The diffusion of phosphorus from a deposited doped oxide into silicon has been found to be consistent with a previously derived model describing doped oxides as diffusion sources. Diffusion coefficients of phosphorus in both silicon and silicon dioxide have been measured in terms of this model over surface concentrations varying from 6×1017 normalto 2×1020 normalatoms/cm3 at diffusion temperatures of 1000°, 1100°, and 1200°C.

The results of diffusion experiments carried out in a tellurium atmosphere suggest that nickel reacts with tellurium vapor forming a second phase at the surface. The effects on the distribution of nickel in PbTe on exposure to a tellurium atmosphere confirm this result as do diffusion studies in the absence of a tellurium atmosphere. These results also suggest an explanation for differences observed in the diffusion of nickel into PbS at low' temperatures (8) in an inert atmosphere and those obtained in the 700~ region under sulfur atmosphere (9). It is also suggested that the diffusing nickel species is different in a tellurium atmosphere than under inert or reducing conditions.The implications of these results on general doping experiments and practice in the lead salts are also of interest. If other impurities in the lead salts behave to change in the crystal stoichiometry as does nickel, methods of doping the lead salts must be reinvestigated. Changes in the crystal composition brought about by changes in the atmosphere above the crystal could induce changes in an impurity distribution which will not return to equilibrium in the time normally assumed reasonable to properly equilibrate crystals. The result will be a nonuniform doping concentration which will obviously affect other measurements.The analysis of the chlorine diffusion data suggest that migration on the anion sublattice occurs via a Frenkel defect, anion vacancies being responsible for motion in lead rich PbTe and anion interstitials in tellurium rich material.

No abstract

In a recent article (1) we presented and verified a model describing the diffusion from a deposited doped oxide into a semiconductor substrate. Subsequently, in related work dealing with diffusion into polycrystalline silicon, we found additional experimental evidence verifying this model. In short, we find that junctions after diffusion from a deposited oxide are deeper in polycrystalline silicon than in single-crystal silicon because the diffusivity of the dopant is higher in the polycrystalline material. However, the sheet conductivities are the same for both types of silicon since the diffusion process is limited by diffusion within the oxide source. Figure 1 shows a general one-dimensional case for diffusion of a dopant from a uniformly doped oxide into a semiconductor substrate. In the work to be described, the layer of undoped barrier oxide has been reduced to a minimum and may be disregarded. We have shown previously that for the case where the deposited oxide is thick compared to the diffusion length in the oxide, the concentration of dopant in the semiconductor after diffusion is given by C2(x,t) = Cs erfc (x/2%/D2t) + (llm)x/DtlDiwhere Co is the initial concentration of dopant in the oxide, D1 and D2 are the diffusivities of the dopant in the oxide and in the semiconductor, respectively, and m, the segregation coefficient, is the ratio of the surface concentration of the dopant in the semiconductor to the surface concentration in the oxide. If the substrate is silicon with a reasonably high resistivity (about 2 ohm-cm for most cases) of opposite type from the diffusing dopant, Eq.[1] and [2] can be manipulated to * Electrochemical Society Active Member. I Doped OxideCl, DI -X 0 Undoped ~-Oxide C3,DI • -X B 0 X .-.--~. Silicon Cz, D2 yield xj = 2N/D2t argerfc (CB/Cs) [3] and I/V = 8.15 X 10-23 ~-Cs ~/D2t _ Co ~/DI~ [4] = 8.15 X 10-23 ~ -5 (I/m)A/D,/D2where xj is the junction depth resulting from the diffusion, I/V is the sheet conductivity as measured with a standard 4-point probe, CB is the bulk concentration of carriers in the substrate, ~-is the effective mean mobility of the carriers, and the constant, 8.15 x 10 -23, contains the electronic charge and the spreadingcurrent factor for the 4-point probe. Since argerfc (CB/Cs) is only a weak function ofCs, the junction depth for a given diffusion time depends primarily on the diffusivity of the dopant m the semiconductor; on the other hand, if (l/m) x/D1/D2 is small compared to 1 (and this appears to be the case for both boron and phosphorus in silicon), the I/V will depend primarily on the dittusivity of the dopant in the oxide. The same analysis applied to Eq. [2] indicates the surface concentrations will vary inversely with the diffusivities in the silicon. Equations [2], [3], and [4] then predict that if diffusions are made from the same deposited oxide into silicon substrates in which the diffusivities of the dopant vary widely, the resulting junction depths and surface concentrations will reflect this variation while the sheet conductivities will ...

No abstract