Temperature and Metallicity Gradients in the Hot Gas Outflows of M82
Laura A. Lopez, Smita Mathur, Dustin D. Nguyen, Todd A. Thompson, Grace M. Olivier
DDraft version June 17, 2020
Typeset using L A TEX twocolumn style in AASTeX62
Temperature and Metallicity Gradients in the Hot Gas Outflows of M82
Laura A. Lopez,
1, 2, 3
Smita Mathur,
1, 2
Dustin D. Nguyen,
2, 4
Todd A. Thompson,
1, 2 and Grace M. Olivier
1, 2 Department of Astronomy, The Ohio State University, 140 W. 18th Ave., Columbus, OH 43210, USA Center for Cosmology and AstroParticle Physics, The Ohio State University, 191 W. Woodruff Ave., Columbus, OH 43210, USA Niels Bohr Institute, University of Copenhagen, Blegdamsvej 17, 2100 Copenhagen, Denmark Department of Physics, The Ohio State University, 191 W. Woodruff Avenue, Columbus, OH, 43210, USA
ABSTRACTWe utilize deep
Chandra
X-ray Observatory imaging and spectra of M82, the prototype of a star-bursting galaxy with a multiphase wind, to map the hot plasma properties along the minor axis of thegalaxy. We extract spectra from 11 regions up to ± − . − Keywords:
Galactic winds — Starburst galaxies INTRODUCTIONGalaxy-scale outflows driven by star formation areubiquitous (Heckman et al. 1990; Veilleux et al. 2005;Rubin et al. 2014). These galactic winds chemicallyenrich the circumgalactic (CGM) and intergalacticmedium (IGM; Borthakur et al. 2013; Werk et al. 2016)as well as regulate the growth and metal-enrichment ofgalactic disks (Oppenheimer & Dav´e 2008; Peeples &Shankar 2011). The prevailing picture is that outflowsare driven by hot gas shock-heated by stellar windsand supernovae (SNe) that entrain the dust, cold, andwarm gases within the flow (e.g., Chevalier & Clegg1985). Additional mechanisms that have been proposedto accelerate cool clouds in galactic winds include radi-ation pressure of starlight on dust grains (e.g., Murrayet al. 2011; Thompson et al. 2015) and cosmic rays (e.g.,Ipavich 1975; Everett et al. 2008; Socrates et al. 2008;Booth et al. 2013).M82 is the prototype of a starbursting galaxy (Riekeet al. 1980) with a strong, multiphase wind (Leroy et al.
Corresponding author: Laura A. [email protected] ◦ : McKeith et al. 1995), mak-ing it well-suited to observe the biconical outflows alongits minor axis. The outflows of M82 have been stud-ied across the electromagnetic spectrum, tracing the (cid:46)
100 K atomic H i and molecular gas (Walter et al.2002; Salak et al. 2013; Beir˜ao et al. 2015; Martini et al.2018), the ∼ K warm-ionized gas in H α (McKeithet al. 1995; Westmoquette et al. 2009), the ∼ hotgas in X-rays (Watson et al. 1984; Bregman et al. 1995;Strickland et al. 1997), and the entrained dust in theUV, IR, and sub-mm (Hoopes et al. 2005; Leeuw & Rob-son 2009; Kaneda et al. 2010; Roussel et al. 2010). De-tailed comparison of the multi-phase wind has revealedthat the H i and CO confine the hot outflow (Leroy et al.2015), and the H α is well correlated with the diffuse X-rays, with the latter tending to be upstream or interiorto the H α (Shopbell & Bland-Hawthorn 1998; Lehnertet al. 1999; Heckman & Thompson 2017).Many past X-ray studies of M82 have aimed to mea-sure the properties of the hot gas that entrains the coldercomponents. Several works (Ptak et al. 1997; Tsuruet al. 1997; Umeda et al. 2002) analyzed ASCA data a r X i v : . [ a s t r o - ph . H E ] J un LOPEZ ET AL. and found the best-fit model is comprised of two thermalplasmas with enhanced α elements relative to Fe. Cappiet al. (1999) found similar results using BeppoSAX data.Read & Stevens (2002) analyzed
XMM-Newton
Reflec-tion Grating Spectrometers (RGS) data and found near-solar abundances of Fe and O and supersolar abundancesof Mg and S. Origlia et al. (2004) analyzed
Chandra ob-servations as well as
XMM-Newton
RGS data, and theyfound that the hot gas in the nuclear region of M82 wasenhanced in α elements relative to Fe.Strickland & Heckman (2007) analyzed Chandra and
XMM-Newton observations, focusing on hard ( > L − X ∼ . × erg s − arises fromdiffuse gas and that the continuum likely has a non-thermal component. Strickland & Heckman (2009) con-sidered the same X-ray data to constrain the efficiencyof SN thermalization and the mass loading, restrictingtheir analysis to the central 500 pc of M82. Using theirX-ray spectral results as inputs for hydrodynamical sim-ulations, they found evidence of high thermalization ef-ficiency and mild mass loading.While extensive X-ray studies have been conducted,most work focuses on the nuclear region of M82 or onthe integrated spectra from the disk and halo together.Two exceptions are the works of Ranalli et al. (2008)and Konami et al. (2011). Ranalli et al. (2008) per-formed a detailed analysis of how the hot plasma prop-erties vary along M82’s minor axis using a 73-ks XMM-Newton observation. They showed a two-temperatureplasma was necessary, with the hot component temper-ature (of ∼ ≈ × K) staying relatively con-stant with distance from the disk, whereas the warm-hotcomponent decreased from 0.53 keV ( ≈ × K) to0.35 keV ( ≈ × K) from the midplane to ± Suzaku observations. They extracted spectra fromthe M82 disk and three regions north of the disk, andthey modeled the data as a multi-temperature plasma.Their best-fit models included a softer component (oftemperature ∼ Chandra observations of M82. We construct high-resolution im-ages of the diffuse X-ray emission, and we explore howthe hot plasma properties change along M82’s minor axisout to ± Chandra data have been analyzed previ-ously to study an ultra-luminous X-ray source (ULX;
Table 1.
Chandra
ObservationsObsID Exposure UT Start Date361 33 ks 1999-09-201302 16 ks 1999-09-202933 18 ks 2002-06-1810542 119 ks 2009-06-2410543 118 ks 2009-07-0110544 74 ks 2009-07-0710545 45 ks 2009-07-0710925 95 ks 2010-07-2811800 17 ks 2010-07-20
Brightman et al. 2016), to constrain the progenitor ofSN 2014J (Nielsen et al. 2014), and to consider the con-tribution of charge exchange to the soft X-ray flux ofM82 (Zhang et al. 2014). The
Chandra data consideredhere are ∼ × deeper than those analyzed by Origliaet al. (2004) and Strickland & Heckman (2009), and thesuperb spatial resolution of Chandra (with an on-axispoint spread function [PSF] of 0.492 (cid:48)(cid:48) ) facilitates reli-able removal of point sources to study the faint diffuseX-rays of the M82 biconical outflows.The paper is structured as follows. In Section 2, weoutline the
Chandra observations and the analysis toproduce images and spectra of the diffuse emission. InSection 3, we present the results, focusing on the gra-dients in the temperature and metal abundances as afunction of distance from the starburst. We find broadagreement with previous works on the temperature anddensity of the central region, but our abundance profilesin the outflows differ qualitatively and quantitativelyfrom those of Ranalli et al. (2008). In Section 4, wediscuss the implications regarding the mass loading andchemical enrichment of the outflows. In Section 5, wesummarize our conclusions and paths for future work.We adopt a distance of 3.6 Mpc to M82 throughout thispaper; in this case, 1 (cid:48) ≈ OBSERVATIONS AND DATA ANALYSISWe analyze nine
Chandra
ACIS-S observations from1999 (PI: G. Garmire), 2002 (PI: D. Strickland), and2009–2010 (PI: D. Strickland) listed in Table 1 that to-tal 534 ks. We reduced the data using the
Chandra
Interactive Analysis of Observations ( ciao ) Version 4.7and produced exposure-corrected images using the ciao command merge obs . Read-out streaks from ULX M82X-1 were removed using the ciao command acisread-corr . Point sources were identified with wavdetect on https://cxc.harvard.edu/ciao/threads/acisreadcorr ot Gas Gradients in M82 Figure 1.
Exposure-corrected broad-band (0 . − . (cid:38) ∼
12 kpc north of the disk (the diffusesubstructure at the top right of the image). North is up, andEast is left. the merged, broad-band (0 . − . mkpsfmap .These point sources were subsequently removed with dmfilth to construct the image of M82’s diffuse gas (Fig-ure 1).As shown in Figure 1, the diffuse gas extends (cid:38) ∼
12 kpc north of M82 (Lehnert et al.1999; Tsuru et al. 2007). Figure 2 is a three-color im-age of M82, with 8- µ m in red (Engelbracht et al. 2006),H α in green (Kennicutt et al. 2008), and the broad-band X-rays in blue. The H α is well correlated with thediffuse X-rays, with the latter being upstream or inte-rior to the H α , as noted by previous works (Shopbell &Bland-Hawthorn 1998; Lehnert et al. 1999; Heckman &Thompson 2017).To assess the conditions of the diffuse hot gas, weextracted spectra using the ciao command specextract from a 0.2 (cid:48) × (cid:48) region aligned with the M82 major axis Figure 2.
Three-color image of M82, with 8- µ m in red (En-gelbracht et al. 2006), H α in green (Kennicutt et al. 2008),and 0 . − (centered on the position of the M82 nucleus and ori-ented along the starburst ridge: Lester et al. 1990) aswell as ten 0.5 (cid:48) × (cid:48) regions (five north and five southof the disk; as shown and labeled in Figure 3). In ourspectral analysis, we considered only the six 2009–2010observations (totaling 467 ks) to limit systematic dif-ferences arising from different chip configurations andtemporal variations in the years between observations.We excluded all point sources identified by wavdetect within our regions. Background spectra were extractedfrom a 1 (cid:48) × (cid:48) region, either 6 (cid:48) northeast or 5 (cid:48) south ofthe disk depending on the roll angle of the observation,and subtracted from the source spectra.The background-subtracted spectra were modeled us-ing XSPEC Version 12.10.1 (Arnaud 1996). We fit thedata from each observation jointly by including a mul-tiplicative factor (with the XSPEC component const )that was allowed to vary while all other model param-eters were required to be the same between observa-tions. We included two absorption components (withthe XSPEC components phabs and vphabs , respec-tively): one to account for the Galactic absorption of N H = 4 . × cm − (Dickey & Lockman 1990) in thedirection toward M82 and another to represent M82’s in-trinsic absorption N M82H which was allowed to vary andhad abundances of Z (cid:12) (Origlia et al. 2004). We adoptedcross-sections from Verner et al. (1996) and solar abun-dances from Asplund et al. (2009). LOPEZ ET AL. − . . C oun t s s − k e V − Energy (keV)
Disk − − . . C oun t s s − k e V − Energy (keV) − − . . C oun t s s − k e V − Energy (keV)
SouthNorth N e X , Fe L M g X I M g X II S i X III S i X I V S
XVS XV I A r XV II C a X I X Fe XXV
N5N4N3N2N1S1S2S3S4S5 D
Figure 3.
Spectra were extracted from the 11 regions in the middle panel. Region D (in orange; aligned with the starburstingridge) is 0.2 (cid:48) × (cid:48) in size, and the outflow regions are 0.5 (cid:48) × (cid:48) in size (recall at the distance of M82, 1 (cid:48) ≈ Left : Combinedspectrum from region D. Prominent emission lines are labeled; O viii (at ≈ N M82H = (7 . ± . × cm − there. Right : Combined spectra from the North (top) and South (bottom)outflows. Spectra are plotted in the same color as the region box (middle panel) denoting where the data were extracted.
We began by fitting the spectra with a single, ab-sorbed optically-thin thermal plasma component in col-lisional ionization equilibrium (CIE) with variable abun-dances ( vapec ; Foster et al. 2012) and temperature T .However, we found large residuals associated with lineemission (particularly Mg xii and Si xiv ) and at hard( > power-law ) component statistically significantly improved thefits in all 11 regions. F-tests also demonstrated thata second, hot thermal plasma component with temper-ature T was necessary in 8 of the 11 regions. In allregions, the photon index Γ was frozen to Γ = 1 . − .
5. We notethat in their fit of the central M82 region, Ranalli et al.(2008) did not exclude point sources and found a best-fitΓ = 1 . +0 . − . .In addition to the model components described above,we included the AtomDB charge-exchange (CX) modelcomponent vacx to account for line emission pro-duced when ions capture electrons from neutral material (Smith et al. 2012). A CX component is necessary be-cause previous studies of XMM-Newton
RGS data havedemonstrated that ∼
25% of the 0 . − vii , Ne ix , andMg xi originates from CX (Liu et al. 2011). We tiedthe individual abundances together in the CX and twothermal components, and we let the abundances of met-als with detected emission lines (O, Ne, Mg, Si, S, andFe) vary. The abundances of metals not detected in the0 . − Z (cid:12) , consistent with theM82 disk metallicity (Origlia et al. 2004). Given the el-evated N M82H in the three central regions S1, D, and N1,the detection of O was limited there, and we froze theO abundance to solar metallicity in those locations.Putting all of the model components together, thecomplete XSPEC model for regions S1–S4, and N1–N3was: const*phabs*vphabs*(vapec+vapec+vacx+powerlaw) . The models for the southern-most andtwo northern-most outflow regions (S5, N4–N5) did notinclude the second vapec component as they did notstatistically improve the fits. As detailed below, a third vapec component was added to region D to account forthe detected Fe xxv line. RESULTSFigure 3 shows the extracted spectra from the 11 re-gions, with data from all six observations combined us- ot Gas Gradients in M82 Table 2.
Spectral Fit Results a Reg. r b N M82H kT kT O/O (cid:12) c Ne/Ne (cid:12)
Mg/Mg (cid:12)
Si/Si (cid:12) d S/S (cid:12) d Fe/Fe (cid:12) χ /d.o.f.(kpc) ( × cm − ) (keV) (keV)N5 2.5 < ± +0 . − . +1 . − . +0 . − . +0 . − . < ± +0 . − . +1 . − . +0 . − . +0 . − . < ± +0 . − . +0 . − . +1 . − . +1 . − . +0 . − . +0 . − . ± ± +0 . − . +0 . − . +0 . − . ± +0 . − . ± +0 . − . ± ± +0 . − . ± ± ± ± ± e ± ± +0 . − . ± +0 . − . +0 . − . +0 . − . +0 . − . − ± ± +0 . − . ± ± ± ± ± − ± ± +0 . − . ± ± ± +0 . − . ± +0 . − . − ± +0 . − . +0 . − . ± +0 . − . ± +0 . − . < +0 . − . − < +0 . − . +0 . − . +0 . − . +0 . − . +0 . − . +0 . − . < +0 . − . − +0 . − . ± +0 . − . +1 . − . +0 . − . +0 . − . a Results reported here are the best-fit parameters of the vphabs and vapec components of the XSPEC model described in Section 2. b r is the distance from the starburst ridge, along the minor axis. Positive values are North of the midplane; negative values are South. c High N M82H in regions D, N1, and S1 precludes detection of the oxygen lines. Thus, the oxygen abundance for these regions is frozento solar values. d In regions where Si and S lines are not statistically significantly detected, the abundances are frozen to solar values. e A third vapec component not listed here was added to the region D model to represent the very hot component. The best-fittemperature of that component was kT = 6 . +1 . − . keV. ing the ciao command combine spectra . In the spec-trum from the disk (region D), prominent emissionlines are evident from Ne, Mg, Si, S, Ar, Ca, and Fe.Among these features, a Fe xxv line is apparent at ≈ +0 . − . keV line (based on a Gaussian fit to thatfeature) that is not detected in the outflow regions, con-sistent with Strickland & Heckman (2007) who foundthat it extends <
100 pc along the M82 minor axis. Toaccount for the Fe xxv line, we added a third vapec component to the region D model that yielded a best-fitvery hot temperature of kT = 6 . +1 . − . keV.The outflow spectra show emission lines from O, Ne,Fe L, Mg, Si, and S. The spectra from region S1 has thestrongest signal, even moreso than region D, but S1 stilllacks the Fe xxv line. Regions N1, D, and S1 appear tohave significant absorption, precluding the detection ofsoft X-ray lines, e.g. O viii . However, at larger r , the ab-sorption diminishes, and an O viii feature at ≈ Note that we show the combined spectra as a qualitativedemonstration of the detected features. However, in our spec-tral analysis, we opted to fit the six observations simultaneously(i.e., without combining the data) due to uncertainties intro-duced by the combining process. See the caveats outlined athttps://cxc.harvard.edu/ciao/ahelp/combine spectra.html. the hard X-rays (above 2 keV) are less significantly de-tected in the outer regions, especially S5 and N5.Spectra were fit jointly using the model described inSection 2, and the results are listed in Table 2. The fitsyielded reduced χ values of 0 . − .
65 with 800 − − number of free parameters). Figure 4 shows the best-fitparameters for the 11 regions plotted as a function of r ,the distance from the starbursting ridge. We find N M82H and the warm-hot T and hot T temperature compo-nents peak in the center and are lower at greater dis-tances to the north and the south.The Si and S abundances are also elevated in the cen-tral regions and level out to solar or slightly sub-solarvalues ∼ (cid:12) and 0.42 Fe (cid:12) , respectively. The O and Ne abundancesare uncertain and have larger error bars arising from theassociated soft X-ray lines being attenuated by N M82H and the Ne features being blended with Fe L lines. Thus,it is challenging to discern the trends in the O and Neabundances, but they may be constant. As seen in Fig-ure 3, prominent O viii lines are apparent in the north-ern outflow spectra but absent in the southern spectra.Based on our best-fit models, we deduce that this dis-parity arises because of temperature differences in the apec components that lead to less thermal continuum
LOPEZ ET AL. -3 -2 -1 0 1 2 3
Distance from Starburst (kpc) N H M ( × c m - ) -3 -2 -1 0 1 2 3 Distance from Starburst (kpc) O / O -3 -2 -1 0 1 2 3 Distance from Starburst (kpc) S i / S i ⊙⊙ -3 -2 -1 0 1 2 3 Distance from Starburst (kpc) N e / N e -3 -2 -1 0 1 2 3 Distance from Starburst (kpc) S / S -3 -2 -1 0 1 2 3 Distance from Starburst (kpc) k T ( ke V ) -3 -2 -1 0 1 2 3 Distance from Starburst (kpc) k T ( ke V ) T ( x K ) ⊙⊙ -3 -2 -1 0 1 2 3 Distance from Starburst (kpc) M g / M g -3 -2 -1 0 1 2 3 Distance from Starburst (kpc) k T ( ke V ) -3 -2 -1 0 1 2 3 Distance from Starburst (kpc) k T ( ke V ) T ( x K ) -3 -2 -1 0 1 2 3 Distance from Starburst (kpc) F e / F e ⊙⊙ N HM82 kT kT O Ne MgSi S Fe -3 -2 -1 0 1 2 3 r (kpc) S X ( e r g k p c - s - ) TotalWarm-HotHotVery HotPower LawCX-3 -2 -1 0 1 2 3 r (kpc) S X ( e r g k p c - s - ) TotalWarm-HotHotVery HotPower LawCX-3 -2 -1 0 1 2 3 r (kpc) S X ( e r g k p c - s - ) TotalWarm-HotHotVery HotPower LawCX -3 -2 -1 0 1 2 3 r (kpc) S X ( e r g k p c - s - ) TotalWarm-HotHotVery HotPower LawCX-3 -2 -1 0 1 2 3 r (kpc) S X ( e r g k p c - s - ) TotalWarm-HotHotVery HotPower LawCX-3 -2 -1 0 1 2 3 r (kpc) S X ( e r g k p c - s - ) TotalWarm-HotHotVery HotPower LawCX -3 -2 -1 0 1 2 3 r (kpc) S X ( e r g k p c - s - ) TotalWarm-HotHotVery HotPower LawCX-3 -2 -1 0 1 2 3 r (kpc) S X ( e r g k p c - s - ) TotalWarm-HotHotVery HotPower LawCX-3 -2 -1 0 1 2 3 r (kpc) S X ( e r g k p c - s - ) TotalWarm-HotHotVery HotPower LawCX
Figure 4.
Best-fit parameters for the 11 regions plotted as a function of r , the distance from the M82 major axis. A distanceof zero is defined as the location of the M82 nucleus (Lester et al. 1990). Negative distances are toward the South, and positivedistances are toward the North. Filled circles represent measurements, upside-down triangles are upper-limits, and open squaresare fixed to solar values because associated lines are not well constrained. Abundances are relative to solar by number, anderror bars are 90% confidence intervals. and greater equivalent widths in the O viii in the north-ern outflow compared to the south.As noted in Section 2, we included a power-law com-ponent with Γ = 1 . − S5 and N1 − N5), the addition of a third thermalcomponent to our fits (rather than a power-law) yieldedunphysical best-fit temperatures of kT = 10 −
70 keV.Moreover, F-tests favored a power-law component overa very hot thermal component with > kT = 6 . +1 . − . keV was included, the power-law improved the fit with 96% confidence. We note that asteeper power-law with Γ = 2 . − . − . S X as afunction of r in Figure 5 (left panel). To compute S X ,we measured the absorption-corrected X-ray luminosity L X of each component in the best spectral fits, and wedivided by the area of the 11 regions. The total S X and all of the individual components peak centrally anddecline with increasing r . Figure 5 (right panel) showsthe relative contribution of the model components. The ot Gas Gradients in M82 -3 -2 -1 0 1 2 3 r (kpc) S X ( e r g k p c - s - ) TotalWarm-HotHotVery HotPower LawCX -3 -2 -1 0 1 2 3 r (kpc) F r ac t i on o f S X Figure 5. Left : Broad-band (0 . − . S X profile of the spectral model components along theM82 minor axis. S X is computed as the absorption-corrected luminosity of each component divided by the area of each region:0.66 kpc for region D and 1.65 kpc for the outflow regions. Right : Fraction of the total S X contributed by the spectral modelcomponents. The warm-hot component dominates, producing ≈ −
60% of the total S X , and the other components produce ≈ −
26% of the total S X . warm-hot plasma is the dominant component, produc-ing ≈ −
60% of the total S X , whereas the other com-ponents produce ≈ −
26% of the total S X .In particular, we find that the CX component con-tributes 7 −
22% of the total broad-band (0 . − . ≈ − N M is greatest in these regions, attenuatingsoft X-rays where CX features are best detected.We estimate the electron number density n e ofthe thermal plasmas using the best-fit normaliza-tions norm of the apec components, since norm =(10 − EM ) / πD , where D is the distance, EM = (cid:82) n e n H dV is the emission measure, and n H is thehydrogen number density. Setting n e = 1 . n H andintegrating over the volume V , then n e = (1 . × normD /f V ) / , where f is the filling factor (here,we assume f = 1). To compute V for each region, weassume a cylindrical geometry and a height equal to theminor axis side of the rectangular regions. To estimatethe radius of the cylinder R cyl , we produced surface-brightness profiles of the broad-band (0 . − (cid:48)(cid:48) slits along themajor axis. We define R cyl as half of the size thatenclosed 99% of the emission.In Table 3, we list the best-fit norm , norm , and norm of the warm-hot, hot, and very hot components,respectively, as well as R cyl , V , n e , , n e , , and n e , . We find that the densities peak in region D (with n e , =1 . × − cm − and n e , = 7 . × − there) and fallwith r . We also compute the thermal pressure P =2 n e kT and the radiative cooling time t cool = 3 kT / Λ n e of the three components, as listed in Table 3. Λ is theradiative cooling function in units of erg s − cm , and wecalculate Λ at solar metallicity assuming an optically-thin thermal plasma in CIE (as in Figure 1 of Rosenet al. 2014 using chianti ; Dere et al. 1997).The warm-hot and hot thermal pressure P /k and P /k peak in region D, and these quantities decreasealong the minor axis. The very hot thermal pressure P /k is the dominant term in region D compared tothe warm-hot and hot components, with P /k = 1 . × K cm − . This value is consistent with the estimatesfrom past X-ray studies (Bregman et al. 1995; Strickland& Heckman 2009) as well as the central pressure mea-surements of the ionized gas from optical spectroscopy(Heckman et al. 1990; Smith et al. 2006; Westmoquetteet al. 2007).Regions D and S1 have the shortest t cool , of ∼
30 Myr,and the timescale increases to ∼
100 Myr ≈ t cool , is lowest in region S1 at ≈
83 Myr and approaches ∼
400 Myr in the outer re-gions; t cool , in region D is the longest of all of the com-ponents, with t cool , ≈
500 Myr. In all regions, t cool of the three components is longer than the advectiontimescale t adv ∼ r/v (where v is the velocity) in themodels discussed in Section 4.2, indicating the windsare not radiative. LOPEZ ET AL.
Table 3.
Physical Parameters of the Disk and Outflow Regions a Reg. norm norm R cyl V n e , n e , P /k c P /k c t cool , t cool , ( × cm) ( × cm − ) (cm − ) (cm − ) (K cm − ) (K cm − ) (Myr) (Myr)N5 1.6 × − – 4.8 11.9 1.6 × − – 1.4 × – 121 –N4 3.8 × − – 4.8 11.9 2.4 × − – 2.1 × – 75 –N3 1.8 × − × − × − × − × ×
113 399N2 6.5 × − × − × − × − × ×
65 156N1 2.0 × − × − × − × − × ×
50 397D 1.7 × − × − × − × − × ×
31 153S1 3.7 × − × − × − × − × ×
30 83S2 6.5 × − × − × − × − × ×
70 182S3 4.9 × − × − × − × − × ×
76 233S4 2.0 × − × − × − × − × ×
130 356S5 4.1 × − – 4.8 11.7 2.6 × − – 2.3 × – 77 – a The parameters of the very hot component kT in region D are: norm = 1 . × − , n e , = 9 . × − cm − , P /k =1 . × K cm − , and t cool , = 501 Myr. b The normalizations are defined as norm = (10 − EM ) / πD , where EM = (cid:82) n e n H dV . Thus, columns 2 and 3 are in units of10 − cm − . 4. DISCUSSION4.1.
Comparison to Previous Work
A similar analysis was conducted previously byRanalli et al. (2008) who analyzed a 73-ks
XMM-Newton observation of M82 and measured how the temperatureand abundance pattern varied (cid:46) (cid:48) diame-ter. They fit a two-temperature plasma model and founda very hot component of ∼ ≈ × K that wasrelatively constant along the minor axis. Their warm-hot component decreased from 0.53 keV ( ≈ × K)to 0.35 keV ( ≈ × K) from the center to their outerregions. Ranalli et al. (2008) found that several elements(O, Ne, Mg, and Fe) were substantially more abundantin the outflows than in the disk: e.g., O and Ne were ∼ × more abundant ∼ ∼ × enhancement in the outflows rela-tive to the disk), whereas S was most abundant in thecentral regions (though large errors precluded reliablemeasurements in the outer regions).In addition to the two thermal plasmas in their modelof the central region’s spectra, Ranalli et al. (2008)included a power-law component (with spectral indexΓ = 1 . +0 . − . ) to account for point sources as well asGaussian functions at 0.78 keV and 1.2 keV to representtwo spectral lines associated with CX (we note that aphysical model of CX emission was not publicly available when that work was completed). Ranalli et al. (2008)did not include the power-law component in their fits tothe outflow spectra, based on the reasoning that pointsources were resolved sufficiently to be excluded fromthe extraction regions.We find a similar warm-hot temperature profile andvery hot temperature in the starburst as Ranalli et al.(2008), but the abundance profiles between the twoworks are different. Thus, it is worth considering whythese discrepancies have arisen. In contrast to Ranalliet al. (2008), we have included a CX and power-law com-ponent in the outflow spectral models in this work. Theformer is necessary for reliable measurements of O, Ne,and Mg abundances as the CX flux contributes ∼ ix and Ne x at energies ≈ . − . XMM-Newton
RGS data using an absorbed apec +CX model,and their values were consistent (within the errors) ofour region D values. Thus, the different abundance pro-files (particularly O, Ne, Mg, and Fe) from this workand Ranalli et al. (2008) likely arise from our inclusionof a CX component.Unlike Ranalli et al. (2008), we only find the very hotcomponent (with kT = 6 . +1 . − . keV) in the M82 cen-tral region, and F-tests show with > xxvot Gas Gradients in M82 -3 -2 -1 0 1 2 3 r (kpc) T ( K ) data6data12data14R = 100 pcR = 200 pcR = 300 pcR = 400 pcData Figure 6.
Flux-weighted hot-gas temperature T (left) and density n e (right) profile of M82 (dashed black lines) com-pared to four adiabatic model predictions (solid lines) with R = [100 , , , α = [0 . , . , . , . β = [0 . , . , . , . T ∝ r − / and n e ∝ r − , and these profiles are steeperthan are observed and may suggest mass loading in the hot superwind. (associated with the very hot component) only extends <
100 pc along the M82 minor axis. A likely explanationfor these discrepant results is that our power-law com-ponent (which was not included in the Ranalli et al. fits)accounted for the hard X-ray flux such that a very hotcomponent was not necessary. F-tests confirmed thata power-law with photon index Γ = 1 . xxv line in the outflow spectra (see the rightpanels of Figure 3) – which was also noted by Ranalliet al. (2008) from their XMM-Newton outflow spectra –reinforce the possibility that the hard X-ray emission isnon-thermal in nature.The other previous study that explored how the hotgas temperature and abundance profiles vary in the M82outflows was conducted by Konami et al. (2011) using101 ks of
Suzaku observations. They reported that two-or three-temperature plasma components (including asoft ∼ . − xxv line. Additionally, Konami et al. (2011) found nospatial variation in the abundance ratios (O/Fe, Ne/Fe,Mg/Fe) of the hot plasmas, similar to our results.4.2. Comparison to Wind Models
The profiles presented in Section 3 can be compared tosuperwind model predictions to constrain outflow prop-erties. Chevalier & Clegg (1985) developed a simplespherically-symmetric wind model to explain the ex-tended X-ray emission of M82, whereby SNe inject massand energy at rates ˙ M and ˙ E , respectively, in a re-gion of size R . The energy injection rate is ˙ E = α ˙ E SN (where ˙ E SN is the energy injection from SNe, ∼ ergper 100 M (cid:12) of star formation), and the mass injec-tion rate is ˙ M = β ˙ M ∗ (where ˙ M ∗ is the star forma-tion rate). By energy conservation (neglecting radia-tive cooling and gravity; see Thompson et al. 2016 andreferences therein for the effects of radiative cooling),the asymptotic velocity of the v hot , ∞ = (2 ˙ E/ ˙ M ) / (cid:39) ( α/β ) / km s − and the temperature at r = R is T hot = ( m p /k )(3 / v , ∞ (cid:39) × ( α/β ) K.Based on their observational measurements from hardX-ray lines detected in the central 500 pc of M82, Strick-land & Heckman (2009) found a central temperature of T c = (3 − × K, central density of n c ∼ . − ,and central pressure of P c /k = (1 − × K cm − .From these values, assuming an injection region of R = 300 pc and ˙ M ∗ = 6 M (cid:12) yr − , they estimated0 . ≤ α ≤ . . ≤ β ≤ . Our results forthe best-fit parameters associated with the very hotcomponent in region D are consistent with these pre- Note that we have converted Strickland & Heckman (2009)’sestimates using our definitions of the mass loading and thermal-ization parameters for consistency. LOPEZ ET AL. vious findings. Assuming ˙ M ∗ = 6 M (cid:12) yr − and aninjection region of R = [100 , , , α = [0 . , . , . , . β = [0 . , . , . , . R = 300 pc, thesenumbers reflect relatively high values of the thermaliza-tion efficiency and low values of the mass loading.The numbers change significantly if we instead fit tothe flux-weighted values of the temperature and den-sity in the core. Figure 6 shows the flux-weighted tem-perature T and density n e profiles of our data and offour Chevalier & Clegg (1985) models tuned to the cen-tral values with α = [0 . , . , . , . β =[0 . , . , . , . α and β relative to those derived from only thehottest component in the core result from the fact thatthe warm-hot component dominates in the flux-weightedtemperature and density (see Figure 5).Outside this central region (for radii r > R ), theChevalier & Clegg (1985) model assumes that the out-flowing gas experiences adiabatic expansion such that T ∝ r − / (for an adiabatic index γ = 5/3), n ∝ r − ,and pressure P ∝ r − / once the asymptotic velocityof the wind has been reached. Near the core region R , where the flow accelerates, the predicted profiles aresteeper. As shown in Figure 6, the data diverge substan-tially from the models > Metal Abundance Profiles
As shown in Figure 4 and discussed in Section 3, theO, Ne, Mg, and Fe abundances are relatively constantalong the minor axis. While the O and Mg abundancesare solar metallicity, Ne is about 1.5 × solar. These val-ues are consistent with the abundances measured in M82stars and H ii regions (Achtermann & Lacy 1995; Origliaet al. 2004). The Si and S are enhanced to 2.5 × and3.5 × solar, respectively, in the central 500 pc, indicativeof enrichment from SN ejecta there. These abundances level out at greater distances from the midplane, sug-gesting the hot wind fluid is mixing with and sweepingup colder, lower-metallicity ISM material. This expla-nation is consistent with mass loading leading to broadtemperature and density profiles (see Figure 6).Three-dimensional hydrodynamical simulations of su-perwinds by Melioli et al. (2013) predict that the hotgas component can be metal-rich with average metal-licities of ¯ Z ∼ . Z (cid:12) , whereas the colder, denser mate-rial will maintain solar metallicity. Melioli et al. (2013)suggested that a large fraction of the ejected metals isretained in and around the galactic disk, with ∼ ≈
500 pc from the starbursting ridge.The stellar disks of spiral galaxies are believed to besurrounded by the CGM out to the virial radius (e.g.Oppenheimer et al. 2016). The CGM, which is be-lieved to be the largest baryon reservoir in a galaxy, isdominated by warm-hot gas of temperatures ∼ K.Highly ionized metal lines are detected from the warm-hot CGM, indicating a metal-rich CGM (e.g., Guptaet al. 2012 and references therein) that originates fromthe stellar disk and is ejected in galactic outflows. Theobservations reported here show that metals are beingexpelled in these superwinds. Our results are consistentwith the super-solar Ne/O and Ne/Fe ratios detected inthe Milky Way CGM (Das et al. 2019a,b) and supportthe direct connection between the wind and the CGM.4.4.
Charge Exchange
As described in Section 2, we include a CX componentin our spectral fits that contributes 7 −
22% of the totalbroad-band flux (see Figure 5) in our 11 regions. Previ-ous work using high-resolution X-ray spectroscopy hasmeasured the CX contribution to M82’s X-ray emissionoverall. Using
XMM-Newton
RGS data, Ranalli et al.(2008) found that the O lines could not be accountedfor properly with a multi-temperature thermal plasmamodel, suggesting a significant flux from CX. Liu et al.(2011) quantified the CX contribution to K α triplets ofHe-like O, Ne, and Mg: 90, 50, and 30%, respectively.Zhang et al. (2014) analyzed XMM-Newton
RGS spec-tra of the central outflow and EPIC-pn spectra of theCap, adopting a physical model (from Smith et al. 2012)that included one thermal plasma and a CX component.They found that ≈
25% of the flux from 0 . − ± ot Gas Gradients in M82 Figure 7.
Exposure-corrected image of the diffuse hard X-ray emission in the 4 − − σ = 3 pixels. Whiteboxes denote the 11 regions where spectra were extracted(see Figure 3). The emission peaks in the starburst ridgeand falls off with distance along the minor axis. North is up,and East is left. the mixing and heating of cool gas into the warm-hotphase persists out to large distances.4.5. Non-thermal Emission
As described in Section 3, a power-law componentis necessary to adequately fit the spectra in all 11 re-gions. The combined luminosity of the 11 regions fromthe power-law component is L X , PL = (5 . ± . × erg s − , ≈
14% of the total L X in the 0 . − − − L X , PL shouldbe viewed as upper-limit on a diffuse non-thermal com-ponent.The presence and nature of diffuse hard X-ray emis-sion in M82 is debated. Using the first Chandra ob-servations of M82, Griffiths et al. (2000) attributed thediffuse hard X-rays detected in the nuclear region tothermal bremsstrahlung from a ∼ × K plasma be-cause of the detected Fe xxv line. Strickland & Heck- man (2007) also examined the diffuse hard X-rays in thecentral 500 pc of M82 using
Chandra and
XMM-Newton data. They concluded that 20 −
30% of the emissionwas truly diffuse and that the continuum was better fitas a power-law with a photon index of Γ = 2 . − XMM-Newton observations (as describedin Section 4.1), Ranalli et al. (2008) used a power-lawcomponent to account for unresolved X-ray binaries, butthis component was not included in their models of theoutflow region spectra.While some of the emission may arise from pointsources, the power-law component in the outflows mayarise from diffuse non-thermal X-rays. Past studies haveproposed that inverse-Compton (IC) scattering of IRphotons by relativistic electrons may produce significantdiffuse hard X-rays in M82 (Schaaf et al. 1989; Moran &Lehnert 1997). In this scenario, the intense IR emissionof the starburst up-scatters target photons to ∼
100 MeVenergies, and the spectrum extends down to X-ray en-ergies with a hard photon index of Γ ≈ . − .
5. Lacki& Thompson (2013) suggest that the non-thermal dif-fuse hard X-ray emission is produced by synchrotron.Although with substantial uncertainties, in their fidu-cial models they find that IC and synchrotron are ableto produce only ∼
1% and ∼ galprop modelsfrom Buckman et al. (2020) predict that ≈ −
90% ofthe non-thermal core and halo emission is from IC inthe 1 −
10 keV band, but the total luminosity falls shortof that observed by a factor of order ∼ CONCLUSIONSWe analyze deep
Chandra observations of M82 to pro-duce images and spectra of the diffuse X-ray emission ofthe starburst and outflows. Based on fits to the spectrafrom 11 regions up to ± >
500 pcaway. By contrast, the O, Ne, Mg, and Fe abundancesare relatively constant between the starburst and theoutflows, indicating effective transport of stellar diskmetals to the CGM. We compare the observed tempera-ture and gas density profiles to superwind model predic-tions, and we show that these profiles are much shallowerthan expected for adiabatic expansion (see Figure 6).2
LOPEZ ET AL.
This result suggests that the hot winds are being massloaded due to the mixing and heating of cooler gas intothe hot phase.In addition to the thermal components in our spec-tral models, we find it is necessary to include charge ex-change and a power-law component in all of the regionsconsidered in this work. The CX contributes 7 −
22% tothe total broad-band flux (consistent with the work ofZhang et al. 2014), with the smallest contribution within ∼
500 pc of the M82 nucleus. However, the high intrin-sic column density there attenuates soft X-rays whereCX features are best detected. We also find that a hardpower-law component is required throughout the cen-tral and outflow regions. This power-law componentaccounts for ≈
14% of the total broad-band flux in the11 regions, and the spectral fits favored a shallow pho-ton index of ∼ XRISM mission; Tashiro et al. 2018) will facilitate measurementsof the hot gas velocity v hot , which will provide strongconstraints on the energy content and mass loading ofthe hot wind. When calorimeters achieve arcsecond spa-tial resolution, as in the proposed Lynx
X-ray Observa-tory (The Lynx Team 2018), they will enable measure-ments of the metal abundances and velocities along theminor axes of starburst-driven winds in many galaxies.Hard X-ray sensitivity (at ∼ xxv in the starburst cores, andsoft X-ray capabilities (below 1 keV) are vital to probethe interplay of warm-hot and cooler gas in the galaxyoutflows (Hodges-Kluck et al. 2019).We thank Adam Leroy, Paul Martini, David Wein-berg, and the OSU Galaxy/ISM Meeting for useful dis-cussions. LAL is supported by a Cottrell Scholar Awardfrom the Research Corporation of Science Advance-ment. DDN and TAT are supported in part by NationalScience Foundation Grant Software:
CIAO (v4.7; Fruscione et al. 2006), XSPEC(v12.9.0; Arnaud 1996)REFERENCES