Transient Response of Optical Fibers Exposed to Pulsed Radiations

Bremsstrahlung and electron-beam radiation experiments have been performed to measure transient radiation darkening and recovery at room temperature in seven commercial optical glass fibers. Several of these fibers were also tested at temperatures approaching the temperature of dry ice, 216 K. Fiber responses measured during and immediately after the radiation pulse have been reduced to delta function response parameters using curve fitting techniques. Most fibers examined demonstrated a complex recovery history that suggests the presence of multiple anneal mechanisms with significantly different characteristic times. At room temperature, peak darkening coefficients ranged from 10-3 to 10-5 dB/m-R; typical recovery rates ranged from 104 to 107 s-1. At temperatures between 216 and 247 K, most fibers demonstrated increased darkening and markedly slowed recovery rates, althotugh exceptions to this result were observed.


Introduction
An optical fiber waveguide exposed to a fast pulse of ionizing radiation experiences a transient optical darkening that can be very much greater than the permanent darkening produced by the same pulse. The amplitude and time history of the transient effect appe ar to depend only on the radiation dose and time hist.ory, the fiber temperature, and the fiber compositiori. 1 4 There are applications for fiber optics (e.gj., in nuclear radiation effects tests and nuclear diagnostics tests) where transient darkening is a serious concern in the time interval of 0 to 500 ns, measured from the beginning of the radiation pulse. This paper describes an experiment performed to characterize early-time darkening in a number of commercia.lly available fibers. An Ion Physics FX 45 Flash X-r-ay machine, operating initially in the bremsstrahlunag mode, and later in the electron-beam mode, was us ad to obtain pulsed ionization doses in the test samples, all of which were glass-core fibers. Three ITT fibers were examined: the step-index T103 and T3-03 (a PCS fiber), and the graded-index T223. Four Ccrning fibers were also tested: the step-index shLort-distance fiber (SD), and the graded-index IVPO, OVTPO, and double-window fiber (DW). All fibers were e:<posed at room temperature, 295 K, and in addition, t ie IVPO, OVPO, DW, T223, and T303 fibers were also e'Kposed at temperatures approaching the temperature of dry ice, 216 K. The measured data have been reduced t.o empirical response parameters that are independent cf sample length and the radiation dose time history.
Eixperimental Configuration Fig. 1 shows the nominal configuration of the apparatus used to measure the transient radiation response of optical fibers. Test samples were noncabled fibers carrying the manufacturer's standard protective coatings, all of which were "thin" to the radiations used in the experiment. Two provisions were made for circumventing the optical interference expected during the radiation pulse: 1) interference filters with passbands of 10 nm FWHM centered on the laser wavelength (820 nm) were interposed between the output ends of the test fibers and the avalanche pho-* Sponsored by the Defense Nuclear Agency under Subtask J24AAXYX959 +Harry Diamond Laboratories Defense Nuclear Agency, Washington, D.C. todiodes (APDs) in the receiver, thus blocking most of the Cerenkov light, and (2) a high frequency sine-wave modulation of the laser signal was used to provide a direct means of tracking the darkening of the fiber through and beyond the radiation pulse. In the electron-beam experiments, the five fibers were cooled by the simple expedient of holding a cylindrical block of dry ice in forced contact with the rear surface of the fiber holder. The front surface of this aluminum piece was provided with a shallow fiber groove (20 cm circumference), six small indentations inside and outside the groove circle for implacement of bare thermoluminescent dosimeters (TLDs), and a space for mounting an iron-constantin thermocouple junction in direct contact with the metal surface near the fiber.
During the bremsstrahlung tests, several halfkiloroentgen exposures of a given fiber segment were made to compile a complete response history. This was possible because the radiation pulses were very reproducible, and the prior dose history had no apparent affect on the transient response of the sample, the first 120 ns of which was recorded on every shot.
Moreover, no cumulative permanent darkening after several exposures was observed. During the electronbeam tests, however, the radiation fluence per shot was in the range of 40-60 kR, and permanent darkening was not negligible. For this reason, a fresh fiber segment was advanced into the holder for each electron-beam exposure. In both bremsstrahlung and electron-beam phases of the experiment, TLDs were routinely removed for readout after each radiation pulse. The TLDs used were CaF:Mn dosimeters; they were read out in equipment calibrated to yield radiation fluence rather than dose. Conversion of fluence to dose in the fibers requires a quantitative specification of the fiber core compositions, information generally unavailable because of its proprietary nature. For this reason, the radiation environments are quantified below in terms of fluence (roentgens, R) and energy spectrum rather than dose in rads, even though the latter is more fundamental to the darkening effect. Similarly, the results of the experiment are reported in terms of fluence.
U.S. Government work not protected by U.S. copyright. The Ion Physics FX 45 radiation source was charged to 4.1 MV. The machine's bremsstrahlung pulse shape (20 ns FWHM) was measured with an ionization chamber located in the test volume and is given in fig. 2. The photon fluences measured with TLDs at the sample location ranged from 0.4 to 0.7 kR. The photon energy distribution is given in fig. 3.5 In the electron-beam phase of the experiment, the FX 45 was first operated in its standard mode, producing a nominal 45-kR fluence at the fiber location. The waveform of this radiation pulse is given in fig. 4 and has a pulse width of 18 ns (FWHM). It was measured with a current sensor located in the diode space of the machine and shows evidence of a cavity ring at late time, which should not be considered a part of the electron output. The electron energy spectrum for this pulse is given in fig. 5.6 The FX 45 was then operated in a crowbarred mode, which shortens the pulse to 12 ns (FWHM) as shown in fig. 6 and provides the electron energy spectrum given in fig. 7.6 The radiation fluence at the fiber obtained in this case was approximately 55 kR. in such a way that the earliest part of the trace recorded on the shot film gives the preradiation amplitude of the transmitted sine-wave, followed directly by the portion of the sine wave signal that shows the effects of radiation darkening of the fiber. In this way, unequivocal attenuation ratios can be drawn from each shot record, with no concern for gain drift in the system from one shot to the next. The dynamic range of measurements made on both dual beam and single beam oscilloscopes is approximately 20 dB. This is less than what might otherwise be expected because oscilloscope preamp gains had to be reduced to accommodate the APD's voltage shift at zero time; this effect, due to the sudden reduction of light intensity at the receiver caused by fiber darkening, is apparent in fig. 8.

Data Reduction
A quantitative theoretical model explaining radiation-induced transient darkening in fiber glasses has not been established in the literature. For this reason, an empirical approach to data reduction was chosen for this experiment. Its objective was to extract differential response parameters from the raw data. In this way, the integrating influences of finite sample length and radiation pulse length on the measured result could be stripped away, leaving quantities that more directly reflect the fundamental radiation response of the glass. To obtain these parameters from the raw data, the following sequence was followed: (1) an empirical model was chosen for the intrinsic response of the glass to a finite radiation fluence delivered in an infinitely short pulse, (2) this expression was then convolved over the length of the fiber sample, (3) the length convolution was then folded over the normalized radiation time history, (4) the double convolution was compared to the measured attenuation history, and (5) the model parameters were adjusted and the procedure was repeated until a satisfactory curve fit was obtained. The following paragraphs describe how these steps were carried out.
After an inspection of data produced during the bremsstrahlung phase, ag empirica model was selected with the form (1-a0e 0 ) (1-a1e 1 The subscripts refer to what appear to be two separable components in the observed fiber recoveries, each component characterized by its own recovery time constant. The convolution of this expression over the length of the fiber sample was accomplished according to the expres- ao, A0, a,, and A1 represent the component recovery processes; and the product operator X provides for the progressive attenuation of the light signal as it travels through the fiber. The next step was to convolve T over the normalized radiation time history F(t): where A(t) is now the amplitude modulation of the signal after its transmission through a length L of the fiber, which has received a total fluence D during the time interval of the radiation pulse. This calculated modulation history is then compared with the measured result, and the a's and A's are adjusted until a suitable curve fit is obtained. The final values for the a's depend on the total radiation fluence D; this de-Abadence is removed by transforming the a's into the desired differential parameters according to the expression: where Ai is in units of dB/m-R, and P is the number   Figs. 11 and 12 are curve-fit comparisons for the SD and T103 fibers, respectively. These data were obtained at room temperature during high-fluence electron-beam exposures of fibers 20 cm long. There is a 20-30% agreement in the time constants and a factor of three agreement in the attenuation constants obtained for the T103 fiber in the low-fluence bremsstrahlung and high-fluence electron-beam experiments. The SD fiber was one of the two slowest fibers tested at room temperature and showed an unexplained "knee" 30 ms into its recovery following exposure to the electronbeam pulse.
The remaining figures also pertain to the electron-beam experiment and show not only comparisons between room temperature measurements and their curve fits for IVP0, OVP0, DW, T223 and T303 fibers, but also similar comparisons for data obtained when these fibers were cooled with dry ice. give the T223 responses in the earlyand late-time regimes, respectively. The effect of cooling the fiber 70 K is pronounced in both figures. The early portion of the curve fit in fig.  13 was obtained by selecting an artificially small time step for the convolution calculation, 0.2 ns. An interesting aspect of the experimental data shown in fig. 14 is that once the cooled fiber began to transmit a measurable signal (at 0.5 s), its recovery rate was very close to the initial recovery rate of the room-temperature fiber.  Fig. 17 shows transient darkening in the OVPO fiber. Like the short-distance fiber, it is very slow to recover at room temperature and shows an even slower recovery at 235 K. The cold-temperature response for this fiber was unusual in that a measurable signal was transmitted during and after the radiation pulse, but its amplitude did not change noticeably until almost a second after the pitse. The   The slow sweep record (500 ns/div) provided response data after 100 ns, which is plotted as solid dots in fig. 18. eight and a half minutes later, a second radiation pulse was applied to the same cooled fiber segment and the same set of TLDs that were exposed in the preceding shot (the fluence given in fig. 18 is half of the two shot TLD readout). The open asterisks show the attenuation history measured during this second shot, normalized to the signal amplitude observed just nanoseconds before the arrival of the second pulse of radiation. At that moment, the fiber showed residual darkening from the first exposure in that the signal amplitude prior to the second pulse was 75% of the value measured before the first exposure. A calibration carried out seconds after the second shot showed that the signal had dropped another 2% to 73%. The transient responses measured on the two cold-temperature irradiations are in good agreement with each other, but, unlike the cold-temperature responses observed in the other fibers, appear smaller in amplitude than the room-temperature response.  Fig. 19 Early responses of T-303 fiber to electron fiber Fig. 19 presents an early-time comparison for the T303 fiber. The time history of the transient darkening measured at 237 K is qualitatively very similar to the room-temperature result, but the darkening amplitude is greater. Not unexpectedly, this low-bandwidth pure-glass-core fiber showed less sensitivity to radiation than the other fibers tested, all of which contained one or more dopants.
The uncertainty bars in the figures discussed above are estimates based on the trace resolution on the film, the random noise fluctuations in the laser output, and the slope of the sine-wave signal. In general, the largest measurement uncertainty belonged to the portion of the data trace containing the avalanche photodiode's voltage shift and the Cerenkov noise pulse. The response parameters used to curve fit the measurements are listed in the table. Each entry carries a superscript indicating the estimated uncertainty in the parameter value. It should be noted that the rather large uncertainties listed for certain of the Ai's are the results of small-signal measurements. When these measurements are converted to attenuation ratios and then to their logarithmic equivalents, their uncertainties take on large values.
The largest uncertainties in the A's are a consequence of curve fitting a slow recovery component to data of limited time duration. In such a case, a relatively wide range of curve slopes appears to fit th e measurement.

Conclusion
The radiation responses of seven fiber s have been presented, both graphically as measured tirAe histories, and in terms of the A, X parameter va'lues required to make an exponential recovery mod.el fit the measurements. The A parameters are normal.ized to unit sample length and radiation fluence and re present the initial peak optical attenuation at 820 mni in a unit length of fiber exposed to a delta function radiation pulse. The X parameters are the exponential time constants that describe the fiber's recovery from the instantaneous darkening producing the optical attenuation.
If a fiber's response to pulsed rad:iation is linear in length and in dose up to the levels obtained in this experiment, then the sets of normalized response parameters listed in the table for the given fiber and temperature should agree, at l east to within measurement uncertainties. The comparis;ons.of parameter sets drawn from the 4 m and 20 m eBxposures of T103 and T223 fibers for similar bremss-trahlung pulses show agreement to the extent in that corresponding uncertainty intervals overlap. From this, one may conclude that the length and time convc)lutions described earlier offer a use-ful means oiaccounting for the effects of fiber length and radiation time history on transmitted signals. In general, tVae same cannot be said for the influence of radiation fluence. In the case of the T103 and T223 fibers, the parameter sets drawn from the low-fluence brems<;trahlung and the high-fluence electron beam experiment<; at room temperature agree within the uncertai-ities. But for the Corning IVPO and OVP0 fibers, recovery time constants observed in the electron-beam experiments are clearly longer than those observed during the bremsstrahlung experiments. The cons;ensus in the literature cited earlier is that induiced darkening is dependent on dose, irrespective of the radiation component delivering the dose. If this is the case, then the longer recovery constants for the electron beam experiments are attributable to the very large doses delivered with the electron be!am, not the fact that electrons were used to darken the fibers. In retrospect, it is clear that the elactron-beam doses were too large to obtain accurate eairly-time attenuation measurements for several of the more radiationsensitive fibers. The most accurate measurements and curve fits were obtained when the peaak attenuation factors fell in the range of 0.3 to 0.7.
The effects of temperature on the radiation responses of the fibers also followed a complex pattern. For the most part, colder temperatvLres produced an increase in the peak attenuation, sucTgesting that recovery mechanisms may be present with characteristic times much shorter than the radiation pulse length, at least at room temperature. Recovery rates for most of the fibers were slower at the cold er temperatures, as expected. An apparent exception i-s the Corning double-window fiber, which showed an anomalous feature at the colder temperature. While the fast-component attenuation parameter, AO, for this fiber is significantly larger at cold temperaturE than at room temperature, its recovery rate A0 apzpears to be a factor of 10 faster than the room-temper-ature value (this faster recovery explains why in fig. 18 the coldtemperature transmission curve l ies above the roomtemperature result). Another ex,-eption was the T303 fiber, which showed almost no difference in the recovery rates observed at the t:hree test tempera-tures. The variety of behaviors observed when nominally simil,ar fibers were cooled makes an explanation of results in terms of simple physical models quite diffic-ult, if not impossible. Certainly, an extension of this work to measure fiber responses across a continuous range of temperatures and radiation fluences would 'be very useful in modeling phy-sical processes, as would be the identification of the roles the various ctlopants and impurities play in determining the net response of the fiber to pulsed radiation.