Evolution of dopants and defects in silicon under various annealing sequences

The activation and diffusion in self-amorphizing (arsenic) and non-amorphizing (boron) implants into silicon has been investigated in a discrete time survey from nanoseconds to seconds. Rs results for the arsenic implants show that using a wide range of heat-up rates from 10<;sup>;2ο<;/sup>;C/s (type C anneal) to 10<;sup>;9ο<;/sup>;C/s (type A anneal) display the same behavior of activation driven primarily by SPER. Defect free re-crystallization and a stable Rs of ~500Ω/Π is observed within 20-60 μs accumulated exposure near 1100<;sup>;ο<;/sup>;C with negligible diffusion. De-activation is observed when accumulating an exposure longer than 10 ms at 900°C with a type B anneal, and the mechanism requires further investigation. For a non-amorphizing boron implant, data indicates that even with 10<;sup>;ο<;/sup>;C/s temperature ramp-up rates of a type A anneal, the presence of an ultra-fast boron diffusion pulse with a mean diffusion enhancement of 10 is observed within 60 μs near 1000<;sup>;ο<;/sup>;C. However, a type A adiabatic anneal sequence does provide significant activation advantages over equivalent peak temperature type B thermal flux anneals. Rs reaches a stable value within ~100μs when accumulated using a type A anneal sequence. Approximately 200°C higher peak temperature and ~48x longer exposure duration within similar Tpeak-50°C is needed with a type B anneal to achieve the same Rs. Also with type B anneal sequence, the Rs does not stabilize but monotonically decreases in time with diffusion enhancements of 100. Although this work uses only Rs and SIMS results without quantitative defect analysis to interpret the observed trends, it is clear that performing the fastest heat-up (type A before type B before type C), directly after a non-amorphizing implant minimizes trapping of dopants in defect clusters and thus permits high and stable activation in durations of <;100μs. Permitting only 15% of the first 100 μs of a thermal sequence to dwell within Tpeak 50°C (as in type A) is far more effective at defect free activation than it is to first heat up "slowly" over a millisecond and then spend as much as 50% of the exposure duration within Tpeak 50°. In conclusion, Sedgwick remarked nearly three decades ago that extrapolation of the temperature dependence of dislocation loop removal rates to very short times may not be valid because it is based on defects "that are first formed and then dissolved" and that in the future, short time annealing may produce different results.