by D. K. Gaskill
The combined Device Research Conference and Electronic Materials Conference (DRC/EMC) was held at the University of California, Santa Barbara, 24-28 June 1996. Over 30 papers were presented on the III-N semiconductor system out of a total of over 230 papers. A few of the highlights are herein described.
Y.-F. Wu of the Denbaars and Mishra UCSB group presented results on "GaN HFETs and MODFETs with Very High Breakdown Voltage and Large Transconductance." For both HFET and MODFET device types, a 200 Å AlN nucleation layer was grown by OMVPE on a c-plane sapphire substrate, upon which the transistor channels were directly grown. The channel thicknesses were 0.4 and 0.3 µm for the HFET and MODFET, respectively, and was unintentionally doped to near 4 x 1016 cm-3. The gate barrier of the HFET was an undoped 300 Å Al0.15Ga0.85N layer whereas for the MODFET it was modulation-doped with a 30 Å spacer, a 150 Å doped layer (3 x 1018 cm-3) and a 120 Å doped cap. Cl2 RIE was used for mesa isolation and Ti/Al ohmic and Au gate metallizations were used. Gate-drain separation was 2 µm. Gate lengths were 1.5 µm and 1.0 µm, breakdown voltages were 340 and 230 V, and turn-on voltages were 2.0 and 1.7 V for the HFET and MODFET, respectively. Highest external transconductances were about 140 mS mm-1 for both device types and an intrinsic value of about 250 mS mm-1 was estimated. A power output of 1.1 W mm-1 at 2 GHz was observed. Z. Fan of University of Illinois also reported on the transconductance and breakdown voltages of MODFETs. Extrinsic transconductance in the dark was 210 mS mm-1 and a 90 V breakdown voltage for 1 mµ gate-drain spacing were reported.
B. W. Lim of APA Optics, Inc., described the responsivity of OMVPE grown AlGaN photoconductors as a function of Al mole fraction. Compositions ranged from 5 to 61 % and the layers were all highly resistive and thus were well suited for photoconductive detectors operating in the DC mode. R. D. Underwood of UCSB presented results on selective OMVPE regrowth of GaN pyramids for electron field emission. A 2000 Å PCVD SiO2 layer was used for masking and standard contact lithography was used for patterning. Field emission from a few of the the pyramids was observed.
B. J. Skromme of Arizona State University gave a very interesting presentation on the "Luminescence and Raman Properties of GaN Epilayers Grown on Sapphire and 6H-SiC." The luminescence data derived were from high resolution spectroscopy that took into account the effects of strain on the samples examined. Reflectance spectroscopy, as well as selection rule analysis of spectral features obtained from magnetospectroscopy resulted in convincing identification of various transitions, both old and new. In particular, evidence was presented for the misidentification of a bound exciton line, at 3.365 eV which is shown to be due to the underlying sapphire substrate. Additional details were recently distributed by Skromme via Max Yoder's (ONR) email mailing list and are reproduced below:
From: skromme@asuvax.EAS.ASU.EDU (B. Skromme)
We have recently observed the n=2 excited state of both the "A" and "B" free excitons in GaN and identified them using a combination of low temperature photoluminescence, reflectance, and magnetospectroscopy in applied fields up to 12 T. The observation of the peak in reflectance with the appropriate strength and its large diamagnetic shift and Zeeman splitting clearly demonstrate that it is indeed an n=2 free exciton. From the n=1 to n=2 separation, a preliminary rough estimate of the free exciton binding energy in GaN is 26.4 meV (using a crude hydrogenic model). More detailed analysis is in progress. Several previous identifications based only on luminescence measurements, which deduced much smaller binding energies of 18-22 meV, are incorrect because they neglected strain effects and confused intrinsic with extrinsic transitions.
We also observed the "two-electron" replica of the neutral donor-bound exciton in the low temperature photoluminescence spectrum of GaN for the first time in material grown by MOCVD and metalorganic MBE on sapphire and SiC substrates. The transition was studied in magnetic fields up to 12 T and shows the expected splitting corresponding to the Zeeman splitting of the n=2 state of the donors. The separation of this peak from the normal neutral donor-bound exciton allows us to determine a donor binding energy of about 29 meV (neglecting the axial splitting of the n=2 state). More detailed analysis is in progress. The binding energy is significantly less than that previously reported by infrared measurements on material grown by hydride VPE (35.5 meV), implying that the residual donor is different in the two cases.
Finally, we have found that the 3.365 eV photoluminescence peak that has been reported a dozen or more times for the last 26 years or so in GaN grown on sapphire substrates is actually sometimes the "F2+" luminescence band of sapphire (which involves excitons bound to a pair of oxygen vacancies), and is sometimes a different, unidentified system in sapphire that occurs at almost exactly the same energy. The two systems are distinguished by their very different degrees of phonon coupling. The occurrence of these peaks is strongly dependent on the sapphire vendor, etc. Previous attributions of these peaks as bound excitons in GaN appear to be incorrect.
H. Amano of Meijo University presented information on the fabrication and properties of OMVPE grown GaN/InGaN inner stripe geometries, which he claimed would be useful for reduced p-type contact resistivity in laser structures. His analysis suggested the influence of piezoelectric fields in the well red shifts the emission. Superlinear emission intensity as a function of electron injection current was reported by M. Asif Khan of APA Optics for an OMVPE grown GaN/InGaN multiple QW structure. J. W. Yang, in a collaboration between APA Optics, Colorado State University, and Carnegie Mellon University, described the growth of GaN-InGaN heterostructures on Si substrates. The best films were formed after a two step buffer layer process. The first was the deposition of GaAs on Si utilizing techniques that are well known, followed by the usual GaN or AlN low temperature buffer layer. G.-C. Yi of Northwestern described Se-doping of GaN films using H2Se. Electron concentrations up to 6 x 1019 cm-3 were obtained, but at such high concentrations compensation effects became important.
S. A. Safvi, University of Wisconsin, presented a finite element-based model for the HVPE growth of GaN which revealed that during growth the ammonia mole fraction near the center of the samples was low due to reactor geometry effects and high total flow rate. Modifications to the growth was performed to correct the situation. This group is now growing thick layers (for potential substrate use) with growth rates of 60-120 µm hr-1, XRD rocking curve FWHM of 200-500 arcseconds, and PL results with strong excitonic emission and little yellow band. Further experimental investigation suggested that parasitic (upstream) reactions between the ammonia and gallium chloride plays a significant role in affecting film morphology and other properties. H. Lee of Stanford demonstrated improved HVPE GaN morphology and XRD FWHM when 500 Å AlN buffer layers grown at 1050 C are used.
C.-K. Sun of UCSB presented details on a time-resolved PL study of InGaN. This approach is being used to assess the quality of InGaN QW and the surrounding interface. They find that the PL decay is a function of both exciton and free carrier radiative recombination processes, where at low temperatures the exciton process dominates. At low temperatures, for 1.7 nm QWs, the lifetime is 420 ps rising to 480 ps for 3.4 nm QWs. S. Chichibu of Science University of Tokyo reported a photoreflectance study of the A,B, and C free excitons and derived deformation potentials from the measurements. H. Zull of University of Würzburg reported on the formation of DFB structures for GaN lasers. The gratings were etched by ECR-RIE, 200 V bias, at a rate of 50-100 nm min-1. PL on 26 nm GAN lines was similar to that from large mesas.
J. M. Redwing of ATMI reported on the properties of 2DEG formed at the AlGaN/GaN interface. The heterostructures were grown on both sapphire and 6H SiC. The best mobilities for the 2DEG are found for growths on SiC substrates., 7500 cm2V-1s-1 whereas 5700 cm2V-1s-1 was found on the sapphire substrates. Shubnikov-de Haas oscillations were strong for the epilayers on SiC substrates and observed at temperatures up to 10 K. The effective mass of electrons was found to be 0.18(2) electron masses, the quantum scattering time was 0.8 x 10-13 s, and the data implied that ionized impurities dominated the scattering.
J. Speck, UCSB, described the nucleation process of GaN on sapphire by studying the morphology of 6 samples with increasing thickness, from the 200 Å nucleation layer to 4000 Å. The samples display mix or pure screw dislocations inside the initially nucleated islands and coalescence of islands is found to result in the formation of edge dislocations. G. Bi of UC San Diego presented work on the incorporation of N in InP. Also, M. Skowronski of Carnegie Mellon presented work on the incorporation of B into GaN. In both cases, limited solubility was found, in the 1-5 % range and this limit appears to be due to growth kinetics rather than thermodynamic stability of an alloy in the volume of the film.