Stellar Wind Variations During the X-ray High and Low States of Cygnus X-1
D. R. Gies, C. T. Bolton, R. M. Blake, S. M. Caballero-Nieves, D. M. Crenshaw, P. Hadrava, A. Herrero, T. C. Hillwig, S. B. Howell, W. Huang, L. Kaper, P. Koubsky, M. V. McSwain
aa r X i v : . [ a s t r o - ph ] J a n ApJ, in press
Stellar Wind Variations During the X-rayHigh and Low States of Cygnus X-1 , D. R. Gies , , C. T. Bolton , R. M. Blake , S. M. Caballero-Nieves , D. M. Crenshaw ,P. Hadrava , A. Herrero , T. C. Hillwig , S. B. Howell , W. Huang , , L. Kaper ,P. Koubsk´y , and M. V. McSwain , Based on observations with the NASA/ESA Hubble Space Telescope obtained at the Space TelescopeScience Institute, which is operated by the Association of Universities for Research in Astronomy, Incorpo-rated, under NASA contract NAS5-26555. These observations are associated with programs GO-9646 andGO-9840. Based on data obtained at the David Dunlap Observatory, University of Toronto. Center for High Angular Resolution Astronomy, Department of Physics and Astronomy, Georgia StateUniversity, P. O. Box 4106, Atlanta, GA 30302-4106; [email protected], [email protected], [email protected] Visiting Astronomer, Kitt Peak National Observatory, National Optical Astronomy Observatory, oper-ated by the Association of Universities for Research in Astronomy, Inc., under contract with the NationalScience Foundation. David Dunlap Observatory, University of Toronto, P. O. Box 360, Richmond Hill, Ontario, L4C 4Y6,Canada; [email protected] Astronomical Institute, Academy of Sciences of the Czech Republic, Friˇcova 298, CZ-251 65 Ondˇrejov,Czech Republic; [email protected], [email protected] Instituto de Astrof´ısica de Canarias, 38200, La Laguna, Tenerife, Spain; Departamento de Astrof´ısica,Universidad de La Laguna, Avda. Astrof´ısico Francisco S´anchez, s/n, 38071 La Laguna, Spain; [email protected] Department of Physics and Astronomy, Valparaiso University, Valparaiso, IN 46383;[email protected] WIYN Observatory and National Optical Astronomy Observatory, P. O. Box 26732, 950 N. Cherry Ave.,Tucson, AZ 85719; [email protected] Department of Astronomy, California Institute of Technology, MC 105-24, Pasadena, CA 91125; [email protected] Astronomical Institute Anton Pannekoek, Universiteit van Amsterdam, Kruislaan 403, 1098-SJ Amster-dam, The Netherlands; [email protected] Department of Physics, Lehigh University, 16 Memorial Drive East, Bethlehem PA 18015; [email protected] Guest investigator, Dominion Astrophysical Observatory, Herzberg Institute of Astrophysics, National
ABSTRACT
We present results from
Hubble Space Telescope ultraviolet spectroscopy ofthe massive X-ray and black hole binary system, HD 226868 = Cyg X-1. Thespectra were obtained at both orbital conjunction phases in two separate runs in2002 and 2003 when the system was in the X-ray high/soft state. The UV stellarwind lines suffer large reductions in absorption strength when the black hole is inthe foreground due to the X-ray ionization of the wind ions. A comparison of the
HST spectra with archival, low resolution spectra from the
International Ultra-violet Explorer Satellite shows that similar photoionization effects occur in boththe X-ray high/soft and low/hard states. We constructed model UV wind lineprofiles assuming that X-ray ionization occurs everywhere in the wind except thezone where the supergiant blocks the X-ray flux. The good match between theobserved and model profiles indicates that the wind ionization extends to nearto the hemisphere of the supergiant facing the X-ray source. We also presentcontemporaneous spectroscopy of the H α emission that forms in the high den-sity gas at the base of the supergiant’s wind and the He II λ α emissionstrength is generally lower in the high/soft state compared to the low/hard state,but the He II λ Subject headings: binaries: spectroscopic — stars: early-type — stars: individual(HD 226868, Cyg X-1) — stars: winds, outflows — X-rays: binaries
Research Council of Canada
1. Introduction
The massive X-ray binary Cygnus X-1 is a seminal target in the study of gas dynamics inthe vicinity of a stellar mass black hole. Its X-ray luminosity and energetic jets (Gallo et al.2005) are powered by gas accretion from the nearby companion star HD 226868 (O9.7 Iab;Walborn 1973) in a spectroscopic binary with a 5.6 day orbital period. There are severalways in which mass transfer from the supergiant to the black hole may occur in this sys-tem (Kaper 1998). The O-supergiant, like other massive and luminous stars, has a strongradiatively driven wind that may be partially accreted through the gravitational force ofthe black hole. The supergiant is large and is probably close to filling its critical Rochesurface (Gies & Bolton 1986a; Herrero et al. 1995), so a gas stream through the inner L1point may also be present. The actual gas flow in the direction of the black hole is probablyintermediate between a spherically symmetric wind and a Roche lobe overflow stream, andthere is evidence that the flow is best described as a focused wind (Friend & Castor 1982;Gies & Bolton 1986b; Gies et al. 2003; Miller et al. 2005). The gas ions responsible for ac-celerating the wind may become ionized in the presence of a strong X-ray source, leading to alower velocity, “stalled” wind (Blondin et al. 1990; Stevens 1991). In situations of very highX-ray flux, photoionization may extend so close to the supergiant’s photosphere that thewind never reaches the stellar escape velocity and thus ceases to become an X-ray accretionsource (Day & Stevens 1993; Blondin 1994). However, such a high X-ray flux may heat theouter gas layers to temperatures where the thermal velocities exceed the escape velocity tocreate a thermal wind that may fuel black hole accretion.Important clues about the mass transfer process come from the temporal variations ofthe observed X-ray flux. Cyg X-1 is generally observed in either a low flux/hard spectrumstate, with an X-ray spectrum that is relatively flat, or a high flux/soft spectrum state witha steeper power-law spectrum (Shaposhnikov & Titarchuk 2006). The gamma-ray portionof the spectrum is also elevated during the high/soft state (McConnell et al. 2002). Thehigh/soft state usually lasts for periods of days to months, and the fraction of time observed inthe high/soft state has increased from 10% in 1996–2000 to 34% since early 2000 (Wilms et al.2006). This increase may be related to an overall increase in the supergiant’s radius in theperiod from 1997 to 2003 – 2004 that is suggested by changes in the long term opticallight curve (Karitskaya et al. 2006a). The system sometimes experiences so-called failed-state transitions, when it starts to increase in flux, but then stops at an intermediate stateand returns to the low/hard state. All these transitions probably reflect changes in theinner truncation radius of the accretion disk surrounding the black hole that are causedby a variable accretion rate (largest when the system is in the high/soft state; Done 2002;McClintock & Remillard 2006). Thus, the temporal variations in the X-ray state offer us themeans to compare the black hole accretion processes with observational signatures related 4 –to mass transfer.The H α emission formed in the high density gas at the base of the stellar wind is animportant diagnostic of the mass loss rate in massive stars (Puls et al. 1996; Markova et al.2005). The H α emission variations in HD 226868 over the last few years are documented in in-dependent spectroscopic investigations by Gies et al. (2003) and Tarasov, Brocksopp, & Lyuty(2003). Both of these studies concluded that the H α emission appears strongest when thesystem is in the low/hard X-ray state, while a range of weak to moderate emission strengthsare observed during the high/soft states. This is a surprising result, since taken at face value,strong emission is associated with a large wind mass loss rate, and the simplest expectationthat the X-ray accretion flux increases with mass loss rate is, in fact, not observed. There areseveral possible explanations: (1) A denser wind may be more opaque to X-rays. However,this seems unlikely because the observed inverse relation between H α emission strength andX-ray flux is observed at all orbital phases, not just when the supergiant and its wind are inthe foreground. (2) The X-ray source may photoionize and heat the gas responsible for theH α emission, so that a larger X-ray flux leads to a decrease in H α strength. This clearly oc-curs at some level, but both Gies et al. (2003) and Tarasov et al. (2003) argue that portionsof the wind shaded from the ionizing flux also display significant temporal variations. (3)Changes in the X-ray flux will lead to variations in the ionized volume of gas surrounding theblack hole, and consequently, the total acceleration of the wind in the direction towards theblack hole will vary with the distance traveled before the atoms responsible for line-drivingare ionized. Thus, a stronger, denser wind might reach a faster speed before ionization, andsince the Bondi-Hoyle accretion rate varies as ∼ v − , the gas captured by the black hole(and the associated X-ray flux) declines.This last process can be tested through direct study of the degree of wind ionizationobserved in the ultraviolet P Cygni lines formed in the supersonic part of the wind outflow.When the system is observed with the ionization region in the foreground, the absorptioncores of these P Cygni lines will be truncated at a blueshift corresponding to the high-est projected speed before encountering the ionization zone, the so-called Hatchett-McCrayeffect (Hatchett & McCray 1977). The binary is so faint in the ultraviolet that high disper-sion spectra were very difficult to obtain with the International Ultraviolet Explorer (IUE) satellite (Davis & Hartmann 1983), but a good series of observations were obtained with
IUE at lower spectral resolution that clearly indicate the weakening of the wind lines whenthe black hole is in front (Treves et al. 1980; van Loon, Kaper, & Hammerschlag-Hensberge2001). Most of these spectra were obtained in the low/hard X-ray state ( § Hubble Space Telescope
Space Telescope Imaging Spectrograph(STIS). We obtained observations at the two orbital conjunction phases in both 2002 and2003. These sets of observations were both made during the rare high/soft X-ray state, andplanned observations during the low/hard state were unfortunately scuttled by the STISelectronics failure in 2004. However, we can rebin the high quality
HST spectra made duringthe high/soft state to the lower resolution of the
IUE archival spectra (mostly low/hardstate) in order to test whether or not the wind ionization state does in fact differ significantlybetween states. We describe a program of supporting optical spectroscopy we have obtainedto check the orbital phase ( §
2) and wind strength ( §
3) at the times of the
HST observations.We compare the H α measurements with the contemporaneous X-ray light curve recordedwith the Rossi X-ray Timing Explorer
All-Sky Monitor instrument (Levine et al. 1996) andconfirm our earlier result showing how H α tends to strengthen as the X-ray flux declines( § § HST and
IUE spectra in §
5, and then reassess the question of the mass transfer processin §
6. We will discuss the photospheric features in the UV spectra in a forthcoming paper(Caballero Nieves et al., in preparation).
2. Observations and Orbital Ephemeris
The
HST
STIS spectra were obtained with the first-order G140M grating in a seriesof subexposures at different grating tilts in order to record the UV spectrum over the fullavailable range (1150 to 1740 ˚A). Two full sets were made near each orbital conjunctionphase in runs in both 2002 and 2003. All the spectra were reduced using the IDL softwaredeveloped at NASA Goddard Space Flight Center for the STIS Instrument Definition Team.The spectra were rebinned on a log λ , heliocentric wavelength scale to a spectral resolutionof R = 10000 and rectified to a pseudo-continuum based upon the flux in relatively line-freeregions.The space observations were supported by contemporaneous, ground-based observa-tions of the red spectrum in the vicinity of H α . These 125 spectra were made with theUniversity of Toronto David Dunlap Observatory 1.88 m telescope, NOAO Kitt Peak Na-tional Observatory 0.9 m coud´e feed telescope, Herzberg Institute of Astrophysics DominionAstrophysical Observatory 1.85 m telescope, and the Academy of Sciences of the CzechRepublic Astronomical Institute Ondˇrejov Observatory 2 m telescope. We also obtained asmaller set of 22 blue spectra that record the variations in the He II λ R = λ/ △ λ , where △ λ = FWHM of the line spread function), num-ber of spectra obtained, and details about the telescope, spectrograph, and detector. Thespectra were reduced and transformed to rectified flux on a uniform heliocentric wavelengthscale (as described in Gies et al. 2003).Most of the red spectra also record the He I λ HST observations. The radial velocities were measured in the same wayas outlined by Gies et al. (2003) by fitting a Gaussian to the central line core. For thesake of completeness, we also measured the radial velocity of the supergiant from the
HST
UV spectra using a cross-correlation method (Penny, Gies, & Bagnuolo 1999) with an
IUE spectrum of HD 34078 as the reference template. The results are presented in Table 2 (givenin full in the electronic version) that lists the heliocentric Julian date of mid-observation,orbital phase, radial velocity and its associated error, observed minus calculated velocityresidual, H α equivalent width, and the corresponding run number from Table 1. We notethat independent measurements of the He I λ HST ) using thenon-linear, least-squares fitting method of Morbey & Brosterhus (1974). We made a circularorbital fit with the period fixed at the value obtained by Brocksopp et al. (1999) from dataspanning a 26 yr interval. Our results are compared to those from Brocksopp et al. (1999)and from Gies et al. (2003) in Table 3, and they are consistent with these earlier studies. Thecurrent epoch for the time of supergiant inferior conjunction T (IC) occurs 0 . ± .
012 dayslater than the prediction from Brocksopp et al. (1999), and we will adopt this revised epochfor the definition of orbital phase throughout the paper. We omitted the
HST measurementsfrom the orbital solution because of concerns about possible systematic differences in thevelocities derived from the UV lines and He I λ O − C ) residuals for the HST measurements given in Table 2 show that the UV measurements are in reasonable agreementwith the velocity curve derived from He I λ
3. Wind and X-ray States During the HST Observations
The optical red spectra were obtained with the primary goal of monitoring the gas den-sity at the base of the stellar wind of the supergiant. The H α observations of HD 226868 are 7 –summarized in two panels in Figure 1 according to the X-ray state at the time of observation(see Fig. 2 below). The top portions show plots of the profiles arranged by orbital phasewhile the lower grayscale images show the spectral flux interpolated in radial velocity andorbital phase. The white line in the lower image shows the radial velocity curve of the super-giant (Table 3). These figures show that most of the emission/absorption complex appearsto follow the orbit of the supergiant as expected for an origin in the supergiant wind. Wemeasured the H α equivalent width in the same way as before (Gies et al. 2003) by makinga numerical integration over a 40 ˚A range centered on H α , and these measurements arelisted in column 6 of Table 2. We estimate that the typical measurement error is ± . α equivalent width for a total of 240 measurementsfrom Gies et al. (2003) and the new observations in the top panel of Figure 2. The lower panelof this figure shows the daily average soft X-ray flux over the same interval from the All-SkyMonitor on the Rossi X-ray Timing Explorer (Levine et al. 1996). These flux measurementsare the quick-look results provided by the
RXTE /ASM team . The two arrows in the toppanel indicate the times of the two HST observing runs, which took place when Cyg X-1 wasin the high/soft state. The new measurements confirm the trends described by Gies et al.(2003) and Tarasov et al. (2003) that the H α emission tends to be stronger when the softX-ray flux declines and that there is a considerable range in emission strength when the softX-ray flux is large (see Fig. 1).Figures 3 and 4 show a detailed view of the time evolution of the H α emission strengthand X-ray flux for the week surrounding the HST runs in 2002 and 2003. The bottom panelsin these figures show the X-ray fluxes in both the low energy (1.5 – 3 keV; + signs) and higherenergy (5 – 12 keV; × signs) bands for the individual, 90 s exposure, dwell measurements.Unfortunately, both the H α and X-ray measurements are not exactly coincident in time withthe HST observations, so we have made a time interpolation between the closest availablemeasurements to estimate the H α emission and X-ray flux levels at the times of the HST observations (summarized in Table 4). All four
HST observations occurred when the H α emission was weak and the soft X-ray flux was uniformly strong.The other important optical emission line in the spectrum of HD 226868 is the He II λ II λ http://xte.mit.edu/ HST observation sets, and these observations show profile variations with orbital phasethat are quite similar to those seen previously (Gies & Bolton 1986b; Ninkov et al. 1987;Karitskaya et al. 2006b). We present in Table 5 the observed equivalent width of the He II λ II λ . II λ α emission line.
4. Orbital Variations in the UV Wind Lines
Our primary interest here is how the X-ray flux ionizes portions of the supergiant’swind and how our line of sight through the ionized zones changes with orbital phase. Weshow in Figures 5, 6, and 7 the changes observed between conjunctions in the major UVwind lines of N V λλ , IV λλ , IV λλ , φ = 0 . solid line ) and in the foreground of the supergiant ( φ = 0 . dotted line ), and the bottom panel shows the same for the 2003 run. The spectra are plottedas a function of radial velocity for the shorter wavelength component in the frame of thesupergiant (according to the orbital solution in Table 3). All three of these transitionsdisplay a large reduction in the extent and depth of the blueshifted absorption componentwhen the black hole is in the foreground (the Hatchett-McCray effect). These changes reflectthe X-ray photoionization and resulting superionization of these ions in the wind gas seenprojected against the supergiant. We also find some evidence of the associated reduction inthe strength of the red emission component due to the loss of these gas ions in the outflowaway from our line of sight when the black hole is in the background. The variations between 9 –conjunctions appear to be almost identical for the observations in 2002 and 2003, which isprobably due to the very similar X-ray fluxes that existed at those times of observation ( § φ = 0 . i = 90 ◦ and the orbital phases are φ = 0 . , . i = 33 ◦ − ◦ ; Gies & Bolton1986a; Brocksopp, Fender, & Pooley 2002), our line of sight at the conjunctions will includesomewhat different portions of the occulted and unocculted wind (Fig. 8), so these modelsare first approximations of the predicted variations for a shadow wind. In a companion paper(Vrtilek et al. 2008), we present a more complete calculation based upon a realistic orbitalinclination and the method outlined by Boroson et al. (1999).The line synthesis is based upon a set of adopted parameters and two fitting parameters.Most of the adopted parameters come from the study of Cyg X-1 and similar X-ray binariesby van Loon et al. (2001), and in particular, we assume a wind velocity law exponent of γ = 1 (eq. 2 in van Loon et al. 2001), a semimajor axis equivalent to 2 . R ⋆ (where R ⋆ is theradius of the supergiant; Gies & Bolton 1986a), and a characteristic turbulent velocity in thewind equal to 0 . v ∞ (Groenewegen, Lamers, & Pauldrach 1989) where v ∞ is the terminalvelocity in the undisturbed wind. The photospheric components corresponding to the windtransitions were assumed to be Gaussian in shape with parameters set by Gaussian fits of thephotospheric profiles in the non-LTE, line blanketed model spectra of Lanz & Hubeny (2003)(for T eff = 30000 K, log g = 3 . V sin i = 100 km s − , a linear limb darkening coefficient of ǫ = 0 .
50, and a spectral resolving power of 10000).The final two parameters, the wind terminal velocity v ∞ and the integrated optical 10 –depth through the shadow wind of the blue component of the transition τ , were fit by trialand error in order to match the entire set of the three wind lines at each conjunction. Notethat the value of v ∞ is set mainly by fits of the profiles at orbital phase φ = 0 . v ∞ = 1200 km s − and τ = 6, 4, and 10 for the N V , Si IV , C IV features,respectively. The shadow wind model profiles are presented in Figures 9, 10, and 11 for thesethree wind lines. The top panel in these figures shows the model ( diamonds ) and averageobserved spectra ( solid line ) for φ = 0 . φ = 0 . i = 40 ◦ ), the fastest moving part of the shadow wind that isprojected against the supergiant at φ = 0 . ≈ . R ⋆ where the wind has not yet reached terminal velocity. Thus, ourfit value of v ∞ = 1200 km s − is probably well below the actual value (which is probablycloser to 1600 km s − ).The low orbital inclination and the subsequent limited projection of the shadow windregion against the disk of the photosphere results in P Cygni absorption troughs that areunusually weak for the spectra of supergiants like HD 226868. We show in Figures 12 and 13montages of the Si IV and C IV wind features in four other O9.7 Iab supergiants as seen inhigh dispersion spectra from the IUE archive. The mean spectrum of HD 226868 at orbitalphase φ = 0 .
5. Comparison of the Wind Lines in the X-ray Low and High States
Our original goal was to obtain another set of STIS spectra when Cyg X-1 was in thelow/hard X-ray state in order to determine how the wind ionization conditions change with 11 –X-ray state. With the loss of STIS, such a comparison is not possible at present. However,the low dispersion FUV spectra of HD 226868 made with
IUE were in most cases madewhen Cyg X-1 was in the low/hard state. Thus, we can investigate differences in the windionization properties between X-ray states through a comparison of the
HST high/soft statespectra with the
IUE low/hard state spectra.We collected 30, low dispersion, short wavelength prime camera spectra of HD 226868from the
IUE archive, and transformed these to rectified flux versions on a uniform wave-length grid. Next, we smoothed, rectified, and rebinned the
HST
STIS spectra onto the same
IUE wavelength grid so that the line blended structures would appear the same as they doin the
IUE spectra. We then measured the effective absorption strength by determiningthe mean flux across a spectral range that extends over the full range of the apparent windfeature (as done by van Loon et al. 2001). This average flux will reflect both the changingP Cygni absorption and the other line blends (including interstellar components), but sincethe latter are generally constant in time, the average flux will serve to show the relativevariations in the wind absorption (low flux when the P Cygni trough is deep and high fluxwhen the trough weakens).The average flux measurements for both the
IUE and rebinned
HST spectra are givenin Table 7, which lists the heliocentric Julian date of mid-exposure, orbital phase (fromTable 3), the mean rectified flux across the Si IV and C IV wind lines, the telescope of origin,the X-ray state at the time of the observations, and a code for references discussing thecontemporary X-ray fluxes. Note that we did not measure the N V transition in the IUE spectra because these spectra are poorly exposed at the short wavelength end. The
IUE average flux measurements have a typical error of ±
7% based upon the scatter in the resultsfrom closely separated pairs of spectra.The average fluxes across the wind lines are plotted as a function of orbital phase inFigures 14 and 15 for Si IV and C IV , respectively (see a similar depiction in Fig. 1 fromvan Loon et al. 2001). Different symbols show these measurements for the different X-raystates as observed with IUE and
HST . We suspect that despite our efforts to rectify the
IUE and
HST spectra in the same way, there are probably still some systematic differences sincethe mean fluxes for the
HST spectra appear to be somewhat lower than those for the
IUE spectra. Nevertheless, the amplitude of line strength variation appears to be more or lessthe same in each of the
IUE low/hard state,
IUE high/soft state, and
HST high/soft statespectra. This result indicates that the Hatchett-McCray effect (and the amount of windphotoionization it represents) occurs at about the same level in both X-ray states. 12 –
6. Discussion
The X-ray accretion flux of Cyg X-1 is fueled by mass transfer from the supergiant. Weargue in this section that the mass transfer process is dominated by a wind focused along theaxis joining the stars. However, the accretion of this gas by the black hole may be influencedby the strength of the radiatively driven shadow wind that is directed away from the blackhole. We begin by reviewing the most pertinent observational results from this investigation,and then we consider the interplay between the dynamics of the wind outflow and the X-rayaccretion flux.First, the
HST
STIS spectra of HD 226868 that we obtained at two epochs when thesystem was in the high/soft state show dramatic variations in the wind line strength thatresult from a superionization of the gas atoms illuminated by the X-ray flux. Shadow windmodels, in which the wind ions only exist in the region where X-rays are blocked by thesupergiant, make a reasonably good match to the observed profile variations, so we suspectthat X-ray photoionization dominates much of the zone between the black hole and thefacing hemisphere of the supergiant. A similar degree of wind ionization probably also existsin the X-ray low/hard state since similar orbital variations in wind line strength are foundin
IUE low dispersion spectra made during the X-ray low/hard state.Second, the
HST spectra suggest that stellar wind gas emanating from parts of the pho-tosphere facing the X-ray source attains only a small velocity before becoming photoionized.For example, the highest optical depth wind feature, C IV λλ , φ = 0 . −
400 km s − (and it is possible that this small component results from aminor part of shadow wind projected against supergiant at φ = 0 .
5; see Fig. 8). This verylow wind speed is probably less than the stellar escape velocity ( ∼
700 km s − near thepoles).These results from the UV wind lines indicate that very little mass loss is occurringby a radiatively driven wind for surface regions that are exposed to the X-ray source. Thefact that the wind features appear similarly weak in IUE spectra obtained in the low/hardstate suggests that a spherical, radiatively driven wind from the hemisphere of the supergiantfacing the black hole is probably always weak or absent, and thus, accretion from a sphericallysymmetric wind must play a minor role in feeding the black hole in Cyg X-1.On the other hand, we found that the emission equivalent width of the He II λ II λ II λ II λ α P Cygni line forms mainly in the base of the stellarwind of the supergiant since we observe that H α follows the orbital velocity curve of thesupergiant (Fig. 1). The new observations are consistent with earlier results (Gies et al. 2003;Tarasov et al. 2003) in demonstrating that the H α emission strength is generally weaker inthe high/soft X-ray state. Photoionization and heating may extend down to atmosphericlevels where the gas densities are sufficient to create H α emission, so that the reduction inH α strength in the X-ray high/soft state may partially result from photoionization relatedprocesses. However, Gies et al. (2003) showed that H α emission variability was present inthose Doppler shifted parts of the profile corresponding to the X-ray shadow hemisphereof the supergiant (see their Fig. 15), so part of the H α variations must be related to gasdensity variations at the base of the stellar wind. Thus, the observed H α variations suggestthat the high/soft X-ray state occurs when the global, radiatively driven part of the wind isweaker. Long term, quasi-cyclic variations in wind strength are apparently common amonghot supergiants (Markova et al. 2005).Gies et al. (2003) and Tarasov et al. (2003) suggested that the variations in X-ray stateare caused by changes in wind velocity due to changes in supergiant mass loss rate. Duringtimes when the supergiant’s wind is denser and the mass loss rate is higher, the photoioniza-tion region would be more restricted to the region closer to the black hole. Consequently, aradiatively driven wind could accelerate to a higher speed before stalling when the gas entersthe ionization zone, and thus, the faster wind would result in a lower black hole accretionrate and X-ray luminosity (creating the low/hard X-ray state). Conversely, if the wind massloss rate drops, then the X-ray ionization zone will expand, the maximum wind velocitytowards the black hole will decline, and the net accretion rate will increase (perhaps creatingthe high/soft state; Ho & Arons 1987). This creates a positive feedback mechanism thatmay continue until the wind is ionized all the way down to the stellar photosphere facing 14 –the X-ray source (Day & Stevens 1993; Blondin 1994).If this scenario is correct, then we expect that the outflow velocities in the direction ofthe black hole (as measured in blue extent of the P Cygni lines at phase φ = 0 .
5) will be largerthan the supergiant escape velocity during the X-ray low/hard state. The superb quality
HST /STIS spectra indicate that outflow velocities are too low to launch the wind during theX-ray high/soft state. Moreover, the low resolution
IUE spectra from the low/hard stateappear to show a very similar pattern of the loss of the P Cygni absorption at φ = 0 . IUE results suggest that the spherical component of wind outflowtowards the black hole is weak and slow in both X-ray states, so a wind speed modulation isprobably not the explanation for the accretion variations associated with the X-ray states.We will require new, high quality, UV spectroscopy of Cyg X-1 during the low/hard state inorder to make a definitive test of this idea.The radiatively driven wind of the supergiant leads to effective mass loss only in theX-ray shadowed hemisphere and in the focused wind between the stars in Cyg X-1. Theoutflow in the shadow wind region will experience a Coriolis deflection, so that the trailingregions of the shadow wind will eventually enter the zone of X-ray illumination (Blondin1994). Once photoionized, this gas will stall with the loss of the important ions for radiativeacceleration, and some of this slower gas may extend around the orbital plane to the vicinityof the black hole. Although this deflected wind gas is probably not a major accretion source(Blondin 1994), it may affect the accretion dynamics of the focused wind. For example,when the shadow wind mass loss rate is high (times of strong H α emission), the resultingstalled wind component will create a higher ambient gas density on the leading side of thezone surrounding the black hole. The focused wind flow will make a trajectory towards thefollowing side of the black hole, and while gas passing closer to the black hole will mergeinto an accretion disk, gas further out will tend to move past the black hole before turninginto the outskirts of the disk. The presence of the stalled gas on the leading side may deflectaway this outer, lower density part of the flow and effectively inhibit gas accretion from thefocused wind. The subsequent reduction in gas accretion by the black hole may correspondto the conditions required to produce the low/hard X-ray state, while conversely a reductionin the stalled gas from the shadow wind may promote mass accretion and produce thehigh/soft state (Brocksopp et al. 1999; Done 2002; McClintock & Remillard 2006). Clearly,new hydrodynamical simulations are needed to test whether the stalled wind component issufficient to alter the accretion of gas from the focused wind and create the environmentsneeded for the X-ray transitions.We thank the staffs of the David Dunlap Observatory, Kitt Peak National Observa- 15 –tory, Dominion Astrophysical Observatory, Ondˇrejov Observatory, and the Space TelescopeScience Institute (STScI) for their support in obtaining these observations. The KPNOspectra supporting the second HST run were obtained with the assistance of participantsin the NOAO Teacher Leaders in Research Based Science Education program, includingJoan Kadaras (Westford Academy, Westford, MA), Steve Harness (Kingsburg Joint UnionHigh School, Kingsburg, CA), Elba Sepulveda (CROEM, Mayaguez, PR), and Dwight Taylor(Goldenview Middle School, Anchorage, AK). We also thank Saku Vrtilek and Bram Borosonfor helpful comments, and we are especially grateful to an anonymous referee whose reportwas pivotal to our discussion of the results. Support for
HST proposal number GO-9840was provided by NASA through a grant from the Space Telescope Science Institute, which isoperated by the Association of Universities for Research in Astronomy, Incorporated, underNASA contract NAS5-26555. The X-ray results were provided by the ASM/RXTE teams atMIT and at the RXTE SOF and GOF at NASA’s GSFC. The
IUE data presented in thispaper were obtained from the Multimission Archive at the Space Telescope Science Institute(MAST). Support for MAST for non-HST data is provided by the NASA Office of Space Sci-ence via grant NAG5-7584 and by other grants and contracts. Bolton’s research is partiallysupported by a Natural Sciences and Engineering Research Council of Canada (NSERC)Discovery Grant. Hadrava’s research is funded under grant projects GA ˇCR 202/06/0041and LC06014. Herrero thanks the Spanish MEC for support under project AYA 2007-67456-C02-01. This work was also supported by the National Science Foundation under grantsAST-0205297, AST-0506573, and AST-0606861. Institutional support has been providedfrom the GSU College of Arts and Sciences and from the Research Program Enhancementfund of the Board of Regents of the University System of Georgia, administered through theGSU Office of the Vice President for Research. We are grateful for all this support.
REFERENCES
Blondin, J. M. 1994, ApJ, 435, 756Blondin, J. M., Kallman, T. R., Fryxell, B. A., & Taam, R. E. 1990, ApJ, 356, 591Boroson, B., Kallman, T., McCray, R., Vrtilek, S. D., & Raymond, J. 1999, ApJ, 519, 191Brocksopp, C., Fender, R. P., Larionov, V., Lyuty, V. M., Tarasov, A. E., Pooley, G. G.,Paciesas, W. S., & Roche, P. 1999, MNRAS, 309, 1063Brocksopp, C., Fender, R. P., & Pooley, G. G. 2002, MNRAS, 336, 699Davis, R., & Hartmann, L. 1983, ApJ, 270, 671 16 –Day, C. S. R., & Stevens, I. R. 1993, ApJ, 403, 322Done, C. 2002, Roy. Soc. London Phil. Tr. A, 360, Issue 1798, 1967Friend, D. B., & Castor, J. I. 1982, ApJ, 261, 293Gallo, E., Fender, R., Kaiser, C., Russell, D., Morganti, R., Oosterloo, T., & Heinz, S. 2005,Nature, 436, 819Gies, D. R., & Bolton, C. T. 1986a, ApJ, 304, 371Gies, D. R., & Bolton, C. T. 1986b, ApJ, 304, 389Gies, D. R., et al. 2003, ApJ, 583, 424Groenewegen, M. A. T., Lamers, H. J. G. L. M., & Pauldrach, A. W. A. 1989, A&A, 221,78Hadrava, P. 2007, Proceedings of RAGtime 8/9: Workshops on black holes and neutronstars, ed. S. Hledik & Z. Stuchlik (Opava, Czech Republic: Silesian Univ.), 73(arXiv:0710.0758v1)Hatchett, S., & McCray, R. 1977, ApJ, 211, 552Herrero, A., Kudritzki, R. P., Gabler, R., Vilchez, J. M., & Gabler, A. 1995, A&A, 297, 556Ho, C., & Arons, J. 1987, ApJ, 316, 283Kaper, L. 1998, in Boulder-Munich II: Properties of Hot, Luminous Stars (ASP Conf. Series,Vol. 131), ed. I. D. Howarth (San Francisco: ASP), 427Karitskaya, E. A., et al. 2006a, IBVS, 5678, 1Karitskaya, E. A., et al. 2006b, Astron. & Ap. Trans., 24, 383Kemp, J. C., Barbour, M. S., & McBirney, R. E. 1981, ApJ, 244, L73Lamers, H. J. G. L. M., Cerruti-Sola, M., & Perinotto, M. 1987, ApJ, 314, 726Lanz, T., & Hubeny, I. 2003, ApJS, 146, 417Levine, A. M., Bradt, H., Cui, W., Jernigan, J. G., Morgan, E. H., Remillard, R., Shirey, R.E., & Smith, D. A. 1996, ApJ, 469, L33Ling, J. C., Mahoney, W. A., Wheaton, W. A., Jacobson, A. S., & Kaluzienski, L. 1983,ApJ, 275, 307 17 –Markova, N., Puls, J., Scuderi, S., & Markov, H. 2005, A&A, 440, 1133McClintock, J. E., & Remillard, R. A. 2006, in Compact Stellar X-ray Sources, ed. W. H.G. Lewin & M. van der Klis (Cambridge, UK: Cambridge Univ. Press), 157McConnell, M. L., et al. 2002, ApJ, 572, 984Miller, J. M., Wojdowski, P., Schulz, N. S., Marshall, H. L., Fabian, A. C., Remillard, R.A., Wijnands, R., & Lewin, W. H. G. 2005, ApJ, 620, 398Morbey, C. L., & Brosterhus, E. B. 1974, PASP, 86, 455Ninkov, Z., Walker, G. A. H., & Yang, S. 1987, ApJ, 321, 438Oda, M. 1980, IAU Circ., 3502, 2Ogawara, Y., Mitsuda, K., Masai, K., Vallerga, J. V., Cominsky, L. R., Grunsfeld, J. M.,Kruper, J. S., & Ricker, G. R. 1982, Nature, 295, 675Penny, L. R., Gies, D. R., & Bagnuolo, W. G., Jr. 1999, ApJ, 518, 450Perotti, F., et al. 1986, ApJ, 300, 297Priedhorsky, W. C., Terrell, J., & Holt, S. S. 1983, ApJ, 270, 233Puls, J., et al. 1996, A&A, 305, 171Shaposhnikov, N., & Titarchuk, L. 2006, ApJ, 643, 1098Stevens, I. R. 1991, ApJ, 379, 310Tarasov, A. E., Brocksopp, C., & Lyuty, V. M. 2003, A&A, 402, 237Treves, A., et al. 1980, ApJ, 242, 1114van Loon, J. Th., Kaper, L., & Hammerschlag-Hensberge, G. 2001, A&A, 375, 498Vrtilek, S. D., Boroson, B. S., Hunacek, A., Gies, D., & Bolton, C. T. 2008, ApJ, in pressWalborn, N. R. 1973, ApJ, 179, L123Wen, L., Cui, W., Levine, A. M., & Bradt, H. V. 1999, ApJ, 525, 968Wilms, J., Nowak, M. A., Pottschmidt, K., Pooley, G. G., & Fritz, S. 2006, A&A, 447, 245
This preprint was prepared with the AAS L A TEX macros v5.2.
Table 1. Journal of Spectroscopy
Run Dates Range Resolving Power Observatory/Telescope/Number (HJD-2,450,000) (˚A) ( λ/ △ λ ) N Spec., Grating/Detector1 . . . . . . . 2419.8 – 2828.7 6510 – 6710 8900 70 DDO/1.88m/Cass., 1800 g mm − /Thomson 1024 × ×
19 –Table 2. Radial Velocity and H α Equivalent Width Measurements
HJD Orbital V r △ V r ( O − C ) W λ (H α ) Run(-2,450,000) Phase (km s − ) (km s − ) (km s − ) (˚A) Number2419.847 . . . 0.115 45.4 3.1 2.2 − − − − − − − − − − − − − − − − − − − − − − − − · · · − · · · − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − · · · · · · − − − − − − − − − − − −
20 –Table 2—Continued
HJD Orbital V r △ V r ( O − C ) W λ (H α ) Run(-2,450,000) Phase (km s − ) (km s − ) (km s − ) (˚A) Number2453.934 . . . 0.202 59.9 1.3 − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − · · · · · · · · · − − · · · − · · · − − − − − − − −
21 –Table 2—Continued
HJD Orbital V r △ V r ( O − C ) W λ (H α ) Run(-2,450,000) Phase (km s − ) (km s − ) (km s − ) (˚A) Number2826.860 . . . 0.798 − − − − − − − − − · · · − − · · · − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − · · · · · · · · · − − − − − − − − − − − − − − − − − − − − − − − − − − − − − −
22 –Table 3. Circular Orbital ElementsElement Brocksopp et al. (1999) Gies et al. (2003) This Work P (d) . . . . . . . . . . . . . . . . . 5.599829 (16) 5.599829 a a T (IC) (HJD-2,400,000) 41,874.707 (9) 51,730.449 (8) 52,872.788 (9) K (km s − ) . . . . . . . . . . 74.9 (6) 75.6 (7) 73.0 (7) V (km s − ) . . . . . . . . . . . · · · − − σ (km s − ) . . . . . . . . . . . . · · · a Fixed.Note. — Numbers in parentheses give the error in the last digit quoted.Table 4. H α Equivalent Widths and X-ray Flux Counts for the
HST
Observation TimesHJD Orbital W λ (H α ) Flux (1.5 – 3 keV)(-2,450,000) Phase (˚A) (ASM counts)2450.3 . . . . . . 0.55 − − − − II λ W λ △ W λ Run X-ray(-2,450,000) Phase (˚A) (˚A) Number State52448.748 . . 0.276 − .
39 0.03 8 high/soft52448.905 . . 0.304 − .
35 0.02 8 high/soft52449.733 . . 0.452 − .
21 0.03 8 high/soft52449.885 . . 0.479 − .
22 0.02 8 high/soft52449.969 . . 0.494 − .
24 0.03 8 high/soft52450.713 . . 0.627 − .
53 0.03 8 high/soft52450.855 . . 0.653 − .
50 0.02 8 high/soft52450.958 . . 0.671 − .
49 0.02 8 high/soft52451.732 . . 0.809 − .
56 0.07 8 high/soft52452.796 . . 0.999 − .
47 0.08 8 high/soft52452.930 . . 0.023 − .
40 0.02 8 high/soft52453.709 . . 0.162 − .
49 0.05 8 high/soft52453.889 . . 0.194 − .
47 0.03 8 high/soft52453.965 . . 0.208 − .
44 0.03 8 high/soft52824.884 . . 0.445 − .
38 0.06 9 high/soft52824.891 . . 0.447 − .
23 0.06 9 high/soft52826.841 . . 0.795 − .
25 0.04 9 high/soft52912.769 . . 0.140 − .
46 0.02 10 low/hard52912.773 . . 0.140 − .
49 0.02 10 low/hard52913.768 . . 0.318 − .
46 0.02 10 low/hard52914.766 . . 0.496 − .
23 0.02 10 low/hard52915.761 . . 0.674 − .
31 0.02 10 low/hard 24 –Table 6. He II λ W λ σ ( W λ )(BY) (˚A) (˚A) Source1971–1981 . . . . . . . . − .
36 0.12 Gies & Bolton (1986a)1980–1984 . . . . . . . . − .
29 0.09 Ninkov et al. (1987)1996–1998 . . . . . . . . − .
25 0.12 Brocksopp et al. (1999)2002–2003 . . . . . . . . − .
39 0.12 This paper, high/soft state2003 . . . . . . . . . . . . . − .
39 0.11 This paper, low/hard state 25 –Table 7. Mean Flux Across the Wind LinesHJD Orbital Mean Flux Mean Flux(-2,400,000) Phase (Si IV λ IV λ IV λ IV λ O R B I T A L P H ASE -1000 -500 0 500 1000VELOCITY (km s -1 )0.0 0.5 1.0 0.0 0.5 1.0 O R B I T A L P H ASE -1000 -500 0 500 1000VELOCITY (km s -1 )0.0 0.5 1.0 Fig. 1.— Two depictions of the H α profiles as a function of heliocentric radial velocity andorbital phase and grouped according to X-ray state at the time of observation. The X-rayhigh/soft group was selected from observations made between HJD 2,452,100 – 2,452,550and between HJD 2,452,770 – 2,452,880, and all other times were assigned to the X-raylow/hard group (see Fig. 2). The upper panel in each shows the profiles with their continuaaligned to the orbital phase of observation, while the lower panel is a grayscale version ofthe profiles with the first and last 20% of the orbit repeated to improve the sense of phasecontinuity. The grayscale intensities represent the rectified spectral flux between 0.92 (black)and 1.12 (white). The white line in the grayscale image shows the orbital velocity curve ofthe supergiant. 28 – W λ ( H α ) ( A ng s t r o m s ) AS M R A T E ( . - k e V ) Fig. 2.— The long term variations in the H α emission strength ( above ) and daily averagesoft X-ray flux ( below ). The two arrows in the upper panel indicate the times of the STISobservations that occurred during the X-ray high/soft state. The recent, more denselysampled observations show clearly how the H α emission increases as the soft X-ray fluxdeclines. 29 – W λ ( H α ) ( A ng s t r o m s ) AS M R A T E Fig. 3.— The variations in H α emission strength ( above ) and X-ray flux for each dwellobservation ( below ) for the week surrounding the first HST run in 2002. The arrows in thetop panel show the times of the STIS observations for each orbital conjunction phase. Thesymbols in the lower panel indicate the count rates in the 1.5 – 3 keV (+) and 5 – 12 keV( × ) bands. 30 – W λ ( H α ) ( A ng s t r o m s ) AS M R A T E Fig. 4.— The variations in H α emission strength ( above ) and X-ray flux for each dwellobservation ( below ) for the week surrounding the second HST run in 2003 (in the sameformat as Fig. 3). 31 – I N T E N S I T Y φ = 0.08 φ = 0.55-3000 -2000 -1000 0 1000 2000 3000RADIAL VELOCITY (km s -1 )0.00.20.40.60.81.01.2 I N T E N S I T Y φ = 0.95 φ = 0.60 Fig. 5.— The spectral variations observed between conjunctions for the N V λλ , top ) and 2003 ( bottom ) runs. The spectra are plotted asa function of Doppler shift for the blue component of the doublet in the rest frame of thesupergiant (which leads to small offsets in the positions of the narrow, interstellar lines thatare stationary in the absolute frame). Vertical line segments indicate the rest wavelengthpositions of both components of the doublet. 32 – I N T E N S I T Y φ = 0.08 φ = 0.55-3000 -2000 -1000 0 1000 2000 3000RADIAL VELOCITY (km s -1 )0.00.20.40.60.81.01.2 I N T E N S I T Y φ = 0.95 φ = 0.60 Fig. 6.— The spectral variations observed between conjunctions for the Si IV λλ , I N T E N S I T Y φ = 0.08 φ = 0.55-3000 -2000 -1000 0 1000 2000 3000RADIAL VELOCITY (km s -1 )0.00.20.40.60.81.01.2 I N T E N S I T Y φ = 0.95 φ = 0.60 Fig. 7.— The spectral variations observed between conjunctions for the C IV λλ , CM XSHADOW φ =0.0 φ =0.5 Fig. 8.— A cartoon view of the binary from within the orbital plane. The normal windions are assumed to exist only within the shadow wind zone to the left of the supergiant(bounded by a thick, dashed line) where the X-ray flux from the vicinity of the black hole(marked by “X”) is fully blocked, while the gas is ionized to higher levels in the rest of thewind. The simple model predictions are based upon viewing the system along the line ofcenters (shown by arrows), while in fact we observe the supergiant at a lower inclinationangle ( i = 40 ◦ assumed). The dotted lines show the part of the line of sight that is projectedonto the disk of the star at the two orbital phases for an observer at the correct directionwith respect to the orbital plane. The shadow wind region projected against the star in eachcase is the area bounded by the surface of the star and the dashed and dotted line for thatphase. CM marks the center of mass position. 35 – I N T E N S I T Y φ = 0.0-3000 -2000 -1000 0 1000 2000 3000RADIAL VELOCITY (km s -1 )0.00.20.40.60.81.01.21.4 I N T E N S I T Y φ = 0.5 Fig. 9.— The average observed spectrum ( solid line ) and the shadow wind model spectrum( connected diamonds ) for the N V λλ , α wings that depress thecontinuum in the observed spectrum towards the left hand side of the diagram. 36 – I N T E N S I T Y φ = 0.0-3000 -2000 -1000 0 1000 2000 3000RADIAL VELOCITY (km s -1 )0.00.20.40.60.81.01.21.4 I N T E N S I T Y φ = 0.5 Fig. 10.— The average observed spectrum ( solid line ) and the shadow wind model spectrum( connected diamonds ) for the Si IV λλ , I N T E N S I T Y φ = 0.0-3000 -2000 -1000 0 1000 2000 3000RADIAL VELOCITY (km s -1 )0.00.20.40.60.81.01.21.4 I N T E N S I T Y φ = 0.5 Fig. 11.— The average observed spectrum ( solid line ) and the shadow wind model spectrum( connected diamonds ) for the C IV λλ , -5000 0 5000RADIAL VELOCITY (km s -1 )0123456 I N T E N S I T Y Fig. 12.— The mean Si IV wind profile for orbital phase φ = 0 . -5000 0 5000RADIAL VELOCITY (km s -1 )0123456 I N T E N S I T Y Fig. 13.— The mean C IV wind profile of HD 226868 for orbital phase φ = 0 . top, thickline ) compared to those of other O9.7 Iab supergiants (in the same format as Fig. 12). 40 – -0.2 0.0 0.2 0.4 0.6 0.8 1.0 1.2Phase0.700.800.901.001.10 < F l u x > IUE Low State IUE High State HST High State Si IV λ Fig. 14.— The mean flux across the Si IV wind profile plotted as a function of orbitalphase. The different symbols represent measurements of IUE low/hard state ( diamonds ), IUE high/soft state ( plus signs ), and
HST high/soft state spectra ( squares ). The Hatchett-McCray effect is seen as low flux (deep absorption) when the black hole is in the backgroundand high flux (little absorption) when the black hole is in the foreground. The