External quantum efficiencies η as high as 7.2% have been obtained for gallium phosphide red light emitting diodes. The p-n junctions were prepared using a p-on-n liquid phase epitaxy process in which the dopant levels significantly differ from those previously reported. Doping profiles for these junctions are compared with earlier n-on-p structures (η = 1–2%) and it is suggested that the observed high efficiencies result from (1) more efficient electron injection, (2) increased O concentration in the p-type layers which may result in a higher concentration of Zn–O complexes, and/or (3) fewer free holes in the p region contributing to nonradiative recombination.

It is shown that electron-beam generation of minority carriers at a semiconductor surface can be used to establish as an exponential minority-carrier diffusion profile Δp (z), measured along the z direction perpendicular to this surface, which is insensitive to possible variations in the surface recombination velocity. This profile can be expressed very simply, Δp (z) = C exp (−z/L), where L is the carrier diffusion length. On the basis of this result, a new measurement technique has been proposed, in which the induced junction current is measured as the electron beam is scanned along an angle-lapped surface of a p-n junction. For a small angle θ between the lapped surface and the junction, the induced current I varies exponentially with scan distance x, I = I0 exp (−x sin θ/L), and provides a direct and accurate determination of the carrier diffusion length L. This technique has been successfully utilized to measure diffusion lengths down to 0.2μ (with 10% accuracy) for GaP, and it should be useful for measuring very short diffusion lengths in other semiconductors.

High-effciency GaP green light-emitting p-n junctions have been reproducibly grown employing an overcompensation technique in a vertical dipping LPE system. In this technique b~th the n-and p-type layers are grown in a single LPE crystal growth run, eliminating potential interface problems at the junction. For junction material grown at an initial temperature of 900~ the EL quantum effciencies of encapsulated mesa diodes ranged from 0.08 to 0.14% and averaged 0.10% at 5 mA (,~5-10 A/cm2). Minority carrier diffusion lengths (L~Lh) and relative cathode-luminescence efficiency (CL) were measured in a scanning electron microscope. At excitation levels equivalent to current densities of ,~1-10 A/cm2 in a p-n junction, both Le and Z,h range from 3 to 5~ with maximum measured values of Lh = 7.2# and Le --5.6/~ for material grown from 900~ At the same excitation level the CL efficiency of the p-layers was typically found to be approximately 2-3 times that of the n-layers. Typical values for Le, Lh, CL, and the EL effciencies were found to be larger by --,50% for material grown from 900~ compared to similarly grown material from 1000~The major assets of the overcompensation technique are the high degree of reproducibility in obtaining high effciencies and that only a single growth process is necessary to form the p~ junction.

The doping dependence of the bulk efficiency for both n- and p-type GaP:N has been investigated experimentally and theoretically. Experimental data are presented for p-type Zn,N-doped, n-type Te,N-doped, and n-type S,N-doped GaP over a majority-carrier range 5×1016−2×1018 cm−3. The efficiency data and photoluminescent decay time data on the same samples are compared to a simple equilibrium model for the recombination kinetics in nitrogen-doped GaP. The model predicts that the efficiency should scale linearly with the minority-carrier lifetime, majority-carrier concentration, and the nitrogen concentration in the doping range considered. The comparison of the theoretical results with the experimental data shows that the bulk efficiency of p-type material agrees quantitatively with the analytical prediction. For Te- and S-doped material, which have widely different and varying minority-carrier lifetimes, the bulk efficiency of n-type material is shown to depend linearly upon the minority-carrier lifetime over the entire doping range considered. However, the normalized efficiency η/τmc is shown to depend upon the net donor concentration as η/τmc∝(ND-NA)0.5−0.6 above ND−NA≈3×1016 cm−3, independent of the donor, rather than linearly as predicted by theory. This deviation from theory remains unexplained. The data and analysis suggest that screening is of little importance over the doping rane considered. The lower limit on the nonradiative Auger lifetime of the bound excitons in p-type material is determined to be τxA≥(100−200)(1017cm−3/p) nsec. The strong variation of the minority-carrier lifetime with doping for n-type material is attributed to nonradiative centers extrinsic to the nitrogen center because of the different dependences observed for S- and Te-doped material and no firm conclusion can be drawn about the strength of nonradiative Auger recombination.

A series of variously doped normalGaP liquid phase epitaxy (LPE) layers were analyzed to establish the distribution of Zn,O and residual impurities contributing to the net impurity gradient previously observed in the p‐layer of high efficiency normalGaP LPE diodes. From these experiments we conclude that the net impurity gradient is primarily a consequence of a decreasing Zn concentration along the growth direction, and, to a lesser degree, an increasing residual donor concentration. The distribution of intentional and unintentional impurities was found to be independent of substrate doping level.The effect of cooling rate on the various impurity distributions was also studied, covering the range 0.5°–18°C/min. Residual impurity levels and associated gradients were significantly reduced by decreasing the cooling rate which indicates that residual impurity incorporation is kinetically controlled, probably by slow diffusion of these impurities in front of the growing LPE layer. Similarly, the O level appears to be decreased by reduced cooling rate. In contrast, Zn doping was found to be independent of cooling rate which suggests that its incorporation was essentially under equilibrium conditions. Assuming this to be the case, we have shown that the observed Zn distribution is consistent with the temperature dependence of the Zn solid solubility. Using an equilibrium description for Zn incorporation, we have computed the near‐junction Zn and O concentrations in the previously reported high efficiency diodes to be 6.7×1017 cm−3 normaland 2.7×1017 cm−3 , respectively. The corresponding level of residual impurities yields a net donor concentration of ≲3×1016 cm−3 .