Semiconductor Defect Signal Misidentified for 15 Years... South Korean Researchers Set the Record Straight [Reading Science]

KERI identifies SiC power semiconductor defect emission wavelength as 459 nm, not 424 nm
Analyzed 207 wafers and more than 100,000 defects... Provides basis for revising international standard

South Korean researchers have found an error in an international standard used to identify defects in silicon carbide (SiC) power semiconductors used in electric vehicles and power-conversion devices. The wavelength of light long considered a characteristic of a particular defect for about 15 years does not match its actual atomic structure. By directly comparing the light emitted by the defect with its atomic arrangement, the researchers confirmed that the correct wavelength is 459 nm, not the previously accepted 424 nm (nanometers, or one-billionth of a meter).

The Korea Electrotechnology Research Institute (KERI) announced on October 23 that a team led by Principal Researcher Na Moonkyung of its Next-Generation Semiconductor Research Center had established a new basis for distinguishing stacking faults in SiC power semiconductors through joint research with Chungnam National University, HORIBA STEC Korea, SK siltron, and others.

The Korea Electrotechnology Research Institute (KERI) research team analyzed the characteristics of stacking faults through observations of the atomic structure of silicon carbide (SiC) power semiconductors, photoluminescence (PL) measurements, and theoretical calculations. They confirmed that the actual emission wavelength of the (4,1) Frank-type stacking fault, long believed to be 424 nm for about 15 years, is 459 nm. Provided by the research team.

SiC power semiconductors can withstand high voltages and temperatures, making them useful in electric vehicles and renewable energy equipment. However, stacking faults, in which atomic layers inside a semiconductor crystal become misaligned, can affect device performance and reliability.

To detect defects without destroying wafers, the industry uses photoluminescence (PL) testing, which analyzes the wavelengths emitted after light is shone on a wafer. The method takes advantage of the fact that different defects emit light at different wavelengths. However, different defects can emit light at similar wavelengths, making it difficult to determine the atomic structure based on emission signals alone.

The research team analyzed data on 104,439 stacking faults identified in 207 commercially available SiC wafers from nine suppliers. In particular, it focused on the "(4,1) defect," a specific atomic arrangement among Frank-type stacking faults. Since around 2010, this defect had been known to emit light at 424 nm, and that wavelength was incorporated into the International Electrotechnical Commission (IEC) standard "IEC 63068-3:2020."

Emission analyzed for nine defects and atomic structures confirmed in six... Evidence secured for reviewing international standard

After locating defects through photoluminescence testing, the researchers cut out the corresponding areas and observed them with an atomic-resolution electron microscope. The study analyzed 176 ultrathin specimens. For the wavelength determination, the team examined nine defects emitting light at 459 nm and used findings confirming the atomic structure of actual "(4,1) defects" in six of them.

Results of a comparative analysis by a research team at the Korea Electrotechnology Research Institute of the atomic structures and electronic and optical properties of various stacking faults in silicon carbide (SiC) power semiconductors. The image shows how atomic arrangements and emission wavelengths vary depending on the type of defect. Provided by the research team.

Three theoretical calculations using a supercomputer also supported emission near 450 nm rather than the previously accepted 424 nm. The calculated wavelength was approximately 444–450 nm, while the actual emission wavelength measured at room temperature was 459 nm. Based on these findings, the team identified the error in the previous wavelength assignment and provided scientific grounds for reviewing the international standard.

The findings are expected to help reduce the risk of misclassifying defects during SiC wafer inspection and enable more accurate tracking of defect causes during manufacturing. However, further verification is needed to determine whether inspection criteria will change at manufacturing sites and whether the findings will improve device reliability.

The results were published in Volume 321 of the 2026 edition of the international journal Acta Materialia. Principal Researcher Na also received an Honorable Mention at the 2026 HORIBA Masao Award, organized by Japanese measurement-instrument company HORIBA, in recognition of her research on SiC defect analysis.

"It is significant that we corrected an error in a global identification standard that had been accepted for a long time by directly linking light signals to actual atomic structures," Principal Researcher Na said. "Going forward, we will not stop at identifying defects; we will also determine why they occur and how they can be controlled, helping strengthen the quality competitiveness of South Korea's power semiconductor industry."

This content was produced with the assistance of AI translation services.