Access Official Technical PDF

Share Summary:LinkedInEmail

For engineers evaluating gate materials for silicon carbide devices, the AlN/4H-SiC interface presents a notable trade-off: room-temperature measurements show substantially fewer interface traps than nitrided SiO2/SiC, while electron trapping within the AlN and limited dielectric breakdown remain significant concerns. These results, obtained from metal-insulator-semiconductor (MIS) capacitors with AlN grown by hot-wall MOCVD, help clarify both the potential and the current constraints of AlN-based SiC gate structures.

Why consider AlN as a 4H-SiC gate dielectric?

Thermal oxidation provides SiC with a native silicon dioxide layer, but the SiO2/SiC interface can contain traps that impair electron transport in a MOSFET channel. Interface-related trapping and scattering are particularly relevant when channel mobility is a design priority. Oxidation and nitridation processes have improved SiO2/SiC technology, supporting high-voltage devices, but further reductions in interface-state density remain desirable.

Aluminum nitride is one of several alternative dielectrics being studied for SiC MIS technology. Its wide bandgap and close lattice match to SiC make crystalline growth on SiC possible. The study examined crystalline AlN layers grown by hot-wall metal-organic chemical vapor deposition (MOCVD), measuring capacitance-voltage (CV), conductance-voltage (GV), and current-voltage (IV) behavior in MIS capacitors.

What the AlN/4H-SiC interface measurements show

At room temperature, CV measurements indicated an interface-trap density at the AlN/SiC interface about an order of magnitude lower than at nitrided SiO2/SiC interfaces. The 10 nm AlN sample showed the lowest measured trap density among the AlN structures tested. A 30 nm AlN layer had a higher density than the 10 nm layer, although it remained below that of the thermal-oxide reference samples.

Thickness and dielectric-stack construction mattered. The authors suggest that differences in AlN morphology and the increase of misfit dislocations with layer thickness may help explain the poorer result for the thicker AlN. Adding a dielectric above AlN also generally increased measured interface-state density compared with a single AlN layer. Among the stacks tested, SiO2 on AlN had the best interface-state result; a Si3N4/AlN stack had the highest.

These measurements require careful interpretation. Low room-temperature trap density does not mean that all electrically active states are absent. At 77 K, CV analysis revealed a high density of shallow traps close to the SiC conduction-band edge. The GV response also differed between room temperature and 77 K, indicating that some traps respond quickly at room temperature and are more apparent under low-temperature measurement conditions. Their effect on channel mobility in an operating MISFET was not established.

Electron trapping creates a reliability concern

A separate issue emerged when the capacitors were biased into accumulation, a condition that drives electrons toward the dielectric. Repeated accumulation bias caused a positive shift in the CV curve, corresponding to a shift in flatband voltage. The shift was attributed mainly to electrons captured in bulk traps within the AlN rather than to interface or near-interface traps.

For the tested SiO2/AlN stack, the reported flatband shift was about 8 V after stress at 298 K and about 5 V at 77 K under the stated measurement conditions. The magnitude depended on the maximum accumulation voltage. The shift saturated after repeated sweeps, and the curves were consistent with charge stored within the dielectric. This behavior is relevant to gate-stack evaluation because trapped charge can alter the voltage required to reach a given operating condition.

The trapped electrons were not permanently fixed. Applying depletion bias at elevated temperatures released them back to the SiC. Measurements showed that emission began around room temperature but was limited there; higher temperatures enabled further release. The results indicate a range of trap depths or emission activation energies, although the study did not establish their physical origin. Defects such as point defects or dislocations were discussed as possible explanations, not confirmed causes.

Breakdown behavior and the role of SiO2

IV measurements showed that the single-layer AlN capacitors had substantial leakage before a clear abrupt breakdown event. The estimated breakdown field was approximately 3 MV/cm, while the study’s overall conclusion placed the AlN breakdown field in the range of 3–4 MV/cm. Leakage was already considered unacceptable above about 1 MV/cm in the tested single-layer structure, with trap-assisted conduction identified as a likely contributor.

Adding a 40 nm SiO2 layer above 10 nm AlN substantially increased the gate voltage the capacitor could tolerate. When the electric field was calculated specifically across the AlN layer, the estimated breakdown field was about 4 MV/cm. The added SiO2 therefore improved the voltage-handling capability of the stack without a major loss in interface quality for the best-performing AlN-based stack. The measurements also showed different leakage behavior between single-layer AlN and the SiO2/AlN stack.

Practical implications for SiC gate-stack development

For device engineers and integration teams, the findings point to several considerations when assessing an AlN/4H-SiC gate dielectric:

  • Evaluate more than room-temperature CV. Low room-temperature interface-trap density does not reveal the full population of shallow traps detected at 77 K.
  • Include bias-stress testing. Accumulation can inject electrons into bulk AlN traps, shifting the flatband voltage; depletion bias and elevated temperature can reverse this effect.
  • Assess the complete dielectric stack. A top SiO2 layer improved voltage tolerance in the tested structure, but dielectric additions can also affect measured interface-state density.
  • Consider AlN thickness and morphology. The 10 nm layer performed better than the 30 nm layer in the reported interface measurements, while the authors identified structural defects as a possible factor.

AlN grown by MOCVD shows promise as a route to a lower-trap 4H-SiC interface, but the capacitor results do not by themselves establish suitability for a production MOSFET. Electron trapping, leakage, and breakdown must be addressed alongside interface quality, and the electrical behavior should be assessed under relevant bias and temperature conditions. The study suggests that a thin AlN layer combined with a thicker deposited SiO2 layer warrants further investigation as a way to balance interface properties and voltage handling.

Access Official Technical PDF

About The Author