It is shown that GaP layers grown upon Si at a single temperature of 900ºC can have a crystalline quality superior to that exhibited by previous two—step and one—step growth methods. The layers are characterized by a planar network of misfit dislocations confined to the interface plane an a reduced density of threading dislocations (low 106 cm-2; previously >108). Very few threading defects were observed in areas devoid of amorphous oxide contamination, as shown by HREM examination of cross—sectional samples. A low growth rate during nucleation enhances crystalline perfection, since it decreases the tendency toward three—dimensional islanding.

Since the discovery in 1973 that GaAs/GaAsP superlattices can be grown with low dislocation densities, considerable interest has developed in utilizing superlattices as dislocation filters in multilayer semiconductor device structures. Many attempts to implement this process have been described, with varying degrees of success being achieved. Some investigators have reported favorable results; some have observed no effect; and in some cases the situation was actually made worse. This paper analyzes these reports and attempts to clarify the confusion that has arisen. Suggestions are made for improved effectiveness. Factors considered include the strain between layers, the layer thickness, the concept of critical thickness, the dislocation geometry, and the influence of buffer layers and growth conditions.

The nucleation, annealing and growth of GaAs and GaP layers on Si substrates have been studied. The morphology of these heteroepitaxial layers was examined by SEM while their structural quality was evaluated by conventional and high resolution TEM. SIMS was used to study the cleanliness of the Si/III-V interface and how that affects the epitaxial growth. The mode of growth and defect structure of the GaP layers were found to depend primarily on the quality of the interface. On the other hand, the defect density in the GaAs layers was dictated by the mismatch and the growth conditions.

We have studied the nucleation, annealing and growth of GaP on Si substrates. Our findings are very similar to those reported for GaAs/Si. That is, dislocation density after I μm of growth is usually about 108 cm−2 surface morphology is best when a multi-temperature growth process is used and is dependent on the substrate orientation; antiphase domain density is minimized by misorienting the substrates. This commonality of results leads us to conclude that the elimination of interfacial contamination is more important in achieving good epitaxial growth of III-V compounds on Si than is the overcoming of lattice mismatch. In support of this hypothesis we present SIMS data revealing up to 2% of interfacial carbon and TEM observations of an amorphous interfacial phase. The carbon comes from the organometallic source and, we believe, reacts with the Si to form amorphous SiC, which disrupts the coalescence of GaP grains and produces lattice defects.

GaAs shallow homojunction solar cells were grown far the specific purpose of studying the effect on the ape: CirC”it Katage Of changing the proPile of the n emitter layer. It was found that grading the emitter layer from 2.5x10'8 cm-3 at the junction to

Nonplanar morphology and compositional inhomogeneity are observed in GaAs1−xPx superlattices when the phosphorus content is substantially different in adjacent sublayers. The main characteristic feature of the distortion is a series of ribbonlike regions running through the superlattice which etch at a faster rate and contain more phosphorus than the surrounding material and are also the source of clusters of dislocations. Morphological stability and a general decrease in dislocation density can be effected by decreasing the interlayer strain, i.e., compositional difference, between the superlattice sublayers.

Superlattices of GaAs(1-x)p(x) are being incorporated into cascade solar cell structures in order to reduce the dislocation density in the top cells and thus reduce recombination loss and increase output voltage. For a superlattice to effectively block the propagation of dislocations its average composition must be equal to that of the layer beneath it (from efficiency considerations for a cascade cell, the average composition should be about GaAs(.7)P(.3). When superlattices of this approximate composition were grown on GaAs by MOCVD, severe distortion of the crystal layers was observed. The essential features of this distortion are nonplanar morphology and accelerated etching in regions containing excess phosphorus and clusters of dislocations. Similar observations have been made with superlattices grown with two other MOCVD systems, indicating that the problem is of fundamental technological significance, not just an artifact of one particular growth system. The nature of the distortion effect is described, and several strategies for preventing its occurrence are presented.

Distorted layer growth manifested by nonuniform etching may occur in GaAs1−xPx superlattices grown by MOCVD and VPE. The distortion was found to be more severe for MOCVD growth than for VPE. The distortion is decreased by decreasing the magnitude of the interlayer strain in the superlattice but does not depend upon the layer thickness. The rate of crystal growth, the temperature of growth and the strain rate in the layer supporting the superlattice also influence the distortion. Several possible causes of the effect are discussed, including nonuniform elastic stress and/or compositional nonuniformity.

Several variations of the metal-organic chemical vapor deposition process for growth of polycrystalline GaAs films have been developed. The grain size of the films ranged from less than one to several hundred microns, yet the air mass zero efficiencies of the best Schottky barrier solar cells made from films grown by each procedure were all only about 1–2%. The short-circuit current was quite high, as predicted, for large grain films but in nearly all cases where extra processing and higher temperatures were introduced in order to obtain larger grains the open-circuit voltage and fill factor were reduced. Clearly enhancement of grain size alone is insufficient to yield high quality devices, and the roles of contamination and/or process-induced defects must be thoroughly investigated in order to understand and improve the low efficiencies.

Photoluminescence at 5K is used to measure the radiative recombination in a potential solar cell material -- polycrystalline GaAs. In some samples the elecron-hole pair recombination is extremely efficient, yielding luminescence intensities up to 40% of that of monocrystalline GaAs. These samples are characterized by a peak at approximately 1.49eV, which is similar to that observed in the monocrystalline GaAs. However in other samples a broad, less intense band at approximately 1.47eV is seen.

AbstractDie Herstellung grobkörniger GaAs‐ Filme (mittlerer Korndurchmesser 0.2 mm) auf Mo zur Verwendung für Solarzellen wird beschrieben.

Films of polycrystalline GaAs have been grown on foreign substrates by the metal-organic process. The main objective was to produce films with as large a grain size as possible, so that high-efficiency photovoltaic devices may eventually be fabricated from such thin film/substrate structures. At 973 K the average grain size was less than 1 µm, and was unaffected by the choice of substrate. Increasing the deposition temperature to 1123 K, while maintaining all other conditions the same, resulted in grains as large as 10 to 20 µm in diameter. Grain sizes as large as 10 µm could be obtained by precoating the substrates with thin films of evaporated gold or tin. However, both of these methods gave films that were discontinuous. A two-step procedure in which the films were nucleated at 873 K prior to growth at 1123 K yielded continuous films with an average grain size of 5 µm. Schottky barrier solarc ells fabricated from these films exhibited short-circuit current densities as high as 15.7 mA/cm2, even though the highest conversion efficiency (AMO, uncoated) was only 1.3 percent because of the low fill factor (0.28).