Detailed investigation of surface passivation methods for lifetime measurements on silicon wafers

摘要

The effect of five different common surface passivation techniques on the measured bulk lifetime values of multi- and mono-crystalline silicon wafers is investigated. Mono Cz and neighboring multicrystalline silicon wafers are deposited with a passivating layer of SiNx, Al2O3, or amorphous silicon (a-Si) or are passivated chemically with 0.08 M iodine-ethanol (IE) or 0.07 M quinhydrone-methanol (QM) solutions. Two Cz wafers are annealed at 200°C for 15 minutes and passivated with QM and IE, respectively. Time resolved lifetime measurements show respectively stable and steadily degrading passivation for the QM and IE samples. An EFG wafer is passivated with QM, measured, and rinsed six times without etching a new wafer surface between passivations. μPCD lifetime measurements show a steady degradation of lifetime values with each passivation repetition, suggesting the presence of quinhydrone residues on the wafer even after post-passivation rinsing. Six multicrystalline wafers are passivated by the dielectrics Al2O3, a-Si, and SiNx and then annealed. Lifetime values measured 13 days after the annealing step are lower than those measured directly after the annealing, which is attributed to surface passivation degradation. After a second annealing, the a-Si samples seem to recover and show lifetime values close to the initial values. Four Cz wafers are dielectrically passivated with Al2O3 or a-Si and then annealed. Time-resolved lifetime measurements reveal a rapid lifetime value drop within 30 minutes after the annealing. Possible causes of the drop include surface passivation degradation and/or boron-oxygen complex formation. Neighboring multicrystalline wafers are passivated dielectrically or with IE or QM and characterized with photoluminescence imaging. All mc wafers subjected to dielectric passivation methods that include high-temperature annealing (300-400&x- 0B;0;C) display a greater area of high lifetime values but fewer areas of very high lifetime values, providing visible evidence of internal gettering and/or defect redistribution.