1. Introduction

The deep level defects that act as charge carrier traps have high importance in semiconductor industry and applications of semiconductor devices . These defects are being created during: (a) semiconductor growth process, (b) electronic device fabrication and (c) operation in harsh environments. We focus our attention on defects created in semiconductor devices exposed to irradiation by ions  and electrons   in the MeV energy range.

It is well known that high energy ionizing particles traversing through or being stopped in a sensitive volume of semiconductor device, deposit part of their initial energy in atomic elastic collisions displacing atoms from their lattice sites . Those primary defects might annihilate or reorganize themselves with impurities to form stable deep defects. Defect accumulation in reasonably low concentrations (well below an extended defect formation threshold value) might modify electronic transport properties of charge carriers introduced into active region of a device, and consequently alter or deteriorate its performance .

Silicon carbide is widely regarded as a semiconducting material, which has desirable physical properties (high thermal conductivity, large saturation electron drift velocity, high electric breakdown field, and excellent thermal stability) for manufacturing of electronic devices suitable for applications in harsh environments, i.e., high radiation  , high temperature , and high power applications  . In this study we investigate the radiation hardness of single crystal 4H-SiC material for particle detection in intense radiation conditions. The radiation hardness of SiC detectors has been studied using irradiation with neutrons , protons  , gamma photons  , electrons , light ions   and heavy ions .

The 4H-SiC epitaxial growth technique achieving high growth rate and large area crystal uniformity has been developed recently  . Utilizing a modified epi-reactor setup, a thickness uniformity of 1.1% and a doping uniformity of 6.7% for a 65-mm-radius area has been achieved, while maintaining a high growth rate of 80 μm/h. Epi-layers of lightly doped 4H-SiC, obtained by this technique, show low concentrations of Z_1/2 and EH_6/7 defects and a very good carrier lifetime of ∼1 μs making it very suitable for electronic applications discussed previously. We used the grown epi-layer to fabricate Schottky barrier diodes (SBD in further text) for ionizing radiation detection and monitoring, and exposed them to a radiation hardness test.

2. Experimental

n-Type silicon carbide SBDs were produced on nitrogen-doped (up to 4–5 × 10¹⁴ cm⁻³) epitaxial grown 4H-SiC single crystal layers approximately 47 μm thick . The Schottky barrier was formed by evaporation of nickel through a metal mask with patterned quadratic apertures of 1 × 1 mm, while Ohmic contacts were formed by nickel sintering at 950 °C in Ar atmosphere on the back side of the silicon carbide substrate. The reverse negative bias was connected to the front Schottky contact, and the back Ohmic contact of prepared 4H-SiC SBD was grounded.

The whole testing procedure for this particular 4H-SiC material follows in slightly modified conditions the experimental protocol previously used on silicon diodes  and . The quality of the 4H-SiC SBDs was characterized by I–V and C–V measurements. Only samples with the lowest reverse current have been considered for our radiation hardness study. Additional care was taken during final sample selection by performing the scanning Ion Beam Induced Charge (IBIC) microscopy   in frontal mode (ion microbeam with a very low rate of up to 1000 cps is scanned perpendicular over a front metal contact) on each pre-selected sample to establish a good uniformity of charge collection efficiency (CCE) across the whole active SBD volume. Subsequent irradiation conditions meant to generate defects in 4H-SiC layer have been optimized for the direct single ion detection and on-line fluence monitoring by counting the number of pulses in each pixel of irradiated areas utilizing the IBIC technique. Details about specific microprobe irradiation conditions (ion microbeam rates, pixel dwell times, and irradiation areas) used to achieve the most suitable fluence values in regions of interest of tested 4H-SiC SBD samples for both Deep Level Transient Spectroscopy (DLTS) and IBIC analysis are given in  . It is important to highlight that different samples have been prepared for IBIC and DLTS analysis but irradiation conditions per pixel (unit surface area) were almost the same. The differences specified in  have been used only to keep irradiation times of individual areas and whole samples within reasonable timeline.

The selected samples were homogenously irradiated with a 2 MeV He ion microbeam at the ANSTO heavy ion microprobe facility  . All irradiations have been performed at the room temperature and zero bias. The samples have not been thermally treated after irradiations. Negligible error might have been caused by a dead time of the microprobe DAQ system in given working conditions.

For C–V and DLTS measurements the micro-beam of the 1 μm spot size and 10,000 cps ion rate (∼1 × 10¹² cm⁻² s⁻¹) was rapidly raster scanned multiple times over the total irradiated area of approximately 1 × 1 mm in order to avoid an instantaneous implantation of the full dose as well as ion beam supported self-annealing of primary defects, and also achieve a homogenous single ion implantation over extended time. The scan area was divided in 512 × 512 pixels with a dwell time per pixel equal to 500 or 1000 μs, i.e., on the average up to 5 ions were implanted in each pixel before the micro–beam was moved to the next pixel position. Total time required to homogenously irradiate SBDs was 33 min (#3), 83 min (#8), 166 min (#4) and 333 min (#12), respectively.

Deep traps created in silicon carbide were characterized using DLTS. DLTS measurements were performed at temperatures between 80 and 380 K. No deep traps have been detected in non-irradiated samples. Eight different initial delays from 0.5 to 100 ms were simultaneously obtained from one temperature scan in order to determine the DLTS signature of formed defects (activation energy and the trap concentration).

Detrimental influence of defects formed in active 4H-SiC epi-layer on the CCE of selectively irradiated SBDs has been investigated by the scanning IBIC microscopy. The amplitude of the IBIC signal was recorded for each ion implanted in nine (3 × 3) squares of 100 × 100 μm²each with a 100 μm gap between them  . Each square was irradiated with increasing fluence value from 10⁹ to 5 × 10¹¹ cm⁻² using optimized microbeam conditions  . Two samples have been pattern irradiated in the same way using 4 MeV He (#1) and 2 MeV He (#2) damaging ion (micro) beam (further in text referred as DIB) respectively. After patterned irradiations the different ions probes or probing ion (micro) beams (further in text referred as PIB, PIB = 1, 2, 3, 4, and 6 MeV He) and reverse bias settings of −100, −200 and −400 V have been used for IBIC microscopy of partially damaged SBDs to monitor degradation of the total charge collection. For radiation damage studies we considered only IBIC events originating from a central part of the irradiated square areas 50 × 50 μm²showing uniform IBIC response, with no changes towards edges of irradiated areas. The reported normalized CCE(Φ_i) values represent the centroids of the induced charge pulse spectra relevant to each irradiated regions corresponding to particular fluence value Φ_i and normalized to the signal resulting from the pristine material.

3. Results and discussions

3.1. Electrical characterization of 4H-SiC SBD

The electrical characterization of a typical pristine 4H-SiC SBD sample suitable for studies is shown in Fig. 1. The measured reverse current is below 4 pA over the full reverse bias range (Fig. 1a), while the forward current increases sharply above ca. +0.6 V. The low reverse current through SBD is necessary for IBIC measurements, i.e., detection of the current transient induced by motion of free charge carriers (generated by single ions) in a depleted region of tested device. As prepared the 4H-SiC SBDs were able to withstand a reverse bias of up to −450 V maintaining a low reverse current and electronic noise. The measured C–V characteristic ( Fig. 1b) was used to calculate the free carrier concentration profile in the region of interest of a pristine SBD ( Fig. 1c). It was calculated to be of the order of 4–5 × 10¹⁴ cm⁻³. The calculated depletion thickness (w) as a function of applied reverse bias is shown in Fig. 1(d). Calculated approximate w values for IBIC settings of −100, −200 and −400 V are 15.0, 20.5 and 28.1 μm, respectively. The voltage settings required for the DLTS measurements were chosen from the comparison of SRIM [25] simulations for the extent of disordered region dense with primary displacements, where the threshold displacing energy of 35 eV for Si and 22 eV for C atoms [26] have been used respectively, following the single 2.0 MeV He ion implantation in silicon-carbide (later shown in Fig. 5c) and doping profiles of implanted samples calculated from corresponding C–Vmeasurements ( Fig. 2). The applied reverse bias was varied between −3 and −5 V with a filling pulse of 0.5 V to sample the region of interest in SBDs. The cumulative decrease of the free carrier concentration in the irradiated SBDs across the whole section from surface to the extent of implantation range ( Fig. 2) is supporting the fact that electron traps are formed within implantation range. A variation of the calculated net free carrier concentration can be simulated with an exponentially decreasing function of damaging ion fluence in the tested range (inset of Fig. 2).

3.2. Trap characterization in He ion irradiated 4H-SiC

The effect of alpha-particle implantation on radiation defect production in 4H-SiC n-type epilayer is shown in Fig. 3 which summarizes results of DLTS studies performed at temperatures from 80 to 380 K using 4H-SiC SBD samples irradiated with different fluences of raster scanned 2 MeV He ion microbeam. All shown spectra were measured at a reverse bias of −4 V and the initial delay of 50 ms, and no annealing in order to thermally stabilize produced defects has not been performed. Those measurement settings were chosen, among different we have applied during the experiment, for the clarity and in order to show all the observed defects in our samples.

Implantation of high energy alpha-particles yields quite complicated DLTS spectra showing broad overlapping peaks, which are most likely due to superposition of several states with close activation energies. One broad feature with its maximum at about 280 K has been observed in the sample implanted up to the lowest fluence value of 1 × 10⁹ cm⁻² (black curve). In further text we denote this electron trap as E1. An increase of implantation fluence up to the total value of 1 × 10¹⁰ cm⁻² (red curve) gives rise to a new state represented by a broad peak with its maximum at about 330 K and a shoulder-like peak at about 280 K. Since the shoulder appears at similar position and possibly resembles the state observed in sample irradiated with lower ion fluence of 1 × 10⁹ cm⁻² (black curve), we denote these two states as E1 and E2. Further increase of implantation fluence gives rise to an additional state; for the highest measured value of 2 × 10¹⁰ cm^.2 (blue curve) we observed two broad overlapping peaks at 330 and 370 K with even less pronounced shoulder-like feature at 290 K. We denote those three states as E1, E2 and E3. The E3 state associated with high temperature overlapping peak (around 370 K) observed in the sample with the highest fluence was not investigated due to the cryostat thermal limitations.

Although the obtained DLTS spectra were complicated (thus, each peak might consist of more than one component), activation energies of electron emission for the E1 and E2 traps have been determined from Arrhenius plots of T 2-corrected electron emission rates (inset in Fig. 3) as follows: E1 = (0.48 ± 0.01) eV for the 1 × 10⁹ cm⁻², E1 = (0.49 ± 0.03) eV and E2 = (0.77 ± 0.03) eV for the 1 × 10¹⁰ cm⁻² and E2 = (0.67 ± 0.03) eV for the 2 × 10¹⁰ cm⁻². All traps appear in concentrations from 1 up to 3 × 10¹² cm⁻³, as determined from the DLTS measurements.

It is very difficult to assign the specific microstructure of the defects associated with the detected levels, particularly in the case of E1 state which does not appear to be stable. A broad peak associated with the E1 state shifts its temperature position and changes shape with increasing fluence. A state (or mixture of close lying states)exhibiting similar features in that temperature range was previously observed in 4H-SiC epilayers irradiated with protons and alpha-particles [27], carbon ions [28] and electrons [29] and [30]having energies in the MeV range. They also noted that this state (or mixture of closely lying states) is unstable, due to a low thermal stability of related defects. The shift of DLTS peaks for low implantation fluences was interpreted as a release of the stress produced by defect clusters. During single ion implantation the defect re-arrangement occurs as individual ion cascades begin to overlap for higher ion fluences. Taking everything into account this E1 state could be related to highly mobile defects, as carbon interstitials, but conclusions about the exact microscopic structure cannot be made without further experiments such as annealing temperature dependence of DLTS peaks. Laplace DLTS [31] might be also an useful technique to clarify this issue.

The observed E2 state resembles a defect already reported in the literature, the so-called Z_1/2[32] and [33]. It should be noted that in the case of the fluence 1 × 10¹⁰ cm⁻² the estimated value for the activation energy is a little bit higher than the reported values, but this is due to the fact that the E1 and E2 traps are closely spaced, and the low temperature E1 peak (so called “shoulder”) overlaps with the E2 peak, which leads to the certain errors in energy estimation. However, the situation clears out for the sample irradiated with 2 × 10¹⁰ cm⁻², where without interference of the low temperature peak we have been able to correctly estimate the energy of the E2 trap. The Z_1/2 defect is one of the most studied defects in SiC. It is an intrinsic, carbon-related defect. It is an acceptor like defect with an extreme thermal stability up to 2000 °C [30]. It has been shown that Z_1/2 defect limits the minority carrier lifetime and therefore strongly affects device properties [34]. Very recently the microstructure of Z_1/2 was revealed to be a single carbon-vacancy [35] and [36]. More detailed studies are ongoing to confirm those results.

3.3. CCE degradation in selectively irradiated 4H-SiC SBD

The normalized charge collection efficiency as a function of the irradiation fluence, CCE(Φ_i), for both 4 and 2 MeV He irradiations is shown in Fig. 4 and Fig. 5 respectively. The figures summarize the results for all the probe ion beams used (1, 2, 3, 4 and 6 MeV He) as well as different bias voltages (sections a, b and d of Fig. 4 and Fig. 5). For comparison the SRIM simulated ionization depth profiles of PIBs and the vacancy-recoil depth profile of DIB, as well as the calculated extent of the depletion region for given bias, are shown in Figures c for both irradiation energies. Their interplay is essential for understanding the charge collection mechanism and for the interpretation of the experimental CCE data [4]. The CCE decreases for different (PIB, DIB, BIAS) settings at different rates similar to previous observations in silicon diodes [3]. In contrast to silicon diodes irradiated with even higher He fluences, CCE distributions for 4H-SiC SBD cannot be fully described by the linear model [4] over the entire DIB fluence range (1 × 10⁹–5 × 10¹¹ cm⁻²), even for the highest applied bias voltage of −400 V, associated with a complete generation of free carriers within a depleted region and pure drift motion of carriers towards collecting electrodes. In fact, while within the 1 × 10⁹–1 × 10¹¹ cm⁻² fluence range, particularly for −400 and −200 V bias settings, the CCE decrease exhibits a linear dependence (inset graphs in Fig. 4 and Fig. 5 for partly damaged 4H-SiC SBD with 4 MeV He and 2 MeV He respectively), above a fluence of 1 × 10¹¹ cm⁻²significant deviations from linear dependence occur for some studied cases.

In the following we discuss the decrease of CCE for both irradiation energies, taking into account (i) the previous observation from DLTS measurements that electron traps are created in the ion implanted region of 4H-SiC SBD (3.B) and (ii) the fact that free electrons are moving from the negatively biased Schottky contact towards the grounded back Ohmic contact, while holes are moving in the opposite direction:

(1)

When the probe range is small compared to both the damaged layer depth and the depletion depth (PIB = 1, 2, 3 MeV He and DIB = 4 MeV He), the highest normalized CCE(Φ_i) value is observed, very close to 100% up to fluence value of 10¹¹ cm⁻². Only a part of free electrons drifting through damaged region is trapped, and this results in a small degradation of the normalized CCE. On the other hand, despite all created holes contribute to the induced charge as they do not cross the highly damaged region, their contribution to induced charge is small due to a short drift distance. Thus, the electron contribution dominates, even in case of a partial electron trapping.

(2)

In the same case of deep damage created by 4 MeV He, the normalized CCE(Φ_i) value decreases with the energy of a PIB. As a PIB with higher energy is penetrating closer to the damaged layer, the average drift distance of later captured electrons decreases for more penetrating particles. Although in terms of carrier drift length, a hole contribution concurrently increases to compensate the electron decrease, in irradiated material holes have a larger probability to encounter traps during their drift towards the Schottky contact. This effect is only distinguishable for fluences of 10¹¹ cm⁻² and larger, where carrier trapping becomes more pronounced.

(3)

In the case of relatively shallow damage created by 2 MeV He, the highest normalized CCE(Φ_i) value is recorded for the deepest penetrating particle (PIB = 6 MeV He) at bias of −400 V. In this case the drift distance towards the Schottky contact for holes generated at the end of range of the PIB is larger than that of electrons collected at the back electrode, thus providing the dominant hole contribution to the induced charge signal. Only a small fraction of the created electrons has to drift through the damaged layer in order to reach the collecting electrode; however, those electrons created at a depth above the damaged layer, contribute to the induced charge the most (the largest average drift distance if not trapped). So when the trapping probability of those electrons increases for larger fluence Φ_i values, the normalized CCE(Φ_i) measured by a very deep probe decreases to the lowest CCE(Φ_i) value for each Φ_i separately.

(4)

In the 2 MeV He irradiation case when a probe ionization maximum is just below the shallow damaged layer (PIB = 3 MeV He and DIB = 2 MeV He, green symbols), high values of CCE(Φ_i) are measured across the whole fluence range because in that particular studied case (i) the majority of created free electrons are insensitive to the damage and (ii) the electron contribution dominates over the hole contribution.

(5)

When the PIB range is equal to the DIB range (orange for 4 MeV and cyan symbols for 2 MeV He) the CCE(Φ_i) decrease is larger if compared to lower probe energies, because a large portion of free electrons originating from the damaged layer can be immediately trapped. The CCE decrease becomes more pronounced for fluence values above 10¹¹ cm⁻² corresponding to a larger trapping probability.

(6)

Additionally, in a case of a deep probe (PIB = 6 MeV He) some carriers are generated beyond the extent of depleted region at reverse biases of −100 and −200 V. Their motion is governed by slow diffusion in the field-free region until they reach the edge of the depleted region. The IBIC measurements were performed at the shortest possible shaping time of the spectroscopy amplifier (0.25 μs) to consider only the contribution to the CCE of fast drifting charge carriers, neglecting the effects of diffusing and/or de-trapping charge carriers. Corresponding CCE distributions (PIB = 6 MeV He, BIAS = −100 and −200 V; red symbols) are not considered further here.

Fig. 6 shows selected IBIC spectra for 4 and 2 MeV irradiated samples, figures a–c and d–f respectively. These illustrate the origin of the non-linear behavior manifested as bending towards CCE values lower than estimated by a linear dependence for the two largest fluence values of 2 × 10¹¹ and 5 × 10¹¹ cm⁻². These IBIC peaks are extracted from the recorded event-by-event list files for PIB detection in each partially damaged area of SBD corresponding to particular DIB fluence value. Their position (channel number) corresponds to the measured amplitude of induced charge signal, while their height corresponds to the normalized yield. IBIC peaks are (1) slightly widening (FWHM increases) and (2) slowly moving towards lower channel values (amplitude of recorded IBIC signal decreases). They further illustrate that up to the fluence value of 1 × 10¹¹ cm⁻² the shape of IBIC peak does not differ from the initial Gaussian shape extracted from a non-irradiated area of the SBD. But IBIC peaks significantly change their centroid position (amplitude), peak resolution (FWHM) and even shape (deviations from Gaussian distribution) for DIB fluences of 2 × 10¹¹ and 5 × 10¹¹ cm⁻². This abrupt change in spectral performance of tested SBDs could be related to: (1) significant decrease of charge carrier lifetime due to increasing Z_1/2 concentration in damaged region with increasing fluence (3.B), (2) decrease of free carrier concentration with increasing fluence (3.A) and (3) changes of electric field profile within disordered region created by DIB.

Additionally, formation of new type of defects, i.e., complex (cluster) defects within disordered region created by DIB, even for fluence values in the 1 × 10¹¹ cm⁻² range (not investigated here using the single ion implantation), could have an important role on performance of the tested 4H-SiC SBD. Roccaforte et al. suggested the point defect reorganization/clustering in SiC irradiated with 1 MeV Si ions above the fluence value of 5 × 10⁹ cm⁻²[15] and we already proved a direct formation of the small clusters of di-vacancies in n-type CZ silicon (N_eff ∼ 10¹⁴ cm⁻³) with no thermal treatment after the irradiation with 8.3 MeV Si ions to the fluence value of 1 × 10¹⁰ cm⁻² [8].

4. Conclusions

A detector grade 4H-SiC material underwent a comprehensive radiation hardness test using light He ions in the MeV energy range. The reported CCE(Φ_i) values are normalized to the signal from pristine area of samples and they do not refer to the absolute CCE value.

A very limited CCE performance degradation has been observed using the partially damaged 4H-SiC SBD irradiated up to the fluence value of 1 × 10¹¹ cm⁻². The CCE decrease is linearly dependent of fluence in the same range. Above He irradiation fluences of 1 × 10¹¹ cm⁻² a significant deviation from the linear behavior of CCE decrease has been observed in some cases. Previous radiation resistance studies performed on the SBD made from n-type 4H-SiC material with similar N-doping concentration do not mention a deviation from the linear behavior of CCE if irradiated with electrons [6] and [12], protons[11] and [12] or other light ions [11]. If we restrict ourselves to only the case of the probe being the same as the damaging ion (PIB = DIB = 2 MeV He), and take into account only the results obtained for the highest bias of −400 V ( Fig. 5a, cyan symbols), our experimental normalized CCE values as a function of fluence closely match those reported by Lee et al.[11] and can be described by a linear decrease within statistical error.

We also observed a substantial change in the detection performance of partly damaged 4H-SiC SBD irradiated with He fluence values above 1 × 10¹¹ cm⁻² even for the highest applied bias voltage.

The sudden change in spectroscopy performance and the non-linear behavior of the CCE have not been observed previously in our studies on FZ Si detector diodes irradiated up to 1 × 10¹² cm⁻²[37] and CZ Si diodes irradiated up to 5 × 10¹¹ cm⁻²[4] of He ions having similar energies, in closely matching experimental conditions (both in terms of probing-damaging ion combinations and the electric field distribution, taking into account main differences in doping and applied bias).

Inspite being obtained for the fluence values below 1 × 10¹¹ cm⁻², the results from electrical and DLTS measurements suggest that the observed seemingly abrupt changes to CCE for high fluence values above 1 × 10¹¹ cm⁻², which reflect substantial changes to free carrier transport properties within the ion implantation range, could be related to the initial displacement of carbon atoms and localized formation of deep defects primarily associated with stable carbon vacancies (Z_1/2) in the active region of the 4H-SiC SBD.

The novelty of such effects requires a deeper understanding of device and material properties, and a full explanation will involve the utilization of a wider range of probing and damaging ion species and energies.