X-ray Compositional MicroAnalysis:

EDS and WDS

William Barker and John Fournelle

Department of Geology and Geophysics

University of Wisconsin-Madison

Madison, WI 53706

(Chapter 9 of Electron Microscopy: Theory and Practice, 1996,

UW Madison, Anatomy Department, Anatomy 660)

Abstract: Electron microbeam instruments are useful for determining the chemical compositions of microvolumes of specimens. Highly energetic electrons that impact the sample yield a variety of by-products, one being X-rays. X-rays have characteristic energies/wavelengths and can be detected using either a solid state energy dispersive spectrometer (EDS) detector, or a diffracting crystal in tandem with a gas-filled proportional counter wavelength dispersive spectrometer (WDS) detector. Each detector type has its merits and drawbacks. In practice, a main distinction is made between specimens that are prepared and studied as ultrathin sections (~nm thick) through which the electron beam passes, i.e. in a TEM, and "thick" sections, which are at least a few microns or more thick. In the first case, EDS is the method used, whereas both WDS and EDS can be used in the latter. Depending on specific features of the detector, all elements on the periodic table from Be on up may be analyzed.

A brief historical outline of X-ray analysis

tied with the development of electron microbeam instruments

1895 - X-rays discovered by Roentgen, produced by electron bombardment of inert gas in tubes; gas fluoresces and nearby photographic plates are exposed (X-rays' wavelength = 0.05 - 100 Å)

1912 - Friedrich and Knipping confirmed that X-rays could be diffracted by crystals with lattice spacings of similar dimension

1913 - the Braggs obtained the first X-ray spectrum of Pt using an NaCl crystal (it's the Law: n*lamda = 2d *sin theta )

1913 - Mosely found that there was a systematic variation of the wavelength of characteristic X-rays from various elements ( wavelength inversely proportional to Z squared )

1922 - Hadding used X-ray spectra to chemically analyze minerals

1923 - von Hevesy discovered Hf after noticing a gap at Z=72

late 1920's - in Germany, development of transmission electron microscopes, with first demonstration in 1932 of transmission electron microscopy by Ernst Ruska (belated Nobel prize for it in 1986) prototype build by Siemens & Halske Co but WWII prevented sale and use outside Germany

1930's - scanning coils added to TEM, producing STEM (image produced by secondary electrons emitted by specimen)

1940 - RCA sold first commercial TEM outside Germany

1942 - first use of SEM to examine surfaces of thick specimens at RCA Labs

1949 - Castaing built first electron microprobe for microchemical analysis (with crystal focusing wavelength dispersive spectrometer = WDS) for Ph.D at University of Paris, and developed the basic theory

1956 - commercial production of electron microprobe began (Cameca)

1965 - commercial production of SEM began

1968 - solid state EDS detectors developed

Where do X-rays come from?

When an electron beam interacts with (rather than simply passes through or transmits unchanged) the atoms in a sample, individual incident electrons undergo two types of scatter - elastic and inelastic. In the former, only the trajectory changes and the kinetic energy and velocity remain constant. In the case of inelastic scattering, some incident electrons will actually collide with and displace electrons from their orbits (shells) around nuclei of atoms comprising the sample. This interaction places the atom in an excited (unstable) state. The atom "wants" to return to a ground or unexcited state. One way for an atom to return to ground state is for an electron in a higher orbital to "fall" into the vacant shell and take the place of the displaced electron. When this occurs, energy is lost and a single X-ray is emitted (Figure 1).

The electronic orbits of each element are relatively unique and thus the set of X-rays emitted from these electron interactions are also fairly characteristic with respect to both energy and wavelength for each element. Energy and wavelength are related by the equation

lambda = 12.396/E

where wavelength (lambda) is in Angstroms and energy (E) is in KeV

For example, if an incident electron strikes an inner K shell electron and knocks it out of its orbit, an L shell electron will drop into the K orbit and emit a K alpha X-ray of some diagnostic energy and wavelength; there is a lower probability that an M shell electron will drop in, yielding a K beta X-ray. Similarly, an L shell electron may be displaced by the incident electron and be replaced by an M shell, and in this case emits an L alpha X-ray. One interesting aspect is that even though the energy required to displace a K shell electron is greater than that for an L or M shell electron, the yield of K alpha X-rays is always much greater. This is because the K orbit is much smaller (closer to the nucleus) and thus the statistical likelihood of a collision is greater. Elements of increasing Z give off a greater variety of X-rays (Figure 2), for they have more electrons in a greater number of orbits about their nucleus. The potential overlap of resultant peaks with those of other elements constitutes the source of one potential problem in interpreting X-ray spectra.

Some electrons nick or hit the nucleus and lose variable amounts of energy in the process, yielding continuum X-rays (a.k.a. Bremsstrahlung or white radiation) which range from very low energy up to the accelerating voltage. This continuum is the effective background which has to be taken into account in chemical analysis.

Monte Carlo simulations of electron-specimen interactions

Several software packages have been developed that model the scattering (elastic and inelastic) of electrons that occurs when they hit specimens. These programs demonstrate very graphically the extent of elastic scattering that occur in bulk specimens, producing the ~micron-size "interaction volume", which is the spatial resolution of chemical analysis in EMPA -- and the difference for nanometer thin films in TEM, where there is effectively no room for scattering to occur (Figure 3). Traditionally, textbooks show diagrams of electron scattering in a tear-shaped pattern in the specimen; this is actually a "special case", for a low atomic number plastic -- appropriate for some biological material but not for minerals or metals.

One can model a variety of conditions (sample thickness, accelerating voltage) on a computer prior to using an electron microbeam instrument, to determine, for example, spatial resolution of the X-ray data.

David Joy (joy@utkvx.utk.edu) has a nice, free set of programs for either Mac or IBM. We will have disks available, and they can be FTPed from the Microscopy Society of America (http://www.MSA.Microscopy.Com/1-Public/4-MMSLib/Monte/Joy).

What can these X-rays reveal about a sample?

By placing a suitable X-ray detector coupled to a set of electronic components (amplifiers, counters, analog-digital convertors) and a computer, one can detect and analyze X-rays emitted from a sample undergoing electron illumination. The resulting X-ray spectrum can be displayed according to energy (Energy Dispersive X-ray Spectroscopy - EDS) or wavelength (Wavelength Dispersive X-ray Spectroscopy -WDS). These data can then be either analyzed to give an indication of which elements occur in a sample (qualitative), or in a much more rigorous process, a precise and accurate (quantitative) chemical analysis.

This seems like an appropriate place for a word about safety. Any electron beam of sufficient energy will generate X-radiation. Modern electron microscope are heavily shielded with lead to protect you from irradiation and the chance of exposure is minimal. In your professional career you may at times use older equipment with, shall we say, less than adequate shielding. Newer machines may also suffer from improper repair or modifications. On balance, the likelihood of X-ray exposure is very small, but well worth keeping in mind.

Energy Dispersive X-ray Analysis (EDXRA, EDS)

X-rays emitted from a sample under electron bombardment are collected with a liquid nitrogen-cooled solid state detector (Figure 4) and analyzed via computer according to their energy. Typically, the computer programs used in EDS will display a real time histogram (Figure 5) of number of X-rays detected per channel (variable, but usually 10 electron volts/channel) versus energy expressed in KeV (thousand electron volts).

In practice, EDS is most often used for qualitative elemental analysis, simply to determine which elements are present and their relative abundance. Depending upon the specific investigation's needs, researchers in need of quantitative results may be advised to use the electron microprobe. In some instances, however, the area of interest is simply too small and must be analyzed by TEM (where EDS is the only option) or high resolution SEM (where the low beam currents used preclude WDS, making EDS the only option).

While it is indeed possible to obtain reliable quantitative analyses in an analytical TEM, the process is rigorous, requiring the use of well-characterized standards. The best results are obtained using very thin samples (< 10 nm), and considerations such as absorption and take off angle are unnecessary under these conditions. In this case, one must factor in the specific response of the EDS detector on your microscope to the X-ray spectra generated (e.g., compensation for absorption of light energy X-rays by your detector's Be window, other specific responses of the electronics, etc). In 'thick section' EMPA work, experimentally determined K factors , -- the ratio of X-ray intensities on the specimen compared to a standard -- are used. In the application of EDS to AEM, the mass fraction of an element can be calculated using Hall's "continuum method", where the characteristic X-ray intensity is ratioed to the continuum X-ray intensity taken at a convenient location in the spectrum. Use of pre-analyzed standards can give a K factor (different from the ones used in EMPA, above) which permits calculating the elemental abundances.

Another approach is to use empirically determined Cliff-Lorimer factors. C-L factors are ratios of a particular element's characteristic X-ray intensity to that of one element present in many phases of interest, e.g. Si. Calibrations are done upon reference materials. This ratioing yields atomic ratios - such as Na/Si, Al/Si - which are then normalized, yielding a chemical composition. Please refer to Williams (1987), Joy, et. al (1986) or Chandler (1977) for further details.

Common problems for EDS

All common problems relating to EDS in the AEM occur either during data collection or analysis.

Data collection problems

Poor sample preparation- for thin film EDS, the sample must not be thicker than 10 nm. Otherwise one must correct for sample absorption and exact sample - detector geometries must be mathematically accounted for during data reduction. Additionally, coatings (if any) and composition of the sample support films and grids must be carefully considered.

Incorrect sample geometry - X-rays emerge from a sample and travel line - of -sight trajectories. Thus, if the sample is tilted incorrectly, something may actually block the path between detector and sample. This will manifest itself either as an inordinately low number of X-rays (expressed as counts sec-1) or you may notice an absence of low energy X-rays (either due to blocking or reabsorption) in the spectrum being collected. Correct the problem by repositioning the sample.

Contamination - one of the principal assumptions one makes in thin film EDS analysis is that collected X-rays in fact originate from the area of interest in the sample. Even closely collimated detectors can "see" quite a large area of the instrument (i.e., polepiece, holder). A major source of extraneous X-rays are high energy "hard" X-rays which result from stray electrons in the incident beam striking the interior of the AEM. This is a serious problem for intermediate and high voltage AEMs (e.g., accelerating voltages in excess of 120 keV), for a 200 keV X-ray can pass directly through a C2 aperture, strike the sample at some distance from the area of interest and cause X-ray fluorescence. An additional problem in IVAEM and HVAEM is X-ray fluorescence of the objective lens polepiece or the sample holder. For this reason, one must always use an analytical holder machined from Beryllium (Be X-rays are too weak to be collected in EDS). From a design standpoint, these problems can be reduced or eliminated entirely through the use of thicker apertures, Be holders, and engineering changes. Always run appropriate background analyses (i.e., off the specimen, to check for stray X-rays).

Data analysis pitfalls

Artifacts - escape peaks and sum peaks, peak overlaps, and excessive deadtime are all phenomena that an EDS user must be aware of. Modern analytical software used in processing energy dispersive X-ray spectra can generally take them into account -- but such software is not perfect. Also, many users will look at the raw spectra, where the software may or may not have labelled the artifacts.

Escape peaks - there is a statistical probability that some of the X-rays, generated in the sample and impacting the solid state detector (e.g., a SiLi device), will 'inadvertently' knock out Si K-shell electrons in the detector, reducing that X-ray's energy measured in the detector by the Si absorption edge energy (1.84 KeV). Say you're looking at something will lots of Fe (Ka of 6.40 KeV); the Si-escape peak of Fe Ka will appear at 4.56 KeV. You see this escape peak only for the major elements present.

Sum peaks - this phenomenon occurs where the count rates are moderate to high, when two X-rays impact the detector virually instantaneously; the pulse created and measured is the sum of the two X-ray energies. Say you have a sample with lots of Si (Ka of 1.74 KeV) and Al (Ka of 1.487); a peak at 3.23 KeV is the sum peak, not to be assumed to be a K peak (Ka of 3.31 KeV).

Peak overlaps - the spectral resolution of EDS is not a great as WDS. Resolution is usually defined as the FWHM (full width at half maximum) of pure Mn Ka: ~ 150 eV. Therefore, the separation of some peaks can be poor. Examples include the case where small amounts of Fe are being investigated in the presence of large amounts of Mn (Mn Kb is very close to Fe Ka), and the case where Cu, Zn and Na are present together: the L lines of Cu and Zn are close to the K lines of Na (Table 1).

Excessive deadtime - because of the closeness of the EDS detector to the sample, and the possibility that the user may be using high beam currents, there may be 'pulse pileup' where the electronics cannot keep up with the X-rays impacting the detector. The electronics/software therefore has to try to adjust for the x-rays not counted, by calculating a 'deadtime correction'; the larger the correction, the greater the margin of error. Generally the deadtime should be kept below 20-30% (usually indicated on the monitor), either by lowering the beam current, inserting apertures in front of the detector nosepiece, or retracting the detector (if adjustable). Excessive deadtime can also cause a shift of the peak position.

Wavelength Dispersive Spectrometry (WDS):

the heart of Electron Microprobe Analysis (EMPA)

WDS was the original technique developed to precisely and accurately determine chemical compositions of microvolumes (a few cubic microns) of "thick" specimens, and the instrument used is the electron microprobe. In the 1960-70s there were roughly half a dozen companies commercially producing them; today, there are only two (JEOL and CAMECA). A full-package EMPA today costs $500-$750,000.

The key feature of the electron microprobe (Figure 6) is a crystal-focusing spectrometer, of which there are usually 3-5, although one 'extinct' company's 1970-80 9 spectrometer instruments are still much in use. An electron gun (usually a heated tungsten filament) produces the electrons, which are then focused with 2 'condenser' lenses, filtered through several apertures and finely focused on the specimen ('objective' lens). A high vacuum (10-6 torr) is required, for the life of the filament and to minimize electron and X-ray dispersion in the column.

X-rays (as well as many other energy throw-offs) are produced in the "interaction volume" immediately below the impact zone of the finely focused electron beam. A very small fraction of all X-rays will be at the proper "take-off" angle to head up into the spectrometer (a much smaller fraction compared to an EDS detector mounted a few cms from the sample). Remember each element's characteristic X-ray has a distinct wavelength, and by adjusting the tilt of the crystal in the spectrometer, at a specific angle it will diffract the wavelength of specific element's X-rays. Those diffracted X-rays are then directed into a gas-filled proportional counting tube, which has a tin wire running down its middle, at 1-2 kV potential. The X-rays are absorbed by gas molecules (e.g., P10: 90% Ar, 10% CH4) in the tube, with photoelectrons ejected; these produce a secondary cascade of interactions, yielding an amplification of the signal (x103-105) so that it can be handled by the electronics.

Different diffracting crystals, with 2d (lattice spacings) varying from 2.5 to 200 Å, are used to be able to 'reach' the different wavelengths of various elements. In recent years, the development of 'layered synthetic crystals" of large 2d has lead to the ability to analyze the lower Z elements (Be, B, C, N, O), although inherent limitations in the physics of the process (e.g., large loss of signal by absorption in the sample) limit the applications.

To be able to produce an X-ray of given element, you must hit it with electrons with energy greater than the minimum excitation energy (also known as the absorption edge energy); for optimal operation, the beam energy should be 2 to 2.5 times that lower limit. 10 kV is about as low as you can go, for coated samples; 35-40 KeV is the upper limit of standard electron microprobes.

Calibration and quantitative analysis: the basic theory is quite simple; the ratio of the intensity of X-ray in the unknown to that of it in the standard is approximately equal to the ratio of the concentration of the element in the unknown to that in the standard. The first ratio, of X-ray intensities, is known as the "K". Obviously, the approximate composition of the unknown can be determined, given knowing the other 3 parts of the equation.

A significant complication arises because we are dealing with a "thick" specimen (more than a few microns thick): absorption of X-rays, particularly long wavelength, lower energy ones, can be an important factor in reducing the number of certain X-rays counted, compared to those generated in the sample. In addition to this absorption correction (A), corrections need also be made for fluorescence (F: the generated X-rays may also produce additional X-rays of other lines in the sample) and for 'atomic number effects'(Z). These three corrections are the matrix correction, ZAF, based upon various physical models developed to describe these effects.

The best situation occurs where the standards used are similar in composition to the unknowns, so that there are no large extrapolations.

EDS on an SEM can be used to get quantitative chemical information in many situations, but there are some/many cases where WDS is the preferred technique, if available:

-- where the peaks are too close in EDS to be resolved (typically EDS resolution is ~150 eV, versus WDS which is ~5 eV

-- where trace element levels are desired, WDS has a higher P/B, yielding lower minimum detection limits.

WDS does not usually suffer from pulse pileup (too many counts coming in, i.e. from major elements) that occurs in EDS, and which must be compensated for mathematically.

WDS has different spectral artifacts, compared with EDS: for WDS, the only problem that may occur is if a higher order reflection (n>1) of a line falls near the line of interest. However, WDS can in many cases eliminate that higher order line, by fine tuning the proportional counter electronics, applying "pulse height analysis" and weeding out the unwanted X-rays. Some lines, however, are so close that special 'tricks' may be required, e.g. for V Ka in the presence of abundant Ti (Kb interfernec) or for B Ka in the presence of abundant Mo (M-line interference). See Table 1.

Electron microprobes today usually also contain

· scanning coils, so the beam can raster across a specimen, producing a scanned image (a la SEM),

· secondary electron detectors (which in scanning mode yield SE images, showing surface features),

· backscattered electron (BSE) detectors yielding images where different phases of differing mean atomic number stand out sharply

· cathodoluminescence (CL) detectors, where the light emitted from the electron-specimen interaction can be imaged, and can clearly show features in minerals and semi-conductors that would be difficult to see compositionally (these are features due to differences in specific trace element concentrations, or crystal lattice defects)

· EDS detectors which are mostly used for quick 'look-sees' to see a snapshot of the X-ray spectrum of an unknown, to determine if it should get the full WDS treatment; also integrated EDS-WDS systems can be used to X-ray map up to 15 or so elements in complex specimens.

A comparison of some key features of EDS versus WDS is given in Table 2.

X-ray compositional analysis of biological specimens

For biological materials, one useful application may be using an analytical scanning transmission electron microscope and appropriate hardware and software to collect X-ray maps of cryosectioned samples.

In addition, there is some history of application of the electron microprobe and WDS to biological investigations, although the location of most electron microprobes in geology, physics or engineering departments has not facilitated such use. Where this has occurred, there have been studies made on a variety of features, many being of trace levels of heavy metals in fibers, hair, skin and fluids (see references below). Important considerations are the low thermal conductivity of organic material and the possible hydrated state of the material, where the rapid heat buildup may damage the material being studied. Special sample preparation as well as beam operating conditions are necessary in many (though not all) cases.

In the UW-Madison Electron Microprobe Lab, work is currently occurring on two projects where biological materials being studied: mapping of metals in woods that have been treated with various preservatives, and an attempt to measure the temporal variation in Hg concentration in hair from humans exposed to environmental Hg pollution. Opportunities for additional application of EMPA-WDS to biological/ medical research are welcome.

Sample Preparation

An important consideration is whether or not the sample is electrically conductive. Biological and most mineral samples are not, and require a very thin coating of a conductive material (e.g., Carbon is commonly used, and applied by evaporation) to eliminate the charging -- repulsion of incident electrons -- that would otherwise occur and hinder analysis.

The samples must also be cleaned, to eliminate contamination that might interfere with the analysis (hydrocarbons from fingerprints or polishing compounds; diamond or alumina or other polishing materials).

For quantitative analysis of thick sections using WDS and EDS, it is critical that the material have a mirror polish (1 micron or less final polish) and be in a mount so that this surface is at a known and constant angle to the electron beam (in EMPA, 90°). If this is not so, the path length of the X-rays through the material at the takeoff angle will not be constant, and the key absorption correction will be incorrect.

References

AEM-EDS

Chandler, J.A. (1977) X-ray Microanalysis in the Electron Microscope. North-Holland Publishing Company, Amsterdam, 547 pp.

Joy, D. C., Romig, A. D., Jr., and Goldstein, J. I. (1986) Principles of Analytical Electron Microscopy, Plenum Press, New York, 448 pp.

Williams, D. B. (1987) Practical Analytical Electron Microscopy in Materials Science. Philips Electron Optics Publishing Group, Mahwah, New Jersey, 153 pp.

EMPA: WDS and EDS

Goldstein,J.I., Newbury, D.E., Echlin, P., Joy,D.C., Romig, A.D. Jr, Lyman, C.E., Fiori, C. and Lifshin, E. Scanning Electron Microscopy and X-ray Microanalysis (1992) Plenum Press. New York, 820 pp. Good reference; 'fundamental' sections highlighted; covers SEM in detail as well as EMPA.

Reed, S. J. B. (1993) Electron Microprobe Analysis. Cambridge University Press. Second edition. 326 pp. Revision of classic 1975 edition.

Welton, Joann E. (1984). SEM Petrology Atlas. American Assoc. of Petroleum Geologists, Tulsa. 237 pp. Has a nice and short explanation of SEM and EDS. Shows SEM images and EDS spectra for a few dozen common minerals. (A few copies available for purchase in Probe Lab, $15)

Williams, K. L. (1987) An Introduction to X-ray Spectrometry. Allen and Unwin, London. 370 pp. Good explanations of EMPA and XRF.

Biological EMPA

Goldstein et al (above) Scanning Electron Microscopy and X-ray Microanalysis. Chapter 12: Sample Preparation for Biological, Organic, Polymeric and Hydrated Materials, pp 571-670

Gupta, Brij L. (1991) Ted Hall and the science of biological microprobe X-ray analysis. Scanning Microscopy, 5, 379-426.

Lechene, C.P. and Warner, R.R. (1977) Ultramicroanalysis: X-ray spectrometry by electron probe excitation. Annual Reviews in Biophysics and Bioengineering, 6, 57-85.

Morgan, A.J. and Winters, C. (1987) The contribution of electron probe X-ray microanalysis (EPXMA) to pollution studies. Scanning Microscopy, 1, 133-157.Figure Credits:

Figure 1. Chandler (1977)

Figure 2. Goldstein et al (1992)

Figure 3. Goldstein et al (1992)

Figure 4. Noran Instruments

Figure 5. William Barker

Figure 6. K.L. Williams (1987)