The mass of X-ray Nova Scorpii 1994 (=GRO J1655-40)

T. Shahbaz 2 F. van der Hooft 1 J. Casares 0 P. A. Charles 2 J. van Paradijs 1 0 Instituto de Astrofsica de Canarias , 38200 La Laguna, Tenerife, Spain 1 Astronomical Institute 'Anton Pannekoek', University of Amsterdam and Center for High Energy Astrophysics , Kruislaan 403, 1098 SJ Amsterdam, the Netherlands 2 University of Oxford, Department of Astrophysics, Nuclear Physics Building , Keble Road, Oxford OX1 3RH A B S T R A C T We have obtained high- and intermediate-resolution optical spectra of the black hole candidate Nova Scorpii 1994 in 1998 May/June, when the source was in complete (X-ray) quiescence. We measure the radial velocity curve of the secondary star and obtain a semi-amplitude of K2 215:5 6 2:4 km s 1, which is 6 per cent lower than the only previously determined value. This new value for K2 thus reduces the binary mass function to f M = 2:73 6 0:09 M(. Using only the high-resolution spectra, we constrain the rotational broadening of the secondary star, v sin i, to lie in the range 82.9-94.9 km s 1 (95 per cent confidence) and thus constrain the binary mass ratio to lie in the range 0.337-0.436 (95 per cent confidence). We can also combine our results with published limits for the binary inclination to constrain the mass of the compact object and secondary star to the ranges 5.5-7.9 and 1.7-3.3 M( respectively (95 per cent confidence). Finally, we report on the detection of the lithium resonance line at 6707.8 A , with an equivalent width of 55 6 8 m A . I N T R O D U C T I O N The soft X-ray transient, Nova Sco 1994 (=GRO J1655 40) has been studied extensively over the past 3 yr in X-rays and at optical wavelengths (Tingay et al. 1995; Bailyn et al. 1995a,b; Harmon et al. 1995; Hjellming & Rupen 1995; Paciesas et al. 1996; van der Hooft et al. 1997; van der Hooft et al. 1998, hereafter vdH98). Strong evidence that the compact object in GRO J1655 40 is a black hole was presented by Bailyn et al. (1995b), who initially established a spectroscopic period of 2:601 6 0:027 d, and determined a mass function f M 3:16 6 0:15 M(. An improved value of f M 3:24 6 0:09 M( was presented by Orosz & Bailyn (1997; hereafter OB), who measured the radial velocity semi-amplitude of the secondary star to be K2 228:2 6 2:2 km s 1 and classified it as an F2F6 IV type star. Both vdH98 and OB analysed their quiescent optical light curves of Nova Sco 1994, which they combined with the OB value for K2 to show that the black hole has a mass (M1) in the range 6.297.60 M(. However, in calculating the radial velocity semi-amplitude, OB used both quiescent data (taken during 1996 February 2425) and outburst data (taken during 1995 April 30May 4). Using outburst data in this way can lead to an incorrect result, as substantial X-ray heating of the secondary star shifts the effective centre of mass of the star (see Wade & Horne 1988 and Phillips, Shahbaz & Podsiadlowski 1999). This results in a significant distortion of the radial velocity curve and hence a spuriously high velocity semi-amplitude. The masses of the binary components derived from this will therefore be incorrect. In this paper we determine the radial velocity curve of the secondary star in Nova Sco 1994 using only data taken when the system was in quiescence. From our high-resolution spectroscopy we determine the rotational broadening of the secondary star, which, when combined with the new value for the mass function and limits to the binary inclination (vdH98) allows us to determine more precisely the individual component masses. 29/5/1998 30/5/1998 31/5/1998 2/6/1998 4.2 A 4.2 A 61316764 A 44527015 A 44527015 A 61316764 A Parameter NTT (this paper) NTT + CTIO (OB) Taken from van der Hooft et al. (1998) and fixed. 1j uncertainties are given, see text. rotational broadening much less than the resolution of our highresolution data. The data reduction and analysis was performed using the Starlink figaro package, the pamela routines of K. Horne and the molly package of T. R. Marsh. Removal of the individual bias signal was achieved through subtraction of the mean overscan level on each frame. Small scale pixel-to-pixel sensitivity variations were removed with a flat-field frame prepared from observations of a tungsten lamp. One-dimensional spectra were extracted using the optimal-extraction algorithm of Horne (1986), and calibration of the wavelength scale was achieved using fifthorder polynomial fits which gave an rms scatter of 0.02 A . The stability of the final calibration was verified with the OH sky line at 6562.8 A , the position of which was accurate to within 0.1 A . 3 T H E R A D I A L V E L O C I T Y O F T H E C O M PA N I O N S TA R Our first task was to measure the radial velocity of the F-type secondary star. The radial velocities were measured from the spectra by the method of cross-correlation (Tonry & Davis 1979) with a template star. Prior to cross-correlation the spectra were interpolated on to a logarithmic wavelength scale (pixel size 14.5 km s 1) using a sin x =x interpolation scheme to minimize data smoothing (Stover et al. 1980), and then normalized. We masked the interstellar feature at 6280 A and Ha before correlating the spectra. We used HR 2927 as our template star after we had broadened it by 90 km s 1 in order to simulate the spectrum of Nova Sco 1994 (a more accurate value for the rotational broadening is derived in the next section). The radial velocity of the template star (derived using the position of the Ha absorption line to be 6.7 km s 1) was then added to the radial velocities of Nova Sco 1994. Using the orbital ephemeris given by vdH98, we phasefolded the heliocentric radial velocities and then fitted a sine wave (see Table 2 and Fig. 1), obtaining a xn2 of 0.99. We then combined the data with the quiescent radial velocity data of OB and repeated the fitting procedure obtaining a significantly worse value of xn2 4:9. Note that our high-resolution data has a factor of 4 better spectral resolution than that of OB, the uncertainty in our individual radial velocity measurements being typically 6 km s 1. HR 3325 HR 2927 HR 870 HR 9046 HR 9057 HR 6192 HR 7864 HR 7281 4 T H E S P E C T R A L T Y P E A N D R O TAT I O N A L B R O A D E N I N G O F T H E C O M PA N I O N S TA R We determine the spectral type of the companion star by minimizing the residuals after subtracting different template star spectra from the Doppler-corrected average spectrum. This method is sensitive to the rotational broadening v sin i and the fractional contribution of the companion star to the total flux. The template stars we use are in the spectral range F0F8 VIII, and were obtained during this observing run but were also observed during previous runs at La Palma with comparable dispersion. First we determined the velocity shift of the individual highresolution spectra of Nova Sco 1994 with respect to each template star spectrum, then Doppler-averaged to the rest frame of the template star ( (...truncated)