The intense interest in production of heteroepitaxial quaternary structures of Gax In1-x Asy p1-y on InP for electro-optical telecommunications systems has Stimulated development of non-destructive techniques for their analysis. One of the most important is double axis X-ray diffractometry, a technique originally developed in the 1920s but only now coming into widespread use as a routine assessment tool. The basic theory is well treated by James and discussion of alignment errors are found in references cited by Fewster in a paper describing alignment procedures for the automated diffractometer manufactured by Bede Scientific Instruments of Durham, The application to III-V systems has been discussed by Tanner, Barnett and Bill.

An extensive programme of characterization of optoelectronic device material has been performed at the Synchrotron Radiation Source, Baresbury Laboratory, in collaboration with Plessey Research, Caswell. The material was grown by Plessey Research by liquid phase epitaxy on InP substrates and had quaternary active layers with, usually, four epilayers in total. Some specimens had graded epilayers. This paper reports use of the methods of double crystal topography, rocking curve analysis and simulation, selective etching and Talysurf measurement in order to develop and assess non-destructive methods of evaluation. The destructive methods above were therefore used in order to test and verify the non-destructive X-ray techniques.

In this paper, the application of recently developed x-ray diffraction techniques to the characterization of thin epitaxial films will be discussed. The double-crystal diffractometer, with high resolution in the non-dispersive arrangement, enables the materials scientist to study epitaxial systems having a very small mismatch with high precision. A key part of the characterization of an epitaxial film is the determination of the strain tensor by measuring lattice spacing! in various directions The determination of strain and composition profiles in ion-implanted films, epitaxial layers and superlattices by rocking-curve analysis will also be reviewed. Grazingincidence diffraction, an emerging new technique, can be used to obtain structural details parallel to the interface on films with thicknesses ranging down to a few atomic layers. The synchroton has now become increasingly available as a powerful source of x radiation which will facilitate the application of conventional and grazing-incidence diffraction to ultra-thin films.

Elastic strain significantly affects the electric and optical properties of PbTe/Pb1-xSnxTe - strained-layer superlattices. In the range of 10 - 350K the temperature dependence of the elastic strain present in these superlattices was measured by double-crystal x-ray diffraction. For superlattice periods smaller than 100nm High-angle x-ray interferences were observed. Using a novel method, which makes use of the High-angle interferences both for symmetrical as well as for asymmetrical reflections in a theta-twotheta scan with a narrow detector slit, the relative inclination of equivalent lattice planes due the elastic strain was measured. The components of the complete strain tensor of the constituent layers can be determined seperately even if their unstrained lattice constants are not known with sufficient accuracy as is the case in ternary and quaternary compounds. The lattice mismatch of up to 0.4% for Sn-contents smaller than 20% was found to be accommodated almost completely by elastic misfit strain. The amount of strain is shared between the constituent layers inversely to their relative thicknesses as long as the superlattice as a whole is much thicker than the buffer layer. Below room temperature an additional temperature dependent tensile strain due to differnt thermal expansion coefficients of the film and the BaF2-substrate is measured quantitatively.

The continuing drive in the semiconductor industry towards smaller and more complex electronic devices has placed increasing severe demands on the materials systems that are used in semiconductor circuits. The refractory metal silicides that are commonly used for electrodes and interconnects are particularly susceptible to the sometimes conflicting requirements of good electrical properties, ease of processing, and long term reliability. We have found that the characterization of metal silicides with a fully automated diffractometer based on the Seemann- Bohlin (S∼B) focusing geometry has proved to be an important aid in the development of several semiconductor metallization systems.

Although a conventional Bragg-Brentano (B-B) diffractometer can provide adequate information for some thin film systems, the use of an S-B diffractometer in which the incident x-ray beam makes a very small angle of incidence with the sample surface results in a substantial improvement in the usable signal from very thin samples.

The introduction of various atoms into crystal surfaces by ion implantation has become common practice in the research laboratory and in many industrial processes. The deceleration of the introduced foreign ions during implantation results in extensive lattice damage. Atom displacement from equilibrium crystal sites and pile-up of the new atoms take place. The use of the pulsed laser for annealing these atomic defects in a single crystal is well documented . The absorption of the total energy in the laser pulse can be rapidly transferred to the lattice, approximately 10-9s, for an elevated temperature anneal of the damaged crystal. The photon energy of the monochromatic laser beam can be tailored so that the absorption of the radiation by a specific crystal can “be matched for its penetration into the crystal to that of the depth of the implanted atoms. A highly efficient mechanism is thus utilized to anneal just that portion of the crystal which is damaged by the impinging atoms of the implant.

In epitaxial single crystal growing it happens very often that the composition of the grown film, say AxB1-x C, is not known exactly. Usually it would be possible to find out that compositions imply by determining the lattice constants of the film. The problem of such a measurement lies in the fact of internal and unknown stresses in the film. On the other side the knowledge of the state of stress σ and strain ε can also be of great importance. To calculate the stresses from X- ray measurements we need the lattice constants in the unstrained condition.

Hornstra and Bartels have shown that the lattice constant can be det ermined by X-ray measurements if the strain is isotropic in the film plane. In the following we show that the lattice constant and the full strain tensor can be calculated fromat least four X- ray measurements also in the general case.

The characterization of multi-layer thin films by X-ray fluorescence using the fundamental parameter method and the LAMA-III program is described. Analyses of a double-layer FeMn/NiFe and two triple-layer NiFe/Cu/Cr and Cr/Cu/NiFe specimens show that the complex inter-layer absorption and secondary fluorescence effects were properly corrected. The compositions and thicknesses of all layers agreed to ±2% with corresponding single-layer films, a precisian comparable with bulk and single-layer thin film analyses.

In the semiconductor and electronics industries the processes Involved in producing integrated circuit soften require thin films such as P205/SiO2, Si/Al, Ni/Cr, Fe/Ni, or others to be deposited on a substrate wafer. It is critical to the performance and life time of the device that the composition and thickness of these films be precisely controlled. The ability to quickly analyze wafers provides data which can be used to adjust fabrication parameters before large quantities of devices are lost.

The fundamental parameter method for quantitative analysis of composite elements has been a powerful technique for x-ray spectrochemical analysis in which the x-ray intensity and spectral distribution from x-ray spectrographic tubes are the moste ssential factors in the calculating process based on x- ray physics.

A prototype X-ray fluorescence system for chemical and phase microanalysis of materials has been developed and tested. Preliminary work with this system has indicated X-ray fluorescence detection limits on the order of 40 picograms for heavier elements such as gold when using a 100 micron collimator, 400 second counting time and a silver anode operating at 12 Kw. Phase identification by X- ray diffraction can be obtained from the same spot. A proposed design for an improved system providing greater elemental sensitivities and capable of semi-automated operation has been completed.

Micro and trace element analysis by X-ray fluorescence was carried out using synchrotron radiation from a bending magnet and an undulator for hard and soft X-ray excitation respectively. The minimum detection limits obtained in the hard X-ray region were less than pg, which corresponds to a spatial resolution of less than a hundred micronmeters, with a detection limit of a few ppm. Light elements such as oxygen, nitrogen and carbon in silicon compounds were analyzed by soft X-ray emission spectroscopy using undulator radiation. The minimum detectable amount of the light elements was greatly improved, since undulator radiation is very strong in intensity, and is highly collimated.

In the EDXRF analysis of minor and trace elements in a variety of matrices, the use of a polarized x-ray source incident on a sample will provide minimum detection limits three to five times lower than the use of non-polarized sources (1,2). There are various methods of producing monochromatic polarized x-rays for specimen excitation (3,4,5,6 ). Such x-ray sources may produce the lowest detection limits for a single element or a narrow range of elements. However, if one is interested in simultaneously analyzing a broad range of elements, a polychromatic source is desired (7,8,9). We present here a new method for producing broad-band polarized x-rays.

As a means for quantitative texture analysis, the crystallite orientation distribution function analysis has an important drawback: to bring ghosts as a consequence of the presence of a non-trivial kernel which consists of the spherical harmonics of odd order terms. In the spherical hamonic analysis, ghosts occur in the particular orientations by symmetry operation from the real orientation in accordance with the symmetry of the harmonics of even orders. For recovery of the odd order harmonics, the 9th-order generalized spherical harmonics are linearly combined and added to the orientation distribution function reconstructed from pole figures to a composite function. The coefficients of the linear combination are optimized to minimize the sum of negative values in the composite function. Reproducibility was simulated by using artificial pole figures of single or multiple component textures. Elimination of the ghosts is accompanied by increase in the height of real peak in the composite function of a single preferred orientation. Relative fractions of both major and minor textural components are reproduced with satisfactory fidelity In the simulation for analysis of multi-component textures.

Analysis of sample composition and/or thickness in a routine, process-control or monitoring environment generally requires rapid turn-around time with minimal sample handling. X-ray Fluorescence (XRF) is well-suited for these kinds of analyses and has been applied to various bulk and thin sample applications (1-7). This technique is rapid, precise, non-destructive, and requires minimal sample handling.

X-ray Fluorescence is generally considered a secondary analysis technique, that is, instrumentation must be calibrated using known standards before unknown samples may be analyzed (quantitative analysis). Continuing work on standardless approaches, particularly for Energy-Dispersive XRF (EOXRF) systems, relax this requirement for some samples but the large share of XRF analyses still require calibration with standards. A number of mathematical and non-matheraatical techniques have been devised to calibrate XRF measurements for many different sample types (1,2,8,9).

A method for resolving overlapping X-ray fluorescence spectra by curve fitting is described. The profile shape of an experimental fluorescence line obtained by wavelength dispersive method is represented by a simple pseudo-Voigt function, i.e. a sum of an asymmetric Gaussian and Lorentzian, each of equal width. Results showed that the pseudo-Voigt function matched the experimental profiles with high reliability. The relative Gaussian and Lorentzian contents and the asymmetry of the profiles depended upon the analyzing crystal, coliimating system and the 2θ peak position. For fixed crystal and collimator the smaller the 2θ, the larger the Gaussian content and the lower the asymmetry. The original Gaussian and Loretzian components of the exact Voigt function calculated from the parameters of the fitted pseudo-Voigt function explain the broadening effects of the X-ray emission lines and the instrumental aberrations on observed spectra. Curve fitting method with the psuedo- Voigt function has been used successfully to analyze overlapping fluorescence spectra. Examples and applications include a thin film sample where the Kα and the Kβ lines of adjacent transition elements overlap, and a strontium zirconium oxide specimen where the Zr Kα and the Sr Kβ lines strongly interfere. Concentrations obtained from the resolved individual peak intensities of Zr and Sr Kα lines are within ±1% of the true values.

The total elemental composition of a sample must be measured or specified for successful application of fundamental parameters type matrix corrections in x-ray fluorescence analyses. Unanalyzed light elements can be dealt with in three ways. They can be related to the concentration of another element by stoichiometry, calculated by difference, or their effects minimized by addition of a diluent of known composition, ‘The relative success of these approaches depends on the composition of the sample. Limiting cases tested are high or low concentrations of analytes having long or short wavelength emission lines, in the presence of additional heavy or light elements. Molybdenum oxide is used as the analyte, with the oxygen serving as a typical unanalyzed light element to be treated by stoichiometry, difference, or dilution. The accuracy and precision of the analysis for molybdenum, using either the Kα or Lα line, is the criterion for judging the success of the strategy. Experimental results and theoretical calculations using the XRF11* matrix correction program are employed. Generally, the difference strategy is inferior to stoichiometry or dilution. However, difference methods can give acceptable results except in the case of a high concentration of short wavelength analyte in a light element matrix. Where stoichiometry is not known, an assumed stoichiometry that is correct within one oxidation number gives results comparable to the difference method.

Acquisition of new x—ray fluorescence (XRF) hardware or a data reduction computer can be a particularly frustrating experience for analysts who depend upon programs not supplied by the XRF spectrometer manufacturer. Computerized data collection and reduction pragrams generated for a particular spectrameter/computer system can become virtually useless when a significant part af the spectrometer/computer system is replaced. The problem is compounded in a research environment where many different kinds of samples are encountered and many unique analysis programs developed.

The potential magnitude of this problem was recently demonstrated at the Ford Motor Company Scientific Research Laboratory when the thirteen year old computer-contolled XRF analysis system was replaced by a new state-of-the-art spectrometer/computer system.

This project was Initiated for the purpose of demonstrating the feasibility of on-line x-ray fluorescence (XRF) analysis for the nondestructive assay of fissile elements (SNM) in reactor fuel reprocessing (dlssolver) solutions, using wavelength dispersive x-ray fluorescence analysis because of its high immunity to the intense gamma emissions of the solutions. A prime objective of this project was the identification and dimensioning of.the parameters critical to XRF assays of high accuracy. The concepts presented herein, though directed primarily to assay of solutions with emphasis on low signal-to-noise conditions and low count rates, are applicable to all assays of solids, slurries, and gases.

In recent years there has been substantial development of computer programs which permit the computation of elemental concentration in a variety of samples from basic principles of X-ray absorption and emission. These prog rates are generally called "fundamental parameter" programs. For many years these programs required large main frame computers for execution. An example is the well known NRLXRF program. For the past few years, programs have been available for minicomputers such as the PDP/11 OR LSI/11 systems. The XRF-11 program from Criss Software, the SAP3 program from Batelle Pacific Northwest Laboratory. and a variety of programs from the X-ray instrumentation vendors are examples. Some fundamental parameter programs have been available for several years that can be executed on microcomputers or the so-called personal computers. Some programs use an "effective wavelength" approximation of the tube output to simplify computations.