Significant improvements in CdTe/CdS solar cell efficiency are commonly observed as a result of a postdeposition CdCl2 dip followed by a 400 °C heat treatment during cell processing which increases CdTe grain size. In this paper, we investigate the electronic mechanisms responsible for CdCl2-induced improvement in cell performance along with possible performance-limiting defects resulting from this process in molecular-beam epitaxy-grown polycrystalline CdTe/CdS solar cells. Current density-voltage-temperature (J-V-T) analysis revealed that the CdCl2 treatment changes the dominant current transport mechanism from interface recombination/tunneling to depletion region recombination, suggesting a decrease in the density and dominance of interface states due to the CdCl2 treatment. It is shown that the change in transport mechanism is associated with (a) an increase in heterojunction barrier height from 0.56 to 0.85 eV, (b) a decrease in dark leakage current from 4.7×10−7 A/cm2 to 2.6×10−9 A/cm2 and, (c) an increase in cell Voc from 385 to 720 mV. The CdCl2 also improved the optical response of the cell. Substantial increases in the surface photovoltage and quantum efficiency accompanied by a decrease in the bias dependence of the spectral response in the CdCl2-treated structures indicate that the CdCl2 treatment improves carrier collection from the bulk as well as across the heterointerface. However, deep level transient spectroscopy measurements detected a hole trap within the CdTe depletion region of the CdCl2-treated devices at Ev + 0.64 eV which is attributed to the formation of VCd-related defects during the annealing process after the CdCl2 dip. J-V-T analysis demonstrated that this trap is the probable source of dominant recombination in the CdCl2-treated cells. An inverse correlation was found between the density of the Ev + 0.64 eV trap and cell Voc, suggesting that the heat treatment with CdCl2 may eventually limit the CdTe/CdS cell performance unless the formation of this defect complex is controlled or eliminated.

The effects of reabsorption and band-gap narrowing (BGN) on experimental photoluminescence (PL) spectra of n-InP grown by metalorganic chemical vapor deposition are analyzed. PL spectra show a pronounced widening of the main PL peak and a shift of that peak to higher photon energy with increasing doping due to band filling. However, the magnitude of these effects, both here and in earlier studies of n-type III–V semiconductors, is smaller than expected based upon band filling calculations and electrical measurements. Various explanations for these discrepancies between PL spectra and band filling calculations have been proposed, but little experimental support is currently available. In this article we demonstrate unambiguously that both the n-InP PL peak width and the peak position are significantly reduced by reabsorption, and that reabsorption completely explains the observed discrepancy between the measured PL peak width and the calculated band filling based on electrical measurements. In particular, we show that reabsorption must be accounted for when extracting the Fermi level from experimental n-InP PL spectra, otherwise the Fermi level value is severely underestimated. Since previous studies of the n-InP PL line shape have neglected reabsorption and instead attributed the unexpectedly low extracted Fermi level value to band-gap narrowing effects, we reinvestigate BGN in n-InP by considering only the low-energy tail of the PL spectra. The extent of the low-energy band tail below the intrinsic band-gap energy is observed to be only about half as large as n-InP BGN predicted theoretically. Very similar results have been reported in the literature for n-GaAs and is either due to an overestimation of the BGN by theory or a failure of PL to reflect the full extent of a highly nonrigid BGN shift. In regard to the latter, we demonstrate that a highly nonrigid BGN shift does indeed exist for n-InP, with the BGN shift near zone center being at least three times larger than the energy shift of states near the Fermi surface for n=4×1018 cm−3.

The strain relaxation mechanism and defect properties of compositionally step-graded InAs y P 1−y buffers grown by molecular beam epitaxy on InP have been investigated. InAsP layers having lattice misfits ranging from 1% to 1.4% with respect to InP, as well as subsequently grown lattice matched In 0.69 Ga 0.31 As overlayers on the metamorphic buffers were explored on both ͑100͒ and 2°offcut ͑100͒ InP substrates. The metamorphic graded buffers revealed very efficient relaxation coupled with low threading dislocation densities on the order of ͑1-2͒ ϫ 10 6 cm −2 for the range of misfit values explored here. A detailed analysis via high resolution x-ray diffraction revealed that the strain relaxed symmetrically, with equivalent numbers of ␣ and ␤ dislocations, and to greater than 90% for all cases, regardless of substrate offcut. Further analysis showed the relaxation to always be glide limited in these materials when grown on a graded buffer compared to a single step layer. The threading dislocation density was observed by plan-view transmission electron microscopy to be constant for the range of misfit values studied here in the top layer of the graded structures, which is attributed to the very efficient use of residual dislocations and the dominance of dislocation glide over nucleation in these graded anion metamorphic buffers, suggesting great promise for metamorphic devices with lattice constants greater than that of InP to be enabled by InAsP metamorphic structures on InP.

No abstract

Relaxed, high-quality, compositionally step-graded InAsyP1−y layers with an As composition of y=0.4, corresponding to a lattice mismatch of ∼1.3% were grown on InP substrates using solid-source molecular-beam epitaxy. Each layer was found to be nearly fully relaxed observed by triple axis x-ray diffraction, and plan-view transmission electron microscopy revealed an average threading dislocations of 4×106 cm−2 within the InAs0.4P0.6 cap layer. Extremely ordered crosshatch morphology was observed with very low surface roughness (3.16 nm) compared to cation-based In0.7Al0.3As/InxAl1−xAs/InP graded buffers (10.53 nm) with similar mismatch and span of lattice constants on InP. The results show that InAsyP1−y graded buffers on InP are promising candidates as virtual substrates for infrared and high-speed metamorphic III–V devices.