An unsolved mystery cause of semiconductor defects

Semiconductors are one of the most globally in-demand technologies, used everywhere from personal computers to industrial manufacturing. Scanning spreading resistance microscopy (SSRM) is an industry standard for analyzing a semiconductor’s charge-carrying capacity and the underlying dopants in its structure at very small scales, especially to compare working and failing chips.

Adlmaier et al. experimentally tested SSRM on semiconductor samples prepared with different polishing methods and found that the traditional method using aluminum oxide caused systematic errors by introducing trace defects. A modern method of embedding the sample in epoxy resin and polishing it with diamond created no errors.

“I’m doing my PhD, and this was the best and worst thing that could happen to me,” said author Thomas Adlmaier. “Every time I thought I understood what’s happening, I was wrong.”

To determine the root cause of the SSRM errors, the researchers tested multiple superconductor samples, identical but for their preparation methods. Higher spreading resistance occurred in the aluminum oxide sample, which indicated a defect in the semiconductor crystal. The aluminum oxide introduced trace amounts of copper, which interacted with the dopants in the sample. However, copper alone could not account for the full spread of errors.

“Either we are missing something, or it’s a combination of things,” said Adlmaier. “It’s quite complex, and that’s where my suspicion is at the moment.”

Aside from copper, the team thinks hydrogen from trace amounts of humidity in the air may disrupt the chips during storage. When embedded in resin and polished with diamond, ambient air is less disruptive.

Traditional aluminum oxide polishing need not be fully phased out, however; the team devised recommended steps for reverting most of the defects, including annealing the sample at 150 C for 15 minutes.

Source: “On the impact of sample preparation on scanning spreading resistance microscopy,” by Thomas Adlmaier, Stefan Doering, and Thomas Mikolajick, Journal of Applied Physics (2026). The article can be accessed at https://doi.org/10.1063/5.0353363 .