EMERGENCE OF A BROAD ABSORPTION LINE OUTFLOW IN THE NARROW-LINE SEYFERT 1 GALAXY WPVS 007^*

The spectra were extracted from the HST Data "Online" Web site⁶ for Legacy instruments including the Faint Object Spectrograph (FOS). These are final calibration data products, so no recalibration was necessary. The observation log is presented in Table 1.

Table 1. Observing Log

Spectrometer	Date	Exposure (s)	Bandpass (Å)	Aperturea	Resolution (Å)
HST FOS (G130H)	1996 July 30	3840	1140–1606	086	2.26
HST FOS (G190H)	1996 July 30	1500	1590–2312	086	3.16
HST FOS (G1270H)	1996 July 30	1280	2222–3277	086	4.72
HST FOS (G1400H)	1996 July 30	1000	3235–4781	086	6.94
HST FOS (G570H)	1996 July 30	630	4569–6818	086	10.06
FUSE LIF1a	2003 Nov 6	47,354	987.5–1082.5	30''	0.25b
FUSE LIF1b	2003 Nov 6	48,054	1094.25–1188.75	30''	0.25b
FUSE LIF2a	2003 Nov 6	48,229	1087.0–1181.25	30''	0.25b
FUSE LIF2b	2003 Nov 6	48,466	980.0–1074.75	30''	0.25b
FUSE SIC1a	2003 Nov 6	35,055	1003.25–1090.5	30''	0.25b
FUSE SIC1b	2003 Nov 6	35,259	904.25–992.5	30''	0.25b
FUSE SIC2a	2003 Nov 6	34,982	917.5–1005.5	30''	0.25b
FUSE SIC2b	2003 Nov 6	35,010	1016.75–1103.5	30''	0.25b

Notes. ^aFor the HST FOS spectra, this refers to the size of the round aperture. For the FUSE spectra, this refers to the size of the LWRS square aperture. ^bThe observed resolution of FUSE using the LWRS aperture is R = 20, 000 ± 2000 (The FUSE Observer's Guide: http://fuse.pha.jhu.edu/support/guide/guide.html). However, the low count rate cannot sample this resolution. Therefore, the resolution refers to the final binsize of the spectra.

Download table as:  ASCIITypeset image

We first verify the wavelength calibration of the spectra by measuring the wavelengths of Galactic absorption lines. We used Si ii λ1260.3, C ii λ1334.5, Al iii λ1670.8, Fe ii lines at 2344.2, 2382.8, 2586.6, and 2600.2 Å, and Mg ii λλ2796.4, 2803.5 (note that vacuum wavelengths are used, as appropriate for the HST data). These lines showed that the spectral segments obtained using the G130H, G190H, and G270H gratings were consistent with no anomalous wavelength shifts. There were no convenient Galactic absorption lines available to examine the spectra obtained with the G400H and G570H gratings; since we find no anomalous redshifts or flux offsets (see below), we assume that these spectra are also free of wavelength shifts.

We next merge the spectra starting with the short wavelengths. We examined overlapping regions for flux offsets and found no evidence for any. The final signal-to-noise ratio (S/N) of the continuum varies from ∼3.7 at wavelengths shortward of 1600 Å (observed frame), to 22 in near 2700 Å, and falling to ∼9 around 6600 Å. We smooth the spectra using a three-point scheme in which the center point is weighted three times the adjacent points. We correct for the Galactic reddening of E(B − V) = 0.012 mag (Schegel et al. 1998) using the Cardelli et al. (1989) reddening curve. Finally, we shift the spectrum into the rest frame, adopting the NASA/IPAC Extragalactic Database redshift value of 0.02882.

The HST spectrum is relatively blue longward of ∼2700 Å, but becomes significantly redder shortward of that value. We show the spectrum in Figure 1 overlaid on two comparison spectra. The first comparison spectrum is a mean of HST spectra of two NLS1s that were chosen as follows with the intention of finding objects similar to WPVS 007 in properties other than the absorption lines and continuum shape. As discussed by Leighly & Moore (2004), there is significant range of UV emission-line shapes among NLS1s; focusing on C iv, we found that some are rather narrow, strong, and symmetric around the rest wavelength, while others are broad, weak, and strongly blueshifted. The mini-BALs absorb most of the far-UV high-ionization emission lines in WPVS 007, but the emission-lines profiles outside of the absorption lines appear to be rather symmetric about the rest wavelength. Thus, we sought other NLS1s that had emission lines similar in shape to the portion of the emission lines in WPVS 007 outside of the absorption lines. We found that the WVPS 007 spectrum most closely resembled that of Mrk 493, which had HST FOS spectra stretching from ∼1160 Å to 6818 Å (observed frame), and Mrk 335, which had HST FOS spectra stretching from ∼1160 Å to 3300 Å (observed frame) in this far-UV region of the spectrum. These two spectra were very similar to each other in slope and emission lines in the overlapping region. These spectra were processed in the same way as discussed above, and then uniformly resampled on a logarithmically binned wavelength scale that matched the original binning approximately. They were then rescaled to match with flux and averaged over the common wavelength region. The second comparison spectrum is the FBQS radio-quiet composite spectrum (Brotherton et al. 2001). The composite spectrum has a very similar continuum shape as the Mrk 493–Mrk 335 average, implying that the Mrk 493–Mrk 335 have continua similar to the average quasar. Thus, we make the simplifying assumption that the unreddened intrinsic WPVS 007 spectrum has the same shape as the Mrk 493–Mrk 335 average, and therefore similar to that of the average quasar.

Zoom In Zoom Out Reset image size

Figure 1 shows clearly that WPVS 007 is reddened in comparison with the NLS1 mean spectrum and the composite spectrum. The reddening curve appears to be very unusual, however. WPVS 007 and the comparison spectra have essentially the same spectra longward of ∼2700 Å, which means that there is either no attenuation in the optical bands, or that there is no wavelength dependence in the attenuation. We develop the reddening curve as follows. We remove Galactic absorption lines and then coarsely rebin the WPVS 007 spectrum and the Mrk 493/Mrk 335 average. We remove regions of prominent emission lines and compute the ratio. The ratio was then fit with a spline model. The ratio appears to be approximately flat shortward of ∼1500 Å. The ratio appears to have some structure longward of ∼3000 Å, but it is clear that it is caused by higher equivalent-width optical Fe ii in WPVS 007, so it was assumed to be one longward of ∼2700 Å. Since the ratio is flat in the optical band, we cannot define a reddening curve in the standard way, relative to E(B − V), as has been done for other AGNs by, e.g., Crenshaw et al. (2001) and Crenshaw et al. (2002). Instead, we define it relative to 2000 Å, and assume that the optical bands are not attenuated at all. The extinction curve from the spectra, the spline fit, and the SMC reddening curve (Prévot et al. 1984) are shown in Figure 2. The extinction curve in WPVS 007 is unusually steep between 2700 Å and 1700 Å. An attempt was made to reproduce this reddening curve using silicate dust and a distribution of sizes (Weingartner & Draine 2001), but it was unsuccessful (J. Weingartner 2008, private communication). While the unusual reddening curve is interesting, we note that it has no effect on the absorption-line measurements or analysis.

Zoom In Zoom Out Reset image size

The HST spectrum reveals high-ionization absorption lines in N v, C iv, and Lyα. Figure 3 shows the observed WVPS 007 spectrum (corrected for redshift and Galactic reddening only) between 1185 and 1595 Å in comparison with the average Mrk 335–Mrk 493 spectrum. The presence of these absorption lines was previously noted by Crenshaw et al. (1999). The N v absorption lines are clearly resolved, and we begin our analysis there. A preliminary examination of the lines shows that the absorbing gas must occult both the line-emitting gas and the continuum emission region since the absorption lines are deeper than the continuum level.

Zoom In Zoom Out Reset image size

We use a template analysis approach to analyze the absorption lines. Briefly, we create an apparent optical depth template from the N v profiles and then fit that to the C iv and Lyα. The template approach has strengths and weaknesses. A strength is that it is a very straightforward approach. A weakness is that, because of saturation and partial covering, it yields only a lower limit on the column density of the absorbing ions. Nevertheless, the results provide useful constraints on the properties of the absorbing gas.

We develop the apparent optical depth template as follows. The mean Mrk 335–Mrk 493 spectrum was resampled on the wavelength scale of the WPVS 007 spectrum and the ratio R of the WPVS 007 spectrum and the average spectrum was measured. Assuming that the average Mrk 335–Mrk 493 spectrum is a good representation of the unabsorbed WPVS 007 spectrum, R should be one outside of the range of the absorption lines, and should be less than one in the region of the absorption lines. The region of the N v absorption lines was isolated, and the apparent optical depths (τ = −ln R) for the two N v lines were computed. The apparent optical depth profiles for the two lines were essentially indistinguishable, implying that absorption is saturated but not black. The shape is commonly understood to be the consequence of a velocity-dependent covering fraction. Since the apparent optical depth profiles were indistinguishable, they could be averaged to produce a mini-BAL template (Figure 4). The HST mini-BAL has an approximate maximum velocity of V_max ∼ 900 km s⁻¹ and approximate FWHM of ∼550 km s⁻¹.

Zoom In Zoom Out Reset image size

If the Mrk 335–Mrk 493 mean spectrum is an accurate representation of the unabsorbed WVPS 007 spectrum, we should be able to apply the mini-BAL template to the Mrk 335–Mrk 493 mean spectrum in the region of C iv and Lyα, and reproduce the observed WPVS 007 spectrum in those regions. To investigate this, we fit the spectrum using IRAF Specfit. Specfit allows a template continuum, and for that we use the mean Mrk 493–Mrk 335 spectrum. For the absorption lines, we use the template developed from the N v line and shown in Figure 4. We fix the template wavelengths (in Specfit treated as a redshift) to the rest wavelengths for each line, and allow the scale factors to be free in the spectra fitting. The best-fitting model results are shown in Figure 5, where we see that the N v template reproduces the C iv and Lyα absorption lines well. The scale factors are given in Table 2, which also includes the FUSE spectral fitting results discussed in Section 3.3. The first column of Table 2 lists the line, the second column gives the vacuum wavelength, the third column is the scale factor of the template including the uncertainty from the spectral fitting, the fourth column is the template used with the exception of the alternative deblending of the FUSE spectrum as discussed in Section 3.3.4, and the fifth column is the estimate of the lower limit of the column density obtained by integrating over the line profile (Equation (9); Savage & Sembach 1991). The uncertainties in the column densities are proportional to the uncertainties in the scale factors in the fitting, and thus they give only an estimate of the statistical uncertainty. We use the apparent optical depth to estimate the column densities without accounting for saturation and partial covering. Therefore, the column densities are lower limits. Atomic data were taken preferentially from NIST⁷ as well as from The Atomic Line List v2.04⁸ and Morton (1991).

Zoom In Zoom Out Reset image size

The analysis of broad absorption lines is very complicated. If the lines are overlapping, the continuum may be difficult to identify, and absorption lines themselves may be blended. The opacity of a line depends on velocity, there may be partial covering that may also depend on velocity, and these cannot be diagnosed directly because of blending. The covering fraction can also differ for different lines and ions. In the case of very high signal-to-noise spectra and unblended lines, these factors may be robustly deconvolved (e.g., Arav et al. 2007), but that is not possible for our relatively blended, low signal-to-noise spectrum. Therefore, our goal in this section is to use the P v λλ1117.98, 1128.01 feature in the FUSE spectrum to derive an apparent optical depth profile to be used to fit and deblend the other lines.

As noted in Section 3.1, P v is clearly seen in the FUSE spectrum. Phosphorus has much lower solar abundance than other elements such as carbon, nitrogen, and oxygen that commonly produce BALs (e.g., P/C = 0.00093; Grevesse et al. 2007). Therefore, even if those lines are saturated and partial covering is important, it is possible that the P v will not be as optically thick. P v occurs in a region of the spectrum where absorption from other ions is not generally observed and, therefore, it may not be subject to severe blending. In fact, the doublet ratio (1118 Å component divided by the 1128 Å component) is 2.03, and visual examination of the region suggests that the 1118 Å component has a larger apparent optical depth than the 1128 Å line. Therefore, we first analyze the P v line with the intention of developing an apparent absorption-line template from this feature that can be used for the other absorption lines.

Figure 7 shows an expanded version of the region of the FUSE spectrum near P v, after three Galactic absorption lines marked in Figure 6 had been modeled out. For the continuum, we use a spline fit to the HST FOS composite spectrum with an initial factor of 0.7 scaling (the level shown in Figure 6). We assume that P v is solely responsible for the absorption between 1095 and 1127 Å. Therefore, the termination of the absorption at 1095 Å marks the maximum extent of the 1118 Å absorption trough, implying a maximum velocity of 6179 km s⁻¹ for the broad absorption lines. Using this fact, the region between 1095 and 1127 Å can be broken into three parts: one in which the absorption is produced by the 1128 Å component only (right in Figure 7), by the 1118 Å component only (left in Figure 7), and by a combination of both components (middle in Figure 7). The extents of the 1128 Å absorption trough and the 1118 Å absorption trough are also marked in Figure 7.

Zoom In Zoom Out Reset image size

In terms of velocities, we glean the following information from Figure 7. The velocity profile between zero and −2678 km s⁻¹ can be constructed from the 1128 Å component (right region in Figure 7). The velocity profile between 3555 and 6179 km s⁻¹ can be constructed from the 1118 Å component (left region in Figure 7). Apparent optical depth profiles from these two regions are easily obtained. But the velocity profile between −2678 and −3555 km s⁻¹ must be constructed from the overlap region. The overlap region consists of a superposition of absorption from the −2678 to −3555 km s⁻¹ part of the 1128 Å trough and the 0 to −854 km s⁻¹ part of the 1118 Å trough.

Thus, our strategy to produce the apparent optical depth profile is as follows. We generate the apparent optical depth by dividing the spectrum by the continuum and taking the logarithm (τ_eff = −ln R). We construct the profile in the 0–2678 km s⁻¹ band using the 1128 Å component, and the profile in the 3555–6178 km s⁻¹ using the 1118 Å component. Then, the two estimates for the 2678–3555 km s⁻¹ are generated from the overlap regions by subtracting the known component from the apparent optical depth. The overlap region is then the mean of these two estimates. To test the resulting template, we apply it to the continuum to generate the spectrum.

The two uncertainties in this procedure are the ratio of the apparent taus of the two lines and the placement of the continuum. Specifically, we assume that the derived optical depth profile corresponds to the 1128 Å component, and the 1118 Å component is related by a ratio, to be determined. Initially, we assume that the ratio is the ratio of the gf_ik for the two lines, 2.03, appropriate if the absorption lines were optically thin and the covering fraction is unity. The result is shown as the dashed dark gray line in Figure 7. By definition, the simulated spectrum models the profile well in the low- and high-velocity regions of the feature that are created by either the 1118 or the 1128 Å component each by itself. However, in the overlap region, the synthetic spectrum clearly shows stronger absorption than in the observed spectrum.

There are several possible origins for the discrepancy between the synthetic spectrum and the observed spectrum. First, the continuum level could be lower than assigned. A continuum offset seems unlikely, though, as we set the continuum level at the point where the sharp low-velocity drop on the 1128 Å component occurs. More likely, the absorption lines could be saturated (τ>1) and partial covering could be important so that the ratio between the apparent optical depths of the 1118 component and the 1128 component is less than 2.03 and approaches one.

Since the ratio of the apparent taus of 2.03 is ruled out, we determine a best-fitting ratio as follows. We define a figure of merit (FOM), which is the sum of the absolute value of the difference between the simulated spectrum and the observed spectrum in the overlap region. Simulated spectra are generated using a range of apparent tau ratios between 2.03 and 1. The FOM is a shallow apparently parabolic function of the ratio with a minimum of 1.35. The simulated spectrum using this best-fitting apparent tau ratio is shown by the light gray line in Figure 7. The fit is not perfect, being too shallow for shorter wavelengths and too deep for longer wavelength in the overlap region, but it is clearly better than the fit for a ratio of 2.03. Another possibility is that the apparent tau ratio varies as a function of velocity, with a lower ratio (approaching one) for low velocities and a higher ratio for high velocities. Since the FOM has no statistical significance, we estimate the uncertainty on this value by examining at the residuals and determining the point at which the simulated spectrum clearly does not fit the data, yielding a conservative range of 1.15–1.55.

The constraints on reasonable placement of the continuum are fairly tight. The absorption decreases sharply at low velocity and, thus, the continuum cannot reasonably be placed very far below the position where this joins the continuum. Likewise, unless we do not see the continuum at all in this object, we cannot place the continuum much higher. We examine the cases in which the continuum was set to be about 4% higher and lower than the original value; this range seemed consistent with the constraints above. The results are very similar. A higher continuum yields a higher maximum velocity and perhaps a slightly better fit for longer wavelengths in the overlap region. A lower continuum yields a lower maximum velocity and a notably worse fit at longer wavelengths in the overlap region. The best FOM was obtained for the original continuum level, but the differences are not large. Therefore, we use the optical depth profile obtained using the original continuum level, with an inferred ratio of the apparent optical depths of the 1118 Å component to the 1128 Å component of 1.35. The apparent optical depth profile plot is shown in Figure 4. The profile has an approximate maximum velocity of V_max ∼ 6000 km s⁻¹ and an approximate FWHM of ∼ 3400 km s⁻¹.

As shown in Figure 6, a set of Fe iii transitions occurs near 1126 Å, between the two P v lines. We do not believe that this transition produces significant absorption for several reasons. First, if there were significant absorption from Fe iii, the template that we derive would be wrong; yet, as we show below, this template produces an excellent fit for the S iv absorption features at 1063 Å and 1073 Å. In addition, we noted the similarity to the LBQS 1212+1445 spectrum also shown in Figure 6. The P v troughs appear narrower and are not blended with one another in that object, and Fe iii absorption does not appear to be present. For future use, we measure upper limits on the column density of a multiplet of N ii near 1085 Å, and one of Fe iii near 1126 (Figure 6) by inserting a FUSE BAL template into the model and increasing the normalization until the χ² increases by 6.63, corresponding to 99% confidence for one parameter of interest. These upper limits are also listed in Table 2.

The FUSE spectrum of WPVS 007 has mini-BALs as well as BALs. These can be clearly seen on O vi and Lyβ, and they can also be seen on N iii and C iii. They are not seen on Si iv or P v. They are distinguished from the BALs not only by their distinctive shape, most clearly seen on O vi, but also because they have a lower velocity onset than the BALs, as shown in Figure 8. In this section, we analyze the mini-BALs in the region of Lyβ and O vi. We note that although the mini-BALs are not blended, we are not able to derive the optical depth and covering fraction as a function of velocity, because of both the poor S/N and a zero-offset problem (described below). Thus, as before, our goal is to produce an apparent tau velocity profile to be used as a template for fitting the other mini-BALs.

Zoom In Zoom Out Reset image size

The mini-BALs are superimposed on the BALs and, thus, their profiles are difficult to analyze. To define a pseudocontinuum, we isolate the region of the spectrum between 1020 and 1038 Å, and remove the regions of the spectrum containing the mini-BALs from Lyβ and O vi λλ1031.9, 1037.6. We then fit the remainder with a model intended to parameterize the pseudocontinuum. We find that a lorentzian line model plus a constant work well. Next, we need to find the apparent optical depths, generally done by dividing the continuum by the data, and then taking the logarithm. The problem with that procedure for these data is that the spectrum in the region of the mini-BALs dips below zero. This is, of course, not physical, and may be a consequence of residual background subtraction problems; although as shown in figures in Appendix A, the errors are sufficiently large that the data are consistent with zero in the region of the mini-BALs. Therefore, to obtain the template, we add 0.1 to both the continuum model and the spectrum before taking the logarithm. The result is encouraging, as it shows three apparent optical depth profiles that are very similar in shape and depth, although the O vi λ1032 component has slightly larger apparent optical depth. We resample the Lyβ and O vi λ1038 components to the wavelengths of the O vi λ1032 component and compute the mean profile. The result, as a function of velocity, is shown in Figure 4 in comparison with the mini-BAL template developed from N v in the HST spectra and the FUSE BAL profile developed in Section 3.3. The FUSE mini-BAL template has approximately the same maximum apparent optical depth as the HST mini-BAL; however, this is a lower limit as we had to add a constant to the spectrum in order to compute the apparent optical depth. The FUSE mini-BAL perhaps seems to have a sharper onset, a little lower maximum velocity (∼800 km s⁻¹) and a little smaller FWHM (∼470 km s⁻¹), compared with the HST mini-BAL, perhaps indicating evolution of this component.

Now that we have developed apparent optical depth templates for the BAL (discussed in Section 3.1) and the mini-BAL (discussed in Section 3.2), we can apply the templates to the remainder of the FUSE spectrum. First, though, we emphasize that the template approach has both values and shortcomings. The templates can be used to deblend the complicated absorbed spectrum to identify which lines are present and obtain estimates of the absorption columns of associated ions. But since we have only the apparent optical depth profile, and no information about the covering fraction that may be both velocity and ion dependent, we can only obtain lower limits on the ionic column densities. In addition, we make the assumption that all of the lines have the same apparent optical depth profile, while in reality, the higher ionization lines are sometimes observed to have broader profiles in other objects. We address this final point in Section 3.3.4.

Our approach is as follows. As for the HST spectrum, we fit the spectrum using IRAF Specfit. We use the HST composite continuum (Zheng et al. 1997) with the normalization fixed at the level shown in Figure 6 for the continuum. We use the two templates derived in Sections 3.1 and 3.2, fixing the wavelength (treated in Specfit as a redshift) to correspond to the rest wavelengths of the absorption lines investigated, allowing only the scale factor (effectively, the apparent optical depth) to be free. To obtain the best fit, we allow the apparent optical depth ratios within multiplets to deviate from the values prescribed by atomic physics. This leads to separate limits on the ionic column densities derived from each line. We discuss three regions of the spectrum separately: the P v/S iv region between 1037.5 and 1135 Å, the O vi region between 991 and 1135 Å, and the C iii/N iii/P iv region shortward of 991 Å.

3.3.1. P v/S iv Region

We first fit the P v region by itself, between 1090 and 1135 Å. The results for the fit of the entire FUSE spectra are given in Table 2, where the definitions of the columns were discussed in Section 2.3. The best fit is shown in Figure 9. The upper panel shows the final model superimposed upon the spectrum, and the lower panel shows each absorption component.

Zoom In Zoom Out Reset image size

We next fit the region between 1037.5 and 1135 Å. This region includes a large absorption feature that can plausibly be identified as S iv. This line is composed of three components. The ground-state transition has a wavelength of 1062.7 Å. The other two transitions from this configuration arise from an excited state with E_i = 951.43 cm⁻¹, with wavelengths 1072.96 and 1073.51 Å. The g_if_ik values for the three transitions are 0.69, 2.09, and 0.23. The latter two lines (the transitions from the excited state) are very close together in wavelength, and we model them with the g_if_ik-weighted mean wavelength of 1073.07. Absorption from ions in the excited state is clearly present; it can be recognized by the sharp low-velocity onset near 1073 Å (Figure 9). We obtain an excellent fit, as seen in Figure 6. In particular, the position of the sharp increase in opacity at low velocity matches the spectrum perfectly for both components, verifying that this feature is indeed S iv and that significant absorption from the excited state is present.

The undeniable presence of the excited-state absorption from S iv is interesting because the energy of the lower level is slightly high (0.118 eV) compared with typical values of nebular kT. Population of this level requires n_e > A_ki/q_ki, where A_ki = 7.70 × 10⁻³ is the Einstein A value for the transition from the excited state to the ground state, resulting in the S iv 10.51 μm line, and q₂₁ is the collision deexcitation rate, which can be found from the collision strength ϒ for this transition. For a nebular temperature of 15,000 K, Saraph & Storey (1999) give ϒ(S iv) = 8.48. The resulting critical density is n_crit = 4.7 × 10⁴ cm³. The fitting reveals that the excited-state line is somewhat stronger than the ground-state line (scale factor 1.54 ± 0.03 versus 1.20 ± 0.05), yielding a ratio of 1.28. At densities above critical, the ratio would approach g_ik(excited)/g_ik(ground) = 2.02. This suggests that the gas density is near or above the critical density.

3.3.2. The O vi Region

3.3.3. The C iii, N iii, and P iv Regions

3.3.4. Alternative Deblending

3.3.5. Column Density Lower Limits

We would like to determine the physical state of the absorbing gas and investigate other parameters such as the launching radius in order to understand outflows in AGNs. While this can be done fairly robustly in some cases where the S/N is very good and the individual lines in multiplets are resolved (e.g., Arav et al. 2007), it is generally difficult in BALs because the lines are saturated and there is partial covering of the emission source(s). But the situation is a little better for WPVS 007 because we measure P v in the FUSE spectrum, and it does not seem to be completely saturated. We find that we can obtain some interesting and useful limits through analysis of the BALs using the photoionization and spectral synthesis code Cloudy (Ferland et al. 1998).

4.1.1. Spectral Energy Distribution

In this section, we try to constrain the properties of the absorbers by comparing the results of Cloudy models with the column density limits obtained in Section 3.3.5. Cloudy requires as input a spectral energy distribution (SED). As discussed in Section 2.2, the UV and FUV spectra are clearly reddened, and the X-rays are generally not detected (until recently; see below), so it is not possible to build an SED directly from the WPVS 007 observations. However, as discussed in Section 2.3, the emission lines of the NLS1s Mrk 493 and Mrk 335 are very similar to those of WPVS 007, and these objects have roughly similar optical luminosity as WPVS 007 (log(L₅₅₀₀) = 39.1, 39.4, and 40.2 for WPVS 007, Mrk 493, and Mrk 335, respectively), so as a first approximation, we assume that the intrinsic SEDs are similar. Therefore, we construct an SED for WPVS 007 using data from Mrk 493 and Mrk 335.

Both Mrk 493 and Mrk 335 have been observed by XMM-Newton and, therefore, simultaneous optical/UV and X-ray data are available. Mrk 493 was observed once. During that observation, only the v-band filter was used. The Optical Monitor (OM) image clearly shows the galaxy as well as the bright central AGN, and it is clear that integration of the flux over the nominal XMM-Newton extraction aperture of 12'' will include a significant amount of galaxy light. Yet we need to use this aperture as it is calibrated for conversion to energy fluxes. So, we devise an aperture correction using an observation of the bright quasar PKS 0558 − 504 observed on 2000 May 24. This object is very luminous, and there is no apparent galaxy emission in the V-band image. Using IRAF, we shifted all 10 images using the V-band filter to a common centroid and summed. We determined the FWHM of the point-spread function (PSF) from the PKS 0558 − 504 image to be 3.34 pixels. We then extracted the net flux within one FWHM, considering this to be dominated by the AGN, and within 12'', using a background annulus between 14'' and 25'', and derive an aperture correction of 1.35. We extracted the flux from the Mrk 493 image in a region with a radius of one FWHM and also within 12''. We then correct for deadtime and coincidence losses using the instructions in the XMM-Newton Users Handbook.⁹ Correcting the flux from the small aperture using the correction factor derived from the PKS 0558–504 observation, we find that about 70% of the flux within the large aperture originates in the galaxy. We also compare the radial profiles from the Mrk 493 and the PKS 0558–504 images, finding that they agree with the size of the small aperture of 3.34 pixels and verifying extended emission from the galaxy at larger radii. Finally, we extract the flux and correct for the sensitivity degradation using the instructions.¹⁰ We then scale the HST spectrum to match the flux point. The XMM-Newton EPIC data from Mrk 493 and Mrk 335 were analyzed in the usual way. Finally, each time-averaged spectrum was fitted with a broken power law between 0.3 and 6 keV.

There were two XMM-Newton observations of Mrk 335. The first was performed on 2000 December 25. During this observation, images were obtained using the V, B, U, and UVM2 filters. The host galaxy spectrum is expected to be red and, therefore, do not contribute significantly to the flux in the UVM2 filter. Therefore, we use the results of the standard OM analysis, and then scaled the HST spectrum to the UVM2 filter flux. The X-ray data were analyzed in the usual way except that pileup was present in the data from the PN-CCD instrument and, therefore, an annular extraction aperture was used. The spectra were fitted with a broken power law. The second observation was performed on 2006 January 3. The OM was operated using the UV grism, and the data were processed in the standard way. The HST spectrum was then scaled to the grism spectrum. The X-ray spectrum was analyzed in the usual way and fitted to a double broken power-law model. Finally, the combined UV and X-ray spectra were sampled in line-free regions to construct SEDs consisting of 10–12 points each.

The three SEDs are very similar. The α_ox values for the Mrk 493, and the first and second Mrk 335 observations were −1.33, −1.36, and −1.32, respectively. These are very close to the value of −1.22 predicted for WPVS 007 based on the 5500 Å luminosity obtained from the HST spectrum and the regression published by Steffen et al. (2006) and their cosmological parameters (H₀ = 70 km s⁻¹ Mpc⁻¹, Ω_M = 0.3, Ω_Λ = 0.7). Finally, we obtain a merged SED. Since the HST spectrum of Mrk 493 extended to the optical band, we started with SED for that object. The Mrk 335 SEDs were rescaled to match the Mrk 493 SED in the near-UV. We found that the Mrk 335 points straddled the Mrk 493 points in the X-ray, such that an average would have been very close to the Mrk 493 points. We found that the far-UV of the Mrk 335 spectra were very slightly higher than the Mrk 493 points, and appeared to match the optical power law better, so we replaced the three far-UV points in the SED with the ones from the scaled Mrk 335 spectrum. We interpolated over the unobservable far-UV using a power law between the HST and XMM-Newton PN-CCD spectra. We use the Cloudy AGN kirk spectrum at higher and lower frequencies.¹¹ The final spectrum has α_ox = −1.28.

After completing the analysis outlined above, and associated Cloudy simulations, Grupe et al. (2008) report the first detection of hard X-rays in WPVS 007. They find evidence for a partially covered spectrum, and a plausible deconvolution indicates an intrinsic (i.e., unabsorbed) α_ox of −1.9. Therefore, we construct an additional continuum spectrum with α_ox = −1.9 by simply decreasing the flux of the X-ray and higher energy portion of the SED developed above by a factor of 41.3. The two SEDs are shown in Figure 10.

Zoom In Zoom Out Reset image size

4.1.2. Simulations and Results for the BALs

Initially, we ran Cloudy models for the α_ox = −1.28 continuum for a range of parameters: ionization parameter −3.5 ⩽ log(U) ⩽ 0.5, Δlog(U) = 0.25; density 8.0 ⩽ log(N_H) ⩽ 12.0, Δlog(n) = 0.5; column density parameter log(N^max_H), defined as log(N^max_H) = log(N_H) − log(U), 21.0 ⩽ log(N^max_H) ⩽ 24.0, Δlog(N^max_H) = 0.1 for a total of 4743 simulations. After determining that the absorption lines are not dependent on the gas density (e.g., see also Hamann 1997), we choose log(n) = 10.0 as the nominal density and extend the simulations to log(U) = 1.5. We also ran the simulations for the α_ox = −1.9 continuum for that value of the density. We use log(N^max_H) rather than just log(N_H) because a constant value of log(N^max_H) probes similar depths in terms of the hydrogen ionization front (e.g., Leighly 2004; Casebeer et al. 2006; Leighly et al. 2007). We used solar metallicity initially.

We followed the approach discussed by Arav et al. (2001) and since used by a number of other investigators. Specifically, we compare our estimates of the column density given in Table 2 with ionic column densities from the Cloudy simulations as a function of ionization parameter and column density. We emphasize that since the absorption lines are saturated and partially covered, and since we estimate the column densities from the apparent optical depths, these are the lower limits of the column density. Most of the ionic column densities can be extracted from the "column density" output from Cloudy. There is one relevant exception: columns from low-lying excited states of metal ions such as S iv. These should be populated at typical broad-line region (BLR) temperatures and sufficiently high densities according to their statistical weights 2J + 1. Thus, the proportion in the ground (J = 3/2, the 1062 Å component) and fine structure (J = 5/2, the 1073 Å component) states should be 40% and 60% of the column density given by the Cloudy output. Interestingly, though, the estimated column densities given in Table 2 show that the proportions of the estimated column density upper limits are 44% and 56%, very close to the expected proportions and, therefore, comparing the estimated lower limits with the simulations for both lines would yield indistinguishable results. Thus, for the following analysis, we compare 60% of the Cloudy S⁺³ column with the observed column of the 1073 Å component.

We first examine the results from the α_ox = −1.28 simulations. We consider the BAL deblending, shown in Figure 9, using the columns listed in Table 2. As mentioned in Section 3.1, we also measure upper limits on the column density of a multiplet of N ii near 1085 Å, and one of Fe iii near 1126 (Figure 9) by inserting a FUSE BAL template into the model and increasing the normalization until the χ² increases by 6.63, corresponding to 99% confidence for one parameter of interest. These upper limits are also listed in Table 2. Figure 11 shows the ionic column contours from Table 2 on the ionization parameter/N^max_H plane for log(n) = 10.0. Since the column densities we estimate are lower limits, the solutions should lie above and to the right of any curve. Ideally, if the lines were not saturated, the contours would converge in one area of the plot, producing best-fitting values of log(U) and log(N_H) − log(U). The presence of P v implies that lines of similar ions of more abundant species are saturated (Hamann 1998) and, therefore, we do not expect any region of convergence a priori. Interestingly, we do find one in which a majority of the lines converge: for log(U) ⩾ ∼0.0, the column densities of a number of the intermediate-ionization ions converge around log(N_H) − log(U) = 23.0. Notable exceptions are O⁺⁵ and H⁰, for which simulations show much higher column density than observed at log(U) = 0.0, log(N_H) − log(U) = 23.0, and S⁺² and P⁺², for which simulations show lower column densities than observed at that point. If the lines are saturated, the observed column densities are lower limits and, therefore, the O⁺⁵ and H⁰ columns may be consistent with the other lines at log(U) = 0.0, log(N_H) − log(U) = 23.0, but the P⁺² and S⁺² would not be, because the lower limit exceeds the simulated column densities at that point.

Zoom In Zoom Out Reset image size

We note that, as expected, there is little dependence on the density for values in the range 9.5 ⩽ log(n) ⩽ 11. The column densities of P⁺² are lower at very low densities and increase as the density is increased, but only for the highest values of log(N^max_H). At the highest densities, lower ionization columns increase, primarily at the highest values of log(N^max_H). In all cases, the differences are small, and we cannot obtain a markedly better agreement with the observations by assuming very high or very low densities.

As discussed in Section 3.3, while the deblending in terms of the FUSE BAL template presented in Figure 9 matches the spectrum well, it is possible that the higher-ionization lines have a broader profile, and O vi dominates the region between 990 Å and 1037 Å and, therefore, we presented the alternative deblending in Section 3.3.4. The alternative deblending does not attribute any measurable opacity to hydrogen, P⁺², or S⁺² (since the onset of absorption for these lines cannot be clearly seen in the spectrum), but instead attributes the absorption to O⁺⁵ between 990 Å and 1037 Å, and to C⁺² between 953 and 978 Å. Using these alternative fitting results, we generate the contour plot as before (Figure 11, the right side). Now we see that all measured ions and upper limits are consistent near log(N_H) − log(U) = 23.0 and log(U) ⩾ 0, except for O⁺⁵, which is clearly saturated.

We perform the same procedure as above for the simulations using the α_ox = −1.9 continuum (Figure 10, the middle panel). The results are rather similar in that there is again a region of convergence for the intermediate-ionization ions, and in this region, the columns O⁺⁵ and H⁰ are much larger than our inferred lower limits, while the columns of P⁺² and S⁺² are much lower than our inferred lower limits. The principal difference is that the location of the convergence of many ionic columns occurs at a lower value of N^max_H. Our nominal best solution for this continuum is log(U) = 0.0 and log(N_H) − log(U) = 22.0. This result can be understood by considering that intermediate-ionization ions are created at a smaller column density in gas illuminated by an X-ray weak continuum than in gas illuminated by an X-ray normal continuum because the X-ray weak continuum is unable to produce the usual higher-ionization ions (e.g., Leighly et al. 2007). Most of the observed absorption lines (except O vi) are from intermediate-ionization ions.

Quasars are often inferred to have enhanced metallicities (e.g., Hamann et al. 2002a). The hydrogen column densities derived above are large, and as will be shown in Section 4.2, challenge outflow models. Inferred hydrogen column densities should be lower if the metallicity is enhanced. We therefore try an α_ox = −1.9 model with Z = 5 (that includes helium enhanced by a factor 1.29 and nitrogen enhanced by a factor of 25; Hamann et al. 2002a). The results are shown in the lower panel of Figure 11. Here, the intermediate-ionization ions converge for log(U) = 0 at log(N_H) − log(U) = 21.6.

Since we expect that partial covering is strongly influencing our analysis, the fact that the data and models converge for many of the ions at particular points in these graphs seems somewhat surprising. This convergence could be coincidental; alternatively, it could really be the case that only O vi and H i are very saturated and the intermediate-ionization ions seen in the FUSE spectrum are not very saturated. This idea is supported by the fact that the upper limits for the undetected absorption lines, Fe iii and N ii are seen in Figure 11 to lie either in the region of convergence, or not very far above it. Therefore, we use the column densities of convergence at log(U) = 0 (specifically, log(N_H) = 23.0, 22.2, and 21.6 for the α_ox = −1.28 set of simulations, the α_ox = −1.9 simulations, and the α_ox = −1.9 and Z = 5 simulations) for discussion in the rest of the paper, emphasizing that these columns are lower limits. The transmitted continua for these cases are shown in Figure 10.

Further analysis reveals some interesting facts about these solutions. First, the choice of ionization parameter is driven by P⁺⁴; specifically, the ionization parameter must be sufficiently high to produce a sufficient column of P⁺⁴ to match the data. This means that the ions seen in the UV spectra are the tip of the ion iceberg, since these ions dominate gas with much lower ionization parameters than log(U) = 0. That is, for most elements, most of the atoms are in higher ionization states, including C⁺³ and N⁺⁴, but also much higher ionization states, indicating that most of the opacity is in the extreme UV and soft X-ray band. That this is true can be seen in the absorbed spectra shown in Figure 10; much of the continuum in the extreme UV and soft X-ray is absorbed by gas with this ionization parameter and column density.

In summary, we find that we can explain the FUSE spectrum, in particular, the presence of absorption from the low-abundance element phosphorus, with a highly ionized (log(U) ⩾ 0.0), high-column density absorber. We need to make the assumption, though, that the high-ionization lines such as O vi are much broader than the intermediate-ionization line P v. Strongly enhanced abundance of phosphorus is not needed because some lines are saturated, although they are not black because partial covering is important. This is essentially the same result found by Hamann (1998) for the quasar PG 1254 + 047 (see also Hamann et al. 2002b). A subtlety explored here is the dependence of the result on the spectral energy distribution. The primary distinction for the X-ray weak α_ox = −1.9 simulations is that the simulated column densities of the intermediate-ionization ions including P⁺⁴ are large enough to match the observed column lower limit at significantly lower values of N_H. As noted above, this occurs because the X-ray-weak continuum is not able to create high-ionization ions and, therefore, the gas cools via transitions of intermediate-ionization ions even in front of the helium ionization front (the hydrogen ionization front for the α_ox = −1.9 continuum is located beyond log(N_H) − log(U) = 24). Thus, while partial covering implies that the column densities are probably larger than inferred from the apparent line optical depths, and the outflows are still massive enough to challenge radiative line driving as the acceleration mechanism (e.g., Hamann et al. 2002b; see also Section 4.2), the situation is not as extreme if the continuum is X-ray weak, because sufficient P v can be produced at significantly smaller (a factor of six for solar abundances and a factor of 25 for enhanced abundances) column densities.

4.1.3. Implications for the X-ray Spectra

4.1.4. Simulations and Results for the Mini-BALs

We next consider the mini-BALs. We combine the results for the HST and FUSE observations, noting that these were not simultaneous and it is possible that the apparent optical depths changed between the two observations. Following the same procedure as above, we produce the contour plot shown in Figure 12, showing again the results for three cases: α_ox = −1.28, α_ox = −1.9, and α_ox = −1.9 with Z = 5. An important constraint on the mini-BALs is that we do not detect the mini-BALs in either P⁺⁴ or S⁺³ (Figure 8). We obtain upper limits on the mini-BAL columns of these ions by including the mini-BAL template in the spectral-fitting model, and increasing its opacity until the χ² increases by 6.63, corresponding to 99% confidence for one parameter of interest. Those values are listed in Table 2.

Zoom In Zoom Out Reset image size

Considering first the α_ox = −1.28 continuum, we find that we obtain a reasonable solution for log(U) ⩾ ≈ − 0.3 and a log(N_H) − log(U) lower limit of 22.8, corresponding to a log(N_H) lower limit of 22.5. This solution is consistent with the upper limit on mini-BAL column densities of both P⁺⁴ and S⁺³. This solution indicates that the Lyα, C iv, and N v absorption lines are strongly saturated.

We next present the results for α_ox = −1.9 (middle panel of Figure 12). We find a rough correspondence for log(U) ⩾ 0.5 and log(N_H) − log(U) lower limit of 22.2 corresponding to a log(N_H) lower limit of 22.8. Again, Lyα, C iv, and N v absorption lines are strongly saturated, consistent with the observed absorption-line intensity ratios (Section 2.3). However, this solution is not consistent with the upper limit on mini-BAL column densities of both P⁺⁴ and S⁺³.

Finally, for α_ox = −1.9 and Z = 5, we find a rough correspondence for log(U) ⩾ 0.5 and log(N_H) − log(U) lower limit of 21.5 corresponding to a log(N_H) lower limit of 22.0. In this solution, however, the column densities of P⁺⁴ and S⁺³ would exceed the upper limits. Thus, it does not appear to be possible to produce the mini-BALs and upper limits with either of the two continua we are considering.

It may be possible to reconcile the column density constraints from the detections and the upper limits if the upper limits on the unobserved lines are too low. As discussed in Hamann et al. (2002b), this could be true if the covering factor is higher for strong/detected lines than for weak lines. In a patchy absorber with a range of optical depths (see, e.g., Figure 6 of Hamann et al. 2002b), strong lines could have large optical depth over a larger area, while weak ones could still have τ>1 but over a smaller area, leading to a lower observed equivalent width. Our assumption of complete coverage for all lines would then lead to a too-low upper limit on the undetected weak lines.

As discussed in Section 1, outflows are common in AGN, but their nature differs between high- and low-luminosity objects. High-velocity outflows are generally limited to high-luminosity objects, while lower luminosity objects have typical outflow velocities of only <10³ km s⁻¹. The dependence of absorption on other AGN parameters has been systematically investigated using a sample of low-redshift (z < 0.5) quasars with M_V between ∼−21 and ∼−27 by Brandt et al. (2000) and Laor & Brandt (2002). A primary result of Brandt et al. (2000) is that there exists a significant correlation between α_ox and C iv absorption-line equivalent width, suggesting that the primary cause of X-ray weakness is absorption, with a continuum of absorption columns connecting unabsorbed objects to BALQSOs. In Laor & Brandt (2002), ideas associated with that continuum of absorption columns were further developed; in particular, the question of what makes a soft X-ray weak quasar a BALQSO was addressed. WPVS 007 is soft X-ray weak as a consequence of absorption, and it is apparently also intrinsically soft X-ray weak (Grupe et al. 2008). As shown in this paper, it had broad absorption lines during the FUSE observation in 2003. In this section, we compare some of our results with those of Brandt et al. (2000) and Laor & Brandt (2002).

One of the principal results of Laor & Brandt (2002) is that there is a strong dependence of outflow properties on luminosity. Specifically, there are strong positive correlations between the C iv equivalent width and V_max, the outflow maximum velocity, with the optical luminosity M_V for soft X-ray weak quasars, and at any luminosity, soft X-ray weak quasars had the largest equivalent widths and maximum velocities. As discussed by Laor & Brandt (2002), this behavior is consistent with outflow scenarios for outflows driven by either dust or line opacity.

We compare WPVS 007 with the quasar sample from Laor & Brandt (2002) in Figure 13. The M_V value of −19.8 was derived from the dereddened, rest-frame HST spectrum using H₀ = 50 km s⁻¹ Mpc⁻¹ and q₀ = 0.1 so as to be consistent with the other data. The straight line in that graph shows the best fit to the soft X-ray quiet quasars given by Laor & Brandt (2002); they note that it suggests an upper envelope as would be expected from radiation-driven winds. The filled star labeled WPVS 007 BAL is taken from the FUSE BAL template derived in Section 3.3. The V_max = 6179 km s⁻¹ for that absorption line is seen to be a factor of 13 in excess of the value expected if the regression holds for low-luminosity objects. We note that most of the data shown in Figure 13 were derived from C iv absorption lines, and it is not clear that the P v absorption line should have the same profile. However, as shown by Junkkarinen et al. (2001), P v absorption is typically better fit by templates derived from Si iv, and templates from Si iv tend to be narrower than templates from C iv. This evidence, including the fact that we infer that the O vi absorption profile in WPVS 007 is most likely significantly broader than the P v profile (Section 3.3.2 & 4.1.2), indicates that the point shown in Figure 13 for the BAL may be a lower limit, and the discrepancy between that point and the Laor & Brandt (2002) regression may be even larger. We also plot the V_max for the mini-BALs, where we show the mean of the values for the HST mini-BAL (Section 2.3) and the FUSE mini-BAL (Section 3.2). That V_max is very near the Laor & Brandt (2002) regression, indicating the outflow maximum velocity that might be expected for this relatively low-luminosity object.

Zoom In Zoom Out Reset image size

Such a large maximum velocity in a low-luminosity object such as WPVS 007 may cause problems for acceleration models. In a radiative line-driving scenario, fundamental limits can be placed by simply considering F = ma, where the force of the radiative line driving, essentially turning continuum luminosity into wind momentum, is opposed by the force of gravity due to the black hole. The F = ma equation, with the acceleration a = v(dv/dr) integrated to get the terminal velocity and converted to parameters suitable for this situation, is presented as Equation (3) of Hamann (1998; see also Hamann et al. 2001):

$$\begin{equation*} V_{\infty } \approx 9300 R_{0.1}^{-1/2} \left(\frac{f_{0.1} L_{46}}{N_{22}} - 0.1M_8\right)^{1/2}, \end{equation*}$$

where R_0.1 is the launch radius in units of 0.1 pc, f_0.1 is the fraction of the total flux incident on the flow that is absorbed or scattered by the wind, relative to 10%, L₄₆ is the luminosity in terms of 10⁴⁶ erg s⁻¹, N₂₂ is the column density relative to 10²² cm⁻², and M₈ is the black hole mass in terms of 10⁸ M_☉.

To use the Hamann (1998) equation, we need an estimate of the black hole mass. We estimate the black hole mass using standard methods using the HST spectrum. We model the region including Hβ with a linear continuum, an Fe ii template, a Lorentzian profile for broad Hβ, and two Gaussians for O iii constrained to have flux ratios of three to one, equal width, and fixed separation. We also use a Gaussian for the narrow-line region (NLR) component of Hβ, fixing the flux to be one-tenth that of [O iii] λ5007 (Cohen 1983; Leighly 1999). The [O iii] lines are slightly blueshifted with respect to the peak of Hβ, as is sometimes found in Narrow-line Seyfert 1 galaxies (e.g., Aoki et al. 2005; Bian et al. 2005), and relative to lower ionization narrow-line region lines and, therefore, we fix wavelength of the narrow component of Hβ to match that of broad Hβ. The resulting fit is very good; e.g., substituting a Gaussian for the Lorentzian to model broad Hβ yields a much worse fit. The width of Hβ is measured to be 1190 km s⁻¹. We obtain the rest-frame flux at 5100 Å from the HST spectrum (1.1 × 10⁻¹⁴ erg s⁻¹ cm⁻² Å⁻¹). We compute the BLR radius using the regressions found by Bentz et al. (2006) using the flux and a luminosity distance of 126.2 Mpc using their (inferred) cosmological parameters of H₀ = 70 km s⁻¹ Mpc⁻¹, Ω_M = 0.27, and Ω_Λ = 0.73. That is found to be log(R_BLR) = 1.062 in units of light-days. Finally, we compute the dispersion of the Hβ line profile obtained after subtracting the other fitted components, and referring to Collin et al. (2006), we use a scale factor of 1.2 to yield a black hole mass of 4.1 × 10⁶ M_☉.

We estimate the bolometric luminosity by integrating over the Cloudy continuum spectra discussed in Section 4.1.1, after they had been normalized to match the rest-frame optical spectrum (recall, as discussed in Section 2.2, the UV continuum is heavily reddened). These are given in Table 4. We obtain a rough estimate of the fraction of the continuum absorbed in the wind by integrating over the absorbed continuum and comparing with the unabsorbed value. This includes only thermal velocity for the lines, so it provides a lower limit on the fraction of the continuum able to accelerate the wind. These estimates vary from f_0.1 = 4.1 for the α_ox = −1.28 continuum, from which we infer the highest column density, to f_0.1 = 0.67 for the α_ox = −1.9, Z = 1 case and f_0.1 = 0.44 for the α_ox = −1.9, Z = 5 case, from which we require lower column densities, as discussed in Section 4.1.2. We emphasize again, though, that only lower limits on the column densities were obtained. Finally, using the opacity profile in Figure 4, we conservatively estimate a terminal velocity of about 4000 km s⁻¹; that is, we observe measurable opacity of about 6000 km s⁻¹, but the opacity has decreased significantly from the maximum by 4000 km s⁻¹. It should be noted, however, that as discussed in previously in this section, and in Section 3.3.4, the O vi absorption line may have a much higher terminal velocity, up to 12,000 km s⁻¹.

The estimate of the launch radii for the α_ox = −1.28 model, the α_ox = −1.9 model, and the Z = 5, α_ox = −1.9 model is given in Table 4. For the α_ox = −1.9, Z = 1 model, we obtain a negative radius, which means that the momentum of the absorbed continuum is not sufficient to counteract gravity, and an outflow would not be predicted. The largest outflow radius is found for the α_ox = −1.9, Z = 5 model, of R_0.1 = 0.0070, or 7 × 10⁻⁴ pc, or 2.16 × 10¹⁵ cm. A 4.1 × 10⁶ M_☉ black hole has a Schwarzschild radius of R_S = 1.2 × 10¹², implying that the launch radius is on the order of 1780R_S.

The radius of the BLR was estimated above to be log(R_BLR) = 1.062, corresponding to 3 × 10¹⁶ cm, about 14 times the inferred launch radius. This could be a problem for the radiative line-driving model, since the broad absorption-line region is thought to lie at a larger radius than the broad emission-line region. This may imply that a magnetic component of acceleration is necessary (e.g., Bottorff et al. 1997; Everett 2005). On the other hand, the radius of the observed BAL gas may be much larger than the launch radius, and the BAL simply does not intercept the line of sight until it reaches the BLR or beyond.

Variability in broad absorption lines in quasars is common. Usually, changes in apparent optical depth, coupling real changes in opacity and covering fraction, rather than changes in velocity profile are observed. Recently, more dramatic changes have been reported (see, e.g., Lundgren et al. 2007; Gibson et al. 2008, and references therein), although it is important to note that these variability studies were limited to known BALs. The BAL in WPVS 007 is distinct in several ways. First, it is one of the very few known cases in which a BAL flow appeared; other possibly similar examples have been found in the quasars TEX 1726+344 (Ma 2002) and J105400.40+034801.2 (Hamann et al. 2008). Second, as shown in Figure 13, it has quite a low luminosity for the maximum velocity of the outflow.

The development of the BAL in WPVS 007 may be associated with its low luminosity. WPVS 007 has a small black hole mass and correspondingly small central engine, emission- and absorption-line regions. For example, LBQS 1212+1445, the comparison object shown in Figure 6, has an outflow with similar maximum velocity, but with M_V = −27.6, it is 100 times more luminous and, therefore, the emission regions are expected to be 10 times larger.

Currently, we have only two epochs of UV observations of WPVS 007: one with only mini-BALs (the 1996 HST observation) and one with both mini-BALs and BALs (the 2003 FUSE observation). Thus, we can only speculate about the origin of the variability. In other variable objects, a favored explanation is generally a change in covering fraction. Thus, perhaps WPVS 007 always had a BAL outflow, but rotation of the accretion disk (assuming the BAL arises from a disk wind) may have simply rotated it into the field of view. This argument is used to explain variability in quasar-luminosity BALQSOs in Gibson et al. (2008) over 3–6 yr and, thus, it would certainly be a viable explanation for change in WPVS 007 with its lower luminosity, thus smaller size- and timescales, and longer interval between observations.

Another, perhaps more exciting, possibility is that the BAL outflow developed over the timescale of seven years. The observations were separated by 2.22 × 10⁸ s in the rest frame. If the velocity of the bulk of the outflow was 4000 km s⁻¹, then it could have covered a distance of 8.9 × 10¹⁶ cm in the interval between the observations. As discussed by Hamann et al. (2008) and references therein, the high-ionization absorption plausibly occurs just outside the corresponding broad emission-line radius. We estimate the size of the BLR in WPVS 007 to be 3 × 10¹⁶ cm and, thus, there would have been just sufficient time for the absorbing flow to cover the BLR. In an object 100 times more luminous, such as LBQS 1212+1445, and similar outflow velocity, it would have taken ten times longer, or about 70 years. So WPVS 007 may be unique in that we observed the development of a BAL. Since we can observe such extreme evolution on human timescales, WPVS 007 is an important object for understanding BAL winds physical conditions and driving mechanisms.

This argument is compelling, but cannot simply explain both the X-ray and UV variability. WPVS 007 was already X-ray weak by the time of the HST observation in which only the mini-BAL was observed. So, the emergence of the UV BAL into the line of sight by the FUSE observation could not have been directly associated with the X-ray absorption. Rather, it is possible that the X-ray absorption (perhaps a separate but associated component of "shielding gas"; Gallagher & Everett 2007) moved into the line of sight some time after the ROSAT All Sky Survey observation, and before the HST and FUSE observations. On the other hand, as shown in Section 4.1.3, the gas that produces the UV BALs can nicely explain the Swift X-ray spectrum. Since we have only two epochs of UV spectroscopic data, with no simultaneous X-ray coverage, it is difficult to determine precisely what happened.

Clearly, we need more observations to understand what is happening in WPVS 007. These are now even more urgent with the discovery from Swift that the absorption in the X-ray band is apparently changing. A second FUSE observation was approved, and even though WPVS 007, with a declination of −51 was in the region of the sky that could be observed after the loss of the reaction wheels, the observation was never scheduled before the satellite was decommissioned. An observation using HST COS has been approved, along with a contemporaneous Chandra observation and, therefore, we still have a chance of observing further absorption evolution of this interesting object.