An XMM-Newton Observation of the Local Bubble Using a Shadowing Filament in the Southern Galactic Hemisphere

0.44 MB·English

by  David B. Henley

Draftversion February5,2008
PreprinttypesetusingLATEXstyleemulateapjv.10/09/06
AN XMM-NEWTON OBSERVATION OF THE LOCAL BUBBLE USING A SHADOWING FILAMENT IN THE
SOUTHERN GALACTIC HEMISPHERE
David B. Henley and Robin L. Shelton
Department ofPhysicsandAstronomy,UniversityofGeorgia,Athens,GA30602
and
K. D. Kuntz
HenryA.RowlandDepartmentofPhysicsandAstronomy,Johns HopkinsUniversity,Baltimore,MD21218and
ExplorationoftheUniverseDivision,NASAGoddardSpaceFlightCenter,Code662,Greenbelt,MD20771
Draft versionFebruary 5, 2008
7 ABSTRACT
0
We present an analysis of the X-ray spectrum of the Local Bubble, obtained by simultaneously
0
analyzingspectrafromtwoXMM-Newton pointings onandoffanabsorbingfilamentin the Southern
2
galactic hemisphere (b 45◦). We use the difference in the Galactic column density in these
n ≈ −
two directions to deduce the contributions of the unabsorbed foreground emission due to the Local
a
Bubble, and the absorbed emission from the Galactic halo and the extragalactic background. We
J
find the Local Bubble emission is consistent with emission from a plasma in collisional ionization
9 equilibrium with a temperature log(T /K) = 6.06+0.02 and an emission measure n2dl = 0.018
2 LB −0.04 e
cm−6 pc. Our measured temperature is in good agreement with values obtained from ROSAT All-
1 Sky Survey data, but is lower than that measured by other recent XMM-Newton obseRrvations of the
v Local Bubble, which find log(TLB/K) 6.2 (although for some of these observations it is possible
≈
4 thattheforegroundemissioniscontaminatedbynon-LocalBubbleemissionfromLoopI).Thehigher
3 temperature observed towards other directions is inconsistent with our data, when combined with a
8 FUSE measurementoftheGalactichaloOviintensity. ThisthereforesuggeststhattheLocalBubble
1 is thermally anisotropic.
0 Ourdataareunabletoruleoutanon-equilibriummodelinwhichtheplasmaisunderionized. How-
7 ever,an overionizedrecombining plasma model, while observationallyacceptable for certain densities
0 and temperatures, generally gives an implausibly young age for the Local Bubble (.6 105 yr).
/ ×
h Subject headings: Galaxy: general—Galaxy: halo—ISM: general—ISM: individual (Local Bubble)—
p X-rays: ISM
-
o
tr 1. INTRODUCTION 2004). These models may essentially be divided into
s two classes. In one class of model, the Local Bub-
a The Local Bubble (LB) is a region of X-ray–emitting
ble was carved out of the ambient ISM by a su-
: gasof 100pc extentinwhichthe SolarSystemresides.
v ∼ pernova or series of supernovae (e.g. Cox & Anderson
MeasurementsoftheinterstellarNaiabsorptiontowards
i 1982; Innes & Hartquist 1984; Smith & Cox 2001;
X 456nearbystarsrevealthattheLocalBubbleresidesina
Ma´ız-Apell´aniz 2001; Breitschwerdt & de Avillez 2006).
cavityintheinterstellarmedium(ISM;Sfeir et al.1999).
r
The hot gas thus produced gives rise to the observed
a TheideaoftheLocalBubbleoriginatedinthelate1970s,
X-rays. If the last supernova was recent enough, one
in order to explain the observed anticorrelationbetween
would expect the ions to be underionized. However,
the intensity of the soft X-ray background (SXRB) and
if the Local Bubble is old enough to have begun con-
theGalactichydrogencolumndensityN (Bowyer et al.
H
tracting, the ions will be overionized and recombining
1968; Sanders et al. 1977). The data are inconsistent
(Smith & Cox 2001). In the second class of model, a se-
withtheanticorrelationbeingduetoabsorption,asthey
ries of supernovae in a dense cloud formed a hot bub-
require an interstellar absorption cross-section that is
ble, which burst out of the cloud into the less dense
one-thirdofits expectedvalue (Bowyer et al.1968), and
surroundings and underwent rapid adiabatic cooling
also different energy bands have the same dependence
(Breitschwerdt & Schmutzler 1994; Breitschwerdt 1996;
on N (Sanders et al. 1977; Juda et al. 1991). Instead,
H
Breitschwerdtet al. 1996; Breitschwerdt 2001). In this
aso-called“displacement”modelwasproposed,inwhich
casethe X-rayemissionisdue tothedelayedrecombina-
the hot X-ray–emitting plasma (the Local Bubble) is in
tion of the overionized ions. X-ray spectroscopy of the
the foregroundanddisplacesthe coolgas(Sanders et al.
Local Bubble emission is essential for distinguishing be-
1977;Tanaka & Bleeker1977). IndirectionsofhigherX-
tween the various models, as it enables us to determine
rayintensitytheLocalBubbleisthoughttobeofgreater
the physical properties of the X-ray–emitting gas.
extent,andsothereislesscoolgas(andhencelowerN )
H
Originally,allthe softX-rayfluxwasattributedtothe
in those directions.
LocalBubble,whichmadedeterminingtheLocalBubble
Numerous models have been proposed for the
X-rayspectrumrelativelysimple. However,thediscovery
formation of the Local Bubble (for reviews, see
of shadows in the SXRB with ROSAT (Snowden et al.
Breitschwerdt 1998; Cox 1998; Breitschwerdt & Cox
1991; Burrows & Mendenhall 1991) showed that 50%
∼
Electronicaddress: [email protected] of the SXRB in the 1/4-keV band originated from be-
2 HENLEY, SHELTON, AND KUNTZ
yond the Local Bubble, either in the Galactic halo or
from an extragalactic background. Hence, in order to -4
4
determinetheLocalBubblespectrum,onemustfirstdis-
entangle the contributions of the Local Bubble and the
background. This has been done by modeling ROSAT
All-sky Survey data (RASS; Snowden et al. 1997) with
an unabsorbed foreground component (due to the Lo-
-4
cal Bubble) and absorbed components for the Galac- 6
tic halo and the extragalactic background, using the
100-µm Infrared Astronomical Satellite (IRAS) maps of
Schlegel et al. (1998) as a measure of N . In this way
H
the Local Bubble emission has been mapped out, and
citoslltiseimonpaelraitounriezaetsiotinmaeqteudilitboribuemT(LSBn∼ow1d0e6nKetaasls.u1m9i9n8g, 28-482 280 278 276
2000; Kuntz & Snowden 2000).
It should be noted, however, that the RASS data are
Fig. 1.— ROSAT All-Sky Survey R1+R2 image centered on
presented in just six energy bands between 0.1 and l = 279◦, b= −46◦, showing the absorbing filament used for our
∼
2 keV, several of which overlap. The data therefore observations(datafromSnowdenetal.1997). Thecirclesshowour
∼have fairly poor spectral resolution. The Local Bubble on-filament (upper) and off-filament (lower) pointing directions.
The smaller circles (radius = 14′) show the approximate areas
temperature is inferred from the R1-to-R2 band inten-
from which our XMM-Newton spectra were extracted, while the
sity ratio of the local emission component, as there is largercircles(radius=30′)showtheareasfromwhichourROSAT
very little Local Bubble emission in the higher-energy spectrawereextracted.
bands: theupper limit onthe ROSAT R45intensitydue we compare our results’ prediction of the Galactic halo
to the Local Bubble is a few 105 counts s−1 arcmin−2 O vi intensity with the observed value in 5.2. In 5.3
× § §
(Snowden et al. 1993; Kuntz et al. 1997). However, this wecompareourresultswiththoseofotherXMM-Newton
observational fact may be used to rule out significantly and Chandra observations of the Local Bubble. In 5.4
§
higher temperatures. It should also be noted that we discuss our choice of plasma emission code used in
this inferred temperature is somewhat dependent upon the analysis, and consider the effect this has on our re-
the plasma emission model used (e.g. Kuntz & Snowden sults. Finally in this section we discuss non-equilibrium
2000). models of the Local Bubble: an underionized model in
The large mirror collecting area and CCD detectors 5.5,andarecombiningmodelin 5.6. Weconcludewith
onboard XMM-Newton enable us to obtain high signal- §a summary in 6. Throughout t§his paper we quote 1σ
§
to-noise spectra of the SXRB with greater spectral reso- errors.
lutionthanROSAT.Inparticular,XMM-Newton enables
ustodetectlineemissionintheSXRBspectrum,notably 2. OBSERVATIONSANDDATAREDUCTION
emissionfromOviiandOviiiat 0.55and 0.65keV. The on- and off-filament XMM-Newton observations
∼ ∼
If we can determine how much of this oxygen emission were both carriedout on 2002May 3. The details of the
is due to the Local Bubble, this will enable us to place observations are presented in Table 1.
stronger constraints on TLB. Since our current understanding of the particle back-
We use a shadowing technique to determine the spec- ground of the XMM-Newton PN is relatively poor, and
trum of the Local Bubble emission. We have analyzed the characterization of the background of the XMM-
spectra obtained from XMM-Newton pointings on and Newton MOS cameras is fairly well refined, we have re-
off an absorbingfilament in the southern Galactic hemi- stricted our analysis to the data obtained by the two
sphere. The absorbing filament appears as a shadow MOS cameras. The data were reduced as follows. We
in the SXRB, as shown in Figure 1, which also shows constructed the light curve in the 2.5–8.5 keV band for
our XMM-Newton pointing directions. Penprase et al. the entire field of view. We fitted a Gaussian to a his-
(1998) have estimated the distance of the filament to be togram of the count rates, and set the “quiescent level”
230 30pc. Mapsofthe extentoftheLocalBubble,ob- to the meanof that Gaussian. We removedfrom further
±
tained from RASS data (Snowden et al. 1998) and Na i analysisalltimeperiodsduringwhichthecountratewas
absorptiondata(Sfeir et al.1999;Lallement et al.2003), >3σ above the quiescent level; the higher count rate in
indicate that the boundary of the Local Bubble is 100 those time periods is due to either strong soft proton
∼
pc away in this direction. Thus, the filament is between contamination or an enhanced particle background. Fil-
the Local Bubble and the Galactic halo. We fit spectral tering the data using a lower-energyband (0.4–2.0 keV)
models simultaneously to the on- and off-filament spec- produces results no different from those obtained using
tra,usingthedifferenceintheabsorbingcolumnbetween the above energy band.
thetwopointingdirections(NH =9.6 1020cm−2 versus Sourcesweredetectedinboththe0.3–2.0keVand2.0–
1.9 1020 cm−2; see 3.1)to deduce t×he contributionsof 10.0keVbands. Sourceswith a maximumlikelihoodde-
the×foreground(unab§sorbed)andbackground(absorbed) tection value greater than 40 (corresponding to 10−13
model components to the observed spectra. erg cm−2 s−1) were removed. The region remo∼ved for
Ourobservationsandthedatareductionaredescribed each source was a circle whose radius contained 80% of
in 2. The spectral models used to fit to the data are the total flux of a point source at the source’s distance
§
described in 3, and the fit results are presented in 4. from the optical axis; this radius was typically 24–29
§ §
We discuss our results in 5. In particular, we compare arcseconds. The few remaining faint point sources are
§
ourresultswiththeresultsofRASSanalysesin 5.1,and likely to be background AGN with a power-law spec-
§
AN XMM-NEWTON OBSERVATION OF THE LOCAL BUBBLE 3
TABLE 1
Detailsof theXMM-NEWTON observations
Observation l b Nominalexposure Usableexposure I100a NHb
Observation ID (deg) (deg) (ks) (ks) (MJysr−1) (1020 cm−3)
Onfilament 0084960201 278.67 −45.32 12.8 11.9 7.10 9.6
Offfilament 0084960101 278.73 −47.09 27.8 4.4 1.22 1.9
a 100-µm intensity from the all-skyIRAS maps of Schlegel etal. (1998).b Calculated from I100 usingthe conversion
relationforthesouthernGalactichemisphereinSnowdenetal.(2000).
trumwhichisunlikelytoconfuseouranalysisofthermal Local Bubble emission. For the Galactic halo emission
spectra. Theactualspectrumofthediffuseemissionwas weusetwothermalplasmacomponents(2T),andforthe
extracted from a region with a radius of 14′ approxi- extragalactic background (due to unresolved AGN) we
mately centered on the optical axis after the removal of use a power-law. In this model, the Local Bubble com-
the point sources. ponent is unabsorbed, while the non-local components
We constructed the spectrum of the “quiescent parti- (halo and extragalactic) are all subject to absorption.
cle background” from the “unexposed corner” data and We carried out our spectral fitting with XSPEC1
filter-wheel-closed data (see Snowden et al. 2004). For v11.3.2p (Arnaud 1996). Our primary analysis was
eachofour two observations,andfor eachMOS camera, done using the Astrophysical Plasma Emission Code
the background spectrum is modeled using a database (APEC2) v1.3.1 (Smith et al. 2001) for the thermal
of filter-wheel-closed data, scaled by data from the un- plasma components. For comparison, we also ana-
exposed corners of the CCDs of that particular cam- lyzed the spectra using the MeKaL model (Mewe et al.
era. Thisscalingisenergydependent,andisbasedupon 1995), as discussed in 5.4. For the absorption we
§
the hardness and intensity of the unexposed corner data used the phabs model, which uses cross-sections from
(which varies with time). The background spectrum is Ba lucin´ska-Church& McCammon(1992),exceptforHe,
interpolated over the 1.2–1.9 keV interval before being in which case the cross-section from Yan et al. (1998) is
subtracted from the observed spectrum. This region of used. Our basic XSPEC model was thus apec+phabs
∗
the spectrum contains two bright instrumental lines due (apec + apec + powerlaw). For chemical abundances
to aluminum and silicon (at 1.48 and 1.74 keV, re- we used the interstellar abundance table in Wilms et al.
spectively). In most of the spe∼ctral fits de∼scribed below, (2000)3. These abundances were used both by the ther-
this region of the spectrum was simply excluded. How- mal plasma components and by the absorption model.
ever,leavingthe instrumentallinesintheobservedspec- Thenormalizationandthe photonindexofthepower-
trumandfittingthemwithGaussiansduringtheanalysis law used to model the extragalactic background were
does not significantly affect our results. frozen at 10.5(E/keV)−1.46 photons cm−2 s−1 sr−1
The strength of the residual soft proton flares is not keV−1(Chen et al.1997;fromtheirmodelAfittoASCA
knownapriori,buttheshapeisreasonablywellmodeled andROSAT data). Notethatthesevalueswereobtained
by a broken power-law with a break energy of 3.2 keV, byChen et al.(1997)after removingpointsourcesdown
where the spectrum is convolved with the redistribution to 5 10−14ergcm−2s−1,whichisroughlyequaltothe
matrix but not scaled by the response function. The lim∼it t×o which we have removed sources (see 2). This
contribution of the residual soft proton flares is fitted meansthattheir resultshouldbe applicableto§ouranal-
during the analysis. ysis. Furthermore, the exact values used for the extra-
galacticpower-lawhavelittleimpactonthefittingofthe
2.1. Solar Wind Charge Exchange
thermal emission.
The solar wind was very steady during both of these We simultaneously analyzed the on- and off-filament
observations. The solar proton flux, measured with the spectra. In the fits the temperatures and normaliza-
Advanced Composition Explorer (ACE), was 1.8 108 tionsofallthreeapeccomponentswerefree to vary,but
cm−2 s−1,slightlybelowthemean,andthepr∼oton×speed were constrained to be the same for all spectra. The
was 420–440 km s−1, slightly above the mean. The only difference in the model as applied to the different
O+7/O+6 and O+8/O+7 ratios had typical values. Both spectra was that the on- and off-filament spectra had
observationsweretakenatasolarangleof 80◦,avoiding different values of N for the phabs model. To deter-
∼ H
thehighestdensityportionsofthemagnetosheath. Thus, mine N for our two pointing directions, we obtained
H
any solar wind charge exchange (SWCX) contamination the 100-µm intensities for these directions from the all-
will be relatively low and, more to the point, similar for sky IRAS maps of Schlegel et al. (1998) and converted
the two observations. them to N using the conversion relation for the south-
H
ern Galactic hemisphere given in Snowden et al. (2000).
3. SPECTRALMODELING
The resulting on- and off-filament column densities are
3.1. Spectral Model Description 9.6 1020and1.9 1020cm−2,respectively(seeTable1).
× ×
The basic model we used to fit to our XMM-Newton Note that the on-filament column density is consistent
spectraisbaseduponthatusedbySnowden et al.(2000) with that derived from the color excess of the filament
and Kuntz & Snowden (2000) in their analyses of RASS
1 http://xspec.gsfc.nasa.gov/docs/xanadu/xspec/
data (which is itself a developmentofthe model usedby
2 http://cxc.harvard.edu/atomdb/sources apec.html
Snowden et al. 1998). Thus, we use a thermal plasma 3 ImplementedusingtheXSPECcommandabund wilm.
model in collisional ionization equilibrium (CIE) for the
4 HENLEY, SHELTON, AND KUNTZ
E(B V) = 0.17 0.05 (Penprase et al. 1998), which MeKaL. The advantage of this model is that it should
yields−N = (8.4 ±2.5) 1020 cm−2 when scaled using giveamoreaccuratepictureofthetemperaturedistribu-
H
± ×
the conversionrelationin Diplas & Savage(1994). Also, tion of the gas in the Galactic halo. On the other hand,
our results are not very sensitive to the values of N the2T halomodelenablesaneasiercomparisonwiththe
H
used. earlier ROSAT results.
It should be noted that our absorbing columns do A final variation of our “standard” model was to in-
not take into account any contribution from the warm vestigate the effect of varying the abundances of various
ionized gas in the Reynolds layer, above the disk of elements.
the Galaxy. The average column density of this gas We simultaneously analyzed the MOS1 and MOS2
is 7 1019/sin b H ii cm−2 (Reynolds 1991), which spectraobtainedfromtheon-andoff-filamentpointings.
gives×9.6–9.8 |10|19 H ii cm−2 for our observing di- We used the data between 0.45 and 5 keV, except for
×
rections. For 1/4-keV X-rays, the effective absorption the regionbetween 1.2 and 1.9 keV, where there are two
cross-section per hydrogen nucleus of the ionized gas is bright instrumental lines (see 2). An alternative to ex-
§
62%ofthatforneutralgas(Snowden et al.1994),which cluding these data is to fit these two lines by adding two
means that at low energies the Reynolds layer would ef- Gaussians to the model; however, as stated in 2, there
fectively contribute an extra 6 1019 cm−2 to our ab- is no significant difference between the results §obtained
×
sorbing columns. However, we find that adding this ex- in these two ways.
tra contribution to our absorbing columns affects only In order to better constrain the models at softer en-
the cooler halo component, andthis has no effect on our ergies, we also included two ROSAT spectra in the fits.
conclusions. Furthermore, observations of the Galactic These wereextractedfromROSAT All-Sky Surveydata
absorption towards extragalactic X-ray sources are gen- (Snowden et al.1997)usingthe HEASARC X-rayBack-
erally best fit without the Reynolds layer contribution groundTool5 v2.3. Thespectrawereextractedfrom0.5◦
(e.g. Arabadjis & Bregman1999). We therefore proceed radiuscirclescenteredonthetwoXMM-Newton pointing
using the IRAS-derivedcolumn densities quoted above. directions, and are normalized to one square arcminute.
As stated in 2, we also included a broken power-law While larger circles would have reduced the errors on
§
component in our fit to model the contribution of resid- the ROSAT data, they would result in the on-filament
ualsoft protonflaresnot removedinthe data reduction. ROSAT spectrum being contaminated by off-filament
Thissoftprotoncontaminationevidenceditselfincertain emission, and vice versa (see Fig. 1).
XMM-Newton spectra as excess emission above that ex- We accounted for the difference between the XMM-
pected from the extragalactic background power-law at Newton effective fieldofview andthe solidangleusedin
energies & 2 keV. The broken power-law parameters theROSAT extractionbymultiplyingthemodelapplied
were the same for the MOS1 and MOS2 spectra for a to the XMM-Newton data by the XMM-Newton field of
given pointing, but were allowed to differ between the view( 580arcmin2),andnormalizingtheROSAT spec-
two pointings. tra to∼1 arcmin2, as mentioned above.
We also experimented with variants of our “standard” After our fitting was complete, we converted the
model. One variation used a non-equilibrium ionization XSPEC model normalizations to emission measures
(NEI)modelfortheLocalBubble component,i.e.were- ( n2dl)assumingn /n =0.1,andneglectingthecon-
e He H
placed the first apec in our model with the XSPEC nei tribution of metals to the electron density n .
e
model. In this model, the emitting plasma is assumed R
3.2. A Note on the Abundance Table Used
to havebeenrapidly heatedto sometemperatureT,but
the ionizationbalance does not yet reflect this new tem- As stated in the previous section, we used the
perature (the ions are underionized). Such a model may Wilms et al. (2000) interstellar abundances, which dif-
be characterized by an ionization parameter, τ = n t, fer from those in widely used solar abundance tables
e
where n is the electron density and t is the time since (e.g. Anders & Grevesse 1989; Grevesse & Sauval 1998)
e
theheating. Collisionalionizationequilibriumisreached for many astrophysically abundant elements. For exam-
when τ & 1012 cm−3 s (Masai 1994). The nei model ple,Wilms et al.(2000)giveaninterstellaroxygenabun-
is configured to use the Astrophysical Plasma Emission dance n /n =4.90 10−4, compared with solar abun-
O H
Database (APED) to calculate the line spectrum4. dancesof8.51 10−4×(Anders & Grevesse1989)or6.76
Another variation replaced the 2T halo model with a 10−4 (Grevess×e & Sauval 1998). However, more recen×t
model that used a power-law differential emission mea- measurements of the solar photospheric oxygen abun-
sure (DEM) of the form dance (Allende Prieto et al. 2001; Asplund et al. 2004)
are in excellent agreement with the Wilms et al. (2000)
α
d(E.M.[T]) T if T <T <T , value. Indeed, recent measurements of other metals’
Tmax min max (1) photospheric abundances (e.g. C, N, Fe; Asplund et al.
d(logT) ∝((cid:16)0 (cid:17) otherwise.
2005) are lower than the Anders & Grevesse (1989) and
Grevesse & Sauval(1998)values,andareinbetteragree-
Here the exponent α and the high-temperature cut-off
ment with the Wilms et al. (2000) values. Therefore, in
T arefreeparameters. Thelow-temperaturecut-offis
max
frozenatT =105 K(theresultsarenotverysensitive this paper we take the Wilms et al. (2000) interstellar
min
abundances to be synonymous with solar abundances.
tothevalueofT ,asplasmaatthislowatemperature
min
does not significantly contribute to the observed X-ray 4. RESULTS
emission). ThismodelisbasedupontheXSPECcemekl
4.1. Spectral Fit Results
model, though we modified it to use APEC instead of
4 ImplementedusingtheXSPECcommandset neivers 2.0. 5http://heasarc.gsfc.nasa.gov/cgi-bin/Tools/xraybg/xraybg.pl
AN XMM-NEWTON OBSERVATION OF THE LOCAL BUBBLE 5
The results of fitting the above-described “standard”
(2T halo) model are presented in Table 2. Also shown 10
in this table are the results of using the XSPEC non-
equilibriumneimodelfortheLocalBubble,andofusing pc)
apower-lawdifferentialemissionmeasure(DEM)forthe -6m 1
haloemission(seeeq.[1]). Whenwetriedvaryingvarious T) (c
chemicalabundances,we foundthat the deviations from g
o
sinoclaornsaibstuenndtan(ec.egs.wthereeLeiotchaelrBstuabtibstleicaolxlyyginensiganbiufincadnatnocer EM) / d(l 0.1
wassomewhatdependentuponthehalomodelused). We d(
0.01
thereforedonotpresentanynon-solar-abundanceresults
in Table 2, and for the remainder of this paper we just
discuss our solar-abundance results. 5 5.2 5.4 5.6 5.8 6 6.2 6.4 6.6 6.8
NotefromTable2thattheLocalBubblemodelparam- log (T/K)
eters do not significantly change whether one uses a 2T Fig. 2.— Our best-fitting halo power-law differential emission
measure (DEM) model. The two arrows mark the temperatures
orpower-lawDEMmodelforthehalo. Forcompleteness
obtainedfromthe2T halomodel.
we show the halo emission measure given by the power-
law DEM model in Figure 2, but as this paper is mainly abundance of the vapec component frozen at zero. The
concerned with the Local Bubble, we defer discussion of twoGaussiansmodelthe oxygenline emission,while the
the halo results to a later paper. For the remainder of vapeccomponent models the continuum and the contri-
this paper the results discussed will be those obtained bution of other lines. The widths of the Gaussians were
with the 2T halo model. fixed at zero, and the energies were fixed at 0.5681 and
The spectra and the best-fit “standard” model are 0.6536 keV for O vii and O viii, respectively. For O vii
shown in Figure 3. Note the offset between the total this is the mean energy of the resonance, intercombina-
model emission and the extragalactic component above tion, and resonance lines, weighted by the line emissivi-
2 keV in the off-filament XMM-Newton data. This ties for a 106.06 K plasma. For O viii it is the weighted
∼
is entirely due to soft-proton contamination (see 3.1). meanenergyofthetwoLyαlines. Therelevantlineener-
§
Note also that in the XMM-Newton band the Local giesandemissivitieswereobtainedfromthe APECcode
Bubble componentmakes asignificantcontributiononly usingthe XSPECidentifycommand. Thawingthe en-
to the O vii emission at 0.57 keV. Most ( 90%) of ergies of the Gaussians did not significantly improve the
∼ ∼
the O viii emission at 0.65 keV is due to the hotter fit, nordiditsignificantlyalter the measuredintensities.
∼
halocomponent, withthe extragalacticbackgroundcon- The temperature and normalizationofthe LocalBubble
tributing 8% of the emission at this energy. However, vapeccomponentwereallowedtovary,whiletheparam-
∼
the Local Bubble dominates the spectrum at the low- etersofallthe othercomponentswereheldfixedattheir
est energies in the on-filament ROSAT spectrum, as the best-fit values from the previous section.
halo and extragalacticcomponents are very strongly ab- We fit this new model to our on- and off-filament
sorbed. The absorbing cross-section per hydrogen atom XMM-Newton and ROSAT spectra, as before. We mea-
is 8.96 10−21 cm2 at 0.2 keV, calculated using the sureanOviiintensityof3.4+0.6 photonscm−2 s−1 sr−1,
× −0.4
Ba lucin´ska-Church& McCammon (1992) cross-sections andobtaina3σ upperlimitontheOviiiintensityof1.0
(except for He; Yan et al. 1998) with the Wilms et al. photons cm−2 s−1 sr−1 (see Table 3). We use these re-
(2000)interstellarabundances. Thisgivesanon-filament sults in 5.6.
optical depth of 8.6 at this energy. We ca§n check these values using line emissivity data
Whiletherearesomefeaturesthatarepoorlyfitinone from the ATOMDB database6. The temperature and
of the spectra (e.g. the model underestimates the O vii emission measure of our “standard” model Local Bub-
emission in the off-filament MOS1 spectrum), when one ble component yield O vii and O viii intensities of 2.9
considers all the spectra together there are no features and 0.017 photons cm−2 s−1 sr−1, respectively, includ-
that are systematically poorly fit. Overall, the model ing the contribution of dielectronic recombination satel-
gives a good fit to the data, with χ2ν = 0.99 for 439 lite lines. These values are consistent with the above-
degrees of freedom. measuredvalues. AlthoughtheOviiintensitymeasured
We tested whether or not a Local Bubble component fromthe Gaussianfits is higher thanthat obtainedfrom
is necessary by trying a model without a Local Bubble ATOMDB, this is unlikely to be due to contamination
component. As can be seen from Table 2 such a model by other oxygen lines, as there are no other bright oxy-
givesa poorfitto the datawhether oneusesa 2T model gen lines in that vicinity. Instead, it is possible that the
(χ2ν = 1.82 for 441 degrees of freedom) or a power-law LocalBubbleapecmodelisslightlyunderestimatingthe
DEM model (χ2ν = 1.81 for 442 degrees of freedom) for contribution of the Local Bubble to the observed O vii
the halo. As can also be seen from the table, a 1T halo emission. This is because this component is not being
model gives a poor fit to the data (χ2ν = 1.22 for 441 constrained solely by the oxygen emission, but also by
degrees of freedom). emissionatotherenergies(e.g.theROSAT R12intensity
andtheR1/R2ratio). TheOviiiemissionismorelikely
4.2. Local Bubble O VII and O VIII Intensities
to be contaminated, for example by O vii n = 3 1
To measure the intensities of the O vii and O viii emission at 0.6656 keV. However, this just means→that
emission from the Local Bubble, we replaced the Local theaboveupperlimitmaybeoverlyconservative;itdoes
Bubble apec component in our “standard” model with
a gaussian+gaussian+vapecmodel, withthe oxygen 6 http://cxc.harvard.edu/atomdb/download.html
6 HENLEY, SHELTON, AND KUNTZ
TABLE2
Spectralfitresults
2T halomodel
LocalBubble Halo(cool) Halo(hot)
logT E.M.a τb logT E.M.a logT E.M.a
Model (K) (cm−6 pc) (1010 cm−3 s) (K) (cm−6 pc) (K) (cm−6 pc) χ2/dof
“Standard”(CIELocalBubble) 6.06+−00..0024 0.018 ··· 5.93+−00..0043 0.17 6.43±0.02 0.011 435.86/439
NEILocalBubble 6.21+−00..0180 0.011 3.4+−41..74 5.90+−00..0023 0.22 6.44±0.02 0.010 434.56/438
NoLocalBubble ··· ··· ··· 5.92±0.01 0.37 6.44±0.02 0.012 801.12/441
Onehalocomponent 6.05+−00..0012 0.023 ··· ··· ··· 6.37±0.01 0.018 539.57/441
HotLocalBubblec 6.21(frozen) 0.017 ··· 5.29+−00..1038 160 6.49±0.02 0.0073 444.04/440
LocalBubble DEMhalomodeld
logT E.M.a τb logTmax α
Model (K) (cm−6 pc) (1010 cm−3 s) (K) χ2/dof
CIELocalBubble 6.02+−00..0012 0.018 ··· 6.70±0.07 −2.01+−00..1143 438.62/440
NoLocalBubble ··· ··· ··· >6.80 −2.73+−00..0087 801.64/442
a EmissionmeasureE.M.=Rn2edl.b Ionizationparameterτ =net.c See§5.3.d See§3.1fordetailsofmodel.
TABLE3 TABLE4
LocalBubbleOviiandOviiiintensities DerivedparametersoftheLocalBubble
Energy Intensity Parameter Value
Ion (keV) (photonscm−2 s−1 sr−1)
Electrondensityne (cm−3) 0.013(L/100pc)−1/2
Ovii 0.5681 3.4+−00..64 Numberdensityn(cm−3) 0.026(L/100pc)−1/2
Oviii 0.6536 <1.0a Pressurep/k(cm−3 K) 2.9×104(L/100pc)−1/2
ThermalenergyEth (erg) 7.4×1050(L/100pc)5/2
a 3σupperlimit. Coolingtimea tcool (yr) 1.7×107(L/100pc)1/2
not adversely affect our later analysis and conclusions. Soundcrossingtimetcross (yr) 1.2×106(L/100pc)
McCammon et al. (2002) measured the intensities of
Note. —Calculatedfromthe“standard”modelparame-
O vii and O viii in the soft X-ray backgroundusing the tersfortheLocalBubbleinTable2,assumingtheradiusof
X-rayQuantumCalorimeter(XQC),whichwasflownon taheCLaloccuallatBeudbbulseinisgLt.he Raymond-Smith cooling function
a sounding rocket. They obtainedintensities of4.8 0.8 (Raymond&Smith1977;Raymond1991)withWilmsetal.
and1.6 0.4photonscm−2 s−1 sr−1 forOviiandO±viii, (2000)abundances: Λ(106.06 K)=7.5×10−23 ergcm3s−1.
±
respectively. ThesearetotalintensitiesforthesoftX-ray
background averaged over a large area of sky ( 1 sr), whether or not this can be attributed to our choice of
∼
whereas the intensities in Table 3 are just for the Local plasma emission code in 5.4. Finally we discuss non-
§
Bubble. Our results are therefore consistent with the equilibrium models of the Local Bubble: in 5.5 we dis-
§
McCammon et al. (2002) results. cusstheresultsofusinganunderionized(ionizing)model
ofthe LocalBubble (i.e. the XSPECnei model), and in
4.3. Derived Parameters of the Local Bubble 5.6 we discuss our results in terms of an overionized
§
(recombining) model of the Local Bubble.
If we assume some spatialextent L for the LocalBub-
ble in our pointing direction, and assume that the Local
5.1. Comparison with ROSAT Results
Bubble plasma is uniform along the line of sight, we can
convert the emission measure found above to a density. In Table 5 we compare our measured temperatures
With the measured plasma temperature, this will give with those measured in various studies of RASS data.
us the thermal pressure of the plasma. If we make the Ascanbe seen,thereisexcellentagreementbetweenour
further simplifying assumption that the Local Bubble is values and the ROSAT-determined values. This is not
asphereofradiusL,wecanestimatethethermalenergy surprising, as we use RASS data to constrain our spec-
content and the cooling time of the Local Bubble. tral models at low energies. However, our Local Bubble
The Local Bubble parameters derived from the “stan- emissionmeasure( n2dl=0.018cm−6pc)is3–10times
e
dard”modelparametersinTable2areshowninTable4. largerthanthat derivedfromthe ROSAT data(0.0018–
0.0058 cm−6 pc; SnRowden et al. 1998).
5. DISCUSSION ThisdiscrepancyisduetothefactthatSnowden et al.
In this section, we first compare our results with the (1998) use the Raymond & Smith (1977) plasma emis-
results of analyses of ROSAT All-Sky Survey (RASS) sion code, whereas we use APEC. Also, Snowden et al.
data in 5.1. In 5.2 we discuss our results in terms of (1998) assume a higher metallicity in their study: Z =
§ §
the Galactic halo O vi emission (R. L. Shelton et al., 0.017 (Allen 1973) against Z = 0.012 (Wilms et al.
in preparation). We compare our results with those of 2000). For a T =106.06 K plasma, a Raymond & Smith
other XMM-Newton and Chandra shadowing observa- model with Z =0.017predicts 3 times as much flux in
∼
tions of the LocalBubble in 5.3. We find a discrepancy the0.1–0.5keVbandasanAPECmodelwithZ =0.012.
§
between our Local Bubble temperature and that mea- Hence,foragivenamountofLocalBubbleemission,our
sured by other XMM-Newton observations, and discuss modelwillgiveanemissionmeasure 3timeslargerthan
∼
AN XMM-NEWTON OBSERVATION OF THE LOCAL BUBBLE 7
-1V 1 XMM-MOS1 ON LoHcaall oB (ucbobolel) -1V 1 XMM-MOS1 OFF LoHcaall oB (ucbobolel)
ke Halo (hot) ke Halo (hot)
-1s s ExtragalTaocttaicl -1s s ExtragalTaocttaicl
nt nt
u u
co 0.1 co 0.1
d d
e e
z z
ali ali
m m
or or
N N
0.01 0.01
s 0.4 s 0.4
ual 0.2 ual 0.2
d 0 d 0
si -0.2 si -0.2
Re -0.4 Re -0.4
0.5 1 2 5 0.5 1 2 5
Energy (keV) Energy (keV)
-1V 1 XMM-MOS2 ON LoHcaall oB (ucbobolel) -1V 1 XMM-MOS2 OFF LoHcaall oB (ucbobolel)
ke Halo (hot) ke Halo (hot)
-1s s ExtragalTaocttaicl -1s s ExtragalTaocttaicl
nt nt
u u
co 0.1 co 0.1
d d
e e
z z
ali ali
m m
or or
N N
0.01 0.01
s 0.4 s 0.4
ual 0.2 ual 0.2
d 0 d 0
si -0.2 si -0.2
Re -0.4 Re -0.4
0.5 1 2 5 0.5 1 2 5
Energy (keV) Energy (keV)
-1keV 10-2 ROSAT ON LoHcHaaall oBlo (u c(bhobooltel)) -1keV 10-2 ROSAT OFF LoHcHaaall oBlo (u c(bhobooltel))
-1s s 10-3 ExtragalTaocttaicl -1s s 10-3 ExtragalTaocttaicl
nt nt
u u
o o
d c 10-4 d c 10-4
e e
z z
mali 10-5 mali 10-5
or or
N N
10-6 10-6
s 4×10-4 s 4×10-4
al al
u u
d 0 d 0
si si
Re -4×10-4 Re -4×10-4
0.1 0.2 0.5 1 2 0.1 0.2 0.5 1 2
Energy (keV) Energy (keV)
Fig. 3.—Ourobservedon-filament(left-handcolumn)andoff-filament(right-handcolumn)spectra,withourbest-fit“standard”model
(note thedifferent energyranges onthe XMM-Newton andROSAT plots). Theindividualcomponents thatcompriseourmodel arealso
illustrated. Noteintheon-filamentXMM-Newton resultsthattheLocalBubble(blue)andcoolhalo(green)componentsoverlap. Thegap
intheXMM-Newton dataiswheretwobrightinstrumentalfluorescencelineshavebeenexcluded. ThelargedifferencebetweentheXMM-
Newton and ROSAT count-rates is because the ROSAT data have been normalizedtoone squarearcminute, whereas the XMM-Newton
dataareintegrated overtheilluminatedXMM-Newton fieldofview(≈580arcmin2).
Snowden et al.’s (1998) model, which is what we find. sr−1 (R. L. Shelton et al., in preparation). In compar-
ison, the best-fit temperature and emission measure of
5.2. The Galactic Halo O VI Emission the cooler halo component from our “standard” model
predicts an intrinsic O vi doublet intensity of 3000
As already stated, the main focus of this paper is the photons cm−2 s−1 sr−1 (using data from the ATO∼MDB
Local Bubble emission, with discussion of the Galactic
database7). Here we use the 2T halo model, as the halo
halo emission being deferred to a later paper. However,
differentialemissionmeasureispoorlyconstrainedatlow
wenoteherethatourfitresultsmayalsobeusedtomake
temperatures. Furthermore, we just consider the cooler
predictionsofthehaloOviintensity,whichmaythenbe
halo component, as the contribution of the hotter com-
compared with that measured from an off-filament Far
ponent to the O vi emission is negligible.
Ultraviolet Spectroscopic Explorer (FUSE) observation.
We used a Monte Carlo method to estimate the un-
This provides a useful check on our fit results.
certainty on this O vi intensity prediction. We gener-
AssumingallthehaloOviemissionoriginatesfrombe-
ated1000 randompairs of(T [halo], E.M.[halo]), where
yondtheabsorbingmaterialinthatdirection(whichhas
a transmissivity to O vi photons of 58%), the intrinsic 7 Seefootnote6.
intensity of the doublet is 8070+980 photons cm−2 s−1
−1140
8 HENLEY, SHELTON, AND KUNTZ
TABLE5
ComparisonwithROSAT results
logTLB logTHalo,1 logTHalo,2
Work (K) (K) (K)
Snowdenetal.(1998)a 6.07±0.05 6.02±0.08 ···
Snowdenetal.(2000) 6.08b 6.00b 6.4c
Kuntz&Snowden(2000) 6.11+−00..1057 6.06+−00..1290 6.46+−00..1028
Thisworkd 6.06+−00..0024 5.93+−00..0043 6.43±0.02
a Nosecondhalocomponentused.bNoerrorsquoted.cTemperature
fixedatthisvalue.d “Standard”modelresultsfromTable2.
80 Barnard 68 (see Appendix), versus log(TLB/K) =
6.06+0.02 from our data. Note, however, that for
70 −0.04
MBM 12 there is an additional log(T/K) 6 com-
60 ≈
s ponent (Freyberg 2004). It should also be noted that
e
alu 50 Freyberg & Breitschwerdt(2003)andFreyberg(2004)do
v
of 40 not state whether or not they use ROSAT data to con-
er strain their fits at lower energies.
mb 30 IncontrasttotheseXMM-Newton results,Smith et al.
u
N 20 FUSE (2005)foundthattheycouldnotexplaintheOvii:Oviii
ratio in their Chandra observation of MBM 12 with
10
an equilibrium Local Bubble model with log(T /K) <
LB
0 6.3. However, they do note the possibility that their
0 2000 4000 6000 8000 10000
Halo O VI intensity (ph cm-2 s-1 sr-1) O viii emission is contaminated by emission from an-
other source, such as solar charge exchange emission.
Fig. 4.— Histogram of 1000 randomly generated values of the
These other shadowing observations were carried out
intrinsicGalactichaloOviintensity,predictedfromourmeasured
values of the halo temperature and emissionmeasure, taking into using much thicker absorbers than our observations:
account the errors on these parameters. The arrow and the hor- 4 1021 cm−2 for MBM 12 (Smith et al. 2005) up
izontal bar denote the intrinsic intensity inferred from FUSE, as- to× 1023 cm−2 for Barnard 68 (Freyberg et al. 2004),
sumingalltheOviemissionoriginatesfrombeyondtheabsorbing aga∼inst 9.6 1020 cm−2 for our filament. The optical
materialinthatdirection(seetextfordetails).
×
depth at O viii energies is at least 2 for these other
E.M. [halo] = n2dl is the emission measure of the
e observations, implying a transmissivity for background
cooler halo component. These pairs of numbers were
O viii radiation of less than 14%. In an observation of
drawnat randomR from normaldistributions whose stan-
one of these thicker absorbers, one may confidently at-
dard deviations are given by the errors on the measured
tribute a large fraction of any observed O viii emission
parametersinTable2(E.M. [halo]=0.17+0.05cm−6pc),
−0.04 to the foreground, and thus infer a higher Local Bub-
and were used to calculate 1000 values of the O vi dou-
ble temperature. However, the higher transmissivity of
blet intensity. We fit a Gaussian to the distribution of
our filament (61% at O viii energies) makes it harder
these values, and from the mean and standarddeviation
to determine how much of the observedemission is from
obtaina predictedOviintensity of3100 1000photons
cm−2s−1sr−1(seeFig.4). Theobservedi±ntensityis3.3σ the Local Bubble, and how much is background emis-
sion that has leaked through the filament. Our best-fit
larger than the predicted intensity (though the distribu-
“standard” model attributes only 2% of the observed
tion of predicted intensities is positively skewed, which ∼
on-filament O viii emission to the Local Bubble, com-
will tend to reduce the significance of this difference).
pared with 30% of the observed O vii emission. The
TheseresultsindicatethatthereismoreOviinthehalo ∼
fact that our best-fit model attributes so little O viii
than is expected from the hot gas alone. This is prob-
emission to the Local Bubble leads to our lower value of
ably because O vi can also arise in the warm interfaces
T . However, the question remains, could more of the
between cool clouds and the hot gas. LB
O viii emission in our observation be due to the Local
Bubble? To put this more precisely, are our data also
5.3. Comparison with Other Shadowing Observations of
consistent with log(T /K)=6.21?
the Local Bubble LB
We tested this by re-fitting our “standard” model to
XMM-Newton has been used to make shadowing the data, but with T frozen at 106.21 K. The results
LB
observations of the Local Bubble in the directions of this are shown in Table 2. This model does give
of the MBM 12 and Ophiuchus molecular clouds a good fit to the data (χ2 = 1.01 for 440 degrees of
ν
(Freyberg & Breitschwerdt 2003; Freyberg 2004), and freedom). However, note that the temperature of the
in the direction of the Bok globule Barnard 68 cooler halo component has significantly decreased, and
(Freyberg et al. 2004). Chandra has also been used to its emission measure has increased by three orders of
observe MBM 12 (Smith et al. 2005). magnitude. Both these effects lead to a huge increase
The Local Bubble temperatures inferred from these in the predicted intrinsic halo O vi intensity (cf. 5.2)
XMM-Newton observations are consistently higher than to 107 photons cm−2 s−1 sr−1. As before, we §used
the temperature we measure: log(TLB/K) = 6.21+−00..0067 a M∼onte Carlo method to estimate the uncertainty on
for MBM 12 and Ophiuchus (Freyberg & Breitschwerdt this prediction. In this case, there is a much largererror
2003; Freyberg 2004) and log(T /K) 6.24 for
LB
≈
AN XMM-NEWTON OBSERVATION OF THE LOCAL BUBBLE 9
on the emission measure of the cooler halo component 10-3
(E.M. [halo] = 160+1000 cm−6 pc), resulting in a much O VII O VIII
−20
largerdynamicrangeinthe predictedintensities. Byfit- -1V
ting a Gaussian to the distribution of the logarithms of ke
tlohge(Ipredi/cpthedcmva−lu2ess−,1wser−fi1n)d=t7h.e5pr1ed.1i.ctIendcionmtepnasriitsyonis, -2min 10-4
OVI c
the intrinsic intensity inferred from± the FUSE data is 1 ar
log(IOVI/ph cm−2 s−1 sr−1) = 3.79+−00..0056 (R. L. Shelton -s s
n
et al., in preparation). While we could interpret the dis- o
ot 10-5
crepancy between the predicted and observedintensities h
P
MEKAL
as evidence that the Galactic halo is out of equilibrium,
given the size of the discrepancy (i.e. possibly up to a 3×10-6 APEC
0.4 0.5 0.6 0.7 0.8
few orders of magnitude), we instead interpret it as ev-
Energy (keV)
idence that log(T /K) = 6.21 is inconsistent with our
LB
XMM-Newton spectraandtheFUSE resultswhentaken Fig. 5.— Our best fitting APEC and MeKaL Local Bubble
models,foldedthroughtheXMM-Newton MOS1instrumentalre-
together.
sponse. Note that there is more Local Bubble O viii emission at
It should be noted that two of the clouds discussed ∼0.65keVintheMeKaLmodelthanintheAPECmodel.
above (Ophiuchus and Barnard 68) lie beyond the Lo-
cal Bubble in the direction of Loop I, near the Galactic We testedthis by repeatingour fits usingthe MeKaL
plane. It is therefore possible that the foreground emis- code instead of APEC for the thermal emission compo-
sion (which in the above discussion we have attributed nents. The results are compared with our APEC results
to the Local Bubble) is being contaminated by emission inTable6. NoteinparticularthatMeKaLgivesahigher
from Loop I. However, Freyberg (2004) states that the Local Bubble temperature than APEC: log(T /K) =
LB
weaknessoftheFe-Llineemission(attributedtoLoopI) 6.17+0.06 (MeKaL) versus 6.06+0.02 (APEC). It should
−0.07 −0.04
in the on-cloud Ophiuchus spectrum indicates that the be emphasized that the difference between the APEC
contaminationissmall. Incontrasttothis,MBM12does and MeKaL results for our data is not because the
notlietowardsanyobvioussourceofcontamination,and two codes give different temperatures for the same Lo-
may evenlie within the LocalBubble, implying that the cal Bubble spectrum. Instead it is because MeKaL at-
temperature inferred from these observations is indeed tributes moreofthe O viiiemissiontothe LocalBubble
that of the Local Bubble. (seeFig.5),andcorrespondinglylesstothehalo. Hence,
Our pointing direction is & 80◦ away from the direc- the inferred Local Bubble spectrum is different between
tionsoftheotherXMM-Newton observations. Itisthere- the two codes, and thus so too is T . This discrepancy
LB
fore not implausible that our value of TLB is different is most likely to be due to uncertain modeling of the
from those measured from these observations, as we are lines from L-shell ions of Ne, Mg, and Si, which domi-
observingadifferentpartoftheLocalBubble. Thesere- nate the emission at the lowest ROSAT energies. The
sultsthereforesuggestthattheLocalBubbleisthermally ATOMDB v1.3.1 release notes8 contain a caveat that
anisotropic. there are very few data on lines from these ions (other
than Li-like ions). Differences between the codes in this
5.4. Choice of Plasma Emission Code
energy regime would affect the fits to the ROSAT data,
Inthissectionweconsiderwhateffect,ifany,thechoice which would then affect the fitting to the higher-energy
ofplasmaemissioncodeusedhasonthe measuredLocal XMM-Newton data.
Bubble temperature. To test whether or not this does cause the discrep-
The higher foreground temperatures measured from ancy between the APEC and MeKaL results, we re-fit
the XMM-Newton spectra of MBM 12, the Ophiuchus our“standard”modeljusttotheXMM-Newton spectra,
molecularcloud,andBarnard68shouldbequite robust, without the ROSAT data. These results are also shown
regardless of the plasma emission code used. This is be- in Table 6. Note that these temperatures are unphysi-
cause O viii emission is observed towards these clouds. cal,astheyareinconsistentwiththelow-energyROSAT
Due to the large optical depths of the clouds, a large data. However,theagreementbetweenthecodes’results
fraction of this emission may be attributed to the fore- is much better than before, suggesting that at leastpart
ground, implying a higher temperature than we found of the discrepancy is due to uncertain modeling of lines
towards our filament. Freyberg & Breitschwerdt (2003) in the lowest-energy ROSAT bins.
do not give details of the models used in their analysis Despite this discrepancy, we reiterate that the mea-
oftheXMM-Newton observationsofMBM12andOphi- surement of log(T /K) 6.21 from the other XMM-
LB
uchus, but they do note that log(TLB/K) = 6.21+−00..0067 Newton observations sh≥ould not be strongly code-
“or higher, depending on the actual model.” Also, we dependent, and note that we obtain a lower tempera-
infer log(TLB/K) = 6.24 from Freyberg et al.’s (2004) ture than this whether we use APEC or MeKaL. This
Barnard68 data whether we use APEC or MeKaL (see thereforeimplies thatthe conclusionofthe previoussec-
Appendix). However, since our filament is less optically tion, namely that the Local Bubble appears to be ther-
thick than these other clouds, there is a greaterambigu- mally anisotropic, is not an artefact of our choice of
ity between what is Local Bubble emission and what is plasma emission code. However, we should reiterate the
haloemission,andsowetestthepossibilitythatdifferent possibility that the foreground emission in some of the
codes would attribute different amounts of the emission other XMM-Newton observations may be contaminated
in our spectra to the LocalBubble and the halo, leading
to a different value of TLB. 8 http://cxc.harvard.edu/atomdb/issues improvements.html
10 HENLEY, SHELTON, AND KUNTZ
by non-Local Bubble emission. ized and are in the process of ionizing.
5.5. Ionizing Model of the Local Bubble 5.6. Recombining Model of the Local Bubble
In the preceding sections we have just discussed col- In this section we consider an alternative non-
lisional ionization equilibrium (CIE) plasma models. In equilibrium Local Bubble model to the previous section,
this section we discuss the results of using the XSPEC namely one in which the plasma is overionized and the
nei model to model the Local Bubble emission (see Ta- ions are in the process of recombining.
ble2). Thisisamodelinwhichtheionsareunderionized, Such a model for the Local Bubble was origi-
i.e. the plasma has been rapidly heated, but the ioniza- nally proposed by Breitschwerdt & Schmutzler (1994;
tion balance does not yet reflect the new temperature: see also Breitschwerdt 1996; Breitschwerdt et al. 1996;
the ions are still in the process of ionizing. As stated in Breitschwerdt 2001). In this model, the Local Bubble
3.1,suchamodelmaybecharacterizedbyanionization is assumed to be the relic of an old superbubble formed
§
parameter,τ =net, where ne is the electrondensity and by 10 successive supernovae in a dense ( 104 cm−3)
t is the time since the heating. Equilibrium is reached mol∼ecular cloud. A few million years ago,∼the bubble
when τ &1012 cm−3 s (Masai 1994). burst out of the cloud and underwent rapid adiabatic
In certain situations one may use the F-test to de- expansion into the surrounding medium, assumed to be
termine whether or not adding an extra model pa- 10–100 times less dense than the cloud. During this
rameter leads to a significant improvement in χ2 ∼process the bubble adiabatically cooled to . 105 K (i.e.
(Bevington & Robinson 2003). Unfortunately we can- muchlessthanthetemperatureinferredfromtheX-rays
not use the F-test here to test whether or not the non- assuming collisional ionization equilibrium). However,
equilibriummodelisanimprovementonthe equilibrium owingtotherapidityofthecooling,theionsremainedin
model. This is because the additional parameter (τ) is highionizationstates. TheobservedX-raysarethensup-
on the boundary of the set of possible parameter values posed to be due to the delayed recombinations of these
in the simpler (null) model (i.e. τ = in the equilib- overionizedions, rather than being from a hot ( 106 K)
rium model). In such a case, one cann∞ot use the F-test plasma in equilibrium. ∼
(Protassov et al. 2002). A recombining plasma model has a number of appeal-
However,wecannotethatthedifferenceinχ2between ing features. Firstly, there is an order of magnitude
the two models is very small, suggesting that thawing τ difference between the pressure of the Local Cloud in
does notlead to a significantimprovementin χ2 (even if which the solar system resides (P /k 2000 cm−3 K;
LC
wecannotusetheF-testtodemonstratethisforcertain). Lallement1998)andthe pressureofthe∼surroundingLo-
Ontheotherhand,itshouldbenotedthatthevalueofτ cal Bubble inferred from the X-ray data assuming CIE
twheanmewahsautreis([e3x.4p+−ec41t..74e]d×in10t1h0ecemqu−i3libs)riiusmsigcnaisfiec(aτnt&ly1l0e1s2s (cPmL−B3/kK∼[T1a5b,l0e004]c)m. I−t3isKd[iSffincouwltdetnoeseteahl.o1w99s8u]cthoa29la,0r0g0e
cm−3 s; Masai1994), which means that our data cannot pressure difference can be maintained. The lower Local
rule out an ionizing non-equilibrium model. Bubble temperature in the Breitschwerdt & Schmutzler
In order to distinguish between equilibrium and non- (1994) model greatly reduces the Local Bubble pres-
equilibrium models we would require more spectral in- sure, and thus eliminates this problem. Secondly, the
formation. With our data, the only Local Bubble “line” electron density inferred from the dispersion measure of
(actually a blend of lines) we can clearly see is the O vii the pulsar PSR 0950+08 (d 130 pc) is 0.023 cm−3
≈
feature at 0.57 keV. There are no other obvious Lo- (Reynolds 1990). If this density is representative of the
cal Bubble∼lines in the XMM-Newton bandpass (the ob- Local Bubble, then a 106 K plasma in equilibrium
∼
served O viii feature in the on-filament spectrum is at- would produce too much X-radiation (cf. the electron
tributed to leak through from the Galactic halo), and density inferred from the X-ray data assuming CIE is
we only have two ROSAT spectral bins at energies be- 0.007 cm−3 [Snowden et al. 1998] to 0.013 cm−3 [Ta-
∼
low the XMM-Newton bandpass. In contrast to this, ble 4]). However, a higher n is acceptable within the
e
one needs to compare the flux ratios of lines from sev- Breitschwerdt& Schmutzler (1994) model. Thirdly, a
eral different ions in order to determine whether or not 106 K plasma in equilibrium should produce most of
∼
non-equilibrium effects are important. its emission as lines in the extreme ultraviolet (EUV).
In principle, we could use data from other wavebands However, such emission is not observed. Extreme Ultra-
to help distinguish between the two classes of model. violetExplorer (EUVE)spectraofthediffuseEUVback-
FUSE has observed the Local Bubble in the same di- ground( 17–80eV)placeanupperlimitontheemission
rections as our XMM-Newton observations. Using the measure∼ofalocal106Kplasmawhichisanorderofmag-
on-filament observation, Shelton (2003) has placed a 2σ nitude less than the Local Bubble emission measures in
upperlimitonthe LocalBubble Oviλλ1032,1038dou- Table2(Jelinksy et al.1995;Vallerga & Slavin1998). A
blet intensity of 800 photons cm−2 s−1 sr−1 (including recombining plasma model can explain this observation,
statistical and systematic uncertainties). In compari- as the much lower kinetic temperature suppresses colli-
son, the predicted O vi intensities of our best-fit models sional excitation of the EUV lines (Breitschwerdt1996).
(determined using XSPEC) are 180 photons cm−2 s−1 Hereweexaminewhetherornotarecombiningplasma
sr−1 (equilibrium) and 63 photons cm−2 s−1 sr−1 (non- model can simultaneously explain our observed Local
equilibrium), both of which are consistent with the ob- Bubble O vii intensity and the observedupper limits on
served upper limits. the O vi and O viii intensities. We do this by first cal-
We thereforeconcludethatourdataareunable todis- culating the fractions of oxygen in the ionization states
tinguish between an equilibrium plasma model and a O vi to O ix, and then using these populations to pre-
non-equilibrium model in which the ions are underion- dict the line intensities. These are then compared with