Looking through the photoionisation wake: Vela X-1 at \varphi_\mathrm{orb} \approx 0.75 with Chandra/HETG
R. Amato, V. Grinberg, N. Hell, S. Bianchi, C. Pinto, D. D'Aí, M. Del Santo, T. Mineo, A. Santangelo
AAstronomy & Astrophysics manuscript no. main © ESO 2021February 25, 2021
Looking through the photoionisation wake: Vela X-1 at ϕ orb ≈ . with Chandra /HETG
R. Amato , , , V. Grinberg , N. Hell , S. Bianchi , C. Pinto , A. D’Aì , M. Del Santo ,T. Mineo , and A. Santangelo Dipartimento di Fisica e Chimica - Emilio Segrè, Università degli Studi di Palermo, via Archirafi, 36, 90123 Palermo, Italy INAF-IASF Palermo, via Ugo la Malfa, 153, 90146 Palermo, Italy Institute for Astronomy and Astrophysics (IAAT), University of Tübingen, Sand 1, 72076 Tübingen, Germany Lawrence Livermore National Laboratory, 7000 East Ave, Livermore, CA 94550, USA Dipartimento di Matematica e Fisica, Università degli Studi Roma Tre, Largo S. Leonardo Murialdo, 1, 00146 Roma, Italy– / – ABSTRACT
Context.
The Supergiant X-ray binary Vela X-1 represents one of the best astrophysical sources to investigate the wind environmentof a O / B star irradiated by an accreting neutron star. Previous studies and hydrodynamic simulations of the system revealed a clumpyenvironment and the presence of two wakes: an accretion wake surrounding the compact object and a photoionisation wake trailing italong the orbit.
Aims.
Our goal is to conduct, for the first time, high-resolution spectroscopy on
Chandra / HETGS data at the orbital phase ϕ orb ≈ . Methods.
We perform a blind search employing a Bayesian Block algorithm to find discrete spectral features and identify them thanksto the most recent laboratory results or through atomic databases. Plasma properties are inferred both with empirical techniques andwith photoionisation models within CLOUDY and SPEX.
Results.
We detect and identify five narrow radiative recombination continua (Mg xi - xii , Ne ix - x , O viii ) and several emission linesfrom Fe, S, Si, Mg, Ne, Al, and Na, including four He-like triplets (S xv , Si xiii , Mg xi , and Ne ix ). Photoionisation models wellreproduce the overall spectrum, except for the near-neutral fluorescence lines of Fe, S, and Si. Conclusions.
We conclude that the plasma is mainly photoionised, but more than one component is most likely present, consistentwith a multi-phase plasma scenario, where denser and colder clumps of matter are embedded in the hot, photoionised wind of thecompanion star. Simulations with the future X-ray satellites
Athena and
XRISM show that a few hundred seconds of exposure will besu ffi cient to disentangle the lines of the Fe K α doublet and the He-like Fe xxv , improving, in general, the determination of the plasmaparameters. Key words.
X-rays: binaries – stars: massive – stars:winds, outflows
1. Introduction
The eclipsing high-mass X-ray binary (HMXB) Vela X-1(4U 0900-40) consists of a ∼
283 s period pulsar (McClintocket al. 1976) and a blue supergiant companion star (HD 77851,a B0.5Ia class star, Hiltner et al. 1972). With an X-ray luminos-ity of ∼ × erg s − and a distance of 2 kpc from Earth(Giménez-García et al. 2016), it is one of the brightest HMXBsin the sky. It is a high inclination system ( > ◦ , Joss & Rappa-port 1984), with an orbital period of ∼ ∼ R (cid:12) (Quaintrell et al. 2003). The donor star has a radius of about30 R (cid:12) (Quaintrell et al. 2003), so that the pulsar is constantlyembedded in the wind environment of the companion. The ge-ometry of the accreting stream of matter onto the compact objectis complex, being made up of an accretion wake, a photoion-isation wake, and possibly a tidal stream, as both simulations(e.g., Blondin et al. 1990; Manousakis 2011) and observationsin di ff erent wavebands show (e.g., Kaper et al. 1994; van Loonet al. 2001; Malacaria et al. 2016). A sketch of the binary systemwith the main features is given in Fig. 1. The line of sight inter-sects the di ff erent elements at di ff erent orbital phases, so that the observational data show strong changes in absorption along thewhole orbital period (Doroshenko et al. 2013).X-ray emission from Vela X-1 has already been detected andstudied for several di ff erent orbital phases with di ff erent instru-ments (e.g., Haberl & White 1990; Goldstein et al. 2004; Watan-abe et al. 2006; Fürst et al. 2010; Grinberg et al. 2017). Highresolution X-ray studies of the system are of special interest, asthey allow to draw conclusions on the properties of the com-plex plasma. High-resolution data from the High-Energy Trans-mission Grating Spectrometer (HETGS) (Canizares et al. 2005)of the Chandra X-ray Observatory (Weisskopf et al. 2000) ofVela X-1 during eclipse ( ϕ orb ≈
0) were studied by Schulz et al.(2002), who discovered and identified a variety of emission fea-tures, including radiative recombination continua (RRCs) andfluorescent lines, that led to the idea of the coexistence of ahot optically thin photoionised plasma and a colder opticallythick one. Goldstein et al. (2004) investigated
Chandra / HETGSdata of the system at three di ff erent orbital phases ( ϕ orb ≈ ϕ orb ≈ . ϕ orb ≈ . ϕ orb ≈ .
25, but thenthey appear again at ϕ orb ≈ .
5, when the soft X-ray contin-
Article number, page 1 of 13 a r X i v : . [ a s t r o - ph . H E ] F e b & A proofs: manuscript no. main uum diminishes. The simultaneous presence of H- and He-likeemission lines and fluorescent lines of near-neutral ions can beexplained by contributions from di ff erent regions: the warm pho-toionised wind of the companion star and smaller cooler regions,or clumps, of gas. Watanabe et al. (2006) compared the same Chandra / HETGS data sets to 3D Monte Carlo simulations ofX-ray photons propagating through a smooth, undisturbed wind.Based on this assumption, they deducted that highly ionised ions,which give rise to the emission lines, are located mainly in theregion between the neutron star (NS) and the companion star,while the fluorescent lines are produced in the extended stellarwind, from reflection of the stellar photosphere, and in the ac-cretion wake. More recent results on the same orbital phase byOdaka et al. (2013) with
Suzaku and by Martínez-Núñez et al.(2014) with
XMM-Newton , respectively, highlighted flux vari-ability and strong changes in absorption over periods of the or-der of ks. The same variability is found in
Chandra / HETGS dataat ϕ orb ≈ .
25 from Grinberg et al. (2017), who attributed thechanges in the overall absorption necessarily to the clumpy na-ture of the winds from the companion. Moreover, the high energyresolution of
Chandra allowed the detection of line emission fea-tures from several ionised elements, corroborating the idea of aco-existence of cool and hot gas phases in the system.Hydrodynamic simulations (Manousakis & Walter 2015; ElMellah et al. 2018, 2019) suggest the presence of a more com-plex structure around the neutron star (NS), with a bow shockand eventually the formation of a transient wind-captured ac-cretion disk (Liao et al. 2020). Such features can influence theway clumps accrete onto the compact objects, i.e., reducing theamount of transferred angular momentum or introducing timelags and phase mixing when the clumps are stored in such struc-tures.In this work we present, for the first time, a high-resolutionspectroscopic study of
Chandra / HETG archival data of Vela X-1at orbital phase ϕ orb ≈ .
75, i.e., when the line of sight is inter-secting the photoionisation wake (see Fig. 1). The study of theX-ray spectrum at this specific orbital phase, where the absorp-tion from the wind of the X-rays coming from the NS is high,allows the detection of a large number of lines from di ff erentelements in high ionisation states and, thus, the application ofplasma diagnostic techniques to characterise the accretion en-vironment. The paper is structured as follows: we first look forchanges in the hardness of the flux in Section 2, finding none;then we proceed with a blind search for spectroscopic absorp-tion / emission features, applying a Bayesian Block algorithm tothe unbinned spectrum; we present the identification of all thedetected features in Section 3, while in Section 4 we comparethe observational data with two di ff erent photoionisation codes;in Section 5 we discuss the plasma properties and the geometryof the wind of the companion star; in Section 6 we perform sim-ulations with future X-ray satellites; we present our conclusionsin Section 7.
2. Data reduction and temporal analysis
We analysed the High Energy Grating (HEG) and Medium En-ergy Grating (MEG) data sets of the
Chandra / HETG ObsID14654, taken on 2013-07-30, with ACIS-S, in FAINT mode, fora total exposure time of 45.88 ks. According to the ephemeris ofKreykenbohm et al. (2008), the data set covers the orbital phase ϕ orb = . − .
78, where ϕ orb = Chandra data analysis threads, but we chose P HOTOIONISATION WAKE A CCRETION
WAKE O BSERVER
Fig. 1.
A sketch of Vela X-1 from Grinberg et al. (2017) showing the ac-cretion and photoionisation wakes. The blue circle represents the donorstar HD 77851, while the pulsar is hidden in the accretion wake. Thegrey arrow indicates the verse of the rotation of the binary system. Atthe orbital phase ϕ orb ≈ .
75, the observer is looking at the system fromthe right, so that the line of sight (dashed line) is crossing the photoion-isation wake. . - k e V - k e V H a r d n e ss Fig. 2.
Light curves in units of counts s − , in the soft 0.5-3 keV ( toppanel ) and hard 3-10 keV ( middle panel ) energy bands, and correspond-ing hardness ratio ((3–10 keV) / (0.5–3 keV), bottom panel ). The bluehorizontal line indicates the mean value of the hardness ratio, with the1 σ uncertainty given by the blue area. Data are binned to the spin periodof 238 s, error bars at 1 σ . a narrower sky mask to avoid the overlapping of the extractionregion and to improve the flux at the shortest wavelengths.Following the work of Grinberg et al. (2017), who observeda change in the hardness of the source during phase ϕ orb ≈ . ff erent energy bands, 0.5–3keV (soft) and 3–10 keV (hard), and computed the hardness ra-tio, defined as the ratio between the counts in the hard and softbands. Fig. 2 shows the result, with data binned to the neutronstar spin period of 283 s (errors at 1 σ ). The hardness ratio val-ues at ϕ orb ≈ .
75 are higher than the ones obtained by Grinberget al. (2017) by at least a factor of ten, which is not surprisingconsidered the high absorption expected at this orbital phase.Moreover, the hardness ratio is almost flat for the whole obser-vation, in contrast to Grinberg et al. (2017), where a variabilityof a factor of three was observed. Hence, we extract only onespectrum, in the full energy range of 0.5–10 keV (Fig. 3).
3. High-resolution spectroscopy
We used the Interactive Spectral Interpretation System (ISIS)1.6.2-43 (Noble & Nowak 2008a,b) to perform the spectroscopic
Article number, page 2 of 13. Amato et al.: Looking through the photoionisation wake: Vela X-1 at ϕ orb ≈ .
75 with
Chandra / HETG Å )2 4 6 8 10Energy (keV)0.000.010.020.030.040.05 P h . c m s k e V Fig. 3.
Combined HEG and MEG spectrum of
Chandra
ObsID 14654in the energy range 0.5-10 keV. analysis of the data, with the ISIS functions (ISISscripts) pro-vided by ECAP / Remeis observatory and MIT , cross sectionsfrom Verner et al. (1996), and solar abundances from Wilmset al. (2000). We used Cash statistic (Cash 1979) with the spec-trum binned to the MEG resolution. All uncertainties are givenat 90% confidence level.We performed a blind search of spectral features, using aBayesian Block (BB) algorithm (Scargle et al. 2013), as de-scribed in Young et al. (2007) and as applied to Chandra / HETGSdata by Grinberg et al. (2017). To optimise the line detection al-gorithm, we divided the whole spectrum into five regions of in-terest, named after the most significant element detected in eachof them, as reported in Table 1. These spectral regions were anal-ysed one by one. We locally modelled the continuum with a sim-ple power law and then looked for significant deviations in theresiduals. Given the narrow wavelength ranges of individual re-gions, the continuum is always adequately well fitted by a powerlaw. Similar piece-wise approaches have been previously repeat-edly used for line searches and modelling (see, e.g., Yao et al.2008; van den Eijnden et al. 2019).The BB algorithm determines whether a data point is farfrom the model above a certain significant threshold, definedby a parameter, α , such that each detection has a significanceof p ∼ exp( − α ), corresponding to a probability of P ∼ − exp( − α ) of positive detection. For each new detection, weadded to the model one or more Gaussian components for emis-sion / absorption lines and the XSPEC (Arnaud 1996) functions redge and edge for the RRCs and the Fe K-edge (Sect. 3.1, 5.1),respectively. After each addition, we fit the data and apply the al-gorithm once more. We iterate the process until the significancedrops to 95%, corresponding to α ∼ α are listed in Tables 2–6,while Table 1 shows the best-fit values of the power law param-eters for each spectral region and the goodness of the fit. Table 7displays the best-fit values of the RRCs.In some cases, lines that are too close to be clearly resolvedby the algorithm, such as for example He-like triplets, are de-tected as a single block. In such cases, we use our knowledge ofatomic physics to add the proper number of lines to the model.Moreover, to improve the fit, we fixed the distance of known . lines, since the BB per se is a blind algorithm, i.e., it does nottake into account known line distances. We did this for the H-like Ly α and Ly β lines (Erickson 1977) and the He-like triplets(Drake 1988), assuming that Doppler shifts are the same withinthe same ionic species.Whenever a line appeared unresolved, we fixed its width to0.003 Å, corresponding to about one third of the MEG resolu-tion (0.023 Å FWHM) . Line identification for S and Si ionsaccounts for the most recent laboratory measurements from Hellet al. (2016), while for the other elements we use the AtomDBdatabase (Foster et al. 2012, 2017).For every detected He-triplet, we computed the density-sensitive ratio R = f / i and the temperature-sensitive ratio G = ( i + f ) / r , where f represents the intensity of the forbidden line(1s2s S –1s S ), i the intensity of the intercombination line(1s2p P –1s S ) and r the intensity of the resonant line (1s2p P –1s S ) (Gabriel & Jordan 1969; Porquet & Dubau 2000) .In our case, the intensities of the lines are linked to reproduce G and R as free parameters in the fit. Results are reported in Ta-ble 8.In the following subsections, we present in detail the resultsof the BB procedure for each spectral region of interest. In the Fe region (wavelength range 1.6-2.5 Å, cf. Table 1), theBB method found only one strong line, that we identified with FeK α and one edge, identified with the Fe K-edge. Best-fit valuesfor these features are reported in Table 2. Although the strongFe K α line implies the presence of a strong Fe K β component,our approach did not detect it. We discuss the possible reasonsin Sect. 5.Given the overall strength of the Fe K α line, we attemptedan additional fit, letting the line width free. We obtained a best-fit value of σ = (cid:16) . + . − . (cid:17) × − Å, consistent with our previousassumption and with results by Tzanavaris & Yaqoob (2018).
We studied the S region in the wavelength range 4.5-6.0 Å (Ta-ble 1). Line identification is based on the recent laboratory mea-surements from Hell et al. (2016). The BB algorithm detected asingle block between 5 Å and 5.4 Å, with α =
27. We modelthis block with the S xv He-like triplet, the S xiv , the S xi and theblended fluorescence S ii - viii lines. The second run of the algo-rithm resulted in the detection of the S xvi Ly α , with α = xiii He β ( α = ix ( α = .
7) and an unidentified absorption line at ∼ α = . α andthe lack of any other absorption feature in the whole spectrum,it is most likely that the line is just a statistical fluctuation.In the same region we could also expect to find the Si xiv Ly β line, at 5.217 Å (Erickson 1977). The lack of a significantdetection of this line is probably due to the strong continuum. http://cxc.harvard.edu/cdo/about_chandra Gabriel (1972) refers to the transitions of the lines of the He-liketriplets as w for the resonant line, x and y for the two components of theintercombination line, and z for the forbidden line. With this notation,the ratios for plasma diagnostic are expressed as R = z / ( x + y ) and G = ( z + ( x + y )) /w . Article number, page 3 of 13 & A proofs: manuscript no. main
Table 1.
Best-fit values of the power laws used to model the continuum and values of the Cash statistic per degrees of freedom (d.o.f.) for eachregion of the spectrum.
Region Wavelength range (Å) Γ Norm. (keV s − cm − ) Cash(d.o.f.)Fe 1.6–2.5 − . + . − . . + . − . − . ± . (cid:16) . + . − . (cid:17) × − − . ± . . ± . × − . + . − . (1 . ± . × − . ± . (cid:16) . + . − . (cid:17) × − Table 2.
Features detected in the Fe region (1.6–2.5 Å) with the detection parameter α and the best-fit values. The width of the Fe K α line wasfixed to 0.003Å. Line BB Ref. wavelength Det. wavelength Line flux τα (Å) (Å) (ph s − cm − × − )Fe K α
157 1.9375 a . ± . . ± . b . ± .
003 – 0 . ± . Notes. ( a ) Drake (1988). ( b ) Bearden & Burr (1967).
Energy (keV) P h . c m s Å -- F e K -- F e K - e dg e -- F e K -- -- F e XX V -- -- N i K Wavelength ( Å ) C Fig. 4.
Fe-region spectrum and best-fit model (red line), with residualsshown in the bottom panel. The only line detected by the BB algorithmis identified and marked as a FeK α emission line, as well as the detectedFe K-edge. Arrows mark the position of the expected Ni K α , Fe K β andHe-like Fe xxv lines (in grey). However, since the Si xiv Ly α line is strong in the Si region(see Sect. 3.3), the Si xiv Ly β is most likely present and blendedwith the S xi line. In Fig. 5, we marked the line at 5.224 Å withboth its possible identifications. Given this line confusion, theLy β / Ly α ratio for Si xiv cannot be easily constrained. Only anupper limit of 0.55 can be derived, assuming the minimum fluxfor Ly α (cf. Sec. 3.3) and that all flux of the discussed blendis due to Si xiv Ly β . Moreover, the high absorption constitutesa source of additional uncertainty as it influences the line ratio(Kaastra & Mewe 1995).For this region, all the line widths were fixed to 0.003 Å.Best-fit values are reported in Table 3, together with the Dopplervelocities computed with respect to laboratory reference values(Hell et al. 2016). Fig.5 shows the spectrum, the best-fit model Energy (keV) P h . c m s Å -- S X V I L y -- S X V r -- S X V i -- S X V f -- S X I / S i X I V L y -- S II - V III -- S i X III H e -- S I X -- S X I V -- ? Wavelength ( Å ) C Fig. 5.
S-region spectrum and best-fit model (red line), with residualsshown in the bottom panel. The detected lines are labelled if identified. and the residuals of the fit. From the S xv triplet, we obtained thebest fit ratios R = . + . − . and G = . + . − . (Table 8). We searched for Si lines in the region 6.0-7.4 Å (Table 1). TheBB algorithm highlighted at the first trial ( α = xiv Ly α line and a whole block in the range 6.6-6.8 Å that we mod-elled with the He-like triplet Si xiii , at first. The fluorescent lineblend Si ii - vi is detected with α = α =
32. We addedthree Gaussians to model this block, according to the labora-tory measurements by Hell et al. (2016) (see also Grinberg et al.2017), corresponding to the Si vii , Si viii and Si ix lines. The lastdetections are identified as the Al xiii Ly α line ( α = x and Si xi lines ( α =
5) and the Si xii line ( α = . Article number, page 4 of 13. Amato et al.: Looking through the photoionisation wake: Vela X-1 at ϕ orb ≈ .
75 with
Chandra / HETG
Table 3.
Spectral features detected in the S region. For each feature we report the detection parameter α , the best-fit values (wavelength and lineflux) and the Doppler velocities, computed using reference wavelengths measured by Hell et al. (2016). Line widths fixed to 0.003 Å for all thelines. Line BB Ref. wavelength Det. wavelength Line flux Velocity α (Å) (Å) (ph s − cm − × − ) (km s − )S xvi Ly α
12 4.7329 a . ± .
003 3 . + . − . − + − S xv r
27 5.0386 5 . + . − . . + . − . + − S xv i
27 5.0666 5 . + . − . b . ± . = v (S xv r) S xv f
27 5.1013 5 . + . − . b . + . − . = v (S xv r) S xiv
27 5.0858 5 . ± .
003 2 . ± . − + − S xi / Si xiv Ly β c
27 5.2250 5 . ± .
002 2 . + . − . − ± ix . + . − . . + . − . + − S ii - viii
27 5.3616 5 . ± .
003 2 . + . − . ± . + . − . − . + . − . –Si xiii He β d . ± .
003 1 . + . − . + − Notes.
Hell et al. (2016) reports the statistical uncertainties for each energy, which correspond to an error in wavelength of the order of 10 − −− − Å. However, authors state that there is also a systematic uncertainty of 0.23 eV for S lines, which results in an error on the wavelengthof 0.0008 Å. ( a ) Garcia & Mack (1965). ( b ) Distances between the r line and the i and f lines computed from Drake (1988). ( c ) The referencewavelength of Si xiv Ly β is 5.217 Å (Erickson 1977). ( d ) Kelly (1987).
In the same region, also the RRC of Mg xii is detected, at6.321 Å (1 . ± .
002 keV), with a temperature of 4 . + . − . eV.Lastly, we added one more redge function to model the Mg xi RRC, expected at 7.037 Å (Drake 1988). It results in a tempera-ture of 3 . + . − . eV, consistent with the one of Mg xii RRC (Table7). The width of the lines was fixed to 0.003 Å, except for theSi xiv Ly α line, which has a slightly larger width of (cid:16) . + . − . (cid:17) × − Å. For each line, we computed the Doppler velocities withrespect to the laboratory or literature reference wavelengths. Allthe best-fit values of the emission lines and RRCs are reported inTable 4 and Table 7, respectively, while the spectrum, the best-fitmodel and the residuals are shown in Fig. 6. The best fit valuesof the R and G ratios of the S xiii triplet resulted in R = . ± . G = . + . − . (Table 8). The BB algorithm did not detectthe Mg xii Ly β emission line expected at ∼ . ii - vi lines. The region we took into account to look for Mg emission linesranges from 7.5 Å to 10 Å (Table 1). The first line detected cor-responds to the Mg xii Ly α ( α = α =
89) consisted in a block in the range ∼ xi . Inthe same block, we insert the Ne x RRC (Schulz et al. 2002;Watanabe et al. 2006; Goldstein et al. 2004). We also detectedand identified the Mg xi He β ( α = x Ly γ ( α = xii r He α ( α = x He δ ( α = . xx ( α = . xxiv ( α = .
9) emission lines. Best-fitvalue are reported in Table 5, while the spectrum, the best-fitmodel and the residuals are shown in Fig. 7. A few lines show abroadening that required to let their widths free. This is the casefor Mg xii Ly α whose width of (7 . ± . × − Å is in agree-ment with those of Si xiv (Sect. 3.3) and Ne x Ly α (Sect. 3.5) P h . c m s Å -- S i X I V L y -- S i X III r -- S i X III i -- S i X III f -- S i II - V I -- S i V II -- S i I X -- S i V III -- A l X III L y -- S i X II -- S i X I -- S i X -- RR C M g X II -- RR C M g X I Å )10010 C Fig. 6.
Si-region spectrum and best-fit model (red line), with residualsshown in the bottom panel. lines. Other broadened lines are the Mg xi r and the Ne x He δ , ∼ xxiii line ( ∼ x RRC, at a wavelength of ∼ . + . − . keV) indicatesa temperature of 10 . + . − . eV (Table 7) consistent with previousfindings at di ff erent orbital phases (Schulz et al. 2002; Goldsteinet al. 2004). Doppler shifts of the Ly α , the He β and the tripletlines are around 150 km s − . From the intensities of the Mg xi triplet we obtained the ratios R = . + . − . and G = . + . − . (Table 8) for plasma diagnostic. The region for Ne emission lines goes from 10 Å to 14.5 Å (Ta-ble 1). We detected and identified 11 lines and two RRCs. Best-fit values are reported in Table 6 and 7, the spectrum, best-fit
Article number, page 5 of 13 & A proofs: manuscript no. main
Table 4.
Spectral features detected in the Si region. For each of them, we report the detection parameter α , the best-fit values (wavelength and lineflux). Line widths fixed to 0.003 Å, if not stated otherwise. Doppler velocities of the Si lines, computed with respect to the reference wavelengthsmeasured by Hell et al. (2016). Line BB Ref. wavelength Det. wavelength Line flux Velocity α (Å) (Å) (ph s − cm − × − ) (km s − )Si xiv Ly α
190 6.1817 a . ± .
001 6 . ± . b ± xiii r
190 6.6483 6 . ± . . + . − . ± xiii i
190 6.7195 6 . ± . c . + . − . = v (Si xiii r) Si xiii f
190 6.7405 6 . ± . c . ± . = v (Si xiii r) Si xii . ± .
003 0 . ± . ± xi . ± .
004 0 . + . − . ± x . ± .
004 0 . + . − . ± ix
32 6.9279 6 . ± .
003 1 . ± . ± viii
32 7.0008 7 . ± .
005 1 . ± . ± vii
32 7.0577 7 . + . − . . ± . − + − Si ii - vi d
121 7.1172 7 . ± .
001 2 . ± . − ± xiii Ly α e . ± .
003 0 . ± . + − Notes.
Hell et al. (2016) report a systematic uncertainty of 0.13 eV for Si lines, corresponding to an error on the wavelength of 0.0005 Å. ( a ) Garcia & Mack (1965). ( b ) This line results in a best line width of 7 . + . − . × − Å. ( c ) Distances between the r line and the i and f lines computedfrom Drake (1988). ( d ) The Mg Ly β (7.1037 Å, Erickson 1977) might be blended with the Si ii - vi line. ( e ) Erickson (1977).
Table 5.
Spectral features detected in the Mg region (7.5–10 Å). For each of them, we report the detection parameter α , the best-fit values(wavelength and line flux) and the Doppler velocities, computed with respect to reference wavelength from literature. Line widths fixed to 0.003 Å,if not stated otherwise. Line BB Ref. wavelength Det. wavelength Line flux Velocity α (Å) (Å) (ph s − cm − × − ) (km s − )Al xii He α a . + . − . . ± . b + − Mg xi He β
48 7.850 c . ± . . ± . ± xxiv d . + . − . . + . − . − + − Mg xii Ly α
220 8.42101 e . ± . . + . − . f ± xi r
89 9.16896 a . ± . . ± . g ± xi i
89 9.2312 a . ± . a . + . − . = v (Mg xi r) Mg xi f
89 9.3143 a . ± . a . ± . = v (Mg xi r) Fe xx h i . ± .
004 1 . ± . ± x Ly δ e . ± .
006 0 . ± . ± x Ly γ
15 9.708 e . ± .
005 1 . + . − . j ± Notes. ( a ) Drake (1988). ( b ) Line width of 0 . + . − . Å. ( c ) Kelly (1987). ( d ) Wargelin et al. (1998). ( e ) Erickson (1977). ( f ) For this line the best-fitline width value was (7 . ± . × − Å. ( g ) Line width of 0 . ± .
003 Å. ( h ) Close to the same wavelength there is also the Ne x Ly ζ emissionline at 9.291 Å, but with a lower intensity ratio. In this case the resulting Doppler velocity would be ( − ± − . ( i ) Unpublished atomicdata from Liedahl (1997). ( j ) Line width of 0 . + . − . Å. model and residuals are shown in Fig. 8. The first line to be de-tected by the BB procedure ( α =
49) was the Ne x Ly α , at awavelength of 12.1398 Å and with a width of (9 . + . − . ) × − Å.The successive detection ( α =
29) was a line at ∼ x Ly β . Hence, we fixed the distance ofthe latter line with respect to the corresponding Ly α according toErickson (1977). The next detection ( α =
17) was a block from13.4 Å to 13.9 Å, that we modelled with the Ne ix triplet (Grin-berg et al. 2017; Goldstein et al. 2004; Watanabe et al. 2006). Lastly, we detected six more lines, corresponding to Ne ix He β ,at 11.549 Å ( α = ix He γ at 11.005 Å ( α = ix He ε at 10.644 Å ( α = . xi Ly α at 10.023 Å ( α = . xix at 10.814 Å ( α = .
8) and Fe xxi at 12.285 Å ( α = . ix RRC at 10.374 Å was detected with α = . + . − . eV, while the O viii RRC at14.22 Å was detected with α = . . + . − . eV (Table 7). This is the first detection of the O viii RRCin
Chandra data for Vela X-1. It was implied in ASCA observa-
Article number, page 6 of 13. Amato et al.: Looking through the photoionisation wake: Vela X-1 at ϕ orb ≈ .
75 with
Chandra / HETG P h . c m s Å -- M g X II L y -- M g X I r -- M g X I i -- M g X I f -- RR C N e X -- M g X I H e -- N e X L y -- A l X II H e -- N e X L y -- F e XX I V -- F e XX Å )10010 C Fig. 7.
Mg-region spectrum and best-fit model (red line), with residualsshown in the bottom panel. P h . c m s Å -- N e X L y -- N e I X r -- N e I X i -- N e I X f -- N e X L y -- RR C N e I X -- N e I X H e -- N e I X H e -- N e I X H e -- N a X I L y -- F e X I X -- F e XX I -- RR C O V III
10 11 12 13 14Wavelength ( Å )10010 C Fig. 8.
Ne-region spectrum and best-fit model (red line), with residualsshown in the bottom panel. tions (Sako et al. 1999), suggested by Schulz et al. (2002), andonly recently detected using
XMM-Newton data (Lomaeva et al.2020). We note that the O viii
RRC might be also blended with aFe xviii line at 14.208 Å (Brown et al. 1998).We computed Doppler shifts for all the lines, obtaining ve-locities consistent with each other (Table 6). The intensities ofthe lines of the Ne ix triplet gave diagnostic best fit ratios of R = . + . − . and G = . + . − . (Table 8).
4. Photoionisation models with CLOUDY and SPEX
We attempted a more physical modelling of the detected fea-tures using photoionisation models with the latest release ofCLOUDY (Ferland et al. 2017; Chakraborty et al. 2020) andSPEX (v3.05, Kaastra et al. 1996, Kaastra et al. 2018). Inboth cases we used proto-Solar abundances from Lodders et al.(2009). Both these codes require an input ionising continuum.We approximated such a continuum with a sum of two com-ponents, as previously done in Grinberg et al. (2017) and Lo- maeva et al. (2020). The emission from the star, that dominatesin the UV, was modelled with a black body, while the emissionfrom the accretion onto the NS with a power law modified by aFermi-Dirac cuto ff . Both components have the same parametersas employed in Lomaeva et al. (2020). In particular, the shape ofthe power law continuum cannot be well constrained at energiesbelow 10 keV, especially when strongly a ff ected by absorption,as is the case with our observations. We thus used parameters de-rived from non-simultaneous NuSTAR observations (Fürst et al.2014). We note that there are some indirect hints that the illu-minating continuum assumed here may not reflect the true con-tinuum seen by the plasma in the system, such as, in particular,the large ratio between the Fe and Si / S fluorescence lines and thestability curves, which are unstable over wide ranges, especiallyat the ionisation parameters of interest. This emphasises the im-portance of strictly simultaneous observations at high resolutionbelow 10 keV and at energies above this range for the future.In our modelling, we left free to vary the electron density n e (cm − ), the ionisation parameter ξ (erg cm s − ), the absorp-tion coe ffi cient N H (10 cm − ), and the turbulent velocity v turb (km s − ). We explored the parameter space with CLOUDY in theranges 5 . ≤ log n e ≤ .
5, 0 . ≤ log ξ ≤ .
0, 20 . ≤ log N H ≤ .
3, and 80 km s − ≤ v turb ≤
160 km s − . For SPEX we assume amuch larger parameter space since its PION model calculates theionisation balance instantaneously and does not require a prede-fined grid of models.We modelled the data with an absorbed partially coveredpower law, with spectral index Γ =
1, corresponding to the inputpower law of our photoionisation models (Fürst et al. 2014), inaddition to the CLOUDY / SPEX photoionisation model. The ab-sorption due to the interstellar medium was fixed to 3 . × cm − (HI4PI Collaboration et al. 2016), while the local absorp-tion was left free to vary. The local partial absorption was appliedonly to the continuum, since both the geometry of the system,with localised wakes of material, and previous high resolutionstudies (Schulz et al. 2002; Watanabe et al. 2006; Grinberg et al.2017; Lomaeva et al. 2020) imply that the line producing regionis not experiencing the same high absorption as the vicinity ofthe neutron star, where the continuum is produced. We furtheradded three more Gaussians for the fluorescence Fe K α line,centred at 1.9388 Å (cfr. Sect. 4), and for the near-neutral flu-orescence emission lines of S ii - viii and Si ii - vi , which are notreproduced by CLOUDY and SPEX.The best fit CLOUDY model resulted in log n e = . + . − . , log ξ = . ± . N H = . ± .
020 cm − and a turbulent velocity of ∼
160 km s − . The modelrequired a redshift, with a best fit value of z ∼ − , corre-sponding to a velocity of v ∼
100 km s − , consistent with theDoppler shifts previously obtained. The Cash (d.o.f.) statisticvalue was 1.58 (2584). The modelling of the whole spectrumwith SPEX resulted in the best fit values of log ξ = . + . − . and N H = (4 . ± . × cm − , with a line broadening of160 ±
16 km s − and a Cash (d.o.f.) value of 1.57 (2382). Alsoin this case the model is redshifted with respect to the data, witha best fit velocity along the line of sight of 130 + − km s − . Bestfits are shown in Fig. 9. We also tried to add a second CLOUDYcomponent, obtaining no statistical significant improvement ofthe fit.We noticed that the electron density n e is degenerate with theabsorption of the interstellar medium (ISM): the larger the ISM N H , the larger the n e (see discussion in Sect. 5.2). Article number, page 7 of 13 & A proofs: manuscript no. main
Table 6.
Spectral features detected in the Mg region (10-14 Å). For each of them, we report the detection parameter α , the best-fit values (wave-length and line flux) and the Doppler velocities, computed with respect to reference wavelength from literature. Line widths fixed to 0.003 Å, ifnot stated otherwise. Line BB Ref. wavelength Det. wavelength Line flux Velocity α (Å) (Å) (ph s − cm − × − ) (km s − )Na xi Ly α a b . ± .
005 0 . + . − . ± x Ly β
29 10.23887 c . ± . d . + . − . ± ix He ε e b . ± .
006 0 . + . − . ± xix . b . + . − . . + . − . − + − Ne ix He γ f g . + . − . . + . − . + − Ne ix He β h g . + . − . . + . − . + − Ne x Ly α
49 12.132 c . ± . . + . − . i ± xxi b . + . − . . + . − . + − Ne ix r
17 13.4476 j . ± .
005 1 . + . − . ± ix i
17 13.553 j . ± .
005 2 . + . − . = v (Ne ix r) Ne ix f
17 13.699 j . ± .
005 2 . + . − . = v (Ne ix r) Notes. ( a ) Possible line blends include Fe xx at 10.024 Å and Ni xxiv at 10.027 Å. ( b ) Reference wavelength taken from AtomDB database ( ). ( c ) Erickson (1977). ( d ) Computed from the Ne x Ly α best fit wavelength, as from Erickson (1977). ( e ) Anotherpossible identification is the Fe xix at 10.648 Å. ( f ) Possible line blends are: Fe xx at 11.007 Å, Na x He ι at 11.003 Å and Fe xix at 11.002 Å. ( g ) Kelly (1987). ( h ) Another possible line is Fe xx at 11.546 Å. ( i ) Best-fit value of the line width of (cid:16) . + . − . (cid:17) × − Å. ( j ) Drake (1988).
Table 7.
Best-fit values of the RRCs in each region, with temperature reported in K and eV. The wavelengths in the last column are simply theconversion of the threshold energy from keV to Å and are meant just for convenience to the reader.
RRC Region Threshold energy (keV) Temperature (10 K) Temperature (eV) Wavelength (Å)Mg xii
Si 1 . ± .
002 5 . + . − . . + . − . xi Si 1 . ± .
001 3 . + . − . . + . − . x Mg 1 . + . − . . + . − . . + . − . ix Ne 1 . + . − . . + . − . . + . − . viii Ne 0 . ± . . + . − . . + . − . Table 8.
Best-fit values of the G and R ratios of the He-like triplets and correspondent electron temperatures and densities (Porquet & Dubau2000). The electron density of the He-like Si xiii triplet (marked as ∗ ) is an upper limit. Element
G R
Temperature (K) Temperature (eV) Electron density n e (cm − )S xv . . − . . + . − . – –Si xiii . + . − . . ± . ×
860 1 × * Mg xi . + . − . . + . − . ×
600 2 × Ne ix . + . − . . + . − . − × . ×
5. Discussion
We performed, for the first time, high-resolution spectroscopyanalysis of
Chandra / HETGS data of Vela X-1 at the orbitalphase φ orb ≈ .
75. A first look at the hardness ratio (Fig. 2) re-vealed no significant continuum spectral variability during theobservation. The mainly flat shape of the hardness ratio is notsurprising, since the line of sight at this orbital phase is expectedto lie well within the photoionisation wake, a denser stream-likeregion that trails the NS (Doroshenko et al. 2013; Malacaria et al.2016) and acts as a constant absorber (see Fig.1).The analysis pointed out the presence of Fe, S, Si, Mg andNe, as well as of less intense emission lines from Al and Na. Contrary to previous observations (Schulz et al. 2002; Goldsteinet al. 2004; Watanabe et al. 2006), there is no evidence of thepresence of Ar ( λ ∼ .
359 Å), Ca ( λ ∼ .
186 Å) and Ni ( λ ∼ .
660 Å) fluorescence lines. Upper limits of their fluxes resultedin 5 . × − ph s − cm − for Ar, 2 . × − ph s − cm − for Ca,and 3 . × − ph s − cm − for Ni, respectively.In the next subsections, we discuss in details the Fe region(Sect. 5.1), carry out plasma diagnostic (Sect. 5.2), and investi-gate the geometry of the wind of the companion star (Sect. 5.3). Article number, page 8 of 13. Amato et al.: Looking through the photoionisation wake: Vela X-1 at ϕ orb ≈ .
75 with
Chandra / HETG
Energy (keV) P h . c m s Å All componentsCLOUDYGaussiansPower law
Wavelength ( Å ) C . F l u x ( P ho t on s − c m − s − Å − ) All components Power law Gaussians Pion5 10 − ( O b s − M od ) / E rr Wavelength (Å)
Fig. 9.
Fit of the whole spectrum with the photoionisation model (dotted green line) from CLOUDY ( left panel ) and SPEX ( right panel ), plus apartially covered power law (dotted red line), and three Gaussians for the fluorescence lines of Fe K α , S ii - viii , and Si ii - vi (solid magenta lines).The total fit function is represented in black. Residuals of fit in the bottom panels. Spectra rebinned for clarity’s sake. The Fe region (1.6–2.5 Å) is dominated by a Fe K α line, centredat 1 . ± . x (Palmeri et al.2003), consistent with the results of Grinberg et al. (2017) (be-low Fe xii ), and di ff erent from the case of an irradiated wind, asshowed by the hydrodynamic simulations of Sander et al. (2018)(where the wind is mainly driven by Fe iii ions). However, theline may be redshifted so that a higher ionisation state could beexpected. A more refined calculation is beyond the goal of thispaper.The only other relevant feature detected in this region is theFe K-edge at 1 . ± .
003 Å (see Table 2), which is not signif-icantly Doppler shifted.The BB algorithm did not detect the Fe K β line, expected at ∼ β and FeK α lines is 0.13–0.14 (Palmeri et al. 2003, for the charge statesFe ii - ix ), we can estimate an expected flux of (1 . ± . × − ph cm − s − , which might not be su ffi cient to let the line emergefrom the continuum underneath. To verify this assertion, we gen-erated 1000 Monte Carlo simulated spectra adding to the best fitmodel a Gaussian at the correspondent wavelength of the Fe K β with the expected flux. We then run the BB algorithm on all thesimulated spectra (cf. Sect. 3.1). In no case the line was detected,confirming its weakness with respect to the X-ray continuumand the K-edge, which precluded a detection in the observationaldata. The Fe K β / K α ratio depends on the ionisation of iron (seethe detailed discussions in Molendi et al. 2003; Bianchi et al.2005). For higher charge states, the expected line ratio is evensmaller, i.e., the Fe K β line would be even weaker than what oursimulation showed as undetectable. Therefore, we cannot ruleout that the ionisation state is higher than what we assumed. Wediscuss the prospects of detecting Fe K β with future instrumentsin Sect. 6.Results from Goldstein et al. (2004) at φ orb ≈ φ orb ≈ . α line at λ ∼ .
660 Å, while Schulz et al. (2002)propose the presence of a Fe xxv emission line at λ ∼ .
85 Å( φ orb ≈ λ ∼ .
66 Å and λ ∼ .
86 Å, for the Ni Ly α and a Fe xxv respectively, and fitthe spectrum again. The Fe XXV is actually a He-like triplet, butthe resolution of the MEG of 0.023 Å FWHM, adopted consis-tently through the paper, is not good enough to resolve the linesindividually. Hence, we use just one Gaussian to fit the wholeion, letting the width free to vary. The width of the Ni Ly α linewas fixed to the usual value of 0.003 Å. The fluxes of these latterGaussians resulted in (cid:16) . + . − . (cid:17) × − ph cm − s − for the Ni Ly α and (3 . ± . × − ph cm − s − for the Fe xxv lines, while thewidth of the He-like Fe xxv had a best fit value of 0 . + . − . Å.From the Fe edge (Table 2), we computed the equivalent hy-drogen column as N H = τ edge / ( Z Fe σ Fe ), where Z Fe = . × − is the solar Fe abundance (Wilms et al. 2000) and σ Fe = . × − cm is the photoelectric absorption cross sec-tion for Fe xxv at the wavelength of the K-edge (Verner et al.1996). Using the best-fit value optical depth τ edge = . ± . N H = (3 . ± . × cm − , which is nearly con-sistent with the best-fit value of N H = (2 . ± . × cm − obtained fitting the spectrum in this region with a simple ab-sorbed power law, with solar abundances and cross sections asspecified in Sect. 3. These values are of the same order of mag-nitude as the best fit values found for observations using MAXI (Matsuoka et al. 2009) by Doroshenko et al. (2013) and
NuSTAR (Harrison et al. 2013) by Fürst et al. (2014) at the same orbitalperiod. However, we must bear in mind here that the model weused does not account for the Fe K β line, which may contributeto larger uncertainties on the Fe K-edge parameters. The presence of five narrow RRCs (Mg xi , Mg xii , Ne ix , Ne x ,and O viii ) suggests that the plasma is photoionised, with a tem-perature between ∼ G = . + . − . ofthe Ne ix triplet (Table 8), consistent with 4 in spite of the largeuncertainties (Porquet & Dubau 2000).However, the G ratios of S xv ( G = . + . − . ), Si xiii ( G = . + . − . ) and Mg xi ( G = . + . − . ) are all smaller than 1, in-dicating that collisional processes are not negligible and mayeven dominate (Porquet & Dubau 2000; Porquet et al. 2010).Under the hypothesis of a collisional equilibrium plasma (CIE), Article number, page 9 of 13 & A proofs: manuscript no. main we can estimate the temperature from the G ratio values (Por-quet & Dubau 2000). From the He-like Si xiii and Mg xi tripletswe obtain temperatures of ∼ × K and ∼ × K, respec-tively, which are two orders of magnitude higher than the onesfrom the Ne RRCs.This inconsistency between temperatures derived from theRRCs and the He-like line ratios is likely due the known is-sue that relative level populations between the upper levels ofthe He-like triplet lines can be shifted by other physical phe-nomena, which are likely present in HMXBs, thus making the G ratio unreliable. In particular, two processes can enhance a res-onant r line stronger than the intercombination i or forbidden f lines: photoexcitation and resonance line scatter. Photoexcitationcan be important in photoionisation equilibrium (PIE) plasma,when many photons with the right energy excite the electrons tothe resonant level. This clearly enhances the resonance line and,then, alters the G ratio with respect to the pure recombinationcase (see the comprehensive explanation in Kinkhabwala et al.2002). The presence of a few weak iron L emission lines (Fe xix - xxiv ) also seems to point in this direction (Sako et al. 2000).Resonant line scattering occurs when a photon is absorbedand re-emitted in the same wavelength, but in the direction ofthe lowest optical depth. This phenomenon is well explained byWojdowski et al. (2003) for the HMXB Centaurus X-3, observedduring eclipse. In the case of Vela X-1, though we are not inthe eclipsing phase, the dense streams of matter surrounding theNS can act as a strong absorber, enhancing the resonance linescattering into the line of sight.Concerning the R ratio, the values of Mg xi ( R = . + . − . ) andNe ix ( R = . + . − . ) He-like lines implies an electron density ofthe plasma of ∼ × cm − and ∼ . × cm − , respec-tively, considering a plasma temperature of 7 × K and 2 × K, as previously estimated . On the other hand, the R ratios ofSi xiii ( R = . ± .
6) and S xv ( R = . + . − . ) are much higher thanthe respective values at the low density limit, when the relativeintensities of the He-like lines are in fact independent of the elec-tron density of the plasma. In the case of Si, for instance, the lowdensity limit value is R =
3, corresponding to a maximum den-sity of the order of 10 cm − (Porquet & Dubau 2000), whichcan be addressed here as upper limit. On the other hand, the fitwith CLOUDY and SPEX photoionisation models highlightedthe degeneracy of the electron density n e with the model chosenfor the continuum, and, in particular, with the absorption fromthe ISM. The best fit value of n e = . × cm − , for instance,can be treated only as a lower limit. The analysis underlines thatthe estimation of the density is influenced in opposite directionsby the R ratio and the continuum and the real value is somewherein between those limits.Also the UV radiation of the companion star can alter theplasma (the so-called “UV-pumping” mechanism, Gabriel & Jor-dan 1969; Blumenthal et al. 1972; Mewe & Schrijver 1978; Por-quet et al. 2001). UV radiation mimics a high density plasma,favouring the population of the P levels against the S level,leading to an increase of the intensity of the intercombinationline, against the forbidden line and, hence, to smaller values ofthe R ratio. The influence of the UV emission is taken into ac-count in both, CLOUDY and SPEX based photoionisation mod-els, through our choice of the continuum. Such models shouldalso, if applicable to the given data at all, give better constrains We note here that the R ratio depends upon the relative ionic abun-dance of the H-like and He-like ions ( χ ion parameter), but in the range ofour interest the dependence is so small that we can neglect it (see Fig. 9of Porquet & Dubau 2000). orb v ( k m / s ) Si IX (S02)Si VIII (S02)Si VII (S02)Si VI (S02)Si IX (G04)Si VIII (G04)Si VII (G04)Si VIII (G17)Si VII (G17)Si VI (G17)Si IX (RA)Si VIII (RA)Si VII (RA)Si VI (RA)
Fig. 10.
Doppler velocities at di ff erent orbital phases of Si ix (circles),Si viii (squares), Si vii (diamonds) and Si vi (reverse triangles), fromSchulz et al. (2002) (S02, blue ) and Goldstein et al. (2004) (G04, or-ange ), as adjusted for laboratory reference values by Hell et al. (2016),from Grinberg et al. (2017) (G17, green ) and from the present work(RA, red ). The solid and dashed lines stand for the radial velocities ofthe NS and the giant star, respectively. on the underlying plasma parameters than the more empiricalconsideration of G and R ratios. The quality of our fits in Sect. 4imply that this is the case.Overall, both the self-consistent photoionisation codes pro-vided a satisfactory fit of the data (Fig. 9), implying that, at thisspecific orbital phase, the plasma is mainly photoionised. How-ever, a closer inspection at the residuals hints to the presence ofat least another phase of the plasma. The near-neutral emissionlines of S ii - viii and Si ii - vi , as well as the Fe K α line are notreproduce by the photoionisation models that are driven by thepresence of highly ionised lines. This naturally suggests that theplasma cannot be a single component plasma.In a possible scenario, colder and denser clumps of plasma,from either the wind or larger scale accretion structures such aswakes, can cross unevenly the line of sight, adding to the PIEemission of the wind of the companion star a further componentwith a lower ionisation. Our data do not allow to constrain theorigin of this component that could be, for example, a further,colder PIE component, a collisionally ionised component or amore complex mix with a temperature gradient as is the case,e.g., in Cyg X-1 (Hirsch et al. 2019). We also note that our resultsemphasise the necessity of an accurate treatment of intermediateand low ionisation ions in atomic codes used for high resolutionX-ray spectroscopy. Doppler velocities at di ff erent orbital phases can reveal the lo-cation and dynamics of the line emitting material. Fig. 10 showsthe velocities for the ions of Si vi - ix from Schulz et al. (2002)and Goldstein et al. (2004) at the orbital phases φ orb ≈ φ orb ≈ .
5, adjusted with respect to the laboratory measurementsof Hell et al. (2016), together with the ones from Grinberg et al.(2017) at the orbital phase φ orb ≈ .
25, and with those in thepresent work ( φ orb ≈ . φ orb ≈ .
25 are neg-ative (blueshift), while velocities at the other orbital phases arepositive (redshift) and / or consistent with no shift. The same be-haviour is observed also for all the others lines of S, Si, Mg andNe (Fig. 11), even though there are no recent laboratory mea-surements that allow us to validate the Doppler shifts found bythe previous studies (Schulz et al. 2002; Watanabe et al. 2006;Goldstein et al. 2004; Grinberg et al. 2017). Most of the veloc-ities are consistent with the radial velocity of the NS, as well Article number, page 10 of 13. Amato et al.: Looking through the photoionisation wake: Vela X-1 at ϕ orb ≈ .
75 with
Chandra / HETG v ( k m / s ) S Si orb v ( k m / s ) Mg orb Ne Fig. 11.
Doppler velocities at di ff erent orbital phases of Ly α lines andHe-like triplets of S, Si, Mg and Ne from Schulz et al. (2002) ( blue ) andGoldstein et al. (2004) ( orange ), from Grinberg et al. (2017) ( green ) andfrom the present work ( red ). Di ff erent symbols stand for di ff erent ioni-sation stages. The solid and dashed lines represent the radial velocitiesof the NS and of the companion star, respectively. as of the companion star (solid and dashed lines in Figs. 10-11),computed as: v rad = π a sin i [cos ( ϑ + ω ) + e cos ω ] / ( T √ − e ) (1)where a is the semi-major axis, i is the inclination, T is the orbitalperiod, e is the eccentricity, ϑ and ω are the true anomaly and theargument of periapsis, respectively.The overall behaviour is consistent with the material co-moving with the NS, though the lack of more observational datafor each orbital phase prevent us to assert it definitively. How-ever, this behaviour has already been observed for the blackhole HMXB Cygnus X-1 (Hirsch et al. 2019; Miškoviˇcová et al.2016), where the Doppler shifts show a clear modulation withthe orbital phase. It has already been suggested for Vela X-1 thatthe wind velocity at the distance of the NS is ∼
100 km s − andlower than typically estimated from prescribed simple β -laws(Sander et al. 2018). The large spread in the range of observedDoppler shifts within the same orbital phases may be due radia-tion coming from regions further downstream the wind or due toa more complex velocity structure in the accretion region.More considerations on the geometry of the emitting regioncan be drawn from the ionisation state of the plasma. The ioni-sation parameter can be expressed as in Tarter et al. (1969): ξ = L X n e r (2)where L X is the X-ray luminosity of the source, n e the particledensity of the plasma, and r the distance at which the lines areproduced. From the photoionisation models, we computed thedistributions of the relative abundances of all the ions as a func-tion of ξ . For the H-like ions, the ionisation parameter was in therange 3 . ≥ log( ξ ) ≥ .
2. Assuming that each ion is producedat the peak of its distribution, for a luminosity of 4 × ergs − and a best-fit value of log( n e ) ∼
8, we obtained a distancein the range (1 . − . × cm = − R (cid:12) . Consideringthat the orbital separation of the system is ∼ R (cid:12) and the com-panion star has a radius of about 30 R (cid:12) , the region where theH-like emission lines come from seems to be very close to thesurface of the companion star, rather then to the surface of theNS. The other ionisation stages have lower values of log( ξ ), im-plying even higher distances, compatible with the idea of a wake expanding after the passage of the NS. We note, however, thatthis estimate assumes a constant density, that is most likely notthe case for an expanding wind, even without taking into accountpossible clumping and wake structures.Our result is in agreement with simulations of X-ray photonsin a smooth wind from Watanabe et al. (2006), for H-like Si, inthe case of a mass loss rate ˙ M ≤ . × − M (cid:12) yr − , consistentwith the latest estimation for Vela X-1 of ∼ . × − M (cid:12) yr − (Sander et al. 2018). In the end, our simple calculation wouldsuggest that the photoionised plasma is produced at the orbitalseparation of the system, in a region close to the surface of thecompanion star.Nonetheless, given the uncertainty in the electron densitydriven by the continuum, we repeated the calculation using in-stead the n e derived from the R ratio ( n e ∼ cm − ). This n e value, however, does not take into account the presence of thestrong UV radiation from the stellar wind (Sec. 5.2) and thushas to be considered as an overestimate. For the same valuesof the ionisation parameter as before, the resulting distance is r (cid:46) . R (cid:12) , comparable with the Bondi-Hoyle-Littleton radius ofthe NS in Vela X-1 of ∼ cm (Manousakis & Walter 2015).The assumption of using the same ionisation parameters holdsbecause log( ξ ) is primarily driven by the ionisation state and thushardly changes with the electron density, which is instead drivenby the absolute line strength (i.e., distance and continuum) andthe triplet shape, if the lines are well resolved. In this case, ofcourse, the ionisation of the wind would be due almost entirelyto the gravitational pulling of the NS.From this analysis, we cannot infer the presence of clumps.
6. Future perspectives with
XRISM /Resolve and
Athena /X-IFU
High-resolution spectroscopy is a powerful tool to study X-rayemission from any kind of astrophysical plasma. Currently, lim-itations of X-ray satellites are due, for instance, to their intrinsicresolution and sensibility. New generation X-ray satellites willgo beyond these limits. The X-Ray Imaging and SpectroscopyMission (
XRISM , formerly
XARM , Tashiro et al. 2018) and theAdvanced Telescope for High Energy Astrophysics (
Athena ,Nandra et al. 2013) will host on-board microcalorimeters withan energy resolution down to a few eV, thus exceeding the reso-lution of Chandra gratings in the Fe K region.We performed simulations of this region (1.6–2.2 Å, cfr.Sect. 5.1), including the Fe K-edge and the Fe K α as detected inthe Chandra observation, and the Fe K β , the He-like Fe xxv andthe Ni K α with the upper limit on the flux as in Sect. 5.1. Bothmicrocalorimeters should be able to resolve the Fe K α doubletand the Fe xxv triplet. To assess this in more detail, the inputspectrum of our simulation included two Gaussians for the FeK α , at 1.9399 Å for the Fe K α and at 1.9357 Å for Fe K α ,respectively, with a 1:2 ratio (Kaastra & Mewe 1993), and fourGaussians for the Fe xxv , with line centroids as in Drake (1988)and a flux ratio of 2:1:1:2 ( w : x : y : z ). The width of all the lineswas fixed to 0.0007 Å ( ∼ XRISM will be provided with the soft X-ray spectrometerResolve, with a nominal energy resolution of 5–7 eV in the 0.3–12 keV bandpass. We used the ancillary and response files of
Hitomi / SXS (Kelley et al. 2016) for the energy resolution re-quirement of 7 eV. Simulations show that an exposure of only300 s (comparable with the pulse period of 293 s) is su ffi cientto clearly detect the Fe K β line with a significance of α = . Article number, page 11 of 13 & A proofs: manuscript no. main measured Fe K β/ K α ratio of 0 . + . − . . With an exposure of 2.5ks, the probability of a positive detection of the Fe K β line raisesup to > .
99% ( α = α doublet is resolved, whileamongst the lines of Fe xxv only the f line is clearly resolved. Athena will be equipped with the X-ray Integral Field Unit(X-IFU, Barret et al. 2018), a cryogenic X-ray spectrometerworking in the energy range 0.2–12 keV, with a nominal en-ergy resolution of 2.5 eV up to 7 keV. Moreover, thanks to thehigher collecting area of
Athena (1.4 m at 1 keV), high qual-ity spectra will be acquired in much shorter exposures. Alsofor the Athena / X-IFU, we performed a 300 s simulation of theFe region (Fig. 12). Running the BB algorithm on the simulatedspectrum, the K β line is detected with α =
9, corresponding to99.99% probability of positive detection. If the exposure times isincreased up to 2.5 ks, then the K β line is detected with a signif-icance of α =
69. The measured intensity ratio between the FeK β and Fe K α is 0 . + . − . . The Fe K α doublet is fully resolved,as well as the f line of Fe xxv . The i line, which is made by twolines (( x + y ) in the nomenclature of Gabriel 1972), is partiallyresolved, with the most energetic one blended with the r line. Athena ’s capabilities will significantly improve also plasmadiagnostic, even at shorter exposures. To test how well wecan determine R and G ratios, we performed simulations with Athena / X-IFU at di ff erent exposure times. Fig. 13 shows the ra-tios of the Si regions at di ff erent exposures, in comparison withthe ratios obtained from the analysis of the 45.88 ks Chan-dra / HETGS observational data set. With an exposure of only 2.5ks the uncertainties on R and G are reduced of the ∼ R and G , from ∼ Athena / X-IFU were per-formed using standard response matrices and background files .A more thorough exploration of possibilities to observe Vela X-1 with Athena , including a detailed modelling of the e ff ects ofdefocussing necessary to avoid pile-up for bright X-ray binariesand the right choice of event grades to address certain scientificquestions, is beyond the scope of this work and will be addressedin a dedicated publication.Overall, the achievement of good-quality spectra with suchshort exposure times imply that the lines can be traced on shortertimescales, i.e., of the same order of magnitude as the pulsar pe-riod. Moreover, because of Athena ’s resolution, the energy ofthe Fe K α line can be better constrained so that we can be ableto determine the ionisation stage of iron with a higher precision.It is clear, then, that upcoming X-ray satellites will considerablyimprove the knowledge of HMXBs, of stellar winds and, in gen-eral, of any kind of astrophysical plasma, as well remarked byXRISM Science Team (2020).
7. Conclusions
We conducted, for the first time, X-ray high-resolution spec-troscopy of Vela X-1 at the orbital phase φ orb ≈ .
75, i.e., whenthe line of sight is going through the photoionisation wake thattrails the neutron star along the orbit.The data did not show any significant variability of the con-tinuum for the duration of the observation. A blind search forspectral features lead us to detect emission lines from Fe, S, Si, Response matrices for the
Athena / X-IFU can be found at: http://x-ifu-resources.irap.omp.eu/PUBLIC/RESPONSES/CC_CONFIGURATION/ . Background files are available at: http://x-ifu-resources.irap.omp.eu/PUBLIC/BACKGROUND/CC_CONFIGURATION/ P h . c m s Å ---- F e K -- F e K -- -- F e XX V -- -- N i K Å )200 C Fig. 12.
Simulated spectrum of the Fe region with the
Athena / X-IFUand best fit model, with residuals in the lower panel. Exposure time of300 s, data binned with a minimum of 15 counts / bin. R Exp. times (ks) G Fig. 13. R and G ratios for the He-like triplet of Si as obtained fromsimulations with Athena / X-IFU with di ff erent exposure times. Solidlines correspond to the best fit values with the error ranges given bythe coloured areas obtained from the present work. Mg, Ne, and, to a lesser degree, from Al and Na. We detectedand identified five narrows RRCs (Mg xi - xii , Ne ix - x , O viii ) andHe-like triplets of S, Si, Mg and Ne.From plasma diagnostic techniques and from fits with pho-toionisation models from CLOUDY and SPEX, we concludethat the plasma at this orbital phase is mainly photoionised, butdata suggest the presence of at least another component, with asmaller ionisation parameter. The presence of a collisional com-ponent cannot be excluded, as well as a mixture of ionised andcollisional phases. This is in agreement with the idea of colderand denser clumps of matter, embedded in the hot, optically-thinwind of the donor star. The complex geometry of the system isalso reflected by the spread of the distribution of the Doppler ve-locities, as well as in the indetermination of the emission region.The future X-ray instruments Athena / X-IFU and
XRISM / Resolve will considerably enhance the detectionand the resolution of spectral features. We showed throughsimulations that, thanks to higher energy resolutions, they willresolve single lines in the Fe K α doublets and Fe xxv triplet and, Article number, page 12 of 13. Amato et al.: Looking through the photoionisation wake: Vela X-1 at ϕ orb ≈ .
75 with
Chandra / HETG thanks to higher collecting areas, will allow plasma diagnosticfor time scales as short as few hundreds of seconds.
Acknowledgements.
Authors acknowledge financial contribution from the agree-ment ASI-INAF n.2017-14-H.0, from INAF mainstream (PI: T. Belloni). VGis supported through the Margarete von Wrangell fellowship by the ESF andthe Ministry of Science, Research and the Arts Baden-Württemberg. SB ac-knowledges financial support from the Italian Space Agency under grant ASI-INAF 2017-14-H.O. Work at LLNL was performed under the auspices of theU.S. Department of Energy under contract No. DE-AC52-07NA27344 and sup-ported through NASA grants to LLNL. This research has made use of NASA’sAstrophysics Data System Bibliographic Service (ADS) and of ISIS functions( isisscripts ) provided by ECAP / Remeis observatory and MIT. For the ini-tial data exploration, this research used the
Chandra
Transmission Grating DataCatalog and Archive ( tgcat ; Huenemoerder et al. 2011). This research alsohas used the following Python packages:
Matplotlib (Hunter 2007),
Numpy (Oliphant 2006),
Pandas (McKinney et al. 2010), and the community-developed
Astropy (Price-Whelan et al. 2018). We in particular thank M. Nowak for theimplementation of the Bayesian Block algorithm used in this work, M. Guainazzifor his input on
XRISM simulations, and I. El Mellah for helpful discussions.
References
Arnaud, K. A. 1996, in Astronomical Society of the Pacific Conference Series,Vol. 101, Astronomical Data Analysis Software and Systems V, ed. G. H.Jacoby & J. Barnes, 17Barret, D., Lam Trong, T., den Herder, J.-W., et al. 2018, in Society of Photo-Optical Instrumentation Engineers (SPIE) Conference Series, Vol. 10699,Proc. SPIE, 106991GBearden, J. A. & Burr, A. F. 1967, Reviews of Modern Physics, 39, 125Bianchi, S., Miniutti, G., Fabian, A. C., & Iwasawa, K. 2005, MNRAS, 360, 380Blondin, J. M., Kallman, T. R., Fryxell, B. A., & Taam, R. E. 1990, ApJ, 356,591Blumenthal, G. R., Drake, G. W. F., & Tucker, W. H. 1972, ApJ, 172, 205Brown, G. V., Beiersdorfer, P., Liedahl, D. A., Widmann, K., & Kahn, S. M.1998, ApJ, 502, 1015Canizares, C. R., Davis, J. E., Dewey, D., et al. 2005, PASP, 117, 1144Cash, W. 1979, ApJ, 228, 939Chakraborty, P., Ferland, G. J., Bianchi, S., & Chatzikos, M. 2020, ResearchNotes of the American Astronomical Society, 4, 184Doroshenko, V., Santangelo, A., Nakahira, S., et al. 2013, A&A, 554, A37Drake, G. W. 1988, Canadian Journal of Physics, 66, 586El Mellah, I., Sander, A. A. C., Sundqvist, J. O., & Keppens, R. 2019, A&A,622, A189El Mellah, I., Sundqvist, J. O., & Keppens, R. 2018, MNRAS, 475, 3240Erickson, G. W. 1977, Journal of Physical and Chemical Reference Data, 6, 831Ferland, G. J., Chatzikos, M., Guzmán, F., et al. 2017, Rev. Mexicana Astron.Astrofis., 53, 385Forman, W., Jones, C., Tananbaum, H., et al. 1973, ApJ, 182, L103Foster, A. R., Ji, L., Smith, R. K., & Brickhouse, N. S. 2012, ApJ, 756, 128Foster, A. R., Smith, R. K., & Brickhouse, N. S. 2017, in American Institute ofPhysics Conference Series, Vol. 1811, Atomic Processes in Plasmas (APiP2016), 190005Fürst, F., Kreykenbohm, I., Pottschmidt, K., et al. 2010, A&A, 519, A37Fürst, F., Pottschmidt, K., Wilms, J., et al. 2014, ApJ, 780, 133Gabriel, A. H. 1972, MNRAS, 160, 99Gabriel, A. H. & Jordan, C. 1969, MNRAS, 145, 241Garcia, J. D. & Mack, J. E. 1965, Journal of the Optical Society of America(1917-1983), 55, 654Giménez-García, A., Shenar, T., Torrejón, J. M., et al. 2016, A&A, 591, A26Goldstein, G., Huenemoerder, D. P., & Blank, D. 2004, AJ, 127, 2310Grinberg, V., Hell, N., El Mellah, I., et al. 2017, A&A, 608, A143Haberl, F. & White, N. E. 1990, ApJ, 361, 225Harrison, F. A., Craig, W. W., Christensen, F. E., et al. 2013, ApJ, 770, 103Hell, N., Brown, G. V., Wilms, J., et al. 2016, ApJ, 830, 26HI4PI Collaboration, Ben Bekhti, N., Flöer, L., et al. 2016, A&A, 594, A116Hiltner, W. A., Werner, J., & Osmer, P. 1972, ApJ, 175, L19Hirsch, M., Hell, N., Grinberg, V., et al. 2019, A&A, 626, A64Huenemoerder, D. P., Mitschang, A., Dewey, D., et al. 2011, AJ, 141, 129Hunter, J. D. 2007, Computing in science & engineering, 9, 90Joss, P. C. & Rappaport, S. A. 1984, ARA&A, 22, 537Kaastra, J. S. & Mewe, R. 1993, A&AS, 97, 443Kaastra, J. S. & Mewe, R. 1995, A&A, 302, L13 http://tgcat.mit.edu/ Kaastra, J. S., Mewe, R., & Nieuwenhuijzen, H. 1996, in UV and X-ray Spec-troscopy of Astrophysical and Laboratory Plasmas, ed. K. Yamashita &T. Watanabe, 411–414Kaastra, J. S., Raassen, A. J. J., de Plaa, J., & Gu, L. 2018, SPEX X-ray spectralfitting package, Zenodo, 10.5281 / zenodo.2419563Kaper, L., Hammerschlag-Hensberge, G., & Zuiderwijk, E. J. 1994, A&A, 289,846Kelley, R. L., Akamatsu, H., Azzarello, P., et al. 2016, Society of Photo-OpticalInstrumentation Engineers (SPIE) Conference Series, Vol. 9905, The Astro-Hhigh resolution soft x-ray spectrometer, 99050VKelly, R. L. 1987, Journal of Physical and Chemical Reference Data, 17Kinkhabwala, A., Sako, M., Behar, E., et al. 2002, ApJ, 575, 732Kreykenbohm, I., Wilms, J., Kretschmar, P., et al. 2008, A&A, 492, 511Liao, Z., Liu, J., Zheng, X., & Gou, L. 2020, MNRAS, 492, 5922Lodders, K., Palme, H., & Gail, H. P. 2009, Landolt Börnstein, 4B, 712Lomaeva, M., Grinberg, V., Guainazzi, M., et al. 2020, arXiv e-prints,arXiv:2007.07260Malacaria, C., Mihara, T., Santangelo, A., et al. 2016, A&A, 588, A100Manousakis, A. 2011, Dissertation, Université de GenèveManousakis, A. & Walter, R. 2015, A&A, 575, A58Martínez-Núñez, S., Torrejón, J. M., Kühnel, M., et al. 2014, A&A, 563, A70Matsuoka, M., Kawasaki, K., Ueno, S., et al. 2009, PASJ, 61, 999McClintock, J. E., Rappaport, S., Joss, P. C., et al. 1976, ApJ, 206, L99McKinney, W. et al. 2010, in Proceedings of the 9th Python in Science Confer-ence, Vol. 445, Austin, TX, 51–56Mewe, R. & Schrijver, J. 1978, A&A, 65, 99Miškoviˇcová, I., Hell, N., Hanke, M., et al. 2016, A&A, 590, A114Molendi, S., Bianchi, S., & Matt, G. 2003, MNRAS, 343, L1Nandra, K., Barret, D., Barcons, X., et al. 2013, arXiv e-prints[ arXiv:1306.2307 ]Noble, M. S. & Nowak, M. A. 2008a, PASP, 120, 821Noble, M. S. & Nowak, M. A. 2008b, PASP, 120, 821Odaka, H., Khangulyan, D., Tanaka, Y. T., et al. 2013, ApJ, 767, 70Oliphant, T. E. 2006, A guide to NumPy, Vol. 1 (Trelgol Publishing USA)Palmeri, P., Mendoza, C., Kallman, T. R., Bautista, M. A., & Meléndez, M. 2003,A&A, 410, 359Porquet, D. & Dubau, J. 2000, A&AS, 143, 495Porquet, D., Dubau, J., & Grosso, N. 2010, Space Sci. Rev., 157, 103Porquet, D., Mewe, R., Dubau, J., Raassen, A. J. J., & Kaastra, J. S. 2001, A&A,376, 1113Price-Whelan, A. M., Sip˝ocz, B. M., Günther, H. M., et al. 2018, AJ, 156, 123Quaintrell, H., Norton, A. J., Ash, T. D. C., et al. 2003, A&A, 401, 313Sako, M., Kahn, S. M., Paerels, F., & Liedahl, D. A. 2000, ApJ, 543, L115Sako, M., Liedahl, D. A., Kahn, S. M., & Paerels, F. 1999, ApJ, 525, 921Sander, A. A. C., Fürst, F., Kretschmar, P., et al. 2018, A&A, 610, A60Scargle, J. D., Norris, J. P., Jackson, B., & Chiang, J. 2013, ApJ, 764, 167Schulz, N. S., Canizares, C. R., Lee, J. C., & Sako, M. 2002, ApJ, 564, L21Tarter, C. B., Tucker, W. H., & Salpeter, E. E. 1969, ApJ, 156, 943Tashiro, M., Maejima, H., Toda, K., et al. 2018, in Society of Photo-OpticalInstrumentation Engineers (SPIE) Conference Series, Vol. 10699, Proc. SPIE,1069922Tzanavaris, P. & Yaqoob, T. 2018, ApJ, 855, 25van den Eijnden, J., Degenaar, N., Schulz, N. S., et al. 2019, MNRAS, 487, 4355van Loon, J. T., Kaper, L., & Hammerschlag-Hensberge, G. 2001, A&A, 375,498Verner, D. A., Ferland, G. J., Korista, K. T., & Yakovlev, D. G. 1996, Astrophys-ical Journal, 465, 487Wargelin, B. J., Beiersdorfer, P., Liedahl, D. A., Kahn, S. M., & von Goeler, S.1998, ApJ, 496, 1031Watanabe, S., Sako, M., Ishida, M., et al. 2006, ApJ, 651, 421Weisskopf, M. C., Tananbaum, H. D., Van Speybroeck, L. P., & O’Dell, S. L.2000, in Society of Photo-Optical Instrumentation Engineers (SPIE) Confer-ence Series, Vol. 4012, Proc. SPIE, ed. J. E. Truemper & B. Aschenbach,2–16Wilms, J., Allen, A., & McCray, R. 2000, ApJ, 542, 914Wojdowski, P. S., Liedahl, D. A., Sako, M., Kahn, S. M., & Paerels, F. 2003,ApJ, 582, 959XRISM Science Team. 2020, Science with the X-ray Imaging and SpectroscopyMission (XRISM)Yao, Y., Nowak, M. A., Wang, Q. D., Schulz, N. S., & Canizares, C. R. 2008,ApJ, 672, L21Young, A. J., Nowak, M. A., Marko ff , S., Marshall, H. L., & Canizares, C. R.2007, ApJ, 669, 830, S., Marshall, H. L., & Canizares, C. R.2007, ApJ, 669, 830