Measurement of emission spectrum for gaseous argon electroluminescence in visible light region from 300 to 600 nm

Liquid Ar is widely used as a radiation detector in particle physics experiments [1, 2, 3]. The most important feature of a liquid Ar detector is that it induces two signals: scintillation photons and ionization electrons. The combination of these signals provides useful information for applications such as particle identification and position reconstruction. In particular, the double-phase argon detector technique is employed in experiments for dark matter detection [2]. The ionization electrons drifting upwards are extracted from the liquid to gas phases under a high electric field (few kV/cm). Subsequently, the ionization electrons emit proportional electroluminescence (EL) by scattering on the gaseous Ar (GAr) atoms.

Figure 1 is a simplified diagram of the EL mechanism of GAr. It describes three emission processes for pure GAr (hereinafter called “ordinary EL”). Two of them are well-known processes: 128-nm vacuum ultraviolet (VUV) light because of transitions of Ar excimers,

and 700–850 nm infrared light (IR) owing to Ar atomic emission,

The third one is a 170–300 nm ultraviolet light (UV) emission that occurs due to the transitions of Ar molecular ions (hereinafter called “third continuum”) [4],

which is not well understood.

The EL process was explained by the ordinary mechanism, but without VL emissions. Recent measurements, however, demonstrated the presence of VL components in the EL of GAr, which was explained by Buzulutskov et al. [5] using a new mechanism called neutral bremsstrahlung (NBrS). Because VL photons are easier to detect using photosensors than VUV photons, the use of VL components in particle detectors has been actively discussed [6, 7], and a detailed understanding of the wavelength spectrum is important for such applications. Figure 2 illustrates the results of theoretical calculations of the EL emission yields as a function of a reduced electric field ($E/N$) at room temperature (300 K) for the ordinary EL model (top) [8] and the NBrS model (middle) [5]. Although ordinary EL lights emit only above the 4-Td threshold, NBrS light also emits below it. The theoretical calculations of the NBrS light wavelength spectra are depicted in the bottom plot of Figure 2 [5]. The spectrum is a continuous distribution from 200 to 1,000 nm, which is different from that of ordinary EL light.

As shown in Fig. 1, VL emissions (300–450 nm) for the Ar-nitrogen gas mixture due to the nitrogen excimer are given by the following process [9]:

Nitrogen emission has a similar $E/N$ dependence as the ordinary EL with a 4-Td threshold, as it occurs due to the transition of Ar^∗. It is known that more than 1 ppm of nitrogen impurity in liquid Ar degrades the Ar scintillation light [10]. Thus, the nitrogen impurity inside the liquid phase is usually controlled at less than 1 ppm for liquid Ar double-phase detectors. However, the nitrogen impurity in the gas phase can be higher because the boiling point of N (77 K) is less than that of Ar (87 K), and nitrogen impurities tend to concentrate in the gas phase after a certain amount of time. Thus, the effect of nitrogen impurities should be carefully considered in order to understand the VL components.

In this study, detailed measurements of the wavelength spectrum of GAr EL in the VL region (300–600 nm) were conducted under electric fields up to $E/N$ = 10 Td at room temperature (300 K) and normal pressure (1 bar) using a spectrometer. Then, the spectra obtained using ordinary EL and NBrS and nitrogen impurity are interpreted. Note that these emission processes are described as functions of a reduced electric field $E/N$ when they are weakly associated with temperature and pressure at the same $E/N$. Thus, the results of the measurement at room temperature and normal pressure are also applicable to the liquid Ar temperature (87 K) for the same $E/N$. $E/N$ is defined as the electric field per molecular density of the Ar gas. The molecular density is inversely proportional to the temperature. $E/N$ at temperatures $T_{1}$ to $T_{2}$ is given by $(E/N)_{T_{1}}/(E/N)_{T_{2}}=T_{2}/T_{1}$ with the same electric field and pressure. Thus, 1 Td at 87 K is equivalent to 3.4 Td at 300 K with the same pressure and electric field.

Figure 3 shows a schematic of the experimental setup of a GAr TPC with a fiducial volume of length 130 mm and diameter 64 mm. The TPC was constructed with polytetrafluoroethylene (PTFE) and Cu field shaping rings. No wavelength shifter was used to measure pure Ar luminescence. The grid was placed 10 mm away from the anode, after which the anode and cathode were used to generate luminescence and drift fields, respectively. The anode was transparent indium-tin-oxide (ITO)-coated quartz, which had a transmittance of more than 95% for VL ($>$300 nm). An $\alpha$-ray source (²⁴¹Am, 300 Bq) was placed inside the TPC for signals.

The TPC was placed inside a vacuum vessel with a quartz viewport. A spectrometer (Shimadzu Corporation SPG-120S) was placed outside the vessel near the viewport. We attached 2 mm slits to both the inlet and outlet of the spectrometer. The distance between the anode and the outlet slit was approximately 250 mm. Figure 4 shows the responses of the spectrometer to monochromatic laser light with wavelengths of 450 nm (left) and 520 nm (right). These responses were approximately modeled by triangular functions $T(\lambda)$ around the peak wavelength $\lambda_{peak}$,

where the width of the function $w$ = 14 nm, and the height of the peak $A$ = 1. We observed a 5 nm shift in the responses caused by the combination of performances of the spectrometer and input laser. This shift was a source of systematic uncertainties observed in the measurements.

Two PMTs with different wavelength sensitivities were arranged near the spectrometer and cathode. The PMT near the spectrometer (Hamamatsu-R11065 “VL PMT” in this study), which had a quartz window, was sensitive to UV and VL, while that near the cathode (Hamamatsu-R6835 called “VUV PMT”), which had a MgF₂ window, was only sensitive to VUV (Fig. 5). A flash analog-to-digital converter (ADC; Struck Innovative Systeme SIS3316-250-14) with a sampling rate of 250 Ms/s and a resolution of 14 bits was used to read the PMT signals.

The vacuum vessel was depressurized to $10^{-2}$ Pa using a molecular turbo pump for at least one day before each measurement. Subsequently, the detector was filled with GAr at room temperature and normal pressure. The gas was pumped at a constant flow rate of 10 L/min during the experiment to avoid impurities from outgassing. To evaluate the effect of nitrogen impurities, three admixtures of Ar and nitrogen gases (G1, N10, and N100) were used in this measurement. The specifications of the gases are listed in Table 1.

Three datasets were used in this study.

• 1.

  Data for wavelength spectrum
  High-purity G1 gas was used to obtain these data. The spectrometer wavelength was scanned from 240 to 660 nm with a 20 nm pitch at two luminescence fields of 4.6 and 8.3 Td. These $E/N$ values were chosen to match the measurements and theoretical calculations in [5].

• 2.

  Data for electric field dependence
  High-purity G1 gas was used to obtain these data. The luminescence field was scanned from 1 to 8 Td with 1-Td pitch at 300, 400, and 500 nm. Additional data for the 300 nm wavelength were obtained at 10 Td.

• 3.

  Data for nitrogen effect
  N10 and N100 gases were used for acquiring these data. The spectrometer wavelength was scanned from 240 to 660 nm with a 20-nm pitch at a luminescence field of 8.3 Td. The spectrometer wavelength for the N100 gas was scanned from 300 to 450 nm with a 2.5-nm pitch. Six additional wavelength datasets (316, 337, 358, 381, 406, and 434 nm) corresponding to the resonant wavelength of nitrogen emission were taken for the G1 and N10 gases [9].

Note that 1 Td of the reduced electric field in the luminescence region (room temperature, normal pressure, and 1 cm luminescence region) corresponds to 245 V/cm.

The primary scintillation light (S1) and ionization electrons were produced by the interaction between the $\alpha$-ray and the GAr atoms. The typical flight distance of the $\alpha$-ray was 3 cm. The ionization electrons drifted in the direction of the grid under a fixed drift field of 100 V/cm (0.4 Td). The electrons emitted secondary EL lights (S2) when they were in the 1 cm gap between the grid and the anode under a luminescence field (1–10 Td).

The typical waveform distributions measured by the Flash ADC are illustrated in Fig. 6. The top and bottom plots correspond to the VUV PMT and VL PMT signals, respectively. These distributions were obtained without a spectrometer. The actual light yield of the VL PMT with a spectrometer was less than one photoelectron (PE). The sharp peak for the VUV PMT at t = 0 corresponds to the S1 signal, while the broad peak at approximately t = 15 $\mu$s corresponds to the S2 signal.

Five light yields (S1, S2, Signal, BG1, and BG2) were calculated by integrating the waveform distributions of the VUV and VL PMT, as illustrated in Table 2. The integration intervals are indicated by the hatched regions in Fig. 6. Events were selected using S1 and S2 in the red box in the top-left plot in Fig. 7. Data for the plot were obtained at a spectrometer wavelength of 400 nm and a luminescence field of 8.3 Td using G1 gas, and 137,684 events were selected (N₀). Distributions of Signal, BG1, and BG2 for the selected events are depicted in the top-right, bottom-left, and bottom-right plots of Fig. 7, respectively. As mentioned earlier, no photons were observed most of the time, and the peaks around 100 counts corresponded to single PE events. The VL PMT was operated with gain for single PE events of approximately 130 counts, and the PMT had distinct separation between the noise level and the single PE events [11]. The single PE events were selected based on the light yields within the range of 30–300 counts, and the selected events were N_Signal = 943, N_BG1 = 166, and N_BG2=502 for Signal, BG1, and BG2, respectively. The signal region contained background events, mainly due to the accidental coincidence of the PMT dark counts. Background contamination was estimated using BG1 and BG2 normalized to the window width, $\mathrm{N_{Background}}$ = $(25~{}\mu s/60~{}\mu s)\times\mathrm{(N_{BG1}+N_{BG2})}$ = 278, and the number of events after background subtraction was $\mathrm{N_{Signal}}$ $-$ $\mathrm{N_{Background}}$ = 665. Assuming that the light yield followed a Poisson distribution, the light yield in the unit of PEs equaled the fraction of events in the single PE peak is $\mathrm{(N_{Signal}}$ $-$ $\mathrm{N_{Background})/N_{0}}$ = 0.00483.

By repeating the same analysis steps for different configurations (spectrometer wavelength, luminescence field, and gas type), we obtained the desired spectrum. The left plot in Fig. 8 illustrates the obtained wavelength spectra using G1 gas at a luminescence field of 8.3 Td. The blue, red, and black points denote the light yields in the signal, background, and signal–background region, respectively. As the light yield in the background region was steadily constant, the average value (red line) was used for subtraction. Finally, the wavelength spectrum (right plot in Fig. 8) was obtained after correcting for the VL PMT quantum efficiency in Fig. 5). Two sources of systematic uncertainties were considered for the observed light yield.

• 1.

  PMT quantum efficiency.
  As discussed in the previous section, the wavelength of light after passing through the spectrometer followed the distribution described by Eq. 6, and a systematic shift of $\pm$5 nm was observed for the spectrometer. Thus, the wavelength to be used to calculate the quantum efficiency of the VL PMT is uncertain. The difference in quantum efficiency when changing the wavelength $\pm$7 nm was assigned as the systematic uncertainty of the observed light yield.

• 2.

  Reproducibility of measurement
  The stability of measurement during the data-acquisition period was estimated by repeating the measurement. A relative uncertainty of 5% was assigned to the observed light yield.

The red inner error bars in the right plot of Fig 8 demonstrate statistical uncertainty, and the outer error bars with black color represent the total ($\sqrt{\mathrm{statistical^{2}+systematic^{2}}}$) uncertainty.

The wavelength spectra obtained are shown in Fig. 9. The left plot shows the spectra for the G1 gas with electric fields of 4.6 Td (green points) and 8.3 Td (black points). The right plot shows the spectra for the G1 gas (black points), N10 gas (blue points), and N100 gas (red points) at $E/N$ = 8.3 Td. The transmittance of the ITO-quartz degraded for the wavelength below 300 nm, and there were background contributions from the secondary light of the spectrometer for the wavelength above 600 nm (300 nm$\times$2). Thus, a wavelength range of 300–600 nm was used for the rest of the analysis.

The GAr EL in the VL region (300 to 600 nm) was studied using the GAr TPC, with ²⁴¹Am $\alpha$-rays as the signal source at room temperature and normal pressure. The secondary emission light waves from the TPC luminescence region were dispersed using a spectrometer. The wavelength spectrum and luminescence-field dependence of the light yield were compared with those of the ordinary EL and NBrS models. The effect of nitrogen impurities on the light yield was evaluated using the Ar-nitrogen mixture gas. We conclude that the ordinary EL model and nitrogen emission alone cannot explain the wavelength spectrum and electric field dependences of the observed light in the VL region. The inclusion of the NBrS model can enable a comprehensive explanation of the phenomena.

The authors thank Dr. A. Buzulutskov for helpful discussions on the NBrS model.

This work is part of the research performed under the Waseda University Research Institute for Science and Engineering (project number 2016A-507/2020N-006), supported by JSPS Grant-in-Aid for Scientific Research on Innovative Areas (15H01038/17H05204), Grant-in-Aid for Scientific Research(B) (18H01234), and Grant-in-Aid for JSPS Research Fellow (20J20839). The authors acknowledge the support of the Institute for Advanced Theoretical and Experimental Physics, Waseda University.