A model for the Balmer pseudocontinuum in spectra of type 1 AGNs
aa r X i v : . [ a s t r o - ph . C O ] N ov A model for the Balmer pseudocontinuum in spectra oftype 1 AGNs
Jelena Kovaˇcevi´c ∗ , Luka ˇC. Popovi´c Astronomical Observatory, Volgina 7, 11060 Belgrade, Serbia and Wolfram Kollatschny
Institut f¨or Astrophysik, Universit¨at G¨ottingen, Friedrich-Hund Platz 1, 37077,G¨ottingen, Germany
Abstract
Here we present a new method for subtracting the Balmer pseudocontinuumin the UV part of type 1 AGN spectra. We calculate the intensity of theBalmer pseudocontinuum using the prominent Balmer lines in AGN spectra.We apply the model on a sample of 293 type 1 AGNs from SDSS database,and found that our model of Balmer pseudocontinuum + power law con-tinuum very well fits the majority of the AGN spectra from the sample,while in ∼
15% of AGNs, the model fits reasonable the UV continuum, buta discrepancy between the observed and fitted spectra is noted. Some of thepossible reasons for the discrepancy may be a different value for the opticaldepth in these spectra than used in our model or the influence of the intrinsicreddening.
Keywords:
Galaxies: active, quasars: emission lines
1. Introduction
One of the interesting features in spectral energy distribution of AGNstype 1 is so-called the λ ∗ Jelena Kovaˇcevi´c
Email addresses: [email protected], [email protected] (Luka ˇC. Popovi´c ), [email protected] (and Wolfram Kollatschny)
Preprint submitted to Advances in Space Research August 4, 2018 almer lines and Balmer continuum emission. Namely, as the number of up-per level of the Balmer lines increases, levels become more and more dense,that results in overlapping of high order Balmer lines. Blended Balmer linesturn into the Balmer continuum at the Balmer edge ( λ λ λ λ e and density – n e ) gives very different results for the Balmercontinuum intensity (for detailed review see Jin et al., 2012, and referencestherein). Some authors try to quantify the ratio between the Balmer contin-uum intensity and some strong Balmer lines (H β or H α ) for different physicalparameters, and they found a large range in flux ratios (e.g. see Kwan andKrolik, 1981; Hubbard and Puetter, 1985). Also, the observed range of theseratios are very large (see Wills et al., 1985).The accurate determination and subtraction of the Balmer continuum isa very difficult task because of the large number of free parameters. Thismakes analysis of the UV spectrum very uncertain, as eg. investigation ofthe spectral energy distribution or calculating of the black hole mass usingthe continuum luminosity in the UV range. One of the free parameters inall previous Balmer continuum models was the Balmer continuum intensity,which determination depends on the fit of the power law and numerous ofUV Fe II lines, as well.In this paper, we try to find the best model to calculate the intensity ofthe Balmer continuum, using only the prominent Balmer lines in spectra.2n this way, we try to eliminate the intensity of the Balmer continuum asthe free parameter in the fitting procedure, and to get a simplified and lessuncertain estimation of the Balmer continuum.
2. The Balmer continuum model
To estimate the Balmer continuum, we assume partially optically thickclouds with a uniform temperature. We use the Balmer continuum functiongiven in Grandi (1982) for λ < e =15 000 K and optical depth at the Balmer edge fixed to be: τ BC =1, as it is estimated in Kurk et al. (2007).The function of the Balmer continuum for the case of optically thick cloudsis given in Grandi (1982) as: F G ( λ ) = F BaC × B λ ( T e )(1 − e − τ λ ) , λ BaC is the estimate of the Balmer continuum flux et the Balmer edge,B λ (T e ) is the Planck function at the electron temperature T e , τ λ is the opticaldepth at λ , which is expressed as: τ λ = τ BE (cid:0) λ BE λ (cid:1) − , where τ BE is the opticaldepth at the Balmer edge λ BE =3646 ˚A.We assume as Wills et al. (1985) and Dietrich et al. (2003), that at wave-lengths λ > λ = 3646 ˚A) is equal to the sum of intensities of all high orderBalmer lines at the same wavelength (see Wills et al., 1985; Dietrich et al.,2003), to calculate the Balmer continuum intensity.The high order Balmer lines with n >
5, are arising very close to eachother. Since they are broad (in AGN type 1 spectra) there are overlapping,and practically forming the continual emission blueward the H ε , giving inthis way the smooth rise to the Balmer edge. Note that Wills et al. (1985)used 70 and Dietrich et al. (2003) used 50 high order Balmer lines to explainsmooth Balmer edge, but they did not use them to estimate the Balmercontinuum intensity. We found that this number of high order Balmer linesis not sufficient since the slope they form starts to decreases before it reachesthe Balmer edge, so at λ = 3646 ˚A it does not represent the intensity of3he Balmer edge. For that reason, we adopt up to n=400 high order Balmerlines, with central wavelengths less than λ = 3645.1593 ˚A.In order to determine the sum of high order Balmer lines at the Balmeredge, first we need to have the clean broad profiles of prominent Balmer lines(H β , H γ and H δ ), without any narrow components or some contaminationlines. After we obtain the clean profiles, we fit these lines and obtain theintensities for high order Balmer lines, since all Balmer lines are connectedwith fixed relative intensities.We fit each Balmer line from H β to n=400, with one Gaussian, wherethe widths and shifts of each Gaussian are the same. For the Balmer linesfrom the level 1 < n <
50 we obtain the relative intensities given by Storeyand Hummer (1995), for the T=15000 K, n e =10 , case B. For the rest ofthe Balmer lines (51 < n I I = b ( T, N e ) b ( T, N e ) ( λ λ ) f f · g g · e − ( E − E /kT (2)where I and I are the intensities of lines with the same lower term, b ( T, N e )and b ( T, N e ) represent deviation from thermodynamic equilibrium, λ and λ are the wavelengths of the transition, g and g are the statistical weightsfor the upper energy levels, f and f are the oscillator strengths, E and E are the energies of the upper levels of transitions, k is the Boltzmannconstant, and T is the excitation temperature.Assuming that: b ( T, N e ) b ( T, N e ) ( λ λ ) f f · g g ≈ I I ≈ e − ( E − E /kT (4)The Eq 3 and Eq 4 are following the principal thermodynamic equilibrium,i.e. that population of higher levels in the Balmer series is leaded by electrontemperature, and that the excitation temperature is similar by electron one(Popovi´c , 2003; Ili´c et al., 2012). In principle, in BLR plasma one can notexpect that partial thermodynamic equilibrium is present, especially in thelow excitation levels (except in some cases, see Ili´c et al., 2012). However,going to higher level in the series, one can expect that the population of the4evels depends very strongly from the T e , and the assumptions in Eq 3 andEq 4 can be applied.After we calculate relative intensities of all Balmer lines which flux con-tributes to the Balmer edge, we use their sum for the parameter calculationof the Balmer continuum intensity at the Balmer edge.We assume that at λ =3646 ˚A function given in Grandi (1982) (see Eq 1)is equal to the sum of all high order Balmer lines at the same wavelength, soit will be: F G (3646) = F BaC × B ( T e )(1 − e − τ ) = X i =6 G i (3646) (5)where G i is the Gaussian function which describes the Balmer line with upperlevel n=i: G i ( λ ) = I i × e − ( λ − λi − d × λiWD ) (6)where I i is the relative intensity, λ i the central wavelength, d is the shift ofthe Gaussian relative to the central wavelength ( d = ∆ λλ i ) and W D Dopplerwidth of the Gaussian. The values d and W D are the same for all Gaussians.The parameter of the Balmer continuum intensity F BaC may be found as: F BaC = P i =6 G i (3646) B ( T e )(1 − e − τ ) . (7)Finally, the function which describes our model is: F ( λ ) = P i =2 G i ( λ ) λ > A P i =6 G i (3646) B ( T e )(1 − e − τ ) × B λ ( T e )(1 − e − τ λ ) λ A We fit simultaneously the power law as: F pl = F ( λ α , and the Balmer continuum (high order Balmer lines). We have four freeparameters: the exponent of the power law ( α ), and the width, shift and5ntensity of the one prominent Balmer line (for example H β ). The parametersof a broad Balmer line are obtained from the best fit. After we have thewidth, shift and intensity of one Balmer line, the intensities of all others aredetermined using relative intensities from Storey and Hummer (1995) andEq 4, as well as the intensity of the Balmer continuum using Eq 7.It is very important to have a clean profile of strong broad Balmer lines(H β , H γ and H δ ) without any contamination with narrow component orother emission lines (numerous Fe II lines and [O III] lines which overlap withBalmer lines). In order to get the clear broad profiles we perform the fittingprocedure described in Kovaˇcevi´c et al. (2010) and Popovi´c et al. (2013) atthe spectral range 4000-5500 ˚A. First, we have to remove the optical part ofthe continuum using the continuum windows given in Kuraszkiewicz et al.(2002) in order to obtain only emission lines for fitting. The used continuumwindows are: 3010-3040 ˚A, 3240-3270 ˚A, 3790-3810 ˚A, 4210-4230 ˚A, 5080-5100 ˚A. It is assumed that all narrow lines have the same widths and shifts,since these lines are coming from the same narrow emission line region. Inthis way, we use of the prominent narrow [O III] lines, to fix the width andthe shift for all narrow components of the Balmer lines, which may be weak insome cases, or hardly distinguishable from the broad component of the lines.The optical Fe II lines are fitted with the template described in Kovaˇcevi´c etal. (2010), Shapovalova et al. (2012) and Popovi´c et al. (2013) .After that, we fit the spectra with the Balmer continuum model, usingthe 4 free parameters (the exponent of the power law, intensity, width andthe shift of the H β ). We mask for fitting all parts of spectra except thepseudocontinuum windows at 2650-2670 ˚A, 3020-3040 ˚A, Balmer edge at3646 ˚A, and the part of spectra with λ > χ minimization routine. The pseudocontinuum windows at 2650-2670˚A, 3020-3040 ˚A are chosen because we assume that contributions of the Fe IIand Mg II 2800 ˚A lines are weak in these ranges (see Sameshima et al., 2011).
3. Applicability of the model
We test our model using the AGN type 1 spectra obtained from Sloan Dig-ital Sky Survey (SDSS) Database, Data Release 7 (DR7). We obtained the On line fit of the Fe II template can be found at http://servo.aob.rs/FeII AGN/ R e l a t i v e i n t en s i t y Wavelength (in A)
Figure 1: The original spectrum of object SDSS J020039.15 − < z < > β and Mg II 2800 ˚A emission lines. The redshift range is chosen in order toinclude the Mg II 2800 ˚A line from the blue side and whole iron shelf (5150-5500 ˚A) from the red side of spectral range. After rejecting the spectra withthe strong absorption lines, our final sample contains 293 AGNs. First wecorrected spectra for reddening and cosmological redshift. In order to removethe narrow lines and Fe II lines which overlap with the broad Balmer lines,we fit the spectra with multiple Gaussian functions (see Kovaˇcevi´c et al.,2010). Then, the narrow and Fe II lines obtained from the fit are removedfrom the original spectra, so we get a sample with cleaned broad profiles ofBalmer lines (see Fig. 1).We found that for the majority of the AGN spectra this model of thecalculated Balmer continuum gives a satisfactory fit (see Fig 2). However,there are some cases where the model cannot describe well the observedspectra (see Fig. 3). In these cases there is a discrepancy blueward of theMg II line ( ∼ ∼ ∼ ∼ ∼ ∼ ∼ ∼
4. The physical cause of discrepancy between model and observa-tion
Although the mismatch between model and observations in UV pseudo-continuum is seen in small percentage of objects from the sample, it is in-teresting to examine what could be the physical cause of these discrepan-cies. Several physical parameters may change the continuum shape: the hostgalaxy contribution, intrinsic reddening or optical depth, etc.In this sample, the host galaxy contribution is not removed, and it is pos-sible that it has influence to the continuum shape. Since we expect the host8 R e l a t i v e i n t en s i t y Wavelength (in A)
Figure 2: Up: An example of the fit (SDSS J092423.42+064250.6) with the Balmer con-tinuum model: dotted line - Balmer continuum with high order Balmer lines, dashed lines- power law and solid line - the sum of the Balmer continuum, high order Balmer linesand power law. Down: the same object, but after subtraction of the pseudocontinuum(Balmer continuum+power law) from the total spectrum. R e l a t i v e i n t en s i t y Wavelength (in A) (a)
Wavelength (in A) R e l a t i v e i n t en s i t y (b) R e l a t i v e i n t en s i t y Wavelength (in A) ( ) C -200204060801001201401601802500 3000 3500 4000 4500 5000 5500 6000 6500 7000 Wavelength (in A) -20-15-10-50510152025302000 2500 3000 3500 4000 4500 5000 5500 6000
Wavelength (in A) -100102030405060702000 2500 3000 3500 4000 4500 5000 5500 6000 6500
Wavelength (in A) (b1)(a1)(c1)
Figure 3: Left column: Typical cases where a discrepancy between model and observationis present. The model is overestimating or underestimating the flux at blue part of thespectrum (a, b), or there is a discrepancy in the continuum slope at ∼ igure 4: Distribution of the difference between the observed flux (F obs ) and calculatedpseudocontinuum - Balmer continuum (F BC ) + power law (F pl ), normalized to the ob-served flux, and measured for 2650 ˚A (left) and 4240 ˚A(right). galaxy contribution to be stronger in the objects which are less luminous, wecompared the luminosities between the objects which have different percent-age of discrepancy between model and observations, measured at wavelengths ∼ ∼ > > α is not included in thespectral range of the sample.Optical depth has influence to the Balmer continuum shape as well. Inour model the optical depth at the Balmer edge is fixed to be τ BC =1. We ex-amined possibility that the different value of the optical depth at the Balmeredge in some objects could be the reason of the discrepancies with the model.Therefore, we fit several spectra, with the strongest discrepancy in UVpart between model and observations, with the new model where the opticaldepth at the Balmer edge is taken to be the free parameter instead the fixedvalue. The example of the fit is shown in Fig. 5. We found that in the11 igure 5: The comparison between the fit using the model with τ BC =1 (up) and withmodel where τ BC is the free parameter (down). The obtained value τ BC from fit for thisspectrum is τ BC =46. few cases the extremely large optical depth improves the fit in UV part, near ∼
5. Discussion and conclusion
We found that the sum of high order Balmer lines at the Blamer edge mayreproduce very well the intensity of the Balmer continuum at the 3646 ˚A. Inthis way, the model given in Grandi (1982), may be used in simplified form,with one degree of freedom less: for the intensity of the Balmer continuum.12he Balmer continuum intensity could be calculated at any wavelength λ τ BC =1). Acknowledgements
This work is a part of the project (176001) ”Astrophysical Spectroscopy ofExtragalactic Objects,” supported by the Ministry of Science and Technolog-ical Development of Serbia. We are grateful to the Alexander von HumboldtFoundation for support in the frame of the program Research Group Linkage.13 eferences
Baldwin, J. A., Spectrophotometry of low-redshift quasi-stellar objects, ApJ201, 26-44, 1975.Dietrich, M., Hamann, F., Appenzeller, I., Vestergaard, M., Fe II/Mg IIEmission-Line Ratio in High-Redshift Quasars, ApJ 596, 817-829, 2003.Grandi, S. A., The 3000 ˚A bump in quasars, ApJ 255, 25-38, 1982.Hubbard, E. N., Puetter, R. C., Theoretical quasar emission-line ratios. VII- Energy-balance models for finite hydrogen slabs, ApJ 290, 394-410, 1985.Ili´c, D., Popovi´c, L. ˇC., La Mura, G., Ciroi, S., Rafanelli, P., The analysisof the broad hydrogen Balmer line ratios: Possible implications for thephysical properties of the broad line region of AGNs, A&A 543, A142,2012.Jin, C., Ward, M., Done, C., A combined optical and X-ray study of un-obscured type 1 active galactic nuclei - III. Broad-band SED properties,MNRAS 425, 907-929, 2012.Kovaˇcevi´c J., Popovi´c L. ˇC., Dimitrijevi´c, M. S., Analysis of Optical Fe IIEmission in a Sample of Active Galactic Nucleus Spectra ApJS, 189, 15-36,2010.Kuraszkiewicz, J. K., Green, P. J., Forster, K., Aldcroft, T. L., Evans, I. N.,Koratkar, A., Emission Line Properties of Active Galactic Nuclei from apre-COSTAR Faint Object Spectrograph Hubble Space Telescope SpectralAtlas, ApJS, 143, 257-276, 2002.Kurk, J. D., Walter, F., Fan, X., Jiang, L., Riechers, D. A., Rix, H.-W.,Pentericci, L., Strauss, M. A., Carilli, C., Wagner, S., Black Hole Massesand Enrichment of z 6 SDSS Quasars, ApJ 669, 32-44, 2007.Kwan, J., Krolik, J. H., The formation of emission lines in quasars andSeyfert nuclei, ApJ 250, 478-507, 1981.Popovi´c L. ˇC., Balmer Lines as Diagnostics of Physical Conditions in ActiveGalactic Nuclei Broad Emission Line Regions, ApJ 599, 140-146, 200314opovi´c L. ˇC., Kovaˇcevi´c J., Dimitrijevi´c, M. S., Spectral properties of asample of type 1 AGNs: influence of star formation, 2013arXiv1301.6941P.Sameshima, H., Kawara, K., Matsuoka, Y., Oyabu, S., Asami, N., Ienaka,N., Implications from the optical to ultraviolet flux ratio of Fe II emissionin quasars, MNRAS 410, 1018-1026, 2011.Shapovalova, A. I., Popovi’c, L. ˇC., Burenkov, A. N. et al., Spectral OpticalMonitoring of the Narrow-line Seyfert 1 Galaxy Ark 564, ApJS, 202, 10,2012Storey, P. J.; Hummer, D. G., Recombination line intensities for hydrogenicions-IV. Total recombination coefficients and machine-readable tables forZ=1 to 8, MNRAS 272, 41-48, 1995.Wills, B. J.; Netzer, H.; Wills, D., Broad emission features in QSOs andactive galactic nuclei. II - New observations and theory of Fe II and H Iemission, ApJ 288, 94-116, 1985. 15 able 1: The percentage of the sample where discrepancy between observed flux (F obs )and calculated flux (power low F pl + Balmer continuum F BC ) is less than 5%, 10% or15%, for both 2650 ˚A and 4240 ˚A wavelengths. percentage of the sample (F obs -(F pl +F BC ))/F obs
39 % < <
10 %94 % <
15 %
Table 2: The average luminosity at 5100 ˚A ± SD (standard deviation) for different dis-crepancy bins. The discrepancy is measured at 2650 ˚A and 4240 ˚A. discrepancy at 2650 ˚A number of objects average log( λL ) ± SD0 % – 5 % 191 44.729 ± ± ± >
15 % 13 44.596 ± λL ) ± SD0 % – 5 % 155 44.716 ± ± ± >
15 % 5 44.603 ±±