Effect of 20 MeV Electron Radiation on Long Term Reliability of SiC Power MOSFETs

The effect of 20 MeV electron radiation on the lifetime of the silicon carbide (SiC) power MOSFETs was investigated. Accelerated constant voltage stress (CVS) was applied on the pristine and irradiated devices, and time-to-breakdown (<inline-formula> <tex-math notation="LaTeX">${T_{\mathrm {BD}}}$ </tex-math></inline-formula>) and charge-to-breakdown (<inline-formula> <tex-math notation="LaTeX">${Q_{\mathrm {BD}}}$ </tex-math></inline-formula>) of gate oxide were extracted and compared. The effect of electron radiation on the device lifetime reduction can be observed at lower stress gate-to-source voltage (<inline-formula> <tex-math notation="LaTeX">${V_{\mathrm {GS}}}$ </tex-math></inline-formula>) levels. The models of <inline-formula> <tex-math notation="LaTeX">${T_{\mathrm {BD}}}$ </tex-math></inline-formula> and <inline-formula> <tex-math notation="LaTeX">${Q_{\mathrm {BD}}}$ </tex-math></inline-formula> dependence on the initial gate current (<inline-formula> <tex-math notation="LaTeX">${I_{G0}}$ </tex-math></inline-formula>) are proposed, which can be used to describe the device breakdown behavior.

environments causes a reduction in lifetime of these devices. The effect of electron and gamma irradiation on electrical stress response for silicon-based technologies has been found to be relatively weak below 10 kGy total ionizing dose (TID) levels showing mainly increased leakage current through the gate dielectric [9], [10], [11].
However, the effect of electron radiation on the SiC power device long term reliability is still unknown. The main degradation mechanism caused by the electrons on SiC MOSFETs is commonly considered to be the TID and a resulting degradation in transfer characteristics [12], [13], [14]. Fortunately, the effects of TID on the transfer characteristics of SiC power MOSFETs have been found to be moderate, except at high dose levels [15], [16]. However, energetic electrons can interact with the material also through the non-ionizing energy loss (NIEL) resulting in knock-on damage, which refers to displacement of atoms in the target material [17]. Especially, during the space missions reaching beyond the Earth orbit, such as in the Jovian environment, the electron energy can reach hundreds of MeV and, thus, contribute on the NIEL in the electronic component material.
Regarding the overall reliability of SiC MOSFETs, the gate oxide degradation remains an issue [18], [19]. Even though the intrinsic reliability of the gate SiO 2 has improved over the years, the material defects and the radiation impact have a significant effect on the oxide reliability [4], [5], [6], [20], [21]. The lifetime of the devices at their normal operating voltage conditions is commonly predicted through accelerated time-dependent dielectric breakdown (TDDB) measurements and extrapolation [21], [22].
In this work, we investigate the effect of 20 MeV electron radiation on long term reliability of SiC power MOSFETs through accelerated wear-out experiments by performing TDDB measurements on the pristine and electron-irradiated devices. Moreover, the radiation induced-leakage current (RILC) has been observed and linked to the radiation-induced lifetime degradation. Finally, the models are proposed to describe the stress current dependence on the device lifetime under stress.

A. Device Under Test
The device under test (DUT) is a commercial SiC power MOSFET manufactured by Wolfspeed (part number C3M0280090D). A total of 118 samples were tested, of which 53 were irradiated. All the devices were electrically characterized before and after irradiation.

B. Radiation Exposure
Electron exposures were performed at RADiation Effects Facility (RADEF) at the University of Jyväskylä, Jyväskylä, Finland, with a Varian Clinac (Clinical LINear ACcelerator) 2100 CD. The electron beam is pulsed and consists of 5-µs long pulses every 5 ms (pulse frequency of 200 Hz). An electron energy of 20 MeV was used, which is the maximum energy of the Clinac. Devices were irradiated in air with an average flux of electrons corresponding to a dose rate of 10 Gy(H 2 O)/min. The devices were irradiated for 100 min, reaching an electron fluence corresponding to the total dose of 1 kGy(H 2 O), which corresponds to 0.87 kGy in Si. This was calculated based on stopping powers of electrons in those materials obtained from [23] and also described in [24]. This TID value is approximately representative of ten years in low earth orbit (LEO) [25]. During irradiations, V GS = −4 V and V DS = 900 V were applied on the devices.

C. Accelerated Wear-Out Experiment
In order to investigate if the electron radiation has an impact on the long term reliability of SiC power MOSFET, an accelerated wear-out procedure was applied on the devices. For each test configuration, two sets of devices (irradiated and pristine) were exposed to a constant voltage stress (CVS) at the gate terminal, while the drain and the source terminals were grounded.
The gate-to-source voltage (V GS ) values for CVS were chosen based on the Fowler-Nordheim (FN) curves measured for three randomly picked devices. Four V GS values (V GS = 33, 35.5, 37, and 38.3 V) were defined at the points where the gate-to-source currents (I GS ) were ≈1, 10, 50, and 100 µA, respectively. Such V GS values are well below the instantaneous breakdown voltage of the gate oxide, but, at the same time, above the normal operating voltage in order to accelerate the wear-out. The CVS was applied, and I GS was monitored until an abrupt increase in I GS was observed. The time at which that increase occurred was then defined as the time-to-breakdown (T BD ). By integrating the gate current over time until T BD , we extract the amount of charge (charge-to-breakdown, Q BD ) injected through the gate oxide during the CVS. The procedure of the global test approach is illustrated in Fig. 1.

A. Radiation Induced Degradation and Oxide Breakdown
It is known that the gate leakage current can be utilized as an indicator of the gate oxide degradation in SiC MOSFETs [18]. In order to assess the effect of radiation on the gate degradation, we first analyze the leakage current through the gate oxide by sweeping V GS up to the FN tunneling region before and after the irradiation (Fig. 2). After electron exposure, the gate current exhibits increased leakage behavior, which can be related to the creation of defects in the gate oxide volume by the electron impact [26]. Similar RILC was originally observed for thin oxides by Scarpa et al. [11]. We express the total gate current after irradiation (I GS(post) ) as the sum of the preirradiation current (I GS(pre) ) and RILC (1) The RILC component was calculated from (1) and plotted in Fig. 2 for one example case. At V GS values below the FNregion (V GS ≲ 22 V), the RILC component stays constant. However, at the FN-region, the contribution of RILC in the total I G increases with applied V GS . Such increase is expected to play a role in the oxide breakdown process, which will be discussed in Section III-B.
As mentioned in Section II-C, four V GS values were used in the breakdown measurements. That allows us to compare the effects of not only the electron radiation but also the stress voltage on the breakdown characteristics. While assuming 50-nm oxide thickness, a resulting E ox of 6.6, 7.1, 7.4, and 7.7 MV/cm, respectively, was applied across the oxide layer. By choosing such values for E ox , we are able to collect T BD data in an accelerated manner while staying below the critical E ox (>10 MV/cm) [27] in order to avoid immediate device failure during CVS. Also, the electric field is below the value where the electric field acceleration factor for higher fields plays a role, when performing the lifetime projection based on the TDDB results [21], [28].
The gate current behavior during the CVS for one device from each category is illustrated in Fig. 3. The breakdown characteristics during CVS are strongly dependent on the applied stress voltage. The initial gate current I G 0 plays a role in how the device fails during the CVS. When applying a stress voltage of V GS = 35.5 V, we can see three different phases in the gate current evolution in Fig. 3, also reported in [20]. In the first phase, we observe increasing gate current over time indicating hole trapping process in the gate dielectric layer. In the second phase, the gate current starts to decrease, likely due to increasing electron trapping in the dielectric. In the third phase, the gate current suddenly increases indicating the oxide breakdown. However, when applying higher voltage on the gate terminal during CVS, the gate current keeps increasing until the device breakdown, and it never reaches the decreasing phase. At higher V GS , the kinetic energy of the injected charge carriers is high enough to initiate impact ionization, which results in a lower injected charge needed for breakdown. On the other hand, for V GS = 33 V, the increasing phase in the gate current does not exist, and the gate current keeps on decreasing until the breakdown occurs. Moreover, T BD is dependent on the applied V GS and, therefore, on the gate current as well. It will be discussed in more detail in Sections III-B and III-C.

B. Effect of Electron Radiation on the Wear-Out Life
In order to extract the lifetime of the devices based on the TDDB results, we applied Weibull statistics on the T BD data. We used the two-parameter Weibull cumulative distribution function (cdf) where β is the shape parameter, and η is the scale parameter of the Weibull distribution. The scale parameter η is also known as the characteristic lifetime, which refers to the time when 63% of the population has failed. In order to compare the empirical data with the distribution function, a common way to obtain the y-coordinate for each failure is to apply the Benard approximation (3) where i is the running number of the failure (first, second, and so on), and N is the sample size. Then, rewriting (2) gives For each device failure, ln(− ln(1 − F)) was then plotted as a function of elapsed time until corresponding failure (Fig. 4). Equation (2) was then fit to the T BD data, and the β parameter, also known as shape parameter, and the η parameter, representing the characteristic lifetime, were extracted. The parameters were extracted by using a software library [29] and by using maximum likelihood estimation (MLE) for the fitting. The extracted Weibull parameters for different test configurations are presented in Table I. The characteristic lifetimes of the different sets are presented in Fig. 5. The lifetime of the devices is strongly dependent on the applied V GS during TDDB test. The difference in characteristic lifetimes between V GS = 33 V and V GS = 38.3 V configuration is four orders of magnitude. Similar, strong electric field dependence has been observed for both Si and SiC based technologies [21], [22], [30]. The increasing injected electron energy with increasing electric field in the oxide during the stress will result in faster accumulation of trapped charge, which will eventually lead to oxide breakdown [30].
Regarding the irradiation effect, when the applied V GS during CVS is set to 35.5 V, irradiated devices exhibit ≈70% lower characteristic lifetime compared with non-irradiated devices. The same trend can be observed also for V GS = 33 V. It indicates clearly that electron irradiation has weakened the gate oxide. Also, it suggests that the radiation-induced defect creation is revealed only at lower stress levels (V GS ≤ 35.5 V). At higher V GS values (≥37 V), higher initial stress current and stress-induced defect creation might contribute more on the number of defects in the oxide volume and, therefore, causing an additional weakening due to the stress test itself.  In order to estimate, how the electron-induced degradation affects on the wear-out life of the components, we have extrapolated the lifetime of the components under the rated voltage condition from the accelerated test results (Fig. 6). The dashed vertical line in Fig. 6 represents the maximum rated V GS = 19 V given by the manufacturer. Based on the extrapolation, the device lifetimes at safe operating V GS for irradiated and non-irradiated devices were found to be 1 × 10 9 and 5 × 10 11 years, respectively. However, regarding the effect of the electron radiation on the long term reliability and the lifetime of the device, it should be kept in mind that V GS during CVS should be low enough in order to be able to conclude on possible lifetime reduction. As discussed in [21] and [22], the lifetime extrapolation should be performed with gate voltages for which the electric field in the oxide is below the impact ionization current threshold. Such threshold value for the E ox is suggested to range from 8.5 MV/cm [21], [22], [28]. Above that threshold value, the field acceleration factor for lifetime extrapolation is approximately double compared with the value when performing the test below the threshold. Therefore, performing the TDDB tests only at high V GS , the lifetime predictions can be strongly overestimated. Moreover, it should be noted that the experiments have been performed at room  I  TEST CONFIGURATIONS AND EXTRACTED WEIBULL PARAMETERS temperature. Therefore, the lifetime predictions presented here can be optimistic, and the lifetime during operation conditions should be significantly reduced due to the temperature stress, as observed in [18], [31], and [32].
Here, the electron interaction with the gate oxide material will contribute to the number of defects in the oxide volume. One possible molecular precursor of oxide breakdown is the oxygen vacancy, which is suggested to play a role in a low-field TDDB process [33]. Those defects then act as a precursor for a current path through the oxide, which results in an increase of the total leakage current through the gate oxide, as shown in Fig. 2. Such increase in leakage current results in shorter T BD for irradiated devices. However, such behavior is only observed at the lowest used V GS values during CVS, when the defect creation and trapping dominate the breakdown. At higher V GS , the impact ionization current dominates the breakdown current, and since it is a fast process, the contribution of irradiation-induced trapping and defect creation in the breakdown characteristics is relatively smaller compared with low V GS settings. When applying higher V GS during CVS, the stress current will cause additional stress to the dielectric material, and the radiation-induced defects are masked out.

C. T BD and Q BD Dependence on Initial Gate Current
After extracting T BD for each device from the TDDB data, we pay attention to the initial gate current (I G 0 ), which is defined as the gate current value at the beginning of the CVS. Fig. 7 shows that the T BD is dependent on I G 0 during CVS and follows a power law: where k and s are the model parameters. A higher I G0 during the CVS results in a shorter breakdown time due to a faster accumulation of the injected charge. Such behavior seems to be present for both irradiated and nonirradiated devices. On the other hand, Q BD has an exponential dependence on I G0 where a and b are the model parameters. It seems that the required injected charge for oxide breakdown saturates below certain I G0 level. It indicates that below certain stress voltage level, T BD could be estimated based on I G0 during CVS. Such saturation is also observed in Fig. 8, where Q BD is plotted as a function of T BD . At lower V GS , Q BD seems to reach a saturation level around 3 × 10 −1 C, also observed in [34].  It seems that below a certain V GS , the breakdown is dominated by the injected charge through the oxide layer.
Also, a simple Q 0 model [34] is shown in Fig. 8, where Q BD is obtained by multiplying I G0 by T BD . However, especially with longer breakdown times (lower V GS ), such a model does not well describe the Q 0 behavior. The shape of the stress current curve in Fig. 3 shows decreasing trend over time due to charge trapping, and therefore, a simple multiplication does not give a correct result for Q BD .
The similar behavior between fresh and irradiated devices in Fig. 7 suggests that I G0 during CVS may be used as an indicator of the device lifetime, whether the devices have been irradiated or not. Therefore, an electron radiation-induced lifetime reduction of SiC MOSFETs could be estimated based on the RILC at the FN region, thus avoiding the time-consuming TDDB tests.

IV. CONCLUSION
A long term reliability degradation of the SiC power MOSFET due to electron irradiation was observed. Even though SiC power technology has been proven to be robust against TID, electron radiation can create defects in the oxide volume and at the oxide-semiconductor interface and, therefore, result in reliability degradation revealed by the accelerated lifetime test. Even if gate failures were not observed during the irradiation experiment neither during post-irradiation characterizations, stronger stressing through CVS reveals lower T BD of gate oxide for irradiated devices. However, such reliability degradation was observed only at the lower V GS values, when the charge trapping dominates the breakdown characteristics. It is assumed that at higher V GS values, the radiation-induced degradation is masked by the stress-induced defect creation during CVS. Nonetheless, it indicates a degradation of gate dielectric layer due to electron radiation impact and, therefore, a radiation-induced reduction in the long-term reliability of the device.
On top of TID effects, the high energy electrons used in this study are able to induce structural damage in the device material. Therefore, in order to extend these results to the other ionizing radiation environments, more investigations should be performed by modifying the energy and fluence of the electrons. However, regarding the TID effect only, the results presented here should represent the worst case for the used accumulated dose.