Masami Morooka

Indiffusion and Out-Diffusion Profiles of Substitutional Au in Si Affected by Self-Interstitials and Vacancies

Precise indiffusion and out-diffusion profiles of substitutional Au in Si have been investigated by a numerical method for the nonlinear diffusion equation affected by self-interstitials and vacancies simultaneously. In contrast to usual approximate solutions, a small contribution of self-interstitials in the indiffusion results in a profile specific to a kick-out mechanism, and a small contribution of vacancies in the out-diffusion results in that to a dissociative mechanism. Therefore, both experimental profiles of indiffusion and out-diffusion are required to determine the contribution rate of these point defects. The experimental results indicate that the contribution rate of vacancies for the effective diffusion of Au is about 90% at 900°C, which is very different from previous result.

Annealing of supersaturated low-temperature substitutional gold in silicon

An annealing experiment of supersaturated low-temperature substitutional gold in silicon was made by an ordinary annealing method. The gold concentration was determined from the resistivity by a four-point probe method, the Hall coefficient and DLTS with no appreciable difference in the results. The dependences of the annealing characteristics upon the temperature and the initial value of the gold concentration were examined. The annealing was represented by a firstorder reaction and the time constant decreased with an increase in the initial value. From these results, it was concluded that homogeneous nucleation occured during the annealing and that a strain energy of low-temperature substitutional gold decreased with nucleus formation.

Essential difference in concentration profile of Au in Si after annealing above or below 850°C

Indiffusion and out-diffusion profiles of Au in Si have been measured after a heat treatment at a higher or lower temperature than 850°C. The diffusion is slow and the profiles show a typical kick-out diffusion, in both heat treatments for the indiffusion above and below 850°C, and for the out-diffusion of Au supersaturated above 850°C. On the other hand, the out-diffusion of Au supersaturated below 850°C is very fast and the profile shows that it is flat. The essential difference in the concentration profiles are consistent with the change of the Au configuration supersaturated below 850°C from high-temperature substitutional Au to a low-temperature one.

Limiting process for gold in-diffusion in silicon with and without extended defects

Gold was in-diffused at 1150° C into silicon containing extended defects (dislocations or stacking faults) and into defect-free silicon single crystals, and the diffusion-limiting process for the gold in-diffusion was investigated. Gold diffusion in the crystals without defects and that very near the specimen surface in crystals with defects are limited by the diffusion of self-interstitials to the surface. Diffusion in the crystals with defects is limited by the diffusion of interstitial gold atoms from the surface. In crystals with an extremely low concentration of extended defects, the diffusion is limited by the diffusion of self-interstitials to the defects.

Stacking Fault Induced by Gold Diffusion in Silicon

Gold was diffused into several silicon single crystals at 1150°C and hexagonal stacking faults were induced in some crystals by the gold diffusion. The relationship of the stacking fault formation to the gold diffusion was investigated. The stacking faults were formed in a region greater than about 100 µm from the specimen surface and increased in the inner part to a constant value greater than 300 µm. The stacking fault size and number increased with the substitutional gold concentration but only their number decreased with the decrease of the gold concentration. The number of point defects contributing to the growth of stacking faults was nearly equal to that of substitutional gold atoms diffused into silicon, and the stacking fault contained one atomic layer of point defects. The reaction between the nucleus of stacking faults and the point defects was not uniformly distributed, but was instead localized.

Effect of Polishing on Time Dependence of Average Substitutional Impurity Concentration in Dissociative and Kick-Out Mechanisms

From the standpoint of experimentally distinguishing the dissociative mechanism from the kick-out mechanism, the time dependence of the substitutional impurity concentration in the diffusion is useful. In a dislocation-free sample, the reaction of point defects with sample surfaces is the limiting process, and the impurity concentration may depend on distance from the surface. In the experiment, the surface is polished before the concentration measurement. Therefore it is important to know the effect of amount of polishing on the time dependence of the average substitutional impurity concentration. In this paper, the effects for both in-diffusion and annealing processes are calculated. It is concluded that, when a dislocation-free sample is used, the effect is strong for both in-diffusion and annealing processes in terms of the dissociative mechanism and for the annealing process in terms of the kick-out mechanism.

Effective Diffusion Coefficient and Controlling Process of P Diffusion in Si Based on the Pair Diffusion Models of Vacancy and Interstitial Mechanisms

Based on the pair diffusion models of vacancy and interstitial (V and I) mechanisms, the V and I components of effective P diffusion coefficient, DP+,Veff and DP+,Ieff, and the controlling process of P diffusion in Si are obtained. Assuming that the I mechanism is dominant, not only the I- concentration, CI- , but also its gradient, d CI- /d λ, is effective on DP+,Ieff at high CP+ . DP+,Ieff is large at d CI- /d λ 0. P+ and I- are generated by the dissociation of P–I pair. When excess I- thus generated is removed, d CI- /d λ 1×1020 cm-3 (s: surface).

Brief measurement of diffusion profiles of deep impurities by moving Schottky contact

Long-range diffusion profiles of deep impurities, such as gold in silicon are usually obtained by repeating capacitance measurements after each removal of a considerable surface layer followed by fabrication of a Schottky junction on the surface. However, this method is time-consuming. By moving the Schottky electrode on an angle-lapped surface, the profiles have been obtained in less than one-tenth of the time required using the usual method.

Concentration Profiles of Deep Levels Induced by Gold Diffusion in Silicon

The concentration profiles of the deep levels (E c-0.54 eV, E c-0.24 eV, E c-0.30 eV and E c-0.16 eV) induced by gold diffusion at 1150° C for 6 h were measured by the deep level transient spectroscopy (DLTS) method. The profile of E c-0.54 eV agreed well with that of total gold atoms measured by neutron activation analysis. The shape of the profile of E c-0.24 eV is very similar to that of E c-0.54 eV, and the concentration ratio of E c-0.24 eV to E c-0.54 eV is nearly constant at 0.12 from the surface to the inner part. Whether the level of E c-0.16 eV is observed or not, depends on the fabrication of the Schottky diode for the DLTS measurement, and the shape of the profile of the observed E c-0.16 eV is similar to that of E c-0.54 eV. The profile of E c-0.30 eV is quite different from those of other levels, and the level concentration is nearly constant from the surface to the inner part.

Temperature Dependence of ESR Lines Related to Phosphorus in Silicon

Temperature dependence of ESR in a silicon crystal containing 1.3×1017 phosphorus atoms/cm3 has been investigated at 4–40 K. The ESR signals depend strongly on the specimen temperature. Typical hyperfine structures of the phosphorus ground state were found in the low-temperature region. The hyperfine lines vanished at 20 K and a single line corresponding to the first excited state of phosphorus was observed in the high-temperature region. The amplitudes of the hyperfine lines and the single line show sharp maxima at 11 K and 30 K, respectively. The change of the signal amplitudes is due to the temperature dependence of spin relaxation time. The change of the linewidth for the hyperfine lines can be explained by the temperature dependence of spin relaxation time, and that for the single line is considered to be caused by motional narrowing.