a r
project fionaX
i
v
:a
s t
r o
-the power of dream
p
h
/
2
1
2
2
13
v
1
9
D
e
c
funk you是什么意思
2
2
The Evolution Of LMC X-4Flares:Evidence For Super-Eddington Radiation Oozing Through Inhomogeneous Polar Cap Accretion Flows ?Dae-Sik Moon,Stephen S.Eikenberry 1,&Ira M.Wasrman Department of Astronomy,Cornell University,Ithaca,NY 14853;ll.edu,ll.edu,ll.edu ABSTRACT We prent the results of two extensive Rossi X-ray Timing Explorer obrvations of large X-ray flaring episodes from the high-mass X-ray binary pulsar LMC X-4.Light curves during the flaring episodes compri bright peaks embedded in relatively fainter regions,with complex patterns of recurrence and clustering of flares.We identify pre-cursors preceding the flaring activity.Pul profiles during the flares appear to be simple sinusoids,and puld fractions are proportional to the flare intensities.We fit Gaussian functions to flare peaks to estimate the mean full-width-half-maximum to be ∼68s.Significant rapid aperiodic variability exists up to a few hertz during the flares,which is related to the appearance of narrow,spiky peaks in the light curves.While spectral fits and softness ratios show overall spectral softening as the flare intensity increas,the narrow,spiky peaks do not follow this trend.The mean fluence of the flare peaks is (3.1±2.9)×1040ergs in the 2.5–25keV energy range,
with its maximum at ∼1.9×1041ergs.The flare peak luminosity reaches up to (2.1±0.2)×1039ergs s −1,far above the Eddington luminosity of a neutron star.We discuss possible origins of the flares,and we also propo that inhomogeneous accretion columns onto the neutron star polar caps are responsible for the obrved properties.
Subject headings:accretion,accretion disks —pulsars:individual (LMC X-4)—stars:neutron —X-rays:bursts —X-rays:stars
1.INTRODUCTION
校准英文
belongtoOut of ∼200known neutron star/black hole X-ray binaries,more than 40sources have been obrved with a wide range of coherent spin period of ∼0.07–1400s,identifying the central source as a rotating,highly magnetized neutron star powered by accretion (van Paradijs 1995;Bildsten et al.1997).The accretion-powered X-ray binary pulsars (hereafter,X-ray pulsars)often show
abrupt increas in X-ray luminosity known as“bursts”or“flares.”About30%of the X-ray pulsars have been obrved to show bursts and/orflares,including(1)the low-mass X-ray binary 4U1626–67(McClintock et al.1980),(2)the bursting pulsar GRO J1744–28(Fishman et al.1995; Kouveliotou et al.1996),(3)the high-mass X-ray binaries such as LMC ,Levine et al. 1991;Woo,Clark,&Levin
e1995;Levine,Rappaport,&Zojcheski2000;Moon&Eikenberry2001) and SMC X-1(Angelini,Stella,&White1991;Moon,Eikenberry,&Wasrman2002),and(4) the transient Be binary system EXO2030+375(Parmar et al.1989).Although they are easily identified as sudden increas of X-ray luminosity in most cas,the detailed properties of the bursts/flares differ much from source to source.For instance,whereas bursts from the high-mass system SMC X-1have been detected only once despite extensive studies of the source2,another high-mass system LMC X-4has been obrved with large,mi-regularflaring episodes over the last two decades by various X-ray satellites.This diversity in bursts/flares suggests that many different mechanisms are responsible for the obrved phenomena.
Most theoretical studies of the origin of bursts/flares from X-ray pulsars have relied upon one of three following mechanisms:(1)accretion disk instabilities,(2)nuclear burning on the surface of the neutron star,and(3)non-uniform stellar winds from mass-donating companions.Obviously the most well known ca for the accretion disk instability is the X-ray bursts from the bursting pulsar GRO1744–28,for which a thermal-viscous instability in the accretion disk has been investigated to explain the origin of the ,Cannizzo1996,1997).The X-ray bursts from GRO1744–28 share many properties with the Type II X-ray bursts from the Rapid Burster(Lewin et al.1996) and the X-ray
flares from SMC X-1(Moon et al.2002).On the other hand,the prence of a He-rich dwarf companion led Brown&Bildsten(1998)to suggest the possibility that∼1000-sflares from4U1626–67are caud by carbon burning under the neutron star surface.Taam,Fryxell,& Brown(1988)attributed the origin of theflares from EXO2030+375to an accretion disk instability caud by the inhomogeneity in the high-velocity(∼550km s−1)winds from a Be star companion.
LMC X-4is a persistent,disk-fed,high-mass system with a pulsational and an orbital period of∼13.5s and∼1.4d,respectively(Kelley et al.1983).It is one of a few X-ray binaries showing a super-orbital period(∼30.3day;Ilovaisky et al.1984)possibly caud by a precessing,tilted accretion disk.The optical companion is a14th magnitude O-type star of∼15M⊙(Sanduleak &Philip1977).Over the last two decades,LMC X-4has exhibited large X-rayflaring episodes which have made it one of the most regular and reprentativeflaring X-ray sources.During the flares,the X-ray spectrum softens considerably,and the pul profiles become simple,sinusoidal (Levine et al.1991,2000).Recently,Moon&Eikenberry(2001)reported two relatively long-time scale structures in theflares:quasi-periodic oscillations of∼0.65–1.35and∼2–20mHz.They also found that the amplitudes of theflares are almost exactly modulated by tho of the neutron star’s coherent pulsations(or vice versa),indicating that theflares occur near the magnetic pole of the neutron star.Although the LMC X-4flares have been obrved for many years with various X-ray
satellites,no convincing mechanism has been propod to explain the origin of thefl, Levine et al.2000).
On the other hand,LMC X-4flares have been known as super-Eddington phenomena:some-times theflare X-ray luminosity reaches up to∼1039ergs s−1(while its normal state luminosity is ∼2×1038ergs s−1,comparable to the Eddington luminosity of the∼1.4M⊙neutron star of LMC X-4).Becau the key physical parameters of LMC X-4–such as mass,distance,and magneticfield –are relatively well determined to be1.4±0.3M⊙(Levine et al.1991),50±,Kov´a cs 2000),and∼1013G(La Barbera et al.2001),respectively,the LMC X-4flares may provide a rare example for studying super-Eddington phenomena thoroughly without assuming the key parame-ters.This is not the ca for most super-Eddington sources owing to relatively large uncertainties in their distances and mass.Considering the recent important controversy over the existence of the so-called“intermediate-mass black holes”(e.g.,King et al.2001;Begelman2001,2002),for which mass(∼102–104M⊙)is determined under the assumption that the obrved X-ray luminosity is the Eddington luminosity,the investigation of the origin of the super-Eddington radiation from the LMC X-4flares can also be uful to investigate the origin of the claimed intermediate-mass black holes.
In this paper,we prent by far the most extensive obrvations of the LMC X-4flares made with the
Rossi X-ray Timing Explorer(RXTE).We describe the RXTE obrvations and our data analys in§2;we prent the results of timing and spectral analys in§3.We discuss the possible origin of the LMC X-4flares,as well as their super-Eddington radiation,in§4,We summarize our results in§5.coolgay
裕兴新概念英语第三册
2.Obrvations and Data Analys
RXTE(Bradt,Rothschild,&Swank1993)obrved two largeflaring episodes of LMC X-4on 1996August19and1999December19during its long-exposure(49and42h)obrvations of the source.The HEAsoft3package(version5.1)was ud for analyzing the data from the Proportional Counter Array(PCA;Jahoda et al.1996).The data within30minutes after passages through the South Atlantic Anomaly and/or with a higher(>0.1)electron ratio were ignored.Only the data obtained with three or more active Proportional Counter Units(PCUs)were ud for timing analys;only the data obtained with allfive PCUs were ud for spectral analys.The photon arrival times from the Good Xenon mode were transformed to the solar system barycenter using the JPL DE400ephemeris.The Very Large Event Models of Epoch3and4were ud to subtract backgrounds for the data obtained in1996and1999,respectively.The Standard2data obtained only from the top xenon layers of the PCUs0,1,2,and4were ud in spectralfits due to responsivity problems of PCU3(R.Remillard2002,private communication).Only the spectrum
gr
between2.5–25keV range was considered for the same reason,and a systematic uncertainty of1 %was assumed in spectralfits.A total of45data gments(24from the1996obrvations and21 from the1999ones)of∼1-h length was obtained,andflares were obrved in4and6gments of the two obrvations.
3.Results
3.1.Light curves and Rapid Aperiodic Variability
3.1.1.Light curves
Becau no reliable ephemeris of the∼30.5-d super-orbital period of LMC X-4around the two obrvations reported here was available,we ud the RXTE All Sky Monitor(ASM;Bradt et al. 1993)data to examine the pha of the super-orbital motion.Figure1shows the ASM light curves around the two obrvations,indicating that they are likely for the high state of the source.
职称英语报名时间Figure2prents background-subtracted light curves of the PCA data–(a)for the1996 obrvations;(b)for the1999ones.The Modified Julian Dates(MJDs)of the start of the two light curves are50314.07905and51531.48489,respectively.The binary orbital pha in the upper x-axis was deter
mined by the LMC X-4ephemeris of Levine et al.(2000).Theflares are easily identified in thefirst four data gments of Figure2a and six data gments of Figure2b as significant increas in the photon count rates.Whereas thefirst half of the data of the1996obrvations shows strong flaring activity in the orbital phasφ≃0.26–0.6,the cond half shows no noticeable activity at the same orbital phas(although some phas are missing due to data gaps).The LMC X-4eclip appears aroundφ=1.0as an apparent decrea in the photon count rates.The1999obrvations showflaring activity atφ≃0.4–0.8.The mean photon count rates of the normal state is∼44and ∼54counts/c/PCU for the1996and the1999obrvations,respectively,and that of the eclip is∼13counts/c/PCU.(We refer to the normal state in this paper as a state abnt offlaring activity and eclip.)
Figure3prents the magnified light curves of the ten data gments containingflares with 4-s resolution.We shall call them“FL1”to“FL10”in time order.Figure3shows the complex flaring activities of LMC X-4summarized as follows.First,the light curves with strongfl, FL1–5)appear to consist of multiple peaks embedded in the relatively low-intensity regions that are still brighter than the normal state;tho with weakfl,FL6–10)show simple,small peaks connected through the regions of which brightness is comparable to that of the normal state. Secondly,FL1and FL2clearly show recurrences of intenflares;FL3and FL5show a strong concentration of intenflares with a ti
me scale of∼2000s.Finally,while most of theflare peaks show symmetric,Gaussian-like profiles,the largest ,tho at t≃3240s of FL3,t≃1770 s of FL4,and t≃2010s of FL5)appear to show rather asymmetric profiles with a slow decay.
For quantitative analys,we performed multi-component Gaussianfits to theflare peaks
in Figure3.Table1summarizes parameters of the68Gaussian-fittedflare peaks of which peak intensity is greater than90counts/c/PCU.The mean peak intensity and full-width-half-maximum (FWHM)are172±103counts/c/PCU and68±31s,respectively.The median and mean values of the reduced chi-square(=χ2ν)of thefits are1.6and3.8±5.7,respectively.This indicates that while most of theflare peaks arefitted reasonably ,χ2ν<2)with a Gaussian function, some significantly deviate from a Gaussian.If we exclude the sixflare peaks withχ2νgreater than 10(including the three aforementioned largestflare peaks that apparently show slow decays),the meanχ2νdecreas to2.2±1.9.The meanfluence of theflare peaks is(3.1±2.9)×1040ergs, with the maximum at1.9×1041ergs.Thefluence is the integrated luminosity over the FWHM, and the luminosity was computed by spectralfits(e§3.2).(All the errors quoted here reprent 1-σrms deviation.)
In Figure3,the light curves FL2,FL3,and FL4begin with a normal state,while other light curves begin
with aflaring state.The mean photon count rates of thefirst500s of FL2,FL3,and FL4are∼35,∼39,and∼40counts/c/PCU,respectively,similar to tho of the normal state. This offers an opportunity to examine the light curve transition from a normal states to aflaring state.In fact,the magnified views around the beginnings of theflares in FL2,FL3,and FL4(Figure 4)identify the existence of the ratherflat precursors of theflares lasting for a few minutes just before abrupt increas in photon count rates.
In contrast with the complicated differences in theflare light curves(Figure3),the pul profiles of FL1–10show a remarkable similarity(Figure5),all having simple,sinusoidal profiles. Figure6compares the puld fraction with the peak intensity of the10pul profiles in Figure5. The puld fraction is proportional to the peak intensity of the pul profile with a linear correlation coefficient of∼0.76.
3.1.2.Rapid Aperiodic Variability
In order to examine the possible existence of any rapid variability associated with theflares, we prent45-s light curves of FL1–10with0.0625-s resolution around the brightest peak in each light curve(Figure7).The light curves are modulated by∼13.5-s pulsations of the neutron star of LMC X-4,and exhibit strong narrow peaks of which intensities are significantly greater than tho expected fr
om the Poisson noi of the given light curve.The Leahy-normalized power-density spectra(PDSs)of the light curves of the central10-s gments reveal the existence of significant powers up to a few hertz,especially for the brightflares such as FL2–5.(Figure8;We only ud the central10-s gments in order to avoid the contamination by the source’s pulsations in PDSs. Note that the Leahy power owing to a Poisson noi is2[Leahy et al.1983]).)We investigated this possible correlation between the existence of the significant powers in the PDSs and theflare intensity more as follows.First,we divided the FL1–10light curves(Figure3)into gments of which lengths are equal to the pulsational period(∼13.5s)of LMC X-4determined by Fourier transformations of the light curves.Next,we computed the Leahy-normalized PDSs of the all
∼13.5-s gments,as well as their integrated intensities.The correlation between theflare intensity and the existence of the significant powers in PDSs is evident in Figure9,where we compare the distributions of the integrated intensities with tho of the significant powers(>10)(in PDSs)of the∼13.5-s gments.This confirms that the rapid aperiodic variability during theflares increas along with theflare intensity,which sometimes appears as strong narrow peaks in the light curves (Figure7).
3.2.Spectral Evolution of the Flares
3.2.1.Softness Ratio Distribution
We investigated the spectral evolution of theflares via the softness ratio,which we defined to be the ratio of the soft X-ray(2–8keV)photon count rates to tho of the hard X-ray(10–20 keV).Figure10compares the total photon count rates(integrated over the2–25keV range)of the flares with the softness ratios obtained with32-s resolution,identifying a strong linear correlation between them with a linear correlation coefficient of∼0.95.
However,Figure11a,which is the same as Figure10but obtained with0.0625-s resolution, shows that the strong correlation identified with32-s resolution disappears when time resolution improves–the linear correlation coefficient decreas to∼0.30.The data points with large photon count rates(so with small Poisson noi)show more significant deviations from the linear correlation (identified with32-s resolution)than tho with small photon count rates,indicating that the deviation is not simply owing to the incread Poisson noi caud by the improved time resolution. We investigated this further via comparison of the correlations obtained from two parate groups of the data:one is the“faint group”having small total photon count rates;the other is the“bright group”having large total photon count rates.We cho the faint group as the group of data points of which total photon count rates belong to the lower90%;the bright group of which total photon count rates belong to the upper1%.Th
e dotted and dashed lines in Figure11a reprent the correlations obtained in the faint and bright groups,respectively.The correlations obtained in the two groups are quite distinctive:while the faint group(dotted line)shows a positive correlation with a slope similar to that obtained with32-s resolution(Figure10),the bright group(dashed line)does not show any apparent correlation,indicating that the strong linear correlation between theflare intensities and the softness ratios obtained with32-s resolution does not hold for this ,the narrow,spiky peaks in Figure7).英语辅导机构加盟
In order to examine the effect of the incread Poisson noi caud by the improved time resolution in Figure11a more thoroughly,we ud a Monte Carlo simulation by re-sampling the each data point obtained with32-s resolution(Figure10)with0.0625-s resolution.We re-sampled the32-s resolution data points using a Gaussian distribution with the1σstandard deviation equal to the Poisson noi of0.0625-s resolution.We performed the simulation1000times,and prent a typical result in Figure11b.Although the correlation coefficients of the three lines reprenting
