Stable room-temperature thallium bromide semiconductor radiation detectors

Thallium bromide (TlBr) is a promising room temperature wide bandgap semiconductor radiation detector material. At present, indium-doped cadmium zinc telluride (CdZnTe:In) is the incumbent room temperature compound semiconductor material which is being widely used. Compared to CdZnTe, TlBr offers several unique advantages. First, TlBr has a larger high-energy γ-ray absorption coefficient, resulting in 50% lower attenuation length than CdZnTe at 662 keV.¹ Second, the high electrical resistivity (∼10¹⁰ $Ω$ cm) is achieved without a dopant compensation, and therefore TlBr crystals do not experience the compositional variations innate to In-doped ternary CdZnTe. Lastly, TlBr growth temperature is lower (480 °C), which reduces the complexity of the growth and purification equipment. Also, TlBr has a cubic CsCl structure (space group Pm3m) and melt-based crystal growth by horizontal and vertical solidification methods producing single crystalline boules, which is a significant challenge for CdZnTe crystals. Progress over the last decade in TlBr purification and growth techniques has greatly improved the detector quality. Purification by zone refining has substantially increased the carrier lifetime, with electron mobility-lifetime product values now in the range of mid-10⁻³ cm²/V.² Unfortunately, TlBr is an ionic material, and the diffusion of Br⁻ (and to a lesser extent Tl⁺) ions under an applied electric field causes charge polarization in the detector and a concomitant degradation of the detector performance.^3–7 

This electro-migration phenomenon has severely hampered the adoption of TlBr as a room temperature semiconductor radiation detector for field applications. Under an appropriate electric field and ionic vacancy concentration in the lattice, these positive and negative ions migrate toward the respective electrodes through electro-diffusion. Since both types of intrinsic vacancies are present in the TlBr matrix, this time dependent migration of ions in accordance with the electric field is continued through a vacancy-hopping mechanism. At a certain temperature, the relationship between the ionic mobility and diffusion barrier of the highly mobile Br⁻ ion species was analyzed in a different study.⁴ Using the relation

where μ_Br is the Br⁻ ion mobility at temperature T, we obtained the migration energy of Br⁻ ions, E_A,_Br, to be 0.33 eV.

Electro-migration has two collective effects. First, chemical reaction between the Br⁻ ions and the metal anode, forming non-conducting metal bromide islands in the contact metal. The electric field in the contact region deteriorates gradually and eventually the electrode is corroded to the extent that it becomes impossible to create a stable electric field which leads to detector failure. Second, the accumulation of electrically active ions at the electrodes resulting in the perturbation of the electric field within the device, adversely influencing charge collection.^3,5 This phenomenon has been reported for other ionic semiconductors such as HgI₂,⁸ and we have directly observed its manifestation through Pockel’s effect experiments.⁹ The former exhibits itself over longer time scales, and the latter is a short-term effect (few hours to a few days). Although both these phenomena manifest themselves as deterioration of charge collection over time, the primary cause of failure for TlBr devices is now considered to be the chemical reaction occurring at the anode.^4,5,10

Several strategies have been used to stabilize TlBr detectors. By operating the device at temperatures as low as −20 °C, the metal-bromine reaction and the diffusivity of ions can be significantly reduced.¹¹ For room temperature applications, several approaches for reducing the rate of Br⁻ reaction with the contacts have been explored: HCl etching of the TlBr surface prior to contact deposition to produce a heterojunction of TlBr_1−xCl_x¹⁰ and use of various metals as contact electrodes including Pt, Au, Pd, and Tl.^3,12 None of these approaches have been demonstrated to achieve room temperature lifetimes exceeding beyond 100 days. Moreover, these solutions are not reproducible, and they yield significantly varying device lifetimes.

In this paper, we report the room temperature performance of TlBr detectors under periodic switching of the applied bias. Using this technique, we overcame the two major effects of polarization as described above. Bias switching is also shown to overcome the short-term deleterious effect of charge accumulation at the contact on the electric field in the detector.² Stable long-term room temperature performance of TlBr detectors with Pt-contacts for tens of thousands of hours under different bias-switching schemes is presented. It allows for maintenance-free field operation of TlBr radiation detectors.

TlBr detector grade crystals used in this study were purified and grown by the Travelling Molten Zone (TMZ) technique using 5N purity anhydrous TlBr beads from EMD performance materials.⁴ The beads were melted and zone refined at 5 cm/h. After 100 passes, single crystal growth was achieved by reducing the translation rate to 1 mm/h. Stoichiometry of the molten phase was controlled by using a mixture of HBr and an inert gas in the growth atmosphere. The TlBr crystal was cut into detector-size pieces with a diamond wire saw. The crystals were then grinded using SiC paper and polished mechanically. The chemical etching was done using a solution of 5% Br-methanol for 5-10 min. Before metalizing the samples, the polished surfaces were cleaned using methanol and dried using argon inside the sputtering chamber. No post-growth thermo-chemical treatment was performed on the detectors. 1.26 cm² planar Pt electrodes were sputtered near the center of 2.5 cm semi-circular detectors. In the pixelated detectors, pixels with 1.1 mm pitch were deposited. The bulk resistivity of the planar TlBr devices was of the order of 10¹⁰ Ω-cm. The electron mobility-lifetime (μ_eτ_e) values of the detectors were in the range of 1–4 × 10⁻³ cm²/V.

Charge collection was accomplished by an automated bias-switching electronic system, as shown in Fig. 1. An NXP LPC series microcontroller was used to switch the applied bias using high voltage power supplies and solid-state relays. The stabilization time between each polarity switching is approximately 2–5 s. The bias-switching frequencies used for the experiments cover the low frequency range of ∼5–1660 μHz (periods of 5 min–24 h). Custom built low-noise FET-input charge preamplifiers were used to detect and process gamma-interaction events under both positive and negative biases. For control experiments, the same electronics circuitry was used with bias-switching turned off. The control unipolar biased devices were fabricated using the same procedure as the bias-switched devices. Consistently high lifetimes for unipolar-biased detectors were obtained using Pd-metal contacts.¹² 

The results presented in this paper are extracted from extensive studies on short-term and long-term polarization of TlBr detectors performed over a course of more than two years. Figure 2 shows the short-term polarization effect in a TlBr radiation detector. After the conditioning period, this 2 mm-thick planar TlBr device was kept under a constant bias (1000 V/cm) and 10 μCi ¹³⁷Cs irradiation for 116 h. Due to weakening of the internal electric field, the channel number and counts from 662 keV gamma rays continuously degraded with time. After 116 h, the detector was subjected to bias switching at 1667 μHz. The 662 keV channel number and the FWHM started to improve after just one hour of bias switching. After 46 h of bias switching, initial performance of the detector was fully recovered. This demonstrates effectiveness of the bias switching technique in minimizing short-term polarization effects.

Figure 3 shows the timelines of all the TlBr devices that have been continuously running under bias-switching. The experiments were started at different times, and Fig. 3 shows them in a chronological order (earliest at the bottom and most recent at the top). These planar (including a 4 cm² large area planar detector) and pixelated detectors are being operated at 1000 V/cm under continuous gamma irradiation using 10 μCi ¹³⁷Cs sources (photon flux ∼3500 photons/s). Figure 3 also includes the lifetime of TlBr detectors in the control experiments. All the devices are subjected to the same biasing field and irradiation unless noted otherwise. The performance of a TlBr device under higher electric field is discussed in Sec. S1 of the supplementary material. The unipolar-biased devices stopped working not long after the start of the experiments, whereas the detectors under bias switching have remained fully functional. The lifetime data for the control devices are from our measurements as well as the literature data and are in the range of 100–2400 h (5–100 days).^4,12 Here, we define functionality of the detectors as its ability to exhibit a fully resolvable photopeak at 662 keV without any digital corrections. The dashed line shows 1.6 yrs of 24 × 7 operation, which is equivalent to five years of 8 × 7 operation.

Figure 4(a) shows the time history of the normalized 662 keV centroid positions of various planar and pixelated detectors. The lower bias switching frequency results in higher shift in the centroid position (40 calibrated channels and 2.8% standard deviation), and thus a relatively poorer performance in the 10 000 h time frame. For all the other detectors, this standard deviation was close to 1% (6 calibrated channels). Figure 4(b) shows the change in the FWHM of the 662 keV peak over the duration of the detector operation. Results of Figs. 4(a) and 4(b) indicate that there is a lower limit to the optimal switching frequency, beyond which the device deteriorates rapidly. Our results indicate that the lowest desirable switching frequency is 17.4 μHz. Section S3 of the supplementary material shows the physical conditions of a Pt electrode in the unipolar bias, 5.8 μHz, 17.4 μHz, and 1667 μHz bias-switched planar devices. The observed performance deterioration over long periods of use can be compensated through device recalibration, say after every 1000 h of operation. Then, the channel drift is less than 0.03% (2 calibrated channels). This is a common practice applied in most scintillator-based commercial off-the-shelf (COTS) handheld devices.

The lowest degradation is observed at the highest switching frequency. The transition frequency from no-to moderate-degradation appears to be in the range of 17.4–5.8 μHz, corresponding to switching times of 8–24 h. The variation in FWHM shows a perceptible drift only for the 5.8 μHz experiment; we have observed the same trend in total counts per second. However, even at frequencies where some drift is observed, the variations in any of the performance metrics over any 1000 h interval is not appreciable, and, therefore, recalibration of the detector every 1000 h results in a nearly constant performance parameter over the 10 000 + hours of collected data.

The observed sensitivity of degradation to switching frequency can possibly shed light on the degradation mechanism(s) and their time scales. The degradation time constant can be broadly considered to consist of three significant components: (a) electro-diffusion of Br⁻ ions in the TlBr bulk, (b) electro-diffusion of Br⁻ ions through the region close to the metal contact which may consist of TlBr and/or bromide of the contact metal, and (c) the reaction of bromine with the contact metal. With unipolar biasing, the combined characteristic time of the three processes yields a lifetime of less than 2400 h. However, with bias switching the overall lifetime is increased, with data suggesting nearly infinite lifetime at switching frequencies above 1667 μHz. We hypothesize that the reaction rate of Br⁻ ions with contact metals is relatively fast, and the rate-determining processes are the first two of the above list.

Repeated changes in the polarities of the applied bias significantly minimize the short- and long-term effects of ionic diffusion polarization. The rate of the short-term process is limited by the applied electric field and that of the latter is limited by availability of free Br⁻ ions near the metal electrode. For unipolar bias, the Br⁻ ion migration through V_Br⁺ sites is highly pronounced and increases over time.⁴ This higher diffusion flux not only facilitates the polarization of the device but also makes more Br⁻ ions available near the TlBr-metal interface. But with the bias-switching technique, the direction of driving force for the migrating Br⁻ species changes periodically. Therefore, polarization in the device is minimized, and the reaction time between the metal and Br⁻ ions is prolonged due to the unavailability of free Br⁻ ions (Fig. 5). With high frequency bias-switching, this reaction time can be prolonged indefinitely. However, that is not practically possible for operation of a radiation detector. In our experiments, we tested a wide range of bias-switching frequencies. Although all the devices are functioning as a radiation detector, the conditions of the platinum contacts are not the same. Thus, it is important to formulate a recipe with a trade-off between the optimum device lifetime and the practical field application constraints. As higher frequency of bias-switching is better for device lifetimes, we must analyze the lower limit of the applicable frequency.

Figure 5 shows the Pt-TlBr interfacial region in a typical TlBr device after a certain biasing period under unipolar and switched polarity devices. We define a bromide depleted layer of thickness L₀ near the Pt contact where there are no free or bound Br⁻ ions present. Under an applied field, if a Br⁻ ion enters this region, it is bound to react with the metal. So, under equilibrium conditions, this layer acts as a source for Br⁻ ions, whereas the metal acts as a sink. The ionic current can be written as j = n_ionμ_ionE, where n_ion and μ_ion are the number density and mobility of Br⁻ ions under an electric field, E, respectively. The following equation can be obtained using the charge conservation principle for planar detectors:

From Eq. (2), we can relate the Br⁻ ions entering the bromide depleted region to a characteristic time scale T₀ for a bias-switched electric field of magnitude E₀ and frequency ω,

In the alternating bias approach, the frequency of switching applied to the device should be lower than 1/T₀. It will ensure the minimum rate of Br⁻ ion diffusion into the region L₀. The mobility of Br⁻ ions in the TlBr bulk is in the range of 10⁻¹¹ cm²/V s,^4,14 which at 1000 V/cm yields an electro-diffusion velocity (i.e. estimate for the quantity μE₀) of 10⁻⁸ cm/s. For a lifetime of 5 days, the excursion distance of Br⁻ ions is about 50 μm. On the other hand, for the bromine depleted layer with a thickness of 3 μm, the switching frequency must be higher than 17.4 μHz. As per ANSI requirements for several types of portable radiation detection and identification devices (PRD, SPRD, RIID, etc.), a continuous field operability of 16 h is required. So, for these kinds of applications, only one bias-switching cycle is required (i.e., a dead time of 30 s every 16 h) during the entire operation time, which can be pre-programmed.

Additionally, increasing the resistance of the near-contact region to electro-diffusion of Br⁻ can complement the alternating bias method in increasing the device lifetime. In principle, this can be approached in a number of ways. First, by etching the detector surface by HCl to produce a TlBrCl layer (up to 8 μm thick) prior to deposition of the contact metal. Section S2 of the supplementary material demonstrates a TlBr planar device with unipolar bias and HCl-etched surface. The second approach would be to reduce crystalline defects such as dislocations that can provide a faster path for electro-diffusion of Br⁻ ions.¹² The slowest electro-diffusion pathway for Br⁻ ions is by vacancy hopping, and crystalline defects can provide alternative faster pathways. Thus, TlBr detectors with low crystalline defects, polished appropriately to remove the cutting-related damages can be expected to offer higher resistance to electro-diffusion of Br⁻. As the electro-diffusion resistance is increased, lower switching frequencies will be required to achieve higher lifetimes. It should be noted, however, that neither of the above approaches under constant bias has produced devices that have performed beyond 2400 h (about 100 days).

In conclusion, we have successfully implemented a bias polarity switching approach to counteract the challenge of ionic polarization in TlBr semiconductor radiation detectors. Using this technique, we have demonstrated a stable behavior of planar and pixelated TlBr devices for tens of thousands of hours. This is a significant improvement from the existing reports of continuous operation of TlBr devices. Bias-switching extends the workability of any TlBr device with reproducibility, alongside relaxing the strict surface condition constraints for manufacturing devices with a longer lifetime. This technique not only stabilizes and enhances the detector performance at room temperature but also decreases the cost of operation. Using this technique, TlBr-based compact room temperature hand-held devices can be deployed for field applications in homeland security, medical imaging, and physics research. Several other approaches including alternative contact metals and lower operating voltages are being implemented with the bias switching configuration to further improve the stability of the detectors.

See supplementary material for the performance of a TlBr planar device under higher electric field (2000 V/cm) (S1), performance of HCl-etched TlBr devices under continuous bias (S2), and the condition of the platinum electrodes after tens of thousands of hours under continuous electric field and gamma irradiation (S3).

The work discussed in this document has been supported by the US Department of Homeland Security, Domestic Nuclear Detection Office. This support does not constitute an express or implied endorsement on the part of the Government.