X-ray fluorescence, or X-ray spectrometry, is a dynamic analytical method; it undergoes constant change and improvement. Because of the early successes for the analysis of difficult samples such as metals, minerals and other solid materials, the technique was widely accepted in industrial applications. This occurred prior to the onset of large government funding to develop instrumental methods such as was provided for the development of inductively coupled plasma emission spectrometry. As a result, many of the important developments in XRF in North America have resulted from research and development by the instrumentation vendors and developments from Industrial and mission oriented government laboratories. There are exceptions, of course.

The European contributions to X-ray fluorescence analysis (XRF) have shown remarkable fluctuations with regard to time and can be divided into three groups, each of them covering about thirty years.

At the beginning there was the discovery of X-rays by Röntgen in 1895. He received the first Nobel Prize in physics (1901). Barkla (Nobel Prize in physics 1917) gave a description of the interaction of X–rays with matter–scattering (1904) and polarization (1906), absorption (1909) and fluorescence (1911). In 1912 the crucial experiment of Friedrich, Knipping and von Laue brought the confirmation on the nature of X–rays as part of the spectrum of electromagnetic waves (Nobel Prize in physics 1914).

This report will describe scientific activity in XRF as well as some practical application topics for Japan and China.

Like most instrumental methods of analysis, X-Ray Fluoresence (XRF) has progressed through the various stages of Innovation, Frustration, Application and Consolidation. My own association with the method dates back to the late 1950's which would correspond to the transition period between the innovation and frustration stages alluded to above. At that time, there were probably less than 500 XRF spectrometers in the whole world and to myself, as a young analyst with a limited background in x-ray diffraction methods, the potential power which the XRF method seemed to offer far outweighed any problems from the ‘matrix effects' that people were starting to talk about. This talk covers my own impressions of the highlights of the 20 year period of frustration and application which were to culminate with the general acceptance of the XRF method, in the mid-1970's, as a versatile and reliable analytical tool.

Over the past many years, x-ray fluorescence analysis (XRF) has developed into a very mature field. The Plenary Session at this, the 35th Annual Denver Conference on Applications of X-Ray Analysis, deals with an evaluation of developments around the world, as viewed by experts from the three major areas, North America, Europe and the Far East. In this brief report, I would like to convey that which I consider to be the state-of-the-art at the present time, and make some prognostications, which may or may not be dignified as predictions. Many of the specifics that I mention will refer to work in which we, at the Naval Research Laboratory, have been involved. To be fair, I will also deal with some areas in which others have worked. In my mind, the field divides itself naturally into four areas: Sources, Dispersion Devices, Detectors and Data Reduction, I would like to deal with each of these in turn.

Elemental analyses of geologic materials were obtained using X-ray microfluorescence combined with simultaneous elemental digital mapping. The analytical results corroborate the information obtained by electron and optical microscopy methods. Zoned garnet and ore deposit specimens were analyzed and compared to information obtained from other analytical methods. The penetrating nature of the X-ray beam using this method provides elemental information from beneath the sample surface. The combination of this depth information with the surface information supplied by electron microscopy provides a more complete analysis of the sample than is possible with each technique alone. Minimal or no sample preparation along with the ability to examine materials in air enhances the microfluorescence and digital mapping analysis technique.

X-ray Microfluorescence (XRMF) analysis uses a finely collimated beam of X-rays to excite fluorescent radiation in a sample (Nichols & Ryon 1986). Characteristic fluorescent radiation emanating from the small interaction volume element is acquired using an energy dispersive detector placed in close proximity to the sample. The signal from the detector is processed using a computer-based multi-channel analyzer.

XRMF imaging is accomplished by translating the sample through the small X-ray beam in a step or continuous raster mode. As the sample is translated, a pixel by pixel X-ray intensity image is formed for each chemical element in the sample. The resulting digitized image information for each element is stored for subsequent processing and/or display. The images, in the form of elemental maps representing identical areas, may be displayed and color coded by element and/or intensity and then overlayed for spatial correlation.

The present study of parameters affecting the performance of an X-ray microfluorescence system has shown how such systems use X-ray beams with effective spot sizes less than 100 micrometers to bridge the gap in analytical capabilities between predominately surface micro analytical techniques such as SEM/EDX and bulk analytical methods such as standard XRF analysis. The combination of XRMF spectroscopy with digital imaging allows chemical information to be obtained and mapped from surface layers as well as from layers or structures beneath the sample surface. Simultaneously, it provides valuable high resolution chemical information in a readily interpreted visual form which displays the homogeneity within a given layer or structure. XRMF systems retain the advantages of minimal sample preparation, non-destructive analysis and high sensitivity inherent to XRF methods.

Electron induced X-ray mapping together with modern SEM/EDX analysis systems has reached a high level of perfection due to established methods of beam deflection and focusing and today's standard in energy dispersive X-ray detection and data processing. X-ray analysis of specimens based on X-ray excitation (XRF/XRD) is routinely performed on comparatively large specimen areas without conserved spatial information. XRF-/XRD-imaging capabilities are not yet commonly available on standard spectrometers, since both suitable X-ray optical elements are missing and there is a large intensity loss due to the necessary primary beam collimation.

The characterization of alloys is strongly requested from metallurgists as basic data for computer aided alloy design systems for duplex type alloys such as α-γ type stainless steels, γ-γ’ type Ni-base superalloys, α-β type Ti alloys, etc. The characterization falls into the following three categories.

1. 1) Determination of entire average composition of bulk specimens.

2. 2) Identification and the determination of phases including deleterious phases in the alloys.

3. 3) Determination of the composition of all phases in the alloys and the accuracy of the evaluation of the results obtained without standards.

During the past eight years or so there has been growing interest in using a polarized x-ray source in energy dispersive x-ray fluorescence spectrometers (1,2,3,4). The effect is to annihilate the source x rays before they scatter into the detector, thus significantly increasing the signal to noise ratio.

Both characteristic or Bremsstrahlung radiation can be polarized by 90° scattering from crystals (Bragg angle = 45°) or from amorphous materials respectively. This 90° polarizing scatter event greatly reduces the incident source radiation on a sample. In an effort to regain some intensity use is made of concave surfaces to utilize a manifold of beams (5,6,7).

In the event of an accident in a light water moderated nuclear plant, the fission products escape from the water moderator in form of gas bubbles. One of the most important fission products is Iodine. Presently there are only rough estimations of the escape of Iodine. The aim of the experiment planned at the Swiss Federal Institute for Reactor Research (EIR) is to simulate the conditions of an accident and to measure the amount of Iodine which escapes from the moderator water into the space inside the reactor containment.

It is supposed that at 5 m depth in a water pool, the canning of the fuel element explodes releasing 1-3 liter large gas bubbles containing the volatile fission products. The Iodine vapor concentration, saturated in the gas bubble, will be about 3 mg/l. It is expected that the water strongly absorbs the Iodine vapor and the I concentration in the gas bubble arriving at the water surface will be strongly reduced to a few ug/l.

X-Ray Fluorescence (XRF) Analysis is an established analytical technique, widely used in industry and research laboratories for accurate, reproducible, and timely analysis of liquid (aqueous and non-aqueous) and solid samples. Modern X-ray spectrometers are of many types, ranging from inexpensive, simple systems for single element analysis to complex, expensive, automated systems capable of providing thousands of determinations per week. Automated data handling is now relatively commonplace, as are sample changers and matrix correction techniques. For proper application of the technique, it is still necessary to prepare the sample in a suitable manner and present it to the spectrometer so that quantitative information can be obtained. This step, sample preparation, is the most labor intensive portion of the analysis.

The fundamental-parameter technique is an important tool for quantitative x-ray chemical analysis and is routinely applied for the quantification of bulk specimens and single layer films. A method extending it to multiple film layers has recently been introduced by Mantler and results from such applications have been reported by Huang and Parrish. In addition, fundamental-parameter methods can be employed to predict intensity ratios of fluorescent lines as well as the spectral distribution of radiation scattered by the specimen (shape of the background in the vicinity of emission, lines). This is useful for accurate quantitative analysis in the case of a poor peak-to-background ratio, where the precise determination of net intensities is difficult.

The analysis of many types of geological samples by X- ray fluorescence (XRF) spectroscopy has become routine. For example, the analysis of major elements in silicate rocks is not a problem because of the large number of standard reference materials available, and because of the benefits of the fusion preparation. Even simple linear regression analysis provides accurate results when standards and samples are similar.

The feasibility of utilizing the Inverse Monte Carlo (IMC) method to determine elemental amounts in homogeneous samples from energy-dispersive X-ray fluorescence (EDXRF) measurements has been investigated. IMC is a novel application of standard Monte Carlo that, in principle, allows a rather large class of inverse problems to be solved noniteratively, in the sense that the simulation is not repeated. IMC is implemented by hypothesizing values for the unknown parameters, executing a direct Monte Carlo simulation using these values and scoring the results with factors that contain the unknown parameters. By equating the resulting Monte Carlo estimators to the known measured responses, a system of algebraic equations is formed that may be soluble by standard numerical techniques. Advantages of IMC include the facts that it can be applied to complex (e.g., multidimensional) problems and that it is relatively efficient since the simulation is performed only once. In the EDXRF case, simplified assumptions have been used to construct an approximate IMC solution. The model provides a means to determine the elemental amounts in an unknown sample by treating the unknown composition as a perturbation around the composition of a fixed "reference" sample that contains the same elements as the unknown sample. Primary and secondary X-rays are included in the model. The resulting system of algebraic equations is nonlinear and solutions are obtained via an approximation. Since the model is based on correlated sampling techniques, several unknown samples can be modeled simultaneously provided that they contain the same elements and their compositions are sufficiently close to that of the assumed reference sample. Direct Monte Carlo was used to generate the relative intensities for eighteen different samples in a ternary (Ni-Fe-Cr) system; these results were then used in the IMC model to recover the sample weight fractions. Excellent results were obtained, demonstrating the feasibility of the approach.

A Monte Carlo model that predicts the entire photon, spectrum for energy-dispersive X-ray fluorescence (EDXRF) analyzers excited by radio-isotope sources from multielement homogeneous samples is developed and demonstrated. The components of the photon spectrum include: (1) the and Kα and Kβ characteristic primary, secondary and tertiary X rays from both the unscattered and scattered source photons, (2) the characteristic X rays excited by other characteristic X rays that have been scattered, and (3) the scattered source photons from single, double, and multiple scatters in the sample.

The computer code NCSMCXF based on this model has been developed. It is capable of handling up to 20 elements per sample and provides a detailed account of the intensities of the X rays and backscattered source photons per unit source decay as well as a summary of the relative intensities from all elements present in the sample. Cubic splines are used within the code for photoelectric and total scattering cross sections and two-variable cubic splines for angular coherent and incoherent scattering distributions for efficiency in both computation time and storage. The code also provides the pulse-height spectrum of the sample by using the appropriate Si(Li) detector response function. The Monte Carlo predictions for benchmark experimental results on two alloy samples of known composition indicate that the model is very accurate. This approach is capable of replacing most of the experimental work presently required in EDXRF quantitative analysis.

A previous Monte Carlo model that uses the simple assumption of spherical homogeneous particles to approximate sample heterogeneities has been modified to improve the computer execution time requirements for the heterogeneous sample case. A new technique for photon tracking in this medium is used and reduces the computation time requirement by half.

In industrial laboratories, a great diversity of samples is usually presented for analysis (alloys, powders, solutions, …). Fast results are invariably required so that methods with minimal preparation times have to be developed. This paper discusses a method for quantitative analysis of metallic samples in the form of clippings or odd-shaped fragments where the specimen surface does not cover the total surface of the sample holder. In some laboratories it is common practice to use sample holders with a small orifice to analyse sample fragments, but this method strongly reduces the fluorescent intensities. We have the feeling that a method that uses the whole surface of the specimen should be much welcomed.

In monochromatically excited energy-dispersive X-ray fluorescence spectrometry, the spectral background has mainly been investigated for a better understanding of its origin and nature, or with the aim to reduce its magnitude for an improvement of detection limits.

The observation that the background intensity can largely be ascribed to a residual polychromatic fraction of the exciting radiation has led to the derivation of an empirical relationship between the scattering characteristics of a sample and the observed background intensity. This dependence constitutes the basis of a matrix correction procedure useful for samples with a considerable fraction of light elements.

Two algorithmic variations are presented: one that relies on the definition of a single matrix element, and a second one that can be used in conjunction with the measurement of the incoherently scattered exciting radiation. These options are useful if direct quantification of either or both scatter peaks is not feasible.

Results are presented for a variety of Standard Reference Materials.