Composition, structure and optical properties of group III nitride films

We are applying optical and structural characterization techniques to measure critical properties of thin films and substrates of the III-nitride materials (GaN, AlN, InN and their alloys). Ten years ago, most researchers considered the III nitrides to be laboratory curiosities, unlikely to be of great practical importance. However, after several breakthroughs in growth and doping technologies, the III nitrides have emerged as the leading semiconductor materials for short-wavelength (green, blue, and violet) visible-light-emitting diodes and laser diodes, as well as solar-blind photodetectors, and show promise for high-power, high-frequency, and high-temperature microelectronics. The III nitrides are probably now the third most economically important class of semiconductors, behind only Si and GaAs.

A primary goal of this project has been to develop a better understanding of how chemical composition and compositional inhomogeneity affect the optical properties of III-nitride alloys, especially indium gallium nitride (In_xGa_1-xN), which is an important alloy because it forms the active layers of visible-light-emitting devices. Thermodynamic models suggest that In_xGa_1-xN is unstable with respect to phase separation into high-indium and low-indium phases at typical growth temperatures. Phase separation and compositional inhomogeneity are thus expected to be significant effects in In_xGa_1-xN films. Several types of inhomogeneity may be present, including like-atom clustering (an increased frequency of In-In and Ga-Ga second-neighbor pairs as compared to a random alloy), quantum dot formation (1 nm to 5 nm diameter regions of nearly pure InN), spinodal decomposition, and long-range atomic ordering (alternating Ga-rich and In-rich planes along one crystal direction).

Some results of our measurements of a set of relatively thick, strain-relaxed In_xGa_1-xN/GaN/sapphire films with alloy compositions in the range 0.04<x<0.47 are discussed in a recent manuscript [1](click here for manuscript).The In_xGa_1-xN/GaN/sapphire and GaN/sapphire samples for this study were grown by Prof. Salah Bedair and coworkers at North Carolina State University in an atmospheric-pressure metalorganic chemical vapor deposition (MOCVD) system. Key results are summarized below.

The average In/(In+Ga) ratio within each In_xGa_1-xN film, denoted x_avg, was determined from wavelength dispersive x-ray spectroscopy in an electron probe microanalyzer (WDS/EPMA) by John Armstrong of the NIST Microanalysis Research Group. X-ray diffraction theta-2theta scans shows that Vegard's law holds; more specifically, the out-of-plane lattice constant (c) is a linear function of x_avg from x=0 (GaN) to x=1 (InN). From analysis of indium K-edge EXAFS of a subset of films with 0.12<x_avg<0.38, measured at the NIST/NSLS beamline, the In/(In+Ga) ratio in the indium second shell, denoted x_N=2, agrees with x_avg within the EXAFS measurement uncertainty. This result indicates that the types of compositional inhomogeneity that would cause a large increase in x_N=2, such as In-In clustering or quantum dot formation, are not major effects in the films with x_avg<0.38. However, other inhomogeneity-related effects such as long-range atomic ordering are not ruled out by the EXAFS, and in fact both we and the Bedair group have seen evidence of atomic ordering from TEM lattice-imaging and diffraction studies (not discussed in detail here).

Optical properties of the films were characterized by room-temperature transmittance spectroscopy, and both room-temperature and low-temperature cathodoluminescence spectroscopy in a scanning electron microscope (CL-SEM). In agreement with other researchers, we found that the optical band gap (E_G) of strain-relaxed In_xGa_1-xN is a quadratic function of x_avg with a large, negative second-order (bowing) parameter. For each film, E_G was determined from the energy of the optical absorption edge in the transmittance spectrum. For films with x_avg<0.4, the band gap was determined independently from the highest-energy peak in the room-temperature CL spectrum. The band gap vs. composition results, together with the fitted quadratic function, are shown in Fig. 1.

The fitted value of the bowing parameter is -4.57±0.75 eV. The large uncertainty of the bowing parameter arises from the lack of data in the x_avg>0.5 region. There is one published report that the intrinsic band gap of InN is 1.89 eV; this agrees well with the extrapolation to x=1 of our fitted function. The mechanism for the band gap bowing is not yet fully understood, but this effect is believed to be correlated with the large (10%) lattice mismatch between GaN and InN. Theoretical models suggest that the large bowing parameter is a property of the random alloy and thus is not an indication of (non-random) compositional inhomogeneity.

Low-temperature (15 K) CL spectroscopy of these films provides an important new result: an additional CL peak appears at high energy, 0.15 eV to 0.4 eV above the main band-edge CL peak (which shows almost no temperature shift from 295 K to 15 K). The intensity of the new high-energy peak, relative to the main band-edge peak, is highest in the alloys with the smallest indium fraction (x_avg<0.1). Spatially resolved CL measurements on a 200 nm length scale reveal that the spatial distribution of the high-energy peak is highly non-uniform, i.e. the intensity of this peak is much higher in specific "hot spots" with sub-micrometer dimensions than in the bulk of the film. Spatially resolved spectra from several films, selected to show the most intense high-energy peaks (i.e., spectra from the "hot spots") are plotted in Fig. 2.

Several models are under consideration for the localized "hot spots" that give rise to the high-energy peak. We do not think that the peak arises from indium clusters or quantum dots, because these structures would be expected to emit light at photon energies lower than the surrounding indium-depleted matrix. Further, structural measurements provide no evidence for indium-rich clusters in the films where the high-energy peak is observed. We suggest instead that the "hot spots" may arise from either (a) regions of high compressive strain (which would cause the band gap energy to increase); or (b) regions of low piezo-electric field strength relative to the surrounding matrix (piezo-electric fields produce a red-shift of the luminescence; thus the absence of piezo-electric fields in a localized region should produce a relative blue-shift). The spatial variation of the strain or piezo-electric field strength implied by these models is likely correlated with compositional fluctuations, through the composition dependence of the lattice constants. (If there are compositional fluctuations, then regions of differing composition must occur in proximity to each other. The occurrence of adjacent regions with differing equilibrium lattice constants necessarily implies the presence of strain, and strain in a Wurtzite structure crystal implies piezo-electric fields.)

Some early results of the In_xGa_1-xN study were published online in the MRS Internet Journal of Nitride Semiconductor Research [2]; a more recent report appears in the proceedings of the 27th International Symposium on Compound Semiconductors (ISCS-27, Monterey CA, 2000) [1]. The spatially resolved low-temperature CL measurements will be discussed in a future publication.

We are currently working with several other research divisions to strengthen and broaden the overall III-nitride measurement program at NIST. Three focus areas for ongoing and future work been identified: (1) the influence of lattice defects (such as stacking faults, polytypes, and inversion domains) and surface processing (polishing and etching) on the structural, optical and electronic properties of bulk single-crystal GaN, being developed as a substrate material for device fabrication, in collaboration with Norman Sanford and coworkers in the Optoelectronics Division - Optoelectronic Manufacturing Group; (2) optimization of ohmic contacts to GaN, based on correlation of the composition and structure of the metal-GaN interface with the contact electrical resistance, in collaboration with Albert Davydov and coworkers in the Metallurgy Division - Materials Structure and Characterization Group; and (3) composition and strain dependence of the optical and electronic properties of Al_xGa_1-xN layers, being developed for solar-blind photodetectors and for dielectric mirror stacks within laser devices, in collaboration with Norman Sanford and other researchers.

To demonstrate the applicability of our measurement techniques to one of the newer focus areas, depth-resolved CL measurements (performed by varying the incident electron energy, which determines the electron penetration depth) of a GaN/Ti(10nm) structure are shown in Fig. 3. The depth-resolved CL reveals an infrared emission band localized near the metal-semiconductor interface. TEM and related structural and compositional measurements (not discussed in detail here) suggest that a TiN layer is formed by a chemical reaction at the GaN-Ti interface. The infrared CL band may then originate either from the interfacial TiN layer, or from defects in the nitrogen-deficient GaN layer near the interface.

[1] L.H. Robins, J.T. Armstrong, R.B. Marinenko, M.D. Vaudin, C.P. Bouldin, J.C. Woicik, A.J. Paul, W.R. Thurber, K.E. Miyano, C.A. Parker, J.C. Roberts, S.M. Bedair, E.L.Piner, M.J. Reed, N.A. El-Masry, S.M. Donovan and S.J. Pearton, "Optical and structural studies of compositional inhomogeneity in strain-relaxed indium gallium nitride films", 2000 IEEE International Symposium on Compound Semiconductors (Institute of Electrical and Electronics Engineers, Inc., 2000), pp. 507-512

[2] L.H. Robins, A.J. Paul, C.A. Parker, J.C. Roberts, S.M. Bedair, E.L.Piner, and N.A. El-Masry, "Optical absorption, Raman, and photoluminescence excitation spectroscopy of inhomogeneous InGaN films", MRS Internet J. Nitride Semicond. Res. 4S1, G3.22 (1999)