galaxies: interactionsgalaxies: ISMgalaxies: starburstinfrared: galaxies
106, 1998 September 1We discuss near-infrared spectra of 20 interacting galaxies from the Arp Catalog and analyze the properties of similar galaxies for which only optical spectra are available. We find excellent agreement between the types of activity determined from the infrared and optical, demonstrating that obscuration does not seriously bias the optical results. None of the 20 galaxies show infrared spectral characteristics that differ from expectations for isolated galaxies; the very strong shock-excited lines seen in a few interacting systems must be uncommon. Most of the galaxies in our infrared sample are the sites of starbursts that appear to have had durations of 1 to 5 Myr and to be observed 3 to 10 Myr after the peak star-forming episode. Four of the 20 galaxies have LINER or composite starburst/LINER spectra that are likely to arise from shocks due to supernovae in late phase starbursts. In the full interacting galaxy sample, there is a substantial excess of Seyfert 2 nuclei, supporting previous indications that this type of activity tends to occur in interacting host galaxies.
Subject headings: galaxies: activegalaxies: interactionsgalaxies: ISMgalaxies: starburstinfrared: galaxies
Interaction and merging play an important role in the evolution of galaxies. Larson & Tinsley (1978) showed that the Arp Catalog galaxies (Arp 1966) are characterized by large scatter in a U-B versus B-V diagram, which they interpreted as the result of recent bursts of star formation. Rieke et al. (1980) first pointed out a tendency for interacting galaxies to have strong infrared emission, as has been confirmed by both ground-based (Cutri & McAlary 1985) and IRAS data (e.g., Telesco, Wolstencroft, & Done 1988). Interacting systems show a wide range of nuclear activity, including active nuclei (AGNs), nuclear starbursts, ultraluminous IR emission, and poststarburst activity. These phenomena presumably all arise from the ability of interactions to produce conditions in which gas is transported into the galactic nuclei (e.g., Barnes & Hernquist 1991), as required to fuel both AGNs and starbursts. In addition, numerical simulations indicate that mergers may be one of the possible progenitors of dwarf and elliptical galaxies (Elmgreen, Kaufman, & Thomasson 1993).
Optical spectroscopy of a sample of interacting systems has been analyzed by Liu & Kennicutt (1995a; 1995b, hereafter LK95). They addressed the question of how the properties compare with those of isolated galaxies and how they depend on the morphological type of the merger. The characteristics of the optical spectra presented by LK95 range from those of an old quiescent stellar population to those of an AGN.
Although optical spectroscopy can access a wealth of emission lines at high sensitivity, the strong infrared emission of interacting galaxies demonstrates that their energy sources are dust enshrouded. To probe their intrinsic properties, complementary infrared spectra are required to penetrate the obscuration. In addition, near-infrared spectra are useful for investigating additional aspects of both the physical conditions of the gas and the properties of the stellar population. For example, the [Fe II] and H2 emission lines probe the shocked gas. Some merging galaxies show dramatic effects in these lines: NGC 6240 displays unusually strong [Fe II] and H2 lines that indicate a galaxy-wide shock of the interstellar medium as the two parent galaxies merge (Rieke et al. 1985; van der Werf et al. 1993).
In this paper we present NIR spectroscopy of a sample of interacting and peculiar galaxies. In § 2 we describe how the sample is drawn from the Arp Catalog. In § 3 we present the observations and data reduction. In § 4 we use infrared and optical features to study the nuclear activity in our sample of galaxies. Section 5 presents our conclusions.
Our sample of galaxies has been selected from Bushouse & Stanford (1992, hereafter BS92). Red excesses in the NIR colors of galaxies due to hot dust or reddening are often associated with different types of nuclear activity. Our first selection criterion involves drawing objects from the BS92 sample with red excesses in the JHK photometry. The typical NIR colors of a normal galaxy are J-H 
0.7 and H-K 0.2 (see, e.g., Glass 1984; Willner et al. 1984); thus, we selected all the objects in the BS92 sample showing nuclear colors of J-H 
0.75 and H-K 0.40. We find that 96 systems (137 individual members of the systems) meet this constraint. This sample will be referred to as the color-selected sample. A further criterion is imposed by our instrumentation. We have estimated an apparent magnitude of K = 12.1 as the limit for obtaining good-quality NIR spectroscopy with reasonable integration times. Therefore, all the objects from the color-selected sample with apparent magnitudes of K 12.1 were rejected. The second criterion left us with a total of 54 individual galaxies (members of 47 systems). These objects will be referred to as the magnitude-limited sample.
We now address the question of whether the magnitude selection biases our sample toward the most luminous objects. For each galaxy, the nuclear K-band luminosity is calculated from the apparent K-band magnitudes measured through a 5
aperture in BS92 and redshifts available in the literature (taken from NED), assuming a Hubble constant of H0 = 75 km s-1 Mpc-1. Figure 1 shows histograms representing the nuclear K-band luminosity distribution of the sample before (solid lines, 119 objects) and after (shaded bars, 54 objects) the magnitude criterion is applied. Redshifts are not available for all the galaxies in these samples, so the figure represents only a portion of the totals. A Kolmogorov-Smirnov test on the distributions of K-band luminosities indicates only a 45% probability that they could be different at a significant level. The average nuclear absolute magnitudes are 
K
C = -21.8 and KM = -22.3 for the color-selected and magnitude-limited samples, respectively, again supporting the conclusion that the magnitude-limited sample is not strongly biased toward luminous objects compared to the color-selected sample.
Fig. 1
The different subsamples (i.e., BS92 and our color-selected and magnitude-limited samples) drawn from the Arp Catalog are examined further in the histograms of Figure 2, which shows the distribution in redshift; in this case, we plot systems of galaxies, rather than individual galaxies. The average distances for each sample are: dArp = 68 Mpc, dBS92 = 75 Mpc, dC = 78 Mpc, and dM = 55 Mpc. The BS92 sample is slightly more biased toward distant objects than the Arp Catalog, since the field of view of their images misses very extended systems (which are likely to be closer). Our magnitude-limited sample tends to favor closer objects. Combined with the results given in the preceding paragraph, magnitude selection leads primarily to the inclusion of galaxies that are relatively close, with a secondary tendency to favor higher NIR luminosity than the full Arp sample.
Fig. 2
Figure 3 compares the above samples in terms of their total infrared luminosity (between 8 and 1000 
m). 1 Because of the spatial resolution of the IRAS observations, the luminosities generally comprise the whole system. For this analysis, we have included systems with both IRAS detections and upper limits. The total number of objects with IRAS data in each sample are: 239 systems in the Arp Catalog, 129 systems in the BS92 sample, 71 systems in the color-selected sample, and 42 systems in the magnitude-limited sample, with average IR luminosities log LIRArp = 10.2 L
, log LIRBS92 = 10.3 L, log LIRC = 10.7 L, and log LIRM = 10.7 L, respectively. We find that while the BS92 selection criteria leaves the sample practically unbiased in far-infrared luminosity, our color selection tends to include more objects with high infrared luminosities (one third of our color-selected sample are galaxies with LIR > 1011 L), as expected, since luminous infrared galaxies show NIR excesses (see, e.g., Carico et al. 1988). This effect is no more apparent in the magnitude-limited sample, showing again that the primary effect of magnitude selection is to favor nearby systems.
Fig. 3
We first address whether our magnitude-limited sample is equally distributed among the morphological classes in the Arp Catalog. Systems in the catalog are grouped into five main classes: class I (from Arp 1 to Arp 36) = peculiar spiral galaxies; class II (from Arp 37 to Arp 101) = spiral galaxies with a companion attached to a spiral arm; class III (from Arp 102 to Arp 145) = systems containing at least one elliptical or elliptical-like galaxy; class IV (from Arp 146 to Arp 268) = systems in which the components are not clearly classified as spirals or ellipticals (because of distortions produced by the interaction); class V (from Arp 269 to Arp 332) = groups of galaxies. From Table 1, we can see that our criteria may tend to select more systems belonging to class IV and V. However, a 
2 test shows that the Arp Catalog and the magnitude-selected sample have only a 68% probability of having different distributions over the Arp morphological classes; any difference is not highly significant.
The morphological types defined by LK95 may represent interaction types better than those of Arp. These are: type 0 = a system with two elliptical galaxies; type 1 = a system with an elliptical and a disk galaxy; type 2 = a merger in which the progenitors are not longer classifiable as disk or elliptical galaxies; type 3 = a system of two disk galaxies. Since the Arp Catalog contains not only pairs of interacting galaxies but also peculiar objects and groups of galaxies, we add two types: type -1 = peculiar but isolated galaxies, and type 4 = groups of three or more galaxies. We visually inspected the photographic plates in Arp (1966) to classify the galaxies according to the LK95 morphological types. Table 1 gives a summary of the results. A 2 test applied to the Arp Catalog and our magnitude-limited sample shows that there is a 50% probability that both samples are different, and this probability decreases to 25% when the BS92 and magnitude-limited samples are compared. That is, the differences are at best marginally significant. Thus, our samples appear to be representative of the parent sample of interacting galaxies in the Arp Catalog.
We have searched in the literature for optical spectroscopic data for all Arp galaxies that show clear evidence of interaction (i.e., systems from type 1, type 2, type 3, and also close pairs in groups of type 4) to derive the class of nuclear activity. The results are presented in Table 2, which gives in the first column the total number of galaxies in each sample and in subsequent columns the number of galaxies of each class, with the percentage this represents of the total number of galaxies with available data given in parentheses. Care has been taken to classify LINERs and Seyfert galaxies using the line-ratio diagnostic diagrams given in Ho, Filippenko, & Sargent (1993). We find that among the interacting galaxies in the Arp Catalog, there are 75 starbursts, 14 LINERs/starbursts, 18 weak LINERs, 2 13 LINERs, two LINER/Sy, 11 Seyfert 2's, and two Seyfert 11.5's.
When compared to all the Arp interacting galaxies, the magnitude-limited sample tends to select more Seyferts and Seyfert/LINER galaxies at the expense of starburst galaxies, but because of the small number statistics this difference is of marginal significance. The portion of galaxies with no nuclear activity (i.e., weak or absent emission lines) is virtually the same among all the samples. Summarizing, the sample including both color and the magnitude selection is typical of the whole Arp Catalog in terms of the distribution of morphological types, and probably also in terms of the nuclear activity.
The detectability of Seyfert nuclei depends significantly on the distance of the sample, and hence on the typical projected slit size on the galaxy and the sensitivity of obtainable spectra (Maiolino & Rieke 1995). The full sample of Arp interacting galaxies is at an average distance similar to that of the CfA Seyfert sample (Huchra & Burg 1992), so we have used the latter sample as a control to represent active galaxies where interaction was not a selection criterion. The CfA Seyferts are 1.9% of the parent sample of galaxies (Huchra & Burg 1992), and are distributed roughly equally between types 1 and 2. The exact division depends on how type 1.8 and 1.9 galaxies are counted; if they are grouped with type 2, there are 54% of this type, but if they are grouped with type 1, there are only 33% (Osterbrock & Martel 1993). Assuming that these portions apply to the Arp interacting sample, of 181 galaxies with spectra, we would predict there would be 1.7 each of type 1 and type 2 Seyferts. The number of type 11.5 Seyferts is in excellent agreement with this estimate. However, the number of type 2 Seyferts far exceeds it. Using the Poisson distribution, we find a probability of only 2 × 10-6 (i.e., a statistical significance of 0.999998 that the effect is significant) of finding 11 type 2 objects if the Arp sample behaves identically to the CfA. These results confirm the findings by Dahari (1985) and Maiolino et al. (1997) that interactions are more frequently present in hosts of Seyfert 2's than in Seyfert 1's and field galaxies.
Although very high quality spectra reveal LINER characteristics in a large portion of galaxies (Ho et al. 1993), given the average distance of the Arp interacting sample we would not expect a high detection rate for them. Nonetheless, 26% of the members with spectra are either LINERs or composite starburst/LINERs. LK95 also found indications of enhanced Seyfert and LINER activity in a similar sample of galaxies, but because of its smaller size (total 40 objects), they were uncertain of the statistical significance of the effect.
1 Here fIR = 1.8 × 10-14(13.56S12 + 5.26S25 + 2.54S60 + S100) in W m-2.
2 A galaxy is classified as a weak LINER when it is either a weak [O I] LINER, i.e., [O I]/H
< 0.16, or classified as a LINER by Keel et al. (1985) based only on the [N II]/H line ratio.
In this paper we present H- and K-band spectroscopic observations of 14 galaxies belonging to the magnitude-limited sample (see Table 3). Spectra were obtained in 1996 FebruaryMarch at the 1.5 m Italian infrared telescope TIRGO (Baffa et al. 1995), located at an altitude of 3100 m in the Swiss Alps. All the spectra were taken with the long-slit NICMOS-based spectrometer LonGSp (Vanzi et al. 1997) at a resolution of about 11 Å pixel-1, and a pixel size of 1
7. The slit was 3.4 × 70 arcsec2, aligned north-south. Spectra were obtained with the source placed at three different positions along the slit, so that the sky signal could be removed by subtracting the on-source spectrum taken immediately before or after a given observation. Only NGC 3690 was extended enough to require on-off observations. Single 60 s integrations were combined to reach the required total integration time. Flat-field spectra were obtained by observing the dome illuminated by a halogen lamp and used to correct for the nonuniform response of the detector along the slit direction. No dark correction is necessary, since the detector dark current is less than 1 e s-1 and no offset is introduced by the electronics.
During the data reduction, particular attention was paid to the subtraction of the sky from the source spectra. Each on-source spectrum (containing the sky spectrum) was multiplied by an appropriate factor and shifted in wavelength to account for both the sky brightness variations and the mechanical instabilities of the spectrometer (Vanzi et al. 1995). For each on-source spectrum, the sky was the average of two adjacent observations. The mechanical instabilities of the spectrometer produce variations in the line position of few hundredths of a pixel, on timescales on the order of a minute. Although the effect is very small, it may considerably affect the final spectrum once sky subtraction has been performed. A good level of sky subtraction was accomplished in most cases using the task SKYSUB in the IRSPEC context of MIDAS. This task allows for an automatic optimization of both the multiplying factor and the shift. We found, however, that the best result was achieved by a manual optimization through visual inspection of the spectra. Those few spectra for which a good sky subtraction was not achieved were discarded. Bad pixels were rejected in the final combined two-dimensional spectra. From each two-dimensional image, a one-dimensional spectrum was extracted with an aperture chosen to optimize the S/N ratio (see Table 3, col. [4]). The wavelengths were calibrated with airglow emission lines (Oliva & Origlia 1992) reaching an accuracy better than 2.2 × 10-3 m at all wavelengths. The reddest OH line used for calibration in the K band is (97)P2(6) at 2.2307 m. The atmospheric absorption features were removed by dividing the object spectrum by a G-type star spectrum observed at similar air mass; such spectroscopic reference stars were observed approximately every hour during the night to have a good sampling of variations in atmospheric transmission with time. The best correction was usually obtained using the averaged spectrum of stars observed before and after the source. The stellar features artificially introduced in the spectra were then removed by multiplying the source spectrum by the solar spectrum (Maiolino, Rieke, & Rieke 1996). No attempt was made to flux calibrate the spectra.
The sample and the log of the observations are given in Table 3. The system Arp 299 was observed with two different slit positions. In one, we obtained a spectrum of the nucleus of Arp 299E, while the second position allowed us to extract two spectra of the western component, one centered on source C and the second one centered on source B1, partially contaminated by source B2 (in the Wynn-Williams et al. 1991 notation). Hereafter, these last two spectra obtained for the western component will be referred to as NGC 3690C and NGC 3690B, and that of the eastern component as IC 694.
In addition, we will discuss NIR spectra for six more galaxies taken from the literature, namely, Arp 157/NGC 520, Arp 186/NGC 1614, and Arp 215/NGC 2782 (Engelbracht et al. 1998b), and Arp 182W/NGC 7674, Arp 245 N/NGC 2992, and Arp 298S/NGC 7469 (Ruiz 1997).
Spectra (in raw counts) are presented in Figure 4. We do not detect any emission lines in the H-band spectra of Arp 232, or in either the H- or K-band spectra of Arp 294S, and therefore these are not shown in this figure or discussed any further in this paper. In Table 4 we present the measured equivalent widths, EW (in Å), of the brightest lines detected, and the photometric CO index (see below). In addition to the observed galaxies, we also list these quantities for galaxies in our sample observed by other authors. In our final sample (i.e., observed galaxies plus data from the literature), we find 11 SB galaxies, two SB/LINERs, three LINERs, and five Seyfert galaxies.
Fig. 4
Some possible trends in the behavior of the NIR lines are suggested by the correlation between the morphological type of the system and the equivalent width of H + [N II] found by LK95 . We confirm this trend for [Fe II] 1.644 m, Br
, and He I(2.058). However, for the EW of H2(1, 0)S(1) at 2.121 m, we find that the tendency may be present but with lower significance.
The [Fe II] 1.644/Br line ratio (or, in an equivalent way, [Fe II] 1.256/Pa
) combined with the optical line ratio [O I] 6300/H can be used to distinguish between starburst and Seyfert activity, i.e., between stellar and nonstellar activity (Mouri, Kawara, & Taniguchi 1993; Goodrich, Veilleux, & Hill 1994; Simpson et al. 1996; Alonso-Herrero et al. 1997), since there is a progression with increasing values of these ratios from pure photoionization to pure shock excitation. The [Fe II] line at 1.644 m is well detected in most of our spectra and therefore can be used to constrain and/or confirm the type of nuclear activity in our sample of interacting galaxies.
Since our spectra are not flux calibrated, special care was paid to the determination of a reliable value for the [Fe II] 1.644/Br line ratio. For the two galaxies where Brackett 11 is well detected (NGC 3690C and Arp 238E), we measured the [Fe II] 1.644/Br11 line ratio, and then calculated the [Fe II] 1.644/Br ratio by assuming Br11/Br = 0.25, as derived from Hummer & Storey (1987). This method ensures a good correction for possible reddening and is not affected by any calibration uncertainties. For those galaxies with a CO index above 0.15 (i.e., those for which the NIR continuum is dominated by evolved giants and supergiants), we used the typical color of an evolved stellar population (H-K = 0.2) for a relative calibration of the H and K spectra. The choice of this value is justified by both the predictions of our model and data on unreddened galaxies. In addition, in this case the calculated value of the [Fe II] 1.644/Br line ratio is automatically corrected for reddening. Finally, for those objects with low CO index (IC 694, Arp 160, Arp 94S, Arp 155, Arp 193, and Arp 215), we use the observed H-K nuclear color (for a 5 diameter aperture) given by BS92. In these cases, the [Fe II] 1.644/Br line ratio is not corrected for extinction, and therefore the values must be considered as a lower limit. In the two cases in which we could use two methods (Arp 238E and NGC 3690C), the differences in the computed [Fe II] 1.644/Br line ratio are between 0.1 and 0.3 in log. The log of the [Fe II] 1.644/Br line ratio along with the method used (as described above) are given in column (7) of Table 4. The [O I] 6300/H line ratios are taken from the literature; see Table 3, column (8) for references.
Figure 5 gives a [Fe II] 1.644/Br versus [O I] 6300/H diagram (see Alonso-Herrero et al. 1997). In this diagram, the [Fe II] 1.644 fluxes have been corrected for blending with Br12 emission by assuming Br12/Br = 0.19. Only in galaxies with very high extinction would this correction not be appropriate. The boxes indicate approximate regions for starburst galaxies, Seyferts, and composite objects (i.e., LINERs and galaxies with evidence for both AGN and starburst activity); these boxes are merely indicative, especially for composite objects (see discussion in Alonso-Herrero et al. 1997). All the galaxies in our sample are located in the overlapping region of the classes mentioned above. The Seyfert galaxies and LINERs, as classified from optical data (Arp 94S, Arp 155, Arp 182, Arp 243, and Arp 245N), are located in the upper part of the overlapping region of Seyfert galaxies and composite objects, where Alonso-Herrero et al. (1997) find approximately half of the AGN plotted in their Figure 3. Most of the interacting galaxies classified as starbursts or H II regions from the optical data (Arp 80, Arp 84N, Arp 90E, Arp 160, Arp 238E, and NGC 3690C) are located in the upper right corner of the starburst region, where again most of the galaxies of this class lie in Figure 3 of Alonso-Herrero et al. (1997). The objects showing a combination of star formation and AGN or LINER activity in the optical spectra (Arp 91E, Arp 193, and Arp 298S) are in the region populated by composite objects. Finally, Arp 90E, IC 694, and NGC 3690B are in the composite-object zone but very close to the starbursts (the optical line ratios of the first two put them close to the boundary between starbursts and LINERs in the diagrams of Ho et al. 1993).
Fig. 5
Despite the reddening independence of the classification method used in Figure 5, the determinations of activity type are all consistent with the results of previous optical spectroscopy. Thus, for example, the evidence for an enhanced incidence of Seyfert 2 and LINER activity in the total interacting galaxy sample discussed in § 2 appears to be intrinsic and not a result of misclassifications due to obscuration. At the same time, identification of morphological peculiarities is difficult in edge-on galaxies, so they are effectively excluded from our samples. Consequently, our samples are also subject to the tendency for nearly all optically selected samples of AGNs to underestimate their true space density by factors of 2 or more (e.g., Keel 1980; McLeod, & Rieke 1995).
CO absorption bands attributable to red giant and supergiant stars are detected in the observed spectra of most of the galaxies in our sample. For the 12CO (2,0) band at 2.293 m, we measure a spectroscopic CO index according to the definition of Kleinmann & Hall (1986), which is transformed into a photometric index (see Vanzi & Rieke 1997 for more details). Our results are presented in Table 4, column (6).
The EW of Br can be used as a measure of the time since the last period of strong star formation, since it measures in an extinction-independent manner the proportion of young hot stars through the Br flux relative to the evolved red supergiants through the K-band continuum. A correlation between the CO index and EW(Br) is therefore expected in objects in which star formation plays a significant role in powering the nuclear activity. In Figure 6 we plot the CO index against log (K/Br) for the galaxies in our sample, along with the blue dwarf galaxies (BDGs) from Vanzi & Rieke (1997). We also display the predictions of a starburst model (Rieke et al. 1993) for a solar-neighborhood IMF and a Gaussian star formation rate (SFR) for three different durations: 1 Myr (solid line), 5 Myr (dot-dashed line), and 100 Myr (dashed line).
Fig. 6
All but three of these objects are in substantial agreement with the predictions of the models for short-duration bursts of star formation. The CO index and EW(Br) of the BDGs seem to be well reproduced by the model for very short bursts (duration 
1 Myr). The galaxies from the Arp sample are located between the curves for this model and bursts of moderate duration (5 Myr) and appear to be observed 3 to 10 Myr after the peak star formation rate.
The three exceptional galaxies are Arp 94S, Arp 155, and Arp 91E, which are classified respectively as a Seyfert, LINER, and SB/LINER from their optical spectra. Arp 91E (SB/LINER) would move into the zone of old starbursts with a slight filling-in of its CO absorption bands by emission from hot dust. The SB/LINER systems Arp 193 and Arp 157 both have extremely strong Balmer absorption features (LK95), suggesting that they also harbor late-stage starbursts. The former galaxy appears in the young starburst zone of Figure 6, suggesting either filling in of its CO with emission by hot dust or the presence of a very recent episode of star formation in addition to an older one. The latter galaxy is not plotted because we do not have a CO band measurement. The LINER Arp 243 falls in the SB region of the diagram but in a position that suggests that it is observed longer after the peak SFR than any of the SB galaxies. It also has extremely deep Balmer absorption (LK95), consistent with this placement. Thus, four of the galaxies with SB/LINER or LINER characteristics behave consistently with the argument of Engelbracht et al. (1998a) and Alonso-Herrero et al. (1998) that some weak [O I] 
6300 LINERs are late-stage starbursts whose LINER characteristics are produced in supernova-driven shocks.
In SB galaxies, the [Fe II] lines are almost entirely produced by SNR and therefore directly related to the SB activity (see Greenhouse et al. 1991; Forbes & Ward 1993; Vanzi & Rieke 1997; Engelbracht et al. 1998a), whereas for AGNs they probably arise largely in photoionizing shocks driven by the nuclear activity (Ruiz 1997).
Is all the [Fe II] emission in interacting galaxies similar to that in isolated galaxies, or is an extra component present? To investigate this issue, in Figure 7a we plot the [Fe II] 1.644 + Br12/Br line ratio against log (K/Br), along with the output of the 1 Myr and 100 Myr starburst models. In the model, the [Fe II] is assumed to be produced entirely by SNR and has been calibrated using M82, as described in Vanzi & Rieke (1997). From this diagram it is clear that all our starburst galaxies lie very close to the model lines, as also found for some BDGs and bright starbursts (see Figure 3 in Vanzi & Rieke 1997; some BDGs lie below the model curves, but this behavior is consistent with their low metallicity). Deviations to the right side of the model lines are well understood in terms of dilution from the underlying evolved stellar population. The age estimates from the galaxy placements on this diagram are in good agreement with those from Figure 6.
Fig. 7
A similar analysis is presented in Figure 7b, where the EW of [Fe II] + Br12 is plotted against that of Br. Again, the model predictions are shown for the two limiting cases, the 1 Myr burst (solid line) and the 100 Myr burst (dashed line). Symbols are as in Figure 7a. In this diagram the time elapsed after the burst peak increases from right to left. The largest deviations from the model in this case are found for those galaxies with the youngest bursts of star formation (NGC 3690C, IC 694, Arp 238E, and the BDG NGC 5253). The detection of the 1.7 m He I line in NGC 5253 (Vanzi & Rieke 1997) and probably in NGC 3690C (this work) indicates the presence of stars hotter than 40,000 K, demonstrating very recent star formation directly, while a similar inference can be made for the other galaxies from the large values of the EW of Br. Because of the youth of the star formation episode in these objects, it is possible that emission by dust and/or gas adds significantly to the stellar continuum, so that the colors predicted by the model are not realistic. The other deviant point is for Arp 182W, a strong Seyfert galaxy (NGC 7674). We conclude that the behavior in this figure is also generally consistent with the starburst model for the excitation of [Fe II].
The sample studied is not large enough to identify truly exceptional cases (such as NGC 6240). However, it appears that no significant excesses in the [Fe II] emission (for instance as a result of interaction-induced shocks) are present in the majority of interacting systems when compared with isolated starburst galaxies.
It is unclear whether molecular hydrogen in galaxy spectra is excited through shocks (via SNRs, winds from young stars, or molecular cloud collisions), UV fluorescence, or X-ray heating. The situation appears to be even less clear for interacting galaxies and mergers, in which an extra component of molecular hydrogen emission associated with the interaction process might be present. For instance, it has been suggested that the very bright emission in molecular hydrogen observed in NGC 6240 and Arp 220 could be associated with shocks related to the ongoing merger (e.g., Rieke et al. 1985; van der Werf et al. 1993).
In Figures 8a and 8b, the H2 (2.121)/Br line ratio and the EW(H2) are compared with log K/Br and log EW(Br), respectively. Since the two lines are very close in wavelength, EWs, fluxes, or ratios of lines are equally good tools for the analysis, and in principle extinction effects are negligible. The symbols are as in Figure 7a. Again, in these figures we display the predictions of the starburst model for two burst durations, 1 Myr and 100 Myr. The starburst model assumes that all the molecular hydrogen is excited by SNR expanding into the interstellar medium formed during the episode of star formation, and again has been calibrated assuming that all the molecular hydrogen emission in M82 arises from SNRs (see Vanzi & Rieke 1997). We have expanded the data set by including the measurements of Goldader et al. (1997a), which include both isolated galaxies (Fig. 8, open circles) and interacting galaxies (filled circles). As expected, the scatter in this diagram is much larger than that observed for the [Fe II] emission, since it appears that mechanisms other than supernova-driven shocks contribute to the H2 emission in starbursts. Figures 8a and 8b confirm the tendency found in BDGs (Vanzi & Rieke 1997) for shock excitation by SNR to define only a lower limit to the total H2 (2.121) emission.
Fig. 8
No significant differences appear to be present between isolated and interacting galaxies, although some interacting galaxies show slightly larger EWs of both Br and H2 than the isolated ones, consistent with their tendency to have more luminous starbursts. Again, we find that the largest deviations from the model in our sample tend to occur for those galaxies with the youngest episodes of star formation (IC 694, Arp 193, and Arp 238E). NGC 6240 is the anomalous point near the top of Figure 8b. As with the [Fe II] emission, it appears to be highly exceptional, and the H2 of other interacting and merging galaxies is very similar in strength to that for starbursts in isolated galaxies.
Interestingly, NGC 3690C lies in Figure 8b very close to the model line, even though we show that its H2 may have a large fluorescent component. Such a situation may also hold for other galaxies in this sample. Useful line ratios for distinguishing among different excitation mechanisms involve H2(1, 0)S(1) at 2.121 m, H2(2, 1)S(3) at 2.073 m, H2(1, 0)S(0) at 2.223 m, and H2(2,1)S(1) at 2.247 m. Some of these lines are detected at a lower level in the K spectrum in some of the galaxies in our sample. EW for these lines are given in Table 5 along with the ratios with the H2(1, 0)S(1) line. We compare the above ratios with the predictions from models for thermal and optically thin fluorescent excitation (Black & van Dishoeck 1987). Results from the models are summarized in Table 6. The detection of the H2(2, 1)S(1) line in IC 694 and Arp 243 and the detection of the H2(2, 1)S(3) line in NGC 3690C and Arp 193 could be interpreted as evidence for fluorescent excitation of the molecular hydrogen. However, when we compare other H2 line ratios, such as H2(2, 1)S(3)/H2(1, 0)S(1) and H2(1, 0)S(0)/H2(1, 0)S(1) or H2(2, 1)S(1)/H2(1, 0)S(1) versus H2(1, 0)S(0)/H2(1, 0)S(1) (see Fig. 12 in Goldader et al. 1997b), we find that NGC 3690C is the only convincing case for fluorescence dominating the H2 spectrum.
We have drawn a sample of interacting and peculiar galaxies from the Arp Catalog of Peculiar Systems and the BS92 NIR study of a subsample of Arp galaxies on the basis of the presence of NIR excesses (the color-selected sample) with a limiting K-band apparent magnitude K 
12.1 (the magnitude-limited sample). We show that these samples are generally representative of the full sample of interacting Arp galaxies, except for a bias toward enhanced nuclear activity. In addition to discussing optical data from the literature on the full interacting Arp galaxy sample, we analyze infrared spectra of 20 galaxies from our magnitude-limited sample. We demonstrate that
1. Classification of nuclear activity type using infrared spectroscopy agrees well with classification using optical spectroscopy, showing that obscuration is not seriously biasing the optical results.
2. Nearly all the galaxies with H II type spectra appear to be the sites of powerful starbursts with star formation durations of 1 to 5 Myr, observed 3 to 10 Myr after the maximum star formation rate.
3. Four of the galaxies with LINER or composite starburst/LINER spectra have characteristics suggesting that the LINER characteristics are shock excited by supernova explosions in a late-phase starburst.
4. None of the 20 galaxies for which we have infrared spectra exhibit anomalously strong shock-excited [Fe II] or H2 emission, demonstrating that the extensive shock excitation of the interstellar medium in the merging systems NGC 6240 and Arp 220 is exceptional.
5. There is a significant excess of type 2 Seyfert nuclei in the full Arp interacting galaxy sample, supporting previous indications that Seyfert 2 nuclei tend to be found in interacting or distorted host galaxies.
We wish to thank Alessandro Marconi, Giovanni Moriondo, and Francesca Ghinassi for assistance during the observations. We also thank Chad Engelbracht and Milagros Ruiz for making their data available before publication. L. V. acknowledges support from the Agenzia Spaziale Italiana under grants 95-RS-120 and ARS-96-66. During the course of this work, A. A.-H. was supported by the National Aeronautics and Space Administration on grant NAG 5-3042 through the University of Arizona. The work of G. H. R. was supported under NSF grant AST 95-29190. This research made use of the NASA/IPAC Extragalactic Database (NED), which is operated by the Jet Propulsion Laboratory, California Institute of Technology, under contract to the National Aeronautics and Space Administration.
. 1997b, ApJ, 474, 104 First citation in article | IOPscience | ADS. 1995b, ApJ, 450, 547 (LK95) First citation in article | Crossref | ADS
Full image (53kb) | Discussion in textHistogram showing the distribution in absolute K-magnitude for the color-selected sample (119 galaxies) and the magnitude-limited selected sample (54 galaxies).
Full image (72kb) | Discussion in textHistogram showing the distribution in redshift for the Arp interacting galaxies (310 systems), the BS92 sample (162 systems), the color-selected sample (85 systems), and the magnitude-limited sample (47 systems).
Full image (70kb) | Discussion in textHistogram showing the distribution in infrared (81000 m) luminosity (in solar units) of the Arp galaxies, the BS92 sample, the color-selected sample, and the magnitude-limited sample. Note that we include galaxies with both IRAS detections and upper limits.
Full image (107kb) 
Full image (104kb) 
Full image (105kb) 
Full image (105kb) 
Full image (54kb) | Discussion in textSpectra in the H- and K-bands (in raw counts) of the sample of galaxies. The brightest emission lines are marked.
Full image (42kb) | Discussion in text[Fe II] 1.644/Br vs. [O I] 6300/H diagram. Filled triangles, starburst galaxies; open squares, Seyfert galaxies; crosses, intermediate objects (i.e., LINERs and galaxies with evidence for both star formation and AGN or LINER activity); open circles, SNRs.
Full image (44kb) | Discussion in textPhotometric index 12CO (2.293 m) vs. log (K/Br). The three lines represent models with a Gaussian burst of star formation with FWHM of 1 Myr (solid line), 5 Myr (dashed-dotted line), and 100 Myr (dashed line). The symbols on the curves represent intervals of 1 Myr for both the 1 and 5 Myr models, and intervals of 10 Myr for the 100 Myr model. The first point represents 1 Myr after the peak of star formation for the 1 Myr model curve, 3 Myr after the peak for the 5 Myr model, and 100 Myr before the peak for the 100 Myr model. Time increases as one moves from left to right. Filled triangles, starburst galaxies; open squares, Seyfert galaxies; crosses, intermediate objects (i.e., LINERs and galaxies with evidence for both star formation and AGN or LINER activity); stars, BDGs from Vanzi & Rieke (1997).
Full image (40kb) 
Full image (37kb) | Discussion in text(a) log ([Fe II] + Br12/Br) vs. log (K/Br). Lines and symbols as in Fig. 6. The output of the starburst model (Rieke et al. 1993) is calculated for two Gaussian bursts (FWHM = 1 Myr and FWHM = 100 Myr) and solar neighborhood IMF, plotted as solid and dashed lines, respectively. In this case, the second point on the 1 Myr model represents the peak of star formation. In the starburst model, the [Fe II] flux is assumed to be produced entirely by SNR and has been calibrated using M82 data as described in Vanzi & Rieke (1997). The symbols plotted on the model curves are spaced in intervals of 1 Myr for the short-duration burst and 10 Myr for the long-duration burst; the peak of star formation occurs at 1 and 10 Myr, respectively, which is the first point (to the left) plotted on each curve. (b) log EW([Fe II] + Br12) vs. log EW(Br). Lines and symbols are as in Fig. 7a.
Full image (40kb) 
Full image (41kb) | Discussion in text(a) log H2(1, 0)S(1)/Br vs. log (K/Br). Lines and symbols as in Fig. 7a. In addition, the open circles show isolated galaxies and filled circles show interacting galaxies, both samples from Goldader et al. (1997a). (b) log EW(H2) vs. log EW(Br). Lines and symbols are as in Fig. 8a.
| Arp Class (1) |
Number of Objects (2) |
Mag/Arp (%) (3) |
LK95 Type (4) |
Number of Objects (5) |
Mag/Arp (%) (6) |
| I... | 36 | 3 | -1 | 76 | 12 |
| II... | 65 | 12 | 0 | 8 | 13 |
| III... | 44 | 9 | 1 | 45 | 11 |
| IV... | 123 | 17 | 2 | 89 | 13 |
| V... | 64 | 19 | 3 | 66 | 24 |
 |
|
4 | 48 | 8 | |
| Total... | 332 | 14 |
Col. (1): Arp class. Col. (2): Number of objects of each class in the Arp Catalog. Col. (3): Percentage of objects of each class selected in the magnitude-limited sample, referred to the number of objects of the same class in the Arp Catalog. Col. (4): LK95 type. Col. (5): Number of objects in the Arp Catalog for each LK95 type. Col. (6): Objects of each type in the magnitude-limited sample as percentage of the number in the Arp Catalog.| Sample | No. | SB (%) | SB/L (%) | L (%) | L/Sy + Sy (%) | Weak-Lined | No Data |
| Arp Catalog... | 302 | 75 (42) | 14 (8) | 31 (17) | 15 (8) | 44 (25) | 123 |
| Color-selected... | 118 | 26 (36) | 6 (8) | 8 (12) | 11 (15) | 21 (29) | 46 |
| Magnitude-limited... | 48 | 12 (29) | 4 (10) | 8 (19) | 9 (22) | 8 (20) | 7 |
| Galaxy (1) |
tH (minutes) (2) |
tK (minutes) (3) |
Aperture (arcsec2) (4) |
v (km s-1) (5) |
Other Names (6) |
Nuclear Type (7) |
References Optical, NIR (8) |
| Arp 80... | 40 | 50 | 3.4 × 10 | 2160 | NGC 2633 | SB | 8, 12 |
| Arp 84N... | 50 | 50 | 3.4 × 8.8 | 3427 | NGC 5394 | SB | 7, 12 |
| Arp 90E... | 80 | 60 | 3.4 × 6.8 | 2672 | NGC 5930 | SB | 4, 12 |
| Arp 91E... | 80 | 30 | 3.4 × 10 | 1959 | NGC 5954 | SB/L | 8, 12 |
| Arp 94S... | 70 | 60 | 3.4 × 12 | 1157 | NGC 3227 | Sy | 4, 12 |
| Arp 155... | 80 | 50 | 3.4 × 8.5 | 2869 | NGC 3656 | L | 2, 12 |
| Arp 157S... | |
|
|
2266 | NGC 520 | SB | 9, 13 |
| Arp 160... | 24 | 24 | 3.4 × 8 | 2506 | NGC 4194 | SB | 1, 12 |
| Arp 182W... | |
|
|
8713 | NGC 7674 | Sy2 | 9, 14 |
| Arp 186... | |
|
|
4778 | NGC 1614 | SB | 9, 13 |
| Arp 193... | 45 | 50 | 3.4 × 10 | 7000 | UGC 8387 | SB/L | 2, 12 |
| Arp 215... | |
|
|
2562 | NGC 2782 | SB | 10, 13 |
| Arp 232... | 80 | 27 | 3.4 × 5 | 3183 | NGC 2911 | L | 6, 12 |
| Arp 238E... | 70 | 60 | 3.4 × 6.0 | 9313 | UGC 8335 | SB | 1, 12 |
| Arp 243... | 60 | 50 | 3.4 × 8.5 | 5535 | NGC 2623 | L | 2, 12 |
| Arp 245N... | |
|
|
2311 | NGC 2992 | Sy2 | 5, 14 |
| Arp 294S... | 68 | 26 | 3.4 × 8.5 | 2678 | NGC 3786 | Sy | 3, 12 |
| Arp 298S... | |
|
|
4892 | NGC 7469 | Sy1 | 11, 14 |
| Arp 299W... | 50 | 50 | 3.4 × 5.1 | 3132 | NGC 3690B | SB | 1, 12 |
| 50 | 50 | 3.4 × 3.4 | 3033 | NGC 3690C | SB | 1, 12 | |
| Arp 299E... | 32 | 34 | 3.4 × 8.5 | 3111 | IC 694 | SB | 1, 12 |
Cols. (2) and (3): Total integration times at H and K respectively. Col. (4): Size of the aperture (in arcsec2) used for extracting the final one-dimensional spectra. Col. (5): Heliocentric velocity from NED. Col. (6): Other names. Col. (7): Nuclear activity. Col. (8): References for the optical and NIR data. REFERENCES. (1) Armus, Heckman, & Miley 1989; (2) LK95; (3) Goodrich & Osterbrock 1983; (4) Keel et al. 1985; (5) Durret & Bergeron 1988; (6) Véron-Cetty & Véron 1986; (7) Leech et al. 1989; (8) González-Delgado & Pérez 1996; (9) Veilleux et al. 1995; (10) Kinney et al. 1984; (11) Morris & Ward 1988; (12) This paper; (13) Engelbracht et al. 1998b; (14) Ruiz 1997.| Galaxy (1) |
EW(Br) (Å) (2) |
EW(H2) (Å) (3) |
EW([Fe II]) (Å) (4) |
EW(He I) (Å) (5) |
(CO)phot (6) |
log [Fe II]/Br (7) |
Morphological Type (8) |
| Arp 80... | 5.0 ± 0.5 | 3.0 ± 0.5 | <3 | <2 | 0.21 | 0.07, b | -1 |
| Arp 84N... | 5.5 ± 0.5 | 4.5 ± 0.5 | 4.0 ± 0.5 | 1.5 ± 1 | 0.19 | 0.23, b | 3 |
| Arp 90E... | 7.5 ± 0.5 | 3.0 ± 0.5 | 3.0 ± 1 | 2.7 ± 0.5 | 0.21 | -0.02, b | 3 |
| Arp 91E... | 3.5 ± 0.5 | 4.5 ± 0.5 | 3.0 ± 1 | 2.0 ± 1 | 0.16 | 0.31, b | 3 |
| Arp 94S... | 2.5 ± 0.5 | 6.0 ± 0.5 | 3.5 ± 1 | 1.1 ± 0.5 | 0.11 | 0.41, c | 1 |
| Arp 155... | 2.0 ± 0.5 | 3.5 ± 0.5 | <2 | |
0.14 | 0.22, c | 2 |
| Arp 157S... | 12.5 | 9.2 | 10.8 | 5.8 | |
0.31, b | 2 |
| Arp 160... | 17.0 ± 0.5 | 4.3 ± 0.5 | 9.5 ± 1 | 7.8 ± 0.5 | 0.14 | 0.05, c | 2 |
| Arp 182W... | 3.0 ± 1 | 3.0 ± 1 | 7.5 ± 1 | |
|
0.77, d | 3 |
| Arp 186... | 20.7 | 4.5 | 10.7 | 10.1 | 0.19 | 0.09, b | 2 |
| Arp 193... | 20.0 ± 0.5 | 18.0 ± 0.5 | 11.0 ± 0.5 | 6.5 ± 1 | 0.12 | -0.09, c | 2 |
| Arp 215... | 8.8 | 4.7 | 8.7 | 4.0 | 0.13 | 0.37, c | -1 |
| Arp 232... | |
|
|
|
0.14 | |
2 |
| Arp 238E... | 34.0 ± 1 | 11.0 ± 1 | 20.0 ± 0.5 | 17.1 ± 1 | 0.19 | -0.17, a | 3 |
| Arp 243... | 6.5 ± 1 | 7.0 ± 1 | 7.0 ± 1 | 1.3 ± 0.7 | 0.23 | 0.41, b | 3 |
| Arp 245N... | |
3.1 ± 0.5 | 15.0 ± 1 | |
|
|
3 |
| Arp 294S... | |
4.0 ± 1 | <1 | |
|
|
3 |
| Arp 298S... | 3.8 ± 0.5 | 3.1 ± 0.5 | 4.4 ± 0.5 | |
|
0.44, d | 3 |
| NGC 3690C... | 46.0 ± 1 | 6.0 ± 0.5 | 35.0 ± 1 | 27.2 ± 0.5 | 0 | -0.14, a | 3 |
| NGC 3690B... | 12.5 ± 0.5 | 5.5 ± 0.5 | 4.0 ± 2 | 5.0 ± 0.5 | 0.19 | -0.11, b | 3 |
| IC 694... | 27.7 ± 0.5 | 26.8 ± 0.5 | 21.3 ± 0.5 | 15.0 ± 1 | 0.13 | 0.13, c | 3 |
Cols. (2), (3), (4), (5): Observed equivalent widths in Å of Br, H2 2.121 m, [Fe II] 1.644 m + B12, and He I 2.058 m. Col. (6): Photometric CO index. Col. (7): log [Fe II] + Br12/Br along with the method used: (a) flux of Br11; (b) color of an evolved population H-K = 0.2; (c) observed H-K; (d) calibrated fluxes. Col. (8): Morphological type (see § 2.2).| Galaxy | EW(1,0) S(0) | EW(2,1) S(1) |  |
 |
| Arp 243... | 2 ± 1 | 2 ± 1 | 0.3 ± 0.2 | 0.3 ± 0.2 |
| IC 694... | 5 ± 1 | 5 ± 1 | 0.2 ± 0.05 | 0.2 ± 0.05 |
| EW(1,0)S(0) | EW(2,1)S(3) | |
 |
|
| Arp 193... | 5.5 ± 1 | 2.5 ± 1 | 0.3 ± 0.1 | 0.1 ± 0.05 |
| NGC 3690C... | 3.5 ± 0.5 | 2 ± 1 | 0.58 ± 0.1 | 0.3 ± 0.2 |
| EW(1,0) S(0) | |
|||
| Arp 94S... | 3 ± 1 | 0.5 ± 0.2 | ||
| Arp 238E... | 6 ± 1 | 0.5 ± 0.1 | ||
| NGC 3690B... | 4 ± 1 | 0.7 ± 0.2 |
| Model | |
|
|
| Thermal S1... | 0.005 | 0.27 | 0.002 |
| Thermal S2... | 0.08 | 0.21 | 0.08 |
| UV (1)... | 0.52 | 0.57 | 0.22 |
| UV (14)... | 0.55 | 0.46 | 0.35 |
| UV (31)... | 0.55 | 1.18 | 0.26 |