Particle Analysis By Computer Controlled Scanning Electron Microscopy and Energy Dispersive X-Ray Analysis

Scanning electron microscopy (SEM) and energy dispersive X-ray analysis (EDX) when operating in tandem under the control of image analysis software are capable of providing unparalled information on the particle composition of environmental samples. Under computer control the data collection process is fully automated resulting in a uniquely powerful analytical tool. The automation of the SEM/EDX technique makes it possible to acquire data on large numbers of particles which are representative of the particulate matter content of the specimen. Sample analysis by automated SEM/EDX proceeds in three stages.

  1. Data Collection

    Initially, a binary image of the particles is formed (typically from the BEI), and data on individual particle size and shape is obtained by the image analysis software. Then, under computer control, X rays are collected from each particle. This process continues until a statistically significant number of particles have been analyzed.

  2. Data Processing

    X-ray data for each particle is processed in order to remove the X-ray background and to correct for any peak overlaps. The data is then normalized to give the percentage of each measured element present in the individual particles. The compositional and morphological data are then combined for exploratory data analysis.

  3. Data Analysis


    Analysis of the individual particle data may take two forms. If the types of particle present in the sample are known, and the objective of the data analysis phase is to determine the proportions of each particle type present, then this step simply becomes an exercise in type assignment. Alternatively, if the particle types which are present are not known, then the aim of the data analysis stage is to identify them. Typically, this is accomplished using some form of supervised or unsupervised pattern recognition technique (e.g. cluster analysis). An example of the results of this form of particle type identification is illustrated opposite. In this example a sample of lead-rich mine waste soil (sieved to <40 µm) from Leadville, CO, was analyzed in the SEM until data had been collected on in excess of 500 metal-rich particles. A divisive hierarchical clustering procedure was used to identify groupings of compositionally-similar particle types. These major groups are identified on the scattergram (by color), which plots the percentage Pb content for each particle against the sum of the Fe, P and S content for each particle. SEI and BEI images of examples of several types have also been included. The cluster analysis identified several major Pb-bearing particle types, including: a Pb-only type (probably Pb carbonate) located at bottom right on the scatter plot, a Pb-P-Ca (± Cl) type (a Pb phosphate mineral), a Pb+Fe-S particle phase, and several minor Pb types (Pb-Mn, Pb-P, and Pb-Si types). Other metal particle types (containing Fe, Zn, Mn) were also identified).

    Once identified, these particle-type groupings serve as the basis for the future classification of additional particulate material from this, or similar sources. The matching of similar particulate sample material is then achieved through the use of some form of source apportionment model (e.g. the particle class balance receptor model).