© ESO, 2014

1. Introduction

Arp 102B, a low-ionization nuclear emission-line region (LINER) like object, was the first active galactic nuclei (AGN) where the broad line double-peaked profiles have been modeled with emission of an accretion disk (Chen et al. 1989; Chen & Halpern 1989). After that, this galaxy has been widely accepted as a prototype of an AGN with broad lines emitted from the disk (see Chen & Halpern 1989; Sulentic et al. 1990; Eracleous & Halpern 1994; Eracleous et al. 1997) and has been studied intensively through different monitoring campaigns (see, e.g., Miller & Peterson 1990; Newman et al. 1997; Sergeev et al. 2000; Gezari et al. 2007; Shapovalova et al. 2013).

In general, the monitoring campaigns agree that the broad-line region (BLR) in Arp 102B, where the double-peaked broad emission lines are forming, seems to have a disk-like geometry, but there are still some open issues and contradictions concerning the disk model (Miller & Peterson 1990; Newman et al. 1997; Sergeev et al. 2000; Gezari et al. 2007; Shapovalova et al. 2013). The broad Hα line profile variation, observed by Sergeev et al. (2000) from 1992 to 1996, corresponds to gas rotating in the disk with inhomogeneity in the surface brightness, which is in agreement with Gezari et al. (2007), who found that the accretion disk is the most favorable model. However, Gezari et al. (2007) concluded that most of the observed facts fail to explain the variability of the profiles assuming processes in the accretion disk. Additional disagreement with the disk model was reported by Miller & Peterson (1990), who showed that, at least at some epochs, the long-wavelength side of the line profile is higher than the short-wavelength side. Miller & Peterson’s finding is contrary to what is expected from a relativistic disk, i.e., an asymmetric relativistic disk cannot explain the peak ratio observed in some periods. A sinusoidal variation of the red-to-blue peak flux ratio is present in the Hα line profile (Newman et al. 1997; Shapovalova et al. 2013). This variation can be explained as a transient orbiting hot spot in the accretion disk¹.

Additionally, it is confirmed that high ionized lines, such as Lyα and CIV λ1550, do not show the disk-like profile (two peaks) and that the Lyα line is single peaked (Halpern et al. 1996), i.e., that the Lyα/Hβ ratio is less than 0.12 in the displaced peaks. The lack of double-peaked high ionization lines from the disk can indicate that the disk medium is too dense for those lines and that they cannot be emitted from the disk. However, it is interesting that in the near-infrared there is no trace of double-peaked HI lines (Riffel et al. 2006). Moreover, both the Bracket and Paschen series are almost absent apart from Paα and Paβ.

One way to confirm the disk emission in the broad lines is to observe polarization in broad lines (see, e.g., Corbett et al. 1998, 2000; Afanasiev et al. 2014), since the expected position angle of polarization should be parallel to the disk plane. This angle is the same as the direction of the radio jet (Antonucci et al. 1996; Corbett et al. 1998, 2000), which is not expected for the broad lines emitted from an accretion disk. Also, Chen et al. (1997) found discrepancies between the disk model and the spectropolarimetric data, but they found that if some specific conditions are taken into account, the observations could be explained with emission of the accretion disk.

Although the double-peaked broad lines of Arp 102B indicate an accretion disk emission, there are some deviations in broad line profile variations that may indicate another geometry. In Paper I (Shapovalova et al. 2013), we present observations and variation in line fluxes and the continuum. In this paper, we discuss the long-term spectral line profile variations in order to clarify the nature of the BLR in Arp 102B. The paper is organized as follows: in Sect. 2, we describe our observations and used methods of analysis; in Sect. 3, we present our results of the line profile variation analysis; in Sect. 4, we discuss our results; and finally, in Sect. 5, we outline our conclusions.

2. Observations and methods of analysis

2.1. Observations

Spectra of Arp 102B were taken during the monitoring period 1987–2010 with five different telescopes: two telescopes in Russia (the 6 m BTA and 1 m Zeiss telescopes of Special Astrophysical Observatory (SAO), Russian Academy of Science), two 2.1 m telescopes in Mexico (Guillermo Haro Observatory, GHO, and Observatorio Astronomico Nacional at San Pedro Martir, SPM), and two telescopes in Spain (3.5 m and 2.2 m telescopes of Calar Alto Observatory, CA). Details of observations and data reductions are given in Paper I, and will not be repeated here.

In addition to the observational data presented in Paper I, we added ten new spectra observed in the 2012–2013 period with 2.1 m telescope of GHO. The measurements of the continuum at 5100 Å, Hα, and Hβ fluxes are listed in Table 1. In Fig. 1, we plot the light curves in the continuum and in the broad lines to compare our new observations with those obtained in Paper I. As can be seen from Fig. 1, the minimum in the line flux is observed in 2012–2013, while the continuum flux is similar to that observed in the 1989–1992 period.

2.2. Method of analysis

To explore the broad line profile variability in detail, we performed the following method of analysis:

• a)

  To subtract the narrow lines and obtain only the broad profiles, we fitted all Hα and Hβ lines with Gaussian functions. There is a problem with removing the narrow lines in Arp 102B, since they are right on top of the red peak in both broad lines (Hα and Hβ) and the measured properties of the red peak can be strongly affected by the narrow line removal procedure. We discuss this in more detail in the Appendix, and here we briefly describe the subtraction procedure. To estimate the narrow line contribution we fitted the entire Hα and Hβ profiles (broad + narrow lines) simultaneously. We used three broad Gaussian functions for the broad component of the Hα and Hβ lines (Fig. 2), while we assumed the same widths and shifts for all narrow lines (see Popović et al. 2004). We assumed that the ratio of [OIII]4959 and [OIII]5007 follows the flux ratio 1:3 (Dimitrijević et al. 2007), and made the same assumption for [NII] doublet: F[NII]6584/F[NII]6548 ~ 3. In the Hα wavelength band, the [SII]6717,6731 doublet, and [OI]6300 and [OI]6363 lines were fitted using single Gaussian. Note here that only one Gaussian cannot properly fit the narrow line wings, therefore, we included an additional, broader, Gaussian (with significantly smaller intensity) to fit the narrow line wings. As can be seen in Fig. 2, the three broad components (red and blue shifted, and central component) can satisfactorily explain the broad line profiles. To check the narrow line subtraction we perform several tests, see Appendix A for details.

• b)

  After subtraction of the narrow lines and continuum, we made month-averaged line profiles using similar profiles (same peak positions and total line flux differs up to ~10%). These month-averaged line profiles are given for both Hα and Hβ line in Figs. 3 and 4, and we continued to measure and analyze properties of these profiles.

• c.

  To determine the peak positions in the Hα and Hβ broad line profiles, we performed another Gaussian analysis, this time on the month-averaged broad profile. The broad profile was fitted with three Gaussian functions, corresponding to the blue, central, and red component (Fig. 5). From the inspection of the parameters obtained from the Gaussian fittings, we note that the central component very often changes its position, which affects other parameters, especially the intensity of the red peak. Therefore, we also fitted the month-averaged broad line profiles using the constraint that the central component should not have a large shift velocity, i.e., setting the limits on the shift of the central component between –1000 km s^-1 and 600 km s^-1 (see Figs. 5 and 6, and the discussion in Appendix A). The obtained error-bars of the parameters from both fittings are often smaller than the difference between the parameters obtained from these two fitting procedures, therefore we accepted the averaged value of the two best-fittings for the line parameters. For their error-bars, however, we took the corresponding standard deviation, i.e., the discrepancy between the averaged parameters and those obtained from the fits. The averaged parameters, with their uncertainties, of the Gaussian fittings of the broad Hα and Hβ components are listed in Tables 2 and 3.

  Fig. 3 Month-averaged profiles of the Hα and Hβ broad emission lines in the period between 1987–2004. The abscissa (OX) shows radial velocities with respect to the narrow component of the Hα or Hβ line. The ordinate (OY) shows the flux in units of 10-16 erg cm-2 s-1 Å-1.

  Fig. 4 Same as in Fig. 3, except for the period between 2004–2013.

• d)

  The broad double-peaked line profiles are usually fitted with some sort of disk model (see, e.g., Eracleous & Halpern 1994; Gezari et al. 2007; Newman et al. 1997; Popović et al. 2011, etc.). Here we use a relativistic disk model given by Chen et al. (1989) and Chen & Halpern (1989) to model the broad Hα profile. In Fig. 7, we present a disk fitting with our parameters and those given in Gezari et al. (2007). We will discuss this in more details in Sect. 3.2.

• e)

  Finally, we measured the full width at half maximum (FWHM) and full width at quarter (1/4) maximum (FWQM). These measurements are also provided in Tables 2 and 3.

3. The broad line profile variability

In Paper I, we measured and analyzed variations in the continuum and line fluxes using total of 118 spectra covering the Hβ wavelength region, and 90 spectra covering the Hα line. We showed that the fluxes in the Hα and Hβ lines and in continuum did not strongly vary (around 20% in line and ~30% in continuum), and that several flare-like events were observed during the monitored period (1987–2010). Considering spectra from new observations, it seems that there is a minimum in the line emission in the 2012–2013 period (see Fig. 1), and also Fig. 4 reveals that both lines are weak, moreover, Hβ is very weak with two weak peaks comparable to noise.

During the monitored period, the broad Hα and Hβ lines have double-peaked profiles (even in the minimum), but, as shown in Figs. 3 and 4, their line shapes and widths are different. Comparing the month-averaged spectra of the Hα and Hβ broad lines (see Figs. 3 and 4), one can note that there is a difference between the peak ratio of Hα and Hβ lines. The Hα line mostly has a stronger blue peak, while very often the Hβ line has almost the same intensities of the blue and red peak. On the other side, the Hβ FWHM and FWQM (FWHM_mean = 16 100 ± 700 km s^-1, FWQM_mean = 17 900 ± 700 km s^-1) are significantly broader than in the Hα line (FWHM_mean = 14 400 ± 400 km s^-1, FWQM_mean = 16 600 ± 800 km s^-1). This contradicts the disk model predictions, i.e., if Hβ originates closer to the central black hole (as indicated by larger line widths), it is expected that it has a more pronounced blue peak than Hα.

On average the line profiles of the Hα and Hβ lines do not change significantly besides some flare-like variations seen in short periods (e.g., as in the period 1987–1992, 1998, and 2012–2013). In order to test the line profile changes, we normalized the blue peak of the month-averaged Hα profiles to unity and calculated the mean and root mean square (rms) profiles (Fig. 8). The normalized Hα rms profile shows that there is practically no change in the line profile. The change is below 10% in the red wing, where some variations may be due to uncertainty of the narrow line subtraction. Thus, considering the whole 26-year period, the averaged Hα line profile variation is practically neglected.

3.1. Gaussian analysis: changes in the broad line profiles

The parameters from the Gaussian fittings are listed in Tables 2 and 3. As shown from Tables 2 and 3, the Hα and Hβ blue-peak positions change during the observed period for about ±500 km s^-1 around the averaged values (see Fig. 9 and Table 2), while the red-peak positions change for ±~1000 km s^-1 around averaged values, which may also be due to the measurement uncertainties.

In general, the Hα profiles are of much better quality than Hβ, thus we plotted in Fig. 10 the results of the peak-measurements of the Hα line: the blue- vs. red-peak intensity (upper panel), and the blue- vs. red-peak velocity (bottom panel). It is interesting that there is no correlation between the velocities of the red and blue peaks (see Fig. 10, bottom panel): there may be a trend toward an anticorrelation, which is statistically not significant (r = −0.28, P₀ = 0.08), and it seems that they vary independently of each other. There is a good correlation, however, between the two peak intensities (see Fig. 10, upper panel), and the blue- and red-peak intensities are responding to each other.

In order to compare the position of the blue peak from our monitoring campaign with previous campaigns, we plot in Fig. 11 (upper panel) our measurements of the blue-peak velocity for the Hα line together with the measurements of Newman et al. (1997) and Gezari et al. (2007). As shown in Fig. 11, the agreement between different measurements is good, and the differences are within the error-bars (±100–200 km s^-1). In Fig. 11 (bottom panel), we compared the blue- and red-peak positions between Hα and Hβ during monitored period. As shown in Fig. 11 (bottom panel), there is a discrepancy between the positions of the red peaks of Hα and Hβ (in some periods for about 2000 km s^-1), while the blue-peak position in both lines has no big change. There are small differences between the Hα and Hβ blue-peak positions, which are below ~500 km s^-1 (see also Tables 2 and 3).

In Fig. 12, we plot the Hα vs. Hβ peak velocity for the central (upper), blue (middle), and red (bottom) peak of Gaussians from the month-averaged profile fits. As shown in Fig. 12, there are no expected correlations, however, it seems that a trend of anticorrelation is present in the red Hα vs. Hβ velocities (solid line on the plot), while a correlation trend may be present between the blue-peak velocities of Hα and Hβ. We should note that a trend of anticorrelation between the red-peak position may be caused by the measurement uncertainties, but that is not the case for the blue-peak positions. A theoretical expectation is that the velocity of the blue peak of the Hα line follows that in the Hβ line, and, as shown in Fig. 12 (middle) there may be a slight dependence between the blue-peak velocities of Hα and Hβ, but it seems to be very weak (even taking the uncertainties in the measurements shown in the plots into account). This confirms that there are no expected correlations between the peak velocities of two broad lines.

3.2. Disk model of the broad emission lines

As we noted above, the relativistic disk model (Chen et al. 1989; Chen & Halpern 1989; Gezari et al. 2007) can explain the broad double-peaked profiles of Arp 102B. We performed several fits, and it was interesting that the month-averaged profiles can be successfully fitted with the simple Chen et al. (1989) disk model. The problem is that in some periods we could only fit the line profiles if we changed the inclination of the disk about 10 degrees, which is not expected. Note here that in the case when the inclination was fixed we obtained the satisfied fit by varying the other parameters (e.g., the emissivity and inner radius). We will not present all fits here, we only present examples of the fits with parameters given by Gezari et al. (2007): the inner radius R_inn ~ 325 R_g, the outer radius R_out ~ 825 R_g, the disk inclination i ~ 31deg, the random velocity in the disk σ = 1050 km s^-1, with the emissivity r^−q, where q = 3.0 is assumed (see Figs. 7 and 13). Additionally, starting from the parameters given above, we find the best fit of the mean Hα profile, by changing the inclination to i = 23deg, and shifting the entire line for –900 km s^-1. As an example, these two fits are compared with observations from two periods for the Hα and Hβ line in Fig. 7 (bottom panels). The agreement between the observed blue peak and both models is good, but we see differences in the red wing (Fig. 7, bottom panels). We emphasize again, however, that the narrow line subtraction may contribute to the uncertainty in the red-peak properties, which can thus affect the obtained disk parameters.

If we consider the mean profile during the whole monitored period, one can conclude that the disk inclination is around i = 23deg, while the disk dimensions are around 500R_g. There is agreement between the disk parameters from our fit of the mean Hα profile and that given in Gezari et al. (2007), apart from the disk inclination. However, by varying other parameters we were able to obtain i ~ 30deg, but in all fittings the disk dimensions seem to be very compact (about several 100 s gravitational radii). In Fig. 13, we plotted both models together with the normalized mean Hα profile.

We conclude the line shape (i.e., double-peaked profiles) can be mainly fitted with a simple relativistic disk model, however, to fit the same line from different periods, one should change the disk parameters, such as inclination or emissivity. This, in principle is not expected, but we should point out that the obtained parameters from the fit are strongly dependent on the positions and intensities of the blue and red peaks, and as we noted above, there are relatively big uncertainties in the red-peak properties.

4. Discussion

Double-peaked broad emission lines signature expected from the disk-like BLR are observed in the spectrum of Arp 102B (Chen & Halpern 1989; Eracleous & Halpern 1994; Sergeev et al. 2000; Gezari et al. 2007; Shapovalova et al. 2013). The double-peaked lines are presented in a small fraction of AGNs (see Strateva et al. 2003). The double-peaked line profiles have been explained with a number of different models: a) binary black holes (see, e.g., Gaskell 1983); b) bipolar outflows (see, e.g., Zheng et al. 1991); c) anisotropically illuminated spherical distribution of optically thick clouds (Goad & Wanders 1996); d) circular or elliptical accretion disk (see, e.g., Chen et al. 1989; Chen & Halpern 1989; Eracleous et al. 1995, 1996), and e) other more sophisticated models like precessing elliptical disk or a circular disk with a long-lived, single-armed spiral or warp (see, e.g., Gezari et al. 2007; Jovanović et al. 2010).

Here we can exclude the binary black hole (see, e.g., Gaskell 1983) model, since there is no significant change in the position of the line peaks. Additionally, the model of an anisotropically illuminated spherical distribution of optically thick clouds (Goad & Wanders 1996) does not explain the big distance between the line peaks (more than 10 000 km s^-1), therefore here we will consider the disk-like geometry of the BLR and the possibility of outflows in the BLR.

4.1. The BLR disk emission: pros and cons

As we noted above, the disk emission has usually been assumed to model double-peaked profiles of Arp 102B, however, some observational facts from monitoring of the broad lines contradicts the disk model. Recall here some of the observational facts (see Miller & Peterson 1990; Gezari et al. 2004, 2007):

• Flares in the broad line flux that can also be seen in our observations (see Paper I), but this can be expected if there are some transient processes in the disk (see, e.g., Jovanović et al. 2010).

• Systematic variations in FWQM reported in Gezari et al. (2004), which also may be explained in the disk structure variation.

• Periodic oscillation of the red-to-blue wing flux ratio observed by Newman et al. (1997) and Shapovalova et al. (2013) can also be caused by rotating structure in the disk.

• Intensity of the red peak is sometimes higher than the blue peak (see Figs. 3 and 4), which can also be explained by perturbations in the disk structure (see, e.g., Jovanović et al. 2010; Popović et al. 2011) as well as by a model of precessing elliptical disk (Gezari et al. 2007).

As one can see, some observed facts mentioned above can be explained by a more complex and nonstable configuration of the accretion disk, i.e., that emissivity enhancements such as transient shocks induced by tidal perturbations could be present, and that the change in line profile can be reproduced by changing the inner and outer radii of the line-emitting portion of the disk.

However, there are some other observational facts that are in disagreement with the disk model hypothesis:

• a)

  A lack of (or slightly anti-) correlations of the blue-to-red intensity ratio with FWQM, as shown in Fig. 14. In principle, one can expect that in the relativistic disk the ratio between the blue and red peak should increase with increasing of the FWQM of the line, since there should be no change in the inclination. The intensive blue peak indicates that the inner radius is closer to the central black hole. This should result in a more extensive (and less intensive) red part, contributing to the broader line measured at FWQM. One would expect the observed tendency in I_blue/I_red vs. FWQM should be correlated, but it seems they are anticorrelated (see Fig. 14). However, Lewis et al. (2010) demonstrated that in the case of a model of the non-axisymmetric disk, in some cases the red-to-blue peak ratio increases as the width of the profile increases. Also, note here that some indication of such anticorrelation in other double peaked AGNs can be seen in Lewis et al. (2010). They showed plots of red-to-blue peak ratio as a function of time and from a quick inspection one can see a few cases where the changes do not follow the expectation that the FWQM is decreasing as the red-to-blue peak ratio increases (or the blue-to-red peak ratio decreases).

• b)

  The distance between the position of the red and blue peak is different for the Hα (~10 500 km s^-1) and Hβ (~12 000 km s^-1), i.e., the Hβ shows larger distance for about 1500 km s^-1 that in the frame of the disk hypothesis should indicate that the line is closer to the central black hole, and consequently that relativistic effects are more observable in the Hβ line profile. As we see in Figs. 3 and 4, however, the blue boosted peak is more intensive in Hα, for almost the entire monitored period. On the other hand, such huge distances between the peaks indicate a fast rotating disk, that is probably close to the black hole. From the rms profile (see Paper I, Fig. 11) one can see that the change in the line profile is also double-peaked, i.e., there is a change in the broad line profile in red and blue peak in both lines, where the variability in the blue wing is significantly bigger than in the red wing. There is one central peak (more intensive in the Hβ line) in the rms profile that may be caused by a central component (also see Fig. 11 from Paper I).

  Fig. 14 Blue-to-red peak intensity ratio as a function of FWQM for Hα (upper) and Hβ (bottom). The correlation coefficients are r = −0.55 (P0 = 0.2E-03) and r = −0.14 (P0 = 0.4) for Hα and Hβ respectively. The best fit parameters are given on the plots. The measurements for the Hα of Gezari et al. (2007) are given on upper plot (open triangles).

• c)

  Additionally, it is confusing that there is no big change in the Hα normalized line profile during this long monitored period (see Figs. 3 and 8), as well as in the displacement of the peaks. If the emission is originating in the disk, the disk emission is very stable during the time, i.e., there is no big change in the disk that affects the emission. On the contrary, the different peak intensity ratio (e.g., in some periods the red peak of Hβ is more intensive than the blue peak) indicates that long-living changes in the disk structure should be present (see, e.g., Jovanović et al. 2010; Popović et al. 2011). These changes should affect a compact disk (~500 gravitational radii, that is several light days), and consequently a stronger variability in the broad line profiles is expected.

• d)

  Finally, the spectropolarimetric observations are also inconsistent with the BLR disk model. Spectropolarimetric observations (see Antonucci et al. 1996; Corbett et al. 1998, 2000) did not confirm the disk-like structure of the BLR in Arp 102B. It specially, is interesting that the angle of the jet direction (~105 degrees) corresponds to the polarization angle in the Hα line (~103 degrees, see Corbett et al. 1998). The Hα polarized shape of Arp 102B indicates a line-emitting biconical outflow surrounded by a cylindrical scattering region (Corbett et al. 1998, 2000). However, there are two facts that prevent us from excluding the disk model. First, that Chen et al. (1997) obtained a qualitative agreement between the disk model, but other possible explanations of the broad-line polarization properties of Arp 102B are possible, e.g., that to the disk emission there is additional outflow emission (see Fig. 15). Second, as it was mentioned in Afanasiev et al. (2014), the role of the interstellar polarization in the polarization properties of the broad lines is very important and after taking it into account, the polarization angle may be changed.

4.2. Outflow(s) in the Arp 102B BLR

Recently, Couto et al. (2013) found a two-armed nuclear spiral (see also Fathi et al. 2011) in an extensive field around the nucleus of Arp 102B (2.5 × 1.7 kpc²) and they estimated that the mass outflow rate along one (the east) arm is between 0.26–0.32 M_⊙ yr^-1, which is higher than the mass accretion rate. Therefore, one can expect that an outflow is present in the BLR. As shown in Fig. 15, during some periods, there is a bump in the blue part of the Hα line (around –2500 km s^-1), which may indicate that in addition to the disk emission, there is an outflow in the BLR.

In principle, a biconical outflow model can explain double-peaked line profiles (as, e.g., Balmer line profiles of 3C 390.3, see Zheng et al. 1991). The blue-shifted and red-shifted peaks are produced by the approaching and receding parts, respectively. Additionally, a model where an accelerating outflow together with an inflow of the emitting gas is dominating in the BLR can explain very complex line profiles (also double-peaked Ilić et al. 2010; León-Tavares et al., in prep.).

Also, spectropolarimetric observations seem to be in favor of the some type of a biconical model (see Corbett et al. 1998, 2000, and corresponding scheme).

As a summary, one can note that the outflow is probably present in the BLR, and seen as a blue bump between the center of the line and blue peak (around –2500 km s^-1, see Fig. 15), but it is not clear if the emission from the outflow is dominant in the broad line profiles. Also, one possible problem is relatively small changes in the broad line profiles, which probably should be highre in the case of outflows.

5. Conclusion

We analyzed the variability in the broad line properties in a long period of 26 years of Arp 102B, an AGN with prominent double-peaked broad line profiles. We investigated the broad line profile variations during this period and from our investigations we outline the following conclusions:

• a)

  The broad line profiles have not been significantly changed in a period of 26 years in the sense that during the monitored time, the shift of peaks stays almost unchanged and there are flare-like changes in the line intensity, and some time in the ratio of the blue-to-red peak intensity. This is unusual if there is an emission of a relatively compact accretion disk (dimension smaller than 1000 Rg), since changes in the intensity ratio of peaks should result in the line widths as well as in the displacement of the peak velocities. Moreover, the intensity ratio of peaks shows a trend of an anticorrelation with the line width, which is opposite from what one can expect from the disk emission. Sometimes, the position of the red peak does not match the accretion disk model predictions (Fig. 6).

• b)

  As an additional conclusion to our earlier one that the variation of fluxes of Hβ vs. Hα has a small correlation (see Shapovalova et al. 2013), there is practically no correlation between the velocities of the blue and red peaks between the Hβ and Hα line. Moreover, the line profiles of Hβ and Hα are different. The Hα line profile has more intensive blue peak, and Hβ has almost the equivalent intensity of peaks in the most of observations. Also, it is unexpected that the Hβ is significantly broader (~1500 km s^-1) than Hα, and that the relativistic effect of the blue-boosting is more prominent in the case of Hα.

• c)

  An outflow in the BLR seems to be present (from time to time it is seen as the blue bump around –2500 km s^-1). Although it is not clear how much the outflow can contribute to the broad line emission, it should be taken into account.

At the end, let us conclude that profiles of the broad double-peaked lines in Arp 102B can be fitted very well with the disk model, but there are several issues in the variation of the broad spectral line properties that are not in agreement with those expected in the variability of the accretion disk structure. The better spectropolarimetric observations/monitoring of Arp 102B are needed to study the polarization properties of the ordinary broad line component in order to clarify the nature of the BLR.

Online material

Appendix A: The line fitting procedure and narrow line subtraction

One of the problems in the measurements of the broad line parameters is the uncertainty in the narrow line subtraction, especially in the red peak of Hα and Hβ, since the narrow lines are right on top of the red peak in both broad lines. Therefore, to find uncertainties in the narrow line subtraction we performed several tests. Additionally, we considered the B-band absorption observed near the [SII] doublet in the Hα wavelength range.

Appendix A.1: Correction of the B-band absorption near [SII]

The B-band absorption is present in the red wing of the Hα line, near the narrow [SII] doublet (see Fig. A.1). To correct this absorption we used the template spectrum of NGC 4339, which was observed at the same night as Arp 102B with 2.1 m GHO telescope with resolution ~8–9 Å on Mar 26, 2003. Then we corrected the B-band absorption near the [SII] lines in the Hα spectral region.

In Fig. A.1, we illustrate the correction of the B-band absorption. As one can see in Fig. A.1 the absorption is well corrected, but in some spectra the residuals are still present. However, the residuals are weak and cannot affect the narrow line estimates, or the estimates of the red-peak position in Hα (see following section and Figs. A.2). Therefore, we accepted the parameters from the fits where the B-band absorption was not corrected.

Appendix A.2: Estimation of the narrow line emission

As we mentioned above, the uncertainties in the estimations of the broad line parameters come mainly from the subtraction of the narrow lines in the Hβ and Hα spectral range. It is hard to handle the removal of the narrow emission lines in Arp 102B, because the narrow lines are right on top of the red peak in the Hα and Hβ broad lines. We used two approaches to explore the uncertainty in the narrow line subtractions: i) fitting the narrow lines with Gaussian functions before and after the B-band absorption was corrected (as described in Sect. 2.2); and ii) estimation of the broad profile using spline fitting (using DIPSO). In Figs. A.2 and A.3, we illustrated the subtraction of the narrow lines using these two methods. In Fig. A.2 (fourth panel), we compared the broad Hα line profile obtained after the subtraction of the narrow lines using these two procedures (note that we fitted the spectra before and after the correction for the B-band absorption). As shown in Fig. A.3 (bottom panel), the fitted position of the blue and red peaks is also practically the same for Hβ for both procedures, therefore we used the Gaussian decomposition in our estimates of the narrow line contribution.

Appendix A.3: Correction of the fits using the ratio of the narrow lines

As we mentioned above, some parameters of the narrow lines are fixed in the spectral fit: first of all, the Gaussian velocity dispersions and shift for all narrow lines in the line wavelength range, as well as the ratio of the [OIII]4959,5007, which is fixed to 1:3 (Dimitrijević et al. 2007). Also, one cannot expect that the ratio of narrow lines changes over the period of several years. Therefore, one very clear way to test the robustness of the narrow line fittings is to compute the emission line ratios for each combined fit.

The timescale of the observations is long enough that there could be some variation in the narrow emission line ratios, but certainly over timescales of ~5–10 years there should not be major changes in the narrow line ratios (except of the random scatter). We calculated the narrow line ratios of [NII]/Hα, [OI]6300/Hα, [SII]6715/Hα, and [OIII]5007/Hβ. The line ratios in the Hα and Hβ wavelength range are presented in Fig. A.4, where dashed lines represent the averaged ratio and shaded regions the ±10% of the averaged value. As shown in the upper panels of Fig. A.4, there is not any trend in the line ratio variability during the monitored period. Consequently, we corrected those fits for which the narrow line ratios were significantly different from the mean value, i.e., the spectra with a big narrow line ratio difference have been refitted taking the constraint in the fit that the line ratio has to be in the frame of ±15% of the corresponding mean value. In the bottom panels of Fig. A.4, the line ratios are shown after the correction of a number of spectra.

Appendix A.4: The broad line fit and estimates of parameter uncertainties

To find the broad line parameters of Hα and Hβ broad lines, we performed a Gaussian analysis, fitting the month-averaged broad line profiles. The broad profile was fitted with three Gaussian functions, corresponding to the blue, central, and red component (Fig. 5). Each Gaussian was described with three free parameters, i.e., in the fit procedure we have nine free parameters: three intensities, widths and shifts of Gaussian functions that describe the blue, central, and red part of the broad line.

Inspection of the obtained parameters from the fit, as well as several additional tests (changing slightly parameters), showed that the central component is often shifted to the blue (especially in Hα) and that, in some cases, this component has a big (unexpected) change in the shift. Additionally, we found that the position of the central component can significantly affect parameters of the peaks, especially the red peak. Therefore, we repeated the fitting procedure with three broad Gaussian functions, but putting the limits on the shift of the central component to be between –1000 km s^-1 and 600 km s^-1. Using this procedure we obtained an additional set of broad line parameters (parameters of the blue, central, and red Gaussians). Inspection of differences between the parameters obtained from the fits with and without constraint of the central component shift showed that the peak positions are not significantly changed, but the intensities of the red peak, the shift of the central component, and the widths of the components have been significantly changed.

We compared the error-bars from the fits with differences between the parameters from these two fitting procedures, and found that the error-bars of parameters are often significantly smaller than the differences between the parameters from both fits. Therefore, we calculated the averaged parameters

from the two fits, and accepted uncertainties (error-bars) as the discrepancy between the parameters from the two fits. The averaged broad line parameters and corresponding estimated uncertainties are given in Tables 2 (for Hα) and 3 (for Hβ).

---

Acknowledgments

This work was supported by INTAS (grant N96-0328), RFBR (grants N97-02-17625 N00-02-16272, N03-02-17123, 06-02-16843, N09-02-01136, 12-02-00857a, 12-02-01237a), CONACYT research grants 39560, 54480, and 151494 (Mexico), PROMEP/103.5/08/4722 grant, the Ministry of Education and Science of Republic of Serbia through the project Astrophysical Spectroscopy of Extragalactic Objects (176001), and DFG grant Ko 857/32-1. L.Č.P., W.K. and D.I. are grateful to the Alexander von Humboldt foundation for support in the framework of program “Research Group Linkage”. D.I. has been awarded L’Oréal-UNESCO “For Women in Science” National Fellowship for 2014. We would like to thank the anonymous referee for very useful comments.

References

All Tables

All Figures