Design principles for photovoltaic devices based on Si nanowires with axial or radial p-n junctions

Semiconductor nanowires (NWs) are a developing platform for electronic and photonic technologies, and many demonstrated devices utilize a p-type/n-type (p-n) junction encoded along either the axial or radial directions of the wires. These miniaturized junctions enable a diverse range of functions, from sensors to solar cells, yet the physics of the devices has not been thoroughly evaluated. Here, we present finite-element modeling of axial and radial Si NW p-n junctions with total diameters of ~240 nm and donor/acceptor doping levels ranging from 10(16) to 10(20) cm(-3). We evaluate the photovoltaic performance of horizontally oriented NWs under 1 sun illumination and compare simulated current-voltage data to experimental measurements, permitting detailed analysis of NW performance, limitations, and prospect as a technology for solar energy conversion. Although high surface-to-volume ratios are cited as detrimental to NW performance, radial p-n junctions are surprisingly insensitive to surface recombination, with devices supporting open-circuit voltages (V(OC)) of ~0.54 V and internal quantum efficiencies of 95% even with high surface recombination velocities (SRVs) of 10(5) cm/s. Axial devices, in which the depletion region is exposed to the surface, are far more sensitive to SRV, requiring substantially lower values of 10(3)-10(4) cm/s to produce the same level of performance. For low values of the SRV (<100 cm/s), both axial and radial NWs can support V(OC) values of >0.70 V if the bulk minority carrier lifetime is 1 μs or greater. Experimental measurements on NWs grown by a vapor-liquid-solid mechanism yield V(OC) of 0.23 and 0.44 V for axial and radial NWs, respectively, and show that axial devices are limited by a SRV of ~7 × 10(3) cm/s while radial devices are limited by a bulk lifetime of ~3 ns. The simulations show that with further development the electrical characteristics of 200-300 nm Si NWs are sufficient to support power-conversion efficiencies of 15-25%. The analysis presented here can be generalized to other semiconductor homo- and heterojunctions, and we expect that insights from finite element modeling will serve as a powerful method to guide the design of advanced nanoscale structures.