Heavy-Ion Effects in SiC Power MOSFETs With Trench-Gate Design

Silicon carbide (SiC) power MOSFETs with trench-gate structure have been irradiated with heavy-ion broad beam and microbeam. Microdose effects resulting in higher subthreshold drain leakage were observed when irradiating the devices at low drain-source voltages and reported for the first time for SiC power devices. Increasing the drain-source bias during the exposure, single-event leakage current (SELC), characterized by microbreaks in the gate oxide, was measured. The accumulation of microbreaks eventually led to a complete gate rupture. The differences with respect to the SiC planar-gate MOSFETs and the impact of these results on the testing procedures for the two technologies are discussed.


I. INTRODUCTION
S ILICON carbide (SiC) power MOSFETs have gained significant attention in recent years due to their unique properties and advantages over traditional silicon-based MOSFETs [1].SiC power devices are capable to withstand higher internal critical fields than their silicon counterparts, making them ideal for high-power applications such as electric vehicles, renewable energy systems, and industrial equipment [2], [3].They are also known for their low ON-resistance and high switching speeds, which results in lower power consumption in the semiconductors and leads to overall higher energy conversion efficiency.SiC power MOSFETs are commercially available in two gate designs: planar and trench [4], [5].While both the designs are based on the same fundamental principles of operation, they differ in their physical structure and performance characteristics.Planar-gate SiC power MOSFETs have a surface horizontal channel; this design is simpler to manufacture and has widely been used in the industry for many years.On the other hand, in the trenchgate architecture, the gate is etched into the SiC epitaxial layer, the channel is vertical, and the JFET resistive region is avoided, leading to superior ON-state performance.However, the trench-gate architecture is more complex to manufacture The authors are with the Advanced Power Semiconductor Laboratory (APS), ETH Zurich, 8092 Zurich, Switzerland (e-mail: martinella@aps.ee.ethz.ch).
Color versions of one or more figures in this article are available at https://doi.org/10.1109/TNS.2024.3379458.
Digital Object Identifier 10.1109/TNS.2024.3379458and can be more expensive than their planar counterparts.Nowadays, symmetric and asymmetric trench designs are available on the market [4].Fig. 1 shows the schematic layouts for a SiC planar-gate MOSFET and an asymmetric-trench MOSFET.
SiC power devices are also becoming increasingly popular in the field of space exploration.The possibility to operate SiC devices at high voltage, current, and temperature, and the ability to switch at comparably high frequencies with reduced losses would greatly improve the overall electrical efficiency of on-board systems, resulting in a more compact design, lower weight, and enormous cost saving [3], [6], [7].For example, due to the capability of SiC to operate at temperatures higher than 460 • C, SiC integrated circuits are currently considered for Venus surface exploration missions [8], [9].However, despite the beneficial characteristics, the adoption on SiC power devices in space is still limited due to the sensitivity to radiation, which increases the risk of single-event effects (SEEs) [10].Galactic cosmic rays (GCRs) represent the higher risk in space, as they are composed of energetic heavy ions that can pass practically unimpeded through a typical spacecraft, inducing SEEs [11], [12].
Depending on the linear energy transfer (LET) of the ion and the bias conditions (i.e., drain-source and gate-source voltages) during the exposure, different types of SEEs have been observed in SiC power MOSFETs [14].When irradiating the device in OFF-state and at low drain-source voltages (V DS ) during the exposure, higher drain current (I D ) is observed due to the ionization generated by the impinging particle.This effect is called charge collection and it is nondestructive [15], [16].Partial degradation such as single-event leakage current (SELC) is observed at higher V DS , characterized by permanent increase in gate (I G ) and drain leakage currents with increasing heavy-ion fluence [13], [14], [15].After the exposure, the device is still operational, but its characteristics are exceeding the specification limits due to increased leakage current.Previous works reported two different mechanisms for SELC [14], [19]; initially, microbreaks are formed in the gate oxide, inducing a drain-gate leakage path, as previously reported for Si power MOSFETs [31].At higher voltages, a second mechanism is observed, characterized by higher increase in I D with respect to I G , which has been associated with the creation of extended defects in the device (i.e., different types of stacking faults) [20], [21].Finally, at higher V DS during the exposure, complete and irreversible loss of the electronic component due to catastrophic failure such as single-event burnout (SEB) and single-event gate rupture (SEGR) are observed [22], [23], [24], [25], [26], [27], [28].
The majority of the data in literature on the reliability of SiC power MOSFETs exposed to heavy ions have been obtained testing planar-gate architectures.Despite the already reported improved electrical performance [4], and higher tolerance when exposed to atmospheric neutrons and 200-MeV protons [26], little data are available on the reliability of SiC trench technology tested with heavy ions.Two previous works reported experimental results [32] and simulation studies [33] on SiC MOSFETs with double-trench architectures.The reliability of the gate oxide was highlighted as the main criticality.
This work explores the effects of heavy-ion irradiation in noncommercial SiC trench power MOSFETs by means of broad beam and microbeam experiments.The components are obtained from the manufacturer of the commercial off-theshelf SiC power MOSFETs.Increased subthreshold leakage current is observed at low drain-source voltages during exposure.Such mechanism was associated with microdose effects caused by localized trapping of positive charges in the gate oxide, which lowers the threshold voltage.Therefore, this drain-source leakage mechanism is related to the MOSFET's channel, as previously reported for Si technologies and as discussed in Section III-B.This effect was not analyzed in previous works [32], [33].At higher voltages during the exposure, SELC is observed, characterized by microbreaks in the gate oxide which eventually induce a complete gate rupture.These mechanisms are described with the respective dependency on the ion LET and bias conditions during irradiation.The differences in the failure mechanism with respect to the planar design are also highlighted.Finally, the impact of these results on the testing procedures for the two architectures is discussed.

II. EXPERIMENTAL METHOD
Two experiments were performed with broad beam: one at the Heavy-Ion Facility (HIF) of the Université Catholique de Leuven, Leuven-la-Neuve, Belgium, and the other at the RADiation Effects Facility (RADEF) in the Accelerator Laboratory of the University of Jyväskylä, Finland.Different ion species were selected for irradiations, as reported in Table I.Yet another experiment was carried out using a microbeam at the UNIversal Linear ACcelerator (UNILAC) micro-probe line at the Helmholtzzentrum für Schwerionenforschung (GSI) in Darmstadt, Germany.The main differences among these experiments are the beam characteristics, as reported in Tables I and II.The broad beams used at HIF and RADEF delivered a wide and uniform beam of radiation with a diameter of 2.5 cm at HIF and squared beam of 2 × 2 cm 2 at RADEF.In contrast, the microbeam used at UNILAC provided a beam with a focal spot of 500 nm to target micrometric areas selected with an optical light microscope situated in the chamber [19], [34].To ensure the irradiation with a preset number of particles and to avoid double hits at the same position, a fast electrostatic beam switch is controlled by the hit detection system for microbeam irradiation.When a hit is detected, the microbeam is switched off and the focal point of the probe is moved to the new x-y-coordinates.
In all, 30 noncommercial bare die SiC power MOSFETs with a trench-gate structure were selected as devices under test (DUTs).The devices are rated for 1.2 kV, with nominal ON-state resistance (R DS(ON) ) of 12 m .The gate and source pads were connected by aluminum wire bonds of 300-µm diameter, while the drain connection was established at the soldered bottom pad.The source terminal was directly grounded on the board.Only two wires (one for the gate and one for the source) were used to minimize the shadowing effect.Two Keithley source measure units (SMUs), models 2636 and 2410, were used to bias gate (V GS ) and drain (V DS ) and monitor the gate and drain currents (I G and I D ), respectively.Each bare die was exposed at normal incidence of the beam with respect to the device surface.The devices were irradiated in a vacuum chamber, at constant V DS (see Section III) and V GS = 0 V during each irradiation run.

A. Heavy-Ion Effects in SiC Trench Power MOSFETs
Three different irradiations performed with krypton (Kr) at HIF were selected as representative of the effects observed in the trench devices when exposed to heavy ions with different LETs.The irradiation discussed in the following paragraphs was performed at V DS = 70 V, 120 V, 400 V, and it is shown in Fig. 2(a)-(c), respectively.For each of them, the panel on the left shows the drain and gate leakage currents during irradiation; the central panel shows the transfer characteristics measured after the run at V GS = 0, . . ., 5 V and V DS = 1 V, and the right panel shows the blocking characteristics measured at V GS = 0 and V DS = 0, . . ., 1000 V.The pristine measurements for the same device are reported using dashed lines.All the measurements were performed in vacuum.
The first effect observed when irradiating the trench devices at a sufficiently high V DS bias (i.e., V DS = 70 V for Kr) is reported in the left panel of Fig. 2(a).Steps associated with the ion strikes are observed in I D current during the exposure, and the resulting leakage current does not recover immediately when the beam is switched off.However, different from the case of SELC observed in SiC devices with planar gate Authorized licensed use limited to the terms of the applicable license agreement with IEEE.Restrictions apply.
structures, no permanent degradation of I G is measured.This is confirmed by the IV measurement reported in the central and right panels of Fig. 2(a), performed directly after irradiation.From the transfer characteristics, it is concluded that the channel is still functional, and there is neither a threshold voltage shift nor an increase in I G .However, a difference is observed in the subthreshold region of the MOSFETs, where an increase in I D leakage current is measured for V GS < V th with respect to the pristine measurement.Conversely, no difference in I D is measured once the device is in ON-state.This effect was also confirmed by repeating the measurement using a Keithley Parametric Curve Tracer PCT-4B and following the specification from the JEDEC 183A procedure for the transfer characteristics measurement (not reported here for brevity) [35].From postirradiation analysis, it was concluded that the described mechanism is caused by microdose effects, as it is further discussed in Section III-B.This effect has not been previously reported for SiC planar MOSFETs irradiated at normal incidence and it is reported for the first time for SiC power devices.The blocking characteristics shown in the right panel of Fig. 2(a) confirmed the increase in I D leakage.
Fig. 2(b) illustrates a second mechanism of damage induced by the ions when V DS during irradiation is set over a second threshold (i.e., V DS = 120 V for Kr).Steps in I D and I G are observed during exposure, indicating the formation of microbreaks in the gate region.In addition to what is described in Fig. 2(a), an increase in I G is also observed in the transfer and in the blocking characteristics in the central and right panels.For the latter, a gate-drain leakage path is observed for the current for V DS > 600 V.This mechanism can be explained as SELC degradation, as previously reported for SiC planar devices in [14] and associated with the formation of microbreaks in the gate oxide.However, in contrast to the results observed for the planar architecture where I D and I G increase with the same rate and in absolute values, a higher I D leakage is observed for the trench-gate architecture due to the microdose effect in subthreshold, which increases with increasing total fluence.
A third and destructive effect induced at higher V DS during irradiation (i.e., V DS = 400 V for Kr) is shown in Fig. 2(c).Higher I D and I G degradation is observed during exposure, an additional increase in I G was measured in the transfer characteristics, whereas a clear gate-drain leakage path is observed in the blocking characteristics in the right panel (i.e., overlap of |I D | and |I G |).For the latter, the measurement was stopped at 440 V as the currents reached the compliance of 1 mA.This effect was classified as a destructive gate rupture induced by an accumulation of micro-breaks caused by SELC, rather than being cause by a single particle, such as in the case of SEGR [31].Finally, the current drop observed around 280 s in the first panel of Fig. 2(c) is an artifact induced by the auto-ranging of the source measure unit.
A summary of the three effects observed during the broad beam and the microbeam tests and their dependency on the drain-source bias during irradiation and ion LET are reported in Table III.The effects are indicated as: 1) microdose; 2) SELC; and 3) gate rupture.Furthermore, no SEB events were measured up to V DS conditions selected for the tests.

TABLE III SUMMARY OF HEAVY-ION-INDUCED EFFECTS (THRESHOLD VALUES)
However, since in this work all the experiments were performed using prototype devices and considering that the design varies between device types and producers, the results cannot be transferred to all the commercial SiC trench MOSFETs without further analysis and testing.

B. Microdose Effects
After exposure, additional measurements were performed on all the samples showing microdose degradation (I D increase in subthreshold).Fig. 3(a) shows the transfer characteristics for five devices irradiated in different conditions.Increased leakage through the channel is observed in the subthreshold region.However, no effect on conduction and no rigid shift of the threshold voltage is observed, such as in the case of X-ray irradiations [36].The channel resistance (R ch ) of the component is moderately decreased, showing an impact on subthreshold, where it has a larger contribution on the total R DS(ON) .A dependency on the LET, the bias applied during the exposure and the total fluence is observed.In Fig. 3(b), the blocking characteristics measured at V GS = 0 V and V GS = −5 V is reported for a device irradiated with Xe at V DS = 60 V and a fluence of 10 6 ions/cm 2 .Increased drain leakage is measured when performing the sweep at V GS = 0 V.However, when repeating the measurement applying V GS = −5 V, the leakage drain current is reduced and remains comparable to pristine levels.These results indicate that the increase observed in the first measurement is caused by the degradation of the subthreshold channel leakage current, and not by the creation of extended defects as previously reported for SiC planar-gate MOSFETs characterized by higher I D increase during exposure [20].
A similar increase in drain leakage current in the subthreshold region of commercial Si trench power MOSFETs has been reported by Felix et al. [37], [38] when irradiating at normal incidence with heavy ions.The mechanism is described as caused by the formation of a parasitic transistor due to the charge deposited by the primary ion along the entire length of the channel.Holes generated by the incoming particle can get trapped in the gate oxide, thereby locally reducing the effective threshold voltage of the channel region close to the ion path.Such an effect belongs to the class of microdose effects [37], [38], [39], [40], [41], [42], [43], which are hybrid between the SEE and the total ionizing dose (TID) effects.As in the case of SEE, the damage created by the single particle is strongly localized.Like TID, the damage induced by different ion strikes is expected to be cumulative and subject to annealing effects.A proportionality with ion fluence is expected due to the cumulative nature of these effects.However, even for an equal number of strikes, a great spread remains in the drain leakage currents due to stochastic nature of this effect [41].The influence of gate bias and ion LET was further analyzed in [38].
Changing the angle of incidence of the ion relative to the die surface has a significant impact on the amount of charge that could be deposited along the entire length of the channel (assuming the channel length significantly larger than the diameter of the ion track) [38], [39].This effect is shown to be reduced for Si trench power MOSFETs irradiated with heavy ions when the angle is increased from 0 • (normal incidence) to 90 • (parallel to the surface of the die and perpendicular to the vertical channel) [38].However, this behavior is not uniquely associated with the trench gate power MOSFETs; Shaneyfelt et al. [42] also reported microdose degradation in planar-gate Si power MOSFETs irradiated with protons and neutrons.In this case, as the secondary particles produced by the elastic and nonelastic interactions between the protons/neutrons and the material can be emitted in all the directions [44], [45], an angular dependence as observed for the heavy-ion irradiations is not expected.
The results reported in this work for SiC trench power MOSFETs are consistent with a model of parasitic leakage current path along the channel length, as described for Si technology.Therefore, it is concluded that microdose effects can also be observed in SiC trench power MOSFETs irradiated at normal incidence.The angular dependence might explain why enhanced degradation associated with microdose effects has not been reported for SiC MOSFETs with planar-gate exposed to heavy ions (i.e., the irradiations are generally performed at normal incidence due to the limitation in many facilities of changing the angle).However, based on the aforementioned results for Si-based device technology [39], [42], it cannot be excluded that microdose effects can be induced in SiC planar-gate MOSFETs as well.Angular studies should be performed to have an overview of such effects in SiC devices and, consequently, their implications for applications in radiation environments.

C. Degradation Rate
The degradation rates as a function of V DS during irradiation with broad-beam and microbeam are presented in Fig. 4(a) and (b) for the drain and the gate leakage currents, respectively.The degradation rate is defined as the difference between the leakage current measured at the end of the run (after exposure to the beam) and at the beginning of it (before exposure) normalized by the fluence and the active area of the die.The degradation rates were calculated considering a single device per ion type, exposed to multiple runs.The values obtained for the drain current are higher with respect to the previously reported degradation rates for SiC planar gate MOSFETs [46].This is most likely due to the microdose effects observed already at low voltages before entering the range where SELC occurs.Conversely, this is not the case for the gate current and the values obtained are lower than the ones previously reported for SiC planar gate MOSFETs [46].
As expected, a proportionality between the degradation rates and the ion LET is observed for both the drain and gate leakage currents.For Xe, irradiations were performed using two different fluxes (i.e., 1000 and 10 000 ions/cm 2 s).The results obtained from the irradiations with the lower flux are represented in dashed lines.The divergence between the two results highlights an effect of the flux selected during irradiation, with higher impact for V DS > 250 V.This effect should be taken into consideration for future radiation tests.

D. Considerations for Radiation Tests
The radiation tests performed in the laboratories do not always match the exact conditions of applications because of time, cost, and testing procedures.Based on the presented data, some aspects should be considered when testing different technologies of SiC power MOSFETs.
The angle of incidence of the primary ion with respect to the die surface has a strong influence on the magnitude of microdose effects depending on the gate and channel geometry [39].Therefore, the testing conditions with heavy ions should be carefully designed based on the technology, Authorized licensed use limited to the terms of the applicable license agreement with IEEE.Restrictions apply.and the angle under which the irradiations are performed should be reported.Conversely, following the suggestions reported for Si technologies [42], protons and neutrons test might be performed at any angles to determine whether a part is susceptible to enhanced degradation associated with microdose.
In addition, due to the cumulative nature of microdose degradation, the testing procedure might strongly impact the results.Attention should be paid to the total fluence accumulated for each bias condition.Furthermore, due to the stochastic nature of subthreshold degradation, differences might be observed among devices tested in the same conditions which reflect in degradation rate evaluation.These aspects should be further investigated to assess the variability among the results; a higher number of DUTs will greatly improve the statistics and reduce the effect of part-to-part variation.
Finally, the flux used during the test is significantly larger than the flux experience during application.This can lead to overestimation of degradation [as shown in Fig. 4(a)] and rejection of parts because of an overly conservative procedure.Therefore, low flux is preferred for components subject to microdose effects.

IV. CONCLUSION
Prototype SiC power MOSFETs with a trench architecture were tested with heavy ions using broad-beam and microbeam irradiations.In all of the DUTs, microdose effects resulting in higher subthreshold drain leakage were observed and associated with the creation of localized parasitic transistors with lowered threshold voltage, as reported in the literature for Si technologies.Such an effect strongly depends on the angle of incidence of the primary ion and can have an impact on the testing procedures.
SELC characterized by the formation of microbreaks in the gate oxide was observed at higher voltages and led to a complete gate rupture.At the voltages selected for the test, none of the devices showed SEB nor the typical signature of damage observed in SiC planar gate MOSFETs characterized by higher degradation of the drain current (i.e., SELC II).
Overall, it has been shown that under heavy-ion irradiation, similar to Si technology, gate oxide damage is the primary cause of failure in SiC trench devices.
The impact of these results on the testing procedures was discussed, highlighting the differences between planar and trench SiC MOSFET technologies, with special attention on the angle of incidence, the flux, and the accumulated fluence.

Manuscript received 5
January 2024; revised 5 March 2024; accepted 10 March 2024.Date of publication 19 March 2024; date of current version 16 August 2024.This work was supported by European Union's Horizon 2020 Research and Innovation Program under Grant 101008126.(Corresponding author: C. Martinella.)

Fig. 2 .
Fig. 2. (a)-(c) Three different types of damage observed when exposing the DUTs to heavy ions at different V DS (broad-beam experiment).The left panel shows the gate and drain leakage currents during exposure.The central and right panels illustrate, respectively, the transfer characteristics (V DS = 1 V) and the blocking characteristics (V GS = 0 V) measured after irradiation.The mechanisms were observed for all the samples and all the ion species but are shown here only for one sample following Kr irradiations (LET = 32.4MeVcm 2 /mg).The total fluence ( ) is indicated in the label.

Fig. 3 .
Fig. 3. Postirradiation analysis of SiC trench power MOSFETs degraded in the first region (I D increase).(a) Transfer characteristics measured at V DS = 50 mV.(b) Blocking characteristics measured at V GS = 0 V and −5 V for a pristine sample and a sample irradiated with Xe at V DS = 60 V and = 10 6 cm −2 .

Fig. 4 .
Fig. 4. Degradation rate for (a) drain and (b) gate leakage currents for the DUTs induced by broad-beam and microbeam irradiations.The results are normalized with the total fluence for each irradiation and the active area of the device.The irradiations with Xe were performed at two different fluxes (i.e., 1000, Xe-LF, and 10 000 ions/cm 2 s, Xe, dashed and solid lines, respectively).

TABLE I CHARACTERISTICS
OF THE IONS AT UCL AND RADEF TABLE II CHARACTERISTICS OF THE MICROBEAM AT GSI