CHANDRA AND SUZAKU OBSERVATIONS OF THE Be/X-RAY STAR HD110432

We have processed and analyzed the Chandra data from the ObsID 9947. The High Energy Transmission Gratings (HETG) spectrometer (Canizares et al. 2005) acquired data during a total of 141.28 ks. Both High Energy Gratings (HEG; 1.55–17.7 Å) and Medium Energy Gratings (MEG; 1.55–31 Å) were used for the analysis. After excluding wavelength ranges with insufficient signal, we obtained spectral coverage from 1.6 to 16 Å. The spectra and the response files (arf and rmf) were extracted using standard procedures with the CIAO software (ver. 4.4). The data were binned to match the resolution of HEG (0.012 Å, FWHM) and MEG (0.023 Å, FWHM) and have a minimum signal-to-noise ratio (S/N) of seven. First dispersion orders (m = ±1) were fitted simultaneously. The peak count rate in a single (MEG) gratings arm spectrum is ≈0.04 counts s⁻¹ Å⁻¹, which is well below the level at which pileup is a concern.³

The analysis was performed with the Interactive Spectral Interpretation System (ISIS) version 1.6.1-24 (Houck & Denicola 2000). For the analysis, we use the abundance table from Anders & Grevesse (1989) and cross sections from Balucinska-Church & McCammon (1992).

We have processed and analyzed public data from instruments on board Suzaku (Mitsuda et al. 2007). We used standard procedures to reduce the XIS (Koyama et al. 2007) CCD detector (0.3–10 keV) and the HXD (Takahashi et al. 2007). The standard screening criteria (satellite out of the South Atlantic Anomaly, Earth elevation angle >5° and Day-Earth elevation angle >20°) were applied to create the Good Time Intervals (GTI). Events were readout from either 3 × 3 or 5 × 5 pixel islands. We created individual spectra and response files for each detector and data mode combination. Response matrices and effective area files were created using the xisrmfgen and xissimarfgen tasks, respectively. All three XIS detectors, XIS 0, 1, and 3, were combined and rebinned to have a minimum S/N of seven (XIS 2 was lost during a micrometeor hit in late 2006).

For the HXD, only data from the PIN diode detector (10–70 keV) presented enough signal to produce a useful spectrum. The PIN data were extracted from the cleaned event files in the hxd/event-cl directories. The source spectra exposure times were corrected with the hxdcorr tool. The PIN spectra were grouped to have an S/N ⩾10 in each energy bin. Table 1 presents the journal of the observations with both telescopes.

In previous works on HD110432 (LO07) and γ Cas (Smith et al. 2004; Lopes de Oliveira et al. 2010), it was shown that the description of the spectrum of these sources requires several optically thin thermal plasma components (mekal, apec) to correctly account for both the shape of the continuum and the emission line strengths. One of the critical points left open by LO07 was whether the different plasmas in HD110432 were affected by distinct absorption columns. The high-resolution Chandra HETG spectrum will allow us to further investigate this issue. In Figure 6, we show the Chandra HETG unfolded spectrum of HD110432.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

It has been shown in Torrejón et al. (2010), that the Fe K fluorescence is always accompanied by the presence of a K edge in massive binaries. This edge is produced by those photons energetic enough to ionize the innermost K shell electrons of the atom. This hole is subsequently replenished by an electron from the upper level producing the observed fluorescence line. Consequently, the Fe K edge is deeper than what is imposed by the curvature of the spectrum at low energies. In general, then, all the models will require an additional edge at 1.74 Å (=7.125 keV). We will fix the wavelength and let the optical depth τ vary free. Likewise, all models require the addition of a fluorescence line at 1.9366 Å, compatible with neutral Fe (i to viii) from the illumination of cold material or reflection onto the tentative WD. This line is not included in the plasma models.

As can be seen in Table 5, the emission lines present in the Chandra spectrum trace a range of plasma temperatures. We have tried a single thermal component modified by a single absorption column, using the vmekal model. As expected, this model does not produce a good fit to the continuum, especially at low energies. The reduced χ²_r = 1.46 for 938 degrees of freedom (dof). In order to account for the soft excess, and in agreement with LO07, we find that at least three plasmas, at different temperatures, are required. However, using a single absorption column to modify all three thermal components yields a reduced χ²_r = 1.25 for 935 dof. The most evident discrepancies are still in the low energy continuum. In order to properly describe the whole continuum, we find that three distinct absorption columns must be used to modify each of the three vmekal components. The fit improves significantly. The reduced chi square is now χ²_r = 1.19 for 933 dof. An F test gives a statistic value of F = 23.5, and the probability for the addition of the individual absorptions not being significant is only 1.08 × 10⁻¹⁰. At this point, the continuum is fairly well described both at high and at low energies by three optically thin thermal plasmas at kT ≈ 22, 8, and 0.2 keV, respectively. Following previous works, we will call them hot, warm, and cool plasmas, respectively. The overall Fe abundance is 0.48 solar. The main discrepancies are now in the line strengths. In particular, it is evident that the Fe L shell transitions (Fe xxiv, Fe xx) that dominate the low energy spectral range are severely underestimated by this model. Therefore, we next investigated the elemental abundances in the several plasmas by letting the Fe abundance to vary independently. We find that the abundance of the hot component is subsolar while the abundances of the warm and cool components are compatible between them. Therefore, we tie the abundances in the warm and cool components and let the abundance in the hot component vary freely. The best-fit parameters are shown in Table 2, under the Model 2 column. Owing to the normalization of the several components, the hot plasma dominates the spectrum, followed by the warm and cool plasmas, respectively. The absorption columns follow this hierarchy as well. The absorption for the cool component turns out to be very low. We have fixed it to the interstellar medium (ISM) value (N_H ≈ 0.15 × 10²² cm⁻²). The Fe abundance is subsolar. However, the Fe abundance of the hot component (Z_Fe = 0.30 Z_☉) is significantly lower than for the warm and cool components (Z_Fe = 0.76 Z_☉), which turn out to be equal within the errors and to have been tied during the fittings.

In Figure 7, we plot the unfolded spectrum with the three vmekal composite model. In the first three panels, we show the resulting fit for the continuum and the lines, while in the fourth panel we show the individual contributions of each component. The reduced χ²_r = 1.07 for 932 dof. As can be seen, the continuum is well accounted for but some discrepancies still exist in the lines. Two cases stand out. First, the observed strength of the Na xi Lyα line at 10.02 Å is much stronger than predicted by the model using solar abundance. We let it vary freely, keeping it tied for the three components. The resulting Na abundance is very high ([∼15 ± 5] Z_☉). Unfortunately, the spectrum starts to be noisy at these wavelengths, but a high Na abundance seems to be necessary to correctly fit the line. Second, the peak emission of the Fe xxvi Lyα and Fe xxvi lines, at 1.78 and 1.86 Å, respectively, seems slightly overpredicted, while the observed profiles are definitely broader than the model. Therefore, the next step is to modify the width of the observed lines with the effects of turbulent velocity. For that purpose, we use the bvapec model, for collisionally ionized plasma based on the Astrophysical Plasma Emission Code (apec).

Zoom In Zoom Out Reset image size

In Table 2, under the Model 3 column, we show the corresponding fit parameters. The hierarchy of normalizations and absorptions is still observed. The temperatures of the hot and warm components are lower, kT = 16.3 and 7.1 keV, respectively, while the temperature of the cool component is a bit higher, kT = 0.28 keV, although its norm is smaller than in the previous model.

In Figure 8, we show the unfolded spectrum for this model. As can be seen, both Fe xxvi and Fe xxv are much better described now. These two lines are the main contributors, within the investigated spectral range, to the v_turb determination. The required v_turb is of the order of 1200 km s⁻¹. No turbulent broadening is needed to modify the warm and cool components. Note, however, that the Mg (9.2 Å) and Na (10.02 Å) lines are underestimated. The bvapec model does not have the Na abundance as variable and cannot fit it well. This reinforces the conclusion from the previous analysis that the Na seems to be highly abundant. However, the reduced χ²_r = 1.04 for 931 dof is the best we have found in our analysis. It is plausible that a correct description of the whole spectrum will be a hybrid collisional/photoionized plasma. In the last panel of Figure 8, we show the relative contribution of the several plasmas. The prominent S xvi, Si xiv, and Mg xii lines as well as the Fe L shell transitions are contributed mainly by the kT = (8–9) keV plasma. Although it also contributes to Fe xxv and Fe xxv lines, these are dominated by the hot component. Finally, the kT = 0.3 keV cool plasma contributes significantly only beyond λ ∼ 12 Å.

Zoom In Zoom Out Reset image size

Given the limited Chandra bandpass, it is necessary to check the consistency of the hard tail with a hot (thermal or no-thermal) emission. Indeed, at such high temperatures, a thin plasma already displays a power-law shape at the energies considered here. To that end, we substitute the hot component by an absorbed power law. The fit is also good and the parameters are given in Table 2, under the Model 1 column. The most significant change is the increase in the norm of the warm component, which is now dominant. The main discrepancy between the data and the model is, again, in the line strengths. Fe xxv (1.85 Å) is overestimated while Fe xx (11.4 Å) is underestimated. In terms of the χ²_r and fit to the continuum, however, this model works equally well.

In order to try to break down this degeneracy, we have tried to establish the underlying broadband continuum using Suzaku. In Figure 9, we present the combined XIS and HXD Suzaku data fitted with three optically thin thermal plasmas. Significant emission is seen up to, at least, 33.5 keV, even though the number of energy bins is very low at high energies. In the BeppoSAX data presented in TO01, this hard tail could be seen up to 50 keV. The S/N of the PDS data on board BeppoSAX was, unfortunately, also low. However, a hard tail is definitely established for this system. Consistent with the previous discussion, the Suzaku data can be modeled with either a sum of thermal plasmas or a thermal-power-law mixture. The Suzaku data, however, do not require the cooler component. In Table 3, we present the details of the fit to the Suzaku data only. As can be seen in Tables 2 (Models 2 and 3) and 3 (Model 2), for pure thermal models, the source exhibits a complex variability, a fact already noted by LO07. The hot component is hotter during the Suzaku observation while the temperature of the warm component is similar. However, the norm of the hot and warm component was lower and higher, respectively, during the Suzaku observation. In the case of the thermal plasma + power-law model, the most noticeable change is in the norm of the power law which again was less important here than during the Chandra observation. This precludes any attempt to obtain physical parameters based on the joint fit to both data sets. Unfortunately, the low S/N of the Suzaku HXD data does not allow us to determine its full extension. The goodness of the fit is comparably good for both models. We cannot unambiguously constrain the very nature of the hard component with the available data sets and hence further observations will be required to establish this point.

Zoom In Zoom Out Reset image size

One of the fingerprints of accretion on to the WD in some cataclysmic variables is through the cooling of the post-shock X-ray plasma (see, for example, Mukai et al. 2003). In order to test this scenario, we have tried to fit the cooling flow model cemekl to the Chandra data. A single absorbed cemekl component does not fit the data at all. An acceptable fit is obtained with the addition of a power law. The best-fit parameters are displayed in Table 4. As can be seen, the maximum temperature displayed by the plasma is very high (∼37 keV), and the α parameter is close to the adiabatic case (α = 1). These parameters are consistent with those found by LO07 for the same model. The photon index of the power law is identical. This model, however, gives a much higher χ²_r = 1.24 for 936 dof, than the models discussed in the previous section. The main discrepancies here are in the fit to the low energy continuum as well as the Fe xxvi Lyα line, which is overestimated. We also tried other combinations like cemekl+mekal, but the quality of the fit does not improve over the one presented here.

In Figure 6, we show the Chandra HETG spectrum of HD110432. Below (in red), we have plotted the HETG spectrum of γ Cas analyzed in Smith et al. (2004), in the same wavelength interval, scaled and offset to allow comparison. In general terms, the spectra of both objects are very similar. There are, however, some noticeable differences. Evidently, the Fe xxvi line is much broader in HD110432 than in γ Cas. The Na xi is much stronger in HD110432. Despite the lower S/N of the HD110432 spectrum, it seems evident that the Ne lines at 12.13 and 13.5 Å, respectively, are much stronger in γ Cas. The same seems to be true for the Fe xvii lines around 15 Å.

The spectrum of HD110432 shows a wealth of emission lines available for plasma diagnostics, including the Fe Kα fluorescence at 1.94 Å. To extract their individual parameters, every line was fitted with Gaussians upon a three bremsstrahlung continuum, following the analysis made in the previous sections. Some lines remain unresolved. In such cases, we fixed the line width to σ = 0.005 Å. The measured equivalent width (EW), though, was found to be independent of this value. All identified and analyzed lines are listed in Table 5.

In the sixth column, we quote the temperature at which the emissivity of the ion is maximum (kT_max), taken from the atomdb⁵ database. This temperature, however, does not necessarily reflect the temperature of the emitting plasma.

In the last column, we quote the emission measure for each ion. This has been computed from the equation:

$$\begin{equation} {\rm EM}_{i}=\frac{4\pi d^{2} F_{i}}{\epsilon _{i}}, \end{equation} \tag{ 1 }$$

where F_i is the measured line flux, _i is the line emissivity at T_max, and d is the distance to the star, which we take as d = 301 pc (Hipparcos; Perryman et al. 1997). As can be seen, values of EM_i are of the order of 10^53–55 cm⁻³. As stated before, the kT_max might not reflect the true temperature of the plasma and, therefore, the emissivity can have a lower value.

The S xvi 4.73 Å line presents a possible P Cygni profile (Figure 10). This possibility is rather interesting, since it would point to the stellar wind as one of the possible sites for X-ray production in the system. P Cygni profiles have been observed by Chandra in the high-mass X-ray binaries Vela X-1 and Cyg X-3 (Vilhu et al. 2009). The blue absorption to red emission peak separation implies a velocity of ∼3400 km s⁻¹. Such high values have only been found in the terminal wind velocity v_∞ of very early type stars like the O3 star HD93129A (Taresch et al. 1997). This is, however, much larger than the wind velocity measured through the UV C iv lines (Smith & Balona 2006) in HD110432 and would rather suggest some violent localized outflow. Unfortunately, no other line presents such a profile in our spectrum and further monitoring will be required to address this issue.

Zoom In Zoom Out Reset image size

4.3.1. He-like Triplets

Three He-like triplets have been identified in the spectrum. Although the S/N is low (see Figure 11), it can provide useful constraints, such as temperature and density diagnostics for the plasma (Porquet & Dubau 2000). The three lines visible in the triplets are called r, i, and f, which stand for the resonance, intercombination, and forbidden transitions, respectively. The models used in the previous section do not fit perfectly all the triplets individually. For example, the bvapec composite model is able to describe the r and f transitions of both S xv and Si xiii (although slightly underestimated) but does not show any detectable i transition. The vmekal, in turn, describes well the r, i, and f transitions in the Si xiii (albeit, again, underestimating all three lines) while they are undetectable in the S xv triplet. Therefore, we use the approach described previously where a Gaussian has been fitted to every transition (when present) over a multiple bremsstrahlung continuum.

Zoom In Zoom Out Reset image size

Using values in Table 2 for the Si xiii triplet, we can compute the index G = (i + f)/r = 0.73, where r, i, and f stand for the fluxes of the corresponding lines in the triplet, respectively. This value is indicative of a high plasma temperature of the order of 10 MK. This is also supported by the presence of Mg xi, which is not too prominent if T < 10⁷ K. On the other hand, the index R = f/i = 0.42 points to a high plasma density of the order of n_e ≈ 10^13–14 cm⁻³. The high density is also supported by the absence of the f component of the Mg xi triplet which tends to be suppressed at high densities. This value is consistent with the high end range deduced by Smith et al. (1998) for the hot plasma in γ Cas, namely, n_e ≈ 10^13–14 cm⁻³.

A strong UV radiation field can mimic a high density plasma because the forbidden line decreases due to the depopulation of the upper (³S) levels via photoexcitation while the intercombination line is increased when the lower (³P) levels are populated with these electrons (Porquet et al. 2001). Consequently, we have considered the possible influence of the intense UV radiation from the B0.5III star photosphere of blackbody temperature T = 33,000 K, diluted by a factor W. For that purpose, we shall use also the R value for the S xv triplet, R = f/i = 0.86. Blumenthal et al. (1972) have calculated the low density, low UV irradiation limit of this quantity, R₀, for Si xiii, 2.51, and for S xv, 2.04. The observed R value will depart from these values according to the equation

$$\begin{displaymath} R=\frac{R_{0}}{1+\frac{n_{e}}{n_{c}}+2W\frac{\phi }{\phi _{c}}}, \end{displaymath}$$

where ϕ is the photoexcitation rate and n_c and ϕ_c are the corresponding critical values for the plasma density and the photoexcitation rate, respectively, and W is the geometrical dilution factor. We see that R → R₀ when n_e ≪ n_c (low density limit) and ϕ ≪ ϕ_c (low UV irradiation limit). Consequently, a value of R lower than (but not orders of magnitude lower than) R₀ means high densities and/or high UV irradiation of the X-ray emitting plasma. In order to gain some insight on the possible plasma emission location, we shall use the values of ϕ/ϕ_c tabulated by Blumenthal et al. (1972) for a blackbody of 10⁵ K and dilution factor W = 0.5 (stellar surface), scaled here to the photospheric temperature T = 33, 000 K and arbitrary W. We obtain ϕ/ϕ_c = 74 for Si xiii and ϕ/ϕ_c = 17 for S xv. We shall assume, further, an average plasma density for HD110432 of the order of n_e ≈ 10¹³ cm⁻³, consistent with the observed R value for HD110432 and that derived for γ Cas. In that case, the n_e/n_c ≈ 1 for Si xiii (for which n_c = 4 × 10¹³) while for S xv it will be n_e/n_c ≈ 0.1 (n_c = 1.9 × 10¹⁴). With these values, we obtain dilution factors of 0.03 and 0.04 for Si xiii and S xv, respectively. Now, recalling that W = (1/2)[1 − (1 − (R_*/r)²)^1/2], where r is the distance from the X-ray plasma site to the stellar center, we can locate the plasma at a distance of r ≃ (2–3)R_*. Although the associated errors are large (∼50%), the small values of R with respect to the low density, low UV flux limit make it clear that neither the density nor the irradiation terms can be neglected in the equation given in Section 4.3.1, locating the emission sites relatively close to the photosphere of the Be star at r < 3R_*.

Alternatively, this intense UV radiation field could come from the radiation of a putative WD companion. For the sake of comparison, we will take the WD surface temperature of the order of 10⁵ K. In such a case, this will be the primary ionizing field and we can use directly the values of ϕ/ϕ_c calculated by Blumenthal et al. (1972) to obtain dilution factors of W = 0.76 for Si xiii and W = 0.2 for S xv. These would translate into emission regions located at r ∼ [1.4–1.5]R_WD, respectively, in units of the (putative) WD radius. That is to say, the X-ray emission would be produced very close to the WD surface, as might be expected if the X-rays are produced via accretion.

4.3.2. The Fe Complex

The available data represent an isolated instance in the life of HD110432 and are insufficient, by itself, to settle this controversy. In principle, the described situation could be produced either by X-ray centers over the Be star photosphere in the active star disk interaction paradigm, or by accretion over a compact object orbiting relatively close to the stellar surface. Notwithstanding, we have presented here several pieces of evidence that can provide valuable new light onto the nature of the system when joined with previous results.

There is some evidence that the source is persistent. Indeed, it has been clearly detected by all the X-ray telescopes that have pointed at it: HEAO, BeppoSAX, RXTE, XMM, Chandra, and Suzaku. None of these observations were performed in any particular time or phase, yet the source was clearly detected and the light curve showed no eclipses. Unfortunately, those observations do not span in time enough to firmly conclude that the source is non-eclipsing. In order to check this hypothesis, we have analyzed the RXTE–All Sky Monitor (ASM) light curve for HD110432. The source is in the very limit of detection of ASM and most of the bins are compatible with zero flux. A continuum coverage with a large area collector ASM would be important to address this point. If we assume that the source is persistent and non-eclipsing, then either the X-ray emission is uniformly distributed over the area or the inclination of the system is such that the putative degenerate companion is never eclipsed.

As we have seen, the available data could favor an origin of the X-ray emission close to the stellar surface, at distances r < 3R_* from the center of the Be star (1R_* is the stellar surface). In such a case, the compact object should be placed in a orbit with radius r < 3R_* with an orbital period between four and eight days. Given that we do not observe periodic outbursts at these periods, the orbit should be coplanar with the circumstellar disk. In order not to observe any eclipse, the inclination of the system should be such that cos i > R_*/r, which delivers the constraint i < 60°–70° for orbital radii of 2–3R_*, respectively. This is very unlikely. The optical spectrum of HD110432 shows all its lines in emission with double peaks (see LO07, Figure 1; Smith & Balona 2006) clearly indicating that the system is seen with i ≈ 90°.

On the other hand, as we have seen in the previous section, we cannot exclude that the X-ray emission is produced very close to the stellar surface r ∼ 1.5R_WD of the putative WD companion. The multiplicity of absorbing columns is, however, difficult to reconcile with a single X-ray source embedded into the circumstellar disk. In any case, at least part of the X-ray emission appears to be absorbed by a column which is compatible with the density column of the circumstellar disk. Since the system is seen edge on, this would be in agreement with an X-ray source located close to the Be stellar surface which, in turn, argues in favor of the Be star photosphere being the main source of ionizing UV photons. The continuum monitoring of the source to detect or rule out the presence of eclipses would thus be important in understanding the nature of HD110432.

On the other hand, an X-ray emission distributed on or over the surface of the star could account naturally for the different absorbing density columns for each component as depicted in Figure 12. Since the system is seen close to edge on and the disk flares outward (in order to be in hydrostatic equilibrium), the lower the latitude of the X-ray source the higher the absorption. Thus, a number of small X-ray sources distributed on or above the stellar photosphere at several latitudes would account naturally for the characteristics of the observed emission.

Zoom In Zoom Out Reset image size

The light curve shows a high variability and flaring but all efforts made so far have failed to detect any coherent pulsation. The persistent lack of detection of any coherent modulation strongly argues against the degenerate companion scenario. In conclusion, with the available data at hand, the X-ray active standalone star seems to emerge as the least problematic explanation. The mechanism behind such energetic emission, however, remains a mystery and is challenging. In particular, the extreme temperatures of the order of 15–30 keV or, alternatively, the presence of nonthermal components is not explained under this paradigm. The fact that other isolated Be stars with similar spectral types and comparably large circumstellar disks do not show any X-ray emission at all, is not explained either. Further observations will be required to establish the nature of this mysterious object. In particular, it would be necessary to (1) obtain a high S/N spectrum of the high energy tail in order to unambiguously establish the nature of the hard component using, for example, NuSTAR. This would require, probably, the simultaneous use of Chandra and/or XMM-Newton. (2) Monitor the star with an ASM to establish or rule out the persistence of the X-ray emission and the presence or lack of eclipses (with MAXI or LOFT), (3) and a monitoring campaign with a large collecting area telescope (to reduce the long exposure times) of the triplets to look for some periodic modulation that could arise from the (co)rotation of the X-ray center(s) around the star.