A FAST X-RAY DISK WIND IN THE TRANSIENT PULSAR IGR J17480−2446 IN TERZAN 5

The full 0.3–10.0 keV light curve of IGR J17480−2446 was searched for bursts. Figure 1 shows a segment of the total light curve; several short, weak bursts are evident. Chakraborty & Bhattacharyya (2011) suggest that the bursts seen in this transient are Type I bursts due to nuclear burning on the stellar surface, not Type II bursts due to an accretion disk process, as previously suggested by Galloway & in't Zand (2010). Theoretical work shows that mHz QPOs can indeed be due to marginally stable nuclear burning (Heger et al. 2007). We employed the search algorithm used by Maitra & Miller (2010) to flag bursts in the 0.3–10.0 keV light curve. In total, 35 bursts were identified, generally confined within 100–200 s intervals. Time-selected spectra were made using the procedure described above.

Zoom In Zoom Out Reset image size

Preliminary fits to the spectra revealed a rather high column density, approaching 10²² cm⁻². For this reason, the MEG spectra are not very sensitive to narrow atomic lines. Moreover, the flux calibration of the MEG in the Fe K band differs from the better-calibrated HEG. As a result, our spectral analysis focused only on the summed first-order HEG spectrum. Below 1.3 keV and above 9.0 keV, plausible continuum models leave residuals that are due to calibration uncertainties. More modest residuals are seen in the 1.7–3.0 keV band, likely owing to the complexity of modeling the effective area of CCDs in the Si K band, the Ir coating on the Chandra mirrors, and chip gaps that fall in this range. These features have little impact on the goodness-of-fit statistics. We therefore fit the HEG spectra across the 1.3–9.0 keV band.

Although numerous bursts are evident in the source light curve, they are short-lived features that represent a small fluctuations on top of a high flux level. Fits to the burst-free spectrum of IGR J17480−2446 yield results that are consistent with fits to the time-averaged spectrum; the 0.5–10 keV flux in the burst-free spectrum is only 5% lower than in the time-averaged spectrum. (The line properties measured during bursting phases, in the burst-free spectrum, and the time-averaged spectrum are also consistent.) Owing to this consistency, we focused on the time-averaged spectrum of IGR J17480−2446.

The continuum can be described well using a simple additive model, consisting of a blackbody component and a power law, both modified by absorption along the line of sight. In XSPEC parlance, this model is "tbabs*(bbodyrad+powerlaw)," and it gives χ²/ν = 3395.2/3255 = 1.04 (where ν is the number of degrees of freedom). This model gives a column density of N_H = (1.17 ± 0.04) × 10²² cm⁻², a blackbody temperature of kT =1.18 ± 0.01 keV, a blackbody normalization of 267 ± 9, a power-law index of Γ = 1.26 ± 0.05, and a power-law normalization of 0.46 ± 0.05. The normalization of the "bbodyrad" model gives a color radius of 9.0 ± 0.1 km (K_BB = R²/D²₁₀, where D₁₀ is the distance in units of 10 kpc); this value is consistent with plausible stellar radii (e.g., Lattimer & Prakash 2006). Extrapolating this simple model to the standard 0.5–10.0 keV soft X-ray band, it gives an unabsorbed flux of 1.01(5) × 10⁻⁸ erg cm⁻² s⁻¹, which corresponds to a luminosity of 3.7 ± 0.2 × 10³⁷ erg s⁻¹. Figure 2 shows the combined first-order HEG spectrum of IGR J17480−2446, fitted with this continuum.

Zoom In Zoom Out Reset image size

The spectrum of IGR J17480−2446 shows clear residuals in the Fe K band (see Figures 2 and 3). A broad emission line, whether a simple Gaussian line or a relativistic line, gives an improvement to the overall fit that is significant at far more than the 8σ level of confidence (the F-statistic is 7.8 × 10⁻²⁴). To characterize the absorption lines, we initially fit the broad line using a simple Gaussian function. The Gaussian model gives a centroid energy of 6.76(3) keV, which is broadly consistent with Fe xxv. The line strength is typical of neutron star low-mass X-ray binaries (LMXBs), with an equivalent width of 112(9) eV. The width is measured to be 0.75(7) keV (FWHM), or roughly 0.1c.

Zoom In Zoom Out Reset image size

Fits to the likely Fe xxv absorption line with a simple Gaussian give a line centroid energy of 6.77(1) keV, indicating a blueshift of 3100 ± 400 km s⁻¹ (for a rest energy of 6.700 keV; Verner et al. 1996). The Gaussian fit nominally indicates that the line is resolved, with a width of 110 ± 20 eV or 4800 ± 900 km s⁻¹. An equivalent width of 12 ± 3 eV is measured. This line may be saturated; its properties may not be accurately measured using a Gaussian.

Fits to the likely Fe xxvi absorption line with a simple Gaussian give a centroid energy of 6.988(4) keV, indicating a blueshift of 1000 ± 200 km s⁻¹ (for a rest energy of 6.966 keV; Verner et al. 1996). As Figures 3 and 4 indicate, this line is quite narrow, and it is not resolved. The 90% confidence upper limit on the width of this line is 13 eV or 600 km s⁻¹ (FWHM). An equivalent width of 6 ± 2 eV is measured for this line. This phenomenological continuum and complex of three Fe lines gives a very good fit: χ²/ν = 3248.3/3248.0. An apparent feature remaining at 6.1 keV is not statistically significant. The inconsistency of the velocity shift in the He-like and H-like lines could be real, but it may reflect possible saturation of the Fe xxv line. If it is real, the fact that the He-like line—which likely arises further from the ionizing flux than the H-like line—has an higher velocity, may signal that the wind is being accelerated. Assuming that the observed absorption occurs on the linear part of the curve of growth, the equivalent width of the Fe xxv line implies a column of N_Z = 1.4(4) × 10¹⁷ cm⁻²; that of the Fe xxvi line implies a column of N_Z = 1.3(4) × 10¹⁷ cm⁻².

Zoom In Zoom Out Reset image size

Modeling the broad feature with a relativistic line function does not give a significantly better fit, compared to fits made with a simple Gaussian function. Detecting asymmetry in a broad line requires high sensitivity (see, e.g., Cackett et al. 2009b, 2010; also see Miller et al. 2010). Fits with the "diskline" model give r_in = 20(2)  GM/c², or 41(5) km for a 1.4 M_☉ neutron star. This model also gives i = 18° ± 2°, E = 6.97_−0.02 keV, and an equivalent width of W = 130 ± 10 eV. The line emissivity index was fixed at q = −3, and the outer line emitting radius fixed at 1000 GM/c².

We note that RXTE observed IGR J17480−2446 on 2010 October 24 (observation 95437-01-10-02). The Proportional Counter Array (PCA) spectra appear to also reveal a broad emission line (Chakraborty & Bhattacharyya 2011). Simple fits to the PCU2 spectrum (including 0.6% systematic errors) in the 3–25 keV band strongly require a line with a similar width (0.9 ± 0.2 keV, FWHM) and strength (equivalent width EW =140 ± 20 eV), but pegging at the lowest plausible energy for Fe K (6.40 keV).

The radius given by the relativistic line fit permits a constraint on the stellar magnetic field strength. Using Equation (1) in Cackett et al. (2009a), making the same assumptions regarding geometry and the accretion efficiency, adopting an unabsorbed 0.1–30.0 keV flux of 1.9 × 10⁻⁸ erg cm⁻² s⁻¹ as the bolometric flux, taking a distance of 5.5 kpc (Ortolani et al. 2007), and assuming a mass of 1.4 M_☉, we obtain μ = 4.5(8) × 10²⁶ G cm⁻³. For a neutron star with a radius of 10 km, this gives a magnetic field strength of B = 9(2) × 10⁸ G at the poles. If a value of 0.5 (Long et al. 2005) is assumed for the conversion factor from spherical to disk accretion when balancing magnetic and ram pressures, the magnetic field at the poles would be B = 3(1) × 10⁹ G. Both values are consistent with estimates based on X-ray timing (Papitto et al. 2011).

In order to better model the absorption found in IGR J17480−2246, we constructed a grid of multiplicative models using XSTAR version 2.2 (Kallman & Bautista 2001). The blackbody plus power-law model detailed above was extrapolated to the 0.1–30 keV band, and supplied to XSTAR as the source of ionizing flux. Solar abundances were assumed. The outflow must be equatorial, owing to the lack of narrow emission lines in the spectra. We therefore assumed a global covering factor of 0.3. Last, the grids considered here assumed a number density of n = 10¹² cm⁻³, consistent with values used to characterize other winds in X-ray binaries.

Direct fits to the combined first-order HEG spectrum with the multiplicative table model strongly suggest that the Fe xxv and Fe xxvi lines observed cannot arise in exactly the same gas. Any single absorption zone that fits the Fe xxv line well also predicts a strong Fe xxvi line at a commensurate blueshift, which exceeds the shift measured from the observed Fe xxvi line. Similarly, any absorption zone that fits the observed Fe xxvi line well, underpredicts the Fe xxv line and fails to match its large observed blueshift.

A plausible fit to the spectrum (χ²/ν = 3306.1/3246) can be obtained using two absorption zones and a relativistic diskline model for the broad emission line (see Figure 5). The Fe xxv line is well described by an absorber with N_H = 3 × 10²² cm⁻², log(ξ) = 3.0, at a blueshift of 3100 km s⁻¹. The Fe xxvi line is well described by a slightly more modest absorber: N_H = 2 × 10²² cm⁻², log(ξ) = 4.3, at a blueshift of 500 km s⁻¹. The higher velocity absorption zone includes a strong Fe xxvi line that is not evident. This may be explained if the blue wing of a relativistic emission line (like that described above) coincides with the Fe xxvi absorption. The model described here is not unique (for this reason, errors are not given), and slightly different relativistic emission line profiles and absorption zones can be combined to give similar outcomes.

Zoom In Zoom Out Reset image size

Assuming that the absorbing gas is a filled volume of uniform density, the ionization parameter can be used to derive a radius within which the wind must be launched (r ≃ L/Nξ, where L is the source luminosity, N is the column density, and ξ is the ionization parameter). Even the more highly ionized absorber is relatively far away from the central engine: r ≃ 3 × 10⁵ km or r ≃ 1.5 × 10⁵ R_Schw. (assuming a neutron star mass of 1.4 M_☉). This is within the binary system of IGR 17480−2446: a sin(i) = 2.498(5) s or 7.5 × 10⁵ km (Papitto et al. 2011), and consistent with the outer disk.

The mechanical luminosity in a wind is given by $L_{W} = 0.5 \dot{m}v^{2}$. In practice, it is necessary to rewrite this equation in terms of the ionization parameter and a geometric factor: L_W = 0.5m_pΩv³(L/ξ), where m_p is the mass of the proton, Ω is the fraction of 4π covered by the wind (Ω = 0.3 is assumed in our XSTAR grids), v is the wind velocity, L is the ionizing luminosity, and ξ is the ionization parameter. The faster, less ionized absorption zone gives a mechanical luminosity of L_W ≃ 2.8 × 10³⁵ erg s⁻¹, or L_W/L_accr. ≃ 8 × 10⁻³.