Time Resolved Spectra of Raman and Thermal Emission during Pulsed Laser Heating of Silicon
- PDF / 471,116 Bytes
- 6 Pages / 420.48 x 639 pts Page_size
- 64 Downloads / 194 Views
TIME RESOLVED SPECTRA OF RAMAN AND THERMAL EMISSION DURING PULSED LASER HEATING OF SILICON
D. VON DER LINDE, G. WARTMANN, M. KEMMLER, AND ZHEN-HE ZHU* Universitgt-GHS-Essen, Fachbereich Physik, 4300 Essen 1, W. Germany
ABSTRACT Laser heating of crystalline silicon is investigated with 10 ns laser pulses at 532 ni. Raman spectra below the transition threshold show distinct shifts to low frequencies. The absence of line shifts at higher energy is due to a time resolution artifact. Temperatures evaluated from frequency resolved anti-Stokes/Stokes ratios are in agreement with the temperature estimated from line shifts, and provide clear evidence that the surface reaches the melting point. These conclusions are confirmed by independent measurements of the thermal emission. Time-resolved pyrometry also provides the temperature evolution of the liquid phase.
INTRODUCTION During the past several years the basic physical processes involved in pulsed laser annealing of semiconductors have been the subject of great interest. According to the thermal melting hypothesis laser energy deposited in a surface layer of the semiconductor leads to a rapid rise of the temperature up to the melting point followed by melting of the heated layer. During the subsequent cooling epitaxial regrowth from the substrate takes place. The thermal melting model is supported by an impressive body of experimental and theoretical work. However, several attemps were made to measure the changes of the lattice temperature of laser-heated surface layers of silicon using spontaneous Raman scattering [1-4]. These Raman experiments seemed to be in disagreement with the thermal melting model. For example, in our previous work [3,4] we measured the time evolution of spectrally integrated Stokes and anti-Stokes Raman scattering in laser heating experiments on crystalline silicon using 10 ns laser pulses. The disagreement with the melting model was based mainly on the following observations: (i) The lattice temperature inferred from the properly corrected [5,6] anti-Stokes/ Stokes ratio was always found to be far below the melting temperature; (ii) Raman scattering was observed to continue for almost 10 ns after the onset of the high reflectivity phase which, according to the melting model, cannot be Raman active. To explain these discrepancies we performed a series of time- and frequency-resolved Raman measurements. Initial results reported recently [7] indicated that the surface does approach the melting point. However, a non-uniform surface temperature had to be invoked to account for the shape of the Raman spectra, and we had to assume that solid and liquid material coexists for about 10 ns before the surface is completely molten. On the other hand, this assumption could not be reconciled with the observations that the transition to the high reflectivity phase takes only about ins to fully develop. The present report is based on additional, extensive experimental data on time-resolved Stokes and anti-Stokes Raman spectra of silicon. These results are now
Data Loading...