Theoretically, the IR radiance intensity can be measured with a passive FTIR detector using Eq. (6) as follows:
$$\text{I}\text{R} \text{r}\text{a}\text{d}\text{i}\text{a}\text{n}\text{c}\text{e} \text{i}\text{n}\text{t}\text{e}\text{n}\text{s}\text{i}\text{t}\text{y}= \text{ϵ}\left(\text{c}\text{d}{\Delta }\text{T}\right)$$
6
Therefore, when some specific environments, such as extremely high concentration, long path length, or large temperature differences occur, the IR intensity in a particular wavenumber correlated with the intrinsic SF6 molecular peak exceeds the limit of the portable FTIR detector. Moreover, SF6 has a strong vibrational IR-active ν3 mode at 947 cm− 1. [25, 26] We conducted IR radiance measurements with a moderate temperature difference (10°C) and path length (1 m) to control the IR radiance of SF6 gases depending on the SF6 concentration. We continuously injected a constant volume of SF6 gas to observe peak distortion through concentration changes. Starting with a low concentration (34.5 mg/m2) of SF6, we measured different concentration ranges of SF6 to confirm the distortion behavior of the BT spectrum. Figure 2a shows the 3D accumulated graph obtained by averaging more than 16 brightness temperatures for each SF6 concentration. The low-concentration region from 34.5 mg/m2 to 600 mg/m2 does not show the peak distortion and shoulder peak at 931 cm− 1. However, at 650 mg/m2 concentration, a shoulder peak was assigned, and peak distortion was observed at 947 cm− 1.
Comparing the two designated concentration (103.5 mg/m2 and 1,380.8 mg/m2), the peak distortion phenomenon is clearly observed in Fig. 2b. At low concentrations of SF6, the intrinsic sharp IR peak is observed at 947cm− 1, which is distinguishable without any distortion of the peak. However, if the concentration of SF6 is increased, the intensity of the IR peak will exceed that of the IR detector. In addition, the low resolution of the portable passive FTIR detector accelerated the peak distortion. As a result, in the region above a certain concentration, the portable passive FTIR detector could not distinguish the position of the exact peak despite acquiring a sufficient IR spectrum of SF6 gases, and the detector acquired only a horizontal spectrum depending on their resolution. As shown in Fig. 2b, only the main peak of the S–F vibration mode is observed in the red line (low concentration of SF6, 103.5 mg/m2), only a main peak of S-F vibration mode is observed. However, in the blue line (high concentration of SF6, 1,380.8 mg/m2), a shoulder peak at 931 cm− 1 is observed, and the main peak area gradually widens as the concentration increases.
Portable passive FTIR detectors are typically used to detect chemicals. Most detection algorithms rely on a probability function that compares BT spectral similarities. Thus, we conducted detection algorithm score comparison focused on peak distortion of SF6 BT spectrum to confirm the effect of false negative alarm.
Figure 3 shows that the non-distortion library exhibits a high detection score in the spectrum for the low-concentration region (under 400 mg/m2). As the SF6 concentration gradually increased, the detection score of the non-distortion library gradually decreased. Moreover, the detection accuracy of the distortion library gradually increased above 600 mg/m2.
Figure 4 shows the receiver operating characteristic (ROC) curves of the detection algorithms with a dual library (non-distortion and distortion) and a non-distortion library. Using the dual library, the performance of the detection algorithm increased, as indicated by the area under the curve. In particular, the detection performance was improved by 15% compared to that of the non-distortion library. However, without the distortion library, the true-positive detection rate distinctly decreased. This is because, as shown in Fig. 3, the distortion library covered the detection of an SF6 broader concentration than the non-distortion library; thus, an increase in the true positive rate included detection results of high concentration SF6 that the detection algorithm with the non-distortion library could not include.
We performed an outdoor test to confirm the detection efficiency of the dual library for a portable passive FTIR detector (Fig. 5a). As shown in Fig. 5a, the proposed portable passive FTIR detector, a miniaturized stand-off chemical agent detector (called MSCAD) [16], was installed in an outdoor flower plantation 3 km from the gas-releasing site. The SF6 gas-releasing site was configured for observation by MSCAD as the sky background. The SF6 gas was rapidly released through a 1/4-inch tube line from the cylinder, observed at the MSCAD site. We conducted release tests under mild wind conditions to reduce the amount of released SF6 as much as possible. Sky is a typical background among the various IR radiation backgrounds owing to its low-temperature conditions under − 10°C. Moreover, released SF6 from the cylinder has a low temperature because it is a compressed gas; however, rapid heat transfer occurs immediately after the SF6 gases are released into the air. This large temperature difference between the background (sky) and gases (SF6) leads to a high IR radiance propagation to MSCAD, as described in Eq. 2. This high IR radiance propagation induced peak distortion of SF6 in the outdoor demonstration. Figure 5b presents the number of detected BT spectra for SF6 depending on the library (distortion and non-distortion library). At the initial elapsed time, the number of detections were similar regardless of the library conditions. However, the heat of SF6 was transferred rapidly, and its concentration also increased because of SF6 release. As shown in Fig. 5b, the number of detections for the distortion SF6 library is larger than that for the non-distortion SF6 library. Although the release of SF6 gas was stopped after 180 s, we confirmed that the SF6 gas was maintained owing to stable atmospheric conditions. After 20 s of release, the IR radiance of the SF6 gas was strong because of the temperature difference and concentration. Therefore, the IR spectrum was observed as a distorted peak. The number of detections of the non-distortion library gradually increased after the SF6 gas cloud dissipated over time.
Finally, we conducted a principal component analysis (PCA) to determine whether the outdoor demonstration data could be distinguished depending on the library types. As shown in Fig. 5c, the non-distortion library detection cases partially overlap with the distortion library detection cases. However, the distortion library covers a wide range of high-concentration SF6, where a non-distortion library cannot be detected. Therefore, the PCA results show that the dual library can effectively detect SF6 with extensive concentration levels during the outdoor operation of MSCAD.