MID-INFRARED PROPERTIES OF OH MEGAMASER HOST GALAXIES. I. SPITZER IRS LOW- AND HIGH-RESOLUTION SPECTROSCOPY

The data were processed using the Spitzer Science Center S17.0 data pipeline. We used basic calibrated data (BCD) products for our analysis, having already been corrected for flat fielding, stray light contributions, nonlinear responsivity in the pixels, and "drooping" (an increase in detector pixel voltage that occurs during non-destructive readouts). The two-dimensional BCD images were first medianed over the data cycles at each nod position to remove transient effects such as cosmic rays. For the SL and LL modules, we subtracted the sky contribution by differencing the BCD images for each nod position with the adjacent position in the same module.

The slit sizes of the SH and LH modules are too small to permit extraction of a sky background during the same observation. The continuum levels in the HR modules, however, contain strong contributions from scattered zodiacal light. Estimations of the flux at 15 μm using SPOT predict zodi contributions within the Spitzer beam ranging from 20 to 80 mJy, which in many cases is of comparable magnitude to the expected signal from the galaxies themselves. The wavelength-dependent brightness of the sky contribution means that it cannot be corrected using a simple scaling, and so we did not attempt to further calibrate either HR module. The calibrated LR spectra were thus used for absolute fluxes and the HR spectra for line ratio diagnostics (which are unaffected by continuum levels).

To obtain more accurate measurements of faint lines at longer wavelengths, we took dedicated off-source sky observations in the LH module for all 24 OHMs in our program. Subtraction of the wavelength-dependent background, however, significantly affected the measured line fluxes. For galaxies with background subtraction in only the LH module, this changes the line ratios measured between different HR modules (e.g., [S iii]λ33/[S iv]λ10.5). We reduced the data both with and without subtraction of the sky backgrounds; the background-subtracted LH spectra are attached as an Appendix.

Following the initial cleaning of the two-dimensional BCD products, we eliminated rogue pixels using the IDL package IRSCLEAN_MASK⁷. We used rogue pixel masks provided by the SSC for each IRS campaign, and supplemented the standard masks with manual cleaning of each nod and module. The one-dimensional spectra were then extracted using the Spitzer IRS Custom Extractor version 2.0. For all modules we used the optimal extraction routine with the standard aperture to improve the S/N ratio in faint galaxies.

The LR modules were stitched together to match continuum levels by using a multiplicative scaling. We fixed the LL1 module and then scaled LL2 to LL1, SL1 to LL2, and SL2 to SL1. The mean scaling factors were 1.03 ± 0.08 for LL2 to LL1, 1.49 ± 0.69 for SL1 to LL2, and 1.37 ± 0.78 for SL2 to SL1. We then calibrated the entire LR spectra as a single unit by scaling to the IRS 22 μm SUR peakups. The required scaling in the majority of cases was quite small, indicating that the sky subtraction and spectral extraction techniques are robust; the mean scaling factor was 0.94 ± 0.08. The accuracy of the overall continuum flux calibration is ∼5% (Houck et al. 2004).

Noisy areas on both ends of the SH and LH orders were trimmed from the one-dimensional spectra. These areas typically encompass a range of 10–30 pixels on the edges of the orders and correspond to areas of decreased sensitivity on the detector. We deliberately trimmed only pixels with an overlapping wavelength range in adjacent orders so that a maximum amount of information is preserved. In isolated cases, we also removed obvious rogue pixels by hand from spectra in which exceptional one-channel features appear in only a single nod.

We calculated a simple figure of merit to measure the S/N in the LR data. The data near λ_rest = 21 μm (a feature-free area near the center of most spectra) are fit with a low-order polynomial; the median flux in that region is then divided by the rms noise to yield the S/N (Table 3). We note that this parameter is a function of wavelength, as well as the performance and integration time in each spectral module; this is intended to give only a rough estimate for each object. The S/N for the samples ranges from ∼10to110, with a median of 35.

Since the archival OHM and non-masing galaxies did not come from a unified observing program, the version of the Spitzer data pipeline and the level of processing varied slightly from object to object—we used the most recent versions available in the archive (version 15.3.0 or later). The reduction process was identical to that for the OHM galaxies in our program, with the exception of the LR photometric scaling; since observations in the archive varied in availability of peakup data, we used a variety of sources to calibrate the spectra. In order of priority, we used the IRS dedicated 22 μm SUR peakups, IRS acquisition 16 and 22 μm DCS peakups, or IRAS 25 μm observations (all lying within the coverage of the LL modules). For galaxies from the archive, 17 are calibrated with DCS peakups, 21 with IRAS 25 μm photometry, and four are left uncalibrated. The mean scaling factor for the objects without SUR photometry was 1.05 ± 0.25, slightly higher than the mean scaling for galaxies in the dedicated OHM program.

Five OHMs and five non-masing galaxies from the archive had both SH and LH sky backgrounds taken simultaneously with the spectroscopic observations; the remainder had no HR sky backgrounds in either module. Since we emphasize uniformity of the observations to the fullest extent possible, all data used for statistical comparisons between the samples use data without HR sky subtraction; line measurements for the background-subtracted objects are given in the Appendix.

The spectra for IRAS 20460+1925 and IRAS 23028+0725 had no flux in the SL modules, most likely due to a pointing error during observations. No SL data from either galaxy are used in our analyses here or in Paper II.

The continuum emission for all objects in the OHM sample has a relatively homogenous spectral shape over the range of the IRS, although differences in spectral shape between the two samples do appear and are explored in Paper II. Figure 5 shows the individual objects overlaid with a template generated by medianing the flux in each wavelength bin from all galaxies. The template bears a close resemblance to starbursting ULIRG spectra seen in previous surveys (Hao et al. 2007; Weedman & Houck 2009). The LR spectrum clearly shows silicate absorption at 9.7 and 18 μm and water ice absorption at 6 μm. LR emission features are dominated by the broad polycyclic aromatic hydrocarbon (PAH) features from 6 to 13 μm, with weaker contributions from neon, sulfur, and molecular hydrogen also visible.

Zoom In Zoom Out Reset image size

The continuum data from λ_rest = 20to30 μm are in most cases well characterized by a power-law fit, with the short wavelength break occurring near the 18 μm silicate feature and the long wavelength end cut off by the spectral range of the IRS. Shortward of 15 μm, the continuum becomes increasingly contaminated by individual absorption and emission features, especially from PAH emission and the deep silicate absorption at 9.7 μm. Following Brandl et al. (2006), we fit a spectral index to the continuum in two components, with α_30–20 measuring the relatively feature-free flux from 20to30 μm and α_15–6 measuring the contribution from 5.3to14.8 μm; the wavelengths are slightly shifted to avoid contamination from water ice at 6 μm and [Ne iii] at 15.6 μm (Table 3).

The mean 15–6 spectral index for the entire OHM sample is α_15–6 = 2.1 ± 0.6; the mean 30–20 slope is α_30–20 = 4.8 ± 1.1. The shallowest 15–6 slope occurs for IRAS 10039–3338 (α_15–6 = −0.1), an object with weak PAH features and very strong silicate absorption; the steepest index occurs for IRAS 12018+1941 (α_15–6 = 3.0), which has moderate PAH and line emission features and a nearly constant spectral index over the entire mid-IR range. Steeper 15–6 indices are likely due to a combination of smaller relative quantities of warm dust (thermal blackbodies of ∼300 K peaking near 10 μm) and larger quantities of cooler dust.

The shallowest 30–20 slope occurs for IRAS 13451+1232 (α_30–20 = 1.9); the LR spectrum for this object more closely resembles that seen in Seyfert galaxies and PG quasars (Schweitzer et al. 2006; Hao et al. 2007), with weak PAH emission and shallower silicate absorption. The continuum emission for this object is also much closer to being uniform over the entire range of the IRS; the difference between the two spectral indices is only Δα = −0.4, compared to an average of Δα = 2.7 for the entire sample. This behavior is more typical of non-thermal emission that can extend over many decades with the same index. The steepest 30–20 emission measured is from IRAS 11028+3130 (α_30–20 = 7.0); the galaxy shows moderate PAH and line emission features, but a flat continuum between 12 and 20 μm.

The non-masing galaxies have average spectral indices of α_15–6 = 1.8 ± 0.6 and α_30–20 = 2.5 ± 0.9. A full statistical analysis of the values for the two samples (particularly α_30–20) is presented in Paper II.

We measured emission from atomic and molecular lines using the standard packages in the Spectroscopic Modeling Analysis and Reduction Tool (SMART) version 6.2.4 (Higdon et al. 2004). A simple Gaussian is a good fit for virtually all HR lines in the sample; in cases where lines are blended (such as the [Ne v]/[Cl ii] and [O iv]/[Fe ii] complexes), we used a multi-Gaussian fit centered at the redshifted rest wavelengths of the expected transitions. To compute upper limits for non-detections, we use the 3σ noise measured from the surrounding continuum and a Gaussian shape with an FWHM estimated from detected lines. Accuracy for all measured line fluxes is on the order of ∼10%.

The most common lines detected are the forbidden [Ne ii] λ12.814 and [Ne iii] λ15.555 transitions (Table 4). [Ne ii] is observed in nearly the entire sample, with detections in 50/51 OHMs and 14/15 non-masing galaxies. The only exceptions are the OHM IRAS 11028+3130 and the non-maser IRAS 20460+1925. [Ne iii] is also common, detected in 43/51 OHMs and 14/15 non-masing galaxies. Other common lines are the [S iii] λ18.713 (detected in ∼80% of galaxies) and [S iv] λ10.511 (∼50%). [Ar iii] λ8.991 is detected in 15 OHMs and 5 non-masing galaxies, but the redshifted line is not visible in the SH module for archived objects at z < 0.1.

The IRS is designed to make accurate measurements of narrow atomic transitions in the SH and LH modules; however, several lines are also detected in the LR modules, most often the powerful neon, sulfur, and H₂ transitions. The only emission lines visible in the LR modules without a corresponding detection in HR are the H₂ S(7) line at 5.5 μm and the H₂ S(5)/[Ar ii] complex near 6.7 μm; this is due to the SH low-wavelength cutoff at λ_rest = 8.2–9.0 μm, depending on the redshift of the galaxy. All other atomic emission features observed in the LR modules have corresponding detections in HR; furthermore, blending of narrow lines makes accurate measurements of flux difficult in the LR modules. For isolated lines with well-defined surrounding continuum, our fluxes are consistent for measurements in both low and high resolution.

We note the presence of two features which have no obvious identifications occurring in the LH spectra for multiple objects: one is an emission feature seen near 29 μm (a prominent example occurs for IRAS 17539+2935) and the second is an absorption feature near 30 μm. The two are often paired and are seen in ∼50% of the galaxies observed. The rest wavelengths of the transitions, however, vary significantly from object to object (with a standard deviation of σ_λ ≃ 0.7 μm), while the observed wavelengths are nearly fixed (σ_λ ≲ 0.05 μm). This implies that the features are either artifacts of the extraction process or that both the emission and absorption come from unidentified foreground features with little to no Doppler shift. Given that we see no evidence for these unidentified lines in any of the LR spectra (which should be detectable, given the high S/N for many of the features), we consider both to be spurious.

We detected multiple emission lines from the pure rotational series of molecular hydrogen in both OHMs and non-masing galaxies. At redshifts of z ≲ 0.1, transitions from H₂ S(0) at 28.22 μm to H₂ S(3) at 9.67 μm are visible in the HR modules; in addition, the LR module is capable of detecting lines as far out as the S(7) transition at 5.51 μm. In each case, the line number (e.g., 0 for H₂ S(0)) indicates the rotational quantum number of the lower state (J = 2 → 0) for the quadrupole S-branch transition. ΔJ = 2 results in two separate branches: ortho (parallel nuclear spin, odd J) and para (anti-parallel nuclear spin, even J).

We detected at least one H₂ line in 49/51 OHMs and 13/15 non-masing galaxies, with S(1) seen in all objects for which at least one molecular hydrogen transition is reported. The higher-order S(2) and S(3) lines are seen in roughly 2/3 of the sample, while the para ground-state S(0) transition is detected in only ∼15% of the sample. Our line detection rate is consistent with results from the SINGS galaxies examined in Roussel et al. (2007) and the ULIRG sample of Higdon et al. (2006). Lower detection rates of S(0) and S(2) are likely due to a combination of the intrinsic ortho–para ratio as well as rising continuum levels near 28 μm that can obscure weak line emission by S(0). Higdon et al. (2006) find that the S(2)/S(3) ratios are consistent with no significant differential extinction for the two lines, which is supported by numerous detections in our sample of S(3) line emission superimposed on optically deep silicate absorption near 9.7 μm. We thus applied no extinction or reddening corrections to the line fluxes (Table 5).

We did not detect the S(6) line in any galaxy, while S(4) showed a single detection in IRAS 16300+1558. Since these lines are excited by hotter gas (T ∼ 1000 K), they are typically weaker than the lower states which probe the larger reservoir of cool gas. In addition, both lines are only visible in the LR modules at z ∼ 0.1. This means that deblending is a significant issue, since both lines lie near broad PAH emission complexes. Eleven OHMs and one non-masing galaxy show the unresolved S(7) ortho line at 5.51 μm in the SL module.

Measurement of the S(5) line presents a particular problem due to its location in a crowded section of the spectra. Its rest wavelength of 6.91 μm lies near the [Ar ii] feature at 6.99 μm; in addition, both features are bracketed by possible hydrocarbon absorption at 6.85 and 7.25 μm. This not only creates difficulties in establishing a reliable continuum, but also in deblending the [Ar ii] and the S(5) emission (see Section 5.5.2). Emission in the [Ar ii]/H₂ S(5) complex is seen in more than half of our sample, however, and so we present measurements for the entire feature, including blended emission from both lines. We caution that these fluxes should be viewed as upper limits for either [Ar ii] or H₂ S(5) emission, since the SL module does not have sufficient resolution to separate the two features.

For galaxies in which multiple H₂ lines are observed, we fit excitation temperatures (T_ex) to the molecular gas following the methods of Rigopoulou et al. (2002) and Higdon et al. (2006). We assume that the emission is optically thin (so that the lines are unsaturated), populations are in local thermodynamic equilibrium (LTE), and that the sources are unresolved in the Spitzer beam. The luminosity of a molecular emission line for the transition from (J +  2) → J is then L_J = A_J × ΔE_J × N_J+2, where A_J is the Einstein-A coefficient, ΔE_J is the energy of the transition, and N_J+2 is the number of molecules in the J + 2 state. The partition function for a given symmetry branch is

$$\begin{equation} Z_{J_{o/p}} = \sum _{J_{o/p}} g_{J} \rm exp[-E_{J} / k T_{\rm ex}], \end{equation} \tag{ 2 }$$

where T_ex is the excitation temperature and we sum only over a single symmetry branch (ortho or para). The statistical weights are g_J = (2J + 1) × J_s, where J_s = 1 for the para branch (even J) and J_s = 3 for ortho (odd J).

Assuming that the lines are in LTE, the ratio of level populations follows a Boltzmann distribution such that $N_J\propto g_J \exp[-E_J/k T_{\rm ex}]$. The inverse slope of the best-fit line of an excitation diagram yields T_ex—Figure 6 shows examples of temperature fits for our data. The total warm H₂ mass can be then calculated from T_ex and the flux (F_J) from any transition as

$$\begin{equation} M_{{\rm tot}} = m_{\rm H_2} \times \phi _{o/p} \times \frac{\big(4 \pi D_L^2\big) F_J Z_{J_{o/p}}}{A_J \Delta E_J g_J \exp[-E_J/kT_{\rm ex}]}, \end{equation} \tag{ 3 }$$

where ϕ_o/p is a numerical factor accounting for the ortho-to-para ratio (assumed to be 3:1), $m_{\rm H_2}$ is the mass of the hydrogen molecule, and D_L is the luminosity distance.

Zoom In Zoom Out Reset image size

For cases where the S(7) line was detected, a single excitation temperature gives a poor fit to the full set of transitions. In these cases, we first fit T_ex between S(3) and S(7), measuring hotter gas. We then subtracted this component from the S(0) to S(3) fluxes, and fit a second T_ex to the warm gas component. This decreased the mean warm T_ex by ∼20 K, with a negligible effect on the gas mass. We calculated the warm H₂ mass using the flux in the S(1) transition and the hot gas mass using the S(3) flux (Table 5).

Both Higdon et al. (2006) and Roussel et al. (2007) suggest that the H₂ emission arises from far-ultraviolet photons from massive stars powering photodissociation regions (PDRs). Detections of the H₂ S(3) transition in nearly all objects implies that the silicate absorption at 9.7 μm must be partially background to the warm molecular gas seen in emission. Since the dust is very optically thick in almost all ULIRGs, this means that at least some molecular gas (and possibly other atomic transitions) actually come from superficial layers at the edge of the merging system. Given that the OHM is typically formed within the central kiloparsec of the host galaxy, a link between the observed warm H₂ gas and the OHM is uncertain.

Eleven objects in the OHM sample and two non-masing galaxies have CO detections published in the literature (Solomon et al. 1997; Gao & Solomon 2004a). The beam width used for CO observations is several times that of the HR slits; since most ULIRGs are unresolved in the Spitzer beam, we consider the gas mass estimates to be comparable. The cold gas masses derived using a ULIRG-calibrated $M_{\rm H_2}/L_{\rm CO}$ ratio of ∼1.4M_☉/(K km s^–1 pc²) give a warm gas mass fraction for the OHMs ranging from 0.04%to0.8%, with the gas fraction of the non-masing galaxies lying in a similar range (0.06%–0.1%). This is comparable to warm gas fractions in ULIRGs from Higdon et al. (2006), implying that the mid-IR H₂ lines probe only a small amount of the total gas mass in these galaxies. The bulk of the remaining portion is likely cold gas without sufficient energy to excite rotational transitions in the mid-IR.

In addition to the atomic and simple molecular emission lines, we also observed multiple features attributed to PAHs; the broad-line emission comes from vibrational modes of C–C and C–H bonds (Draine 2003). PAH features are ubiquitous in the mid-IR emission of starburst galaxies and ULIRGs (Lutz et al. 1998; Genzel et al. 1998; Sturm et al. 2000; Peeters et al. 2004; Desai et al. 2007; Imanishi et al. 2007), and dominate the LR spectra of most galaxies in our sample. Multiple PAH features are seen for all OHMs, encompassing galaxies with very wide ranges in continuum shape and line emission. We detect strong PAH transitions centered at 6.2, 7.7, 8.6, 11.3, and 12.7 μm, several of which are also visible in the HR spectra. Weaker emission features at 13.5, 14.2, 16.4, 17.1, and 17.4 μm are also visible in many galaxies.

We measured the PAH emission via two methods: the first defines a local continuum around the PAH feature using a spline fit and then integrates the total flux after baseline subtraction. The default continuum pivots are located at 5.15, 5.55, 5.95, 6.55, and 7.10 μm for the 6.2 μm feature and at 10.1, 10.9, 11.8, and 12.4 μm for the 11.3 μm feature. These are shifted slightly for each object to avoid both broad absorption features (including water ice and hydrocarbons) and narrow atomic emission lines. We quantify the emission from the two cleanest PAH features appearing in our spectra: the 6.2 and 11.3 μm complexes (Table 6).

A second method uses PAHFIT (Smith et al. 2007), a public IDL package, to fit global mid-IR spectral templates and simultaneously measure the relative effects of overlapping features. The routine decomposes the LR IRS spectra into emission from stellar continuum, dust features (including PAHs), atomic and molecular lines, and blackbodies from thermally heated dust at a variety of temperatures; this is simultaneously fit with an extinction curve including silicate features at 9.7 and 18 μm. Due to the large number of parameters being fit, however, even strong features may overlap sufficiently to affect the accuracy of the fit (Spoon et al. 2002); in addition, PAHFIT does not fit for several features commonly found in ULIRGs, such as absorption features from ice, hydrocarbons, and gas-phase molecules.

PAHFIT fits the dust emission features (including PAHs) with a Drude profile, which has the form

$$\begin{equation} I_\lambda [\lambda ] = \frac{b \gamma ^2}{(\lambda /\lambda _c - \lambda _c/\lambda)^2 + \gamma ^2}, \end{equation} \tag{ 4 }$$

where λ_c is the central wavelength, γ is the fractional FWHM, and b is the (peak) central intensity. The Drude profile is typically broader than a Gaussian, with significant amounts of power in the extended wings. Several dust emission features (e.g., the 7.7 and 12.7 μm PAHs) require more than one component for a reasonable fit. PAHFIT returned positive detections for both the 6.2 and 11.3 μm PAH features for nearly all galaxies; the fit for OHM IRAS 07572+0533 showed no emission at 6.2 μm, while the fits to OHM IRAS 13218+0552 and the non-masing galaxy IRAS 13349+2438 show no emission in either dust feature.

We compared the flux measured in the 6.2 and 11.3 μm PAH features from our baseline-subtracted spline fits to the PAHFIT values; results from PAHFIT are consistently higher than those from the spline fit, indicating significant mixing between the PAH emission and what was previously designated as "continuum" (Figure 7). Fluxes of the 6.2 μm complex measured with PAHFIT are a factor of ∼3–4 greater than the spline-fit fluxes, while the 11.3 μm feature is an average of ∼2–3 times larger. The relative strengths of the two features are consistent using both methods; the mean value of the (6.2 μm PAH/11.3 μm PAH) ratio is identical to within 15% for OHMs.

Zoom In Zoom Out Reset image size

Galliano et al. (2008) also use both spline and multi-component profile fitting approaches to measure the PAH variations within galaxies; they show that both methods yield the same overall trends, although each have their own underlying biases depending on the property being measured. Given the large contributions of overlapping dust emission features to the 5–10 μm spectrum, which are not possible to separate from the underlying blackbody emission using spline fits (Marshall et al. 2007), we consider the PAHFIT results to be the more robust method. Measurements of PAH features in the literature, however, typically use a spline-fit method (e.g., Brandl et al. 2006; Desai et al. 2007; Spoon et al. 2007; Zakamska et al. 2008). In particular, comparisons of PAH data from the OHM galaxies to other samples (e.g., the "fork diagram" from Spoon et al. 2007) must use the same method to return physically meaningful results. While PAHFIT fluxes may thus better represent the absolute PAH luminosity, the spline-fit data are used when comparing the OHMs to objects from the literature (Table 6).

Water ice absorption at 6 μm can have significant effects on the measurement of the 6.2 μm PAH EW; following the method of Spoon et al. (2007), we correct for this by substituting the continuum inferred while measuring the 9.7 μm silicate strength for the measured 6.2 μm continuum (see Section 5.5). In total, 24 OHMs and three non-masing galaxies with SL data showed absorption strong enough to affect the measured EW; the spline fit with the new continuum decreased the EW for all objects except IRAS 11180+1623, for which the continuum levels as measured by the PAH fit and the silicate depth are nearly identical (within error). Since PAHFIT does not fit for ice absorption, we also calculate ice-corrected EW for objects showing absorption at 6 μm by using the flux from PAHFIT and the inferred continuum from silicate measurements.

The 11.3 μm PAH is seated atop the edge of the deep silicate absorption at 9.7 μm; determining an extinction-corrected continuum level for EW measurements is thus also difficult. Spectral mapping of AGN galaxies with ISOCAM has shown that PAH emission can be spatially extended and suppressed near the nucleus (Le Floc'h et'al. 2001; Díaz-Santos et'al. 2010); this means that the PAH emission in AGNs may be largely unaffected by dust absorption. Starburst galaxies, however, can show strong PAH emission in both the 6.2 and 11.3 μm bands in the nuclear regions where the silicate optical depth is at its highest (Galliano et'al. 2008), and is likely to affect continuum levels for PAH features. The measured 11.3 μm PAH data using the spline-fit method are therefore likely to underestimate the luminosities.

5.5.1. Silicates

The LR spectra show near-ubiquitous absorption from amorphous silicate dust, with a strong feature caused by a Si–O stretching mode near 9.7 μm and a weaker feature caused by an Si–O–Si bending mode near 18 μm (Knacke & Thomson 1973). The presence of dust is unsurprising, as the characteristic extreme IR luminosities of ULIRGs are caused by large amounts of heated dust being thermally re-radiated. Hao et'al. (2007) found that ULIRGs nearly uniformly show absorption in the two silicate features, in contrast to QSOs and some Seyfert galaxies which typically show the feature in emission (Siebenmorgen et'al. 2005; Hao et'al. 2005; Sturm et'al. 2005; Schweitzer et'al. 2008).

We measure the strength of the silicate absorption at both 9.7 and 18 μm using the method of Spoon et'al. (2007):

$$\begin{equation} S_{\lambda } = \ln \left(\frac{f_{\lambda }}{f_{{\rm cont}}}\right), \end{equation} \tag{ 5 }$$

where f_λ is the measured flux and f_cont is the interpolated continuum at the feature extremum. More negative values of S_sil represent deeper absorption. The expected continuum is calculated using a combination of spline and power-law fits, depending on the strength of the PAH and water ice features in the spectrum (Spoon et'al. 2007).

All OHMs and ∼95% of the non-masing galaxies showed absorption at both 9.7 and 18 μm (Table'7); the average depth for the OHMs is S_9.7 = −1.8 ± 0.8, while the average depth of the non-masing galaxies is S_9.7 = −0.6 ± 0.4. The deepest absorption is in the OHM IRAS 04454–0838 (S_9.7 = −3.7), while only one object (the non-masing galaxy IRAS 13349+2438) shows emission in both amorphous silicate features.

Table 7. Solid-phase Absorption Features

Object	6.0 μm H2O Ice	6.85 μm HAC		7.25 μm HAC		9.7 μm	18 μm	16 μm	23 μm
	τ	τ	Flux	τ	Flux	Ssil	Ssil	Sresidsil	Sresidsil
IRAS 01355–1814						–2.4	–0.9	–0.3	–0.2
IRAS 01418+1651		0.49	–8.3			–1.3	–0.4
IRAS 01562+2528						–0.7	–0.3
IRAS 02524+2046		0.20	–0.4			–0.9	–0.3
IRAS 03521+0028	0.42	0.13	–0.4			–1.4	–0.2	–0.1	–0.1
IRAS 04121+0223	1.61					–1.0	–0.2
IRAS 04454–4838	0.42	0.35	–9.0	0.10	–1.5	–3.7	–1.0	–0.3	–0.2
IRAS 06487+2208		0.23	–4.6			–1.2	–0.3
IRAS 07163+0817						–1.2	–0.1
IRAS 07572+0533						–0.6	–0.3
IRAS 08201+2801	1.06	0.37	–4.0	0.22	–1.5	–2.2	–0.6	–0.2	–0.1
IRAS 08449+2332						–1.2	–0.5	–0.1	–0.1
IRAS 08474+1813		0.36	–0.2			–1.9	–1.2
IRAS 09039+0503	0.98	0.15	–0.5			–2.0	–0.6
IRAS 09539+0857		0.24	–2.5			–3.1	–1.2	–0.4	–0.2
IRAS 10035+2740						–1.5	–0.8
IRAS 10039–3338	0.23	0.23	–166.3			–3.1	–1.0	–0.4	–0.3
IRAS 10173+0828						–1.9	–0.8	–0.3	–0.2
IRAS 10339+1548						–1.1	–0.05
IRAS 10378+1109	0.72	0.18	–0.6			–2.0	–0.3
IRAS 10485–1447		0.25	–0.4			–2.9	–0.9
IRAS 11028+3130						–2.6	–1.0
IRAS 11180+1623	0.54					–1.7	–0.5
IRAS 11524+1058		0.27	–1.5	0.2:	–0.3:	–1.5	–0.8
IRAS 12018+1941		0.16	–2.6			–1.4	–0.4
IRAS 12032+1707	0.71	0.55	–6.6	0.30	–3.9	–2.7	–0.8
IRAS 12112+0305	0.59	0.41	–3.5			–1.8	–0.3	–0.2	–0.1
IRAS 12540+5708						–0.7	–0.2
IRAS 13218+0552						–0.5	–0.4
IRAS 13428+5608	0.50	0.40	–28.5			–2.0	–0.5
IRAS 13451+1232						–0.5	–0.1
IRAS 14059+2000						–0.8	–0.1
IRAS 14070+0525	0.90	0.24	–1.8	0.15	–1.8	–2.7	–0.9
IRAS 14553+1245						–1.3	–0.5
IRAS 15327+2340	0.68	0.35	–50.1			–3.1	–0.4	–0.2	–0.1
IRAS 16090–0139	0.56	0.45	–10.5	0.24	–5.6	–2.4	–0.6
IRAS 16255+2801	0.54					–2.2	–0.6
IRAS 16300+1558	0.61	0.38	–2.2	0.21	–1.1	–2.7	–0.7	–0.3	–0.1
IRAS 17207–0014	0.31	0.23	–17.4			–1.9	–0.6	–0.2	–0.1
IRAS 18368+3549						–1.8	–0.2	–0.3	–0.2
IRAS 18588+3517	0.72					–2.2	–0.6	–0.3	–
IRAS 20100–4156	1.45	0.23	–2.9	0.23	–4.5	–2.4	–0.7	–0.2	–0.1
IRAS 20286+1846	1.08					–1.6	–0.6
IRAS 21077+3358						–1.9	–0.7
IRAS 21272+2514	1.66	0.21	–0.7			–2.8	–0.7
IRAS 22055+3024						–1.3	–0.3
IRAS 22116+0437		0.04:	–0.1:			–2.6	–0.9	–0.2	–0.2
IRAS 22491–1808	0.43	0.19	–0.2			–1.5	–0.5	–0.2	–
IRAS 23028+0725	...	...	...	...	...	...	–0.6
IRAS 23233+0946	0.44					–1.9	–0.4
IRAS 23365+3604	0.66					–2.0	–0.5
IRAS 00163–1039						–0.5	–0.1
IRAS 01572+0009						–0.2	–0.2
IRAS 05083+7936						–1.1	–0.3
IRAS 06538+4628						–0.5	–0.2
IRAS 08559+1053	0.18					–0.6	–0.2
IRAS 09437+0317						–1.1	–0.3
IRAS 10565+2448						–1.2	–0.3
IRAS 11119+3257	0.19					–0.7	–0.3	–0.2	–
IRAS 13349+2438						0.1	0.07
IRAS 15001+1433	0.30					–0.9	–0.4
IRAS 15206+3342						–0.4	–0.2
IRAS 20460+1925	...	...	...	...	...	...	–0.4
IRAS 23007+0836						–0.3	–0.1
IRAS 23394–0353						–0.7	–0.3
IRAS 23498+2423						–0.6	–0.3

Notes. Silicate strength is defined in Equation'(5); the 9.7 and 18 μm features are depths for amorphous silicates, while the 16 and 23 μm crystalline features are residual depths measured after the 18 μm feature was subtracted. Fluxes are given in 10^–21 W cm^–2; objects marked with a colon ":" represent uncertain detections.

Download table as:  ASCIITypeset images: 1 2

In addition to the amorphous silicate, we detect weaker features from crystalline silicate absorption in 19 OHMs, including bands at 11, 16, 19, 23, and 28 μm (see Figure'8 for an example). We use the method of Spoon et'al. (2006) to subtract off both the dust continuum and amorphous component to measure the residual optical depth at 16 and 23 μm, typically the strongest crystalline features (Table'7). The deepest S₁₆ occurs for IRAS 10039–3338, at −0.4; however, the S/N ratio means we are only sensitive to an absorption limit of S₁₆ ≃ −0.1 and S₁₈ ≃ −0.05. Only one detection of crystalline silicates is made in a non-masing galaxy, in IRAS 11119+3257. Since all detections of crystalline silicates in OHMs have S_9.7 < −1.4, the lower total dust column in non-masing galaxies is a likely contributor to the detection rate.

Zoom In Zoom Out Reset image size

Table 8. Gas-phase Absorption Features

Object	13.7 μm C2H2		14.02 μm HCN		15.0 μm CO2
	fnorm	Flux	fnorm	Flux	fnorm	Flux
IRAS 10039–3338	0.93	–0.9	0.86	–4.0
IRAS 12018+1941	0.95	–0.7	0.94	–0.8
IRAS 12540+5708	0.95	–9.5
IRAS 13218+0552	0.82	–1.1
IRAS 13428+5608	0.93	–0.8
IRAS 14070+0525	0.79	–0.4
IRAS 15327+2340	0.84	–5.8	0.90	–5.6	0.94	–2.0
IRAS 16090–0139	0.86	–1.1
IRAS 17207–0014	0.93	–0.7
IRAS 20100–4156	0.82	–1.4	0.84	–1.1
IRAS 22491–1808	0.92	–0.7

Notes. f_norm gives the peak depth of absorption features plotted in normalized flux units. Fluxes are measured in 10^–21 W cm^–2.

Download table as:  ASCIITypeset image

5.5.2. Aliphatic Hydrocarbons

5.5.3. Ices

5.5.4. C₂H₂, HCN, and CO₂

Previous mid-IR surveys have also identified bands of molecular gas absorption in ULIRGs (Spoon et'al. 2006; Armus et'al. 2007; Lahuis et'al. 2007), including the vibration–rotation bands of acetylene (C₂H₂; 13.7 μm), hydrogen cyanide (HCN; 14.02 μm), and carbon dioxide (CO₂; 15.0 μm). Lahuis et'al. (2007) reported the detection of both C₂H₂ and HCN in 15 (U)LIRG nuclei, with detections of CO₂ in four objects. Eight of the objects in the Lahuis sample are OHMs in our sample; we confirm detections of C₂H₂ in all eight galaxies, in addition to the OHMs IRAS 10039–3338 and IRAS 12018+1941 (Figure'8). HCN is detected in only 4/10 archival galaxies, meaning that we cannot confirm the HCN detection of four galaxies; since the optical depth of HCN is typically much weaker than that of C₂H₂, however, it is possible that our lower detection rate is a result of improved S/N in their reduction process. We also confirm the detection of CO₂ in IRAS 15327+2340 (Arp 220). No galaxies in the non-masing control sample showed absorption in any molecular band, nor did any of the OHMs observed in our dedicated program.

Since these gas-phase absorption features are actually a blend of multiple absorption lines, the peak optical depth measured is a function of the velocity resolution of the spectrograph. We therefore report the integrated flux and the peak depth in normalized flux units (f_norm, where the spectrum has been divided by the adopted continuum; Spoon et'al. 2004) in Table'8.

Lahuis et'al. (2007) model abundances for ULIRGs with detections of C₂H₂, HCN, and CO₂, and suggest that they are associated with a phase of deeply embedded star formation, excluding the possibility of the features arising from an X-ray dominated region (XDR) powered by AGNs. Darling (2007) has also shown that OHMs have the highest mean molecular gas densities among starburst galaxies (traced by the J = 1 → 0 rotational HCN transition) and also possess high dense molecular gas fractions, comprising a distinct population in the IR–CO relation. The results of Lahuis et'al. (2007) show that 9/15 ULIRGs with absorption in both C₂H₂ and HCN are known OHMs in our Spitzer sample; this dense gas fraction (∼50%) is nearly identical to the observed OHM fraction in starbursts with dense (ULIRGs with L_HCN/L_CO>0.07) fractions of molecular gas (Gao & Solomon 2004b; Darling 2007; Baan et'al. 2008).

Given that the S/N ratio for the OHM and non-masing galaxies are of comparable magnitude, the lack of detection of any gas-phase species in the non-masing galaxies is a striking difference compared to the OHMs. Figure'9 shows the median stack of both samples near the regions of gas-phase absorption; while the C₂H₂ feature at 13.7 μm can be clearly seen in the median OHM spectrum, neither the HCN nor the CO₂ transition is prominent. Since the data have been median stacked (as opposed to a mean, which can be dominated by a few deep absorbers), this suggests low levels of C₂H₂ present in a significant fraction of the OHM host galaxies. No molecular absorption appears in the medianed spectrum for the non-masing galaxies.

Zoom In Zoom Out Reset image size

While a connection between OHMs and dense molecular gas is known to exist, the lack of detected molecular absorption in the mid-IR for the majority of OHMs is not entirely unexpected. OHMs occur in merging galaxies where different populations of gas may be kinematically and thermally distinct, yet are observed as a single unresolved region within the Spitzer beam. Lahuis et'al. (2007) suggest that high abundances of warm, dense gas are associated with deeply embedded star formation, where H ii regions are prevented from expanding by large pressure gradients and extend the lifetime of the star formation process.

Baan et'al. (2008) interpret dense gas abundances as excluding very hard radiation fields (such as those found in XDRs) that dissociate the molecules; they suggest that the molecular emission arises from PDRs surrounding H ii regions. Although few OHMs show absorption from dense molecular gas, the OHM sample also has few identified AGNs or XDRs (only 4/51 OHMs show [Ne v] at 14 μm). The connection between dense molecular gas and the presence of an AGN is thus unclear based on this data alone.

5.5.5. Gas-phase OH

For three OHMs, we report detection of the ²Π_1/2 J'= 5/2→ ²Π_3/2 J'= 3/2 OH absorption doublet near 34.616 μm: III Zw 35 (IRAS 01418+1651), Mrk 273 (IRAS 13428+5608), and Arp 220 (IRAS 15327+2340). This feature is generally difficult to detect since it lies near the noisy, far-red edge of the LH module; for all objects with z>0.08, it is redshifted out of the IRS range. OH absorption at 34.6 μm in Arp 220 was first reported by Skinner et'al. (1997) using ISO and confirmed with the IRS by Farrah et'al. (2007), who incorrectly identified it as the OH⁻ ion. All OH absorption features are well fit with a single Gaussian (Figure 10), since the separation between the doublet features (Δλ ≃ 0.02 μm) is comparable to the resolution element in the LH module. No detection of OH absorption was made for any of the non-masing galaxies.

Assuming the OH transitions are optically thin, we can use the EW to derive a column density for the OH ground state, which is likely to be a good proxy for the total column at typical molecular cloud densities (Bradford et'al. 1999):

$$\begin{equation} N_l = \frac{EW}{A_{ul}} \frac{8 \pi c}{\lambda ^4} \frac{g_l}{g_u}. \end{equation} \tag{ 6 }$$

OH column densities for all galaxies are quite similar, lying between (1–3) × 10¹⁷ cm^–2 (Table'9). The measured N_OH from the IRS data for Arp 220 also agrees within a factor of two of the column measured with ISO (Skinner et'al. 1997). Limits for galaxies in which the 34.6 μm OH feature is not detected are of order N_OH ≲ 1 × 10¹⁷ cm^–2.

Table 9. Properties of OH Gas-phase Absorption

OHM	fOH	τpeak	EW	NOH	γabs	γOHM	ϕpump
	(10–21 W cm–2)		(10–3 μm)	(cm–2)	(photons s–1)	(photons s–1)	(%)
IRAS 01418+1651 (III Zw 35)	–6.1	0.15	5.5	1.1 × 1017	1.7 × 1054	1.8 × 1053	10
IRAS 13428+5608 (Mrk 273)	–28.5	0.14	10.2	2.1 × 1017	1.7 × 1055	1.4 × 1053	0.8
IRAS 15327+2340 (Arp 220)	–172	0.21	15.0	3.0 × 1017	2.3 × 1055	1.6 × 1053	0.7

Note. The pumping efficiency (ϕ_pump = γ_OHM/γ_abs × 100) assumes all pumping comes from the 34.6 μm transition.

Download table as:  ASCIITypeset image

Zoom In Zoom Out Reset image size

We compare the N_OH derived from the rotational 34.6 μm transitions to the OH column density measured in galaxies who show the hyperfine 1667 MHz feature in absorption. The majority of such galaxies are ULIRGs of comparable luminosity to the galaxies in our non-masing sample. Measurements from 10 OH absorbers (Baan et'al. 1992; Darling 2007) give N_OH = T_ex(1.8 ± 1.9) × 10¹⁵ cm^–2, where T_ex is the OH excitation temperature in K. If the dust and gas are well mixed, then the temperature of the dust (∼50–100 K) can be used as a proxy for T_ex. This gives OH column densities for both OHMs (34.6 μm) and non-masing galaxies (1667'MHz) with comparable values of N_OH ≃ 10¹⁷ cm^–2. If so, then this addresses one of the crucial differences between OHMs and non-masing ULIRGs—namely, that differences in the abundance of masing molecules are not a key factor for triggering an OHM.

The amount of OH available in the galaxy can also test models of the OHM pumping mechanism. Skinner et'al. (1997) computed the photon flux (γ_abs = L^OH_abs/hν_OH) absorbed in the 34.6 μm transition from the OHM Arp 220. They found that γ_abs is roughly 1% of the photon flux in the OHM (γ_OHM = L_OHM/hν_OHM). If the 18 cm and mid-IR pumping photons lie along the same line of sight, this means that pumping photons from the 34.6 μm transition alone can power the OHM (given an efficiency of ∼1% or higher). While radiative transfer models from Lockett & Elitzur (2008) suggest that the 53 μm OH transition likely contributes more pumping photons than the 34.6 μm line, the energetics are consistent with the basic accepted mid-IR pumping model.

For the galaxies in our sample with OH absorption, Arp 220 and Mrk 273 would require pumping efficiencies on the order of 1% to power the OHMs from the 34.6 μm transition alone. III Zw 35 shows the weakest OH absorption among our detections and would require an efficiency of ϕ_pump ≃ 10%.