Feasibility Study of the One-Dimensional Radiation Distribution Sensing Method Using an Optical Fiber Sensor Based on Wavelength Spectrum Unfolding

The significant amount of radionuclides was discharged into the environment in the wake of the Fukushima Daiichi nuclear power station (FDNPS) accident, which occurred following the massive tsunami caused by the great east Japan Earthquake in March 2011. Decommissioning operations, including decontamination inside and outside the reactor buildings of the FDNPS, are underway. To carry out decommissioning operations in the FDNPS, locating radioactive hotspots is essential. Clarifying the precise positions of locally existing hotspots is essential for establishing countermeasures for reducing external doses to the workers and ensuring efficient decontamination.

For the radiation distribution measurement inside the FDNPS reactor buildings, dose rate measurement using a handheld survey meter was employed by the Tokyo Electric Power Company Holdings [1]. However, to carry out the radiation distribution measurement over the wide area of the FDNPS reactor buildings, measurement using survey meter is time-consuming, given that workers' exposure to radiation during the measurement is a critical issue. Moreover, locally existing hotspots may be overlooked because of the “at-point” measurement.

Radiation distribution measurement using a gamma-ray imager, such as a pinhole camera or Compton camera, was conducted by several institutes inside the FDNPS reactor buildings [2,3]. The gamma-ray imager is a powerful device, which can measure the radiation distribution of remote places. However, it can only measure the incident direction of gamma-rays and difficult to evaluate the relative intensity of radioactive hotspots inside the field of view. Moreover, it can only measure a gamma-ray emitting source, such as ¹³⁷Cs. Therefore, it is unable to measure the distribution of a beta-ray emitting source, such as ⁹⁰Sr/⁹⁰Y, which is one of the dominant radioactive sources inside the FDNPS site.

In this study, we focus on radiation distribution measurement using plastic scintillating fiber (PSF). PSF is an optical fiber-type plastic scintillator, which is sensitive to various radiation types (e.g., gamma, beta, and neutron) [4]. It can measure the distribution of radioactive material one-dimensionally along a fiber using the time-of-flight (TOF) method. This method measures the incident position of radiation to the PSF from the time difference between two signals reaching both ends of the PSF [5,6]. Taking advantage of this characteristic, a PSF has been applied inside the FDNPS site to monitor the leakage of the contaminated water from a large container [7]. However, under the high dose rate radiation field (e.g., the FDNPS reactor buildings), which exceeds several mSv/h, the pile-up event of timing signal and chance coincidence effect in fiber increase, with the timing spectrum no longer reflecting the incident position of radiation [5]. To apply the PSF to the FDNPS buildings, an alternative method to measure radiation distribution under the high dose rate radiation field is required.

In this study, we propose a novel radiation distribution sensing method that can be applied to the high dose rate radiation field, which uses the wavelength and attenuation of light emitted and transmitted inside the fiber. We focus on the fact that the amount of attenuation of light along the fiber depends on its wavelength. By measuring the wavelength spectrum of every transmission distance using the spectrometer in advance, the radiation distribution along the fiber can be estimated reversely by the unfolding method. The proposed method draws information from the integrated light intensity of specific wavelengths. Thus, signal pile-up and chance coincidence effects under the high dose rate radiation field, which are critical to the TOF method, can be avoided. In this study, we demonstrate the basic characteristics of the proposed method.

where $λ$ is the wavelength of light, $x$ is distance from the fiber edge, $S(λ)$ is observed wavelength spectrum at the fiber edge, $I(λ)$ is initial wavelength spectrum, $μ(λ)$ is the attenuation length of light with wavelength $λ$⁠, and $f(x)$ is radiation intensity at position $x$⁠. $R(λ, x)$ is the response function, which can be known in advance. Therefore, by measuring $S(λ)$⁠, radiation distribution $f(x)$ can be estimated reversely.

The experimental setup used to demonstrate the principle of the proposed method is shown in Fig. 2. In the experiment, a PSF (Kuraray SCSF-81, Kuraray Co., Chiyoda-ku, Tokyo, Japan) was used as an optical fiber sensor. A PSF is expected to have lower radiation tolerance than that of a quartz optical fiber; however, it has higher emission intensity and transmission loss than that of a quartz optical fiber. It can also obtain the emission wavelength spectrum not only by the radioactive source but also by the ultraviolet (UV) irradiation, which shows higher emission intensity. Therefore, it is suitable for verifying the proposed method's principle, which uses the information of light attenuation along the fiber. A PSF's length and diameter are 10 m and 1 mm, respectively. The system comprises a PSF, a quartz optical fiber, and a spectrometer. Assuming the application in the FDNPS reactor buildings, the emission light from the PSF is transmitted to the spectrometer via quartz optical fiber without a severe intensity loss to avoid the spectrometer from exposure to the radiation field. The spectrometer (Ocean Optics QEPro) uses a Peltier-cooling charge-coupled device as the photodetector, which has low dark noise.

The UV excitation source (Emission peak: 370 nm) irradiates PSF every 50 cm to obtain the response function. Figures 3(a) and 3(b) show the obtained response function and evaluated attenuation length, respectively. According to Fig. 3(b), the wavelength dependency of attenuation length was clearly confirmed. As a radiation measurement test, 1 MBq ⁹⁰Sr/⁹⁰Y radioactive point source emitting beta rays was used.

The optimization of $X1−Xj$ was carried out using the generalized reduced gradient method [8], which is the standard equipment in the Microsoft excel solver. This algorithm uses the gradient of the objective function from the initial condition and identifies the optimum solution whereby partial derivatives become zero (Fig. 4). In this method, the global optimum solution may not be obtained depending on the initial value and objective function form. Therefore, the optimum solution was searched from multiple initial values using the multistart function of the solver.

Figure 5 shows the wavelength channel used for the unfolding analysis. Three patterns of wavelength channel were employed to evaluate the effect of attenuation length on the accuracy of the spatial distribution estimation. Pattern #1 is the combination of the wavelength channel with a short attenuation length, pattern #2 is the combination of the wavelength channel with a long attenuation length, and pattern #3 is the combination of the wavelength channel with short and long attenuation lengths. Fiber position index $j$ was divided into 20 parts, which indicates that the source position can be estimated every 50 cm for a 10 m long PSF.

Figure 6 shows the wavelength spectra observed at the fiber edge when the beta-rays from 1 MBq ⁹⁰Sr/⁹⁰Y radioactive point source and UV excitation source were irradiated 0.5 m away from the fiber edge. The integration time of ⁹⁰Sr/⁹⁰Y and UV irradiations were set at 10 min and 500 ms, respectively. Both spectra were normalized at the wavelength range from 460 to 470 nm. The wavelength spectrum originating from beta-ray irradiation by ⁹⁰Sr/⁹⁰Y was clearly confirmed by the integration time of 10 min. Additionally, the wavelength spectra obtained by ⁹⁰Sr/⁹⁰Y and UV irradiation showed good agreement. This result suggests that the wavelength spectrum obtained by the UV irradiation can be applied as the response function.

Figure 7 shows results of the spectra unfolding when the PSF was irradiated with UV light at 1.5 and 7.5 m from the fiber edge. Wavelength channel pattern #3 in Fig. 5 was applied for analysis. Wavelength spectra with the irradiation point of 1.5 and 7.5 m were used as inputs.

According to Fig. 7, the irradiated positions were estimated reasonably. However, the result of the 7.5 m irradiation point was estimated broadly compared with that of the 1.5 m irradiation point. This result suggests that the spatial resolution of the proposed method depends on the irradiated position. The position dependency of spatial resolution may be due to the disappearance of the shorter attenuation length component. Figures 8(b) and 8(c) compare the wavelength spectrum with point irradiation and multiple point irradiation described in Fig. 8(a). The wavelength spectrum with multiple point irradiation was imitated by the summation of the response function. According to Figs. 8(b) and 8(c), the difference between the spectrum of point irradiation and multiple point irradiation appears to be small. However, as shown in Fig. 8(d), the ratio of the wavelength spectrum of 1.5 m point irradiation to the multiple point irradiation of 1.0 m from 2.0 m is clearly below 1.0 in the wavelength region of shorter attenuation length (see the shaded part). However, as shown in Fig. 8(c), this short attenuation length component already disappeared at the 7.5 m position, and there is almost no difference between the wavelength spectrum of 7.5 m point irradiation and the multiple point irradiation of 7.0 m from 8.0 m. This feature suggests the difficulty involved in estimating the source position of 7.5 m and source position 7.0 + 7.5 + 8.0 m separately, leading to the position dependence of spatial resolution.

Figure 9 shows results of the spectra unfolding when the UV rays were irradiated at multiple points of the PSF. Wavelength channel pattern #1 to pattern #3 in Fig. 5 was applied for analysis. According to Fig. 9, the estimated position using the wavelength channel pattern #1 and pattern #3 showed good agreement with the actual irradiated position. However, as shown in Figs. 9(a-2) and 9(b-2), the estimation using the wavelength channel pattern #2 showed poor agreement with the actual irradiated position. The estimated irradiated position indicates no irradiation at the position where irradiation was conducted. Consequently, to estimate the complex irradiation distribution accurately, a wavelength with short attenuation length (attenuation length shorter than approximately 4.0 m in the analysis condition of this study) should be included in the analysis.

According to results of the spectra unfolding using the wavelength channel pattern #1 and pattern #3 in Fig. 9, position dependency of the spatial resolution was confirmed, which corresponds to the same feature discussed in Fig. 8. Although the spatial resolution of the proposed method has position dependency, the actual irradiation positions were reproduced reasonably using wavelength channel #1 and #3.

Figure 10 shows results of the spectra unfolding when the PSF was irradiated with beta-rays from 1 MBq ⁹⁰Sr/⁹⁰Y radioactive point source at multiple points. Wavelength channel pattern #1 and #3 in Fig. 5 were applied for the analysis. According to the unfolding results using the wavelength channel Pattern #1 in Figs. 10(a) and 10(d), the actual source position can be estimated roughly. However, in Fig. 10(b), the sources at 1.5 and 3.5 m were almost overlooked and estimated at the 2.5 m position. In Fig. 10(c), the source at 5.5 m was overlooked and misidentified at the 9.5 m and 10.0 m positions. On the other hand, according to the unfolding results using the wavelength channel pattern #3 in Fig. 10, the source positions were successfully estimated near the actual source position, which was overlooked in the analysis using the wavelength channel pattern #1, as shown in Figs. 10(b) and 10(c). Moreover, as shown in Figs. 10(a) and 10(c), the separation of sources at the 0.5 and 1.5 m positions was successful. To understand the reason for the difference between the unfolding results using the wavelength channels #1 and #3, further investigation should be conducted. The difference in the attenuation lengths of adjacent wavelength channels may be the reason. Although slight misidentifications of the source position near 9.0 to 10.0 m appeared using the wavelength channel pattern #3 in Figs. 10(a) and 10(b), we confirmed that the source position can be estimated roughly using the proposed method by the appropriate setting of the wavelength channel used for the unfolding analysis.

We proposed a new approach to measure the radiation distribution one-dimensionally using an optical fiber sensor based on the wavelength spectrum unfolding for the application under a high dose rate environment. By using the fact that the attenuation length of light along the fiber depends on its wavelength, we demonstrated that the radiation distribution can be estimated reversely using the unfolding method through basic experiments using an UV excitation source and a ⁹⁰Sr/⁹⁰Y radioactive point source. The proposed method has position dependency in spatial resolution, which shows worse spatial resolution at the long distance from the fiber edge. It was suggested that this feature is due to the disappearance of the shorter attenuation length component. There is a possibility that the position dependency of the spatial resolution can be improved by optimizing the wavelength channel used for analysis and controlling the wavelength dependence of light transmittance. Despite these features, it was confirmed that radiation distribution can be estimated reasonably using light with a short attenuation length for unfolding analysis.

For future work, the application of the quartz optical fiber should be considered as the optical fiber sensor for the improvement of radiation tolerance. The wavelength spectrum can also be obtained using the quartz optical fiber by the detection of Cerenkov photons. Thus, the method proposed in this study may apply to the quartz optical fiber.

• JAEA Nuclear Energy S&T and Human Resource Development Project (Grant No. JPJA19B19206529; Funder ID: 10.13039/501100005118).

$f(x)$ =

radiation intensity distribution

$i$ =

wavelength channel index

$I(λ)$ =

initial wavelength spectrum

$j$ =

position index

$Ri,j$ =

response function that stands for the light intensity of the $i$th (⁠$i=1, 10$⁠) wavelength channel when irradiating the $j$th (⁠$j=1, 20$⁠) fiber position with radiation

$Si$ =

intensity of light at the $i$th wavelength channel in the observed wavelength spectrum

$x$ =

position, m

$Xj$ =

radiation intensity at position $j$

Greek Symbols

$λ$ =

wavelength, nm

$μ$ =

attenuation length, m

$σ$ =

standard deviation

$Δ$ =

sum of squared differences

Abbreviations

FDNPS =

Fukushima Daiichi nuclear power station

PSF =

plastic scintillating fiber

TOF =

time-of-flight

UV =

ultraviolet