Concentration-depth calibration and bombardment-induced impurity relocation in SIMS depth profiling of shallow through-oxide implantation distributions: a procedure for eliminating the matrix effect

Abstract

Secondary ion yields are known to be strongly enhanced by the presence of oxygen in the analysed sample. The magnitude of the yield enhancement is often significantly different for impurity and matrix ion species. This kind of SIMS matrix effect severely aggravates concentration calibration in depth profiling through regions of transiently varying oxygen concentration. To eliminate the matrix effect, a procedure has been developed that allows the differences in yield enhancement to be corrected in a quantitative manner. The procedure will ultimately be required to calibrate profiles extending through native surface oxide layers. The calibration exercise was carried out for boron in silicon. The dependence of the B+/Si+ sensitivity ratio, RB,Si, on the oxygen content of the sample was explored in situ by implanting 1.9 keV O2+ ions at O° (normal incidence) into a uniformly B-doped reference sample, followed by sputter profiling through the synthesized oxide with the same beam incident at 75°. All measurements were performed at base pressure. During oxygen build-up after initial sputter cleaning the Si+ and SiO+ yields increased by factors of 200 and 500, respectively, whereas for B+ the yield increased only 40 times. Almost inverse yield changes were observed during oxide removal. Bombardment-induced mixing caused a broadening of the oxide/Si interface and some relocation of B atoms. Under internally consistent assumptions the relatively small boron mixing effect could be separated from the oxygen-induced B+ yield enhancement effect. The normalized SiO+ signal ĨSiO+, was used as a measure of the oxygen content of the samples bombarded at the two different impact angles. The B+ yields and the sensitivity ratios RB,Si(ĨSiO+) could be fitted very well by polynomial functions. The polynomials were employed to quantify the depth profiles of 0.5 and 2 keV 11B implanted in Si test samples covered with 6 nm layers of thermal SiO2 (i.e. thinner than the synthesized oxide layer that can be produced by the 1.9 keV O2+ beam at O°). The compositional changes encountered in passing from the thermal oxide into the Si substrate had be taken into account, not only for time-to-depth conversion but also for concentration calibration based on the measured sensitivity ratios. The changes in erosion rate and Si density around the interface were modelled by error functions. Direct evidence is presented that, for accurate calibration, density and sensitivity changes must be treated separately. Even though the through-oxide variations of RB,Si are quite different for O° and 75°, the calibrated 2 keV 11B profiles derived from measurements at these two vastly different impact angles agree very well, even at the interface. This implies that the large matrix effect occurring in through-oxide profiling at 75° can be eliminated using the new calibration procedure. Minor differences (<10%) between the calibrated 2 keV 11B profiles from measurements at 0° and 75° can be attributed to differences in bombardment-induced relocation. The mixing effect is particularly severe for the profiles of the very narrow and shallow 0.5 keV 11B implantation distributions, which turned out to be heavily distorted at depths below 10 nm, both at 0° and 75°. Hence it is mandatory, for reasonably accurate profile measurements, to use O2+ energies that are significantly (∽50%) lower than the implantation energy, both for normal and oblique incidence of the probing beam. © 1998 John Wiley & Sons, Ltd.