1. Introduction
Ion implantation has been widely applied in the semiconductor industry, as introducing dopants is an easy and fast way to engineer the electrical and optical properties of semiconductors. Over the last decades, deterministic single-ion implantation has attracted wide interest in the semiconductor field because of its application in solid-state quantum technology across various material systems, including silicon and diamond. Some non-exhaustive examples of Si-based single-dopant devices include donors coupled to quantum dots [
1] for charge [
2], electron [
3,
4], and nuclear spin [
5,
6] qubits (quantum bits). Alternatively, single-color centers in diamond substrates, including nitrogen-vacancy (NV) centers [
7] and coupling silicon-vacancy (SiV) centers [
7,
8], are studied as quantum electrodynamics (QED) devices based on diamond technology. Motivated by the proposed quantum applications, the demand for deterministically placing single dopants into nanostructured devices has prompted the development of various techniques related to silicon and diamond material systems [
9,
10]. Single-ion implantation is achieved through accurate control over the ion’s final position and the number of implanted ions. Single-atom lithographic techniques based on scanning probes have successfully achieved the positioning of single dopants with nanometer-scale precision [
11,
12,
13]. However, this technique is currently limited to a small number of species and is relatively slow. In contrast, direct ion implantation offers less precision in terms of dopant atom positioning but offers more flexibility in the choice of ion species, potentially allowing for faster and more scalable processes [
9,
14]. The primary challenges in direct ion-implantation methods lie in accurately counting individual ions as they approach the substrate and precisely predicting their final position within the sample. Several techniques have been developed to monitor the number of ions reaching the sample during single-ion-implantation processes. One of the most commonly used methods involves the detection of secondary electrons emitted upon the impact of ions on the sample [
15,
16]. This approach requires the presence of a secondary electron detector and can be applied to nearly all samples (provided that the ion energy is sufficiently high to yield a detectable secondary electron emission). Alternatively, other frequently used techniques exploit integrated structures within the implanted sample to generate detectable signals during the process. Integrated structures such as PiN diodes, which utilize electron–hole pairs generated by ion–matter interactions, can produce a detectable signal upon free-charge capture [
17,
18]. Additionally, integrated field-effect transistors (FETs) are employed, where ion implantation modulates the drain current [
18,
19]. However, these methods are exclusively applicable to samples featuring PiN or FET structures, thereby limiting their utility in various areas.
In the present work, we present an innovative single-ion detection method using an independent silicon carbide sensor to be placed ahead of the sample to be implanted. The sensor is a sub-micrometer SiC membrane realized through a state-of-the-art, doping-selective electrochemical etching process [
20,
21]. The described sensor geometry was employed here for detecting ion beams in the MeV energy range. In this setup, the ions lose only a portion of their energy (ΔE) in the device and are transmitted further without significant influence on the impact trajectory and total energy (E) of the ions. The electron/hole pairs generated through the ion–sensor interaction are amplified and collected, resulting in a distinct signal corresponding to each ion passing through the SiC membrane. The results were compared with measurements obtained using a reference silicon detector. Moreover, a study of the alteration in the ion beam after crossing the membrane was conducted, and the lateral ion straggling, a crucial parameter for single-ion-implantation applications, was calculated and compared with simulations.
2. Materials and Methods
The device tested is an advanced silicon carbide ultra-thin radiation sensor engineered as a free-standing membrane with a parallel-plate electrode configuration. This device is a semiconductor Schottky barrier diode consisting of an ultra-thin n
− silicon carbide active layer characterized by a low doping concentration of 10
14 cm
−3 on top of an inert n
+ highly doped silicon carbide substrate approximately 370 µm thick with a doping concentration of 10
18 cm
−3. The fabrication of the free-standing membrane at the core of the sensor was accomplished through a state-of-the-art, doping-selective electrochemical etching (ECE) technique, which allowed for precision material removal down to sub-micrometer thickness levels. In more detail, the electrochemical etching of silicon carbide in hydrogen fluoride (HF)-based solutions consists of two steps: the first step is the oxidation of SiC driven by holes, and the second step is the dissolution of the formed SiO
2−x in HF [
22]. In the case of the highly doped 370 µm substrate, holes were generated by tunneling effects. In contrast, for the low-doped n-type SiC epitaxial layer, tunneling was negligible. Therefore, the thin epitaxial layer acted as a stopping layer for the etching process, hence resulting in the formation of the free-standing membrane [
20]. The total sensor area (5 × 5 mm
2) was divided into four independent pads, and the ECE process was carried out within a 2 mm diameter circular region in the central area of the device. In this study, only one of the four pads was connected to the data acquisition system and analyzed.
The metal contact, needed for both the ECE process and for the subsequent sensor operation, was established by depositing a 30 nm aluminum layer on the front surface of the device to create a Schottky contact. In contrast, the back contact was applied after the ECE process and involved a 100 nm aluminum layer. The metal depositions were conducted using an electron-beam (E-beam) evaporation system. A schematic structure of the SiC membrane sensor is presented in the inset of
Figure 1b.
The main characterization of the sensor was performed by exposing the device to accelerated ions of different masses and energies in the MeV range using the ion beam facility of the Ruđer Bošković Institute [
23]. Techniques based on the interaction of MeV ion beams with materials offer a powerful analytical framework for semiconductor detector characterization [
24,
25]. The device was mounted in a vacuum irradiation chamber attached to the 6 MV Tandem Van de Graaff electrostatic accelerator. The accelerator is equipped with a sputtering ion source used for the production of a wide range of ion species, from light ions such as H or Li to very heavy ions (up to Au). Ions are accelerated and transmitted through a range of ion beam optics elements downstream to the experimental end station, where the samples are positioned. In our setup, an ion microprobe end station was employed, allowing the focusing of ion beams to a micrometer-sized spot and enabling the scanning of the beam across the sample surface. These experimental conditions were utilized to acquire spatially resolved information about the ion–sample interaction as determined by the beam spot size. This setup was crucial for testing the device for single-ion detection, as the ion beam current could be reduced to ~ Hz rates and positioned in different spatial regions of interest of our device.
In our experimental scenarios, the ion beam was transmitted through the SiC membrane portion of the sensor, leaving only a portion of the energy inside, and was stopped 6 cm downstream on the in-beam-positioned Si detector, which was a Passivated Implanted Planar Silicon (PIPS) detector with very thin top dead layer. The PIPS detector will be referred to as a Scanning Transmission Ion Microscopy (STIM) detector (Canberra Semiconductor, Olen, Belgium), as it was used to detect transmitted ions. Both devices were connected to the same low-noise signal-processing chain, which was based on a charge-sensitive preamplifier (ORTEC 142A) and a shaping amplifier (ORTEC 570), both provided by ORTEC, Oak Ridge, TN, USA. This geometry enabled independent detection of ions by the device under test (the SiC membrane) and a well-characterized solid-state Si detector (the STIM detector). When an ion interacts with a semiconductor detector, it produces electron–hole pairs that are collected by the electric field applied through the electrodes. This technique is often referred to as the ion beam induced charge (IBIC) technique [
24]. Signals collected from the sensor electrode can be used to quantify different parameters, such as the deposited energy, transient collection behavior, timing properties, and so on. When combined with a scanning microbeam setup, the IBIC technique can be seen as a 3D-like microscopic technique for the investigation of charge transport properties in semiconductor detectors.
A 4 MeV O
3+ ion beam was employed to precisely determine the thickness of the silicon carbide membrane. A schematic representation of the experimental setup used during this investigation is presented in
Figure 1a. The incident ions that passed through the SiC sensor deposited a portion of their energy (ΔE, approximately 45%) within the free-standing membrane sensor. Subsequently, these ions were collected by the STIM detector, and the acquired signal was represented as a count-versus-energy plot (
Figure 1b). After subtracting the energy deposited in the STIM detector, which is represented by the peak position in
Figure 1b, from the initial ion beam energy, the energy deposited in the SiC membrane sensor (ΔE) was determined. Furthermore, the full-width half maximum of the peak in
Figure 1b was used to calculate the uncertainty in the membrane thickness. The Stopping and Range of Ions in Matter (SRIM) Monte Carlo simulation tool [
26] was used to estimate the energy loss ΔE/Δx (eV/nm) of the beam in the sensor, enabling the calculation of the membrane sensor thickness T
SiC. Using this method, a total sensor thickness of T
SiC = 727.3 ± 57.6 nm was calculated. Considering the thickness of the aluminum electrodes, the membrane active layer thickness resulted in about 597 nm. The relatively high 8% error associated with this measurement can be primarily attributed to the surface roughness of the membrane resulting from the doping-selective ECE process used for the formation of the SiC membrane [
20,
21]. The ion-counting fidelity of the SiC membrane was determined using the same oxygen beam and experimental setup while employing the IBIC technique.
The electron–hole pairs generated as a result of energy deposition during ion–membrane interactions were collected, and 2D-IBIC maps were generated using the homemade software SPECTOR v2.0 [
27]. During the acquisition of IBIC signals, a reverse bias of −5 V was applied to the SiC Schottky diode. After this first interaction, the ions had enough energy to reach the STIM detector, thereby allowing for a simultaneous generation of a second IBIC map corresponding to transmitted ions. A comparison between the two acquired maps was performed to determine the number of recorded events in the two devices while assuming a 100% collection efficiency in the STIM detector. This comparison enabled the evaluation of the single-ion detection efficiency of the SiC membrane sensor.
The decrease in ion energy was not the sole effect of the interaction between the ion beam and the SiC membrane sensor. As the ions collide with the atomic electrons of the solid sensor, the trajectory angle of the ions in the material can be altered [
25]. This phenomenon, commonly denoted as “ion lateral straggling”, increases the uncertainty in the final position of ions within the implanted sample. In our experiment, we quantified ion lateral straggling resulting from the interaction between ions and the SiC sensor by using a finely machined metal grid with a defined pitch dimension. The grid was positioned between the membrane and the STIM detector, allowing the scanning transmitted ion beam to form a projection image of the grid. Using this experimental setup, a 10 MeV C
4+ ion beam was scanned across the grid to acquire 2D-IBIC maps both with and without the presence of the membrane. This enabled the determination of the beam spot dimension in the two cases using the knife-edge analysis technique based on the grid projection.
4. Discussion
In deterministic ion implantation, the exact counting of ions as well as its spatial precision represent an ongoing challenge. The device presented here utilizes a membrane solid-state sensor and a low-noise, charge-sensitive electronic chain. This system collects a signal generated by ions transmitted through the sensitive membrane volume. In our experiments, the energy loss in the membrane active layer was about 1.5 MeV, and the number of generated pairs was on the order of 105 pairs per ion. However, typical ion-implantation energies are in the range of a few hundred keV, which requires low energy loss inside the membrane and the detection of a signal derived by 103 ÷ 104 electron–hole pairs. These limits impose the use of a nanometric-thin membrane and a very low noise generated both by the detector and by the stage electronics.
Concerning the thickness of the sensor, a 100 nm SiC free-standing membrane can be produced using the ECE process described earlier. Thin membranes of this nature have already demonstrated favorable mechanical properties, including a high fracture strength and deformation [
28]. The energy loss in ionization, i.e., the energy that ions lose in collisions with atomic electrons generating free charges in the solid material, depends on the mass and energy of the implanted ion. Using a SiC sensor with a 100 nm SiC epitaxy sandwiched between 20 nm and 70 nm aluminum electrodes, a typical dopant such as P at 250 keV loses a total energy of 190 keV in the sensor and generates approximately 9 × 10
3 electron–hole pairs in the sensor active layer. With low noise, this system will allow the implantation of a deterministic number of 60 keV P ions. The detector noise is mainly determined by the leakage current and by the detector capacitance. The leakage current in our detector was sufficiently low (a few pA) thanks to the wide bandgap of the silicon carbide semiconductor. The capacitance, on the other hand, considerably influenced the noise level due to the large sensor area (~1.6 mm
2) and the low sensor thickness (~730 nm). Although a 100-nanometer thickness may have a negative impact on the sensor capacitance, this effect can be substantially alleviated by reducing the surface area of the sensor, resulting in an enhanced signal-to-noise ratio. To minimize electronic noise, a custom charge-sensitive amplifier with effective capacitance matching of the input stage to the detector capacitance can be employed. Further experiments are in the planning stage to utilize an even thinner device along with upgrades to the signal-processing electronics.
Concerning the measured lateral ion straggling, the obtained value of 8.15 µm seems to be relatively high for single-ion-implantation applications. This high value introduces a significant level of uncertainty in determining the final position of the ions, potentially compromising the deterministic nature of the implantation process. However, it is important to note that the primary contribution to the final ion beam size is attributable to the divergence of the beam in the region between the SiC membrane sensor and the sample. In this work, this distance is represented by the distance between the SiC sensor and the metal grid (DGrid = 0.76 ± 0.14 mm). The membrane straggling contribution was very low (rSRIM = 2.42 ± 0.30 nm calculated with SRIM) compared to the final beam dimension. Therefore, by reducing the distance between the SiC sensor and the implanted sample, a higher determination of the ion’s final position can be achieved. For example, by reducing the DGrid distance to a few micrometers (5 ÷ 10 µm), the membrane straggling contribution on the sample will be 19.1 ÷ 35.7 nm (calculated with the same θSRIM angle). In this case, the initial beam dimension rBeam will strongly affect the final beam lateral profile (in this experiment, rBeam was 3.43 ± 0.54 µm).
Hence, through the reduction in the initial beam size to a few tens of nanometers and the detector thickness to 100 nm, it becomes possible to attain a final beam size of approximately 100 nanometers. This reduction significantly mitigates the uncertainty associated with the final position of the implanted atom.