ emission, H I emission, or a model of the circular rotation, according to whichever yielded the most complete coverage and coherent reference velocity field for each galaxy. The velocity window is adapted to the disk-average line width of each individual galaxy and can vary from 20 to 200 km/s. We additionally ensure that all significant CO emission is included by combining the fixed-width velocity mask with the “strict” mask from Leroy et al. (2021). These “flat” masks will capture all of the CO emission along each line of sight and have well-defined noise, making them ideal to calculate the mean CO (2-1) intensity, $I_{\rm CO}$, as a function of $I_{\rm PAH}$.

We analyze how the CO-to-PAH correlations vary as functions of environment within galaxies, focusing on three specific environments: 1) galaxy centers; 2) regions outside galaxy centers with prominent nebular emission (“nebular regions”), which are $\gtrsim 80\%$ H II regions but also include a $\sim 10\%$ contribution from supernova remnants (Li et al., 2024); and 3) regions outside of centers and not covered by the nebular region mask (“diffuse regions”). Following Pathak et al. (2024) these three regions show distinct distributions of mid-IR intensity. These three regions likely show differences in radiation field strength and/or hardness, PAH abundance, $\mathrm{H_{2}}$/H I ratio, and dust-to-gas ratio. Additionally, PAH emission is found to be systematically suppressed inside H II regions (Chastenet et al., 2023a; Egorov et al., 2023; Sutter et al., 2024; Chown et al., 2024).

We define galaxy centers using the masks in Querejeta et al. (2021), which are based on near-IR stellar morphology, and nebular regions following Santoro et al. (2022); Groves et al. (2023), which are based on ionized gas emission observed by the VLT/MUSE as part of the PHANGS-MUSE survey (Emsellem et al., 2022). The galaxy center masks are available for all targets, though only 53 of our targets have well-defined central regions. The rest of the targets do not have any pixels classified as being in a “center”, and all of their pixels are included in the “disk” category. The nebular masks are available only for the $19$ Cycle 1 targets, which restricts a subset of our analysis to that subsample. See Pathak et al. (2024) for further details on a similar application of these environment masks.

To quantify the observed relationship between CO and PAH emission, we compile all measurements of individual pixels for all 70 galaxies into a single table, with columns for each NIR and MIR band (§3.1), CO(2-1) intensity (§3.2), and the values of each mask (§3.3). To avoid introducing systematic uncertainties, we analyze CO(2-1) intensities and do not adopt a CO-to-H₂ conversion factor. In Appendix C we describe how to use our results to estimate H₂ surface densities.

The CO (2-1) data have the coarsest angular resolution in this work. The CO resolution varies slightly from galaxy to galaxy. Over the range of distances of the galaxies in the sample, the working resolution for PHANGS–ALMA corresponds to median $98$ pc with a $16{-}84\%$ range of $63{-}129$ pc (Table 15 in Leroy et al., 2021). Using webbpsf⁵⁵5https://webbpsf.readthedocs.io/en/latest/-generated JWST PSFs, we convolve all JWST PAH images to share the same resolution of the corresponding CO (2-1) data following Aniano et al. (2011) and Williams et al. (2024). Then we define an astrometric grid with pixels $1/3\times$ the CO PSF FWHM in size so that approximately 9 pixels represent an independent measurement. We reproject the CO (2-1) intensity (in K km s^-1), JWST PAH intensity (in MJy sr^-1), and mask information onto this grid, noting whether the majority of the area in each pixel corresponds to a galaxy center, a nebular region, or neither. We correct all intensities to face-on values by scaling by $\cos i$, where $i$ is the galaxy inclination from Lang et al. (2020) and/or Leroy et al. (2021). The sample consists of nearly-face-on galaxies (median $\cos i=0.7$), and so this correction does not have a large impact on our analysis.

PAH emission is also expected to emerge from dust mixed with atomic gas (e.g., see Sandstrom et al., 2023a). To avoid lines of sight where H I is expected to make up most of the ISM, we consider only regions that have inclination-corrected $I_{\mathrm{F770W}}^{\rm PAH}\geq 0.5~\mathrm{MJy~sr^{-1}}$ (see Leroy et al., 2023a, for arguments about this specific value for “bright” emission). In Table 1 we report the percentages of flux or area that are classified as “bright” according to this criterion. Focusing on F770W-bright pixels means that we analyze the significant majority of flux in all bands across the sample. However, we do exclude a majority ($49\%$) of the area in the maps outside galaxy centers and nebulae. Analysis of this faint, extended diffuse emission requires the inclusion of H I and will be the topic of future work.

The $3.3~\mu$m PAH feature is $>10\times$ fainter than the 7.7 and 11.3 $\mu$m ones (e.g., Chastenet et al., 2023a; Dale et al., 2025), and the maps tend to be more limited by noise and systematic uncertainties related to continuum subtraction. As a result, we impose an additional threshold for analyzing the $3.3~\mu$m emission, considering only lines of sight with $I_{\rm F335M}^{\rm PAH}>0.1$ MJy sr^-1. This corresponds roughly to $I_{\rm F770W}^{\rm PAH}\gtrsim 2$ MJy sr^-1 and was selected to catch the lines of sight where the $3.3~\mu$m map yields robust detections at resolution matched to ALMA. Our analysis of F335M images is limited to lines of sight with surface brightness above this threshold.

Using the matched-resolution measurements, we analyze the correlation between CO (2-1) emission and emission in PAH-dominated JWST filters. For each MIR band $X$, we record the Spearman rank correlation coefficient ($r$), as well as the median and scatter in the ratio $I_{\rm CO(2-1)}/I_{\rm X}$. We then calculate a best-fit power law relating $I_{\rm CO(2-1)}$ to $I_{\rm X}$. Because the PHANGS–ALMA CO maps are much less sensitive than the JWST F770W and F1130W maps at matched angular resolution (see Leroy et al., 2023b), we treat the mid-IR imaging as the independent ($x$) axis for this calculation. We construct logarithmically-spaced bins in $I_{\nu}^{\mathrm{X}}$ and then compute the median and scatter, captured by the 16-84% range, of $I_{\mathrm{CO(2-1)}}$ within each bin.

We perform linear regression on these binned measurements⁶⁶6Because of the large number of individual pixels, the mean CO (2-1) intensity is detected at high signal-to-noise in all bins. using linmix⁷⁷7https://linmix.readthedocs.io/en/latest/index.html a hierarchical Bayesian method described in Kelly (2007). It performs a linear regression of $y$ on $x$ while incorporating measurement errors in both variables. We model the CO-vs-PAH emission relationship as

where the pivot $x_{0}\equiv\mathrm{median}(x)$. Re-centering the fit at $x_{0}$ ensures minimal covariance between the best-fit $m$ and $b$.

A number of studies have already demonstrated suppression of PAH emission, likely due to PAH destruction, within H II regions (e.g., Madden et al., 2006; Povich et al., 2007; Gordon et al., 2008; Montillaud et al., 2013; Chastenet et al., 2023a; Egorov et al., 2023; Pedrini et al., 2024; Sutter et al., 2024), while the intense radiation fields around H II regions also lead them to stand out as the brightest features in MIR maps of galaxy disks (Pathak et al., 2024). We note that at the physical resolution of PHANGS/JWST, PAH emission is reduced but still detected inside H II regions.

The top-right panel in Figure 1 separates sight lines in galaxy disks into those near H II regions and diffuse (i.e., all other) regions, and Table 2 presents our correlation analysis for these two regions separately. Perhaps surprisingly, we find median CO/F770W_PAH ratios in H II and diffuse regions to be quite similar, with H II regions exhibiting a ratio only $\approx 0.10$ dex ($1.25\times$) lower than diffuse regions (i.e., PAHs are slightly brighter relatively to CO near H II regions). The correlation strength near H II regions appears stronger than in diffuse regions ($r\approx 0.8$ vs. $r\approx 0.5$), but this partially reflects that the sight lines near H II regions tend to be brighter, and so less affected by the high statistical noise in the CO maps. The slightly steeper relationship observed for the diffuse regions reflects that diffuse sight lines with $I^{\mathrm{PAH}}_{\rm F770W}$ $\gtrsim 10$ MJy sr^-1 show slightly elevated CO/PAH ratios compared to nebular regions with similar intensities, which may be due to preferential destruction of CO compared to PAHs in nebular regions. This preferential destruction may be related to recent work demonstrating that PAH emission is slightly more long-lived than CO emission in gas-poor regions (Kim et al. in preparation). Sight lines with $I^{\mathrm{PAH}}_{\rm F770W}$ $\lesssim 10$ MJy sr^-1 in Figure 1 show almost identical CO-to-PAH ratios for diffuse and nebular regions.

The similarity of the CO-vs-PAH relationship in diffuse and nebular regions likely results from the fact that the H II region masks that we use (from Groves et al., 2023) are based on data with coarse resolution (median $70$ pc) compared to the actual size of most H II regions (e.g., see Barnes et al., 2022). As a result, these regions often include CO and PAH emission from ISM material projected towards, but not actually inside H II regions. The CO and PAH emission may come from the well-shielded material surrounding the H II regions. Sutter et al. (2024) discuss a scenario in which the Groves et al. (2023) regions contain a mixture of diffuse material and true H II regions to explain observed F770W/F2100W ratios. Resolved comparisons of high ($\lesssim 10$ pc) physical resolution H$\alpha$, Pa$\alpha$, CO, and MIR emission (e.g., Pedrini et al., 2024) should help test this hypothesis in the near future. Subtracting hot dust continuum emission from F770W, which we do not do, may also be important, since the filter still captures such emission in regions where PAHs have been destroyed to the point where their emission is not detectable.

Galaxy centers, especially bar-fed central molecular zones, differ from the disks of galaxies in ways that may also affect CO-to-PAH ratios. Galaxy centers exhibit some of the most intense star formation found in galaxies, with correspondingly high interstellar radiation fields, gas column and volume densities, and some host active galactic nuclei (e.g., Schinnerer et al., 2023, and see review in Schinnerer & Leroy 2024). As a result, CO in galaxy centers often exhibits broader line widths, lower opacity, and low $X_{\rm CO}$ (e.g., Bolatto et al., 2013; Teng et al., 2023).

The bottom-left panel of Figure 1 shows the CO-vs-PAH relationship separating sight lines towards galaxy centers compared to those in disks. Both CO and PAH emission appear systematically brighter in galaxy centers than in disks, likely due to the higher densities and stronger radiation fields in galaxy centers. This is expected from previous analyses of these galaxies (e.g., Sun et al., 2020; Pathak et al., 2024). Sight lines in galaxy centers also appear offset towards higher CO intensity compared to disks at matched $I^{\mathrm{PAH}}_{\rm F770W}$. On average the CO-to-PAH ratio is $\approx 0.16$ dex ($\approx 60\%$) higher in galaxy centers compared to disks, and the median $I_{\rm CO(2-1)}$ in centers appear higher than that found for disks at fixed $I^{\mathrm{PAH}}_{\rm F770W}$.

We note one caveat here. The stellar continuum can also be bright in galaxy centers, and the starlight subtraction can become correspondingly more difficult (e.g., Sutter et al., 2024; Baron et al., 2024b). If our standard correction oversubtracts the starlight from F770W in galaxy centers, then we might artificially underestimate F770W_PAH. However, considering a similar situation, Baron et al. (2024b) found that the range of plausible starlight SED variations cannot impact the F770W emission enough to explain observed variations in the F770W/F1130W ratio in the inner parts of galaxies. We also note that we find a similar contrast between disks and centers in the CO-to-F1130W ratio, where F1130W is less affected by starlight. Finally, we note that we do not subtract any hot dust continuum from F770W. Hot dust continuum is likely stronger in galaxy centers, implying that our reported CO-to-PAH ratio measurements are lower than the true ratios.

A higher CO-to-PAH ratio in bright galactic centers may imply a large impact of $X_{\rm CO}$ on the measured ratio. A number of studies have shown that $X_{\rm CO}$ in galaxy centers can be 5–15 times lower than the standard Milky Way $X_{\rm CO}=2\times 10^{20}$ cm^-2/(K km s^-1), and $\approx 2\times$ lower than the galaxy mean on average (e.g., Sandstrom et al., 2013; Israel, 2020; Teng et al., 2023; Chiang et al., 2024). Following Equation 2, centrally suppressed $X_{\mathrm{CO}}$ will lead to centrally enhanced CO-to-PAH ratios. On the other hand, the intense radiation fields expected in galaxy centers should increase PAH emission and decrease the CO-to-PAH ratio, unless PAH destruction occurs. There is some evidence for slightly lower $q_{\mathrm{PAH}}$ in galaxy centers (Chastenet et al., 2023b), but a general trend is not so clear. Metallicity and DGR tend to be highest in galaxy centers, which may also decrease the CO-to-PAH ratio (see Eq. 2). Considering that most of these effects would decrease CO-to-PAH, the enhancement seen in Figure 1 and Table 2 may indicate that $X_{\rm CO}$ represents the dominant effect, offsetting these other PAH-enhancing effects.

So far, we have treated all galaxies together, separating sight lines into categories but not otherwise differentiating between targets. However, our sample spans a range of stellar mass ($M_{\star}$), star formation rate (SFR), metallicity, morphology, and more. These factors will influence $U$, DGR, $q_{\mathrm{PAH}}$, and $X_{\rm CO}$ and so we might expect differences among the CO-PAH relationships for different galaxies (as in Chown et al., 2021). To explore the impact of these variations, the bottom-right panel of Figure 1 shows the CO-to-F770W relationship varies from galaxy to galaxy. The fits for individual galaxies are provided in Appendix B.

The bottom right panel of Figure 1 shows that individual galaxies exhibit strong CO-PAH correlations, mostly parallel to our best-fit overall relation. Broadly, the agreement among the 70 individual galaxies appears good, supporting the potential use of PAHs to trace CO. Specifically, the normalization of the CO-PAH relation scatters by $\pm 0.17$ dex from galaxy to galaxy. Simply applying our best fit overall relation to an individual galaxy with no additional information can be expected to yield a map biased by a factor drawn from this galaxy-to-galaxy scatter. This is similar to the pixel by pixel scatter observed within each galaxy, $\sigma_{\mathrm{pix}}$ in Table 4, implying that galaxy-to-galaxy variations and internal scatter contribute about equally to the total observed scatter.

In Figure 2, we test how these galaxy-to-galaxy offsets correlate with integrated galaxy properties. We compute the normalization of the best-fit CO(2-1) versus F770W_PAH power law for each galaxy, $C_{\mathrm{F770W}}^{\mathrm{PAH}}$, by performing a linear fit of $\log_{10}I_{\mathrm{CO(2-1)}}$ versus $\log_{10}I^{\mathrm{PAH}}_{\rm F770W}$ for each galaxy. $C_{\mathrm{F770W}}^{\mathrm{PAH}}$ is then given by the value of the best-fit relation at $I^{\mathrm{PAH}}_{\rm F770W}=1$ MJy sr^-1. We plot $C_{\mathrm{F770W}}^{\mathrm{PAH}}$ against $M_{\star}$ and specific star formation rate ($\mathrm{SFR}/M_{\star}$) (WISE+GALEX-based galaxy-integrated SFR and $M_{\star}$ are drawn from Leroy et al., 2021). Stellar mass correlates with the H₂/H I ratio, gas-phase metallicity, DGR, $X_{\rm CO}$ and more (e.g., Saintonge & Catinella, 2022). Meanwhile, $\mathrm{SFR}/M_{\star}$ anti-correlates with $q_{\mathrm{PAH}}$, and correlates with the mean interstellar radiation field, $\bar{U}$ (Chastenet et al., 2024). Therefore following Equation 2 both parameters might be expected to impact the CO-to-PAH ratio. Correlations between the galaxy-integrated CO-to-12$\mu$m ratio (with the $12\mu$m data from WISE) with star formation activity, stellar mass, and $\mathrm{SFR}/M_{\star}$ have previously been observed (Chown et al., 2021; Leroy et al., 2023a).

Consistent with previous work we find a mild positive correlation between the CO/PAH ratio at fixed $I^{\mathrm{PAH}}_{\rm F770W}$ and $M_{\star}$ and an anti-correlation between the CO/PAH normalization and $\mathrm{SFR}/M_{\star}$. As seen in the bottom right of Figure 1, the galaxy-to-galaxy offsets are thus not random but agree with physical expectations. The $\log_{10}$$C_{\mathrm{F770W}}^{\mathrm{PAH}}$ vs. $\log_{10}$$\mathrm{SFR}/M_{\star}$ trend may indicate the higher $U$ associated with high $\mathrm{SFR}/M_{\star}$ galaxies leads to stronger PAH emission, offsetting any suppression due to lower $q_{\rm PAH}$ at high $\mathrm{SFR}/M_{\star}$. As mentioned in §2, the anti-correlation between $C_{\mathrm{F770W}}^{\mathrm{PAH}}$ and $\mathrm{SFR}/M_{\star}$ is expected because the dust heating rate increases with $\mathrm{SFR}/M_{\star}$. The $\log_{10}$$C_{\mathrm{F770W}}^{\mathrm{PAH}}$ vs $\log_{10}M_{\star}$ trend goes in the sense that galaxies with higher metallicity, DGR, and molecular-to-atomic gas ratios (e.g., Saintonge et al., 2017) show higher CO-to-PAH ratios. As a result, perhaps the low observed CO-to-PAH ratios seen in low $M_{\star}$ systems reflects that the PAH emission is coming from regions dominated by H I or CO-dark H₂ despite our selection of only bright emission.

We also checked for correlations with distance and inclination to test how orientation and resolution might bias our results. We found that the normalization is uncorrelated with distance ($r=-0.04,p=0.78$), and weakly correlated with $\cos i$ ($r=0.24,p=0.06$). Variations in $\cos i$ from galaxy to galaxy simply slide the data points parallel to a 1:1 relation. Since the best-fit slopes of CO(2-1) vs. PAH emission for each galaxy and for the sample as a whole are within a few percent of 1.0, and furthermore that the sample covers a range of inclinations across the full $M_{\star}$ range, it is understandable that $\cos i$ does not have a significant effect.

We fit functional forms to the two trends and note the $\mathrm{SFR}/M_{\star}$ as the stronger, clearer trend. These predict the normalization of the CO vs. PAH relation as functions of galaxy-integrated $M_{\star}$ or $\mathrm{SFR}/M_{\star}$

We provide best-fit parameters for normalizations versus $\log_{10}M_{\star}$ and $\log_{10}\mathrm{SFR}/M_{\star}$ for the other PAH bands in Table 5.

In the bottom panels of Figure 2 we normalize the data from each galaxy by the values predicted by this fit, aiming to remove galaxy-to-galaxy scatter. Then we re-fit the relation to all data (lower right panel) and find

with scatter $\sigma=0.43$ dex, where $x\equiv\log_{10}I^{\mathrm{PAH}}_{\rm F770W}-\log_{10}C_{\mathrm{F770W}}^{\mathrm{PAH}}$ shown in Table 6 along with prescriptions for the other two PAH bands. We describe how to use these prescriptions in Appendix C.

So far we have focused only on the correlation between CO(2-1) and F770W_PAH. Do the other PAH-tracing filters show similar correlations? In Figure 3 we show CO versus F335M_PAH and versus F1130W for our $19$ Cycle 1 targets. The median CO/PAH ratios, fits and statistics for these bands are shown in Table 2. The correlation remains strong in both of these bands, and the best fit binned CO-PAH slopes are also very close to linear for the $3.3~\mu$m and $11.3~\mu$m features. Of the three PAH bands, F335M_PAH is the faintest (Chastenet et al., 2023b; Sandstrom et al., 2023b), and therefore shows the largest CO-to-PAH ratios ($10^{1.24}\approx 17.4$ times higher at F335M_PAH than F770W_PAH; while F1130W is a factor of $10^{0.15}\approx 1.4$ times lower than F770W_PAH); see top rows of Table 2.

These three PAH bands (3.3, 7.7, and 11.3 $\mu$m) are dominated by different species of the PAH population – smaller, neutral PAHs for 3.3 $\mu$m, ionized PAHs for a range of sizes for 7.7 $\mu$m, and larger, mainly neutral PAHs for 11.3 $\mu$m. Their intensity ratios also respond to changes in the interstellar radiation field spectrum (e.g., Maragkoudakis et al., 2020; Draine et al., 2021). All three bands yield reasonable first-order estimators of CO intensity, but we would not expect all three to show identical or even equally good correlations with CO. We anticipate that future work will explore optimal combinations of bands to trace CO (and H₂) and examine the impact of conditions in the molecular gas on PAH band ratios.

From a practical perspective, each band has advantages. The F770W band traces the brightest PAH feature and is available now for $>70$ nearby galaxies. The F335M filter offers even sharper resolution but harbors the fainter 3.4 $\mu$m PAH feature, the Pfund-$\delta$ line (e.g. Peeters et al., 2024), and is more strongly affected by contamination by starlight. Meanwhile, the F1130W filter is largely unaffected by starlight and also captures a bright feature in a relatively narrow filter.