X-ray Diraction

X-ray diraction has acted as the cornerstone of twentieth-century science. Its development has catalyzed the developments of all of the rest of solid-state science and much of our understanding of chemical bonding. X-ray powder diraction (XRPD) plays a critical role in materials research and development because many materials and minerals are not readily available in single-crystal form. Some 10 00020 000 powder diractometers are in use worldwide, and of the order of 160 000 powder patterns are available in the Powder Diraction File (PDF) electronic database. Although the technique is termed powder diraction, any polycrystalline specimen may be studied using the technique, including monolithic solids, thin lms, and powders. Among the applications of powder diraction are: $ Qualitative identication of the phases present in a specimen using the powder pattern as a ngerprint of the phase which may be compared to known powder patterns stored in a database. $ Quantitative analysis of the concentrations of each phase present in a multiphase specimen can be carried out. $ The unit cell metrics, i.e., the size and shape of the unit cell, of any crystalline material can be determined, and then rened to very high accuracy. Because the lattice parameters of a material change with chemical composition, XRPD nds widespread use in solid solution analysis, phase diagram work, and for measuring the true density of materials. $ Full orientation distribution functions that describe the crystallite orientation in polycrystalline materials can be determined. $ Residual and applied stress can be determined by measuring the strain present in the crystallites composing a monolithic body. $ Microstructural features of a polycrystalline material including the crystallite size, microstrain within crystallites, and defect densities may be determined. $ In situ techniques under controlled gaseous atmosphere, temperature, pressure, eld, and mechanical loading provide a means of determining phase diagrams and reaction kinetics, and provide a means of tracking changes in materials during synthesis and processing. $ Full structure solution from powders. The applications listed above are strictly for polycrystalline materials. It is important to note, however, that powder diractometers may be equipped to study thin lms using x-ray reectivity and high-resolution diraction. 1. Origins of X-ray Powder Patterns X rays are relatively short wavelength, high-energy electromagnetic radiation. Diraction experiments

require x-ray wavelengths of the order of the interatomic spacing to produce interference, and hence typical wavelength ranges are 0.070.2 nm. The most common radiation used in the laboratory is Cu K radiation with l 0.15406 nm. A typical x-ray powder diraction pattern is shown in Fig. 1 and consists of diraction peaks, or lines, as a function of the diraction angle, 2. Diraction occurs when waves scattering from an object constructively and destructively interfere with each other. The diraction peaks in Fig. 1 are a result of constructive interference of waves scattered from the atoms or ions composing a crystal. M. von Laue in 1912 rst determined that x rays diract from the periodic arrangement of atoms in a crystal, exactly analogous to the diraction of visible light from a grating. Based on the work of von Laue, W. H. and W. L. Bragg showed in 1913 that diraction from a crystal is described by the equation now known as Braggs law: l 2dhkt sin (1) This equation allows one to measure the perpendicular distance (dhkl) between imaginary planes which form parallel families and which intersect the repeating unit cell lled with atoms in a way described by the Miller indices (hkl ). X rays of wavelength may be thought of as reecting from these imaginary planes at the measurable angle , where is one-half of the diraction angle as shown in Fig. 1. A powder pattern therefore contains a set of diraction peaks at 2 positions that correspond to the interplanar spacings in the crystal. Although dhkl may be calculated for any set of (hkl ) planes in any unit cell using geometry, the interplanar spacings are most easily calculated in reciprocal space. P. P. Ewald developed what is by far the most useful method for describing diraction phenomena using reciprocal space. Ewald in 1913 showed that a reciprocal lattice can be constructed by using the reciprocals of all of the distances in a real unit cell. The reciprocal lattice has the same symmetry as the real crystal lattice, but the points in the reciprocal lattice are equally spaced and hence simple to manipulate. In an orthogonal lattice (cubic, tetragonal, orthorhombic), all distances are the reciprocals of the real space distances. a*, b*, and c* are the reciprocal lattice parameters, while the interaxial angles are *, *, and *, where the reciprocal of an angle is dened as the complement or 180m minus the angle. Likewise, the reciprocal of dhkl is d* . In the hkl nonorthogonal systems, one must account for the interaxial angles when calculating the reciprocal lattice parameters. The convenience of using reciprocal space is demonstrated when calculating the d* values. Since the hkl length of d* is simply related to the real space dhkl # equation: spacing, one simply uses the d* hkl # l h#a*#jk#b*#jl #c*#j2hka*b* cos * d* hkl j2hla*c* cos *j2kl b*c*cos* (2) 1

X-ray Diraction
10.0

8.0

Intensity (Counts 103)

6.0

4.0

2.0

10-0173> Corundum - A12O3 27-1402> Silicon - Si 20 30 40 50 60 70 2 () 80 90 100 110 120

Figure 1 A powder diraction pattern for a mixture of two phases, with the standard patterns from the PDF database shown below the data.

This expression relates the square of the inverse of dhkl to the size and shape of the reciprocal unit cell for any plane in any crystal system. For many applications then, Eqn. (2) reduces to a particularly simple form. # l (h#jk#jl#)a*# For example, Eqn. (2) becomes d* hkl for a cubic crystal. The intensity of each diraction peak in a pattern is governed by the structure of the crystal and the physical conguration of the diractometer. First we must consider how much intensity a single electron will coherently scatter (the more massive nucleus is a very inecient scatterer and may be ignored). Next we need to consider the interference eects which will occur due to the electrons being distributed in space around atoms and the fact that atoms are not stationary in a lattice but vibrate in an anisotropic manner. Then the interference eects of atoms scattering from dierent regions of the unit cell must be allowed for, and nally the optical conguration of the diractometer and absorption eects must be considered. The intensity of peak i of phase in a powder pattern measured using BraggBrentano geometry (see Fig. 2) is described by the following equation: Ii l Io$e% Mi Q Fi Q # 32rm# c% 2V # e
F E

where Fhkl l
No. of atoms in cell
A

 j="

B E

( fjoj f hjj f hh ) exp(kBj) j

isin#
F

exp 2i(hxjjkyjjlzj) (4)

1jcos# 2i cos# 2m sin# i cos i

(3) 2

The rst term in Eqn. (3) describes classical scattering from an electron as described by J. J. Thompson in 1906, while the last term in parentheses is a correction for the polarization of the x-ray beam on diraction. The term Mi accounts for the multiplicity of (hkl ) planes with identical d-spacings that diract at the same angle; for example, there are six (h00) peaks in a cubic crystal. The structure factor, Fhkl, is the most important quantity in crystallography and contains the atomic scattering factor, fj, and the atom positions in the unit cell, xj, yj, and zj. The exponential term in Fhkl accounts for the phases of the scattered waves, while the atomic scattering factor accounts for the amplitudes scattered from each atom. fj completely describes scattering from an atom by including the number of electrons and their spatial distribution around the atom, including thermal vibrations (Bj). It is important to note that fj explicitly includes the atomic number of each element in a crystal, dem-

X-ray Diraction
(a) Detector X-ray tube line focus Detector slit Focusing circle Sample (b) Parabolic mirror X-ray tube line focus Soller slit collimator Measuring circle

Detector

Sample

Figure 2 Schematics of two x-ray powder diractometers: (a) BraggBrentano parafocusing geometry; (b) parallel beam geometry using a parabolic multilayer optic.

onstrating that x rays scatter more strongly from heavy elements than from light ones. It is a triumph of our understanding of physics that computations of diraction intensities, using Eqn. (3), produce the observed values. This triumph has been heavily exploited in determining the crystal structure of materials. This ability, rst used by the Braggs in 1913, has established the basis of modern solid-state science. The most common application of our ability to compute a powder pattern, in the materials laboratory, is to compare it with any features of the experimental pattern (like preferred orientation of the crystallites, solid solution eects, etc.) to determine any crystal structure or microstructural modications which may have taken place.

through an angle of while the detector is scanned through an angle of 2. This 2 motion implies that the radius of the focusing circle is continually changing throughout a diraction pattern scan. The need to make the beam monochromatic follows directly from Braggs law. If more than one wavelength is present in the beam, each set of planes will yield a diraction peak for each wavelength. This would hopelessly complicate the diraction pattern. The most common method to produce monochromatic radiation is to employ a diracted-beam monochromator. Most BraggBrentano instruments include a quasi single-crystal graphite monochromator which allows only the K and K radiation to reach " # elimination of all the detector, as shown in Fig. 2. The spectral wavelengths from the x-ray tube except the K and K greatly improves the quality of the powder " # pattern. Placing the monochromator in the diracted beam not only removes spectral impurities from the xray tube but also any uoresced x rays from the specimen. Although BraggBrentano geometry is by far the most common conguration for powder diraction, the parafocusing nature of the measurement makes it prone to signicant systematic errors. By far the most serious of these is due to a displacement of the sample from the focusing circle. This, of course, can never be avoided because even if the sample surface is perfectly tangent to the focusing circle, the x rays penetrate into the sample to an average depth dependent on the absorption coecient of the sample. Thus, there is always a sample displacement (or sample transparency) error present. One can remove these errors to obtain high-accuracy measurements of peak positions by using certied line position standards available from NIST.

2.2 Parallel-beam Diractometers Revolutionary new x-ray optics have been developed that have proved to be eective for laboratory measurements. Multilayer optics, polycapillary optics, and high-performance channel-cut crystals have been used independently or together to build high-intensity and\or high-resolution parallel-beam diractometers. A parallel-beam measurement minimizes or completely eliminates systematic errors arising from sample displacement and transparency. In addition, a parallel beam is required for asymmetric diraction experiments, including thin-lm diraction. A parallel-beam diractometer, therefore, lends itself to powder and thin-lm measurements with little or no systematic errors. The only potential disadvantage of a parallel-beam diractometer is termed particle statistics. Compared to a divergent beam Bragg Brentano instrument, a parallel-beam instrument will reduce the number of crystalline grains in a powder specimen that are exactly oriented to produce a 3

2. Instrumentation 2.1 Traditional Parafocusing Instruments Beginning in 1945 and continuing through the 1950s, William Parrish developed the powder diractometer for recording powder patterns electronically. The diractometer is shown schematically in Fig. 2. The parafocusing geometry is called BraggBrentano but should more appropriately be called Parrish geometry. The parafocusing principle in the diractometer comes from keeping the x-ray source (i.e., the x-ray tube target), the specimen, and the receiving slit on a common circle, called the focusing circle. To meet the condition of Braggs law we must move the sample

X-ray Diraction diraction peak. One must therefore take special care to ensure that powder specimens are nely ground or polycrystalline samples are rotated or oscillated to increase the number of grains bathed in the x-ray beam. Both graded multilayer and polycapillary optics may be used to convert a divergent beam into a parallel beam, or to focus an x-ray beam. Multilayer optics consist of thin-lm stacks of alternating high atomic number materials with low atomic number materials. The alternating thin lms therefore act as a crystal, the diraction angle of which is determined by the thickness of the lms and is typically below 3m. Herbert Go $ bel rst showed that by grading the lm thickness of each layer along the length of a multilayer optical element and bending it to a parabola or ellipse, either parallel or focusing beams may be produced. This optic is now called a Go $ bel mirror. Polycapillary x-ray optics use hollow capillaries as waveguides for x rays, where multiple total external reection of x rays from the smooth inner walls of capillary channels can be used to redirect an x-ray beam. The curvature of the capillary determines the nal direction of the x-ray beam, and bundles of capillaries may be used to produce parallel or focusing beams. One optical assembly includes a parabolically curved graded multilayer coupled to channel-cut silicon or germanium crystal monochromators. The multilayer produces a beam with divergence of the order of 100 seconds of arc, which is then diracted by the latest generation of two- or four-bounce channelcut crystals. The result is a high-intensity, highly monochromatic beam that allows extremely highresolution measurements in the laboratory. In fact, diraction peak widths as narrow as 0.035m 2 have been measured, which is commonplace when using a synchrotron source but unheard of for a laboratory instrument. A schematic of a parallel-beam diractometer is also shown in Fig. 2 for comparison with the Bragg Brentano geometry. incident x-ray photon directly to an electrical signal by creating electronhole pairs in a doped semiconductor crystal. Solid-state and scintillation detectors are both point detectors because there is no means of determining where the photon collides with the detector. In contrast, one-and two-dimensional position-sensitive detectors sense where the x-ray photons strike the detector and can therefore be used to collect some angular range of data in parallel. Collecting diraction data in parallel signicantly decreases the measurement time, often down to seconds and even microseconds. The common one-dimensional position-sensitive detectors (PSDs) cover angular ranges of some 5m to 120m 2 and work using gas ionization to absorb the photon and a timing signal to detect the arrival of the electrical pulses at a position analyzer. Two-dimensional detectors include gas proportional designs, charge-coupled devices, or image plates to collect a large area of the cone of diracted intensity from a powder specimen. The diracted intensity that is recorded as a cone is then electronically integrated to produce the familiar x-ray powder pattern. Data collection times using two-dimensional detectors are often on the order of seconds or less.

3. Applications of Powder Diraction 3.1 Phase Identication by X-ray Diraction The set of diracted d and I values for any material is as characteristic as a ngerprint is for a human. We must now face the same problem the crime solvers have had to facean observed set of ngerprints is useless unless you have an extensive collection of reference patterns for comparison. Our problem is even more complex in that unknown materials are commonly composed of more than one phase. When this occurs, the observed pattern contains all the diraction peaks of each phase present. Scientists have answered the need for a powder diraction pattern database by creating an organization called the International Centre for Diraction Data (ICDD), which publishes the Powder Diraction File (PDF) in annual installments. The powder patterns published in the PDF include experimentally determined patterns from the literature and calculated patterns from crystal structure databases. They are reviewed by editors and published in book and computer formats (e.g., CD-ROM). There are about 160 000 patterns in the le. An example of phase identication is shown in Fig. 1, which is a pattern for a two-phase mixture. The phases are identied using the PDF entries shown in Fig. 1. A number of manual and computer searching methods have been developed over the years to accomplish phase identication. These methods place

2.3 Modern X-ray Detectors The nature of x-ray detectors and the electronics used with them is an important part of understanding the approaches to, and limitations of, x-ray applications. Most modern x-ray diractometers employ point detectors that collect data through a very ne receiving slit, but one- and two-dimensional detectors are becoming common in powder diraction. The standard point detectors work using a scintillation crystal to convert x-ray photons to blue light, which is then converted to an electrical signal using a photoelectric material. Solid-state detectors are another variety of point detector, and convert the 4

X-ray Diraction an emphasis on the accuracy of the intensities (Hanawalt method ) or the dhkl values (Fink method ) or on the elements present (alphabetical search). Modern highly automated computer search routines use a combination of chemical information, intensity, and peak position, generally coded into heuristic algorithms, to quickly search the database for possible matches. A modern computer search requires only a few seconds, and will often correctly identify at least two phases in any unknown on the rst attempt. 3.3 The Internal-standard Method of Quantitati e Analysis The internal-standard method is the most general of any of the methods for quantitative phase analysis, and is based on elimination of the absorption factor ( \)m, by dividing Eqn. (5) by itself, giving Ii X lk Ijs Xs (6)

3.2 Quantitati e Phase Analysis Quantitative phase analysis by x-ray powder diffraction dates back to 1925 when Navias at GE quantitatively determined the amount of mullite in red ceramics, and to 1936 when Clark and Reynolds reported an internal-standard method for mine dust analysis. In 1948 Alexander and Klug presented the theoretical background for quantitative analysis based on diracted intensities. Since then there have been numerous methods developed, based on their basic equations. Methods applicable to a wide range of phases and samples include the method of standard additions (also called the spiking method) the absorption diraction method, the internal-standard method, and the Rietveld method (described below). Quantitative analysis using any method is a dicult undertaking. It requires careful calibration of the instrument using carefully prepared standards and will require a minimum of a few days and often a week of setup time. After the technique and standards have been established, routine analyses may be completed in as little as a few minutes to an hour per sample, depending on the precision desired. The intensity of a diraction line i from a phase in a specimen containing a mixture of phases can be rewritten from Eqn. (3) as Ii l K iX E G H F (5)
m

which is linear in the weight fraction of phase . The subscripts i and j refer to dierent hkl reections. k is the slope of the internal standard calibration curve: a plot of Ii\Ijs vs. X\Xs. The addition of a known amount, Xs, of an internal standard to a mixture of phases, which may include amorphous material, permits quantitative analysis of each of the components of the mixture by rst establishing the values of k for each phase (i.e., the slope of the internal standard plot of (Ii\Ijs) Xs vs. X), from standards of known concentration. Use of a pre-established calibration curve, dening the slope k, permits the weight fraction of any phase in the original mixture to be computed.

3.4 Unit Cell Indexing The indexing of a unit cell with unknown lattice parameters used to be a formidable problem. Indexing is the procedure used in assigning hkl values to the set of planes which give rise to each diraction line, or, in other words, determining the lattice parameters that dene the unit cell. The problem becomes more dicult as one goes to lower symmetry because the number of unknowns one must determine increases. Nonetheless, computer methods can often be used to determine the lattice parameters of an unknown crystal if, and only if, the patterns are measured very accurately and corrected for systematic errors. The problem of indexing can then be stated as # equation for up six unknown having to solve the d* hkl # equation using lattice parameters. Solving the d* hkl experimental data will result in some amount of error, and it is convenient to have a means of quantifying the goodness of t of the experimental data to the unit # equation. A gure of cell determined by solving the d* hkl merit (FOM) is used to determine the goodness of t. The most common FOM was developed by Smith and Snyder in 1979 and is dened as FN l N Q 2 Q Nposs (7)

where Ki l a set of constants describing the diracted intensity, including the Thompson equation, polarization, etc., X l weight fraction of phase , l density of phase , and ( \)m l the mass absorption coecient of the mixture. ( \)m is calculated using tabulated values of ( \) for each element as a weighted average. The fundamental problem in quantitative analysis lies in the ( \)m term in Eqn. (5). To solve for the weight fraction of phase , we must be able to compute ( \)m and this requires a knowledge of the weight fractions of each phase. Special techniques are therefore required to eliminate the unknown (\)m and quantify the phase concentrations.

where N is the number of observed lines used in the calculation, Nposs is the number of possible diraction peaks allowed by the space group out to the 2 of line 5

X-ray Diraction N, and Q2Q is the average error between the observed and calculated 2 values of the rst N lines in the pattern. This gure of merit is useful both in evaluating the quality of powder patterns and in establishing the correctness of an indexing. The advantage for indexing is that its value is independent of the crystal system, so a common scale can be applied to all crystal systems. With only one unknown lattice parameter the problem of indexing a cubic powder pattern is always solvable. Lower symmetry crystals are signicantly more dicult to index, and three fundamental approaches have been useful. Since the problem gets progressively more dicult numerically as we decrease symmetry, all three approaches employ computers. The three approaches are (i) intelligent algorithms, (ii) searches of solution space and (iii) searches of index, or (hkl ), space. The intelligent procedures try to determine the lattice parameters by applying known relationships from both geometry and crystallography and even previous experience with indexing problems. Two quite successful methods of this type have been developed by Visser and Werner. Search methods do a systematic search of solution space (i.e., look at all possible solutions within some boundaries). There are two types of searching index procedures: those that work in parameter space and those that examine index or hkl space. Some of the search methods make assumptions about the regions of solution space that should be examined while others are exhaustive. We should pause here to discuss the concept of a solution space. This space for an exhaustive parameter search is envisioned as having up to six dimensions in a plane (corresponding to functions relating to a, b, c, , , and ) and the FN gure of merit plotted perpendicular to this plane. For cubic problems this means a two-dimensional problem: vary all possible values of a along the x axis with FN plotted vertically. One needs to use caution, however, because there are a great number of incorrect cells that will almost describe the crystal within reasonable error limits. Each of these solutions will correspond to a maximum of FN in solution space. Our assumption is that the highest maximum in FN corresponds to the correct unit cell. An exhaustive indexing procedure working in parameter space examines all possible values of a, b, c, , , and within some reasonable limits dened by the user. The lattice parameters are divided into small increments and each possible set is used in an attempt to index the unknown pattern. Loue $ r and Loue $ r in 1972 established an extremely successful program based on this approach. Exhaustive procedures in index or hkl space examine all possible combinations of h, k, and l which can be used to index each of the lines in the unknown pattern. Here our solution space is four dimensional for all symmetries: axes in integer units of h, k, and l in a 6 plane with FN perpendicular to that plane. A very powerful algorithm based on hkl space searching was developed by D. Taupin in 1973. Before the days of powerful, fast PCs, exhaustive search procedures were only practical in relatively high symmetry systems and in a few exceptional cases. For example, the classic case of a manual orthorhombic indexing was performed by Jacob and Warren in 1937 in the determination of the crystal structure of -uranium. As the symmetry decreases to triclinic, the amount of computation required became impractical. Today, however, monoclinic and triclinic problems are routinely solved on researchers desktop workstations.

3.5 Rening Accurate Lattice Parameters Once a powder pattern is indexed we may routinely # equations with all terms write from one to six d* hkl known except the one to six lattice parameters. These equations may be solved simultaneously to give values for the lattice parameters, but these will not be the most precise values. The precise and accurate measurement of lattice parameters, or unit cell dimensions, have been used for various purposes such as: determining thermal expansion coecients, nding the true density of a material, providing a direct measure of interatomic distances, and studying interstitial and substitutional solid solutions. In order to obtain accurate lattice parameters, the minimization and elimination of a number of errors is necessary. All sources of error in 2 aect the accuracy of the lattice parameters, and may be classied into two categories: (i) random errors and (ii) systematic errors. Random errors are those made in locating and recording the center of the diraction peak. These errors vary randomly with the angle 2. Systematic errors are those whose magnitude depends upon the position of the line, that is, upon the angle 2. Systematic errors are best eliminated by using an internal standard, which is a material certied by NIST to have well-known diraction peak locations. With systematic error handled in this manner, the best approach for obtaining the most precise parameters from observed data is to use the method of least squares. Least squares minimizes the dierences between the observed dhkl values and dhkl # equation by varying the values calculated from the d* hkl lattice parameters. This very powerful method will simultaneously produce estimates of the standard deviations of the lattice parameters.

3.6 Preferred Orientation Preferred orientation of crystallites in polycrystalline materials is a vital subject to many materials industries.

X-ray Diraction A number of industrial materials are based on directional physical properties of the crystallites. For example, barium hexaferrite ceramic magnets, used commonly as seals on refrigerator doors, are polycrystalline materials in which only the (00l ) crystallite directions have a magnetic moment. Thus, the fabrication and quality control procedures must involve manipulating and measuring the degree of preferred orientation in the ceramic. Extruded wires show a characteristic preferred orientation as do most rolled metals, pressed powders, and thin-lm materials. The most common way of evaluating the type and extent of preferred orientation is to measure the pole gure for a particular crystallographic direction. The pole gure is simply the intensity of a particular Bragg diraction peak plotted as a function of the three-dimensional orientation of the specimen. It is determined on a pole gure diractometer which is essentially the same as a single-crystal diractometer, able to rotate the specimen through all orientations while monitoring the diraction intensity of a reection. The results are displayed as a pole gure which is a two-dimensional stereographic projection. By collecting several pole gures, a complete orientation distribution function (ODF) can be calculated, which describes the orientation of the crystallites in the coordinates of the specimen geometry. The full ODF is often critical for understanding the variation in properties as a function of direction in monolithic parts. ally, as the crystallite gets smaller, the destructive interference near the Bragg angle decreases and the diraction peak becomes wider. The crystallite size broadening ( ) of a peak can usually be related to the crystallite size () via the Scherrer equation. l k\ cos (8)

The additional broadening in diraction peaks beyond the inherent peak widths due to instrumental eects can be used to measure crystallite sizes as low as 1.0 nm. However, a second cause of broadening, due to microstress within the crystallites, can complicate the picture. Whatever the cause of microstress in a crystallite, the eect will cause a distribution of d-values about the normal, unstrained, dhkl value. The broadening in a peak due to stress has been shown to be related to the residual stress, , by l 4 tan (9)

3.7 Crystallite Size When a well-crystallized powder sample with an average crystallite size of $ 10 m diracts, the diffraction peaks are sharp, symmetrical, and well dened. However, if the specimen contains crystallites that are smaller than $ 500 m, possess large internal microstrains, or have a large density of stacking faults or other defects, the diraction peaks will be broadened and\or asymmetric. The simple case of crystallite size peak broadening can be understood by considering the interference of x rays scattered from dierent positions in the crystallite. A large crystallite, i.e., larger than a few micrometers, contains some 50 000 successive (hkl ) planes and produces a narrow diraction peak. The diraction peak is narrow because at angles slightly above and below the Bragg angle, complete destructive interference occurs. A photon scattered from the top (hkl ) plane will be exactly out of phase with a photon scattered from an (hkl ) plane deep within the crystallite; for example 40 000 planes from the surface, at angles very close to the Bragg angle. However, if the crystallite is smaller than the 40 000 planes used in our example, complete destructive interference will not occur and some diracted intensity will be observed at angles slightly dierent from the Bragg angle. Natur-

The fact that stress-induced diraction peak broadening follows the tangent of while crystallite size broadening follows 1\cos allows us to separate these eects. The most common procedure for accomplishing this was developed by Warren and Averbach in 1950. In this method, the Fourier transformation of the diraction peaks of a broadened sample and a standard material exhibiting no sample broadening are computed. A deconvolution procedure is then carried out in Fourier space to determine the peak broadening that results from crystallite size and\or microstrain. A signicant amount of research is underway to determine appropriately the microstructural features of a powder or polycrystalline body from powder diraction data. Most of these approaches involve rather detailed computations of the peak widths as a function of diraction angle. The angular dependence of the peak width is then used to quantify the crystallite size, microstress, stacking faults, and other atomicscale defects present in a specimen.

3.8 Residual Stress If the stress in a monolithic body is uniformly compressive or tensile, it is called a macrostress and the distances within the unit cell will become smaller or larger, respectively. This will be observed as a shift in the location of the diraction peaks. Essentially, the individual crystallites in a polycrystalline body act as strain gauges, where the diraction peak location indicates the strain value. Since most stresses in materials are a result of thermal cycling, machining, etc., x-ray measurements of macrostress are normally referred to as residual stress measurements. 7

X-ray Diraction Careful measurement of the locations of diraction peaks in several orientations of the specimen directly provides the strain present in the specimen as a function of direction. By using a goniometer that allows the specimen to be rotated in three directions, one can determine a full triaxial strain distribution. Given the diraction elastic constants, the full secondrank stress tensor may be determined. Like lattice parameter renements, residual stress measurements are prone to peak location errors and special care must be taken to ensure instrument alignment. When using a BraggBrentano diractometer, stress measurements are best done using highangle peaks to minimize systematic errors. Typically, a stress-free powder specimen is measured to conrm the instrument performance. Parallel beam measurements oer the advantage of reduced systematic peak location errors and allow thin-lm and stress depth proling measurements. Surface and near-surface stresses are important in most technological applications because surfaces and lms often dene the performance of materials. served prole. The Pearson VII function also allows for the variation in shape from Lorentzian to Gaussian and can rather easily be split to describe asymmetric x-ray diraction peaks. Another approach for distributing intensity in a prole is called fundamental parameters convolution. Instead of using a mathematical function to approximate the shape of each diraction peak, the fundamental parameters approach builds each diraction peak by convolution of the x-ray emission prole with the instrumental and sample contributions to the prole shape. Hence, an accurate peak prole is obtained that explicitly includes all aspects of the experiment and sample; the x-ray source size, receiving slit width, beam divergence, crystallite size, etc. The development of the fundamental parameters method is underway for tting individual peaks and for use in Reitveld renement. In a typical Rietveld renement, individual scale factors (related to the concentration of each phase) and prole, background, and lattice parameters are varied. In favorable cases, the atomic positions and site occupancies can also be successfully varied to obtain the most correct models of the structures of the phases. Preferred orientation, polarization, and systematic errors may also be modeled. Figure 3 includes the results of a completed Rietveld renement. The observed and calculated patterns are shown, along with the dierence pattern and the peak position markers. The renement in Fig. 3 includes all variables except the atomic positions and thermal parameters, i.e., the background, prole shapes, preferred orientation, sample displacement, and scale factors. Thus the Rietveld method is used to describe all of the features of the specimen, from the crystal structure to full orientation distribution functions in the case of monolithic polycrystalline specimens. It is the most accurate method available for quantitative analysis and provides estimated uncertainties for each of the rened values.

3.9 Prole Fitting and Riet eld Analysis The most exciting of the prole tting developments has been the continuing development of a whole pattern x-ray renement method to obtain crystal structure information, begun by Rietveld in 1969, who applied it to neutron diraction. Rietveld renement is conducted by tting a calculated diraction pattern to the observed data by adjusting each of the variables that describe the diraction pattern. Rietveld computations are relatively intensive, and require minimizing the sum of the weighted, squared dierences between observed and calculated intensities at every step in a digital powder pattern. The Rietveld method requires knowledge of the approximate crystal structure of all phases present in the specimen. The input data to a renement include the space group, atomic positions, site occupancies, and lattice parameters. This method goes beyond the simple computation of intensities from Eqn. (3) by distributing the calculated intensity over the kind of prole found using a particular instrument. Appropriately describing the shape of each peak in an x-ray pattern remains a challenge, but several successful analytical prole shapes have been developed. Of interest are the Voigt, the pseudo-Voigt, and the split Pearson VII prole functions. The Voigt function is the result of an analytical convolution of a Gaussian and a Lorentzian. It therefore ranges from pure Lorentzian to Gaussian type, depending on the ratio of both half-widths. The pseudo-Voigt conveniently allows the renement of a mixing parameter determining the fraction of Lorentzian and Gaussian components needed to t an ob8

3.10 In Situ Analysis In situ powder diraction methods can be used to characterize complex systems under controlled conditions of temperature, atmosphere, pressure, applied electric or magnetic eld, mechanical loading, or time. Powder diraction lends itself well to both basic science and materials processing problems. Crystal structures, phase equilibria, anisotropic thermal expansion tensors, volume expansion in response to gaseous atmospheres, and reaction kinetics are examples of fundamental studies which are enormously simplied using in situ diraction. In situ diraction is quite often the only means to determine structures and phase equilibria in systems that cannot be quenched to ambient conditions; for example, in high-pressure

X-ray Diraction

1.0 Counts (104)

0.5

0.0

2.0

3.0

4.0 2 ()

5.0

6.0 X10E 1

Figure 3 Rietveld renement of barium zirconate powder.

geological studies or peritectic melting systems that crystallize on quenching. Typical materials science problems can often be investigated using laboratory in situ instruments tted with furnaces, helium displexes for cooling, and controlled-atmosphere chambers. Generally, in situ measurements involve multiple measurements as a function of an external variable and position-sensitive detectors that reduce the data collection time are convenient. Reaction kinetics can also be investigated using one- and two-dimensional detectors. Most high-pressure studies incorporate diamond anvil cells to apply pressures up to several hundred kilobars, often at high temperature, and are typically conducted using synchrotron sources. An elegant example of an in situ measurement is shown in Fig. 4 where a peak of a PZT piezoelectric was recorded dynamically, into 120 dierent banks of MCA channels, which were addressed according to the instantaneous value of the electric eld by Zorn and his colleagues in 1985. The full electrostriction tensor was derived from measurements like these. 3.11 Structure Solution Ab initio structure solution from powder diraction data has become more popular as a result of our abilities to extract intensities from powder patterns. The integrated intensities from a powder pattern are treated as single-crystal intensities and the traditional single-crystal structure solving and renement procedures can be employed. The one-dimensional nature of a powder pattern, however, complicates the extraction of intensities because of overlapping reections. Several pattern

332 332 323 332

332 323

123 124 125 Parallel to electric field

123

124 125 2h Perpendicular to electric field

Figure 4 Dynamic XRD of the 400 peak of a PZT as a function of electric eld.

decomposition methods have been developed, including individual prole tting, the Pawley method, and the LeBail method. The former two methods involve tting peak proles with positions that are not constrained by the lattice parameters, while the LeBail 9

X-ray Diraction method relies on the lattice parameters to constrain the peak positions. After extracting the intensities for each reection in a powder pattern, traditional methods such as the Patterson method or direct methods can be used to phase the crystal, and Rietveld renement is then employed to rene the structure. Structure solution from powders has been quite successful for many inorganic materials with small unit cells, and is an emerging technique for larger organic structures.
Cullity B D 1978 Elements of X-ray Diraction, 4th edn. Addison Wesley, Reading, MA Hahn T (ed.) 1983 Space-group symmetry. International Tables for Crystallography. Reidel, Dordrecht, The Netherlands Jenkins R, Snyder R L 1996 An Introduction to X-ray Diffractometry. Wiley, New York Klug H P, Alexander L E 1974 X-ray Diraction Procedures, 2nd edn. Wiley, New York Noyan I C, Cohen J 1998 Residual stress: Measurement by Diraction and Interpretation, Springer, New York Snyden R L, Fiala J, Bunge H-J 2000 Defect and Microstructure Analysis by Diraction. University Press, Oxford Stout G H, Jensen L H 1968 X-ray Structure Determination. Macmillan, New York Young R A 1993 The Riet eld Method. Oxford University Press, Oxford

Bibliography
Bish D L, Post J F (eds.) 1989 Modern Powder Diraction. Reviews in Mineralogy, Vol. 20. Mineralogical Society of America, Washington, DC

S. T. Misture and R. L. Snyder

Copyright ' 2001 Elsevier Science Ltd. All rights reserved. No part of this publication may be reproduced, stored in any retrieval system or transmitted in any form or by any means : electronic, electrostatic, magnetic tape, mechanical, photocopying, recording or otherwise, without permission in writing from the publishers. Encyclopedia of Materials : Science and Technology ISBN: 0-08-0431526 pp. 97999809 10