1. Introduction

Ionizing radiation produces point defects (radiation-induced color centers, RICC) in the optical fiber glass network, which absorb the light propagating in the fiber. Therefore, by measuring the value of the radiation-induced attenuation (RIA) of light in the fiber, it is possible, in principle, to determine the absorbed radiation dose [1], just as with the conventional glass dosimeters based on the coloration effect [2]. For a fiber glass chemical composition to be suitable for dosimetry, it should ensure, first of all, a relatively large RIA. It is also necessary that RIA should be independent of temperature and dose rate and should not decay on termination of irradiation. The latter requirements can be met at once, provided the effect of RICC decay is negligibly small as compared to the effect of their formation (i.e. RICC should be long-lived).

Ge-doped silica used in ordinary fibers for optical communication and undoped silica are unsuitable for fiber dosimetry, because RIA is too small and RICC are not sufficiently long-lived. Silica fibers with various dopants, such as lead [3], rare-earths [4,5], and aluminum [5,6], have been shown to be promising for dosimetry; however, most papers on the subject have been devoted to phosphorus-doped silica fibers (e.g. see [1,4,6–16]). Unfortunately, RICC decay does occur in P-doped silica fibers as well [7] to make the RIA dose-rate- [9,13] and temperature-dependent [7–9] in the course of irradiation and time-dependent in the course of post-irradiation recovery [4,9,14] virtually at all wavelengths.

The lifetime of the numerous RICC in P-doped silica is different; in addition, some RICC are known to interconvert during the decay process [7]. Such RICC peculiarities allow one to suppress the effect of RICC decay by measuring RIA at a properly selected single wavelength or at several wavelengths. In particular, radiation-induced phosphorus oxygen hole center (POHC) with a few absorption bands peaking in the visible region [13] converts into the P1-center with an absorption band centered at 1.57 µm [7]. As a result, at the interface between the POHC and the P1-center absorption bands, in the wavelength region ~1.2–1.4 µm, RIA does not decrease on termination of irradiation and, consequently, must not strongly depend on dose rate and temperature. Dose reconstruction from RIA measured in a P-doped silica fiber at λ = 1.37 µm with an accuracy of no less than 20% was demonstrated experimentally in ref [8]. Unfortunately, radiation sensitivity of P-doped silica fibers in the near-IR region reaches its minimum (< 1·10⁻³ dB·m⁻¹·Gy⁻¹ [1,8]), which is too low for most practical applications.

It is also possible to avoid the RICC decay effect by using the RIA difference Δα at two properly selected wavelengths to calculate the dose. In this way, in ref [9]. by measuring Δα in a P-doped fiber at wavelengths of 880 and 770 nm, rather accurate dose reconstruction was obtained with a much higher radiation sensitivity (20·10⁻³ dB·m⁻¹·Gy⁻¹) than that obtained from RIA at λ = 1.37 µm in ref [8].

Despite numerous investigations, there has been, to our knowledge, only one practical realization of fiber-optic dosimeter [16], in which P-doped silica fibers were used to determine the dose distribution along the wall of TESLA Particle Accelerator by measuring the RIA distribution along the fiber length by means of optical time domain reflectometry (OTDR) [17] and by measuring RIA in fiber spools situated in separate important places of the accelerator [16]. It should be noted that the measurements were performed at a single wavelength of 850 or 670 nm, at which the RICC decay significantly influences the dose determination accuracy.

Radiation sensitivity of P-doped silica fibers increases with decreasing wavelength and reaches maximum in the visible and UV-regions, where most RICC manifest themselves [12–15]. A sensitivity as high as ~0.6 dB·m⁻¹·Gy⁻¹ was measured in a P-doped silica fiber at λ = 300 nm [13]. However, according to [14], the absorption band at 4.21 eV determining RIA at λ = 300 nm significantly decays on termination of irradiation; therefore, RIA at λ = 300 nm is likely to vary with dose rate and temperature.

One may assume that RICC interplay in the visible range will make it possible to find two wavelengths at which RIA difference Δα will not be subject to post-irradiation fading. It follows from ref. [14] that such two wavelengths do exist: Δα does not vary on termination of irradiation, provided the two wavelengths are equally spaced from λ_0$\simeq$440 nm: λ₀–λ₁ = λ₂–λ₀ (see Fig. 2(b) of [14]). An explanation for this interesting RIA feature can be gained from the analysis of the behavior of the RICC determining RIA in the visible region presented in refs [7,13,14]. In particular, it was shown therein that RIA at λ>440 nm is mainly associated with one of the two POHC forms, whereas at λ<440 nm, with the other. The two POHC forms (POHC^s and POHC^m according to [13]) possess absorption bands centered at 2.54 eV (488 nm) and 3.17 eV (391 nm), respectively, the latter band amplitude being ~40% greater during irradiation [14]. What is important is that the difference of the amplitudes of these bands remains virtually unchanged in the course of their post-irradiation decay (see Fig. 4 of [14]) making it possible to select two appropriate wavelengths for Δα-dosimetry.

We fabricated a tentative version of a fiber-optic dosimeter operating at two wavelengths nearly equally spaced from λ₀ = 440 nm: 413 and 470 nm. In this paper, we describe the dosimeter design and the results of its first test to compare the dose sensitivity, accuracy, dynamic range and dose-rate dependence when using RIA at a single wavelength or Δα.

2. Dosimeter design

Figure 1 shows a schematic of our fiber-optic dosimeter. Two LEDs 6 and 7 with central wavelengths λ₁ = 413 nm and λ₂ = 470 nm (full width at half maximum of the emission bands of ~20 nm) alternately launched light into fiber couplers 8 and 9 fusion spliced with the sensor head (fiber pigtail), which was formed by an undoped-silica-core F-doped-silica-cladding fiber 13, 12 m in length, fusion spliced, in turn, with a P-doped-silica-core undoped-silica-cladding fiber 14. The P-doped was MCVD-produced and had the core and cladding diameters of 32 and 125 µm and the numerical aperture NA = 0.16 (the P₂O₅ concentration was ~9 mol. %). Its length of 146 cm was taken rather arbitrarily, but with regard to the expected RIA under our experimental conditions indicated below. An aluminum mirror layer 15 was applied onto the endface of the P-doped fiber; thus, the probe light was reflected back from the mirror layer to arrive at a photoreceiver 16. The intensities of the LEDs and of the probe signal arriving from the sensor head as well as the dark current of the photoreceivers 10 and 16 were measured continually with a certain time step to determine RIA at the two wavelengths α(λ₁) and α(λ₂) and their difference Δα = α(λ₁)–α(λ₂). The fibers and the couplers were multimode with a step-index profile, but their core diameters and numerical apertures were somewhat different.

 

Fig. 1 Dosimeter schematic: 1 – microcontroller; 2 and 3 – electric current sources; 4 and 5 – switches; 6 and 7 – LEDs; 8 – Y-coupler; 9 – X-coupler; 10 and 16 – photoreceivers; 11 and 17 – amplifiers; 12 and 18 – analog-digital convertors; 13 – undoped-silica-core fiber; 14 – P-doped fiber; 15 – aluminum mirror layer on the P-doped fiber endface, 19 – USB interface.

Download Full Size | PDF

Light of our spectral range is known to both photobleach and induce RICC (e.g. see Fig. 7 of [18]). To rule out those unwanted effects, the input light intensity in the P-doped fiber was made as low as possible: 0.6 and 0.05 nW at 413 and 470 nm, respectively. The 413-nm light intensity was made an order of magnitude lower so as to have comparable electric signals of our silicon photoreceivers at both the wavelengths.

3. Experimental procedures

The dosimeter was tested at a ⁶⁰Co gamma-ray source, which was constituted by 16 parallel cobalt rods, 50 cm in height and 0.9 cm in diameter. The rods were situated in such a way that after lifting up from the underground lead shield they formed the lateral surface of a cylinder, 22 cm in diameter, neighboring rods being equidistant.

In experiment No. 1, the P-doped fiber was situated 226 cm away from the cylinder axis and was additionally shielded by a concrete wall. In experiment No. 2, it was situated 126 cm away from the cylinder axis with no shielding. The dose rate measured at the two selected points by glass dosimeters proved to be 0.00064 and 0.0066 Gy/s in experiments No. 1 and 2, respectively. In both the experiments, the P-doped fiber was loosely spooled to a diameter of ~10 cm, and we believe the dose rate uniformity over the fiber spools was no worse than 1%.

In experiment No. 1, we made two cycles of irradiation and post-irradiation recovery. Next, we replaced the sensor head (fiber pigtail) and made experiment No. 2, which included only one cycle of irradiation and recovery. The time step in both the experiments was 16 s.

We believe the P-doped fiber temperature varied during the experiments by no greater than ± 2 °C, which did not noticeably influence our measurements.

4. Results and discussion

Figures 2(a) and 2(b) show the RIA evolution at the two working wavelengths (α(λ₁) and α(λ₂)) and their difference Δα in experiments No. 1 and 2, respectively.

 

Fig. 2 Experiment No. 1 (a) and No. 2 (b): RIA measured at wavelengths of 413 and 470 nm (left Y-axes) and the RIA difference at these wavelengths (Δα = α(413 nm) – α(470 nm), right Y-axes). Label “γ” indicates periods of irradiation each followed by a period of recovery.

Download Full Size | PDF

One can see from Fig. 2(a) that the three curves behave similarly during two successive runs of irradiation and recovery, the RIA at individual wavelengths being ~5–6 times larger than their difference and far less noisy. At the same time, the post-irradiation decay is tangibly less for the Δα-curve than for the α(λ₁)- and α(λ₂)-curves, just what we tried to achieve. Thus, taking the RIA difference to calculate the dose will allow suppression of the RIA post-irradiation decay, although, the suppression in our case is still incomplete. Owing to the above peculiarities of the lifetimes of the POHC^s and POHC^m, it is not impossible that a more precise selection of the two working wavelengths will lead to more complete decay suppression. For this, measuring RIA with the help of a Si-diode-array spectrometer would be of much use.

It is no wonder that the rise of our three curves with dose is sublinear, which is most evident in the Δα-curve [Fig. 2(a)]. As with fibers of any other chemical composition, RIA growth with dose in P-doped fibers is inherently sublinear (e.g. see [4,8]) and can be considered as linear only at the very beginning of irradiation. However, this fact does not call the fiber dosimeter concept into question: it is just necessary that RIA be uniquely related to dose.

Experiment No. 2 allowed us to assess the dynamic range limit of the dosimeter [Fig. 2(b)]. We notice that as soon as α(λ₁) and Δα reach ~10 and 1.8 dB, respectively, the Δα-curve loses smoothness and we turn out beyond the dynamic range. Closer inspection shows that the Δα-curve distortions are caused by the distortions of the α(413 nm)-curve, which are virtually invisible in Fig. 2(b). Obviously, such a narrow dynamic range is, first of all, due to the very low probe light intensity. Neither RICC photo-bleaching nor RICC photo-induction appear to have influenced our experiments, and, therefore, the light intensity can be increased. The admissible level of the intensity is to be established experimentally.

Figure 3(a) shows the α(413 nm)- and Δα-dependences on dose for the points of both the experiments taken in the top position of the γ-source. In calculating the dose values, we also took into account the dose absorbed in the process of lifting the γ-source up and down, which allowed us to combine the points of both the irradiation runs of experiment No. 1 on one graph.

 

Fig. 3 (a) Dose dependences of α(λ = 413 nm) and Δα measured in experiments No. 1 (dots) and No.2 (open circles) in the top position of the ⁶⁰Co source (the red straight lines give the power-law approximations); (b) dose reconstructed from the Δα-values (open circles) and α(λ = 413 nm)-values (dots) measured in experiment No. 1, the straight red lines depicting the real dose.

Download Full Size | PDF

One can see that the Δα-points of both the experiments obey the same power-law dependence on dose (exponent of 0.903 ± 0.006), except the last high-dose points of experiment No. 2 falling beyond the dosimeter dynamic range. The excellent coincidence of both the experiments testifies that the dosimeter readings will not depend on dose rate variation by as great as an order of magnitude, and the above power-law dependence can be used as the calibration algorithm. The sensitivity of the Δα-dosimeter (~0.75 dB·m⁻¹·Gy⁻¹) is only a little less than that obtained at λ = 300 nm in ref [13]. (it would be ~1.2 dB·m⁻¹·Gy⁻¹ in the case of two probe light passes through the fiber as is in our dosimeter).

The α(413 nm)-dependence in experiment No. 1 is described by a power-law with a little different exponent of 0.925 ± 0.005 [Fig. 3(a)]. What is significant is that the Experiment No. 2 points lie 10%-17% higher than those of experiment No. 1 and are approximated with a little higher exponent of 0.936 ± 0.004. In other words, α(413 nm) depends on dose rate. The sensitivity of the α(413 nm)-dosimeter would be a factor of ~5.4 higher – 4.0 dB·m⁻¹·Gy⁻¹

The α(470 nm)-dependence is very similar to the α(413 nm)-dependence (not shown for figure legibility reasons) and also features larger RIA in experiment No. 2 than in No. 1

In Fig. 3(b), we reconstructed the dose in experiment No. 1 from the measured Δα- and α(413 nm)-values and using the corresponding power-law coefficients in Fig. 3(a). The average error of Δα- and α(413 nm)-dose reconstruction due to noise amounted to ± 3% and ± 0.4%, respectively. Apart from noise, there was a systematic error due to decay on termination of irradiation, which was ~4% during the first recovery and ~3% during the second one for the Δα-dose reconstruction and ~7% and ~8% for the α(413 nm)-dosereconstruction. The maximum overall error proved to be about ± 8% for both the curves.

Although with our tentative dosimeter version both Δα and α(413 nm) yielded the same accuracy of dose reconstruction, we, nevertheless, believe Δα-dosimetry to be more promising. The accuracy of the latter can be improved by increasing the light intensity, which must lead not only to dynamic range extension, but also to noise reduction. Secondly, it is possible that it can be improved by optimizing the working wavelengths. Lastly, noise can be reduced by additional data averaging at each time point. At the same time, there are no potentialities for improvement of the dosimetry based on RIA at a single wavelength (413 or 470 nm), because it depends on dose rate and exhibits noticeable post-irradiation decay.

From Fig. 3(b) the dose resolution and the minimum detectable dose of the Δα-dosimeter can be assessed as ~0.01 Gy. Clearly, this value can be decreased not only by reducing noise, but also by increasing the P-doped fiber length. Typically, the optical loss in as-drawn P-doped silica fibers at λ = 413 nm is ~70 dB/km and at λ = 470 nm ~20 dB/km [19]. Therefore, it is possible to increase the P-doped fiber length by two orders of magnitude to reduce the minimal detectable dose to ~10⁻⁴ Gy.

Further experiments on Δα-dosimetry should include, among other things, investigation of the possible temperature dependence and of post-irradiation decay at longer times.

5. Conclusion

RIA at λ = 413 nm and λ = 470 nm and their difference Δα were found to grow sublinearly with dose with a power-law exponent of 0.90-0.94. In contrast to RIA at the individual wavelengths, Δα proved not to vary with dose-rate variation and to decay on termination of irradiation only slightly, which was evidently due to the favorable relation of the lifetimes of the two RICC dominating at the two wavelengths, which was found out elsewhere.

To rule out RICC photo-induction and photo-bleaching, we reduced the probe light intensity to an extremely low level of 0.6-0.05 nW, which turned out to be the main cause of the dosimeter narrow dynamic range and of the Δα-noise. We believe the light intensity can be increased without affecting RICC, which will extend the dynamic range and lower the noise.

Although the dose reconstruction accuracy from α(λ = 413 nm) and Δα was found to be roughly the same ( ± 8%), we expect Δα-dosimetry to be more promising, because the random point-to-point noise – the main cause of inaccuracy in Δα-dosimetry – can be reduced by increasing the light intensity and by data averaging, whereas the accuracy limitation in using RIA at an individual wavelength to calculate dose appears to be insuperable.