Improved temporal performance and optical quantum efficiency of avalanche amorphous selenium for low dose medical imaging

Abstract.

Purpose

Active matrix flat panel imagers (AMFPIs) with thin-film transistor arrays experience image quality degradation by electronic noise in low-dose radiography and fluoroscopy. One potential solution is to overcome electronic noise using avalanche gain in an amorphous selenium (a-Se) (HARP) photoconductor in indirect AMFPI. In this work, we aim to improve temporal performance of HARP using a novel composite hole blocking layer (HBL) structure and increase optical quantum efficiency (OQE) to CsI:Tl scintillators by tellurium (Te) doping.

Approach

Two different HARP structures were fabricated: Composite HBL samples and Te-doped samples. Dark current and optical sensitivity measurements were performed on the composite HBL samples to evaluate avalanche gain and temporal performance. The OQE and temporal performance of the Te-doped samples were characterized by optical sensitivity measurements. A charge transport model was used to investigate the hole mobility and lifetime of the Te-doped samples in combination with time-of-flight measurements.

Results

The composite HBL has excellent temporal performance, with ghosting below 3% at 10 mR equivalent exposure. Furthermore, the composite HBL samples have dark current $<{10}^{-10}\mathrm{A}/{\mathrm{cm}}^2$ and achieved an avalanche gain of 16. Te-doped samples increased OQE from 0.018 to 0.43 for 532 nm light. The addition of Te resulted in 2.1% first-frame lag, attributed to hole trapping within the layer.

Conclusions

The composite HBL and Te-doping can be utilized to improve upon the limitations of previously developed indirect HARP imagers, showing excellent temporal performance and increased OQE, respectively.

1. Introduction

Active matrix flat panel imagers (AMFPI) have revolutionized medical imaging due to their improved image quality and real-time readout capabilities.¹ Indirect AMFPIs use a scintillator to convert x-rays into optical photons, which generate electron-hole-pairs when detected by photodiodes and are then readout by a thin-film transistor (TFT) array. In low-dose applications (e.g., $0.1\mu \mathrm{R}$ detector exposure for fluoroscopy), indirect AMFPIs image quality is degraded by electronic noise associated with the TFT array.² One proposed strategy is to reduce electronic noise utilizing active pixel sensor (APS) designs. APS designs incorporate pixel amplification by adding additional TFTs at each pixel.^3–6 The electronic noise could be reduced by up to a factor of $\sim 2$, however it does not completely eliminate the degradation of image quality and adds substantial cost to the manufacturing. Our proposed solution is the scintillating high-gain avalanche rushing photoconductor AMFPI (SHARP-AMFPI), which replaces the photodiodes with a high-gain avalanche rushing photoconductor (HARP) amorphous selenium (a-Se) layer to amplify signal prior to the introduction of electronic noise.⁷

We previously demonstrated the feasibility of a large area SHARP-AMFPI by constructing a $24\mathrm{cm}\times 30\mathrm{cm}$ prototype detector, and achieved uniform avalanche gain of up to 75.⁸ The initial prototype was limited by a low-transparency bias electrode and poor temporal performance due to lack of electron transport in the hole blocking layer (HBL). These limitations were improved upon in the second SHARP-AMFPI prototype by utilizing a transparent indium tin oxide (ITO) bias electrode in combination with a metal oxide HBL with improved electron transport.⁹ This prototype also allowed for back-irradiation (BI) geometry to improve the intrinsic performance of CsI:Tl by reducing the impact of depth dependent effects.^10–12 The remaining limitations of the second prototype include: (1) the increased dark current and poor reliability at avalanche fields associated with the metal oxide HBL, and (2) the relatively low inherent optical quantum efficiency (OQE) of a-Se to CsI:Tl emissions. In addition to overcoming secondary quantum noise, our group has shown that the OQE of HARP needs to be maximized for optimal gain matching between CsI and a-Se for back-irradiated SHARP-AMFPI.¹³ In this work, we investigate the performance of HARP samples with a composite HBL structure to improve temporal performance while maintaining reliable avalanche gain. Additionally, we fabricated and characterized HARP samples with a thin layer of tellurium (Te) alloyed a-Se to improve the OQE for the optical emission spectrum of CsI:Tl.

2. Methods

HARP samples with an active area of $6.35\times 6.35{\mathrm{cm}}^2$ were fabricated on glass substrates, consisting of an electron blocking layer (p-layer), a-Se (i-layer), and a HBL (n-layer). Two types of HARP structures were fabricated: (1) with a composite HBL for improved electron transport and (2) with an additional layer of Te-doped a-Se to increase the OQE to columnar CsI:Tl scintillators. Guided by previous work on Te-doped HARPICON,¹⁴ the Te concentration ranged from 2.5 to 18.2 weight percent (wt %) (determined via Auger electron spectroscopy). Both HARP structures were deposited in a p-i-n sequence, as shown in Fig. 1. The glass substrate is coated with $\sim 25\mathrm{nm}$ of ITO, which acts as the negative bias electrode. A thermally deposited $2\mu \mathrm{m}$ thick p-layer traps injected electrons while permitting hole transport. The i-layer, for optical sensing and avalanche multiplication, is stabilized a-Se with a thickness of $15\mu \mathrm{m}$. The n-layer for the composite HBL samples consists of a thin (100 nm) n-type metal oxide (injection barrier layer), which provides an energy barrier to prevent hole injection from the positive bias electrode,¹⁵ and a thin layer (350 nm) of organic HBL (hole trapping layer) to trap any injected holes while minimizing the effect on electron transport. The n-layer for the Te-doped samples is a 100 nm thick n-type metal oxide. The top electrode is a $\sim 75\mathrm{nm}$ thick transparent ITO. The entire structure is encapsulated with a transparent organic insulator to improve device reliability. The samples contain three separate high-voltage (HV) electrodes of different sizes: 0.2284, 0.9784, and $2.9784{\mathrm{cm}}^2$. The smallest electrode sensor is used for optical excitation experiments while the larger two sensors are used for dark current characterization.

Fig. 1.

(a) Cross-sectional schematic of composite HBL HARP sample. (b) Cross-sectional schematic of Te-doped HARP sample. Thicknesses are not drawn to scale.

2.1. Composite HBL: Dark Current and Temporal Performance Measurements

Dark current in avalanche a-Se is dominated by charge injection from the positive bias electrodes.¹⁶ A HV power supply (Stanford Research Systems PS365) was used to apply an external bias voltage to the HARP samples in a light tight box and the dark current was measured with an electrometer (Keithley 6514). The measurement was performed over a range of electric fields ($E_{\mathrm{Se}}$) of $\sim 10$ to $60\mathrm{V}/\mu \mathrm{m}$, and the results with the composite HBL and the previously used HBLs were compared.

The temporal performance (e.g., lag and ghosting) of the composite HBL samples were characterized using optical excitation by a pulsed blue light emitting diode (LED) (${\lambda }_{\text{peak}}=458\mathrm{nm}$, $\mathrm{FWHM}\pm 10\mathrm{nm}$) light source driven by a function generator (Tektronix AFG 3021B) function generator at a frequency of 30 Hz and a 50% duty cycle. The bandgap of a-Se is $\sim 2.2\mathrm{eV}$.¹⁵ This leads to the collected charge from a single light pulse being equivalent to what is anticipated of a 10 mR x-ray exposure to a $600\mu \mathrm{m}$ thick high-resolution (HR) CsI:Tl scintillator under RQA5 beam quality conditions ($70{\mathrm{kV}}_p$, 6.8 mm Al HVL).¹⁷ This exposure level was chosen to replicate extreme detector conditions (i.e., near the saturation region of the detector). The induced photocurrent was captured by a digital oscilloscope (Tektronix TDS 7104) with a 1 MΩ terminating resistor to integrate the collected charges. The ghosting was quantified as the percentage optical sensitivity of the initial value

$$G_n=\frac{S_n}{S_1},$$ (1)

where $S_1$ and $S_n$ are the optical sensitivity measured at the first and $n$’th optical exposure, respectively. Larger ghosting means more drop in optical sensitivity. The first frame lag was quantified as the residual signal 16.7 ms after optical exposure. This should correspond to the lag measured at real-time frame rate of $30\text{frames}/\text{second}$ commonly used in fluoroscopy.

2.2. Avalanche Gain Measurements

Optical excitation measurements using the pulsed blue LED (${\lambda }_{\text{peak}}=458\mathrm{nm}$, $\mathrm{FWHM}\pm 10\mathrm{nm}$) light source were also used to measure avalanche gain of the composite HBL sample by acquiring the photocurrent amplitude from a single light pulse as a function of $E_{\mathrm{Se}}$. Prior to exposure, the target HV bias was held for several seconds to allow the dark current to stabilize. After each measurement, the HV bias was ramped down to 0 V and the sample was exposed to white light to depolarize the sample. Active probes (Tektronix P6245) with a 1 MΩ input resistance were used to protect the oscilloscope against device failure at high $E_{\mathrm{Se}}$.

2.3. Optical Sensitivity Measurements

The OQE of Te-doped samples were determined by finding the ratio between collected charge and the amount of incident photons delivered by a single green LED (${\lambda }_{\text{peak}}=532\mathrm{nm}$, $\mathrm{FWHM}\pm 17\mathrm{nm}$) light pulse. The number of incident photons was determined from optical sensitivity measurements using a calibrated silicon photodiode (Thorlabs, PDA10A2). The resulting photon fluence was found to be $2.17\times {10}^7{\mathrm{mm}}^{-2}$, which is equivalent to the light output expected of a 3.5 mR x-ray exposure to a $600\mu \mathrm{m}$ thick HR CsI:Tl scintillator at RQA5 beam quality.

2.4. TOF Measurements and Modeling

To understand the temporal performance of the waveforms measured during optical sensitivity experiments, we investigated the charge transport characteristics of Te-doped samples via modeling in combination with time-of-flight (TOF) transient photoconductivity experiments.^18,19 We used our previously established charge transport model to simulate the trapping, release, and recombination of charge carriers in a multilayer structure.²⁰ The current continuity equations given as²¹

$$\frac{\partial n}{\partial t}=\frac{\partial {\mu }_{e}E(x)n}{\partial x}-\frac{n}{{\tau }_e}(1-\frac{n_t}{N_{te}})-rn{p}_t+\frac{n_t}{{\tau }_{re}}+g{e}^{-\alpha x},$$ (2)

$$\frac{\partial p}{\partial t}=\frac{\partial {\mu }_{h}E(x)p}{\partial x}-\frac{p}{{\tau }_h}(1-\frac{p_t}{N_{th}})-rp{n}_t+\frac{p_t}{{\tau }_{rh}}+g{e}^{-\alpha x},$$ (3)

represent the motion of charge carriers in one dimensional space ($x$) and time ($t$) across the thickness of the sample, where $n$ and $p$ are the concentrations of electrons and holes, respectively; $n_t$ and $p_t$ are the concentrations of trapped charges; $E(x)$ is the electric field; $\mu$ is the field dependent mobility; $N_t$ is the number of traps; $\tau$ is lifetime of the charge carrier; ${\tau }_r$ is the trap release time; $g(x,t)$ is the electron-hole pair generation rate; and $\alpha$ is the linear attenuation coefficient of the photoconductor. The amount of modeled incident photons, $g(x,t)$, was adjusted to account for expected OQE to the laser spectrum for the different Te concentrations. The recombination coefficient follows that for the Langevin mechanism: $r=q\mu /\epsilon$, where $\epsilon$ is the dielectric constant.²² A similar set of rate equations describing the trapped charges can be created²³

$$\frac{\partial n_t}{\partial t}=\frac{n}{{\tau }_e}(1-\frac{n_t}{N_{te}})-\frac{n_t}{{\tau }_{re}}-rp{n}_t,$$ (4)

$$\frac{\partial p_t}{\partial t}=\frac{p}{{\tau }_h}(1-\frac{p_t}{N_{th}})-\frac{p_t}{{\tau }_{rh}}-rn{p}_t.$$ (5)

Equations (2) and (3) depend on the internal electric field, therefore Eqs. (2)–(5) must be solved simultaneously with Poisson’s equation

$$\frac{\partial E}{\partial x}=\frac{q}{\epsilon }(p-n+p_t-n_t).$$ (6)

To solve Eqs. (2)–(6) simultaneously, a finite difference method is utilized. The finite difference equations are convergent on the condition $dx/dt\gg {\mu }_{h}E$, so a time step $dt$ is chosen such that the distance the carrier travels during $dt$ is less than the modeled slice thickness $dx$. Location parameter value $x=0$ is defined as the HV electrode and $x=L$ is the location of the ground electrode, where $L$ is the total thickness of all layers. The charge density boundary conditions are set to be $p(x,0)={\rho }_p$ and $n(L,t)={\rho }_n$, as given by the charge injection equation ²⁴

$${\rho }_{p,n}=N_{v,c}\exp (-\frac{{\phi }_{h,e}-{\beta }_s\sqrt{E}}{kT}),$$ (7)

where $N_{v,c}$ is the respective density of states in the valence and conduction band ($N_{v,c}=1\times {10}^{19}$), ${\phi }_{h,e}$ is the hole and electron injection barrier at the interface (${\phi }_e=0.83$ and ${\phi }_h=0.80$), and ${\beta }_s$ is the Schottky coefficient (${\beta }_s=1.52\times {10}^{-4}$). We assume $p(x,0)=n(x,0)=0$ for all other locations. Finally, the current density, $J$, is calculated as

$$J=q\int ({\mu }_{h}pE+{\mu }_{e}nE)\mathrm{d}x.$$ (8)

Figure 2 shows the modeled p-i-n structure for samples without and with the Te-doped layer. The model is composed of the a-Se (I-layer), the HBL (N-layer) and the electron blocking layer (P-layer). An additional I-layer with tunable material properties was added to represent the Te-doped a-Se layer. The charge transport properties for each layer are detailed in Table 1. Parameter values for a-Se were taken from a range of values found in literature.²⁵

Fig. 2.

p-i-n HARP structure (a) without and (b) with Te-doped layer. The n-layer and p-layer trap holes and electrons, and therefore, contain positive and negative charges, respectively. The modeled Te-doped layer contains hole traps and can result in the trapping of photogenerated or injected holes. Thicknesses are not drawn to scale.

Table 1.

Parameters used for the different layers in the charge transport model.

	I-layer (a-Se)	N-layer	P-layer	I-layer (TeSe)
${\mu }_h$ (hole mobility)	$0.16{\mathrm{cm}}^2{\mathrm{V}}^{-1}{\mathrm{s}}^{-1}$	$0.005{\mathrm{cm}}_{ }^2{\mathrm{V}}^{-1}{\mathrm{s}}^{-1}$	$2\times {10}^{-8}\times \mathrm{E}{\mathrm{cm}}_{ }^2{\mathrm{V}}^{-1}{\mathrm{s}}^{-1}$	0.0035 to $0.008{\mathrm{cm}}_{ }^2{\mathrm{V}}^{-1}{\mathrm{s}}^{-1}$
${\mu }_e$ (electron mobility)	$0.003{\mathrm{cm}}_{ }^2{\mathrm{V}}^{-1}{\mathrm{s}}^{-1}$	$4.7\times {10}^{-3}{\mathrm{cm}}^2{\mathrm{V}}^{-1}{\mathrm{s}}^{-1}$	$1\times {10}^{-5}{\mathrm{cm}}^2{\mathrm{V}}^{-1}{\mathrm{s}}^{-1}$	$0.001{\mathrm{cm}}_{ }^2{\mathrm{V}}^{-1}{\mathrm{s}}^{-1}$
${\tau }_h$ (hole lifetime)	$1.5\times {10}^{-4}\mathrm{s}$	$8\times {10}^{-8}\mathrm{s}$	$1\times {10}^{-4}\mathrm{s}$	$1.05\times {10}^{-7}$ to $4.8\times {10}^{-7}\mathrm{s}$
${\tau }_e$ (electron lifetime)	$1\times {10}^{-3}\mathrm{s}$	$6\times {10}^{-5}\mathrm{s}$	$1\times {10}^{-5}\mathrm{s}$	$1\times {10}^{-3}\mathrm{s}$
${\tau }_{\mathrm{rh}}$ (hole trap release time)	300 s	300 s	300 s	$2.0\times {10}^{-7}$ to $3.0\times {10}^{-7}\mathrm{s}$
${\tau }_{\mathrm{re}}$ (electron trap release time)	$4\times {10}^4\mathrm{s}$	$4\times {10}^4\mathrm{s}$	$4\times {10}^4\mathrm{s}$	$4\times {10}^4\mathrm{s}$
$N_{\mathrm{th}}$ (hole trap concentration)	$1.3\times {10}^{12}{\mathrm{cm}}^{-3}$	$1\times {10}^{18}{\mathrm{cm}}^{-3}$	$1\times {10}^{18}{\mathrm{cm}}^{-3}$	$1.3\times {10}^{12}{\mathrm{cm}}^{-3}$
$N_{\mathrm{te}}$ (electron trap concentration)	$1.3\times {10}^{12}{\mathrm{cm}}^{-3}$	$1.3\times {10}^{12}{\mathrm{cm}}^{-3}$	$1\times {10}^{18}{\mathrm{cm}}^{-3}$	$1.3\times {10}^{12}{\mathrm{cm}}^{-3}$

TOF experiments were performed using a pulsed laser light source (STV-01E-12C, ${\lambda }_{\text{peak}}=355\mathrm{nm}$, 0.4 ns pulse width), and the resulting waveforms were fitted with results from our charge transport model. The fitted simulated waveforms were subsequently run through a circuit simulation in Cadence OrCAD Capture CIS Lite 17.2 (EMA Design Automation, Inc., Rochester, New York, United States) to account for sample capacitance and electrode resistance. The induced photocurrent was captured by a digital oscilloscope with a 50 Ω terminating resistor to minimize the RC constant of the circuit ($\sim 25\mathrm{ns}$). A diagram of the simulated circuit is shown in Fig. 3. The circuit consists of three major components: the HV power supply, the HARP structure, and the oscilloscope. The HV component is comprised of a direct current voltage source, HV, and a small output impedance, $R_{\mathrm{HV}}$. The HARP structure contains the top electrode resistance, $R_{\mathrm{TE}}$, the pixel capacitance, $C_{\mathrm{Se}}$, the pixel resistance, $R_{\mathrm{Se}}$, the photogenerated current (current output of the charge transport model), $I_{\mathrm{Se}}$, and the bottom electrode resistance, $R_{\mathrm{BE}}$. The oscilloscope contains a terminating resistor, $R_{\mathrm{Osc}}$, and a capacitor, $C_{\mathrm{Osc}}$.

Fig. 3.

Simulated circuit diagram to replicate experimental apparatus used in TOF experiments.

3. Results

3.1. Composite HBL: Dark Current and Temporal Performance

Figure 4 shows the measured dark current as a function of $E_{\mathrm{Se}}$ for HARP with different HBL structures. The composite HBL dark current is nearly equal, within a factor of two, to that of the thick organic HBL utilized in the first SHARP-AMFPI prototype at all electric fields.⁸

Fig. 4.

Measured dark current as a function of $E_{\mathrm{Se}}$ for different HBLs.

Figure 5(a) shows the change in optical sensitivity of the composite HBL HARP sample over a sequence of light exposure of 90 pulses with 10 mR equivalent exposure per pulse. The observed ghosting is less than 3% over the entire excitation sequence. This is a substantial improvement over the $\sim 30\%$ ghosting observed with the thick organic HBL. The sensitivity decrease is attributed to trapping of the photogenerated electrons within the HBL or at the HBL/a-Se interface, causing a reduction in the effective electric field across the HARP layer.²⁶ The initial rise in sensitivity with the thick organic HBL is attributed to trapping of the photogenerated electrons within the HBL or at the HBL/a-Se interface. This leads to an increase in the effective electric field across the HBL and positive electrode, thereby causing an increase in charge injection. The amount of ghosting is expected to decrease with lower levels of light exposure, as fewer electrons will be generated in a-Se and subsequently trapped within the HBL.

Fig. 5.

(a) Result of ghosting measurement using 10 mR equivalent exposure per blue LED light pulse at $10\mathrm{V}/\mu \mathrm{m}$. (b) Result of lag measurement using 10 mR equivalent exposure per blue LED light pulse at $70\mathrm{V}/\mu \mathrm{m}$.

Figure 5(b) shows the photocurrent as a function of time of the composite HBL HARP sample to a single 10 mR equivalent exposure light pulse. The measured lag (i.e., the residual signal 16.7 ms after optical exposure) indicates negligible levels ($<1\%$) for frame rates utilized in fluoroscopic imaging.

3.2. Avalanche Gain

Figure 6 shows the average integrated signal charge of three different composite HBL HARP samples as a function of $E_{\mathrm{Se}}$. Avalanche gain occurs at $E_{\mathrm{Se}}>70\mathrm{V}/\mu \mathrm{m}$ and reaches a value of 16 at $E_{\mathrm{Se}}=90\mathrm{V}/\mu \mathrm{m}$. The avalanche gain was fit using the equation

$$g_{av}=e^{\alpha d},$$ (9)

where $d$ is the thickness of the a-Se and $\alpha$ is the impact ionization coefficient given as

$$\alpha ={\alpha }_1{e}^{-\frac{{\alpha }_2}{E_{Se}}},$$ (10)

where ${\alpha }_1$ and ${\alpha }_2$ are experimentally determined fitting parameters.²⁷ ${\alpha }_1$ was found to be $8964\pm \mathrm{2,700}{\mu \mathrm{m}}^{-1}$ and ${\alpha }_2$ was found to be $994\pm 27\mathrm{V}/\mu \mathrm{m}$. Previously measured values for ${\alpha }_1$ range from 949 to $4101{\mu \mathrm{m}}^{-1}$ and values for ${\alpha }_2$ range from 849 to $990\mathrm{V}/\mu \mathrm{m}$.^28–30 Variations in these values are expected, as different HBLs modify the field distribution and lead to errors in $E_{\mathrm{Se}}$.³¹

Fig. 6.

Average gain measurements and corresponding fit from composite HBL HARP samples as a function of $E_{\mathrm{Se}}$. The error bars represent the standard error for three different samples.

3.3. Te-doping to Improve HARP OQE

Figure 7 shows the measured OQE of Te-doped HARP samples to green LED emissions as a function of Te concentration at $10\mathrm{V}/\mu \mathrm{m}$. Also shown is the expected OQE of Te-doped a-Se to the green LED optical spectrum as a function of Te concentration for both $10\mathrm{V}/\mu \mathrm{m}$ and $90\mathrm{V}/\mu \mathrm{m}$, computed using a weighted average with data extracted from Hagen et al.³² An increase of OQE is observed, up to 0.43 for 18.2 wt % Te, confirming Te-alloying is feasible for solid-state HARP applications. The measured OQE scaling shows good agreement with previously published literature and suggests that an OQE approaching unity could be achieved with 18.2 wt % Te-doping and sufficient $E_{\mathrm{Se}}$.

Fig. 7.

The expected OQE of Te-doped a-Se to the green LED emission spectrum and corresponding second order polynomial fits as a function of Te concentration for 10 and $90\mathrm{V}/\mu \mathrm{m}$, extracted with data from Ref. 32. Also shown is the measured OQE from HARP samples as a function of concentration to a 3.5 mR equivalent single green LED light pulse at $10\mathrm{V}/\mu \mathrm{m}$. The error bars (smaller than data point) represent the standard error for three repeated experiments.

Figure 8 shows the measured signal response to a green LED light pulse for HARP samples with various Te concentrations. The data show that the addition of Te increases OQE and introduces temporal effects, up to 2.1% first-frame lag for the 18.2 wt % sample. The temporal performance degradation is attributed to the presence of carrier traps in the Te-doped layer.³³ Trapping in the Te-doped layer can result in lag through the following mechanisms: (1) photogenerated electrons trapped in the Te-doped layer modify the field, resulting in increased charge injection from the positive bias electrode. (2) Holes generated in the Te-doped layer have delayed extraction time, determined by the hole trap density and release time. We investigate mechanism (2) in the following section by modeling the charge transport parameters needed to replicate measured TOF waveforms.

Fig. 8.

Comparison of signal response of Te-doped a-Se samples (2.5, 9.2, and 18.2 wt %) and a non-doped a-Se reference sample to a single green LED light pulse at $10\mathrm{V}/\mu \mathrm{m}$.

3.4. TOF Measurements and Modeling

Figure 9(a) shows the simulated TOFs for HARP samples with various Te concentrations. These represent the transport of charges unimpeded by the limited frequency response of the circuitry. The reference TOF is non-dispersive, as the majority of holes transport through the a-Se layer with uniform hole mobility and are unimpeded by hole traps. The observed initial peak in the reference TOF is due to the electron current, which quickly degrades as the electrons get trapped in the HBL or transport through to the positive electrode. The observed transit time ($t_T$) agrees with that predicted by the equation

$$t_T=\frac{L}{{\mu }_h{E}_{\mathrm{Se}}},$$ (11)

where $L$ is the sample thickness ($15\mu \mathrm{m}$) and ${\mu }_h$ is the hole mobility ($0.16{\mathrm{cm}}^2{\mathrm{V}}^{-1}{\mathrm{s}}^{-1}$).

Fig. 9.

(a) Simulated TOF waveforms of Te-doped a-Se samples and a non-doped a-Se reference sample prior to circuit simulation. (b) Experimental TOF waveforms to a single 355 nm laser pulse at $5\mathrm{V}/\mu \mathrm{m}$ and corresponding simulated TOF waveforms after convolving with circuit response.

Reduced hole transport is observed as a function of Te concentration. The Te-doped sample waveforms have an initially low current before increasing rapidly. This is attributed to the carriers traveling through a region of low hole mobility, according to Eq. (8). As the holes transport from the low mobility Te-doped layer into the higher mobility a-Se layer, the current subsequently increases. The introduction of shallow hole traps (parameters listed in Table 2) in the Te-doped layer were necessary to replicate experimental waveforms. The slow release of holes from these traps results in the dispersion of the photogenerated charge packet and the reduced integrated charge collection (85% and for 4.6 wt % Te and 17% for 9.2 wt % Te) compared to the reference sample TOF.

Table 2.

Parameters used for hole transport to model Te-doped samples.

	4.6 wt % Te	9.2 wt % Te
${\mu }_h$ (hole mobility)	$0.008{\mathrm{cm}}_{ }^2{\mathrm{V}}^{-1}{\mathrm{s}}^{-1}$	$0.0035{\mathrm{cm}}_{ }^2{\mathrm{V}}^{-1}{\mathrm{s}}^{-1}$
${\tau }_h$ (hole lifetime)	$4.8\times {10}^{-7}\mathrm{s}$	$1.05\times {10}^{-7}\mathrm{s}$
${\tau }_{\mathrm{rh}}$ (hole trap release time)	$2.0\times {10}^{-7}\mathrm{s}$	$3.0\times {10}^{-7}\mathrm{s}$

Figure 9(b) shows the measured signal response to a 0.4 ns laser pulse for the Te-doped HARP samples and the simulated TOFs convolved with the circuit detailed in Fig. 3. The introduction of the circuit response results in the smoothing of high-frequency details present in the simulated waveforms (e.g., initial current peaks and falling edges). Excellent agreement between the simulated TOFs and the experimental waveforms is observed.

4. Discussion

4.1. Composite HBL

The HARPicon, a video camera employing avalanche a-Se, utilized a HBL structure with cerium oxide (${\mathrm{CeO}}_2$) as the injection barrier layer and lithium fluoride (LiF) doped a-Se as the hole trapping layer.^14,34 This HBL structure is suitable in n-i-p deposition sequence, however the high temperature deposition methods required of this structure makes it unsuitable for use in p-i-n devices (e.g., SHARP-AMFPI). The metal oxide and organic HBLs used in our samples are transparent and capable of being deposited below the crystallization temperature of a-Se.

The composite HBL structure achieved consistent and reproducible avalanche gain for three separate samples (Fig. 5), up to a factor of 16. Zhao et al. demonstrated that a gain of 10 to 20 should be sufficient to overcome electronic noise for most medical x-ray imaging applications.⁷ The required avalanche gain could be further reduced by improving the OQE of a-Se to scintillator emissions via Te-doping.

The improved temporal performance over the original SHARP-AMFPI prototype makes this design more suited for use in real-time imaging. We expect the temporal performance to improve with decreasing detector exposure, as less photogenerated charges will be trapped resulting in reduced field modification.

4.2. OQE Improvement via Te-doping

The use of Te-doping to increase OQE in a-Se to longer wavelength light is well established through extensive investigations in xerography^33,35–37 and camera^38–40 devices. Figure 8 has demonstrated the feasibility of this approach for a-Se based solid-state devices. The addition of Te results in a linear decrease in the band-gap of the material, allowing for the absorption of lower energy photons.^41,42 Figure 10 shows the expected OQE of Te-doped a-Se to a CsI:Tl scintillator as a function of Te concentration for both 10 and $90\mathrm{V}/\mu \mathrm{m}$. These data were calculated using a weighted average of OQE over the entire CsI:Tl emission spectrum with data extracted from Hagen et al.³² An OQE of 0.90 can be achieved with sufficient Te-doping at avalanche fields, minimizing the impact of HARP on the Swank factor. Increased OQE will also reduce the amount of avalanche gain necessary to overcome electronic noise, resulting in reduced dark current noise and risk of breakdown. The 2.1% first-frame lag observed with the 18.2 wt % sample is less than that observed in current commercial detectors ($<5\%$).⁴³ However, an optimization may need to be made to maximize OQE while minimizing the temporal degradation observed with higher Te content.

Fig. 10.

The expected OQE of Te-doped a-Se to the broad CsI:Tl emission spectrum and corresponding second order polynomial fits as a function of Te concentration for $10\mathrm{V}/\mu \mathrm{m}$ and $90\mathrm{V}/\mu \mathrm{m}$, extracted with data from Ref. 32.

4.3. TOF Modeling of Te-doped HARP Samples

The trends observed in experimental waveforms suggest reduced hole mobility and significant hole trapping in the Te-doped layer. This agrees with previously published literature, where a sharp decrease in hole mobility is observed with the addition of Te.^42,44 The dispersive waveforms observed with the addition of Te are a result of the trapping and release of holes in the Te-doped layer prior to transport into the higher mobility non-doped a-Se region results. For longer time scale measurements (e.g., fluoroscopy), the trapped holes will be released and collected, appearing as latent signal (i.e., lag). These results indicate that improving hole transport in this layer can improve the temporal performance of the sample. One known method of improving hole transport is Cl co-doping, as demonstrated by Kasap.²⁵

4.4. Limitations and Future Work

Our experiments used optical light sources, resulting in the omission of direct x-ray interactions in a-Se. When operating in avalanche mode, x-rays absorbed by HARP will experience different avalanche gain depending on their depth of interaction.⁷ This will lead to additional source of signal and noise in x-ray images, primarily at high spatial frequencies. The effect of direct interaction on the performance of SHARP-AMFPI will be determined in a future prototype utilizing the composite HBL. The modeling of the Te-layer focused primarily on transient condition TOF measurements and hole transport. To get a comprehensive understanding of the material, additional modeling should be conducted focusing on steady-state conditions, as well as performing reverse bias experiments to extract electron transport properties.

5. Conclusion

Experiments with HARP samples using the composite HBL show avalanche gain of 16, with improved temporal performance compared to the initial SHARP-AMFPI prototype and reduced dark current compared to the second prototype. Measurements of Te-doped HARP samples demonstrate improved OQE to 532 nm light, up to 0.43 for 18.2 wt % Te at $10\mathrm{V}/\mu \mathrm{m}$. TOF experiments indicate that hole traps within the Te-doped layer are a significant source of temporal performance degradation. Future work will further optimize the metal oxide deposition to reduce the risk of dielectric breakdown at avalanche fields and conduct additional experiments with co-dopants of the Te-doped a-Se layer to reduce temporal effects and retain the benefits of improved OQE to CsI:Tl emissions.

Biographies

Corey Orlik has been a graduate student in DRIL Laboratory at Stony Brook University since 2019. His research interests are in a-Se based flat-panel detector devices and medical imaging physics.

Sébastien Léveillé is the manager of the deposition and encapsulation teams and processes at Analogic Canada Corporation. His research interests are the development of deposition processes required to manufacture medical imaging detectors.

Salman M. Arnab is a scientist innovation at Analogic Canada Corporation. He received his PhD in electronic materials and devices from Concordia University. His research interests are in semiconductor materials and device physics.

Adrian F. Howansky is a clinical assistant professor at Stony Brook University Hospital, Department of Radiology. He received his PhD in biomedical engineering from Stony Brook. His research interests are in medical imaging physics.

Jann Stavro is a medical physics resident at Stony Brook University Hospital, Department of Radiology. He received his PhD in biomedical engineering from Stony Brook. His current research interests include development of energy-resolving photon counting detectors, fabrication of novel amorphous selenium-based sensors, and digital breast imaging applications.

Scott Dow received his PhD in biomedical engineering from Stony Brook University. His research interests are in medical imaging physics.

Safa Kasap is a research professor of optoelecronic materials, electrical, and computer engineering at the University of Saskatchewan. He received his PhD from the Imperial College of Science and Technology, University of London, London, United Kingdom. His research interests are in photoconductors and x-ray detectors, electrical and optical properties of materials, and optoelectronic glasses.

Kenkichi Tanioka is a research professor of radiology at the State University of New York at Stony Brook. He received his PhD in electronic engineering from Tohoku University, Sendai, Japan. His research interests are in highly sensitive image sensors using avalanche multiplication and ultra-high-definition television systems (8K) for medical use.

Amir H. Goldan is an associate professor of radiology at Weill Cornell Medical College, Cornell University (New York, New York, United States). His research is focused on the physics and instrumentation of positron emission tomography and photon counting x-ray imaging.

Wei Zhao is a professor of radiology at the State University of New York at Stony Brook. She received her PhD in medical biophysics from the University of Toronto. Her research interests are in medical image sensors and digital breast tomosynthesis.

Disclosures

The authors have no financial interests to disclose. A portion of this work was previously published in the SPIE Medical Imaging 2022 conference proceedings (C. Orlik, et. al., “Improved Temporal Performance and Optical Quantum Efficiency of Avalanche Amorphous Selenium for Low Dose Medical Imaging,” Proc. SPIE, 12031, (2022)).

Code and Data Availability

All data in support of the findings of this paper are available within the article.