Silicon surface defects are a materials issue with implications for devices built on silicon, including metal–oxide–silicon field-effect transistors. At the boundary between silicon and silicon dioxide (SiO₂), electrically active defects can act as carrier-recombination centers. Understanding their microscopic structure—and how oxidation changes it—requires measurement techniques sensitive enough to detect very low defect concentrations.
A study of commercially available (001)-oriented silicon wafers used electron spin resonance (ESR) combined with spin-dependent recombination (SDR) to examine two interface-related signals, known as Pm and KU1. The results link both signals to a center involving two neighboring silicon dangling bonds and show that the overlying oxide affects the center’s measured properties.
Why silicon surface defects are difficult to measure
Conventional ESR can identify paramagnetic defects by measuring their response to a magnetic field. However, the concentration of centers at a native silicon oxide interface may be too low for conventional detection. Traditional ESR studies of some interface defects have therefore used heavily oxidized samples and multiple wafers to increase the detectable signal.
SDR-ESR provides another route. When light generates free carriers in silicon, their recombination rate can depend on the spin state of a defect. That rate affects the sample’s photoconductivity. Under magnetic resonance conditions, a change in spin state can alter the microwave photoconductivity, which can then be monitored to reveal the defect’s ESR spectrum.
This approach made it possible to observe the Pm and KU1 signals in silicon wafers bearing native or thermally produced oxide layers. Pm has spin S = 1/2, while KU1 has spin S = 1. Both show orthorhombic symmetry and are associated with the (001) silicon surface.
What the Pm and KU1 signals reveal
The two signals have different spin states, but their behavior under changes to the surface oxide points to a shared structural origin. The researchers propose that the center consists of two interacting dangling bonds on neighboring silicon atoms. A dangling bond is an unsatisfied chemical bond at a surface or interface that can contribute an unpaired electron.
In the proposed model, the two dangling bonds form bonding and antibonding molecular orbitals. Their electron occupancy determines the observed spin state. The KU1 signal is attributed to an excited triplet state of the neutral center, in which two electrons have parallel spins. The Pm signal is associated with a state containing one or three electrons and therefore a net spin of 1/2. These assignments are a structural model supported by the measured spectra, rather than direct imaging of the defect.
Evidence for the two-atom structure also appears in the KU1 spectrum’s hyperfine pattern. Hyperfine interaction arises from coupling between electron spins and nearby atomic nuclei. The observed pattern is consistent with interaction involving silicon-29 nuclei at two equivalent neighboring sites, supporting the proposed pair of dangling bonds.
How oxide formation changes the interface center
The study compared samples with surface oxide against samples treated with hydrofluoric acid (HF) to remove the oxide. After HF treatment, the lines associated with KU1 and Pm became narrower. The KU1 fine-structure splitting also increased, while the Pm g-tensor values remained unchanged within the reported measurements.
These changes indicate that the oxide layer affects the local environment of the center. In particular, the KU1 splitting reflects the magnetic dipole interaction between the two electrons. From the measured splitting, the authors estimated a slightly greater separation between the interacting electrons when SiO₂ was present. They attributed this difference to mechanical stress imposed by the oxide at the silicon surface.
The spectra changed again after the treated samples were left in air for several days, returning toward their earlier form as oxide developed. This behavior connects surface preparation and oxidation conditions with the ESR response. It also illustrates why measurements of interface defects should account for sample history, not just wafer orientation or conductivity type.
Implications for semiconductor characterization
For engineers and technical decision-makers concerned with silicon devices, the findings underline the value of interface-sensitive characterization. A defect population may be difficult to detect with a conventional technique yet still participate in carrier recombination. SDR-ESR can help expose such centers by measuring their effect on photoconductivity rather than relying only on a large defect concentration.
The work also separates several considerations that can otherwise be conflated:
- Defect identity: Pm and KU1 have distinct spin signatures but are attributed to the same underlying center in different electronic states.
- Surface orientation: The signals were observed on (001) silicon, with two orientations associated with orthogonal surface directions.
- Oxide condition: Removing or regrowing oxide changes spectral linewidth and KU1 fine structure, consistent with a change in the local mechanical environment.
- Detection method: The spectra depend on photoexcited carriers and spin-dependent recombination, so illumination and measurement conditions are central to observation.
These points are relevant when interpreting interface measurements or comparing samples prepared under different oxidation and cleaning conditions. The study does not establish a direct device-performance penalty for a specific process or quantify production impact; instead, it provides a microscopic model for two recombination-related signals and demonstrates how the oxide modifies their spectroscopic characteristics.
Conclusion
The Pm and KU1 ESR signals offer complementary views of a silicon surface center involving two neighboring dangling bonds. Spin-dependent microwave photoconductivity makes these low-concentration centers observable, while hyperfine structure and changes after oxide removal support the proposed model. For semiconductor characterization, the practical lesson is to treat the Si/SiO₂ interface as a condition-sensitive region: defect signals, spin states, and oxide-related stress all matter when interpreting measurements of silicon materials.