Constraining the age of the NGC 4565 HI Disk Warp: Determining the Origin of Gas Warps
David J. Radburn-Smith, Roelof S. de Jong, David Streich, Eric F. Bell, Julianne J. Dalcanton, Andrew E. Dolphin, Adrienne M. Stilp, Antonela Monachesi, Benne W. Holwerda, Jeremy Bailin
TT HE A STROPHYSICAL J OURNAL
Preprint typeset using L A TEX style emulateapj v. 5/2/11
CONSTRAINING THE AGE OF THE NGC 4565 H I DISK WARP: DETERMINING THE ORIGIN OF GAS WARPS D AVID
J. R
ADBURN -S MITH , R OELOF S. DE J ONG , D AVID S TREICH , E RIC
F. B
ELL , J ULIANNE
J. D
ALCANTON ,A NDREW
E. D
OLPHIN , A DRIENNE
M. S
TILP , A NTONELA M ONACHESI , B ENNE
W. H
OLWERDA , AND J EREMY B AILIN Department of Astronomy, University of Washington, Seattle, WA 98195, USA Leibniz-Institut für Astrophysik Potsdam, D-14482 Potsdam, Germany Department of Astronomy, University of Michigan, Ann Arbor, MI 48109, USA Raytheon, 1151 East Hermans Road, Tucson, AZ 85756, USA European Space Agency, ESTEC, 2200 AG Noordwijk, The Netherlands and Department of Physics and Astronomy, University of Alabama, Tuscaloosa, AL 35487, USA
Received 2013 July 30; accepted 2013 November 13
ABSTRACTWe have mapped the distribution of young and old stars in the gaseous H I warp of NGC 4565. We find a clearcorrelation of young stars ( <
600 Myr) with the warp, but no coincident old stars ( > ∼
300 Myr agorelative to the surrounding regions, is (6 . + . − . ) × − M (cid:12) yr − kpc − . This implies a ∼ ±
20 Gyr depletiontime of the H I warp, similar to the timescales calculated for the outer H I disks of nearby spiral galaxies. Whilesome stars associated with the warp fall into the asymptotic giant branch (AGB) region of the color magnitudediagram, where stars could be as old as 1 Gyr, further investigation suggests that they may be interlopers ratherthan real AGB stars. We discuss the implications of these age constraints for the formation of H I warps, andthe gas fueling of disk galaxies. Subject headings: galaxies: formation – galaxies: individual (NGC 4565) - galaxies: spiral – galaxies: stellarcontent – galaxies: structure – techniques: photometric INTRODUCTION
The neutral hydrogen (H I ) gaseous disks of many nearbydisk galaxies are warped. From a sample of 26 edge-on galax-ies, García-Ruiz et al. (2002b) postulated that all H I disks thatextend beyond their optical counterpart are warped. Unlikestellar warps, the onset of the H I warp is often abrupt, dis-continuous, and in edge-on systems coincides with the breakin the exponential disk profile (Briggs 1990; van der Kruit2007). However, the mechanisms forming these distortionsare still unclear (for review, see van der Kruit & Freeman2011).One possible explanation is that the warp is a manifestationof ongoing gas accretion (e.g., Macciò et al. 2006; Spavoneet al. 2010). Many galaxies show possible signs of accre-tion in their H I outskirts in the form of bridges to compan-ions, asymmetric density or velocity fields, and gas cloudswith anomalous velocities (for review see Sancisi et al. 2008).However, the combined contribution of these phenomena isunable to support the current star formation rates (SFRs)found in the local universe (Tinsley 1980). Indeed, providingsufficient gas to fuel star formation (SF) remains a key issuein Λ CDM cosmology. Recent theoretical work suggests thatfor galaxies up to L * in size, this gas is primarily accretedas cold, unshocked gas (e.g., Dekel & Birnboim 2006). Forlarger systems, the quiescent cooling of shock-heated gas af-ter the last major merger plays the dominant role (e.g., Brooket al. 2004; Robertson et al. 2006). At early times ( z > Although, as noted by Hanish et al. (2006), such estimates of star-formation-rate densities carry large systematic uncertainties. in galaxy formation and evolution (e.g., Ostriker & Binney1989; Binney 1992; Jiang & Binney 1999).However, disk warps may be formed by mechanisms unre-lated to accretion (e.g., Binney 1978; Hunter & Toomre 1969).For instance, dynamical studies have also reproduced warpspurely through gravitational effects. These torques may bedue to a misalignment between the angular momentum of thedisk and dark-matter halo (Debattista & Sellwood 1999) or bytidal interactions with nearby galaxies (e.g., Hunter & Toomre1969), satellites (e.g., Weinberg & Blitz 2006), or dark-mattersubstructure (Kazantzidis et al. 2008). Non-gravitationaltorques are also possible, e.g., from passage through the in-tergalactic medium (Kahn & Woltjer 1959) or intergalacticmagnetic fields (Battaner et al. 1990). Although, such mag-netohydrodynamical processes may not be strong enough tocause the observed warp angles (Binney 1992, 2000).Placing constraints on the age of H I warps, and so de-termining how long they remain coherent, may help deter-mine their origin (e.g., Roškar et al. 2010). Hence, we needto identify the associated stellar populations, which will en-code a record of the formation of the structure. SF beyondthe optical disk was first inferred in the outer disks of nearbygalaxies via the detection of faint H II regions (Ferguson et al.1998a,b). This extreme SF was subsequently confirmed byobservations of outer-disk UV emission (Thilker et al. 2007).However, as discussed in de Jong (2008), the inherent degen-eracies between stellar age, metallicity, and reddening due todust preclude a meaningful measurement of the stellar contentof these outer regions from integrated-light studies. Studies ofresolved stellar populations mitigate many of these problemsand allow us to probe the extremely low stellar densities foundin the outer envelopes of nearby galaxies (e.g., Radburn-Smithet al. 2012).In this paper, we present the first study of the resolved stel-lar populations located within a distinct H I warp. We use a r X i v : . [ a s t r o - ph . GA ] D ec Radburn-Smith et al.
Figure 1.
Distribution of neutral hydrogen (H I , Dahlem et al. 2005), shownas ten logarithmically spaced contours between column densities of 2 × and 2 × cm − , overlaid on a IIIaJ (blue) POSS-II image of NGC 4565.The location of the ACS field is shown as a dashed red box on top of thesignificant northwestern warp. The companion systems NGC 4562 and thesmaller IC 3571 are seen at (R.A.,decl.) ∼ (12 h m . s , + d m s )and (12 h m . s , + d m s ), respectively. Part of a faint H I bridgethat connects NGC 4565 and IC 3571 is identified and extraplanar gas at ∼ (12 h m s , 26 ◦ I cloud. imaging from the Hubble Space Telescope ( HST ) AdvancedCamera for Surveys (ACS) of NGC 4565. This edge-on Sbgalaxy resides at a distance of 11.9 ± I warps before summarizing our findings in Section 7. THE WARP OF NGC 4565
As first noted by Sancisi (1976), the H I disk of NGC 4565is strongly warped, most notably on the northwestern (NW)side of the galaxy. Rupen (1991) calculated the gas massof the NW component of the warp to be ∼ × M (cid:12) (c.f.7 . × M (cid:12) for the entire NGC 4565 system, Heald et al.2011). However, with evidence for a mild warp along the lineof sight, Rupen (1991) suggested that the full warp is con-tinuous around the edge of the disk and encompasses a totalmass of 1 . × M (cid:12) , or about ∼ −
20% of the H I massof the entire system. Rupen (1991) found this structure to beasymmetric, in that the upward and downward bending modesof the warp are not separated by 180 ◦ in azimuth but instead ∼ ◦ . The authors note, however, that this apparent lopsid-edness may be due to a lack of H I in the southern componentof the warp rather than an inherent asymmetry. In Figure 1,the H I data from Dahlem et al. (2005) is overlaid as contourson a IIIaJ (blue) plate scan from the Second Palomar Obser-vatory Sky Survey (POSS-II; Reid et al. 1991).In order to present lower surface-brightness features, wehave reprocessed the H I data, following a similar procedure toWalter et al. (2008). We first created a post-imaging mask byconvolving the data cube to a 30 (cid:48)(cid:48) × (cid:48)(cid:48) resolution and select- ing all pixels with flux densities greater than twice the noisein the convolved cube in three consecutive velocity channels.We then edited this mask by hand to remove spurious signalssuch as sidelobes, noise spikes, and other artifacts. The finalmask was applied to the original unconvolved data and a newzeroth moment map was generated by summing the cleanedvelocity channels.Clearly evident in Figure 1 is the abrupt upturn ( ∼ ◦ ) ofthe NW warp. This structure begins at ∼ . (cid:48) I gas (Rupen 1991). The distinctmorphologies of this warp gas and the in-plane gas led van derKruit (2007) to conclude that the two components are distinct,with the warp representing an accreting component. As evi-dent in the contours of Figure 1, the warp appears to turnoverand realign with the gas disk at greater distances, as first notedby Sancisi (1983).In the optical, the upturn of the NW gaseous warp corre-sponds with a truncation of the main stellar disk (van der Kruit1979; Wu et al. 2002). Internal to this truncation, a mild warpis seen in the optical (van der Kruit 1979; van der Kruit &Searle 1981). Naeslund & Joersaeter (1997) have also arguedfor a possible extension of the optical disk beyond the trun-cation that coincides with the NW H I warp. In their V -bandobservations with the 2.56 m Nordic Optical Telescope, theymeasured a surface brightness of µ v ≈
27 mag arcsec − forthis optical structure. However, they could not rule out thisfeature as a background galaxy.Also evident in Figure 1 are the two companion systems. Tothe southwest is NGC 4562, for which Heald et al. (2011) finda regularly rotating H I disk with a total H I mass of 1 . × M (cid:12) . To the north is IC 3571 ( M H I = 4 . × M (cid:12) ; Healdet al. 2011), which shows evidence of interaction with NGC4565. Specifically, a weak H I bridge is found between thedisk of NGC 4565 and this companion galaxy, which exhibitsa smooth change in velocity between the two systems (van derHulst & Sancisi 2005; Heald et al. 2011). The initial extensionof this bridge from NGC 4565 can be seen as an extendedshelf in Figure 1 at (R.A., decl.) ∼ (12 h m s , 26 ◦ (cid:48) ). Wealso find evidence in Figure 1 for a smaller overdensity, orcloud, of H I lying above the disk at ∼ –(12 h m s , 26 ◦ (cid:48) )as also seen by Zschaechner et al. (2012). This overdensityappears connected to the bridge between NGC 4565 and IC3571. HST
ACS PHOTOMETRY
The ACS field covers the extent of the H I NW warp, asindicated in Figure 1 by a dashed box. The field was ob-served with exposures of 8266 s and 7340 s using the F606Wand F814W filters, respectively (
HST
GO program 12196).Each exposure was dithered to aid cosmic-ray rejection andto cover the chip gap. However, in the subsequent analysis,we removed detections from the chip gap region where thedetection efficiency is less due to the shorter mean exposuretime. We followed the same procedures for the data reductionas detailed by Radburn-Smith et al. (2011). This involved us-ing SE
XTRACTOR (Bertin & Arnouts 1996) to mask out re-solved sources and the ACS module of
DOLPHOT (Dolphin2000) to identify stars and measure their photometry. We alsoused the same crowded-field selection parameters selected byRadburn-Smith et al. (2011) to remove the bulk of unresolvedbackground galaxies. Finally, we generated approximately250,000 artificial stars that mimicked the color and magni-he NGC 4565 Disk Warp 3
Figure 2.
CMD of the stars in the ACS field is shown in the leftmost panel. Filled contours are used at densities greater than 50 stars in a 0.1 mag × . The TRGB is indicated by a red dot-dashed line, whilethe 50% completeness limit measured from the artificial star tests is marked by a blue dashed line. Old RGB stars are indicated in red, while young MS and HeBbranches are colored blue. Ellipses indicate the photometric uncertainties reported by DOLPHOT , while the arrow indicates the direction of foreground reddening,which has been corrected for in these data. The spatial densities of these old and young stars are shown in the middle and right panel, respectively. The densitiesare plotted as Voronoi tessellations on top of an Rc -band POSS image with H I gas column densities spaced logarithmically by 0.2 dex between 5 × and5 × cm − . A clear correlation between the gas and the young stars is seen, but no such coincidence is seen with the older stars. tude distribution of the observed photometry. These artificialstars were added to the ACS image and subsequently passedindividually through the same photometry pipeline and selec-tion criteria. By measuring the recovery rate of these stars,we can assess incompleteness in the real data at fainter mag-nitudes and measure the effects of stellar crowding in the disk.We found that at magnitudes of F814W < . DOLPHOT , reports the typical uncertainty as 0.07 mag. Theresulting color-magnitude diagram (CMD) of the stellar de-tections is shown in the leftmost panel of Figure 2.By analyzing synthetic CMDs similar to Figure 2, we canidentify and date several distinct regions corresponding to dif-ferent stages of stellar evolution. Such models depend criti-cally on metallicity, which in outer disks is found to rangefrom [M/H] = − − . − <
100 Myr. Asshown by the shaded region of the CMD, we have binnedthese two features together to represent young stars. In themiddle and rightmost panel of Figure 2, the spatial density ofthese old and young stars, shown as Voronoi tessellations, arecompared with the reprocessed H I gas map as indicated bythe contours. A clear correlation with the gas warp is seenwith the young stars but not with the old stars. After cor- recting for completeness, the mean density of these youngstars (22 . < F814W < .
6) in the warp is (2 . ± . × − stars arcsec − . This is approximately four times less thanthe mean density of the old stars (26 . < F814W < .
5) at(1 . ± . × − stars arcsec − . THE STELLAR COMPONENT OF THE WARP
To study the star formation history (SFH) of the warp, wefirst defined the warp as the region of the ACS image withcoincident H I detection ( > × cm − ). We excluded anydetections from the main disk, which we defined as the re-gion in the southeast corner of the ACS field bound by a stel-lar density of 0.5 stars arcsec − . The remaining area is at-tributed to the stellar halo. To measure the SFH of these re-gions, we used the software package MATCH (Dolphin 2002),which fits the observed CMD with synthetic stellar popula-tions. These populations were generated using stellar evolu-tion models and a model of completeness and photometric er-rors, which is computed from the artificial star tests describedin Section 3. For this analysis, we used the stellar evolutionmodels of Marigo et al. (2008) with the updated AGB tracksof Girardi et al. (2010) and adopted both a Kroupa (2001) ini-tial mass function (IMF) and a binary fraction of 40%. Thesesynthetic populations were distributed across a 36 ×
25 gridin log(age)-[ Z ] space with bin widths of 0.1 dex. The dis-tance modulus was set to m − M = 30 .
38 (as measured fromthe TRGB, Radburn-Smith et al. 2011), the extinction was setto A V = 0 .
05 (Schlegel et al. 1998), and the chemical enrich-ment history was required to increase in metallicity with time.
MATCH also allows for up to 0.5 mag of differential reddeningfor stars younger than 40 Myr, decreasing to 0 mag for starsolder than 100 Myr. With this setup, we measured the SFR infixed logarithmic age bins of width 0.3 dex. To determinethe relative uncertainties in SFR between different regions
MATCH employs a Markov Chain Monte Carlo approach tosample the probability space around the original data.The regions, and their respective SFHs, are shown in Fig-ure 3. For ages older than ∼
300 Myr, the SFRs of the halo and Radburn-Smith et al.
Figure 3.
Left panel shows all the stellar detections superimposed on the F814W image. The H I column density is plotted as contours spaced logarithmicallyby 0.2 dex between 5 × and 5 × cm − . The smoothed dashed line bounds the region containing stellar detections greater than 0.5 stars arcsec − , whichwe define as the disk population. The warp population is composed of stars residing in regions with coincident H I flux > × cm − and that are not part ofthe disk population. Stars in the remaining area are identified as the halo population. We mask out all stellar detections in the chip gap region, which is markedby a vertical strip. The right panel plots the SFHs of these separate populations in logarithmic age bins of width 0.3 dex. Error bars indicate uncertainties basedon a Markov Chain Monte Carlo exploration of the probability space. Note that the uncertainties in the underlying stellar models are not included as we are onlyinterested here in the relative difference of the SFHs rather than the absolute calibration. A pronounced difference is seen between the star formation rates in thewarp and halo at ages <
300 Myr. warp are remarkably similar. This suggests that the old starsfound in the warp are likely contamination from the halo andare not directly associated with the warp. However, a cleardifference between the halo and warp populations is seen atyounger ages. Specifically, the SFR of the halo rapidly dropsoff and is undefined at ages <
20 Myr while stars in the warpsustain a steady SFR of (6 . + . − . ) × − M (cid:12) yr − kpc − overthe last 250 Myr. Following Dolphin (2012), this measure-ment includes uncertainties in the stellar evolutionary mod-els by incorporating shifts in both luminosity and temperaturewhen computing the Monte Carlo realizations.Figure 3 also shows a dearth of SF between 1 Gyr and 300Myr ago. The subsequent rapid increase in SF across all com-ponents after this period may indicate a significant event thattriggered both the SF and the onset of the warp. However, thisfeature may also be attributed to a lack of sensitivity at theseintermediate timescales. Such stellar ages encompass AGBstars (discussed in Section 5) and faint HeB stars that are atthe detection limit of the ACS exposure. MATCH accounts for interstellar extinction in the fits by al-lowing for up to 0.5 mag of differential reddening for the veryyoung stars found in dusty stellar nurseries. However, as anextreme case, we may ask how the SFH results would changeif interstellar dust affected all populations. To estimate thislevel of extinction, we used the relation between H I columndensity and dust extinction as measured locally by Bohlin et al(1978). Given that the average projected surface density ofH I in the warp is 4 × M (cid:12) kpc − , this relation yields a to-tal extinction of A V ∼ .
26 mag. If the stars are embeddedin the gas, then half this value may be assumed as the typicalstellar extinction. However, we note that such a calculationis likely to overestimate dust extinction in low-metallicity en-vironments such as those expected to be found in the warp(Leroy et al. 2011). To test the impact on the SFH from a red-dening of this order of magnitude, we allowed
MATCH to fitup to 0.5 mag of differential extinction for all stellar popula-tions. The resulting fits were consistent with the best-fit SFHspresented in Figure 3 given the
MATCH reported total uncer- tainties. We thus conclude that interstellar reddening has littleeffect on the SFHs.The projected surface density of H I in the warp is compa-rable to the typical density of H I outer disks in nearby spiralgalaxies (Bigiel et al. 2010b). Using the warp SFR as cal-culated by MATCH , we infer a depletion time of ∼ ± ∼
100 Gyr depletion times ofquiescent outer disks as found by Bigiel et al. (2010b) usingfar-UV (FUV) emission as a proxy for SFR. This suggests thatthe same factors setting SF efficiency (defined as the inverseof the depletion time) in galaxy outer disks are also operatingin the NGC 4565 H I warp. Stellar Populations
Due to the shallow depth of the observations, the SFHs de-rived in Figure 3 may lack sensitivity at intermediate ages.Hence, to further examine the stellar content of the H I warp,we investigated the CMD of individual detections. Specifi-cally, we examined detections in a region of the warp with apronounced overdensity of stars. This includes a massive starcluster lying at (R.A., decl.) ∼ (12 h m s , + ◦ (cid:48) . (cid:48)(cid:48) Galaxy Evolution Explorer (Martin et al. 2005)archive. Although there is some evidence of UV emissionfrom the surrounding warp, the flux in the region is mostly at-tributed to this galaxy, which thus hampers studies of the UVwarp.We defined a new halo region, which is equal in size( ∼ ) after adjusting for the area removed fromthe warp region due to the chip gap. For greater age fi-delity, we split the CMDs into the distinct stellar popula-tions discussed in Section 3. Based on the stellar isochroneshe NGC 4565 Disk Warp 5 Figure 4.
Stellar content of the warp. The left panel shows stellar detections superimposed on the
HST
ACS F814W image, with H I column densities plottedas contours spaced by 0.2 dex between 5 × and 5 × cm − . The chip gap, which we remove from our analysis, is indicated by a nearly vertical redstrip. The sloping yellow line indicates the plane of the disk inferred from the stellar density in the lower left of the image. The colors of the points representthe distinct stellar populations: MS, lHeB, uHeB and AGB stars as labeled in the CMD in the middle panel. RGB stars and detections that do not lie in one ofthese age bins make up the majority of detections, but are not shown in the left panel for clarity. The background galaxy SDSS J123556.05+260601.6, which willcontaminate UV studies of the warp, is labeled. Two rectangular regions inside and outside of the warp are indicated by shaded boxes. Detections within theseregions are displayed as CMDs in the middle and rightmost panel. Regions denoting different stellar populations, as described in the text, are shaded. The 50%recovery rate as measured from the artificial star tests is indicated by a dashed blue line, and arrows indicate the direction of the corrected foreground reddening.The CMD in the warp region shows an abundance of young stars not seen in the halo region. used to construct the SFHs derived earlier, we further splitthe helium-burning branches into an upper helium-burningbranch (uHeB), which corresponds to an age range of 30–200Myr, and a lower helium-burning branch (lHeB) of 100–600Myr. Figure 4 plots the new warp and halo regions, and theirresulting CMDs.Clearly evident in the CMD of the overdensity in the warpregion is an abundance of younger stars (MS, uHeB and lHeB)that are not present in the halo region. This suggests, as alsodetermined by the SFHs, that SF is ongoing in the warp. How-ever, the density of AGB and RGB stars also appears higher inthe warp region. This is likely due to an older, flattened enve-lope surrounding the main disk, such as a thick disk or innerstellar halo. Such a structure would yield higher densities ofold stars closer to the disk axis. To place age constraints onthe H I warp, we thus need to assess the correlation of theseolder stars with the warp.In Figure 5, we define three strips running perpendicular tothe disk plane. Along each of these strips, we plot separatestellar density profiles for the RGB, AGB, lHeB, uHeB, andMS detections, as well as the surface density of H I gas. Asshown by Radburn-Smith et al. (2011), these stellar densitiesare equivalent to surface brightnesses. In the crowded regionof the disk, detections are lost due to the overlapping point-spread functions of the stars. Although this effect can be par-tially corrected by artificial star tests, at extreme densities thecorrection factor can become too large and therefore unreli-able. We thus excluded the high density regions within 20 (cid:48)(cid:48) ofthe disk plane as indicated in the leftmost panels by the lightlyshaded points. We also excluded the chip gap from the analy-sis as the total exposure time in this region is shorter than therest of the field. This results in a brighter completeness limit,which would affect detections in the fainter CMD selectionbins. To illustrate the symmetry of the profiles, dashed linesin the figure correspond to reflections of the profiles aroundthe disk axis.The leftmost panel of Figure 5 lies largely within the dom- inant main disk, and so all populations are found to approx-imately correspond with the gas distribution. We note thatthis distribution is slightly asymmetrical as the main stellardisk itself gradually warps upward. The middle plot showsthe transition from the disk to warp populations, which is de-scribed by a gradual increase in the asymmetry of the pop-ulations with decreasing age. However, the rightmost panel,which covers the strip in the warp, shows a clear discrepancybetween both the RGB and AGB profiles with the youngerpopulations. The RGB stars are symmetrically distributedaround the disk axis, as would be expected from a very ex-tended thick-disk-like component or an oblate stellar halo.The AGB population appears to be similarly distributed ex-cept for a small overdensity coincident with the peak of thegas distribution. However, the peaks of the asymmetric lHeB,uHeB, and MS populations all strongly coincide with the peakof the H I profile. This again suggests that the majority of theRGB and AGB stars are not associated directly with the H I structure but rather with an extended outer envelope that iscoincident with the warp when seen in projection. Future ob-servations to the south of the field, i.e., < − (cid:48)(cid:48) from the diskplane, would help measure the symmetry of this older com-ponent and so rule out coincidence with the H I warp.Given the stochastic uncertainties of the profiles in Figure 5,a sufficiently small fraction of the old RGB stars may be asso-ciated with the warp without producing any detectable peak.However, as the bulk of detections in the warp are RGB stars(see Figure 4), this small fraction may still be significant. Inorder to date the warp using stellar ages, we thus need to as-sess our detection sensitivity to these old stars.Using the SFH analysis presented in Section 4, we con-structed a series of artificial CMDs with increasingly inflatedSFRs at ages older than 1 Gyr. The excess RGB stars gen-erated from these higher SFRs were then spatially distributedin the warp following the density of young stars. Finally, theRGB profile from Figure 5 was recomputed and the signal-to-noise of any peak near the warp was measured. Using this Radburn-Smith et al. Figure 5.
Correlation of the stars and gas across the plane of the disk in the NGC 4565 observations. The shaded regions in the top panels indicate the cutsperpendicular to the plane of the disk, which is marked by a solid yellow line. Blue dots indicate the distribution of MS stars in the ACS region. The chip gapis shown as a vertical red stripe and any detections falling in this region are removed. The first cut (left column) covers the start of the warp from the main disk,while the third (right panel) cuts across the full warp. The second cut (middle panel) is intermediate. Below these plots are the stellar densities along the cuts ofthe old RGB stars (top red points), AGB, lHeB, uHeB, and MS stars. The AGB, lHeB, uHeB, and MS profiles have been offset by − . − . −
3, and − . I gas density along the strips is shown as a gray solid line at the bottom of each column. The stellar profiles are reflected around the disk planeand plotted as dashed lines. A clear correlation between the young stars and gas is seen in the right panel. The corresponding distribution of old stars insteadshows a symmetric distribution around the disk plane, as indicated by a second order polynomial fit to the RGB profile, which is plotted as a thin gray line. assessment, we found that in order to produce a statisticallysignificant peak, the old SFR in the warp needed to be in-creased by an additional ∼ × − M (cid:12) kpc − yr − relative tothe SFR in the halo. This is comparable to the SFR seen inthe warp today. Hence, the lack of a peak in the RGB dis-tribution of Figure 5 does not preclude the existence of olderSF associated with the warp. However, we note that such SFwould likely be seen in Figure 3, as the SFH analysis is sen-sitive to much lower SFRs in the warp relative to the halo atthese stellar ages ( < × − M (cid:12) kpc − yr − ). AGB STARS
In the outermost profile of Figure 5, a small excess of AGBstars is seen in the region of the warp ( ∼ + (cid:48)(cid:48) ). As previ-ously noted, AGB stars are typically a few Gyr old and somay represent the oldest component of the warp. In Fig-ure 6, we replot the CMD of the warp stars in the stellaroverdensity defined in Section 4, which intersects with thethird strip of Figure 5. Overplotted in each panel are binnedregions of constant age using different assumptions. The un-he NGC 4565 Disk Warp 7 Figure 6.
CMDs of the AGB stars in the stellar-overdensity region found in the warp of NGC 4565. Approximate upper-age limits to these stars are indicatedby colored strips for different theoretical models. Stars identified in Figure 4 as AGBs by their color and magnitude are indicated by heavy black dots with ared border. The age bins are generated from the AGB evolutionary models of Marigo et al. (2008) and Girardi et al. (2010) and include the thermally pulsatingstage of the AGB sequence. Panel (a) uses the default parameters with no circumstellar dust, an interstellar extinction of A V = 0 .
042 mag, and a metallicity of Z = 0 . − Z = 0 .
006 ([M/H] = − . ∼ derlying isochrones are extracted from the models of Marigoet al. (2008) using mass-loss corrections to the AGB tracksfrom Girardi et al. (2010). However, we note that as starsevolve upward through the AGB region these bands should betreated as upper-age limits. The bands also average over theeffects of thermal pulsations. Typically the effects of thesepulsations are smaller than the vertical (F814W) extent ofeach age bin. For our baseline instance, we use a Chabrier(2001) log-normal IMF, a total extinction of A V = 0 .
042 magwith no contribution from circumstellar dust, and a metallic-ity of Z = 0 . − ∼ Z = 0 .
006 ([M/H] = − . −
1) but use isochrones that includecircumstellar reddening. To achieve this, Marigo et al. (2008)tracked the C/O ratio in their model stellar atmospheres forstars that experienced significant mass loss. Depending onthis ratio, each star was then reddened using the AGB dust ab-sorption models of Groenewegen (2006), either with a mix of60% aluminum oxide (AlOx) and 40% silicates for oxygen-rich stars, or 85% amorphous carbon (AMC) and 15% sili-con carbide (SiC) for carbon-rich stars. With this addition, astrong dependence on color is now seen, with large extinctionvalues in F814W for redder colors.We explored this age dependence on the spatial distribu-tion of AGB stars across the entire ACS field, as shown inFigure 7. In this figure, we only plot stars brighter than the TRGB that fall in the upper-age bands of Figure 6, specifi-cally the 400 Myr to 1 Gyr bands (shown in the left panels)and 1 Gyr to 4 Gyr bands (right panels). Thus not all starsthat fall in the classical AGB region are shown. Furthermore,as the bands in each panel of Figure 6 cover different regionsof the CMD, selections based on these various models willsample different subsets of the entire AGB population. Us-ing the age definitions from the evolutionary models withoutcircumstellar reddening (upper panels), we found a high spa-tial correlation between the H I warp and stars < < < < I warp only for the cases without circumstellar red-dening, we can reason a priori that these evolutionary modelsare a better match to these data.In Section 4, we estimated based on the H I gas densitythat interstellar reddening from dust may account for up to A V ∼ .
13 mag of extinction. Correcting for this would shiftthe stars by − .
04 mag in color and − .
08 mag in F814Wmagnitude, slightly decreasing the measured ages of the stars.However, we disregarded such shifts in this analysis as theyare smaller than the typical systematic uncertainties in themodels (e.g., the differences between the left and middle pan-els of Figure 6).If these stars do indeed place an upper limit of ∼ I warp structure, we need to assess the validityof using AGB evolutionary models to tag their age. That is,we should assess the likelihood that these stars are AGB starsbased solely on their position in the CMD. Significance
We studied the dependency of stellar age on position in aCMD by using data from the Optical Gravitational LensingExperiment (OGLE; Udalski et al. 1997). This survey pro- Radburn-Smith et al.
Figure 7.
Spatial distribution of stars in the ACS field that are brighter than the TRGB and fall in the upper-limit age bands defined in Figure 6. The leftmostpanels plot only the stars that fall in the 400 Myr to 1 Gyr bands, while the rightmost panels use the 1-4 Gyr bands. The upper two plots use our baseline stellarevolutionary models from Marigo et al. (2008) and Girardi et al. (2010) (no circumstellar dust, an interstellar extinction of A V = 0 .
042 mag, and a metallicity of[M/H] = − < vides photometry and ages of approximately 600 star clustersin the Large Magellanic Cloud (LMC), spanning ages from ∼ ∼ HST /ACSF606W and F814W photometry to V - and I -band photome-try using the conversions listed in Sirianni et al. (2005) anddereddened all photometry using the extinction values pub-lished by Schlegel et al. (1998).We began by arranging the LMC clusters into nine age bins,spaced logarithmically by 0.3 dex. We then binned the pho- tometry of each cluster into a Hess diagram at a resolution of0.1 mag in color and 0.2 mag in the I -band equivalent. Foreach of these clusters, Pietrzynski & Udalski (2000) defined acore radius, marked by a major drop in radial stellar density,as well as a total cluster radius, calculated by fitting profilesto the radial stellar density. We subsequently defined a skyregion for each system as the annulus that extended outwardfrom the total cluster radius and encompassed an area equalto that of the cluster. Stars from the OGLE stellar maps werethen selected in this region by their R.A. and decl., and weresimilarly binned into a 0 . × . Figure 8.
CMDs of the degraded LMC clusters and the stellar-overdensity region in the NGC 4565 warp plotted in the Johnson-Cousins system used by OGLE.Gray points indicate degraded stars that were discarded by the selection criteria used to cull the NGC 4565 warp photometry. The left panel plots the degradeddata from the combined 34 LMC clusters in the youngest age bin. A clear overdensity of stars falling in the AGB region is seen even though the cluster is tooyoung to contain AGB stars. The middle panel plots these data from the overdensity of the warp in NGC 4565, which shows that same overdensity. The rightpanel plots the degraded data from the 64 LMC clusters in the oldest age bin. The main sequence is no longer visible but a clear abundance of old AGB and RGBstars is seen.
We degraded the resulting photometry of each cluster us-ing the artificial stars generated for the NGC 4565 ACS fieldas described in Section 3. For each artificial star that wasinjected into the ACS image and run through the pipeline,both the input and the recovered colors and magnitudes wererecorded. We binned these input magnitudes and colors intothe same 0 . × . ∼ < . < σ spreadof these realizations around the median values is indicated bya shaded region in Figure 9. We note, as expected, a sharp in-crease in this ratio around the onset of the AGB phase ( ∼ ∼ . IMPLICATIONS FOR WARPFORMATION MECHANISMS
Early work by Hunter & Toomre (1969) showed that galaxydisks are unable to sustain warps. Hence, the long-lived warpswe see today are due to ongoing gravitational effects. Initially,the dark matter halo was believed to be the source of thisexcitation. By creating a rigid, flattened halo that was mis-aligned with the disk, many authors were able to create suchwarps (e.g., Dekel & Shlosman 1983; Toomre 1983; Sparke& Casertano 1988). However, such halos are unrealistic, andmore modern simulations with live halos have shown that theinduced warps quickly disappear, primarily due to the shapeof the inner halo realigning itself with the disk (Dubinski &Kuijken 1995; Binney et al. 1998).Consequently, warps are now suspected to be driven by twobroad mechanisms: the tidal interaction of nearby galaxiesand the effects of misaligned infall.
Tidal Interactions
Interactions with satellite galaxies have long been suspectedto induce warps (e.g., Huang & Carlberg 1997) and havebeen investigated in detail to explain the warp of the MilkyWay (MW). However, the masses of the Magellanic Clouds,specifically the LMC, seem insufficient to generate the warp(Hunter & Toomre 1969). Weinberg (1998) proposed thatthe response of the halo to the LMC may amplify the torque;this was further modeled and found to be viable by Tsuchiya(2002) and Weinberg & Blitz (2006). However, despite thiseffect, the orbit of the LMC may be at odds with the loca-tion of the existing warp. An analytical and N -body simula-tion by (García-Ruiz et al. 2002a) suggests that the Magel- Although, as noted by Sparke & Casertano (1988), such models cannoteasily explain the turnover of the NGC 4565 warp at larger radii back to thedisk plane.
Figure 9.
Ratio of the residual number of AGB stars to young stars from the LMC cluster analysis. This ratio is systematically greater than zero at all timesand is consistent with the ratio seen in the NGC 4565 warp, indicated by a dashed blue line. The onset of AGB formation, shown as a vertical dot-dashed line,coincides with a dramatic increase in the ratio. The shaded region indicates the one-sigma spread of 100 Monte Carlo re-simulations of each age bin. lanic Clouds would induce a warp perpendicular to the ori-entation of the actual warp. Instead, the favorable locationof the Sagittarius dwarf galaxy, which has been disrupted inour inner halo, may be the cause of the MW warp (Ibata &Razoumov 1998; Bailin 2003; Gómez et al. 2013).Similar arguments for tidal satellite interactions causingwarps have been made for external galaxies such as NGC5907, which was once thought to be isolated (Shang et al.1998). Additionally, gravitational encounters with dark-matter substructure within the host halo have also been shownto induce warps (Kazantzidis et al. 2008). Hence, the twonearby companions of NGC 4565, as seen in Figure 1, mayin part be responsible for the warp. Indeed, bending of theH I layer toward the closer IC 3571 system has been observed(van der Hulst & Sancisi 2005). However, this is independentof the larger warp.If such tidal interactions are warping a pre-existing gas disk,then they will also have an effect on any stars associated withthat disk. If such mechanisms formed the prodigious H I warpof NGC 4565, then they need to explain the lack of old starsassociated with the warp. The vertical energy of these dynam-ically hotter stars will likely increase in response to a tidal in-teraction. However, whether these stars can be sustained in awarped component, as opposed to simply thickening the olderstellar disk, remains to be shown by simulations.SF could also have been triggered by the same processforming the warp. In such a chaotic event, resonances mayselectively reinforce the warp perturbation such that both thegas and the newly formed stars spatially coincide, as seen inNGC 4565. If this is the case, these tidal models may need toshow that the warp in gas and stars can survive and remain co-incident for ∼ Misaligned Infall
If simulations reveal the satellites of NGC 4565 are unableto induce the warp of the host system, then misaligned infallremains the most likely cause.In a Λ CDM cosmology, material is regularly accreted from all directions, causing the angular momentum of the hostdark-matter halo to continually, and significantly, change overtime in a random walk (Vitvitska et al. 2002). Ostriker & Bin-ney (1989) proposed that the continual slewing of the orienta-tion of a flattened outer halo due to this misaligned accretioncauses warps in the galaxy disk. Indeed, such warping hasbeen shown in numerical simulations (e.g., Jiang & Binney1999; Shen & Sellwood 2006). More generally, Debattista &Sellwood (1999) showed that when the angular momentum ofeven a spherical halo and galaxy disk differ, dynamical fric-tion can cause long-lived warps.As with the tidal interactions, these simulations torque anexisting gas disk. Hence, in order to explain the stellar con-tent of the NGC 4565 H I warp, the relative influence of theformation mechanisms on stars older than 1 Gyr needs to befurther investigated in these simulations.Alternately, rather than torquing an existing disk, the warpitself may have formed from the direct accretion of misalignedmaterial. Sancisi et al. (2008) show several examples ofgas-rich dwarf galaxies interacting with their host galaxies,e.g., through an H I bridge. If these systems merge, withthe higher-angular-momentum material falling into an outerdisk, we might expect a gas warp with coincident stars. How-ever, given that approximately 50% of the stellar mass in thesedwarf galaxies is in place by z = 2 (Weisz et al. 2011), olderstars would likely lie in the warp as well.Instead of a dwarf galaxy, Roškar et al. (2010) suggest thatthe warp forms from infalling gas. In their models, this ma-terial cools and sinks through a hot gaseous halo. The spinof this halo, which is not aligned with the disk, torques theinfalling material. Hence, the angular momentum of the freshgas when it finally reaches the disk is aligned with the spinof the gas halo and thus forms a warped disk. These hydro-dynamic torques overwhelm the tidal torques from the darkmatter halo, which are insufficient for affecting the infallinggas.Roškar et al. (2010) posit that the cooling gas in the long-lived warp will reach densities sufficient for SF, resulting inyoung metal-poor stars associated with the warp. Conse-he NGC 4565 Disk Warp 11quently, their Figure 14 shows a remarkable agreement withthat of NGC 4565 (Figure 2), namely an old stellar componentsmoothly distributed around the galaxy and a young stellarcomponent associated with the warp.Unlike the previous warp formation mechanisms, theRoškar et al. (2010) warp forms from pristine material.Hence, future metallicity measurements of the warp may lendfurther credence to this model. CONCLUSIONS
Using an
HST
ACS observation, we have investigated thestellar distribution of stars in the vicinity of the prodigiousH I warp of NGC 4565. We find no correlation of old ( > <
600 Myr) in the warp.An analysis of the SFH shows that the rate of SF in the warpincreased, relative to the surrounding regions, at least 300 Myrago. The sustained SFR over this time is (6 . + . − . ) × − M (cid:12) yr − kpc − .A slight excess of stars falling in the AGB region of theCMD are also found in the warp. Stellar evolutionary modelslimit the age of these stars to < I gasin the warp of ∼ ±
20 Gyr, similar to the rates found inthe H I outer disks of nearby spiral galaxies. However, theseouter disks are often dominated by RGB populations (e.g.,de Jong et al. 2007; Radburn-Smith et al. 2012). The lack ofthese stars in the NGC 4565 warp may suggest that the gas hasbeen recently accreted. Indeed, simulations of warp formationfrom newly accreted gas by Roškar et al. (2010) predict stellardistributions in qualitative agreement with the NGC 4565 H I warp. In this scenario, the warp represents a large reservoir ofnewly accreted H I gas, which could be a potential fuel sourcefor the galaxy disk it circumscribes (e.g., see discussion inSancisi et al. 2008; Bigiel et al. 2010a)These observations help place constraints on numerical andanalytical models of warp formation. Future observations ofthe metallicity of the warp will help clarify the origin of thegas and therefore the mechanisms driving the warp.The authors thank C. Purcell and V. Debattista for insightfuldiscussions on the theories of warp formation. Support forthis work was provided by NASA through grant GO-12196from the Space Telescope Science Institute, which is operatedby the Association of Universities for Research in Astronomy,Incorporated, under NASA contract NAS5-26555. Facility:
HST (ACS)
REFERENCESBailin, J. 2003, ApJL, 583, L79Battaner, E., Florido, E., & Sanchez-Saavedra, M. L. 1990, A&A, 236, 1Bertin, E., & Arnouts, S. 1996, A&ASS, 117, 393Bigiel, F., Leroy, A., Seibert, M., et al. 2010a, ApJL, 720, L31Bigiel, F., Leroy, A., Walter, F., et al. 2010b, AJ, 140, 1194Binney, J. 1978, MNRAS, 183, 779Binney, J. 1992, ARA&A, 30, 51 Binney, J., Jiang, I.-G., & Dutta, S. 1998, MNRAS, 297, 1237Binney, J. J. 2000, in ASP Conf. Ser. 197, Dynamics of Galaxies: from theEarly Universe to the Present, ed. F. Combes, G. A. Mamon, &V. Charmandaris (San Francisco, CA: ASP), 107Bohlin, R. C., Savage, B. D., & Drake, J. F. 1978, ApJ, 224, 132Bresolin, F., Ryan-Weber, E., Kennicutt, R. C., & Goddard, Q. 2009, ApJ,695, 580Briggs, F. H. 1990, ApJ, 352, 15Brook, C. B., Kawata, D., Gibson, B. K., & Freeman, K. C. 2004, ApJ, 612,894Brooks, A. M., Governato, F., Quinn, T., Brook, C. B., & Wadsley, J. 2009,ApJ, 694, 396Chabrier, G. 2001, ApJ, 554, 1274Dahlem, M., Ehle, M., Ryder, S. D., Vlaji´c, M., & Haynes, R. F. 2005,A&A, 432, 475Debattista, V. P., & Sellwood, J. A. 1999, ApJL, 513, L107de Jong, R. S. 2008, MNRAS, 388, 1521de Jong, R. S., Seth, A. C., Radburn-Smith, D. J., et al. 2007, ApJL, 667,L49Dekel, A., & Birnboim, Y. 2006, MNRAS, 368, 2Dekel, A., & Shlosman, I. 1983, in IAU Symp. 100, Internal Kinematics andDynamics of Galaxies, ed. E. Athanassoula (Cambridge: Cambridge Univ.Press), 187Dolphin, A. E. 2000, PASP, 112, 1383Dolphin, A. E. 2002, MNRAS, 332, 91Dolphin, A. E. 2012, ApJ, 751, 60Dubinski, J., & Kuijken, K. 1995, ApJ, 442, 492Ferguson, A. M. N., Gallagher, J. S., & Wyse, R. F. G. 1998a, AJ, 116, 673Ferguson, A. M. N., Wyse, R. F. G., Gallagher, J. S., & Hunter, D. A. 1998b,ApJL, 506, L19García-Ruiz, I., Kuijken, K., & Dubinski, J. 2002a, MNRAS, 337, 459García-Ruiz, I., Sancisi, R., & Kuijken, K. 2002b, A&A, 394, 769Gil de Paz, A., Madore, B. F., Boissier, S., et al. 2007, ApJ, 661, 115Girardi, L., Williams, B. F., Gilbert, K. M., et al. 2010, ApJ, 724, 1030Gómez, F. A., Minchev, I., O’Shea, B. W., et al. 2013, MNRAS, 429, 159Groenewegen, M. A. T. 2006, A&A, 448, 181Hanish, D. J., Meurer, G. R., Ferguson, H. C., et al. 2006, ApJ, 649, 150Heald, G., Józsa, G., Serra, P., et al. 2011, A&A, 526, A118Huang, S., & Carlberg, R. G. 1997, ApJ, 480, 503Hunter, C., & Toomre, A. 1969, ApJ, 155, 747Ibata, R. A., & Razoumov, A. O. 1998, A&A, 336, 130Jiang, I.-G., & Binney, J. 1999, MNRAS, 303, L7Kahn, F. D., & Woltjer, L. 1959, ApJ, 130, 705Kazantzidis, S., Bullock, J. S., Zentner, A. R., Kravtsov, A. V., &Moustakas, L. A. 2008, ApJ, 688, 254Kereš, D., Katz, N., Weinberg, D. H., & Davé, R. 2005, MNRAS, 363, 2Kormendy, J., & Barentine, J. C. 2010, ApJL, 715, L176Kroupa, P. 2001, MNRAS, 322, 231Leroy, A. K., Bolatto, A., Gordon, K., et al. 2011, ApJ, 737, 12Macciò, A. V., Moore, B., & Stadel, J. 2006, ApJL, 636, L25Marigo, P., Girardi, L., Bressan, A., et al. 2008, A&A, 482, 883Martin, D. C., Fanson, J., Schiminovich, D., et al. 2005, ApJL, 619, L1Naeslund, M., & Joersaeter, S. 1997, A&A, 325, 915Nelson, D., Vogelsberger, M., Genel, S., et al. 2013, MNRAS, 429, 3353Ostriker, E. C., & Binney, J. J. 1989, MNRAS, 237, 785Pietrzynski, G., & Udalski, A. 2000, AcA, 50, 337Radburn-Smith, D. J., de Jong, R. S., Seth, A. C., et al. 2011, ApJS, 195, 18Radburn-Smith, D. J., Roškar, R., Debattista, V. P., et al. 2012, ApJ, 753, 138Reid, I. N., Brewer, C., Brucato, R. J., et al. 1991, PASP, 103, 661Robertson, B., Bullock, J. S., Cox, T. J., et al. 2006, ApJ, 645, 986Roškar, R., Debattista, V. P., Brooks, A. M., et al. 2010, MNRAS, 408, 783Rupen, M. P. 1991, AJ, 102, 48Sancisi, R. 1976, A&A, 53, 159Sancisi, R. 1983, in IAU Symp. 100, Internal Kinematics and Dynamics ofGalaxies, ed. E. Athanassoula (Cambridge: Cambridge Univ. Press), 55Sancisi, R., Fraternali, F., Oosterloo, T., & van der Hulst, T. 2008, A&AR,15, 189Schlegel, D. J., Finkbeiner, D. P., & Davis, M. 1998, ApJ, 500, 525Shang, Z., Zheng, Z., Brinks, E., et al. 1998, ApJL, 504, L23Shen, J., & Sellwood, J. A. 2006, MNRAS, 370, 2Sirianni, M., Jee, M. J., Benítez, N., et al. 2005, PASP, 117, 1049Sparke, L. S., & Casertano, S. 1988, MNRAS, 234, 873Spavone, M., Iodice, E., Arnaboldi, M., et al. 2010, ApJ, 714, 1081Thilker, D. A., Bianchi, L., Meurer, G., et al. 2007, ApJS, 173, 538Tinsley, B. M. 1980, FCPh, 5, 287Toomre, A. 1983, in IAU Symp. 100, Internal Kinematics and Dynamics ofGalaxies, ed. E. Athanassoula (Cambridge: Cambridge Univ. Press), 1772 Radburn-Smith et al.