The giant X-ray outbursts in NGC 5905 and IC 3599: Follow-up observations and outburst scenarios

14 Pages·0.4 MB·English

by  Stefanie Komossa

A&A manuscript no. ASTRONOMY
(will be inserted by hand later) AND
Your thesaurus codes are: ASTROPHYSICS
03 (11.01.2; 11.09.1; 11.17.2; 11.19.3; 13.25.2)
The giant X-ray outbursts in NGC 5905 and IC 3599:
Follow-up observations and outburst scenarios
Stefanie Komossa1, Norbert Bade3,2
1 Max-Planck-Institutfu¨r extraterrestrische Physik,85740 Garching, Germany
2 HamburgerSternwarte, Gojenbergsweg 112, 21029 Hamburg, Germany
9
3 Schwalbenweg 55, 89081 Ulm, Germany
9
9
1 Received 28. July 1998; accepted 4. December 1998

n
a
J Abstract. Huge amplitude X-ray outbursts in a few 1. Introduction
galaxieswerereportedinthe lastfew years.As oneofthe
2
1 exciting possibilities to explain these observations, tidal X-rayvariabilitybyfactorsof∼2–3isacommonproperty
disruption of a star by a central supermassive black hole of active galaxies (e.g., Mushotzky et al. 1993). However,

1 has been proposed. In the present paper, we perform a X-rayoutburstsbyfactorsoforder100areextremelyrare;
v detailed discussion of this and other possible scenarios only very few objects have been reported to show such
1
for the X-ray outburst in NGC5905, and a comparison behaviour. Among these are IC3599 (Brandt et al. 1995,
4
1 of NGC5905 and IC3599 in outburst as well as in qui- BPF95hereafter;Grupeetal.1995a,G&95hereafter)and
1 escence. To this end we present (i) a thorough analysis NGC5905 (Bade et al. 1996,paperI hereafter).
0 ofallROSAT PSPCX-rayobservationsofNGC5905and Both galaxies were detected in the X-ray high-state
9
newHRIdata,(ii)opticalphotometryofNGC5905quasi- during the ROSAT all-sky survey and then declined in
9
simultaneous to the X-ray outburst and on longer terms, intensity within months–years. The peak luminosities ex-
/
h (iii) the first post-outburst optical spectra of NGC5905 ceededL =1042erg/sandtheoutburstspectrawerevery
x
p and high-resolutionpost-outburstspectra of IC3599,and soft (photon indices Γ ≈ –4). Further, both objects ap-
- x
o (iv) photoionization models for the high-excitation emis- pear rather inactive, as judged from optical spectra (this
r sion lines that were discovered in the optical outburst point will be examined in detail below). Here we focus on
t
s spectrum of IC3599. The investigated outburst models NGC5905, and perform a comparison with IC3599. For
a
include, besides the tidal disruption event, a supernova both sources, tidal disruption of a star by a supermas-
:
v in dense medium, an accretion-disk instability, an event sive black hole (SMBH) has been mentioned as possible
i
X of extreme gravitational lensing, the X-ray afterglow of outburstmechanism(BPF95,G&95,paperI).The obser-
a GRB, and the possibility of a warm-absorbed hidden vation of a UV flare in the center of the elliptical galaxy
r
a Seyfertnucleus in the center ofthe galaxy.The successful NGC4552 was interpreted in the same sense (Renzini et
models, all involving the presence of a central supermas- al. 1995).
siveblackhole,areselectedandimplicationsarediscussed. ThepossibilityoftidaldisruptionofastarbyaSMBH
Theopticalspectraofbothgalaxiesinquiescencearecare- wasoriginallyproposedasameansoffuellingactivegalax-
fully examined for signs of permanent low-level Seyfert- ies (Hills 1975), but was later dismissed. Peterson & Fer-
activity. Whereas IC3599 shows several signs of activity, land (1986) suggested this mechanism as possible expla-
noneisrevealedforNGC5905.Atpresent,andamongthe nation for the transient brightening and large width of
X-ray bursts, this makes NGC5905 the only safe candi- the HeII line observed in a Seyfert galaxy. Tidal disrup-
dateforatidaldisruptioneventinanotherwisenon-active tion was invoked by Eracleous et al. (1995) in a model to
galaxy. The physical conditions in the HII emission-line explain the UV properties of LINERs. Rees (1988, 1990)
gasare also investigated.We briefly comment ona search suggested that the flare of electromagnetic radiation pre-
for further highly variable objects on the basis of ROSAT dicted when the stellar debris is swallowed by the black
observations. Several with factors 10-20 are found; all of hole could be used as a tracer of SMBHs in nearby non-
them are known to harbour warm absorbers. active galaxies. This question is of particular interest in
thecontextofAGNevolution;i.e.,whatfractionofgalax-
Key words: Galaxies: active – individual: NGC5905, ies have passed through an active phase, and how many
IC3599 – emission lines – Starbursts – X-rays: galaxies nowhavenon-accretingandhenceunseenSMBHsattheir
centers (e.g., Rees 1989)? Several approaches were fol-
lowed to study this question. Much effort has been put
Sendoffprintrequeststo:S.Komossa,[email protected] in deriving central object masses from studies of the dy-
2 St.Komossa, N.Bade The giant X-ray outburstsin NGC5905 and IC3599
namics of stars and gas in the nuclei of nearby galaxies. HRI. An HRI observation of NGC5905 was performed
Earlier groundbased evidence for central quiescent dark fromNovember18–21,1996,centeredonthetargetsource.
masses in non-active galaxies (e.g., Tonry 1987, Dressler The total exposure time is about 76 ksec. X-ray emission
&Richstone1988,Kormendy&Richstone1992)hasbeen from NGC5905 is detected with a background-corrected
strengthened by recent HST results (e.g., van der Marel countrate of 0.0007±0.0001 cts/s. This indicates a fur-
et al. 1997,Kormendy et al.1996;see Kormendy & Rich- ther drop in brightness by a factor ∼2 as compared to
stone 1995 for a review). Evidence for a SMBH in our the 3-yr earlier last PSPC observation. The longterm X-
galactic center is strengthening as well (Eckart & Genzel ray lightcurve is displayed in Fig. 9. There, we converted
1996). the HRI countrate to PSPC countrate assuming constant
X-raystracetheveryvicinityoftheSMBHandprovide spectral shape.
animportantprobeoftheAGN’scentralregions.Informa-
tion can be extracted from X-ray variability and contin-
2.2. Optical spectra
uumshape(e.g.,Mushotzkyetal.1993)andstrengthsand
profiles of X-ray emission lines (e.g., Tanaka et al. 1995,
In March 1997 we took spectra of NGC5905 and IC3599
Bromley et al. 1998). This X-ray approach was hitherto
with the focal reducer MOSCA at the Calar Alto 3.5m
mainly applied to galaxies that were previously known to
telescope in the green (4150− 6650˚A) and red (5800−
be active.1
8300˚A) wavelength region. For NGC5905 the slit width
Presenting new observations and outburst models in was1′′whichyieldsaspectralresolutionof3˚A,inthecase
the present study we critically assess the possibility of an of IC3599 the slit width was 1′.′5 and the spectral resolu-
event of tidal disruption in NGC5905 (and IC3599) and
tion is 4.5˚A. In both cases the slit was oriented in E-W
contrastthis with otherpossible outburstscenarios,some
direction. The optical data were reduced within the MI-
(but not all) of which involve the presence of a SMBH as
DAS package. The wavelength calibration was performed
well.
with Hg-Ar-Ne comparison frames and the velocity zero
The paper is organized as follows: In Sect. 2 we de-
point was set by the night sky lines onthe galaxyframes.
scribe the data reduction. Section 3 deals with the X-ray
Relativeintensitycalibrationwasachievedbytakingspec-
spectralanalysisandin Sect.4 we providethe opticalob-
tra of the spectrophotometric standard star GD71.
servations of NGC5905 and IC3599. Section 5 presents
photoionization models for the optical outburst emission-
line spectrum ofIC3599.Sections 6–9areconcernedwith
the discussionofthe data.InSect.6 wescrutinizescenar- 3. X-ray spectral analysis
ios for the X-ray outburst in NGC5905, Sect. 7 investi-
gates the properties of NGC5905 in quiescence, in Sect. 3.1. NGC 5905
8 we perform a comparison with IC3599, and in Sect. 9
3.1.1. Standard models
we briefly present results from a search for further X-ray
outbursters. Section 10 provides a summary and the con-
In addition to the analysis in paper I we add several fur-
clusions.
thermodels,whichwelateruseindiscussingpossibleout-
Luminosities are given for H =50 km/s/Mpc if not
o burst scenarios. The results for the powerlaw (pl) and
stated otherwise.
pl+blackbody (bb) model fit of paper I are summarized
in Tab. 1. In the following,we first fit the high-state data
2. Data reduction due to better photonstatistics, andthen commenton the
low-state spectrum. If not stated otherwise,N was fixed
H
2.1. X-ray data to the Galactic value, Ngal = 1.47 1020 cm−2 (Dickey &
H
PSPC. TheROSATPSPC(Tru¨mper1983,1990;Pfeffer- Lockman1990).Applying a single bb, we find a tempera-
ture kT = 0.06 keV, an intrinsic (0.1-2.4 keV) luminosity
mann et al. 1987) data reduction for NGC5905 was de-
L = 2.6 1042 erg/s, and χ2 = 1.2. The corresponding
scribed in paper I. In the same manner, we treated the x red
values for emission from a Raymond-Smith (rs) plasma
PSPC all-sky survey data of IC3599. In brief, photons
are kT = 0.1 keV, L = 2.2 1042 erg/s, and χ2 = 1.2.
were extracted in a circular cell aroundthe targetsource, x red
If N is treated as free parameter we get N = 4.4 1020
the backgroundwas determined from a source free region H H
cm−2, kT = 0.06 keV, L = 44.5 1042 erg/s, and χ2 =
alongthescanningdirectionofthetelescopeandcorrected x red
1.1. Fitting the accretion disk model included in EXSAS
for. The data reduction was performed using the EXSAS
software package (Zimmermann et al. 1994a). yieldsforaBHmassof104 M⊙ anaccretionrateinterms
of the Eddington rate of m˙ = 0.2, L = 2.9 1042 erg/s,
x
1 Oneexception,thepresumedHIIgalaxyIRAS15564+6359 and χ2 = 1.1. Some models are shown and compared in
red
that showed X-ray variability (Boller et al. 1994) turned out Fig. 1.
to be active, as judged by optical spectra (Yegiazarian &
Khachikian 1988, Halpern et al. 1995).
3
Table1.SummaryofsomeX-rayspectralfitstoIC3599andNGC5905(survey=high-stateobservation,pointing=low-
state).
galaxy model → pl, N free pl(1) pl + bb(1) ref
H
N(2) Γ χ2 Γ χ2 Γ kT(3) χ2
H x red x red x bb red
IC3599 survey 5.2 –4.8 1.5 –3.1 6.4 –2.2 91 2.2 BPF95
–3.1 3.1 –2.0(4) 91 1.4 G&95
pointing II 2.9 –4.6 0.9 BPF95
–3.3 1.2 –2.0(4) 65 1.0 G&95
NGC5905 survey 2.9 –5.1 1.0 –4.0 1.26 –1.9(4) 48 0.9 paper I
pointing 8.2 –4.5 1.0 –2.4 1.2
(1)N fixed to NGal (= 1.47, 1.3 1020cm−2 for NGC5905 and IC3599, respectively), (2)in 1020cm−2, (3)in eV, (4)fixed
H H
3.1.2. Warm absorber models be present, being depleted by a factor of ∼20, it becomes
possible to fit the high-state spectrum (with log U ≃ 0.2,
The presence of a warm absorber could explain the ob-
log N ≃ 22.9).
served variability and softness of the X-ray spectrum of w
NGC5905 (Komossa & Fink 1997a). This model is dis-
cussed further in Sect. 6.5. To test for the presence of a 3.1.3. Combined fits
warmabsorber,wehaveappliedourCloudy-basedmodels
Here we fit the low-state data, fixing some of the param-
(describede.g.inKomossa&Fink 1997b-d;fordetails on
eters to the values determined for the high-state observa-
the photoionization code Cloudy see Ferland 1993, 1998).
tion. This is to testwhether the data canbe explainedby
The properties of the ionized material which are derived
requiring the change of one parameter only, or as few as
from X-ray spectral fits are the total hydrogen column
possible. The following descriptions were applied:
density N of the warm absorber and the ionization pa-
w
(i)Pureabsorptionvariability.Isthedeclineinluminosity
rameter U, defined as
caused by increased absorption, e.g., by a cloud blocking
∞
f the line-of-sight (l.o.s.) during the low-state or by a tran-
U =Q/(4πr2n c) , with Q∝ νdν (1)
H Z hν sitory decrease in absorption during the high-state ? To
ν0 test this possibility, the pl index and norm were fixed to
where f is the flux of the incident continuum, ν the thevaluesdeterminedforthehigh-state,andthelow-state
ν 0
frequency at the Lyman limit. We used a mean Seyfert was fit with N as free parameter. The quality of the fit
H
spectrum of piecewise powerlaws with αuv−x=–1.4 in the turns out to be unacceptable (χ2red=2.8).
EUV and Γ = –1.9. Solar abundances (Grevesse & An- (ii) Changes in warm absorption. Ionized material re-
x
ders 1989) were adopted if not stated otherwise. sponds to changes in the impinging photon flux. The ob-
The fitofa warmabsorberto the high-state spectrum served spectrum shows a flattening with increasing time
proves to be successful with an ionization parameter of from the outburst. Whereas a warm absorber is expected
log U ≃ 0.0 and a column density log N ≃ 22.8 (χ2 to first respond by deeper absorption (causing a steeper
w red
= 1.0). In a second step, we applied the model of a dusty observed spectrum) to a decline in intrinsic luminosity it
warm absorber. The presence of dust could obscure the then goes from ‘warm’ to ‘luke-warm’, shifting the edges
nucleus at optical/UV wavelengths, and explain why the to more lowly ionized species, thereby flattening the ob-
optical spectrum of NGC5905 looks like that of a non- served spectrum. In any case, one expects lower U in the
active galaxy (for details see below). low-state. Whereas a warm absorber can formally be fit
The dusty models were again calculated with Cloudy. to the low-state spectrum, we find a higher ionization pa-
We first assumed a Galactic gas/dust ratio, the dust rameter (log U ≃ 0.5) as well as higher column density
speciesgraphiteandsilicate,anddepletedgas-phasemetal (log N ≃23.2).
w
abundances.ThephotonindexoftheintrinsicX-rayspec- (iii) Same bump component, declined in intensity and/or
trumwasfixedtothecanonicalSeyfertvalueofΓ =–1.9. temperature.Singlecomponentfits (like bborrs)already
x
No fit is possible. This can be traced back to the strong show that an increase in temperature (in addition to the
‘flattening effect’ of dust (Komossa & Fink 1997b-d; Ko- decrease in luminosity) is found for all models, if T is left
mossa & Bade 1998)that is in conflict with the steep ob- as free parameter. If constant T is enforced, i.e. fixed to
servedspectrum.Severalmodelswithmodifieddustprop- the value determined for the high-state observation,χ2 is
ertieswereexplored.Ifweallowonlythesilicatespeciesto unacceptably large (χ2 ∼> 2). In the accretion disk de-
red
4 St.Komossa, N.Bade The giant X-ray outburstsin NGC5905 and IC3599
scription, if the black hole mass is fixed, m˙ = 1 in the
low-state, and the fit is still unacceptable.
(iv) Variable bump component on top of constant under-
lyingpl.Fixingthenormoftheunderlyingpltothehigh-
state value (in the pl+bbdescription)it is notpossible to
fitthelow-state.Suchatwo-componentdescriptionofthe
low-state anyway does not lead to an improvement of the
fit.Ifitisneverthelessapplied,asuccessfulfitrequiresthe
powerlawcomponentto decline by abouta factor of10in
flux, and kT to decrease to 12 eV (as compared to 48
bb
eV in the high-state).
Since noneofthe‘tight’descriptionsprovedtobesuc-
cessful, we conclude that the spectral component seen in
the PSPC low-state (with L ≃ 4 × 1040 erg/s and Γ
x x
≃ −2.4) is not the same as the one in the high-state,
andmostprobablydominatedbythe emissionofthe host
galaxy, which is typically of the order 1038−40 erg/s in
non-active galaxies (e.g., Fabbiano 1989, Vogler 1997).
3.2. IC3599
Standardfits (plandbb)aresummarizedinTab.1.Here,
we comment on the possible presence of a warmabsorber
since no such model has been applied previously. Fitting
this model to the high-state X-ray spectrum (with Γ =–
x
1.9 fixed) yields log U ≃0.6 , log N ≃23.3 and cold ab-
w
sorptionconsistentwiththe Galactic value.The fitcomes
with a rather high χ2 = 2.4. However, none of the ap-
red
Fig.1. Comparison of several selected spectral fits to
plied models gives an acceptable fit (cf. Tab. 1)the warm
the high-state and low-state data of NGC5905. Upper
absorber being as successful as the pl + bb description.
panel: Different descriptions of the high-state spectrum;
solid line: warm absorber, thin long-dashed line: black
4. Optical observations
body, dotted line: powerlaw. In all these models, N was
H
4.1. Photometry fixed to the Galactic value. The short-dashed line corre-
sponds to a powerlaw with N treated as free parameter.
H
Depending on the ratio Lopt/Lx of the variable compo- Whereas the models are rather similar in the soft energy
nent, we might expect to see some optical variability as region (e.g., compare the bb with the warm absorber),
well; a factor 100 in the variable component would corre- they show very different high-energy behaviour. In fact,
spondto∼5m (but‘smeared’duetoseeing).Tosearchfor the single thermal models tend to underpredict the ob-
optical variability, we used the photographic plates avail- servedfluxabove1keV.Lower panel:Low-statedataas
able at Sonneberg Observatory. For an overview of the thin lines; solid line: powerlaw, dashed line: black body.
programandplatedetailsseeBra¨uer&Fuhrmann(1992). For comparison, with thick lines the same models for the
We examined all plates taken around the time of the ob- high-state data are re-plotted.
servedX-rayoutburstaswellasarandomsampleofplates
taken on a longer timescale (between the years 1962 and
1995).We mainly used the blue plates due to their better
4.2. Optical post-outburst spectra
quality; in the time-interval of the X-ray observation we
also checked the red plates. 4.2.1. NGC5905
The brightness of the nucleus of NGC5905 was mea-
Emission line strengths. The optical spectrum of
suredrelativetothenucleusofthesimilarlybrightnearby
NGC5905 is displayed in Fig. 3. It shows strong emis-
non-active spiral galaxy NGC5908. The lightcurve is
shown in Fig. 2. Within the error of ∼0.2m the optical sion lines on a continuum with absorption lines produced
by the stellar content of the nucleus. The contribution of
brightnessis constant.This holds for the time aroundthe
the stellar content is higher in NGC5905 than in IC3599
X-ray outburst as well as on long terms, and places tight
(judged by the deeper absorption in the former’s spec-
constraints on the outburst scenarios (Sect. 6).
trum).The Balmerlines arecomposedofanemissionline
plusabroaderabsorptionwhichhintstoalargeportionof
5
Fig.2. Optical lightcurve of the nu-
cleus of NGC5905, based on photo-
graphicplatestakenatSonnebergOb-
servatory. The abscissa gives the Ju-
lian date in JD-2400000 (the data
points bracket the time interval 1962
– 1995). The upper panel shows the
longtermlightcurve,theloweronethe
lightcurve in the months around the
X-ray outburst. Within the error of
about 0.2m the optical brightness is
constant.
Fig.3. OpticalspectrumofthenucleusofNGC5905takenabout6.7yearsaftertheX-rayoutburst.Thearrowsmark
the detected emission lines. The weakness of [OIII]λ5007 compared to Hβ and the narrow lines classify the galaxy as
HII-type. The positions of some high-ionization lines, not detected in the spectrum, are marked as well.
moderatelyyoungstarsinthenucleus ofNGC5905(stars We find clear signs of a rotation curve in the spec-
of spectral type B, A and F). Further detected emission tra of NGC5905. The real (un-broadened) width of the
lines are those of [OI], [OIII], [NII] and [SII] (Tab. 2). In lines isthereforeevensmallerthangivenabove.Toderive
Fig.5emissionlineratiosareplottedinthediagnosticdia- the rotation curve, we extracted one dimensional spectra
gramsofVeilleux&Osterbrock(1987).Nohigh-ionization of NGC5905. The center shows a decreased Hα emission
lines like HeII or [FeX] were detected. line strengthwhichcanbe interpretedwith absorbinggas
along the line of sight. The E-W orientation of the slit
is approximately perpendicular to the orientation of the
Line profiles. The emission line widths were determined bar. To extract line of sight velocities along the slit we
by fitting Gaussians to the lines. The instrumental width measured the position of the emission lines for each CCD
was derived from night-sky lines which yielded 2.9±0.3˚A. rownearthe galaxycenter.AGaussianwasfittedto each
Thelines areverynarrow,oforder270–300km/sFWHM emission line to determine its wavelength. The distance
(Tab.2;notcorrectedforinstrumentalresolution).Inpar- between the rows (0′.′32) is clearly smaller than the see-
ticular, we find no hints for an underlying broad emis- ing of 1′.′2. Therefore fine structure and a higher gradient
sion component that would indicate permanent low-level in the rotation curve is expected for spectra with better
Seyfert activity. All properties of the emission line spec- spatialresolution.Thehighestrotationvelocityisreached
trum are fully compatible with spectra observed in HII 2′′ outside the center which corresponds to 0.72kpc. The
galaxies.
6 St.Komossa, N.Bade The giant X-ray outburstsin NGC5905 and IC3599
Fig.5. Locations of NGC5905 (filled
square) and IC3599 (open square)
in the standard diagnostic diagrams,
derived from single-component Gaus-
sian fits to our emission line spectra.
Table 2.OpticalemissionlineratiosrelativetoHβ (in- 4.2.2. IC3599
qu
dex‘qu’for‘quiescence’)andlinewidths (forthe stronger
Emission line strengths. The optical spectrum of IC3599
lines only) of NGC5905 und IC3599 in quiescence. (For
differs from that of NGC5905 in several respects (Figs.
comparison:BPF95measureforthe outburstspectrumof
IC3599 FWHM ≃ 1200 km/s and FWHM ≃ 1500 6, 7). It shows stronger [OIII] emission and some high-
Hβ Hα
ionization lines, like HeIIλ4686 and [FeVII]λ6078. These
km/s.) Results given in the Tab. were derived by fitting
have not changed in strength in comparison to the 1995
a one-component Gaussian to each emission line; see text
post-outburst spectrum in G&95, indicating these lines
for details.
maybepermanent.Wedonotfindsignsforemissionfrom
FeII.
Line I/IHβqu FWHM[km/s]
NGC5905 IC3599 NGC5905 IC3599
Line profiles. The instrumental width was determined to
be 4.5±0.5˚A. Line widths given below and in Tab. 2 are
Hγ 4340 0.20 0.26
[OIII] 4363 0.21 0.32 corrected for this value (4.5˚A). The Balmer lines have
HeII 4686 - 0.37 870 larger widths than the forbidden lines of [OIII] and [SII],
Hβ 4861 1 1 270 480 but the differences in line widths and the width of Hβ
[OIII] 5007 0.49 3.2 300 260 (480km/s)wouldstillbeconsistentwithaSeyfert2.How-
[FeVII] 5721 - 0.10 440 ever, the complex of Hα and the two [NII] lines is signifi-
[FeVII] 6087 - 0.23 360 cantlybetterdescribedifweassumetwoGaussiancompo-
[OI] 6300 0.30 0.18 430 nentsforHαwithFWHMsof400kms−1and1090kms−1,
Hα 6563 6.4 5.2 280 480 both of similar intensity.2
[NII] 6584 2.7 1.6 290 430
These spectral results obtained by our new spectra
[SII] 6716 0.86 0.3 280 210
[SII] 6731 0.66 ∗ 280 with higher spectral resolution suggest an intermediate-
typeSeyfertclassificationforIC3599,namelyaSeyfert1.9
∗
overlaps with atmospheric O λ6880 absorption.
2 in the scheme of Osterbrock (1981).
5. Photoionization models for the optical
outburst spectrum of IC3599
5.1. Model assumptions
In order to model the high-ionization emission line com-
ponentintheopticaloutburstspectrumofIC3599(taken
∼0.5 yr after the observed X-ray outburst; BPF95),
thereby deriving the physical properties of the illumi-
nated gas, we have calculated a sequence of photoioniza-
tionmodels,againusingthecodeCloudy.Weassumecon-
stantdensitygasionizedbyapointlikecentralcontinuum
Fig.4. Line of sight velocities of NGC5905 measured in
E-W direction with a slit width of 1′′. source and employ the mean Seyfert spectrum described
in Sect. 3.1.2except for modifications in the EUV and X-
ray range to account for the observed X-ray spectral and
variability properties.
velocity curve is shown in Fig. 4 and further discussed in
2 FittingGaussianstoemissionlinestoassessthepresenceof
Sect. 7.3.
abroadcomponentisstandardpracticewhichalsowefollowed;
‘real’ line profiles may be of more complex shape, though.
7
density, the other with high density. The high density so-
lution is needed to successfully account for the strength
of OIλ8446 (which is underpredicted otherwise) and to
avoidthe overpredictionofother low-ionizationforbidden
lines.
The best match is obtained for clouds of column den-
sity N ≃ 1023 cm−2, a gas density n ≃ 109 cm−3, and
H H
a clouddistance fromthe nucleus of r≃0.1 pc. Observed
and modeled emission lines are compared in Tab. 3. The
derived parameters are typical for BLR or CLR clouds4
suggestingthatIC3599permanentlyharboursaBLRthat
isusuallytooweaktobeeasilydetected(orismainlyhid-
Fig.6. Spectrum of IC3599 (bottom) between 4600 and den) but was illuminated by the X-ray outburst.
5200˚A rest wavelength. The one of NGC5905 (top) is
shownforcomparison.IC3599clearlyshowsbroaderemis-
Table 3. Comparison of observed (BPF95) and modeled
sion lines.
outburst emission lines (I/I , ‘out’ for outburst) of
Hβout
IC3599.
I/IHβout
Line observed modeled
HeII 4686 0.36 0.38
Hβ 4861 1.0 1.0
HeI 5876 0.14 0.12
[OIII]5007 < 0.2 0.02
[FeX]6375 0.37 0.45
[FeXI]7892 0.23 0.36
[FeXIV]5303 0.17 0.11
OI 8446 0.23 0.10
Fig.7. Spectra ofIC3599(bottom) andNGC5905(top)
between 6000 and 6750˚A rest wavelength.
Emissionline ratioswereextractedfromthe spectrum In traveling through the inner regions of the galaxy,
of BPF95. A large grid of models was calculated, varying the outburst emission acts as a probe of this region. It
all relevant unknown parameters (i.e., the EUV–X con- will be interesting to see when the emission reaches the
tinuum shape, the density n, the column density N, and distance, where in active galaxies the bulk of the NLR is
the distance fromthe nucleusr; abundanceswerefixedto located.
the solar value) with the aim of matching the observed
line-ratiosduringoutburstwithinafactorof2(note,e.g.,
6. Discussion: NGC5905 outburst scenarios
uncertainties in collisional strengths of coronallines). We
alsocarefullychecked(anditturnedouttobe important) TheX-rayemissionofnormalspiralsistoolowtoaccount
that no other, non-detected, lines were overpredicted by for the maximum X-ray luminosity observed and drastic
the models (like [OIII], [OI])3. variability is not expected (see also paper I). Below, we
scrutinize scenarios that may explain the observations.
5.2. Results
6.1. Supernova
We find that the high observed X-ray flux in form of a
strong ‘hot’ soft excess and a large gas column density X-raysare firstemitted during the SN outburst itself. Al-
are required to produce the observed wide range in ion- though energetically feasible, the timescale is too short
ization states (like O0 and Fe13+). For the emission in (seconds to hours, instead of days) to account for the ob-
the Fe lines there are always two solutions, one with low served X-ray outburst in NGC5905. On a longer time-
scale, further radiation originates in a shock produced by
3 The high-density gas discussed below could contribute
to [OIII], but clearly not to [SII] (much lower critical den- 4 for some recent BLR models for Seyfert galaxies see, e.g.,
sity). Therefore, we have checked that the observed ratio of Radovich&Rafanelli(1994),Baldwinetal.(1995);forcoronal
[OIII]/[SII] is constant from outburst to quiescence, i.e. there lineregion (CLR)models,e.g., Fergusonetal. (1997),Binette
is no strong contribution to[OIII] in outburst. et al. (1997).
8 St.Komossa, N.Bade The giant X-ray outburstsin NGC5905 and IC3599
expansion of the SN ejecta into the nearby circumstellar assumingline coolingtodominatewithacoolingfunction
medium. The X-ray luminosity of the few detected SNe of Λ∝T−0.6, as expected from the X-ray fit.
is much weaker and the spectra are harder than that of
NGC 5905, though.5 Since it is unclear in which evolutionary state
NGC5905 has been observed we make the following as-
sumptions:
6.2. SN in dense medium (1) We assume the observed luminosity to be the peak
luminosity (and N = Ngal), dominantly emitted in the
The possibility and consequences of the existence of SNe H H
X-ray regime. These assumptions go in the direction of
in dense molecular gas was studied by Shull (1980) and
requiring less extreme conditions. We then check, if, and
Wheeler et al. (1980). The idea was revived by Terlevich
underwhichconditions,theobservedhighluminosity,and
(1992)inthecontextofastarburstinterpretationofAGN.
the quick decline in L can be reproduced within this
Thebasicscenarioisthefollowing:Thematterejected
scenario. An energy of E = 1051 erg is typically de-
by the SN explosion expands into a region of high inter-
posited by the explosion. At t=t , and for a measured
stellargasdensity.Theinteractioncausesashockedregion rad
L ≈ 3×1042 erg/s, the above equations can be solved
of hot gas enclosed by two shock waves,on the outside of x
theoutgoingshock,andontheinsideoftheinwardreverse for the density, yielding n4 ≃ Lx/(2.5×106L⊙) ⇒ n =
3×106cm−3. A decrease in luminosity within 5 months
shock. The outwardshock encounters dense circumstellar
(1.45 yr) of a factor of ∼14 (20) is then predicted, which
material. After sweeping up a small amount of gas these
compares to the observed drop of a factor of >5 (100).
remnantsbecomestronglyradiative(t=t )anddeposit
rad
Whereasthesetimescalesroughlyagree,theexpectedtem-
mostoftheirkineticenergyinveryshorttime-scales,thus
perature is T ≃ 108 K, suggesting a much harder X-ray
reachingveryhighluminosities.Becauseofthelargeshock
spectrum than is observed.
velocities most of the energy is radiated in the EUV and
(2) In a second step, we drop the assumption that L was
X-rayregion.Forexpansionintodensematerial,allevolu-
observed at its peak value, and use the temperature de-
tionary phases are substantially speeded up as compared
to the ‘standard’ case of n≈100 cm−3. rived from the X-ray Raymond-Smith (rs) fit as an esti-
mateofT,withT ≈7×105 K,1×106Kforthersmodel
Sinceinthisscenariohighluminositiescanbereached,
with free or fixed N , respectively.In both cases the esti-
andthe evolutionarytime is considerablyshortened(only H
mateddensitiesareunrealisticallylarge,n >1010 cm−3.
of the order of years to few tens of years), a SN in dense H
medium may be an explanation for the observed X-ray
In conclusion, extreme conditions would be required
outburst in NGC5905.In the following we firstgive some
and the predicted T disagree. Additionally, fine-tuning in
estimatesusingtheanalyticaltreatmentpresentedinShull
the column density of the surrounding medium would be
(1980) and Wheeler et al. (1980).
neededinordertopreventtheSNRfrombeingcompletely
Assume SN ejecta of initial total energy E = 1051E51 self-absorbed. Thus, we consider this scenario very un-
erg,expandingintoamediumofdensityn=104n4 cm−3. likely.
Most of the energy is emitted during/shortly after the
onset of the radiative phase, which starts after a time
6.3. Accretion disk instabilities
t=E0.22n−0.56(83y) . (2)
51 4
The evolution of temperature, radius and luminosity of If a massive BH exists in NGC5905, it usually has to ac-
the shock is then given by crete with a very low accretion rate or radiate with a low
efficiency, or be completely dormant. One solution in the
T =E501.14n04.27(t/trad)−10/7(2.7107K) , (3) framework of accretion disks that has recently received a
lot of attention, are advection dominated disks in which
R=E501.29n−40.43(t/trad)2/7(0.29pc) , (4) mostenergyisadvectedwiththeaccretionflowinsteadof
radiated away (e.g., Abramowicz et al. 1988, Narayan &
L=E501.78n04.56(t/trad)−11/7(9.81039erg/s) , (5) Yi 1994;review by Narayanet al. 1998).This may be the
dominant accretion mode in low-luminosity objects (like
5 OnlyafewSNe(nearlyall oftypeII)havebeenobserved
in NGC5905 in ‘quiescence’).
inX-rayssofar(reviewedbySchlegel1995).Themaximalob-
servedluminositiesreachedinthesoftX-raybandare(Schlegel
An instability in the accretion disk may then provide
1995, assuming H = 75) 2.7 1039 erg/s (SN 1980K), 3.2 1040
o an explanation for the observed X-ray outburst. Ther-
erg/s (SN 1986J), 9.5 1039 erg/s (SN 1978K), 2.8 1040 erg/s
(SN1993J,5dafterexplosion),≈1041erg/s(SN1988Z;Fabian mally unstable slim accretion disks were studied by, e.g.,
& Terlevich 1996), and 1 1039 erg/s (SN 1979C; Immler et al. Honma et al. (1991), who find the disk to exhibit burst-
1998). The temperatures assuming a thermal spectrum range like oscillations for the case of the standard α viscosity
from ∼0.5 keVto > 7keV(Canizares et al. 1992, Bregman & description and for certain values of accretion rate. The
Pildis 1992, Schlegel 1995, Zimmermann et al. 1994b). high-luminosity state is characterized by a high mass ac-
9
change in luminosity by a factor of less than 10 is needed
to change the absorption to complete within the ROSAT
band(Fig.8),veilingtheviewofthenucleusinsoftX-rays.
Variability of such order is not unusual in low-luminosity
AGN.
SincethereisnoevidenceforSeyfertactivityintheop-
ticalspectrum,thenucleusmustbehiddeninthisspectral
region.Assuming dust to be mixed with the total column
density of the warm absorber would result in an optical
extinction of A ≈ 34m (for a Galactic gas/dust ratio)
v
and would hide the Seyfert nucleus completely. However,
this dust strongly modifies the X-ray spectrum and such
a model does no longer fit the X-ray observation. As we
saw in Sect. 3.1.2, only fine-tuning in the dust properties
Fig.8. Best-fit warm absorber model of the ‘outburst’
enabled a successful spectral fit in terms of a dusty warm
spectrum of NGC5905 (upper solid curve), compared to
absorber. Alternatively, a steep underlying intrinsic spec-
the predicted spectrum resulting from a decrease in the
trum could be assumed to compensate for the flattening
intrinsic luminosity by a factor of 10 for constant column
effectofdust.Thisisnotanelegantsolution,though,since
densityN (lowersolidcurve).Theverticalbarmarksthe
w part of the originalaim was to explain the steep observed
end of the ROSAT energy band. The predicted countrate
spectrum.
for the low-L model is below the detection limit in the
To summarize, a warm absorbed hidden Seyfert nu-
pointed ROSAT observation. The dotted line represents
cleus only explains the observation when strongly fine-
the dusty modelforcomparison(forsilicate only,0.05de-
tuning the dust-properties.
pleted).
6.5. Tidal disruption of star
cretion rate in the inner disk region, lasting for a time
of Depending on its trajectory, a star gets tidally disrupted
afterpassingacertaindistancetotheblackhole,thetidal
tburst ≃1.9α−0.01.64(M/M⊙)1.36 sec (6) radius, and the debris is accreted by the hole. This pro-
where α = 0.1α is the standard viscosity parameter duces a flare of electromagnetic radiation. The peak lu-
0.1
(e.g., Frank et al. 1985) and M is the mass of the central minosity should be a substantial fraction of the Edding-
black hole. (For assumptions of this analytic result see ton luminosity. The timescale after which the bulk of the
Honmaetal.1991;fornumericalresultsforawiderrange material is swallowed is estimated to be of the order of
of parameters cf. their Fig. 11). If the duration of the months, after that, the luminosity declines as ∝ t−5/3
outburst is less than 5 months (i.e. the survey variability (Rees 1990). Detailed numerical simulations of the dis-
is taken as a real rise in L), the mass of the central black ruption process (e.g., Laguna et al. 1993, Diener et al.
hole in NGC5905 should be less than ∼105M⊙. 1997)and consequences for the surroundings (e.g.,Rauch
Theseburst-likeoscillationsarefoundbyHonmaetal. &Ingalls1998)havebeencarriedoutinthelastfewyears.
only for certain values of the initial accretion rate. A de- Explicitpredictionsoftheemittedspectrumandluminos-
tailedquantitativecomparisonwiththeobservedoutburst ity arestill rare.In particular,it is uncertainwhether the
in NGC5905 is difficult, since the behaviour of the disk emissionpeaksintheUVspectralregion(e.g.,Loeb&Ul-
is quite model dependent, and further detailed modelling mer 1997, Ulmer et al. 1998) or in the EUV – softX-rays
would be needed. (e.g., Rees 1988, Sembay & West 1993).
If the X-ray outburst of NGC5905 is interpreted as a
tidaldisruptionevent,the blackholemasscanberoughly
6.4. Warm absorber
estimated viathe Eddingtonluminosity whichis givenby
In this scenario, NGC5905 would host an active nucleus Ledd ≃1.3×1038M/M⊙ erg/s . (7)
that is usually hidden by a large column density of cold
absorbinggas.TheSeyfertnucleusisintrinsicallyvariable, In case of NGC5905, a BH mass of ≈ 105 M⊙ would be
andbecomesvisible(inX-rays)duringitsfluxhigh-states sufficient to produce the observed Lx (assuming NH =
byionizingtheambientmediumthatbecomesawarmab- Ngal and L to be observed near maximum; this provides
H
sorberandtransparenttosoftX-rays.Thiscouldnaturally a lower limit on the BH mass). In order to reach this
explaintheobservedvariabilityaswellasthesteepnessof luminosity, L = ηM˙ c2, a mass supply of about 1/2000
acc
the outburst spectrum. M⊙/year is necessary, assuming η=0.1. Rees predicts the
Wefoundthatthehigh-statespectrumcanbewellde- event to last of the order of several months, consistent
scribedbya(dust-free)warmabsorber.Asource-intrinsic with the observations of NGC5905 (cf. Fig. 9).
10 St.Komossa, N.Bade The giant X-ray outburstsin NGC5905 and IC3599
Again, a more detailed analysis has to awaitavailabil- ruption of a star because it can account for the high out-
ity of more sophisticated models. In particular, the flares burst luminosity, seems to require least fine-tuning, and
cannot be standardised. Observations would depend on the long-termX-ray lightcurve shows a continuous fading
many parameters, like the type of disrupted star, the im- of the source over the whole measured time interval (Fig.
pact parameter, the spin of the black hole, effects of rel- 9). We re-caution, though, that many details of the ac-
ativistic precession, and the radiative transfer is compli- tual process of tidal disruption of a star and the emitted
cated by effects of viscosity and shocks (e.g., Rees 1994). spectrum are still rather unclear.
6.6. Further possibilities 7. Discussion: Properties of NGC5905 in
quiescence
Gravitational lensing event. High magnification factors
can be achieved by lensing if the lense lies nearly ex- 7.1. Classification
actlyonthelinebetweenobserverandsource(seeSchnei-
Since the optical spectra obtained in 1997 (6.7 yr after
der, Ehlers & Falco 1992).6 If the X-ray variability of
the X-ray outburst) do not show evidence for an high-
NGC5905 was caused by gravitational lensing one would
excitationemissionlinecomponentweassumethesespec-
expect the same magnification factor in the optical as in
tratorepresentthepropertiesofNGC5905inquiescence.
X-rays (assuming emission regions of similar size); and
We have carefully searched for signs of permanent low-
the spectral component observed in the X-ray low-state
level Seyfert activity. This question is important in the
has to be of different originthen, since the spectralshape
contextofAGNformationandevolution(e.g.,Rees1989),
changed. This contradicts the optical observations.Grav-
i.e.,havewefoundevidence(viatheX-rayoutburst)fora
itational lensing therefore seems to be very unlikely.
SMBH in a galaxy that otherwise appears perfectly non-
X-ray afterglow of a GRB. For completeness, we finally active to an observer ?8
note that within the time interval of the observed X-ray Wefindtheemissionlinestobeverynarrow,andlowly
outburst no gamma-ray burst (GRB) has been reported7 excited, as typical for HII-type galaxies; we do not de-
(the 2 detected GRBs ofJuly 8,1990,havedifferentposi- tect high-ionization lines like [FeVII], and we do not find
tions;Castro-Tirado1994onthebasisofGranat/WATCH any signs for the presence of a broad component in the
data). We made the same check for IC3599. No simulta- Balmerlines.(Suchastatement,ofcourse,alwaysreflects
neous GRB is reported. The one nearest the time of the the current instrumental sensitivity limits.) The same re-
X-ray outburst (Dec. 9, 1990) of Dec. 19, 1990, has a dif- sults holdforapre-outburstopticalspectrumgiveninHo
ferent position. et al. (1995).
7.2. HII properties
6.7. Summary of ‘successful’ outburst scenarios
Second, we determine some of the properties of the low-
All those models that fullfil the available observational
ionization (HII-like) emission line component.
constraints plus the order-of-magnitude estimates pre-
Extinction from Balmer lines. The observed Balmer
sented above involve the presence of a SMBH in the cen-
decrement is above its recombination value, indicative of
ter of the galaxy. A particularly tight constraint is the
reddening by dust intrinsic to the emission line gas or
huge observed outburst luminosity L of at least several
x
1042 erg/s.Theactualpeakluminositymayevenbemuch within the host galaxy along the l.o.s. Assuming all dust
tobe locatedinascreenoutsidethe emissionline gasand
highersince(i)wemaynothaveobservedexactlyatmax-
anintrinsicHα/Hβ ratioof2.7(Brocklehurst1971)yields
imum light, (ii) the X-ray spectrum may extent into the
EUV,and(iii)theremaybeexcessabsorption(weusually an extinction of EB−V = 0.75m for a Galactic extinction
law as in Osterbrock (1989).
assumedN =N ). We favourthe scenarioof tidaldis-
H Gal
Densityviasulphurratio.Theobservedintensityratio
6 Ostriker&Vietri(e.g.,1985)andothersdiscussedthepos- [SII]λ6716/6731implies a density of the emission-line gas
sibility to explain the strong and rapid variability of BL Lac of n ≃ 102 cm−3 (Osterbrock 1989), a value typical for
objectsviagravitationallensing.Asagoodcandidateforlens- HII regions.
ing, Nottale (1986) presented the BL Lac object 0846+51W1 Temperature via oxygen ratio. The oxygen intensity
and successfully modeled the optical lightcurve which varied ratio [OIII]λ5007/4363 is unusually high. Taken at face
by about 4m within one month. value, it implies a temperature of T > 20000 K; but see
7 ThepossibleconnectionofGRBswithnearbygalaxieswas
discussed by, e.g., Larson et al. (1996), but see Schaefer et 8 Evidence for a population of non-active galaxies of un-
al. (1997). Recently,evidencehasaccumulated thatGRBsare usually high X-ray luminosity has occasionally been reported.
extragalacticandmayoriginateinstarburstregionsofgalaxies However, optical follow-up observations revealed signs of
(e.g., Halpernet al.1998). Forthesuggestion thatGRBsmay Seyfertactivityineachspectrum(e.g.,Moranetal.1994,1996,
beextreme tidal disruption events,see Carter (1992). Wisotzki & Bade 1997).