Fabrication of a Hydrogenated Amorphous Silicon Detector in 3-D Geometry and Preliminary Test on Planar Prototypes

Abstract

Hydrogenated amorphous silicon (a-Si:H) can be produced by plasma-enhanced chemical vapor deposition (PECVD) of SiH₄ (silane) mixed with hydrogen. The resulting material shows outstanding radiation hardness properties and can be deposited on a wide variety of substrates. Devices employing a-Si:H technologies have been used to detect many different kinds of radiation, namely, minimum ionizing particles (MIPs), X-rays, neutrons, and ions, as well as low-energy protons and alphas. However, the detection of MIPs using planar a-Si:H diodes has proven difficult due to their unsatisfactory S/N ratio arising from a combination of high leakage current, high capacitance, and limited charge collection efficiency (50% at best for a 30 µm planar diode). To overcome these limitations, the 3D-SiAm collaboration proposes employing a 3D detector geometry. The use of vertical electrodes allows for a small collection distance to be maintained while preserving a large detector thickness for charge generation. The depletion voltage in this configuration can be kept below 400 V with a consequent reduction in the leakage current. In this paper, following a detailed description of the fabrication process, the results of the tests performed on the planar p-i-n structures made with ion implantation of the dopants and with carrier selective contacts are illustrated.

1. Introduction

Hydrogenated amorphous silicon (a-Si:H) is a disordered semiconductor obtained via plasma-enhanced chemical vapor deposition (PECVD) of a mixture of silane (SiH₄) and hydrogen at temperatures of 250–300 °C [1]. The resulting material has an irregular arrangement of atoms resulting in not all Si–Si bonds being saturated, leading to the presence of dangling bonds (DBs) that are related to the presence of intermediate states between the valence and the conduction bands. The introduction of hydrogen into the mixture has the purpose of passivating most of these dangling bonds. In amorphous (non-hydrogenated) silicon (a-Si), the density of defects due to DBs is 10¹⁹ cm⁻³, whereas for a-Si:H this density can be as low as 10¹⁵ cm⁻³. Although the amount of hydrogen necessary to reach these levels of DB passivation is approximately 1% atomic, device-grade a-Si:H usually exhibits hydrogen concentrations between 4% and 12%. The hydrogen percentage in the material affects its bandgap (increasing the hydrogen percentage leads to a larger bandgap) and depends on deposition conditions such as processing temperature. However, temperatures above 350 °C lead to hydrogen desorption, transforming the material into a-Si. This relatively low deposition temperature facilitates the adhesion of a-Si:H on many different substrates, like glass (including Pyrex and fused silica), stainless steel, crystalline silicon, silicon oxide, aluminum, coated ceramic, chromium-plated brass, copper-coated printed circuit board (PCB), and organic materials like polyammide, PEN (polyethylene naphthalate), PET (polyethylene terephthalate) and PI (polymide), heat-resistant organics/inorganic polymers (like ormocer^®), and even on top of electronic devices in CMOS (complementary metal-oxide semiconductor) technology [2]. Usual deposition techniques for detector-quality a-Si:H are PECVD with plasma excitation at radio frequency (13.56 MHz) [3], very high frequencies (between 27 and 150 MHz) [4] or even with microwave frequencies [5], and hot-wire deposition [6].

The relatively wide bandgap of a-Si:H and its disordered nature results in a low charge carrier mobility (1–3 cm² V⁻¹ s⁻² for electrons and 0.01 cm² V⁻¹ s⁻² for holes) and a charge collection time in a 30 µm-thick diode of below 15 ns [7]. A-Si:H is a semi-isolating material with a resistivity of above 10¹⁰ Ω cm. This resistivity can be lowered by more than seven orders of magnitude by doping [8]. However, doping of a-Si:H creates additional defects, which consequently lowers the carrier lifetime of the material [9]. Therefore, doped layers cannot be used as active layers in photodiodes or particle sensors. For this reason, p-i-n diode structures, not direct p–n junctions, are preferred with relatively thin p- or n-doped layers. Usually, this material is doped with the addition of PH₃ (n-type doping) or B₂H₆ or trimethylboron (p-type doping) in the process gas mixture. It was also demonstrated that similar doping levels and conductivity can be achieved by ion implantation at low temperatures [10]. This doping technique was demonstrated by the fabrication of various devices [11,12]. The feature of a-Si:H that makes it attractive for particle detector developments is its remarkable radiation hardness. This has led to significant research activities aiming at single minimum ionizing particle (MIP) detection [13,14,15]. In Ref. [16] the radiation tests of a 32.6 µm-thick p-i-n diode irradiated at CERN’s Proton-Synchrotron with 24 GeV protons up to a maximum fluence of 7 × 10¹⁵ p/cm² is reported; an increment in leakage current of a factor of 2 at 9 × 10⁴ V/cm electric field was observed, and this increase disappeared after 24 h of annealing at 100 °C. Although this material has an extremely high radiation hardness, the main limitation of a-Si:H planar detectors is their poor signal-to-noise ratio for the detection of MIPs, which is yet to exceed a value of 5. The reason for this low S/N ratio is the very high depletion voltage (with electric fields up to 10⁵ V/cm) that generates a high leakage current (up to about 1 µA/cm²). Furthermore, the low charge collection efficiency (below 50% for a 30 µm-thick diode) due to the disordered nature of the a-Si:H lattice structure contributes to the decrease in the overall value of the signal-to-noise ratio even if some effort has been performed to increase the charge collection efficiency in these devices [17]. The depletion voltage of an a-Si:H detector V_d is related to the density of DBs (N_db number of dangling bonds/volume in cm³) according to the following formula:

$$V_d=\frac{e\times {\mathrm{N}}_{\mathrm{db}}\times {\mathrm{d}}^2}{2\epsilon }$$

where e is the electron charge and d is the thickness of the depleted layer. If we express d in µm and V_d in volts this formula becomes V_d = k × d², where k ranges from 0.3 to 1.2 when N_db ranges from 5 × 10¹⁴ to 2 × 10¹⁵ cm⁻³. Furthermore, a large thickness of the detector reduces the charge collection efficiency due to charge trapping over a longer distance.

Increasing the depletion voltage increases the leakage current at full depletion, and this increases the noise. To reduce noise and increase the signal we propose fabricating a p-i-n-structured a-Si:H detector in 3D geometry that makes it possible to keep a relatively small collection distance (namely, a 25–35 µm inter-electrode spacing) with a detector thickness of up to 100 µm or more, increasing the total charge generated in the detector by an MIP. Maintaining a small distance between the electrodes is important for keeping the leakage current low, hence reducing the noise. In Figure 1 we show the baseline design for our detector, where a-Si:H is deposited over a low-resistivity p-type silicon substrate that distributes the bias to the p-type electrodes on the detector. The n-type electrode collects the signal for the readout electronics connected on the top contacts. This configuration is the basis of development for Si-3D for high-luminosity (HL) LHC applications. [18]. Figure 2a depicts the electrode arrangement as seen from a top-down perspective, where one n-type finger-type electrode is surrounded by four p-type finger-type electrodes. Additionally, Figure 2 shows alternative electrode configurations: Figure 2a is the baseline configuration, Figure 2b shows a configuration with eight p-type finger-type electrodes surrounding an n-type electrode, Figure 2c shows a configuration with one finger-type n-type electrode surrounded by four trench-type p-type electrodes, and Figure 2d shows all alternate trench n-type and p-type electrodes. Trench electrode configurations have been studied and used for the realization of excellent time-resolution silicon detectors, for example in the TIMESPOT collaboration. [19].

Table 1. Summary of c-Si and a-Si:H characteristics.

Properties	c-Si	a-Si:H	Remarks
Bandgap (eV)	1.14	1.65–1.8 (for device grade)	For a-Si:H given by Tauc’s gap, depends on H content
Density (g/cm3)	2.3290	2.20–2.31 (device grade)
Hydrogen content (at. %)	0	4–15% (device grade)—up to 50% possible
e-h creation energy (eV)	3.6	4–6
e-h creation/µm for MIP	~80 (1)	~80	Stopping power similar (almost same density, effect of H negligible)
e drift mobility (cm2s−1V−1)	1450	1–3
h drift mobility (cm2s−1V−1)	450	0.01
e mobility lifetime (cm2V−1)	~5 × 10−3	≤5 × 10−5	For a-Si:H, depends on deposition and treatments
h mobility lifetime (cm2V−1)	~3 × 10−3	≤10−7	For a-Si:H, depends on deposition and treatments
Full depletion field (V/µm)	<1	≥10	Depends on defect density
Charge collection efficiency	97–99%	40–50% (2)	Depends on charge type, field, and time window
Device leakage current not irradiated (A/cm3)	10−3–10−5	≤10−3 (3)	For a-Si:H, thickness and field dependent (a-Si:H bulk, for 10 µm-thick diode at field of 105 V/cm, non-irradiated)
Device leakage current irradiated 5 × 1015 p (24 GeV)/cm2) (A/cm3)	10−1–10−3	<3 10−3 (4)	For a-Si:H, thickness and field dependent (a-Si:H bulk, for 10 µm-thick diode at field of 105 V/cm,

(1) Experimentally observed value. Theoretical value is 108 for a stopping value of 1.66 MeV cm² g⁻¹ for an MIP in c-Si. (2) For transient experiment. Due to deep trapping, holes are, in principle, not collected. Collection efficiency >80% in steady state. (3) Leakage value for planar diode. Diode on non-planar substrates and unpassivated edges can lead to significantly higher leakage. (4) Experimental value not available. Increase of leakage by 2–3 for 32 µm-thick diodes.

shows a comparison between a-Si:H and c-Si physical parameters.

2. The Fabrication of a 3D Detector

The fabrication of a 3D detector in a-Si:H is somewhat different than the construction of a 3D detector in crystalline silicon due to the constraint of keeping the temperature of processing below 250–300 °C in order to avoid hydrogen desorption.

The process starts with the deposition on support wafers (typically a Cz p-type silicon wafer with a resistivity below 10 Ωcm) of the a-Si:H performed via PECVD with a VHF-excited plasma at the frequency of 70 MHz. The plasma is composed of a mixture of silane and hydrogen (ratio of 1 to 1) at temperatures around 200 °C and will produce a layer with a thickness of around 100 µm.

After that, both the ohmic and junction columns are defined and etched using a deep reactive ion etching (DRIE) apparatus based on the Bosh process. This technique allows for the etching of a hole with a diameter of a few microns and with a high aspect ratio (more than 50). The ohmic columns are etched deeper than the a-Si:H layer (partially etching the c-Si substrate) and the junction columns are etched to approximately 20 µm shallower than the a-Si:H thickness.

Afterwards, the etched holes in the a-Si:H layer must be doped in the internal surfaces. In order to create the p-i-n structure in the detector, two independent masks are fabricated to define an n-plus region and a p-plus region. Doping the detector inside the holes cannot be performed using common techniques for planar structures (i.e., PECVD deposition of doped a-Si:H) as these deposition techniques cannot achieve the required conformality in deep trenches [20,21]. Therefore, two doping options are considered:

• Option 1: atomic layer deposition (ALD) of conductive metallic oxides for the creation of selective contacts for each type of charge carriers: titanium oxide for electron-selective contacts and tungsten or molybdenum oxide for hole-selective contacts. Since the oxides for charge-selective contacts are quite resistive, depositing a conductive metal inside the holes or trenches by PLD (pulsed laser deposition) is foreseen.
• Option 2: ion implantation of boron (p-type doping) and phosphorous (n-type doping) followed by a temperature activation annealing process below 200 °C or an additional metal deposition by PLD if needed.

In order to have a reasonable time constant (order of 10 ns) for charge collection, the resistance of the electrodes should be in the order 1 MΩ.

After the doping, a contact hole is opened and the metal (typical Al 1% silicon) is deposited and defined. The passivation, based on a multilayer of Si₃N₄ and SiO₂, is deposited by the low-temperature PECVD technique and subsequently opened to allow the contact of the metal layer. The final step is the metal deposition of the backside that is used as bulk bias contact.

3. Results of the Phase I Prototype

The complexity of the proposed 3D detector geometries presents many technological challenges regarding fabrication. The construction of prototype devices allows for the challenges of 3D a-Si:H device fabrication to be addressed in a sequential manner. Phase 1 prototypes feature a planar p-i-n architecture and aim to assess the feasibility of the doping options introduced in the previous chapter. Measurements investigate the performance of the doped device layers to produce an efficient junction for biasing the intrinsic layer and provide an effective charge collection and current rectification. A second generation of prototypes (under design) aims to accurately measure the relation between the hole diameter and the rate at which a-Si:H is etched via DRIE. The proposed 3D geometries require the p-type electrodes to be embedded at a sufficient depth to erase a shallow layer of the a-Si:H substrate, and the n-type electrodes are to be fully contained within the a-Si:H device layer (Figure 3). Therefore, second-generation prototypes aim to assess the feasibility of etching p-type and n-type electrodes in a single process, exploiting the feature of the DRIE process in which larger-diameter holes turn out to have a faster etching rate.

All measurements in this section were taken using a Keithley 237 Source-Measure Unit (SMU).

3.1. Leakage Current Test on Option 1

Concerning option 1, related to the use of selective contacts, vertical structures were fabricated on chromium-plated glass substrates. As electron-selective contacts, ZnO:Al (aluminum-doped zinc oxide (AZO)) and TiO₂ (respectively, 60 nm and 10 nm thick) layers were implemented on the substrate side and compared. As a hole-selective contact, MoO_x (20 nm thick) protected by an indium thin oxide layer (60 nm) was deposited. All selective contacts were deposited by sputtering. Note that for 3D architecture, ALD should be used to enable conformal deposition in the holes or trenches.

Figure 4 depicts the initial I/V measurements performed on two vertical diodes with a sensitive area of 0.25 cm², providing comparisons between the two different materials employed as electron-selective contacts. The results of both vertical diode structures displayed exceptional rectifying behaviors. The generated leakage currents in TiO₂ and AZO electron-selective contact devices at −5 V applied bias were measured as two and three orders of magnitude larger than the leakage currents under +5 V bias, respectively. Furthermore, the obtained leakage current of approximately 4 nA/cm² in both devices at +40 V bias is comparable to that obtained in samples fabricated via ion-implantation doping. Thus, this doping technique is identified as an acceptable alternative to implantation and will be explored in detail in future studies.

3.2. Leakage Current Tests on Doping Option 2

In order to verify the feasibility of performing a doping process inside the holes fabricated via the DRIE process to form effective 3D p- and n-type electrodes, the two doping options explained in the previous section were verified with the construction of two types of prototypes: vertical diodes and lateral diodes in a planar configuration, obtained from the deposition of a layer of 10 µm of a-Si:H on a heavily doped p-type c-Si substrate (300 µm thickness).

Vertical diodes, shown in Figure 5, used the p-type c-Si silicon substrate as the p-type biasing electrode and the n-type electrode was obtained via ion implantation of phosphor ions. Diodes with several geometries were produced, namely, single diodes, 2 × 3 diode arrays, 2 × 10 diode arrays, and strip detectors.

Figure 6 shows an I/V test on two samples of a 4 mm² vertical diode. As shown in the figure, the diode acted as a rectifying junction and the leakage current with 50V bias was about 147 pA with a current density of 3.6 nA/cm² on the best sample.

Figure 7 presents the various configurations and geometries of the lateral diode structures. These lateral diodes were fabricated with the p-type and n-type electrodes deposited on the surface of the a-Si:H layer and a thin (500 nm) layer of silicon oxide separating this a-Si:H layer from the silicon substrate. The sensitive region of the lateral diodes existed between the p- and n-type electrodes, which were doped with boron and phosphorous, respectively.

In comparison to the I/V characteristics of the vertical diodes (Figure 6), the I/V curves of the lateral diodes in Figure 8 displayed an even more pronounced rectifying behavior and exceptionally low leakage current values, ranging from 339 pA (best small detector) to 1030 pA (worst large detector).

4. Testing with X-ray

Since these devices were too thin for MIP signal detection, the radiation detection capabilities of these diodes using X-rays was investigated. Figure 9a,b show the dose linearity of the response of devices with different thicknesses of the substrate and two different areas of the junction. The sensors were irradiated by a 6 MV medical linear accelerator in standard conditions (source-to-surface distance of 100 cm, field size of 10 × 10 cm², and an equivalent water depth in a plastic phantom of 1.5 cm). Despite the sensors being configured in photovoltaic mode and the high degree of disorder of the substrate material, the adopted configuration of a doped area above an intrinsic amorphous layer allowed for the collection of the charge in proportion to the substrate thickness and the area of the sensor as expected.

In Addition, the timing response of the sample with 10 µm-thick substrate and an area of 5 × 5 mm² (Figure 9c) showed a stable response during the irradiation (beam-ON) within 2.2%, well within the expected fluctuations of the machine output [22].

Additional measurements were taken in order to determine the stabilization time after turn-on and the linearity of x-ray flux measurements. Figure 10 presents the measurements of the dark current versus time in a vertical diode under 25 V bias possessing a 10 µm active layer thickness and a 2 × 2 mm² active area. Figure 10a shows the measurement on a 2700 s time scale after turn-on and Figure 10b shows the same data on a 200 s time scale. These results shows that the measurement of the dark current became stable after about 200–250 s, with current fluctuations observed after this time remaining in the order of 80 pA.

Using the same device, we measured the linearity of response to X-rays using an X-ray tube with a tungsten cathode biased at 10 kV. Figure 11 shows the diode current response (after dark current subtraction) to irradiation versus the tube current for diode bias at 25 V (Figure 11a), 50 V (Figure 11b), and 100 V (Figure 11c). The resulting diode responses at all three applied biases displayed a good linearity of induced diode currents for X-ray tube currents in the range of 0 to 200 µA, with excellent linearities observed for tube currents above 30 µA.

5. TCAD Simulation of a Full 3D Detector

Different geometrical configurations are being studied by means of device-level simulator Synopsys Sentaurus TCAD [23]. A proper description of the a-Si:H material has been included in the material library of Sentaurus TCAD [24]. In this study, full column electrodes, hybrid trench/column electrodes (mini trenches) and full trench electrodes were considered (Figure 12). For sake of computational effort, a 3D slice of the whole device depth was considered, looking at the electric field map distributions (steady-state simulations) and at the current vs. time response to a particle hit (transient simulation) of the different configurations.

Depending on the electric field distribution at the particle hit position, different amplitude and timing characteristics of the current response of the device can be appreciated. A faster and higher-current response can be seen with trench configurations (full lines in Figure 13) with respect to full column configurations (small dashed lines in Figure 13), in particular for particle hits close to the read-out electrode (blue and green curves in Figure 13). All currents in the graph were calculated for a 300 V bias.

By integrating the time responses at the read-out electrode, the charge collection properties of the different configurations at different biasing conditions can be evaluated. This study enabled the choice of the optimal detector configuration.

6. Conclusions and Outlook

The fabrication of a 3D a-Si:H particle detector as described in this paper presents a new detector technology based on the peculiar properties of this material. A preliminary description of this fabrication process has already been given in another paper [25]. The fabrication process has been described here in detail and the R&D process towards the clarification of the feasibility of the detectors has been explained. The early results of the first generation of prototypes have been presented, demonstrating the capability of producing a working planar p-i-n diode with a quite low leakage current. We consider the first phase of prototype development to be satisfactory and have begun the second phase.

Detectors will be also tested after fabrication with radioactive sources like ⁹⁰Sr, electron accelerators, and collimated X-ray sources. After this, preliminary test detectors will be irradiated with protons, neutrons, and gamma rays from a cobalt-60 source for the evaluation of displacement and ionization damage. A new test run with the above-mentioned sources is foreseen.

In order to obtain charge collection efficiency maps of charge particles interacting with the detector, a test at the Australian Nuclear and Science Technology Organization (ANSTO—Lucas Heights—NSW, Australia) Ion Beam Induced Charge facility (IBIC) is also foreseen. The IBIC technique will be used to measure the charge-collection efficiency of the 3D device under test (DUT) using a ~1 µm spot size-focused ion beam scanning the DUT surface. The response of the DUT stimulated by the particle is correlated with the position of the beam, allowing a map of the efficiency of the device to be created [26,27]. The IBIC measurements will be performed using the heavy ion microprobe upon the ANTARES 10 MeV Tandem Accelerator. Various sources are available, including the 5.5 MeV helium ion beam with a range of 28 µm in silicon, which has already been used successfully in previous works [28].

The device will be also evaluated as a dosimeter using X-ray beams for imaging (keV energy range) and therapeutic applications (MeV energy range). The presence of two doping species allows for the use of these devices in photovoltaic mode (with no bias applied across the contacts). This property is particularly important for dosimetry, where a continuous current is read out from the device and dark currents limit the minimum threshold of the detectable dose delivered. The prototype test structures fabricated within the 3D-SiAm collaboration have demonstrated exceptionally low leakage currents and stable, linear responses to delivered doses via X-ray irradiations. These promising results verify the quality of the substrate, and the employed implantation methods are sufficient to guarantee the required MV X-ray response sensitivity and reproducibility outlined in the Code of Practice for dosimetric measurements [29].

The wide range of materials onto which a-Si:H can be deposited on opens up the possibility of the direct deposition of the detector material onto the readout chip. This feature has already been exploited in the past [2]. This detector technology, called TFA (thin film on ASIC), gave the best results in terms of signal-to-noise ratio in a planar a-Si:H detector for MIP detection (S/N ratio of 5 for an MIP). The deposition of a planar detector on the present technology front-end chip may increase this signal-to-noise ratio to higher values. Furthermore, a new technological frontier for the future may be the direct deposition of a 3D a-Si:H detector on the readout chip. The usage of DRIE technology for the fabrication of electrodes with two different lengths and selective implantation on different finger-type electrodes will allow for the fabrication of a detector like the one shown in Figure 14, where the n-type electrode reaches the readout pad of the readout chip and the p-type electrode does not reach the readout chip. After ion implantation, a passivation and a metallization layer to distribute the biasing contacts to the p-type electrodes will be added.

7. Patents

The a-Si:H 3D detector is patented in Italy (patent number 102018000010735) and is patent pending in Europe (application number EP19831899.0) and in the U.S. (application number 17297441).