Near-infrared Hydrogen and Helium QSO Emission Lines

The total observed flux of any emission line is dependent on the strength of the incident ionizing radiation and the range of densities locally present in the emitting gas. Using photoionization modeling, Ruff et al. (2012) were able to derive a set of interesting line flux ratios which placed constraints on the two most important ionization parameters for helium and hydrogen BELR gas: local gas number density and incident ionizing flux.

A complete model of the BELR region should be capable of explaining the observed flux ratios and properties consistently, including scattering, photoionization, and recombination mechanisms. An estimate of the photoionizing UV flux is therefore required to ensure that the overall energy budget of the system remains physically tenable.

The observations presented below consist of 14 NIR quasar spectra, spanning the combined JHK wavelength regions of approximately 0.97–2.52 μ, and thus include the following emission lines: H α, He i 6678, He i 7065, Pa , Pa δ, He ii 10120, He i 10830, Pa γ, and Pa β. Exposure times and associated parameters were calculated so as to achieve an anticipated continuum S/N ≥ 10 ensuring that emission lines were adequately discernible from the continuum and that equivalent widths (EWs) could be properly measured and compared.

Quasar targets were selected on the basis of having redshifts of 0.58 ≤ z ≤ 0.63, as this includes the hydrogen and helium emission lines of interest within the covered grism spectral ranges with minimal interference from atmospheric opaque regions. Importantly, this will allow for the detailed analysis of helium flux ratios.

The spectra were reduced, and instrument and sky signatures were removed, using IRAF (Tody 1986, 1993) and the Gemini IRAF package version 1.13. The procedure used is fairly standard for NIR spectroscopy, though there were some key differences. Flamingos 2 displays a rather strong dark current pattern. For that reason, we chose to take a conservative approach and opted to remove the dark current using dark exposures of the same exposure time as the calibration (flat and telluric standard star) and science observations. A telluric standard star was observed as close as possible in time before or after each science observation. This was done to ensure a close match between the airmass of the telluric standard star and that of the science target. We chose to use as telluric standards hot A and B stars, a summary of which can be found in Table 1. Both the science and telluric observations were taken in an ABBA sequence, where the instrument was nodded back and forth along the slit in a quad sequence, in order to improve sky subtraction. The telluric standard star and science observations were flat-fielded, combining and shifting the ABBA sequence images in the process, and then wavelength calibrated and extracted.

The wavelength calibration required the use of the OH lines present in the telluric and the science spectra, before sky subtraction was performed. Lamp arcs were obtained just before or after the observations, with the telescope in the same position as the observations. However, we found that the arcs were not useful in calibrating our spectra. Flamingos 2 suffers from optical distortion toward the edges of the field of view and the arc lines become visibly asymmetrical toward the red and the blue ends of a spectrum. We believe that this optical distortion combined with a different optical path between the lamp observation and the on-sky observation was sufficient to render the arcs unusable. Fortunately, both the telluric observations and the science observations had sufficiently strong OH emission throughout the region of interest to allow us to use those lines to calculate the wavelength solution. The process was performed interactively and led to reliable results.

The hydrogen lines in the telluric standard star spectra were removed with Spectool, a routine within IRAF, by fitting a Voigt profile to each line. Each telluric spectrum was spectrophotometrically calibrated via a blackbody spectrum referenced to the temperature of the telluric star and then scaled in flux and shifted in wavelength to match the science spectrum with IRAF's telluric task and then divided through the science spectrum producing our final science spectrum.

Two of the sources (SDSS J034025.48-000819.8 and SDSS J231535.04+000127.7) had additional JH exposures taken on different dates. The two different exposures were reduced using the usual Gemini IRAF procedure described above and were then combined at the end using the IRAF procedure scombine.

The spectra have been corrected for foreground extinction by dusty gas in the Milky Way. This was performed using line-of-sight extinction values for each quasar location from Schlafly & Finkbeiner (2011) and applying extinction curve models provided by Fitzpatrick & Massa (2007) using a combination of bespoke routines and the publicly available code, extinction (Barbary 2016). Linear V-band extinction values for each source (as listed in Schlafly & Finkbeiner 2011) are quoted in Table 1.

Figure 1 is an example of the final spectral plots produced for each source (in this case, SDSS J003251.46-002748.0). The spectra are not absolutely flux calibrated as the observing conditions were nonphotometric. The complete figure set of 28 spectral plots is available.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

View figure set (14 panels)

Tarball of figure set images

EWs of relevant emission lines were measured using a combination of bespoke routines and the python package lmfit.py (Newville et al. 2014). The measurement process involved co-fitting line profiles with a local straight line continuum, overlaid with a skewed Voigt profile described by

$$\begin{eqnarray}&&f(x;A,\mu ,\sigma ,\gamma )=\displaystyle \frac{A{\rm{Re}}[w(z)]}{\sigma \sqrt{2\pi }},\end{eqnarray} \tag{ 1 }$$

where

$$\begin{eqnarray}&&z=\displaystyle \frac{x-\mu +i\gamma }{\sigma \sqrt{2}}\end{eqnarray} \tag{ 2 }$$

$$\begin{eqnarray}&&w(z)={e}^{-{z}^{2}}\mathrm{erfc}(-{iz})\end{eqnarray} \tag{ 3 }$$

and erfc is the complementary error function, μ is the peak centroid, A is the amplitude, σ is the Lorentzian contribution to the characteristic width, while γ represents the Gaussian contribution to the characteristic width. Re[w(z)] denotes the real part of the complex function w(z). Skew values relate to the cumulative distribution function of the standard normal distribution function. As an examination of the 28 AGN spectral plots in the online figure set will demonstrate, emission line profiles in the 14 AGN sources did not have any discernible unusual structures. Therefore, fitting a Voigt profile made sense pragmatically. There were difficulties in achieving a good (or any) fit to some profiles, but this was due to poor local S/N, rather than the particular shape chosen to model the individual line.

Initially, each line was fit iteratively by eye to obtain reasonable starting values for parameters such as slope and line intercept for the continuum and line center, γ, σ, and amplitude for the skewed Voigt profile, in addition to the wavelengths over which to measure the line profile. The default fitting process sets σ (the Lorentzian contribution to the skewed Voigt function) equal to γ (the Gaussian contribution to the skewed Voigt function). In this initial phase, a continuum range of 300 Å on both sides of the measured peak was used for H α (and/or He i 6678), He i 10830, and Pa γ, while a continuum range of 500 Å on both sides of the measured peak was used for the other emission lines. The continuum was linearly fit using these ranges on each side of the peak. All absorption, cosmic rays, and other extraneous emission lines were masked. Each continuum and line profile was then simultaneously fit via bootstrapping and resampling. For each measurement, the flux was randomized within a normal distribution of the rms error and the properties of each line profile were measured for 1000 different realizations. Additional parameters including the FWHM and EW were calculated from the resulting fits. The median, 16th percentile, and 84th percentile were taken to be the parameter value, lower error, and upper error, respectively.

There are two narrow nitrogen emission lines—the [N ii] λλ6548, 6583 doublet—flanking the dominant H α line, which appear in the JH spectra for each source. EW values for these narrow emission lines are expected to be 0.22%–1% of that measured for the H α line and are, therefore, not deblended from the H α line measured for each source. The resulting measurements for each source are found in Table 3 and discussed further in Section 5. Light from the host galaxy is assumed to be negligible as these are moderately high-redshift quasars.

A brief examination of each spectrum clearly demonstrates that some emission lines are blended with other emission lines and in many of these cases, these emission lines cannot be extricated from one another with any certainty. One example of this is the previously mentioned nitrogen emission lines on either side of the H α line. As stated, these lines are narrow and likely to contribute very little to the overall emission of the dominating H α line, with the exception of a few of the sources (SDSS J033202.33-003738.9, SDSS J222515.32+010340.4, SDSS J231535.04+000127.7, and SDSS J231645.08-001129.4), where the emission lines clearly complicate the hydrogen profile. It is important, therefore, to note that listed errors are likely to be overestimated. There also exists a broad emission line, He i λ6678, in the redward wing of the H α line. In many spectra, this line can be seen to clearly broaden the red wing of the H α complex, however, deblending remains problematic. Thus, there has been no attempt in the measurement process to deblend the nitrogen emission lines or the He i line, and the resulting measurement is of the H α complex rather than the individual emission lines.

Blended emission lines occur within the spectra in two other cases and their treatment was adapted on an individual basis. In the cases where the emission lines could not be clearly separated from each other, they were profiled and measured as a complex. This was often the case for the Paschen δ + He ii λ10120 and He i λ10830 + Paschen γ complexes. In spectra where the individual emission lines were indiscernible and/or the emission lines could not be profiled accurately as a complex, the lines were measured individually.

Limitations with the Flamingos-2 instrument lead to a fish-eye effect at the edges of some of the spectra. This is particularly obvious in the redmost wavelengths of the JH grisms of these sources, creating difficulties in the confident measurement of the He i λ10830 + Paschen γ complex. In these cases, although the consistency in the shape of the emission line complex can be seen, the continuum is likely not representative of the true value at these wavelengths, changing rapidly toward the longest wavelengths and severely affecting any fitting of the line. To minimize these effects, continua for the spectra were fitted locally with the emission lines and only over a small wavelength region (using a straight line) rather than attempting to fit the continuum for the entire spectrum. Additionally, this effect has led to the omission of the measurements of some of the He i λ10830 + Paschen γ complexes in the tables above, despite the emission lines being easily discernible by eye in the plotted spectra.

In some cases, particular emission lines were not visible above the noise in the spectrum. This tended to affect the He i λ7065 line the most, although at times Pa was also difficult to discern and measure. Generally, the H α complex and the Pa β emission lines were well measured, although these tended to fall at the sides of the spectrum and therefore were most susceptible to any fish-eye effect of the instrument.

This data set represents a significant improvement in the number of NIR quasar spectra with a suitable S/N for in-depth photoionization analysis of the BLR spectrum, for example via methods pioneered by Ruff et al. (2012) and further explored in a recent paper by Schnorr-Müller et al. (2016). In a future paper, we will compare the spectral measurements made in this paper with predictions from photoionization simulations with the aim of placing better constraints on the physical conditions of the hydrogen emitting regions (such as number density and incident ionizing flux) and also, for the first time, constrain the physical conditions involved in creating helium broad emission lines in the spectra of quasars.