Boron carbide based solid state neutron detectors: The effects of bias and time constant on detection efficiency

Neutron detection in thick boron carbide(BC)/n-type Si heterojunction diodes shows a threefold increase in efficiency with applied bias and longer time constants. The improved efficiencies resulting from long time constants have been conclusively linked to the much longer charge collection times in the BC layer. Neutron detection signals from both the p-type BC layer and the n-type Si side of the heterojunction diode are observed, with comparable efficiencies. Collectively, these provide strong evidence that the semiconducting BC layer plays an active role in neutron detection, both in neutron capture and in charge generation and collection. 1 digitalcommons.unl.edu 2 Hong et al . in Journal of Phys ics D: Appl ied Phys ics 43 (2010) This fundamental difference between the two types of detectors leads to differing efficiencies, pulse height signatures and detection thresholds as shown both in GEANT4 Monte Carlo simulations [15] and a simple physical model [14, 15]. There are two striking differences—the models indicate that for a BC/Si diode, the energy deposition spectra are weighted to higher energies, with a cut-off below the lowest energy signature at 0.84 MeV, whereas in the BC conversion layer, the pulse height spectra are weighted towards lower energies with a cut-off at the highest energy of 1.78 MeV. In both cases, a thickness of 1 μm BC leads to a substantial smearing of the individual peaks associated with the 4He and 7Li ions. In addition, in the case of the diode detector, peaks at the summed energies of the 7Li and 4He ion should be present, an effect absent in the case of a conversion layer detector since charge capture and neutron capture occur in two different layers. Both predictions are modified in the presence of electronic and statistical noise and both depend on the assumption of 100% charge capture, an assumption we will show is highly dependent on the parameters of the associated processing electronics and the properties of the semiconductor. In thin BC/Si heterojunctions, much of the charge capture occurs in the Si layer due to geometrical constraints, and hence the neutron detection signatures are very similar to those of conversion layer detectors. Measurements on a 232 nm thick BC/n-type Si heterojunction diode [2] resulted in an efficiency of 3.35 × 10−3, even at zero applied bias, almost exactly equal to the calculated neutron capture efficiency. The effect of increasing the bias led to small increases in the detection efficiency of less than 10%, because such a large amount of charge is liberated on neutron capture, that capturing even small fractions of it leads to a charge pulse above threshold noise [2]. Subsequent efforts on neutron detection in BC/Si heterojunction diodes have focused on the parameters important for increased efficiency. Although increasing the thickness of the BC layer must increase the neutron capture rate, the entire process from neutron capture to charge collection (in both the BC and the Si) and the subsequent pulse processing are found to play important roles in the overall detection efficiency. In this paper, we describe neutron detection experiments on thick (1.0 μm and 1.8 μm) layers of BC on n-type Si. This five to nine fold increase in thickness over previous samples leads to a greater proportion of charge being captured in the semiconducting BC layer, resulting in a larger dependence on the properties of the BC in the neutron detection efficiencies. As we demonstrate below, the charge pulse rise time is governed by the characteristics of the semiconducting material and is substantially different for the two materials. Both the applied bias voltage and the integration time constants are shown to lead to much improved detection efficiencies. The neutron capture reaction in the BC/Si heterojunction detector for thermal neutrons is described in equation (1) and illustrated in figure 1. Neutron capture occurs only in the ptype BC side, with the neutron capture site serving as the origin for highly energetic 7Li and 4He ions, emitted back-to-back. Electron–hole pairs are generated by the passage of these secondary ions through the semiconductor. The signal at either electrode arises mainly from charges generated within the depletion layer. The charge carriers are accelerated due to the internal electric field, with negative (positive) charge drifting towards the positive (negative) electrode. Depending on the ground and center pin connection of the coaxial cable, the charge pulse consists either of electrons (from the Si side) or holes (from the BC side). Hence the charge pulse we collect consists of only one type of charge. Since the mobility of carriers on the BC and Si side differ vastly, the choice of time constants has a significant impact on the charge collection, and consequently on the neutron detection efficiency. We will show that there is a significant difference in the signal, depending on whether electrons or holes are collected at the center pin. 2. Experimental details The 1.0 μm and 1.8 μm BC layer were grown using PECVD from an orthcarborane closo-1,2-dicarbadodecaborane (C2B10H12) precursor [16] on n-type Si(1 1 1) substrates (resistivity ~100 Ω cm) in a custom designed parallel plate 13.56 MHz rf PECVD reactor [1]. The base pressure was 5.3 × 10−5 Pa and the working pressure was 27 Pa Ar gas. The substrate temperature was held at 330 °C during the deposition with 30 W rf-power supplied. The film growth rate, 80 nm/10 min, was obtained by ex situ grazing incidence x-ray reflectivity. Ohmic contacts for BC and Si layers were sputter deposited using Cr and Au metal targets [1] with contact areas ranging in size from 0.785 to 19.625 mm2 on the BC side. The Si side contacts covered the entire area of the chip, ~1 cm2. The neutron source used is a Thermo Electron Corporation MP320 [17]. The 1.0 × 108 n s−1 fluence of high energy 14 MeV neutrons is moderated using a 10 cm thick beryllium block followed by 8 cm of paraffin. The thermal neutron fluence rate (3.0 × 103 n cm−2 s−1) was determined using a 3He detector [18]. Each pulse height spectrum was obtained from a total incident neutron count of ~423,900 for the 1.0 μm, and ~1,356,480 for the 1.8 μm neutron detector. A coaxial cable connects the detector to a Canberra 2004 charge-to-voltage preamplifier and subsequently to a Canberra Digital Signal Processor (DSP 9600) for pulse counting. The outer shield of the coaxial cable is grounded and connected to either the p or n side of the junction. The scope traces were obtained on a Tektronix TDS520D oscilloscope. Three different trapezoidal time constant filters were used to process the signal from the preamplifier. Background measurements were obtained using a 0.025 mm Cd foil, as a thermal neutron shield [19]. All experiments with and without Cd foil were performed under identical conditions of neutron flux and applied bias to obtain the most reliable measurement of background. 3. Semiconducting properties Current–voltage (I–V) curves for both p-type BC/n-type Si heterojunction diodes are shown in figure 2. The low leakage currents are crucial in suppressing the noise peak, which enables the detection of neutron capture signals. The combination of low leakage current and large breakdown voltage (<−40 V) allow for the application of large Boron carbide based sol id state neutron detectors 3 bias voltages, without significant increases in the noise level. This enhancement of the device properties for neutron detection was accomplished by more stable plasma control, low doped (100 Ω cm) Si wafers and thick BC deposition. The 1.0 μm BC layer with a contact area of 0.785 mm2 and a leakage current of 0.098 μA (corresponding to a current density of 12.5 μA cm−2) at an applied bias of −19 V shows a 103 fold decrease in the leakage current compared with the Figure 1. (Top) A neutron capture event in a BC/Si heterojunction diode detector. (Diagram not to scale.) The charge sensitive preamplifier may be connected to either side of the junction, collecting either a hole or an electron current. (Bottom) Schematic diagram of the neutron detection system used showing the associated electronics. Oscilloscope traces taken directly after the preamplifier are dominated by the rise time associated with charge capture within the semiconductor whereas oscilloscope traces taken after the DSP are a convolution of the rise time with the selected time constant. Figure 2. I–V curves of (a) 1.0 μm BC and (b) 1.8 μm BC heterojunction diodes on n-type Si at room temperature. Inset (a): C–V curve for the 1.0 μm BC heterojunction detector showing the V –1⁄2 dependence of the capacitance. Both I–V and C–V curves indicate a built-in voltage of 0.7 V. 4 Hong et al . in Journal of Phys ics D: Appl ied Phys ics 43 (2010) thin BC diode used in previous work [2]. In the case of the 1.8 μm BC layer, we were able to utilize a much larger detection area (3.14 mm2) due to the higher resistance of the thicker film. The leakage current at −19 V corresponds in this case to 0.244 μA, leading to a much lower current density of 3.4 μA cm−2. Both the 1.0 μm and 1.8 μm samples showed the same built-in potential, Vbi = 0.7 V. In a diode detector, the active region for charge capture is the depletion region, where charge recombination is highly suppressed. The depletion region increases with increasing reverse bias as expected, shown in the capacitance–voltage (C–V) measurement in the inset of figure 2(a), with the V –1⁄2 dependence expected for a step junction [20]. The C–V curve indicates that the device is not entirely depleted over the range of bias used for neutron detection measurements (from 0 to −19 V). In order to estimate the individual depletion widths in the BC (where the doping density is not well known) and Si, we first assume that at reverse biases <−10 V, the BC region is fully depleted and the slope of the C2 versus 1/(Vbi − Vappl) line is entirely due to increasing depletion in the Si. From this we obtain a doping density of 4.5 × 1012 cm−3 in BC, corresponding to a depletion width of 10 μm at zero bias, well above the thickness of either of the two BC films. Hence over the entire bias range, the BC is always fully depleted. Application of a reverse bias leads to an increase in the depletion width on the Si side of the heterojunction, as well as increasing the internal electric field, both of which have implications for charge collection. Previous measurements of the resistivity of BC range from 108 to 1010 Ω cm leading to a mobility ranging from 1.4 × 10−4 to 1.4 × 10−2 cm2 V−1 s−1, many orders of magnitude lower than the mobility of the n-type Si (1400 cm2 V−1 s−1). This drastically lowered mobility has a pronounced effect on the charge capture efficiency. 4. Pulse height spectra Neutron detection pulse height spectra with and without Cd foil are shown in figures 3 and 4 as functions of applied bias and integration time constants, respectively, for both the 1.0 μm and 1.8 μm diodes. All data shown in these figures were collected from the Si n-type side of the heterojunction diode, with the BC side grounded, and hence correspond to electron collection. Four bias voltages (figure 3) using the longest time constant with a 28 μs rise and fall time and a 3.0 μs flattop time, abbreviated as 28_3.0 μs, and three different trapezoidal time constants (figure 4) at a fixed bias voltage are shown. The smaller detection area of 0.785 mm2 for the 1.0 μm film leads to much lower count rates and less well-defined peaks. The details of the trapezoidal time constants are indicated in figure 4(c). The pulse height spectra for the larger area 1.8 μm thick detector show three peaks corresponding to the 0.84 MeV 7Li ions (first peak), 1.47 MeV 4He ions (second peak) and 1.78 MeV 4He ions (much smaller third peak). At 0 V, part of the 0.84 MeV peak is subsumed within the noise peak but is clearly discernible at an applied reverse bias of 10 V. The low intensity of the 1.78 MeV peak arises from the much lower probability (6%) of the reaction generating the 1.78 MeV 4He ions. A peak from the 1.02 MeV 7Li ions is not distinguishable within our energy resolution but may be embedded within the first peak. 4.1. Effects of increased bias With increasing applied bias, the pulse height spectrum broadens, and shifts to higher channel numbers, an effect also seen in earlier work on thinner detectors [2]. The secondary ions produced as a result of neutron capture within the BC layer are emitted with equal probability in all directions with path lengths of a few micrometers in both the BC and the Si (the path lengths are not identical). The amount of energy lost by the ions within the diode, varies from a minimum (when the ions are emitted normal to the film plane) to maximum (when the ions are emitted parallel to the plane of the sample) and corresponds to the number of electron–hole pairs produced. An increase in the applied bias results in increases in both the depletion width, which allows for charge collection over a larger region, and the internal electric field, resulting in higher accelerations and less charge trapping and recombination, both of which result in higher charge collection and hence higher channel numbers. The broadening of the pulse height peaks is also ascribed to increases in the depletion width, because charge may be collected from a much larger range of solid angles, resulting in a wider range of energies deposited. A quantitative measure of the shift in the pulse height spectra with increasing bias may be obtained by using the second peak of the 1.8 μm film. A plot of the peak position versus the bias across the depletion width (figure 5) shows a (Vbi – Vappl) dependence, similar to that of the C–V curve, further suggesting that as the depletion width (or reverse bias) increases, the fraction of charge collected increases proportionately. The peak position approaches, but does not reach saturation even at 20 V; and since the device is not entirely depleted, we may assume that applying still higher bias voltages would lead to increased charge collection. Because the range of the 4He ion in Si is 5 μm [19] and because all ions must originate in the BC layer, increasing the depletion width beyond 5 μm in Si (reached at a bias voltage of 2.3 V) should lead to no further increase in the charge collection. The increases in peak position at biases >2.3 V must result from increased acceleration of charge on both sides of the junction. However, as will be discussed in figure 8, this does not correspond to increases in efficiency. Unlike in solid state gamma and x-ray detectors [19] the position of the peaks in the pulse height spectra provides no information about the neutron energy and energy resolution is not important; rather it is the intrinsic efficiency that must be maximized. For a particular neutron capture event, incomplete charge capture will still lead to a neutron count provided the signal-tonoise ratio is large enough.