Korean Team Finds 459 nm Signal for SiC Defect, Challenging 15 Years of Inspection Guidance

104,439 defects help expose a mistaken identification signal

A South Korean research team has identified a different light signal for a defect in silicon carbide power semiconductors after examining 104,439 stacking faults across 207 commercial wafers from nine suppliers. The finding provides scientific grounds to revise an identification value included in IEC 63068-3:2020, an international standard used in semiconductor inspection.

The defect, known as the standard (4,1) Frank-type stacking fault, emits light at a characteristic wavelength of 459 nanometers at room temperature. For about 15 years, it had been associated with 424 nanometers, a value carried into guidance from the International Electrotechnical Commission.

Na Moonkyung, principal researcher at the Korea Electrotechnology Research Institute's Next-Generation Semiconductor Research Center, led the work with Professor Hong Soonkoo's team at Chungnam National University, HORIBA STEC Korea, SK siltron and other collaborators. Their investigation connected the light from individual defects with images of their actual atomic arrangements.

The team's Acta Materialia study combines optical measurements, microscopy capable of showing atomic structures and calculations of how electrons behave inside the material. Together, those methods identify 459 nanometers as the characteristic signal of the standard (4,1) fault.

The practical goal is more accurate inspection of materials used in electric vehicles, chargers and renewable energy equipment. Manufacturers need to know which defects they are seeing before they can trace their causes and adjust production. The research supplies a basis for reviewing inspection guidance; formal changes to the IEC standard, adoption in factories and resulting improvements in device reliability still require further work.

Why a tiny change in atomic stacking matters

Silicon carbide, usually shortened to SiC, is useful in equipment that must control electricity while handling high voltages and temperatures. Power semiconductors help switch and convert electrical power, including in vehicle drive systems and renewable energy installations.

A wafer is a thin slice of semiconductor material on which devices are made. The research concerns 4H-SiC, a form of silicon carbide whose atoms follow a particular repeating stacking pattern. An epitaxial wafer has an additional crystalline layer grown on its surface for making devices.

Stacking faults occur when that regular arrangement is interrupted. Some involve missing or inserted layers; others form when parts of the crystal slide relative to one another. A simple comparison is a carefully ordered stack of colored sheets: adding, removing or shifting a sheet changes the pattern around it.

The atomic arrangement affects how electrons move and release energy. A defect can therefore change both the electrical behavior of a device and the light emitted during inspection. The (4,1) designation is shorthand for a particular stacking sequence, rather than a measurement of the defect's size.

How light became a misleading fingerprint

Photoluminescence, or PL, inspection shines light onto a wafer and measures the light coming back from the material. Defects can produce characteristic emission signals, allowing inspection equipment to locate and classify them without breaking the wafer.

A wavelength measures the distance between successive peaks of a light wave. One nanometer is one billionth of a meter. The revised signal, 459 nanometers, is 35 nanometers longer than the old value of 424 nanometers, a difference of about 8.3 percent relative to the previous assignment.

Different atomic defects can emit at nearby wavelengths. An inspection system therefore needs a reliable connection between each optical signal and the structure that produces it. If the connection is wrong, the equipment may find a real defect but give it the wrong name.

The earlier 424 nanometer assignment had become established without the same direct matching of optical signals and atomic structures used in the new investigation. The researchers addressed that problem by returning to the exact locations where the light had been measured.

From whole wafers to individual atomic layers

The team began with the broad survey of 207 commercial wafers. That collection offered a view across nine suppliers rather than a small area of a single research sample. Frank-type stacking faults emerged as an important defect group in the commercial material.

Researchers then selected individual defects for closer examination. After recording their positions and emission wavelengths with PL measurements, they used a focused ion beam to prepare extremely thin samples from the same locations. Across the investigation, they examined 176 specimens containing Frank-type, Shockley-type and other stacking fault structures.

An advanced electron microscopy technique, abbreviated HAADF HR-STEM, revealed the atomic arrangements. Its full name is high-angle annular dark-field high-resolution scanning transmission electron microscopy. In simple terms, it sends electrons through a very thin sample and uses the scattered electrons to build an image detailed enough to distinguish atomic columns.

The targeted investigation included nine independent defects emitting at 459 nanometers. The researchers confirmed the actual (4,1) atomic structure in six of them. That direct connection between a measured light signal and a visible stacking arrangement supplied the experimental foundation for the revised assignment.

Na, the KERI principal researcher who led the project, described the importance of connecting the two measurements:

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.

What the supercomputer calculations found

The researchers also used density functional theory, or DFT, to calculate properties of the defects. DFT is a method for estimating how electrons behave in a material, helping researchers connect an atomic arrangement with its energy levels and possible optical signals.

Three theoretical calculation approaches placed the relevant emission at approximately 444 to 450 nanometers. The measured wavelength at room temperature was 459 nanometers. The calculations therefore favored emission near 450 nanometers rather than the longstanding 424 nanometer assignment, while retaining a numerical gap from the laboratory measurement.

The paper brings these calculations together with microscopy and PL spectroscopy, which measures how strongly a material emits at different wavelengths. The experimental measurement supplies the 459 nanometer inspection value, while the calculations provide another way to examine the connection between structure and light.

The broader calculations also compare stacking fault energies and electronic band structures. Fault energy describes the energetic cost of a particular interruption in the crystal. Electronic band structure describes the energies electrons can occupy, which helps explain why different defects have different optical behavior.

A 2024 study had already challenged the old map

An earlier wafer inspection study, published on June 1, 2024, in Materials Science in Semiconductor Processing, examined approximately 50,000 stacking faults across 70 commercial wafers. It also paired optical measurements with direct examination of selected atomic structures.

That investigation found a single Shockley-type (3,1) stacking fault at 424 nanometers, the wavelength previously associated with the Frank-type (4,1) defect. Researchers searched defects emitting between 420 and 430 nanometers but did not observe the expected (4,1) structure there.

The 2024 work also reassigned the characteristic signal of intrinsic Frank-type (5,2) faults to 482 to 484 nanometers, rather than the previously reported 488 nanometers. Structures examined in that range were predominantly (5,2) faults instead of the triple Shockley-type (5,3) faults that earlier wavelength assignments suggested. No (5,3) faults were observed in that survey.

Other nearby signals became easier to separate: the study measured multilayer Frank-type (4,2) faults at 457 nanometers and quadruple Shockley-type (4,4) faults at 462 nanometers. Accurate calibration and direct structural checks helped distinguish defects whose light signals lie only a few nanometers apart.

The newer survey covers almost three times as many wafers, 207 compared with 70, and a little more than twice as many recorded faults, 104,439 compared with approximately 50,000. The advance is also specific: after the earlier search failed to find the expected (4,1) structure near 424 nanometers, the newer investigation connects it to 459 nanometers.

The paper also revises how faults are classified

The Acta Materialia paper goes beyond replacing one wavelength. It revisits how Frank-type stacking faults are classified as intrinsic or extrinsic, terms describing their relationship to the surrounding crystal structure.

Applying Frank's original crystallographic definition, the authors classify the widely discussed (3,2) fault as intrinsic even though its formation involves insertion. The classification depends on the resulting arrangement of the crystal, while the formation process describes how that arrangement developed.

Atomic observations also show structural connections among standard (5,2), (4,2) and (4,1) faults. These observations support one possible route by which the structures could evolve from one to another. Understanding such connections could help researchers investigate why a particular defect appears during material growth or processing.

The authors found that identical Zhdanov notations, the numerical labels used to describe stacking sequences, can refer to either intrinsic or extrinsic configurations among nonstandard faults. The surrounding crystallographic arrangement is therefore needed to interpret the label.

The calculations identify further differences in the (1,1) and (2,1) faults. Their electronic characteristics resemble the 2H form of silicon carbide, whereas most faults studied show states resembling the cubic 3C form. These results connect the classification work with the paper's wider explanation of how atomic stacking changes electronic and optical behavior.

Publication, recognition and the next practical step

The research was announced in October 2026. Accounts of KERI's announcement give October 23 and October 25 as the announcement date. The study was reported as published on September 8, while Acta Materialia assigns it to volume 321, article 122747, dated December 1, 2026.

The main milestones place the latest finding within a research effort spanning more than two years:

• June 1, 2024: The earlier study examined approximately 50,000 faults across 70 wafers and challenged several accepted wavelength assignments.
• September 8, 2026: The newer Acta Materialia study was reported as published, linking the standard (4,1) fault to 459 nanometers.
• October 2026: KERI announced the findings from the broader investigation of 207 wafers and 104,439 faults.
• December 1, 2026: The date assigned to the Acta Materialia volume containing the paper.

The project received support from the Ministry of Trade, Industry and Energy's Compound Power Semiconductor Advancement Technology Development Project. The paper was accepted without requests for revision and appeared in a journal described as being within the top 5 percent of its field in Journal Citation Reports.

Na also received an Honorable Mention in the 2026 HORIBA Masao Award for research on SiC defect analysis. According to KERI, this was the first recognition in the awards for a researcher affiliated with a Korean research institution since the awards began in 2004.

For manufacturers, the immediate task is to assess how the revised identification fits into existing inspection procedures. Better classification can help engineers trace defects back to growth or processing conditions and decide which changes to test. The intended benefit is a more accurate route from detecting a defect to understanding and controlling its cause.

Key Points

• The standard (4,1) Frank-type stacking fault in 4H-SiC emits at 459 nanometers at room temperature, rather than the longstanding 424 nanometer assignment.
• The investigation covered 104,439 stacking faults in 207 commercial wafers from nine suppliers.
• Researchers examined 176 ultrathin specimens and confirmed the (4,1) structure in six of nine targeted defects emitting at 459 nanometers.
• Calculations estimated emission at approximately 444 to 450 nanometers, supporting a signal near 450 nanometers.
• A 2024 study had already identified a different defect, the single Shockley-type (3,1) fault, at 424 nanometers.
• The findings provide scientific grounds to review IEC inspection guidance and improve defect classification.
• Na Moonkyung received an Honorable Mention in the 2026 HORIBA Masao Award.