Dependence of Nebular Heavy-Element Abundance on H I Content for Spiral Galaxies
Paul Robertson, Gregory A. Shields, Romeel Davé, Guillermo A. Blanc, Audrey Wright
aa r X i v : . [ a s t r o - ph . C O ] J un Dependence of Nebular Heavy-Element Abundance on H I Content for SpiralGalaxies
Paul Robertson , Gregory A. Shields , Romeel Dav´e , Guillermo A. Blanc , and AudreyWright Department of Astronomy, The University of Texas, Austin, TX, 78712; [email protected] Astronomy Department, University of Arizona, Tucson, AZ 85721, USA Carnegie Observatories, Pasadena, CA 91101-1292
ABSTRACT
We analyze the galactic H I content and nebular log(
O/H ) for 60 spiral galaxies inthe Moustakas et al. (2006) spectral catalog. After correcting for the mass-metallicityrelationship, we show that the spirals in cluster environments show a positive correla-tion for log(
O/H ) on DEF, the galactic H I deficiency parameter, extending the resultsof previous analyses of the Virgo and Pegasus I clusters. Additionally, we show forthe first time that galaxies in the field obey a similar dependence. The observed rela-tionship between H I deficiency and galactic metallicity resembles similar trends shownby cosmological simulations of galaxy formation including inflows and outflows. Theseresults indicate the previously observed metallicity-DEF correlation has a more univer-sal interpretation than simply a cluster’s effects on its member galaxies. Rather, weobserve in all environments the stochastic effects of metal-poor infall as minor mergersand accretion help to build giant spirals.
Subject headings: galaxies: abundances — galaxies: evolution — galaxies: spiral
1. Introduction
The relationship between galactic H I content and other properties of giant spiral galaxies is awell-documented phenomenon. Most notably, star formation (e.g. Kennicutt 1998; Rose et al. 2010)and gas-phase metal abundance (Skillman et al. 1996; Ellison et al. 2009; Zhang et al. 2009, amongothers) are known to be intimately connected to a galaxy’s overall H I mass. Many observationalstudies of the H I-metallicity relation interpret the phenomenon as a consequence of environment-driven evolution (namely ram pressure gas stripping or infall cutoff) through either cluster mem-bership (Skillman et al. 1996; Petropoulou et al. 2012) or local overdensity (Cooper et al. 2008;Ellison et al. 2009). On the other hand, cosmological hydrodynamical simulations (e.g. Dav´e et al.2011b) predict a dependence of galactic metallicity on H I content for galaxies in the field as well.In Robertson et al. (2012), we took the approach of the Skillman et al. (1996) analysis of Virgo,examining a single galaxy cluster–Pegasus I–to evaluate the effect of H I content on mean galacticmetallicity for giant spirals. Rather than bifurcate our sample into “gas-rich” and “gas-poor,”as had been done for Virgo, we quantified H I content using the DEF parameter described inSolanes et al. (1996), and considered galactic abundances as a function of this quantity. In the case 2 –of Pegasus I, DEF proved to be an excellent predictor of galactic log(
O/H ). Furthermore, the Virgogalaxies from the Skillman et al. (1996) study agreed nicely with the log(
O/H )-DEF correlationdespite the dramatically different properties (density, number of galaxies, velocity dispersion) ofthe two clusters.The most significant limitation of the observed relation between log(
O/H ) and DEF is the smallnumber of galaxies for which the dependence has been tested. Between the Virgo and Pegasus Iclusters, only 12 cluster galaxies were included in the Robertson et al. (2012) paper. Furthermore,while we included a small number of field spirals from the Zaritsky et al. (1994) sample, the numberof objects and the precision of their associated log(
O/H ) measurements made it impossible to con-clude whether our observed correlation extended to galaxies in the field. In this paper, we remedyboth of these shortcomings by utilizing galaxy-integrated spectra of 60 giant spirals (35 cluster,25 field) from the Moustakas & Kennicutt (2006a) catalog. Here, we show that the abundancesof these galaxies confirm the dependence of galactic log(
O/H ) on DEF for cluster spirals, andthat field spirals are subject to a similar relation, in agreement with cosmological hydrodynamicalsimulations.
2. Data
To expand on the results of Skillman et al. (1996) and Robertson et al. (2012, hereafter PaperI), we sought to obtain H I and metallicity measures for a large number of galaxies in a wide rangeof environments. Because our H I deficiency parameter DEF requires accurate 21 cm H I fluxesand morphological types, we were confined to relatively nearby galaxies. Also, since accuratelog(
O/H ) determinations for spiral galaxies are dependent on spectra covering the entire galacticdisk, very large surveys such as SDSS are unsuitable, as nearby spirals do not fit within a singlefiber.We found a suitable sample of objects in the Moustakas & Kennicutt (2006a) catalog of long-slit galactic spectra. The catalog contains emission-line spectra for 417 galaxies. While Paper Iand other similar studies determine galactic nebular metallicities by fitting abundance gradientsto spatially resolved H II region spectra, Moustakas & Kennicutt (2006b) show that the integratedspectra from these long-slit observations yield equivalent log(
O/H ) values. Taking advantage oftheir result, we derived galactic abundances from this catalog. First, though, we selected thegalaxies suitable for our log(
O/H )/DEF analysis according to the following criteria:I. We selected only objects for which H I 21cm flux measurements, optical diameters, andT-types exist in the Third Reference Catalog of Bright Galaxies (RC3 de Vaucouleurs et al. 1991).II. We eliminated any objects without significant detections of the [O II] λ β emissionlines.III. We selected only massive spirals, with T-types between 0 and 8. Additionally, we elimi-nated any objects known to be in interacting or merging pairs because of the difficulty of assign-ing morphological types to these galaxies, and because of the known metallicity dilution effects(Kewley et al. 2006; Ellison et al. 2008) for interacting pairs. Galaxies known to be in groups (notclusters) have also been eliminated due to their relatively limited number. 3 –After selecting for the above requirements, we are left with a sample of 60 spiral galaxies. Forthese objects, we first calculated the H I deficiency parameter DEF (Giovanelli & Haynes 1985).We computed DEF following Solanes et al. (1996), who define the quantity asDEF = log M H I ,exp − log M H Iwhere M H I ,exp is an expectation value for a galaxy’s H I mass based on its optical diameter andmorphological type. Since DEF is an underabundance relative to the expectation, more positivevalues represent lower H I content.As in Paper I, we used oxygen as a proxy for a galaxy’s heavy-element abundance, and usedthe strong-line R calibration for the [O II] and [O III] emission lines. To facilitate direct com-parison to Paper I, we have again used the Zaritsky et al. (1994) R calibration to compute12 + log( O/H ). Our error bars are obtained from standard propagation of the uncertainties givenfor the Moustakas & Kennicutt (2006a) emission lines.We categorized our galaxies as cluster, group, or field members using the associations listedin HyperLeda (Paturel et al. 2003). In cases where HyperLeda did not offer this information, weconsulted the SIMBAD and SDSS SkyServer Object Explorer databases, and references therein. Ifa galaxy was not listed as a group or cluster member in any available literature or database, weconsidered it a field galaxy.In Table 1, we list the names, DEFs, 12+log(
O/H ) values, and, where applicable, host clustersof the galaxies examined in this study. For cluster members, we have also included approximatesky-projected separations ρ C from the cluster center, using the coordinates and redshifts of clustercenters from Baiesi-Pillastrini et al. (1984), assuming H = 72 km s − Mpc − . The Table isseparated into cluster and field populations, as they will be presented in the following section.
3. Analysis
As in Paper I, we are interested in the dependence of log(
O/H ) on DEF for the galaxies inTable 1. In order to evaluate any functional dependence, it is important that our sample cover asatisfactory dynamical range in DEF. In Figure 1, we show a histogram of DEF for the galaxiesstudied herein. For comparison, we also indicate the DEFs sampled in Paper I. We see thatthese objects cover a broad range of H I deficiency, and include significantly more very high- andlow-DEF galaxies than the targets of Paper I and Skillman et al. (1996). We note that, whilethere are members of both cluster and field samples with very low DEFs, there are considerablymore cluster galaxies with positive DEF values. This is consistent with the results of Solanes et al.(2001) and Levy et al. (2007), among others, who show that the cluster environment drives galacticH I depletion.In order to properly understand the influence of H I content on galactic heavy-element abun-dance, we must first correct for the mass-metallicity relationship (MZR, Zaritsky et al. 1994;Tremonti et al. 2004). To ensure easy comparison to Paper I, we have again removed the ef-fect of the MZR by using inclination-corrected circular velocity as a proxy for galactic mass, andsubtracting the log(
O/H ) versus v C fit derived in Paper I: 4 –12 + log(O/H)= 8 .
57 + 0 . × V C / (200 km / s).We plot the residual log( O/H ) differential for each galaxy in Figure 2. In order to ensure thatour results are not dependent on our MZR correction, we also present the same data, corrected byinstead subtracting the log(
O/H ) versus M B relation from Paper I:12 + log(O/H)= 8 . − . × ( M B + 20).Note that we show our M B -corrected data as a consistency check, and base all of our for-mal conclusions on the v C -based MZR correction. This is because, as mentioned in Paper I andZaritsky et al. (1994), v C is independent of distance and unbiased by recent star formation.Because Paper I showed a clear correlation between oxygen content and DEF for cluster galax-ies, but was unable to confirm or reject that correlation for galaxies in the field, we examine thecluster and field galaxies separately. Considering first the subset of cluster galaxies (Figure 2(a)), we see that the greatly increasednumber of objects contains a considerable amount of scatter in comparison to the Virgo/Pegasussample of Paper I (Figure 6 in that paper). Evaluating the dependence of log(
O/H ) on DEFtherefore requires a careful statistical analysis.As mentioned in the previous section, uncertainties on galactic log(
O/H ) are obtained in astraightforward manner from the errors on the line fluxes. However, understanding the uncertaintyon DEF is considerably more complicated. Because the calcluation of DEF relies on T-type andoptical diameter in addition to 21 cm flux, uncertainties in all of those parameters contribute tothe overall error budget. Additionally, since DEF is calibrated to a finite sample of field galaxies(Solanes et al. 1996), the calculation of expected H I mass is not exact. Rather than assign individ-ual errors to each object, we chose instead to adopt a uniform error σ DEF = 0 .
15 for all galaxies,based on the recommendation of Levy et al. (2007), who estimate a “cosmic scatter” of 0.15 inDEF. Our derived dependencies on DEF will therefore have relatively conservative error estimates,since purely statistical error calculations would result in smaller uncertainties.We began our analysis with a standard linear regression on the cluster subset. Although weexperimented with a number of weighting schemes, because the uncertainties in our data onlydiffer in the estimates of log(
O/H ), which does not by itself dominate the error budget, each of ourweighted fits resulted in unreasonably small errors on the resulting slopes and intercepts. For thisreason, all least-squares fits presented herein are computed with equal weights for all data points.With an ordinary least squares (OLS) fit, our model islog(
O/H ) res = 0 . ± . + 0 . ± . × DEF (1)where log(
O/H ) res is the measured abundance after subtracting our MZR fit.Because our fitted slope is only ∼ σ away from zero, we also performed a Pearson correlationtest on the data in our cluster sample to confirm the statistical significance of the relation between 5 –log( O/H ) and DEF. The correlation coefficient for the cluster galaxies is 0.55. For a sample size of35 galaxies, this coefficient indicates the probability of no correlation is just 0.0006. We see, then,that there is a significant correlation between log(
O/H ) and DEF, and the slope of the relation isconsistent with that found in Paper I.To obtain a better estimate of the actual functional relationship between log(
O/H ) and DEF,we have performed a more statistically rigorous linear fit to the data using the maximum likelihood(MLE) method outlined by Kelly (2007). The code works by creating a likelihood function for thetrue distribution of regression parameters, based on the observed data and errors. The regressioncoefficients and errors are estimated by performing Bayesian inference using 10,000 MCMC samplesof the parameter space, where each chain performs a random walk through the parameter space(using a Gibbs sampler), eventually converging on the posterior distribution. The values of the slopeand intercept to which each chain converges represents a single random draw from the posteriordistribution. Fitting a gaussian to the resulting distribution of slopes, and extracting the meanand FWHM, the resulting “mean fit” to the cluster subset becomeslog(
O/H ) res = 0 . ± . + 0 . ± . × DEF (2)We have included both of the fits above in Figure 2(a). Since the MLE routine allowed usto include uncertainties on both DEF and log(
O/H ), we adopt the MLE fit as our final model.However, it is worth pointing out the agreement between the OLS and MLE fits for the clustersubset, suggesting OLS is actually adequate in this case.When considering our sample using the M B -based MZR, we find results consistent with ourprimary MZR correction. We obtain slopes of 0 . ± .
08 (OLS) and 0 . ± .
13 (MLE), whichagree with the fits above.In addition to being internally consistent, our fits to the log(
O/H ) versus DEF relation, alsoagree with our fits to the Virgo and Pegasus spirals derived in Paper I. For comparison, we haveincluded these fits in Figures 2(a) and 2(b).
Having recovered the log(
O/H ) versus DEF relationship discovered in Paper I, we revisited thequestion of whether the same dependence exists for galaxies in the field. Our field galaxy sampleis plotted in Figure 2(c). Again, there is plenty of scatter, but a positive trend is visible. We againperformed a Pearson correlation test to the field sample, acquiring a correlation coefficient of 0.58.For 25 galaxies, our correlation coefficient gives the probability of no correlation at P = 0 . O/H ) res = 0 . ± . + 0 . ± . × DEF (3)To properly quantify the relationship using our errors on DEF, we again calculate the MLEfit, giving a final model 6 –log(
O/H ) res = 0 . ± . + 0 . ± . × DEF (4)We note that the OLS and MLE fits for field galaxies all agree to within 1 σ regardless of whichMZR correction we use. However, it is worth noting that the MLE fit to the M B -corrected data(Figure 2(d) results in a very steep slope of 0 . ± .
4. We find that when we exclude NGC 4605,which is very metal-poor for its DEF value (0.34), the OLS fit also displays a much higher slope.Since our v C MZR correction places NGC 4605 in better agreement with the observed trend, wedo not exclude it as an outlier.Since the Moustakas & Kennicutt (2006a) selection of spiral galaxies is not a volume-limitedsample, it is prudent to consider whether our observed trends in log(
O/H ) versus DEF couldbe produced by an observational bias. The RC3 catalog is essentially complete for galaxies withoptical diameters greater than 1 arcminute and total B magnitudes brigher than 15.5. Althoughthe surface brightness cutoff may lead to the omission of some edge-on spirals, such a bias shouldnot have a significant influence on our result, as the uncertainty in morphological type and increasein interstellar reddening complicate the determinations of DEF and log( O/H ), respectively. InPaper I, we intentionally avoided edge-on spirals for this reason.As for the Moustakas & Kennicutt (2006a) selection, which is described in Moustakas & Kennicutt(2006b) while it is neither blind nor complete, the galaxies included cover a wide range in M B , B − V , and morphological type. We therefore expect a representative sampling of different galaxymasses, star formation histories, and dust content.In order for a bias to create such an effect, we would somehow have to systematically excludeH I-rich galaxies with high metallicity and/or H I-poor galaxies with low metallicities. We believeboth possibilities are very unlikely. Galaxies with low DEF (high H I content) should producestrong 21 cm radiation, and will also likely have relatively high specific star formation rates (e.g.Rose et al. 2010), leading to strong H β lines. Therefore, low-DEF galaxies should not be excludedfrom our selection at any metallicity, according to our criteria listed in the previous section. As forhigh DEF/low metallicity spirals, their low metal content should lead to strong nebular emissionlines via higher temperatures, ensuring their inclusion from the Moustakas & Kennicutt (2006a)catalog. Furthermore, even our highest-DEF galaxies represent 21 cm detections well above the100 σ level, so we are not excluding any high-DEF galaxies due to nondetections of H I emission.We are therefore confident that our result is not due to an observational bias, despite the fact thatour sample was not specifically chosen to be completely unbiased.
4. Discussion4.1. Comparison to Previous Observations
The slope of the DEF-log(
O/H ) relation for cluster galaxies (0 . ± .
15) remains in goodagreement with the slope derived in Paper I (0 . ± .
1) upon increasing the number of galaxiesin our sample by a factor of 3. While the scatter around the fit in Figure 2(a) is higher thanseen in the Virgo and Pegasus samples, it is important to consider the differences between thegalaxies examined in the two studies. The Virgo and Pegasus galaxies selected by Skillman et al. 7 –(1996) and Paper I were chosen for their abundance of bright individual H II regions, and werealso all nearly face-on. Furthermore, the Virgo/Pegasus galaxies were all very similar in massand luminosity. In this larger sample, there is certainly scatter introduced by inclination effects,ambiguous morphological types, and imperfect mass correction which was largely inconsequentialin the smaller, more homogeneous earlier data sets. In this sense, the Virgo/Pegasus galaxies canbe interpreted as the “ideal case,” and our analysis of the Moustakas & Kennicutt (2006a) sampleextends the preliminary results to a much broader group of objects.Considering most studies of the interrelation between galactic gas content and metallicity (e.g.Skillman et al. 1996; Ellison et al. 2009; Petropoulou et al. 2012) have examined H I deficiencyin the context of cluster environment or local galactic density, it is somewhat surprising to findthat the observed metallicity dependence extends to field galaxies. In fact, our measured slopefor the field subset is actually higher than for the cluster galaxies, although it is doubtful thatdifference is significant. While the distributions of the slopes in our Monte Carlo resamplings aredifferent, given the uncertainties on our fits, we are not confident that the difference in the observedslopes is significant over the range of DEF explored here. We therefore conclude that, within theuncertainties, the log(
O/H ) versus DEF relation applies generally to any non-interacting massivespiral galaxy in a similar way, regardless of environment.
Modern cosmological hydrodynamic simulations can predict the neutral hydrogen and oxygencontent for representative samples of galaxies. Here we compare our DEF-log(
O/H ) res results tothe simulations of Dav´e et al. (2011b). Since these simulations have a box length of 48 h − Mpc ona side that does not contain any cluster-sized objects, their simulated sample is most appropriatelycompared to our field sample. Similar to Dav´e et al. (2011b), we have excluded any galaxies withstellar masses lower than M ∗ = 2 × M ⊙ . While the sample does not include morphological data,the masses, star formation rates, and gas fractions of our simulated galaxies are a good match tothe selection of Moustakas & Kennicutt (2006a), who note in Moustakas & Kennicutt (2006b) thattheir observed galaxies are largely late-type (Sbc and later) spirals.In these models, we compute the deviations in metal and H I content at a given specificstar formation rate (sSFR ≡ SFR /M ∗ ). This is different than our treatment of the observations,where DEF is defined based on the expected HI content of galaxies with similar morphology andsize. Unfortunately, these simulations lack the resolution to predict these parameters, and hencewe must choose a proxy from among the available model-predicted quantities. We choose sSFRbecause Rose et al. (2010) showed that DEF is most tightly correlated with sSFR, as opposedto SFR or M ∗ alone. To verify this approach is qualitatively valid, we have examined 19 spiralgalaxies with measured sSFRs from Howell et al. (2010). Taking T-types, 21cm fluxes, and opticaldiameters from HyperLeda, we computed DEF via the Solanes et al. (1996) formulae. We thenestimated H I deficiency by correcting for a trend in M H I versus sSFR, and adopting DEF as thevertical offset from this trend. We show a comparison between the two estimates of DEF in Figure3. Performing a linear fit to the data, we find a slope of 0 . ± .
19, but also a vertical offset of0 . ± .
09, indicating a systematic difference between the two calculations. Therefore, while weare confident that the two estimates of DEF reflect qualitatively similar trends, the offset prevents 8 –us from directly equating the simulations and our data.Figure 4 shows the correlation between DEF and log(
O/H ) res from the momentum-drivenwind scaling simulation of Dav´e et al. (2011b), defined relative to the mean at a given sSFR. Thegalaxy metallicities are computed as described in Dav´e et al. (2011b), while the H I mass accountsfor self-shielding and conversion to molecular hydrogen as described in Dav´e et al. (2013), broadlyfollowing Popping et al. (2009) and Duffy et al. (2012). The green line shows the best-fit powerlaw to these points, which follows the relation DEF= 0 .
07 + 0 . × [ O/H ] res . Also shown in theFigure is the best linear fit to the constant wind model (red line, see Dav´e et al. 2011b, for detailson the constant wind model).The predicted slope is close to that observed (0.41), and the results display a similar amount ofscatter around the fit. The predicted amplitude is slightly low, but likely within uncertainties giventhe different way in which DEF is computed between the models and the data. We note that had wechosen stellar mass rather than sSFR about which to measure our deviations, the predicted slopewould be shallower, namely 0.27, but still within 1 σ of that observed. Hence the trend in DEF vs. [ O/H ] res appears to be a relatively robust prediction of hierarchical galaxy formation simulations,regardless of the details of feedback .Why do these hierarchical models predict such a trend? The physical origin can be explainedby appealing to the equilibrium model of galaxy evolution (Dav´e et al. 2012). In this scenario,galaxies live in a slowly-evolving balance between accretion, outflows, and star formation. Thisresults in preferred equilibrium relations for the main physical properties of galaxies, such as tightrelations between stellar mass, star formation rate (Dav´e 2008), metallicity (Finlator & Dav´e 2008),and H I content (Popping et al. 2009).Galaxies are perturbed off these equilibrium relations owing to the stochasticity in accretion(e.g. mergers), which governs the scatter around these relations (Finlator & Dav´e 2008). Considera galaxy undergoing a merger with a smaller system. Its metallicity will go down because thesmaller system will tend to have lower metallicity. However, its H I content will rise since smallersystems tend to be more H I-rich. Hence deviations towards low metallicity will be correlated withdeviations towards high H I content. The converse can also happen, where a galaxy experiences alull in accretion (or a dimunition owing to it becoming a satellite in a larger halo), in which caseit will consume its available gas reservoir, increase its metallicity, and lower its H I content. It isstraightforward to see that such perturbations will produce a trend in DEF vs. [ O/H ] res that isqualitatively as observed. Furthermore, the fact that an upward trend exists regardless of windmodel indicates the behavior does not arise as the result of an outflow effect, but rather it appearsbecause of inflow stochasticity, which is independent of outflows.We attribute the slope of the DEF vs. [ O/H ] res relation in the simulations primarily to threephysical phenomena: First, it reflects the characteristic spectrum of mergers and smooth accretionthat drive perturbations off the equilibrium relations. Second, it reflects the tendency of minormergers to enrich spirals with metal-poor gas, decreasing the global nebular metallicity. Finally,it reflects the trend of H I richness vs. sSFR, which analogously sets the typical deviation inH I content when a giant spiral merges with a gas-rich dwarf.The agreement between the models and the data suggests that the simulations are properlycapturing these phenomena. As shown in Dav´e et al. (2011b), this model produces roughly the 9 –correct mass-metallicity relation. In Dav´e et al. (2013) we show that it also broadly matches theobserved mass-H I richness relation. The spectrum of mergers is set by the underlying cosmology,which is assumed to be WMAP7-concordant. Given that all the individual pieces in the modelagree with data, it is perhaps not surprising that the DEF vs. [ O/H ] res is also reproduced. Also,since the constant and momentum-driven wind models yield qualitatively similar trends (Figure 4)for metallicity and H I richness, it is also not surprising that our simulation results are not stronglysensitive to the assumed feedback model.We see that this scenario for what sets the H I deficiency in galaxies offers a mechanism toproduce the log( O/H )-DEF relation separate from traditional scenarios that have posited that itarises from environmentally-driven processes such as ram pressure stripping (e.g. Gunn & Gott1972). Naively, such scenarios would predict that the trends would be stronger in clusters, but ourobservations suggest that the trend of DEF vs. [
O/H ] res is similar in the field. In the simulations,environment does not play a large role (except for satellite galaxies; Dav´e et al. 2011a). Instead,DEF is simply set by the stochastic nature of hierarchical accretion, and galaxies’ response to suchstochasticity generically yields the observed trend in DEF vs. [ O/H ] res . We caution that thesesimulations only produce field galaxies, so environmental processes may still play a major role inextreme environments such as clusters. But the success of these models suggests that at least fortypical field galaxies, it is not necessary to appeal to environmental processes in order to understandthe behavior of [ O/H ] res vs. DEF. Evidently, changes to a galaxy’s nebular metallicity caused byvarying H I content are to some degree insensitive to the specific physical processes (i.e. infall,minor mergers, ram-pressure stripping) responsible for regulating H I richness.
5. Conclusion
Using the spectral library of Moustakas & Kennicutt (2006a), we have conducted an expandedinvestigation into the influence of H I abundance on galactic nebular metallicity analogous tothe analysis of Robertson et al. (2012) for the Pegasus cluster. We have compared these results topredictions based on cosmological hydrodynamical simulations. Our conclusions can be summarizedin three main results: For galaxies in clusters, we recover the previously observed trend of increasing log(
O/H ) withdecreasing H I content. For galaxies in the field, log(
O/H ) is similarly dependent on H I deficiency. Our hydrodynamical simulations for field galaxies predict a metallicity-DEF correlationsimilar to that observed. We interpret this result as the product of a galaxy’s natural “excursions”between H I-rich/metal-poor and H I-poor/metal-rich in response to stochastic fluctuations in theinflow rate. These departures from equilibrium with respect to the mass-metallicity relation canoccur in any environment, and do not require cluster membership or enhanced local galaxy density.We thank the anonymous referee for valuable comments. P. R. is supported by a Universityof Texas Continuing Fellowship. G.S. gratefully acknowledges the support of the Jane and RolandBlumberg Centennial Professorship in Astronomy. This research has made use of the SIMBAD 10 –database, operated at CDS, Strasbourg, France. We acknowledge the usage of the HyperLedadatabase (http://leda.univ-lyon1.fr). 11 –
REFERENCES
Baiesi-Pillastrini, G. C., Palumbo, G. G. C., & Vettolani, G. 1984, A&AS, 56, 363Cooper, M. C., Tremonti, C. A., Newman, J. A. & Zabludoff, A. I. 2008, MNRAS, 390, 245Dav´e, R. 2008, MNRAS, 385, 147Dav´e, R., Oppenheimer, B. D., & Finlator, K. 2011, MNRAS, 415, 11Dav´e, R., Finlator, K., & Oppenheimer, B. D. 2011, MNRAS, 416, 1354Dav´e, R., Finlator, K. & Oppenheimer, B. D. 2012, MNRAS, 421, 98Dav´e, R., Katz, N., Oppenheimer, B. D., Kollmeier, J. A., & Weinberg, D. H. 2013, arXiv:1302.3631de Vaucouleurs, G., de Vaucouleurs, A., Corwin, H. G., Jr., et al. 1991, Third Reference Catalogueof Bright Galaxies. Volume I: Explanations and references. Volume II: Data for galaxiesbetween 0 h and 12 h . Volume III: Data for galaxies between 12 h and 24 h ., by de Vaucouleurs,G.; de Vaucouleurs, A.; Corwin, H. G., Jr.; Buta, R. J.; Paturel, G.; Fouqu´e, P.. Springer,New York, NY (USA), 1991, 2091 p., ISBN 0-387-97552-7, Price US$ 198.00. ISBN 3-540-97552-7, Price DM 448.00. ISBN 0-387-97549-7 (Vol. I), ISBN 0-387-97550-0 (Vol. II), ISBN0-387-97551-9 (Vol. III).Duffy, A. R., Kay, S. T., Battye, R. A. et al. 2012, MNRAS, 420, 2799Ellison, S. L., Patton, D. R., Simard, L. & McConnachie, A. W. 2008, AJ, 135, 1877Ellison, S. L., Simard, L., Cowan, N. B. et al. 2009, MNRAS, 396, 1257Finlator, K. & Dav´e, R. 2008, MNRAS, 385, 2181Howell, J. H., Armus, L., Mazzarella, J. M., et al. 2010, ApJ, 715, 572Giovanelli, R., & Haynes, M. P. 1985, ApJ, 292, 404Gunn, J. E., & Gott, J. R. 1972, ApJ, 176, 1Isobe, T., Feigelson, E. D., Akritas, M. G., & Babu, G. J. 1990, ApJ, 364, 104Kelly, B. C. 2007, ApJ, 665, 1489Kennicutt, R. C., Jr. 1998, ApJ, 498, 541Kewley, L. J., Geller, M. J., & Barton, E. J. 2006, AJ, 131, 2004Levy, L., Rose, J. A., van Gorkom, J. H. & Chaboyer, B. 2007, AJ, 133, 1104Moustakas, J. & Kennicutt, R. C., Jr. 2006, ApJS, 164, 81Moustakas, J. & Kennicutt, R. C., Jr. 2006, ApJ, 651, 155Paturel, G., Petit, C., Prugniel, Ph. et al. 2003, A&A, 412, 45Petropoulou, V., V´ılchez, J., & Iglesias-P´aramo, J. 2012, ApJ, 749, 133 12 –Popping, A., Dav´e, R., Braun, R. & Oppenheimer, B. D. 2009, A&A, 504, 15Robertson, P., Shields, G. A., & Blanc, G. A. 2012, ApJ, 748, 48Rose, J. A., Robertson, P., Miner, J. & Levy, L. 2010, AJ, 139, 765Skillman, E. D., Kennicutt, R. C., Jr., Shields, G. A., & Zaritsky, D. 1996, ApJ, 462, 147Solanes, J. M., Giovanelli, R., & Haynes, M. P. 1996, ApJ, 461, 609Solanes, J. M., Manrique, A., Garc´ıa-G´omez, C., et al. 2001, ApJ, 548, 97Tremonti, C. A. et al. 2004, ApJ, 613, 898Zaritsky, D., Kennicutt, R. C., Jr., & Huchra, J. P. 1994, ApJ, 420, 87Zhang, W., Li, C., Kauffmann, G. et al. 2009, MNRAS, 397, 1243 This preprint was prepared with the AAS L A TEX macros v5.2.
13 –Fig. 1.— Distributions of the H I deficiency parameter DEF for our cluster (top) and field (bottom)selections of galaxies from the Moustakas & Kennicutt (2006a) catalog. The open bins show theDEF distributions from Paper I. 14 – -0.5 0 0.5 1DEF-0.4-0.200.20.40.6 ∆ [ O / H ] Ordinary Least-Squares (OLS)Maximum Likelihood (MLE)Fit to Virgo/Pegasus Data (Paper I)Cluster, v c MZR correction (a) -0.5 0 0.5 1DEF-0.4-0.200.20.4 ∆ [ O / H ] Ordinary Least-Squares (OLS)Maximum LikelihoodFit to Virgo/Pegasus Data (Paper I)Cluster, M B MZR correction (b) -0.6 -0.4 -0.2 0 0.2DEF-0.4-0.3-0.2-0.100.10.20.30.40.5 ∆ [ O / H ] Ordinary Least Squares (OLS)Maximum Likelihood (MLE)Field, v c MZR correction (c) -0.6 -0.4 -0.2 0 0.2 0.4DEF-0.3-0.2-0.100.10.20.3 ∆ [ O / H ] Ordinary Least-Squares (OLS)Maximum LikelihoodField, M B MZR correction (d)
Fig. 2.— Residual log(
O/H ) after subtracting the mass-metallicity relationship (MZR) for ourselected galaxies, plotted as a function of DEF. Our sample is separated into [a,b] cluster and [c,d]field galaxies. Plots on the left [a,c] have been corrected for the MZR using circular velocity, whileplots on the right [b,d] use absolute blue magnitude instead (see text for details). For each subsetof galaxies, we have included linear fits according to ordinary least squares (red) and maximumlikelihood (blue). For the cluster galaxies, we have also included our ordinary least squares fit tothe Virgo/Pegasus data from Paper I (dashed green line). 15 – -0.8 -0.6 -0.4 -0.2 0DEF (via Solanes et al. 1996)-0.4-0.200.20.40.6 D E F ( v i a s SF R ) Fig. 3.— We compare DEF as measured by the (Solanes et al. 1996) method and by using ourestimate relative to a given sSFR. The red line gives the best fit to the data, while the dotted blueline indicates the line y = x . 16 –Fig. 4.— Residual [O/H] as a function of DEF for galaxies from our hydrodynamical simulation.The green line gives our best fit to the relation, while the red line represents the best fit to thesame galaxies with a constant wind model (see text). We note that DEF in this figure is computedrelative to a “normal” H I content at fixed sSFR to account for a lack of morphological informationin our simulations. 17 –Table 1: Galaxy data from Moustakas & Kennicutt (2006a). DEF has been computed according toSolanes et al. (1996), and 12 + log( O/H ) is calibrated using the method outlined in Zaritsky et al.(1994). v C values are taken from the HyperLeda database, and are corrected for inclination. Whereappropriate, UGCl cluster listings have been replaced with their more familiar names according toBaiesi-Pillastrini et al. (1984). Galaxy Name DEF 12 + log(
O/H ) v C Cluster ρ C (kpc) Cluster Galaxies
NGC 0660 − .
22 8 . ± .
19 140.82 UGCl 029 1650UGC 01281 0 .
30 8 . ± .
16 50.11 UGCl 032 2340UGC 01385 − .
29 9 . ± .
04 227.69 Abell 262 513NGC 0784 0 .
34 8 . ± .
11 41.31 UGCl 032 3700NGC 0877 − .
58 9 . ± .
05 272.82 UGCl 035 2090NGC 0976 − .
39 9 . ± .
09 400.9 UGCl 038 6840NGC 0972 0 .
22 9 . ± .
05 145.99 UGCl 038 3640NGC 1003 − .
28 8 . ± .
08 95.49 Perseus 6580NGC 1058 − .
44 9 . ± .
06 13.27 Perseus 6750NGC 1087 − .
01 9 . ± .
04 120.27 UGCl 043 937NGC 1345 − .
18 8 . ± .
13 97.19 Eridanus 369NGC 2893 0 .
01 9 . ± .
06 109.36 UGCl 148 505NGC 3079 0 .
02 8 . ± .
11 208.39 UGCl 163 22100NGC 3310 − .
39 8 . ± .
07 288.38 UGCl 163 9170NGC 3353 0 .
05 8 . ± .
09 57.16 UGCl 189 167NGC 3504 0 .
49 9 . ± .
04 194.09 Abell 1185 1590UGC 06665 − .
69 8 . ± .
09 114.58 UGCl 231 3660NGC 3913 0 .
20 8 . ± .
20 34.07 UGCl 229 3260NGC 3953 0 .
47 9 . ± .
10 215.86 UGCl 229 9490NGC 3972 0 .
67 9 . ± .
18 114.36 UGCl 229 3830NGC 3982 0 .
11 9 . ± .
04 191.83 UGCl 229 4290NGC 4062 0 .
28 9 . ± .
11 140.47 UGCl 263 8531NGC 4085 0 .
11 9 . ± .
04 127.84 UGCl 229 14600NGC 4088 − .
08 9 . ± .
05 167.29 UGCl 229 14200NGC 4102 0 .
48 9 . ± .
09 158.14 UGCl 229 10100NGC 4136 0 .
01 8 . ± .
17 101.3 UGCl 263 3880NGC 4157 − .
03 9 . ± .
12 188.89 UGCl 229 15000NGC 4288 − .
45 8 . ± .
10 114.37 UGCl 265 388NGC 4389 0 .
83 9 . ± .
04 95.47 UGCl 265 47.1NGC 4414 − .
24 9 . ± .
04 217.83 UGCl 267 19.2NGC 5014 0 .
25 9 . ± .
07 85.29 UGCl 281 2530NGC 6052 − .
58 8 . ± .
07 293.49 Hercules 4620NGC 7518 0 .
07 9 . ± .
11 35.61 Pegasus 2590NGC 7591 − .
49 9 . ± .
10 211.21 Pegasus 2480NGC 7625 − .
42 9 . ± .
05 285.57 UGCl 486 1880NGC 7678 − .
05 9 . ± .
04 198.3 Abell 2657 13100
Field Galaxies
NGC 0095 − .
34 8 . ± .
07 203.78NGC 0157 − .
34 9 . ± .
04 154.42
18 –
Table 1 cont’d.
Galaxy Name DEF 12 + log(
O/H ) v C Cluster ρ C (kpc)NGC 0278 − .
05 9 . ± .
03 256.28NGC 0922 − .
42 8 . ± .
08 178.59NGC 1421 − .
26 8 . ± .
07 161.59NGC 2139 − .
54 8 . ± .
08 135.61NGC 2782 − .
05 8 . ± .
06 116.73NGC 2903 0 .
08 9 . ± .
07 186.95NGC 3049 − .
05 9 . ± .
09 102.61NGC 3198 − .
14 8 . ± .
14 142.51NGC 3274 − .
60 8 . ± .
09 79.62NGC 3344 − .
07 9 . ± .
08 222.87NGC 3521 − .
18 9 . ± .
06 244.92NGC 3600 − .
28 8 . ± .
12 86.9NGC 4384 − .
10 9 . ± .
05 102.29NGC 4455 − .
14 8 . ± .
10 56.98NGC 4605 0 .
34 8 . ± .
07 60.87NGC 4670 − .
04 8 . ± .
08 140.4NGC 5104 − .
43 8 . ± .
19 203.18NGC 6207 − .
11 8 . ± .
05 114.83NGC 7137 − .
04 9 . ± .
04 104.46NGC 7620 − .
50 9 . ± .
05 423.69NGC 7624 − .
16 9 . ± .
07 173.88NGC 7640 − .
08 8 . ± .
09 107.92NGC 7742 0 .
04 9 . ± ..