Improved Diffused-region Recombination-current Pre-factor Analysis

Abstract

Photoconductance (PC) measurements of the diffused-region recombination-current pre-factor J0d make the approximation of excess-carrier Δn density uniformity. That is, from the front illuminated surface to the rear the Δn is constant. Kane and Swanson outlined acceptable sample parameters such that the error caused by this approximation is small when performing transient J0d measurements. For quasi-steady state (QSS) PC measurements, where there is generation, the uniform carrier-density approximation leads to a larger J0d under-estimation compared to the transient approach. We avoid this approximation by numerically solving the J0d accounting for depthwise generation, recombination and carrier diffusion in the quasi-neutral bulk that occurs during a QSS PC measurement. We demonstrate the application of the technique to samples with surface-diffusion sheet resistances 10–140Ω/sq., formed by etch back, and a sample where the surface passivation degrades with damp heat exposure. We find, for the samples tested, the direct application of the Kane and Swanson method: i) under-estimates the J0d by 2% to 8% when the sample bulk is lightly doped (boron doping of 1.3×1014 cm-3), and ii) under-estimates the J0d by 10% to 80% when the wafer is moderately doped (phosphorus doping of 1.5×1015 cm-3), for samples with J0d 350 to 2200 fA/cm2; compared with an experimental uncertainty of 6%. For both samples sets the under-estimation increases as J0d increases.