The fundamentals of energy transfer in semiconductors during and following pulsed laser excitation are reviewed within the frame of results obtained in picosecond laser experiments. This paper discusses the theoretical processes as well as the experimental evidence, which sets limits of electron and lattice temperature of irradiated silicon. Simple thermal melting is the most probable explanation for the laser induced phase transitions observed.

Synchrotron x-ray pulses have been used to make nanosecond resolution time-resolved x-ray diffraction measurements on silicon during pulsed laser annealing. Thermal expansion analysis of near-surface strains during annealing has provided depth dependent temperature profiles indicating >1100°C temperatures and diffraction from boron implanted silicon has shown evidence for near-surface melting. These results are in qualitative agreement with the thermal melting model of laser annealing.

The lattice temperature of silicon was measured during pulsed ruby laser annealing (-20 ns pulse, spot diameter ≥ 5 mm) using a classical time of flight method to determine the velocity distribution of Si atoms evaporated from the hot surface. The maximum of this Maxwell type distribution was used to calculate the temperature of the Si surface. The resulting lattice temperatures vary between 1200 and 2500 K for energy densities between 1.0 and 2.0 J/cm2 , i.e., Si is molten for energy densities ≥ 1.4 J/cm2 . This result clearly supports the strictly thermal annealing model [1] and contradicts the non-thermal-equilibrium model [2] as well as Raman measurements [3].

Emission of charged particles from (111) and (100) crystalline silicon surfaces during and following picosecond pulsed laser irradiation in vacuo has been investigated.

No thermionic electron emission is observable, setting an upper limit of 5000°K on the electron temperature during the laser pulses at 532 nm and at 266 nm. Equal number of positive and negative particles are emitted when the laser energy fluence is sufficient to cause vaporization of a few surface layers. Significantly larger amount of electrons than that of positive particles are emitted under irradiation with UV pulses at low energy fluences. This phenomenon can be tentatively explained by thermally enhanced photoelectric emission from a molten silicon surface.

The presence or absence of the emission of charged particles sets important lower or upper limits on the temperature of the electrons and of the lattice. Our data are fully consistent with a model of complete thermalization between carriers and lattice on a time scale of 10–11 sec.

The changes in reflectivity of a silicon surface, irradiated by a green picosecond pulse, are probed during and following that pulse with a spatial resolution of 10μm. The data indicate the development of a liquid phase, and a resolidification either into a single crystal or an amorphous phase. The latter has a characteristic ring-type pattern, and occurs only at locations where the incident picosecond laser fluence lies between 0.2 and 0.26 J/cm2. The reflectivity data appear to be in good quantitative agreement with a “simple heating” model, in which the electrons and phonons maintain a local thermodynamic equilibrium on a picosecond time scale.

Raman measurements of temperature reported earlier have been repeated using a doubled Nd: YAG pulse for excitation and an electronically delayed dye laser pulse. These results, together with a variety of experimental tests of the Raman method, confirm the validity of the small temperature rise during pulsed laser annealing. Transmission measurements spanning the visible and near IR show that there exists a thin (∼ 70 nm) layer at the surface in which the induced absorption coefficient is ∼ 7 × 105 cm−1.

Until recently two theories of what may be occurring during pulsed laser/electron/ion beam annealing of Si, Ge, GaAs, and other semiconductors enjoyed credence. One held that the material became hot and underwent normal thermal melting. The other held that the material annealed in the presence of a hot dense plasma of excited free carriers, electrons and holes. A closer examination of previous experiments, as well as new experiments, particularly those of Marquardt et al., which indicate an absence of free carrier absorption in the infrared, make both these theories untenable. They also make possible a more sophisticated understanding of what really is happening. The present status of this subject is reviewed. In particular the available evidence that the plasma excited by the pulsed radiation undergoes a boson condensation to a superfluid phase is described.

Pulsed Raman temperature measurements by Lo and Compaan on Si samples have been interpreted as proving that the surface region does not melt during intense pulsed laser irradiation. In this paper, it is shown by detailed calculations with the melting model that the choice of experimental parameters in the Raman measurements can severely compromise a straightforward interpretation of the data. Moreover, it is demonstrated that temperatures extracted from Raman measurements are highly sensitive to the temperature-dependent optical properties of the material. Finally, it is pointed out that the very large temperature gradients present during pulsed laser annealing may entirely invalidate the Stokes/anti-Stokes ratio as an accurate temperature probe.

Compaan and co-workers have reported the results of time-resolved optical experiments on ion-implanted silicon which they claim prove the melting model of pulsed laser annealing cannot be correct. These results concern the rapid onset of a Raman signal after a heating laser pulse, the simultaneous occurrence of a Raman signal and the high reflectivity phase characteristic of molten silicon, and the lattice temperature measured by the Raman Stokes/anti-Stokes intensity ratio. In this paper, we show by detailed numerical calculations with the melting model that there is, in fact, excellent agreement between the results of the calculations and the experimental results reported by Compaan and co-workers.

The time resolved optical transmission, T (atλ = 1152 nm), and reflectivity, R (at 633 nm and 1152 nm), have been measured for n-type single crystalline silicon (c-Si) during and immediately after pulsed ruby laser irradiation (λ = 693 nm, FWHM pulse duration 14 nsec), for a range of pulsed laser energy densities, El. The T is found to go to zero, and to remain at zero, for a period of time that increases with increasing El, in apparent disagreement with earlier measurements elsewhere that used semi-insulating Si and a different pulsed laser wavelength. Measured reflectivities during the high R phase agree within experimental error with reflectivities calculated from the optical constants of molten Si. Quantitative agreement is also found between both our T and R measurements and detailed time– and El-dependent results of thermal melting model calculations.

We report measurements of reflectivity and transmission of probe beams at 0.633, 1.06, 1.34, and 3.39 μm during pulsed laser annealing of silicon. The transient infrared optical response of the silicon was monitored with several fast photodetectors. The diversity of detector response to transmitted light indicates that contradictory results obtained previously may be artifacts of photodetector response.

We report simultaneous measurements of time resolved reflection and transmission of low intensity 1.06 μm, 35 ps pulses subsequent to excitation of 50 KeV, 1016 cm−2 boron implanted silicon by 0.53 μm 35 ps pulses of varying energy densities. The samples are examined by optical and scanning electron microscopy in conjunction with defect etching. These data are discussed from the point of view of both the thermal melting model and plasma model.

Time resolved transmission (λ = 1060 nm) and reflectivity (λ = 632.8 nm) during 20 nsec Nd:glass laser annealing in ion implanted Si were measured to study dynamic behaviors of laser annealing. Transmitted laser energy was also measured to complement transient measurements. Transient transmitted light intensity was found to be almost completely quenched as would be expected from absorption by a molten Si layer.

Spectra of photoelectrons and thermionic electrons emitted from silicon during pulsed laser irradiation at energy densities encompassing the thresholds for laser annealing and damage are reported. Annealing is accomplished with a 90-nsec pulse of 532-nm light, which may be accompanied by a 266-nm probe pulse. A cylindrical mirror analyzer is used for energy resolution of emitted electrons. Time-resolved reflectivity at 633 nm verifies attainment of the high-reflectivity annealing phase. Spectral widths and total yields imply a modest electron temperature (T < 3000 K) during annealing. The data suggest that the work function of the silicon (111) face may increase about 0.6 eV upon transition to the molten phase.

Just below threshold for irreversible macrocrystallization there may exist a narrow energy (or power) density region where microcrystallites are formed. In the case of Si and GaAs this region lies just above the threshold for melting and the microcrystallites are formed irreversibly. In the case of GeSe2 when Urbach tail laser energy is used just below the macrocrystallization threshold the photons selectively replace like-atom bonds with chemically ordered bonds, and microcrystallization can be thermally reversed in melt-quenched glass at temperatures as low as Tg/2.

Solidification of the melted region during pulsed laser annealing occurs under conditions that are far from equilibrium at the interface. Studies of laser annealing of ion implanted silicon can be used to provide a wealth of information on nonequilibrium crystal growth phenomena.

Laser melting has been used to controllably vary the Si solidification velocity in the range 1–20 m/sec. The segregation of implanted impurities is found to be critically dependent on the liquid-solid interface velocity and substrate orientation for velocities <10 m/sec. This behavior can be understood in terms of different degrees of undercooling of the melt. While the (100) epitaxy is generally excellent up to velocities ∼10 m/sec, twins are observed for (111) epitaxy in the range ∼5–10 m/sec. Amorphous Si is produced from the melt for velocities in the vicinity of 20 m/sec. The amorphous phase forms at lower velocities on the (111) interface than on the (100) interface. These estimates of interface velocities come from heat flow calculations which do not include undercooling of the melt. Undercooling does not affect interface velocities ∼3 m/sec but significant lowering of the higher velocities could result from such effects.

Short (2.5nsec) pulses of ultra-violet radiation from a Q-switched laser system have been used to induce a range of defect transitions in the surface regions of (001) and (111) Si single crystal by transient melting and resolidification. Relatively large areas of Si have been uniformly processed and this has enabled the measurement of thresholds for both amorphization and extended defect production. The highest quench rates (isotherm velocities of ∼20m/sec) were achieved at the surface melting threshold near 0.2J/cm2 and both (001) and (111) Si could be amorphized with radiation energy densities close to this value. With increasing energy density the quench rate fell and (001) Si ceased to amorphize before (111) Si. Furthermore, over a range of high radiation fluxes the crystalline Si produced on (111) surfaces was highly defective and contained twins due to errors in liquid-phase epitaxy. The various observed structure transitions have been related to the predictions of crystal growth theory with account taken of melt undercooling effects. The amorphous Si produced during this work (up to almost 1000Å thickness) has been shown to be structurally similar to that produced by conventional methods. Both direct analysis and solid-phase regrowth experiments have demonstrated that its impurity content is negligible.

The work on annealing of displacement damage, dissolution of boron precipitates, dopant redistribution and formation of constitutional supercooling cells using pulsed ruby and dye laser irradiation is reviewed in order to provide convincing evidence for melting as a primary mechanism of laser annealing. The nature of the liquid-solid interface and the interfacial instability during laser-induced rapid crystal growth are considered in detail. Solute concentrations after laser annealing can far exceed retrograde solubility limits, but there is a critical concentration above which a planar liquid-solid interface becomes unstable and breaks into cellular structures. The solute concentrations and cell sizes associated with this instability have been studied as a function of crystal-growth velocity and these results compared with calculations of the perturbation theory. Good agreement between the experimental results and the theory was obtained when the dependence of the distribution coefficient upon crystal growth velocity was taken into account in the calculations.

Heat flow calculations, based on the main assumption that the energy of the incident pulse is instantaneously and locally converted into heat, are used to review the results of laser pulse annealing and impurities behaviour in ion implanted semiconductors. The crystallization behaviour of amorphous layers, the velocity of the solid-liquid interface and impurity redistribution are detailed with the main emphasis on the relevance of the parameters (laser wavelen gth, pulse duration and energy density, substrate temperature, etc.) that can be experimentally controlled. Comparison with the latest experimental results is also given.