The stress dependencies of the phonon modes in a 4H silicon carbide (SiC) crystal were investigated. The deformation potentials of the A1(TO), E2, and E1(TO) modes were determined on the basis of uniaxial stress tests. The A1(TO) mode was sensitive to stress along the c axis, whereas the E1(TO) and E2 modes exhibited larger dependencies on the stress perpendicular to the c axis than the A1(TO) mode as expected from their atomic displacements. The coefficient of the frequency shift in the E2 mode for an isotropic biaxial stress in the c plane was determined to be −323 MPa/cm−1. We applied cross-sectional Raman measurements to SiC metal-oxide-semiconductor field-effect transistors. The horizontal and vertical stress components, σ11 and σ33, were derived from the frequency shifts of the E2 and A1(TO) modes using the deformation-potential constants. A compressive horizontal stress was observed near the interface between the SiC chip and the metals. The observed compressive stress increased as the temperature decreased. The main cause for the compressive horizontal stress at low temperature is considered to be the difference in the coefficients of thermal expansion of the SiC chip and the metals. The results suggest that the temperature of the fabrication process is one of the key factors for achieving a reduction in the thermal stress in the SiC power devices.

An experimental method to determine the temperature dependence of residual stress in threedimensional (3D) structures was developed using polarized Raman spectroscopy. Stresses of a copper-filled silicon via at three temperatures, 223, 298, and 413 K were derived by measuring the frequency shift of the optical phonons through the backscattering geometry from the cross-section of the structure and assuming non-isotropic biaxial (horizontal and depth) stresses on the crosssection. Both stress components changed from tensile to compressive in almost all areas as the temperature changed from 213 to 413 K. The absolute stress values increased at both low and high temperatures and were smallest at 298 K, which was nearest to the process temperature of copper filling by plating. The main cause of stress is considered to be the difference in the coefficient of thermal expansion between copper and silicon. These results indicate that the temperature dependence of stress of copper-filled vias is affected mainly by their fabrication temperature. Process temperature is one of the key factors for the reduction of thermal stress in 3D structures such as integrated circuits connected by through-silicon vias. V C 2013 AIP Publishing LLC.

Ga-rich zones created along macrosteps in n-AlGaN plausibly function as electron pathways of AlGaN-based deep-ultraviolet (DUV) LEDs fabricated on AlN templates using 1.0°-miscut c(0001) sapphire substrates toward the m[1-100] axis. This work was performed to clarify AlN mole fractions (xAl) of Ga-rich zones. xAl ≃ (7/12, 6/12, and 5/12) was observed in Ga-rich zones in AlαGa1−αN (α ≃ 0.63, 0.55, and 0.43, respectively) by the method proposed in our previous article in which we showed that Ga-rich zones of Al8/12Ga4/12N were created in Al0.7Ga0.3N. xAl in the Ga-rich zones obtained from an energy-dispersive x-ray signal by scanning transmission electron microscopy calibrated by Rutherford backscattering well agreed with xAl obtained by cross-sectional cathodoluminescence (CL) spectroscopy using scanning electron microscopy. A weak CL shoulder peak corresponding to Al4/12Ga8/12N was also observed for Al0.43Ga0.57N. In addition, xAl ≃ n/12 (n = 6–9) in Al-rich zones appeared in the rest of the Ga-rich zones. Furthermore, nanobeam electron diffraction patterns of the Ga-rich zones indicated a high possibility of a regular configuration of Ga and Al atoms on the c(0001) plane in our samples. Consequently, xAl values in nonflat AlGaN layers with macrosteps were often determined to be near n/12 (n: integer). Thus, Ga-rich zones (xAl = n/12: n = 4–8) formed in our nonflat AlGaN layers, which originated from the macrosteps along [11–20] edgelines normal to the m[1–100] axis, are suggested to be metastable. The creation of discrete xAl in Ga-rich zones should contribute to the stable production of DUV-LEDs using high-miscut sapphire substrates.

Cross-sectional cathodoluminescence (CL) and scanning capacitance microscopy (SCM) measurements were carried out for silicon carbide (SiC) metal–oxide–semiconductor field-effect transistors (MOSFETs) to investigate process-induced defects. The DI defect-related line at 426 nm and a broad luminescence at approximately around 430–470 nm, which were produced by ion implantation, were observed in addition to the near-band-edge emission. CL images showed that the densities of nonradiative recombination and DI centers were high near the source region. Moreover, DI centers existed even in the n-drift region located 10 µm from the surface. These results indicate that many types of defects diffuse and interact with each other during annealing even in the area where dopant atoms are not implanted. The annealing process not only activates dopant atoms but also induces the diffusion of unstable native defects and transforms their structure into more thermally stable defects such as DI centers.

When nonflat Al x Ga1−x N quantum wells (QWs) for producing 285 nm light emitting diodes (LEDs) were fabricated on n-AlGaN on AlN templates with dense macrosteps on c(0001) sapphire substrates with a 1.0° miscut relative to the m[1–100] axis, composite electroluminescence (EL) spectra from both inclined and terrace zones in Al x Ga1−x N QWs (x∼ 1/3) were generated owing to compositional and thickness modulations. The shoulder or main peaks in composite EL spectra tended to locate at fixed discrete wavelengths of ∼287, ∼292, and ∼296 nm from 12 nonuniform 285 nm LED wafers that were involved in nonnegligible run-to-run drift, even though these wafers were fabricated using the same source gas flow parameters for metal-organic vapor phase epitaxy. The discrete wavelengths of ∼287, ∼292, and ∼296 nm were attributed to EL from Al1/3Ga2/3N QWs with thicknesses of 8, 9, and 10 monolayers (ML), respectively, by referring to the results of cathodoluminescence (CL) analysis. Also, when nonflat Al x Ga1−x N QWs (x∼ 1/2) for 265 nm LEDs were grown, single-peak-like EL spectra were mainly generated from the inclined zones in nonflat QWs. The EL spectra taken from four nonuniform 265 nm LED wafers tended to show weak structures or main peaks at ∼257, ∼261, ∼266, and ∼271 nm, which were also attributed to emissions from Al1/2Ga1/2N QWs with thicknesses of 5, 6, 7, and 8 ML, respectively, by referring to CL analysis results. The creation of Al1/3Ga2/3N and Al1/2Ga1/2N in nonflat QWs in this work was in agreement with the results of our previous studies that indicated the creation of metastable Al n /12Ga1−n/12N (n: consecutive natural numbers). Furthermore, QW thicknesses of consecutive n ML may imply that Al1/3Ga2/3N and Al1/2Ga1/2N have 1 ML configurations of Al and Ga atoms on a c(0001) plane.