THE ACCURATE MEASUREMENT OF AMORPHOUS CONTENT

James P. Cline¹, Robert B. Von Dreele² and Brian H. Toby¹

¹National Institute of Standards and Technology , Gaithersburg, MD 20899, USA
²LANSCE, Los Alamos, NM 87545, USA

The measurement of phase abundance by powder diffraction techniques may accomplished by either the Reference Intensity Ratio, RIR, method or, with greater accuracy and precision by Quantitative Rietveld Analysis, QRA. However, in order to determine the amorphous content, both methods require that a specific amount of an internal standard, of known purity, be added to the unknown and that the data obtained from the diffraction trace be unbiased. The amorphous content is then determined by comparing the diffraction analysis results, which do not include the amorphous phase, with the relative mass of the standard and unknown in the prepared specimen, which does. We discuss systematic bias of various powder diffraction methods and the measurement (certification) of the amorphous content of a Standard Reference Material, SRM, 676. SRM 676 is a high purity, non-orienting, fine grained alumina powder certified for quantitative analysis.

The experimental approach was based on the comparison of the phase abundance of mixtures of silicon and SRM 676. The silicon was crushed and jet milled from Czrochralski grown boules. The resulting silicon powder was fractionated into four lots of varying particle size and then annealed. The surface area of each lot was measured via BET adsorption. A total of sixteen specimens were prepared, four from each size range of silicon. Data were collected via neutron time-of-flight, TOF, laboratory and synchrotron x-ray powder diffraction. The Rietveld method was used for data analysis; thus results are anchored by the credibility of the refined crystallographic parameters and the quality of the fit between the calculated and observed data. The refinements included the Sabine model for the effects of extinction.

Implicit in our interpretation of the data is the assumption that all amorphous material associated with the silicon is confined to the crystallite surface and the amorphous layer thickness is invariant with crystallite size. We then model the amount of amorphous silicon based on the known variation in the surface area. Data from the silicon were plotted in the form of refined mass fraction vs. surface area. An extrapolation of a linear fit to this data to the point of zero surface area indicates the true amorphous content of the alumina while the slope of the line indicates the layer thickness of the silicon. Results indicated that the alumina contained 1.58% amorphous material while the silicon layer thickness was 2.6nm. The former value is entirely plausible while the latter corresponds well to the value of 3nm which is generally accepted by the semiconductor industry as the native oxide layer thickness of silicon. However, it was apparent that there were difficulties with the data as excessive scatter was observed. This could be due to difficulties in modeling the extinction exhibited by the silicon or variations in it's surface chemistry. A new suite of specimens are being prepared using intrinsic, float zone, silicon and additional controls on the annealing atmosphere. A systematic bias of about 2% noted in the x-ray refinements is caused by a simplistic modeling of the Lorentz factor. Refinements of the TOF data offered results which appeared to be free of bias; results from the synchrotron data nominally agreed with those from the TOF.