The noncontact ac-surface photovoltage technique is modified to enable reliable measurement of the minority diffusion length in incoming silicon PV wafers. The modifications to overcome very low SPV signal in wafers with saw damage include three elements; 1-increased photon flux in a range that still maintains advantages of low injection level; 2-elevated frequency light modulation with increased averaging that enhances signal to noise ratio and 3-thermal preconditioning of surface to create depletion layer SPV. The technique is tested using comparative measurements on Cz wafers and MC wafers with saw damage and after a saw damage removal etch. Consistently similar results were obtained before and after etching for diffusion length L, and for iron and boron-oxygen defect concentrations. The results demonstrate excellent repeatability that for a L of about 100µm is 0.1%. The repeatability data for Fe and boron-oxygen enable to evaluate defect detection limits in low e9 atoms/cm 3 range. To our knowledge, no other techniques can match such sensitivity for as-cut saw damaged wafers. 1 OUTLINE OF THE APPROACH Surface photovoltage based measurement of minority carrier diffusion length is widely used in silicon IC manufacturing for monitoring heavy metal contamination and micro-defects for evaluation of the cleanliness of crystal growth, IC processing tools and key integrated circuit processing steps. An extension of the technique to silicon PV wafers has been successful in measuring wafers after etching; after diffusion; after passivation, and also for measurements on final solar cells. In this paper, we present a version of SPV that has been developed in order to achieve a similar wafer level monitoring of minority carrier diffusion length for incoming silicon PV wafers. Emphasis is given to reliable measurements of the diffusion length, iron and boron-oxygen defects in as-cut wafers without saw damage removal and without any surface passivation. There are two recognized advantages of SPV for as-cut wafers that can make it especially attractive among carrier lifetime techniques: 1-SPV uses different light penetrations and signal ratio that eliminates the effect of front surface recombination velocity; it also uses algorithm for calculation of L that corrects the effect of back surface recombination velocity, in the case of long diffusion length comparable to or larger than the wafer thickness; 2-SPV measurements are not a subject of ambiguity caused by injection level dependences and nonlinear effects that in other carrier lifetime techniques generate fundamental questions on the true meaning of the measured parameters. The modified SPV technique is based on the most advanced version of the technique, referred to as Ultimate SPV. It uses digitally controlled illumination and signal detection. Simultaneous measurements of SPV signals, generated by light beams of different wavelengths, and different but high modulation frequencies, help to overcome the signal to noise issue. In addition, the approach introduces the best known surface preconditioning achieved with 200C annealing, that is compatible with defect monitoring used to isolate boron-oxygen defects and Fe. The approach is confirmed in a series of measurements performed on the same wafers before and after saw damage removal etch. 2 RESULTS Very good measurement stability and repeatability are demonstrated with this approach; for example, a 1 standard deviation of 0.10µm is obtained in 10 repeats for a Cz as-cut wafer with an average diffusion length of 96µm. To our knowledge such repeatability of 0.1% could not be achieved for saw damaged wafers in any other lifetime techniques. High repeatability leads in turn to precise determination of iron concentration and concentration of LID (Light-Induced Defects) from diffusion length differences after defect activation and deactivation. The measurements were performed with a modified apparatus and a special SPV procedure using a Semilab PV-2000 tool equipped with automated optical Fe activation and accelerated LID station with robotic wafer handling. The results presented include SPV diffusion length, bulk lifetime, Fe and LID concentrations for photovoltaic silicon Cz wafers and multicrystalline wafers measured as-cut with saw damage and then after typical alkaline etching for Cz wafers and acid etching for MC wafers. Very good agreement is shown for measurements before and after etching. For example, for diffusion lengths from 50µm to 170µm, the values before and after etching differ by less than 5% for wafer average and by only about 10% for individual sites. For Fe measurements in the typical range from mid 1E10cm-3 to 1E12 cm-3 the relative differences were about twice larger than that for diffusion length. This is consistent with the differential character of Fe measurements. LID results were similar to those for Fe. 3 CONCLUSIONS Present results prove that the modified SPV diffusion length method provides a reliable means for contamination measurement on incoming as-cut silicon PV wafers. A previous study has already shown very effective application of the SPV method to measurements on wafers after diffusion, passivation, and on the final solar cells [1]. This, combined with the new capability of measuring as-cut and etched wafers, opens a means for 26th European Photovoltaic Solar Energy Conference and Exhibition 979
