تجزیه و تحلیل حساسیت از جدایی دو طیف از سطح و اجزاء عمده ای از طول عمر اقلیت حامل

Performing quasi-steady-state lifetime measurements using two different illuminating spectra provides quantitative information about bulk lifetime (τb) and surface recombination velocity (S). This paper motivates the investigation of this relatively new method by demonstrating that the conventional method of iodine/methanol passivation for the extraction of τb, which is then used to calculate S for a dielectric, may fail for solar-grade materials such as string ribbon silicon. To facilitate the use of the two-spectrum method, first we introduce a novel empirical procedure for the determination of the constant of proportionality between the short-circuit current of the reference cell and the average generation rate (Gav) in the test wafer. Then a sensitivity analysis is performed to show that the method of using a white light spectrum and an infrared spectrum to obtain information about τb and S also has serious limitations in certain cases: only a lower bound can be placed on τb for τb greater than about 10 μs, and only an upper bound can be placed on S for S less than about 1000 cm/s. Our analysis demonstrates that in order to use the two-spectrum method to specify τb and S within a factor of about 2–20 when experimental uncertainty is ±10%, the quality of both the bulk of the material and the surface passivation must be somewhat poor. Precision may be improved by reducing experimental uncertainty. To illustrate the requirement that bulk and surface recombination must be high in order to use the two-spectrum method with the greatest precision, the method was applied to nitride-passivated float zone and cast multicrystalline silicon wafers of different resistivity. Only an upper limit to S (165 cm/s) was inferred for the easily passivated float zone wafer, whereas both upper and lower limits to S were extracted for the less effectively passivated heat-exchanger method (HEM) multicrystalline wafers. The analysis yielded 1200<S<4200 cm/s for the 1.4 Ω cm HEM wafer and 3000<S<20000 cm/s for the 0.2 Ω cm wafer after the nitride was annealed at 850 °C. The 0.2 Ω cm HEM wafer was also measured before the nitride was annealed. The two-spectrum method provided a τb range that remained nearly unchanged, while the S range was much higher for the as-grown SiNx. This indicates that the 850 °C anneal improves surface passivation without passivating the bulk of the HEM material.

Silicon solar cells are frequently coated with a dielectric film such as SiNx to reduce reflection, to hydrogenate bulk defects, and to passivate the surface. There is often a need to determine the resulting bulk lifetime (τb) and surface recombination velocity (S). While the effective lifetime (τeff) determined by a photoconductance measurement is influenced by both of these parameters, it is important to be able to separate their effects in order to deduce their values. If τb is known, S can be calculated from the measured τeff [1], which is always less than τb (1/τeff=1/τb+2S/W for low S). One method for the determination of bulk recombination is the application of an extremely effective surface passivation, such as corona discharge [2] or immersion in an iodine/methanol solution [3], which causes S to be negligibly small for high-quality wafers. With surface recombination practically eliminated, a measurement of τeff yields τb. While the use of iodine has been shown to be convenient, reliable, and effective for reducing S below 5 cm/s on monocrystalline materials [3], it is not obvious that this method should be expected to work on low-quality, solar-grade materials such as string ribbon silicon. The method would clearly fail if τeff measured when the wafer is passivated by a dielectric is greater than τeff measured when the wafer is passivated by iodine. If we identified the latter with τb, then calculated S would be less than zero for the dielectric passivation, which is meaningless. The possible failure of the iodine passivation method motivates the search for new methods for the separation of τb and S. Two recent methods have been suggested for the separation of surface and bulk recombination by two different photoconductance measurements of the same wafer; no temporary surface passivation is required. Nagel et al. [4] showed theoretically that for S>1000 cm/s and s, S and τb can be extracted from two 400 nm illumination τeff measurements: one transient (in which generation can be approximated as an impulse function) and one steady state. Since a transient measurement can only be made when the light pulse is much less than τeff, and since τeff is much less than 10 μs under the given conditions, a light pulse of duration much less than 10 μs must be used to make the transient measurement. This precludes the use of the popular laboratory flash lamp whose decay constant ranges from 18 μs to 2.3 ms. The second method, described by Bail and Brendel [5], uses two quasi-steady-state photoconductance (QSSPC) measurements performed under two illumination spectra: blue and infrared (IR). Based on an exact calculation of the generation profiles and the excess carrier distributions throughout the wafer, they determine which values of S and τb yield the measured photoconductances under blue and IR illumination. While their method is meticulously precise, it requires knowledge of the spectral dependence of the following: the photon flux, the external quantum efficiency (EQE) of the reference cell used to measure the photon flux, and the front and rear reflectance of the test wafer. This paper extends Bail and Brendel's idea as follows: a calibration procedure is proposed to circumvent the need for knowledge of photon flux spectral density and reference cell EQE; the steady-state equations linking S, τb, and τeff are elucidated; and the sensitivity of this method for various combinations of S and τb is assessed. Our sensitivity analysis evaluates for the first time how the uncertainty in the extracted values of S and τb depends on their true values. These results are validated by applying the technique to a high-quality float zone wafer and two lower-quality multicrystalline wafers.

While confirming that the conventional method of iodine passivation for the extraction of τb can reduce S below 2 cm/s on float zone, we demonstrate that the same iodine solution can fail to effectively passivate string ribbon silicon. Our sensitivity analysis has shown that the method of using a white light spectrum and an IR spectrum to obtain information about τb and S has serious limitations: only a lower bound can be placed on τb for τb greater than about 10 μs, and only an upper bound can be placed on S for S less than about 1000 cm/s. Thus, both τb and S must be somewhat poor in order for their values to be extracted within about an order of magnitude. Fortunately, when τb is too good to be extracted with precision using the two-spectrum method, it can be well determined by the conventional iodine passivation method. The two-spectrum method was applied to float zone and HEM wafers of different resistivity. Only an upper limit to S (165 cm/s) was inferred for the easily passivated float zone wafer, as predicted by our sensitivity analysis. Also consistent with the sensitivity analysis is the fact that both upper and lower limits were inferred for S on the less effectively passivated HEM wafers: 1200<S<4200 cm/s for the 1.4 Ω cm wafer, and 3000<S<20000 cm/s for the 0.2 Ω cm wafer. The 0.2 Ω cm HEM wafer was measured before and after the SiNx was annealed. As expected, the two-spectra method provided a τb range that remained unchanged, while the S range was higher for the as-grown SiNx, by a factor of about 3.