Quantum electrometry in a silicon carbide power device

Abstract

For high-bias operation devices such as silicon carbide (SiC) power devices, early detection of failure mechanisms is essential to ensure reliability. This requires a method to map high electric fields with high spatial resolution, which has not been realized until now. Here we report that the silicon vacancy (Vsi) in SiC has outstanding characteristics for detecting electric fields applied in various directions within a high-biased SiC device. Vsi exhibits an equivalent response to electric field components parallel (Epara) and perpendicular (Eperp) to the c-axis, a feature unique among quantum sensors, and the responsiveness to Epara and Eperp enables detection of arbitrary electric fields encountered in cutting-edge SiC power devices. We confirmed high electric field detection of ~2.3 MV/cm, which is ~90% of the breakdown electric field of a 4H-SiC with typical carrier concentration. Selectively formed Vsi enables high-resolution mapping of electric field distribution. Vsi-based quantum sensors bring data-driven research and development methodologies as well as device degradation diagnosis.

Figures (18)

• Figure 1: Conceptual diagram of quantum electrometry in a SiC power device. a, Structure of VSiVC in 4H-SiC. This defect responds well to an electric field parallel to its defect axis. b, Structure of VSi in 4H-SiC. This defect, being a monovacancy with no defect axis, responds equally to electric fields applied both parallel and perpendicular to the quantization axis of VSi (= c-axis). c, An electric field inside a SiC device is measured with high spatial resolution using a selectively formed VSi-based quantum sensors embedded within the device. Selective formation of VSi avoids device degradation and enables electric field measurement while maintaining the device in a normal state. A red laser focused on one of ensemble VSi and a metal wire for applying RF for quantum sensing are depicted.
• Figure 2: Sample structure and electric properties. a, Cross-sectional image of a pn diode with an edge termination region. The device is fabricated on a n-type substrate with a 4$^\circ$-off orientation. b-e, Distribution of electric field components parallel (E||) and perpendicular (E$\perp$) to the c-axis obtained by correcting device simulation results. Applied voltage is set to 1500 V. Point1 and point3 are used to determine the values of d|| and d$\perp$, respectively, and at point2, high electric field measurement is performed to investigate the stability of d|| and d$\bot$ under high electric field conditions. f, Top view of a sample mounted on a PCB. The upper electrode (p-type epilayer) is connected to GND, while a positive voltage is applied to the lower electrode (n-type substrate). To avoid a short circuit between the RF and DC lines, an enameled wire is used for the RF application. The white arrow indicates the off-direction ([11$\bar{2}$0]). g, Photoluminescence mapping image of the VSi dot array. h, I-V characteristics under reverse bias before and after VSi formation. No significant changes were observed after PBW.
• Figure 3: Determination of d|| and its stability under high electric fields. a, ODMR spectra at point1 (diamond mark in fig. 2d) measured under reverse bias ranging from 0 to 1000 V. The black lines are the fitting curves. b, Electric field dependence of the resonance frequency. The values of electric field are based on device simulations. From the gradient, d||/h was determined to be -15.0 ± 0.5 MHz/(MV/cm). c, ODMR spectra at point2 (circle mark in fig. 2b) measured under applied voltages of 1000, 1250 and 1500 V. Electric fields based on device simulation are $\sim$1.8, $\sim$2.1 and $\sim$2.3 MV/cm, respectively. Inset shows the theoretical values (dashed line) considering related parameters (d||, E$\perp$/E|| at the measured point and |d$\perp$/d|||) and the experimental values. It is noted that even at an E$\perp$/E|| ratio of 0.07, nonlinearity in the resonance frequency shift appears, especially at high electric fields exceeding 2 MV/cm.
• Figure 4: Determination of d$\bot$. Electric field dependence of the resonance frequency at point3 (triangle mark in fig. 2e) and the theoretical resonance frequencies calculated by varying d$\bot$/d|| from 0 to 2 under a zero magnetic field. E$\bot$/E|| was calculated to be $\sim$0.19 at the point. Comparing the experimental data and calculated values, |d$\bot$/d||| is estimated to be 1.1 ± 0.06.
• Figure 5: Electrical field distribution measurements. a, Electric field distribution measured under the applied voltage of 750 V. Electric fields were calculated from d|| and the resonance frequencies. Since the effect of E$\bot$ on the resonance frequency is small under the condition, no correction accounting for E$\bot$ was applied at each measurement point. Solid and dashed lines show the simulated electric fields with and without the enameled wire with DC potential of zero, respectively. The discrepancies from the dashed line are significant in the regions z = 3.7 and 5 $\mu$m, and x > 70 $\mu$m. b, c, Electric field distribution obtained from device simulations without and with the enameled wire, respectively. The presence of the enameled wire causes electric field to extend to x > 100 $\mu$m.