Access Official Technical PDF

Share Summary:LinkedInEmail

Diamond Schottky diodes are being investigated for power electronics where high blocking voltage and low conduction losses are important design goals. But semiconductor material properties alone do not determine device performance: the condition of the epitaxial surface and the quality of the underlying crystal also matter. A comparison of vertical diamond diodes made on smooth, hillock-rich and polished surfaces shows how these factors influence reverse leakage, breakdown behavior and forward current transport.

The devices in the study used homoepitaxial diamond layers and achieved blocking voltages of approximately 2.4–2.6 kV, with specific on-resistances of 300–440 mΩ·cm². The results offer a practical reminder for device developers and integrators: surface preparation must be considered alongside layer properties and device design when evaluating diamond for high-voltage applications.

Why surface morphology matters in diamond Schottky diodes

A Schottky diode forms a barrier at the interface between a metal contact and a semiconductor. Variations at that interface can affect how carriers cross the barrier. In a power diode, surface irregularities may also contribute to local electric-field concentration, while crystal defects can provide paths associated with increased reverse current or premature breakdown.

The comparison covered three surface conditions:

  • Smooth as-grown: a low-roughness surface without prominent growth hillocks.
  • Hillock-rich: a surface with growth features associated with defects in the substrate and local roughness.
  • Polished: a hillock-rich grown layer that was mechanically polished and then treated with hydrogen plasma to reduce polishing-related damage.

These conditions are not interchangeable process options in isolation. The observations indicate that both surface topography and the defect population of the substrate can influence the resulting diode, and their effects may appear in different electrical measurements.

Reverse leakage and breakdown reveal different risks

The smooth as-grown and polished diodes both maintained reverse current densities below 10-4 A/cm² at reverse voltages up to 2.2 kV. The hillock-rich diode showed somewhat higher leakage, below 10-3 A/cm² over the reported range. All three representative devices reached breakdown between 2.4 and 2.6 kV.

However, a similar breakdown voltage does not mean the electric field is distributed in the same way. Using a simplified one-dimensional calculation, the smooth as-grown diode had an estimated maximum field of about 1.6 MV/cm, compared with about 1.2 MV/cm for the hillock-rich and polished devices. The authors associate the lower calculated field in the hillock-rich case with field enhancement caused by surface inhomogeneity. This model is an estimate, not a direct measurement of the local field; the study also notes that the devices lacked edge termination, which can affect field distribution.

Polishing produced a surface with low measured roughness, and the representative polished diode showed reverse leakage and breakdown voltage similar to the smooth as-grown device. Yet a statistical comparison across multiple diodes identified a separate concern: the polished sample had a higher probability of increased reverse current and a smaller practical area for fabricating high-blocking-voltage devices. The authors attribute this behavior to the substrate’s higher defect density. In other words, improving topography does not necessarily remove defects within the crystal.

Forward transport depends on the contact barrier

Forward current measurements added another distinction between the samples. The smooth as-grown and polished diodes were well described by a thermionic-emission model, with ideality factors of 1.03 and 1.08, respectively—values close to ideal diode behavior in this analysis. Their fitted Schottky barrier heights differed, and the polished diode had a lower reported specific on-resistance: 300 mΩ·cm² versus 440 mΩ·cm² for the smooth as-grown diode.

The lower resistance of the polished device was not fully explained by the similar doping levels. The authors identify further investigation, including temperature-dependent measurements and Hall measurements, as necessary to understand the difference. This is an important qualification for engineers comparing process variants: a measured device advantage may be clear even when its physical cause has not yet been established.

The hillock-rich diode showed two distinct regions in its forward current-voltage curve. The analysis modeled this behavior as contributions from two parallel current paths with different effective barriers. A modified thermionic-emission expression using the Lambert W-function enabled the two-barrier model to be fitted explicitly. This provides a way to describe non-ideal, multi-region transport; it does not by itself establish the precise physical shape or location of the lower-barrier region.

What the results mean for power-device development

All three diodes combined kilovolt-class blocking with specific on-resistances of 300–440 mΩ·cm². Their reported Baliga Figure of Merit (BFOM), calculated from breakdown voltage and specific on-resistance, ranged from 15 to 21 MW/cm². The polished diode had the highest reported value, followed by the hillock-rich and smooth as-grown devices. BFOM provides a useful comparison of voltage and resistance trade-offs, but it does not capture every factor relevant to a finished power component or its application.

For engineering teams assessing diamond devices, the findings point to several considerations:

  • Evaluate surface roughness and hillock formation, but do not treat a smooth surface as proof of low defect density.
  • Use multiple devices and areas in electrical characterization; individual breakdown results may not represent substrate quality across a wafer or sample.
  • Compare forward and reverse behavior together. Low on-resistance does not, by itself, characterize leakage or breakdown-field limits.
  • Interpret model-derived quantities, such as one-dimensional maximum field and fitted barrier parameters, in light of their assumptions.

Conclusion

Surface morphology and crystal quality both influence diamond Schottky diode performance, but in different ways. Hillocks can be associated with higher leakage and field enhancement, while polishing can yield a smooth surface without eliminating the effects of substrate defects. The practical implication is to treat substrate screening, surface preparation and statistical device testing as connected parts of power-device development. For industrial applications, the reported results demonstrate promising high-voltage performance, while also underscoring the need to control defects and validate device behavior across more than a single diode.

Access Official Technical PDF

About The Author