Argon I Lines Produced in a Hollow Cathode Source, 332 nm to 5865 nm

We report precision measurements by Fourier transform spectroscopy of the vacuum wavenumber, line width, and relative signal strength of 928 lines in the Ar I spectrum. Wavelength in air and classification of the transition are supplied for each line. A comparison of our results with other precision measurements illustrates the sensitivity of Ar I wavelengths to conditions in the light source.

Zarem [J. Quant. Spectros. Radiat. Transfer 53, 1 (1995)]. Their measurements, made with the high-resolution Fourier Transform Spectrometer at the National Solar Observatory (Kitt Peak), have been widely used as a source of wavelength standards. Measurements of neutral argon from the same spectra have been circulated as an unpublished line list. This list has been used in a number of laboratories to identify lines of argon in spectral sources and as a source of calibration wavelengths.
This article presents the comprehensive list of neutral argon observations from the Kitt Peak spectra over a wide wavelength region from the near ultraviolet to near infrared and describes the experimental conditions under which the measurements were made. It was prepared by Ward Whaling for publication in the NIST Journal of Research with the encouragement and assistance of the Atomic Spectroscopy Data Center in order

Introduction
Argon is frequently used to sustain the discharge in a hollow cathode spectral source, and the source typically excites the first, second, and third spectra of argon as well as the spectrum of the cathode material. Consequently, a complete picture of the argon spectrum is useful in avoiding, or correcting for, interference between argon lines and lines from the cathode material. In this paper we provide such a line list for Ar I; similar line lists for Ar II [1] and Ar III [2] have been published earlier.

Experimental Method
Argon lines were obtained by comparing many highresolution spectra of a hollow cathode discharge in argon, with different cathode materials for the different spectra. By extracting lines common to several spectra, regardless of the cathode material, we have generated a list of lines arising from the argon that supports the discharge, or from a common contaminant as discussed below.
All spectra were recorded on the 1 m vacuum Fourier transform spectrometer [3] at the National Solar Observatory (Kitt Peak). Hollow cathode spectra from the Kitt Peak archive with cathodes of Co, Cu, Fe, Mo, Ti, V, and Y were measured for this work; the Kitt Peak National Observatory archive designation of each spectrum is given in Ref. [1]. The cathode cavity was 25 mm long by 8 mm diameter; further details of the water-cooled source will be found in Ref. [1]. The source pressure varied from one spectrum to another between 270 Pa and 530 Pa (2 Torr and 4 Torr) of high purity argon, with discharge currents between 150 mA and 600 mA.
All spectra were measured with the DECOMP [4] line-finding program developed by Brault. This program fits a Voigt profile to the spectral feature and records the line-center, peak amplitude, full width at half maximum intensity (FWHM), and other parameters of the Voigt profile.

Results
In the first column of Table 1 we list the line-center vacuum wavenumber (in cm Ϫ1 ), followed by the experimental uncertainty in column 2. The ratio of the peak amplitude S to the root mean square background noise N appears in column 3 as the log 10 (S /N ) for each line. Note that the peak amplitude has not been corrected for the spectrometer response. Over the range of a given spectrum, the spectrometer response may vary by an order of magnitude, and the comparison of the signal-tonoise ratio of two lines is a reliable indication of the relative strength of the lines only if they are close in wavelength. The full width of the observed line (in 10 Ϫ3 cm Ϫ1 ) at half maximum height appears in the 4th column. The absolute value of the line width depends on source and instrumental conditions, but the increased relative width of lines from levels of high excitation and high angular momentum is an aid in identification.
The uncertainty ⌬ in the line-center wavenumber appears in the second column. This experimental uncertainty (one standard deviation level) comes from two sources: (1) the uncertainty ⌬ L in locating the center of the line in the presence of noise, and (2) ⌬ S from the uncertainty in the scale of the FTS. We combine these to find the overall uncertainty in the wave number: ⌬ = [(⌬ L ) 2 + (⌬ S ) 2 ] 1/2 .
(1) We evaluate the uncertainty in locating the line center from ⌬ L = 0.5(W )/(S /N ), where W (the FWHM) and S /N (the signal-to-noise ratio) appear in Table 1, and the factor 0.5 is an approximation to the more precise factor given by Davis, Abrams, and Brault [5]. For very strong lines (S /N > 10 4 ) this expression may underestimate ⌬ L because it neglects self-absorption that may, if asymmetric, shift the line center, and it neglects the local increase in the noise level under a strong line in the FTS spectrum [6]. We therefore set a lower limit on ⌬ L of 0.0003 cm Ϫ1 . The argument supporting this limiting value is presented in Ref. [1].
(2) Wavenumbers measured with the FTS must be multiplied by a scale factor s to correct for any angular deviation between the path followed by light from the source as it passes through the instrument and the path followed by the laser beam that measures the displacement of the FTS mirror. The scale factor s is a constant (ϳ1) for a given spectrum and is determined empirically by measuring with the FTS standard lines of accurately known wavenumber; s = (standard wavenumber)/(measured wavenumber). Any uncertainty ⌬s in the determination of s contributes a proportional uncertainty in the corrected wavenumber given by ⌬ S = (⌬s /s ). The factor in parentheses was typically (3 to 7) ϫ 10 Ϫ8 for the different spectra that we measured. We adopt as the uncertainty ⌬s for a particular spectrum the standard deviation in the mean value of s derived from measurements of 28 Ar II standard lines in that spectrum.
As standards we used as many as possible of the 28 Ar II lines recommended by Learner and Thorne [7]. These well isolated lines of good strength from transitions between low excitation levels should be minimally sensitive to pressure in the source. Our wavenumbers for these Ar II standard lines were taken from Ref. [1] and derive ultimately from the CO molecular lines used to calibrate the Ar II spectra [8]. Our Ar II wavenumbers are slightly higher than the values reported by Norlén [9]. For the 28 Ar II standard lines, the mean value of the ratio (based on CO)/ (Norlén) is [1 + 67(8) The list of argon wavenumbers collected by comparing different spectra was then compared with line lists in the literature for Ar I, Ar II, and Ar III-notably those of Minnhagen [10], Striganov and Sventitskii [11], and Kurucz and Peytremann [12]-to identify the ion and classify the transition. Additional Ar I lines with wavelengths longer than those in existing compilations were identified by calculating all electric dipole transition energies allowed by parity and J -value selection rules, using Ar I level energies from Moore [13] as expanded and refined by level energies from Minnhagen [10] and Norlén [9]. In Table 1, we identify the transition upper level in columns 5-7 by its energy, J -value, and orbital, and the lower level in columns [8][9][10]. When individual J levels could not be resolved, the [K] value (the result of coupling the total angular momentum J 1 of the 3p 5 core with the orbital angular momentum l of the valence electron) is given. The last column gives the wavelength in air as determined from the vacuum wavenumber by a formula from Edlén [14].
Ar II and Ar III lines were identified and classified in the same way. When all Ar I, II, and III lines had been removed from the list of common lines, we searched for possible contaminants that might be common to all our spectra by searching for the strongest lines in the spectra of H, He, C, N, O, Ne, Na, and Fe. A few H, C, and O lines were found, readily identified as coming from a low-mass contaminant by their broad Doppler widths. On the Cu cathode spectrum between 2000 cm Ϫ1 and 3000 cm Ϫ1 , molecular lines from ArH + were abundant and strong (S /N up to 10 2 ).
With all known Ar and contaminant lines removed, the 114 common lines left over are presumably from Ar. By comparing the relative intensity of these unidentified lines in spectra recorded at different power levels, we conclude that most come from an Ar ion. Four unidentified lines that appear to come from Ar I are included in Table 1 with no designation.

Comparison With Other Line Lists
Our Ar I wavelengths agree with those of Minnhagen [10] within the modest precision he assigns: "better than 10 mÅ for lines that are not too weak." Our agreement with Norlén's [9] wavenumbers is generally within (1 or 2) ϫ 10 Ϫ3 cm Ϫ1 , as might be expected from the close agreement of his calculated wavenumbers for Ar II with the values we used to calibrate the FTS scale. Norlén's calculated wavenumbers are essentially averages over many alternate (Ritz-equivalent) paths that reduce the uncertainty of his values well below that which we are able to achieve in a single measurement, especially for weak lines. Hence Norlén's values, multiplied by [1 + 67(8) ϫ 10 Ϫ9 ], are to be preferred to those in Table  1 when available.
Palmeri and Biemont [15] have measured 100 transitions from 4f , 5g , and 6g levels in Ar I, using spectra recorded with the same FTS used in the present experiment. In their hollow cathode source they used various Ar pressures but all were lower than ours; their average pressure of about 130 Pa (1 Torr) was lower than ours by a factor of 1/3. For the 4f → 3d transitions (66 lines), our wavenumbers agree beautifully with theirs; the mean deviation between the two sets of wavenumbers is 0.4(1.1) ϫ 10 Ϫ3 cm Ϫ1 . For the 5g → 4f and 6g → 4f transitions (33 lines), our wavenumbers are, on average, greater than theirs by 6 ϫ 10 Ϫ3 cm Ϫ1 . We attribute this difference to a pressure effect that appears to be strongly dependent on the angular momentum of the levels involved. This difference did not show up in our comparison with Norlén's wavenumbers, even though his source pressure was lower than ours by a factor of 1/10, because the shift depends on the angular momentum of the levels involved and Norlén measured only transitions between s , p , and d orbitals in Ar I.
Any interaction between an excited Ar atom and neighboring atoms that shifts the energy of an excited level will also broaden it, and this effect is clearly seen in Table 1. For example, transitions near 3800 cm Ϫ1 from the 6g levels in Ar I have an average width ഠ 300 ϫ 10 Ϫ3 cm Ϫ1 , while nearby lines from 4d' and 5p' upper levels have widths between (18 and 22) ϫ 10 Ϫ3 cm Ϫ1 . The dependence on l appears to be greater than the dependence on n ; nine lines from 7p configurations in the same (3600 to 4000) cm Ϫ1 range have an average width of 38 ϫ 10 Ϫ3 cm Ϫ1 . In these examples, the lines are only broadened and shifted at the pressure (270 Pa to 530 Pa) of our source, but the lines may not be seen at all from a source at higher pressure. Palmeri and Biemont [15] report lines near 2100 cm Ϫ1 from 7g and 7h levels of Ar I that are broadened beyond recognition as lines on our spectra.
We conclude that the gas pressure in our hollow cathode source has increased the wavenumber of some lines from levels of high l or n . The shifted lines display a width noticeably greater than that of nearby transitions between s , p , and d levels. The wavenumbers, line widths, and relative signal amplitudes in the table are those one can expect to see from a hollow cathode discharge in Ar at a pressure of a few hundred Pa. 331.9342 a The notation nl (e.g., 6p ) denotes levels of the 3p 5 nl configuration with 3p 5 ( 2 P3/2) core. The notation nl' (e.g., 6p' ) denotes levels of the 3p 5 nl configuration with 3p 5 ( 2 P1/2) core.