This article was received on December 30, 1996 and accepted on May 23, 1997.
1. Experimental
1.1 Sample Preparation
Nitride growth was done in a Riber molecular beam epitaxy system with a custom nitrogen plasma source. The plasma source is described in Ref. Reference Spah, Hess, Stormer, White and Short[7]. The nitrogen plasma is excited with 10-15 watts of RF power in nitrogen at 65-85 mTorr. The typical plume of the plasma source is only 2 cm wide at the 7.5 cm diameter substrate holder, resulting in spatially varying active nitrogen flux. The films described here were doped with magnesium at relatively high levels. The (0001) sapphire substrate was heated to 675°C in vacuum, then exposed to the nitrogen plasma at a temperature of 600°C for 30 minutes before starting the GaN deposition. The reflection high energy electron diffraction (RHEED) pattern was observed through the whole process. The first 30 minutes of growth were done with the Ga effusion cell adjusted to obtain a slow growth rate of about 50 nm/hr, a substrate temperature of 600°C, and a plasma power of 10W. This initial growth produced a streaky RHEED pattern indicative of a substantially smooth, crystalline surface. The growth was then continued with the Ga cell set to grow 150 nm/hr; the substrate temperature was raised to 650°C; and the plasma power was increased to 15W. At the end of the growth, the film had a very streaky RHEED pattern. The film thickness was 0.49 μm thick, as determined by the interference fringes in optical transmission spectroscopy.
1.2 Atomic Force Microscopy
The GaN film was imaged with a Nanoscope III (Digital Instruments) contact mode AFM under ambient conditions. A 200 μm, triangular, silicon nitride cantilever (k = 0.12 N/m, Digital Instruments) was selected to perform the imaging. The sample was secured on a steel disk with double sided tape before being placed on a magnetic, 12 μm scanner. In a random 12 μm scan of the surface only one GaN island or a portion of a GaN island could be observed. However, the x-y translation stage of the cantilever mount allowed several different GaN islands to be “centered” within the range of the scanner. The images were stored as 512 × 512 point arrays and no image processing was necessary.
1.3 Cathodoluminescence
The instrument used for CL measurements is based on a JEOL JSM-6400 scanning electron microscope equipped with an Oxford Instrument CF302 system, which allows measurements to be made between room temperature and 8K. The CL system is described in a previous publication Reference Chao, Freiler, Levy, Lin, Cargill, Osgood and McLane[8]. CL measurements were made at room temperature and at 8K with 15 kV beam voltage and 60 nA beam current.
1.4 Photoluminescence
Photoluminescence (PL) measurements of the films were made at 5K and at room temperature. A He-Cd laser was used as an excitation source; the emitted light was dispersed by a 0.5 m monochromator and detected by a CCD camera. PL spectra were measured at 7 spots along the full sample to determine the spatial uniformity of the film.
1.5 X-ray Diffraction
X-ray diffraction measurements of the films were made using a 4-circle diffractometer with monochromatized Cu Kα radiation.
2. Results and Discussions
Inspection of the film by optical microscopy (Figure 1a) revealed interesting spatial variation of the film morphology. A circular region about 2 cm in diameter was essentially smooth and featureless. Outside this region, the smooth film was punctuated with gallium droplets, about 2-3 μm in diameter, spaced roughly 5-30 μm apart. In an annulus about 1mm wide separating the regions with and without droplets was a region which looked much like the dropleted region, except that instead of droplets there were features with a texture too fine to resolve in the optical microscope.
Figure 1a. Optical micrograph of a GaN film showing the boundary region, indicated by the dashed line, between Ga droplets (bottom, left) and GaN islands (top, right). The field of view is 400 μm × 280 μm.
The SEM micrograph in Figure 1b shows these densely-distributed features on the surface of the as-grown film. These island-like features are 2-3 μm in diameter, and separations between islands vary from 5 μm to 30 μm. The typical morphology of these features, shown in Figure 1c, consists of entangled submicron faceted structures with many hexagonal (120°) facet angles. Observations by AFM, as shown in Figure 1d, give similar results for the feature morphologies, although the micrographs shown in Figure 1c and Figure 1d are from different islands. Figure 1e shows the three-dimensional AFM image recorded from the island used in Figure 1d, which shows that the island is rising above the surface by about 250 nm. As described earlier in this paper, an inhomogeneous surface morphology for GaN films had been reported by Trager-Cowan et al. Reference Trager-Cowan, O’Donnell, Hooper and Foxon[4]. However, the features on the surface of our sample are labyrinthine faceted structures, unlike the simple hexagonal plates or hillocks reported in Ref. Reference Trager-Cowan, O’Donnell, Hooper and Foxon[4], or the raised hexagonal crystals in Ref. Reference Ponce, Bour, Gotz and Wright[5]. The background material for our sample is highly epitaxial, both from RHEED patterns observed during growth, as shown in Figure 2a, and from the electron channeling pattern Reference Joy, Newbury and Davidson[9], as shown in Figure 2b, obtained from an area including both islands and background material. This was further confirmed by the x-ray diffraction measurements, which showed only the GaN (0001) orientation grown on the (0001) sapphire. No other diffraction planes were observed close to the substrate normal, even in scans with wide-open slits.
Figure 1b Top view SEM image showing islands of micron size on the surface of the GaN film.
Figure 1c. Top view SEM image of one of the islands.
Figure 1d. Top view AFM image of one of the islands.
Figure 1e. Perspective view AFM image showing the height of the island.
Figure 2a. RHEED pattern observed during the growth of the film.
Figure 2b. Electron channeling pattern obtained from an area including both islands and background material.
Spatially-resolved CL measurements were made to relate the luminescence efficiency to the microstructures described above. As shown in Figure 3, only a broad, asymmetric peak was observed in the CL spectra obtained at room temperature from (a) one of the islands and (b) its neighboring background material. The peak wavelengths, 364.8 nm and 367.0 for luminescence from island and from background material respectively, are attributed to band-edge emission Reference Ponce, Bour, Gotz and Wright[5] Reference Shan, Xie, Song and Goldenberg[10]. The band-edge luminescence from the island has a narrower spectrum than that of the neighboring background material, and the luminescence intensity of the island is an order of magnitude higher, although both spectra shown in Figure 3 have been scaled to have the same peak height. The appearance of noise spikes in the spectrum obtained from background material is due to the poor signal-to-noise ratio.
Figure 3. CL spectra at room temperature from (a) one of the islands and (b) its neighboring background material.
The ratio of luminescence intensity between an island and its neighborhood could be as high as two orders of magnitudes, as shown in Figure 4 from measurements on five islands and their corresponding neighborhoods. Figure 4 also shows that the luminescence efficiency varies from island to island, while it is spatially uniform in the background material. The monochromatic CL image, shown in Figure 5, recorded at the peak luminescence energy from the same area used for the SEM image in Figure 1b, also confirms the luminescence contrast between islands and background material.
Figure 4. Comparison of luminescence efficiency at room temperature between five islands and their corresponding neighborhoods.
Figure 5. Room-temperature monochromatic CL image at 365 nm from the same area shown in the SEM image in Figure 1a.
In addition to spatial variation of luminescence intensity, a spectral shift of 1-2 nm in peak wavelength was observed between individual islands and the neighboring background material, as shown in Figure 6. The peak luminescence of each island always had a shorter wavelength than that of the corresponding neighborhood. It is well known that high impurity concentration can lead to tails of states extending as impurity bands into the forbidden energy gap, and that this will cause a smaller energy gap and a broader luminescence Footnote [a]. Therefore, the blue shift and narrowing of the luminescence from the islands might result from the islands having lower Mg doping and fewer impurity- or defect-related near band-edge, or band-tail states than the background material. Luminescence from the islands would then be mainly band-to-band, and luminescence >from the background material would be red shifted because it mainly involved band-tail states. We speculate that the islands may lower Mg concentration than the background material because they are formed by liquid phase epitaxy from Ga droplets at the edge of the nitrogen plasma beam, as discussed below. The lower incident nitrogen flux responsible for forming the islands is expected to give a lower flux of oxygen from the ppm level water impurity in the nitrogen source. Less oxygen in the islands is expected to reduce incorporation of Mg as a co-dopant Reference Yamamoto and Katayama-Yoshida[12] relative to Mg incorporation in the background material.
Figure 6. Comparison of peak luminescence wavelength at room temperature between five islands and their corresponding neighborhoods used in Figure 4.
CL spectra recorded at 8K from one of the islands and from its neighboring background material are shown in Figure 7(a) and (b). Three main features appear in the spectrum from the island. (1) An intense peak at 357.2 nm with a FWHM of 5.2 nm is associated with excitons bound to neutral acceptors (BE), which has been previously reported Reference Merz, Kunzer and Kaufmann[13], Reference Fischer, Wetzel, Haller and Meyer[14]. (2) A weaker, broader peak attributed to recombination at donor-acceptor (DA) pairs occurs at 375.4 nm Reference Rieger, Metzger, Angerer, Dimitrov, Ambacher and Stutzmann[15]. (3) An LO-phonon replica of the DA transition occurs at 385 nm. The BE emission at low temperature from background material was also found to be red shifted by less than 1 nm with respect to that from islands in five different measurements. In these measurements, as the electron beam was moved from an island to its neighboring material, the intensity of bound exciton emission was reduced by factors of 80-100, while the relative intensity of bound exciton emission to donor-acceptor-pair transition dropped by a factor of about two. In addition to the spectral features described above, a broad, weak emission band, centered at ~ 490 mn, appeared in the spectra of the background material, which is probably associated with the deep levels from impurities or structure defects. However, our observed impurity-induced luminescence band is more greenish than in other reported data, which place the impurity luminescence band at ~ 550 nm Reference Glaser, Kennedy, Doverspike, Rowland, Gaskill, Freitas, Khan, Olson, Kuznia and Wickenden[16], Reference Freitas, Kennedy, Glaser and Carlos[17]. Our observations suggest that the GaN material in the islands has fewer impurities and/or structure defects than the background material, in agreement with the suggestions by Trager-Cowan et al Reference Trager-Cowan, O’Donnell, Hooper and Foxon[4].
Figure 7. CL spectra at 8K from (a) one of the islands and (b) its neighboring background material.
PL measurements were consistent with the CL measurements. Strong, spatially inhomogeneous band-edge luminescence was observed at 300K and at 5K in the annular region where the faceted islands were seen, whereas only weak band-edge luminescence was observed in the smooth, featureless center region. At 5K the difference is particularly striking as shown in Figure 8. PL from the smooth center region has almost no band-edge peak, with most of the luminescence coming in the broad band between 380 nm and 390 nm. This band is probably associated with Mg acceptors Reference Merz, Kunzer and Kaufmann[13], Reference Myoung, Shim, Kim, Gluschenkov, Kim, Kim, Turnbull and Bishop[18]. The difference between the CL spectrum of Figure 7b and the PL spectrum of Figure 8b may be due to the higher excitation density for CL, or to intrinsic differences between the background material from the halo region and the smooth, featureless center region.
Figure 8. Photoluminescence spectra (a) at 5K from the annular region with faceted islands (b) from the smooth, featureless central region.
The cause of the spatial variation of the film morphology is no mystery; Ga droplets form in the region outside the direct plasma beam as a result of insufficient active nitrogen. GaN is thermodynamically unstable under conditions used in MBE, and it only forms because the decomposition to Ga and N2 is slow compared to the formation of GaN in active nitrogen Reference Newman, Ross and Rubin[19]. In the plasma beam, Ga droplet formation is suppressed, and atomically smooth (0001) facets grow by step flow or layer-by-layer growth typical of MBE. It is the region at the boundary which is a puzzle: what are the faceted islands, why and how do they form, and why is the luminescence so much brighter?
The size and distribution of the faceted islands are identical to those of the droplets in the outer region of the film; this suggests that the two morphologies have a common origin. One possibility is that Ga droplets are formed in regions out of the plasma beam during the low temperature nucleation step. Droplets near the edge of the plasma beam are then consumed, forming high quality GaN from a nitrogen saturated melt in a process similar to that demonstrated by Argoitia et al. Reference Argoitia, Hayman, Angus, Wang, Dyck and Kash[20]. Another possibility is that GaN islands of a different polarity or orientation form in the initial growth step, and then nucleate droplets outside of the beam and high quality GaN at the edge. However, we found no evidence for this in the x-ray diffraction.
3. Conclusions
The morphology of an MBE-grown GaN film on a (0001) sapphire substrate was characterized by SEM and AFM. Multiple-faceted features of 2-3 μm in diameter rose above the film surface by about 250 nm. Differences in luminescence intensity and spectral shifts were observed in CL spectra recorded at both room temperature and 8K from island features and from the background material. The luminescence intensity of any individual island was always two orders of magnitude or more greater than that of neighboring background material.
The high luminescence from the faceted islands is a tantalizing mystery. It might be a result of better crystallinity, or reduced impurities. Determining the cause is the next topic for investigation and could lead to improved growth techniques.
