13R Weil and K Sheppard Mater Char Vol 28 1992 p 103112

14. D.S. Rickerby and S.J. Bull, Surf Coat. Technol, Vol 39/40, 1989, p 315 Experimental Techniques for Microstructural Characterization

It is not possible to determine all the structural features of a material using any one analysis technique. Therefore, many techniques that can be used to provide information about specific microstructural features have been developed. It is important to know what factors are important in any application before undertaking a program of structural analysis.

For coatings, many of the techniques are applicable, but somewhat more care must be taken in preparing representative samples, because the volume of material to be investigated is relatively small. Table 1 identifies the most important microstructural features and defines what techniques can be used to assess them. However, other, more unusual techniques can also provide the same information.

Table 1 Techniques for microstructural analysis

Coating property


Phase composition

X-ray or electron diffraction

Phase distribution

Metallographic sections, SEM, TEM, optical microscopy

Grain size

X-ray diffraction, SEM, TEM (plus image analysis)

Grain shape

SEM, TEM (plus image analysis)

Preferred orientation

X-ray diffraction, TEM

Surface morphology


Note: SEM, scanning electron microscopy; TEM, transmission electron microscopy

Note: SEM, scanning electron microscopy; TEM, transmission electron microscopy

Brief descriptions of some of the most frequently used techniques are described below, as are some of the factors that need to be considered when applying them to coated systems.

Metallography. Most standard metallographic preparation techniques can be applied to coatings, and polished cross sections are commonly used to determine coating thickness and uniformity. Samples are generally cut from a coated component using a low-deformation saw and are mounted in a resin medium prior to polishing. Components should be rigidly clamped and sectioned so that the abrasive wheel enters from the coating side. For most coating-substrate combinations, a hot mounting medium is preferred, but for temperature-sensitive or porous coatings, better results are obtained when a cold-setting resin is used. The section should then be ground back until a flat surface is produced. When dealing with coated samples, it is often observed that some chipping of the coating occurs next to the saw cut. It is therefore essential to grind the section well back to remove any damage prior to polishing. For thin, hard coatings, a large grit size should be avoided during grinding, because it can result in coating detachment (Ref 15). The preferred direction of grinding is from the substrate into the coating because this maintains the best edge sharpness.

After grinding, the section can be prepared using metallographic polishing techniques similar to those used for bulk materials. However, for porous plasma-sprayed coatings, better results can be achieved using more advanced metallographic techniques, as described in Ref 16. The twofold aim of the polishing process is to remove the damage introduced during grinding and to produce the high surface finish needed for metallography. For thin coatings, it is particularly important to use an edge-retaining mount, because edge rounding of the sample can occur during polishing in cases of differential polishing of the coating and the substrate or of sample tilting. For very thin or fragile coatings, the sample can be encapsulated with a thick protective layer prior to sectioning and mounting in order to prevent damage of the coating during preparation. Porous coatings, such as air plasma-sprayed coatings, are best mounted by vacuum infiltration, because the resin can then penetrate the porosity in the coating and will prevent collapse of the coating around the porosity during polishing (Ref 16).

For measuring the thickness and uniformity of the coating, a simple polished cross section may be sufficient. However, particular care must be taken to ensure that the coating has not been damaged during preparation. The importance of grinding back from the cut surface cannot be emphasized enough. Similarly, the porosity revealed in the coating may be an artifact of the preparation process, because grain pullout often occurs in the final stages of polishing materials that have weak boundaries, such as air plasma-sprayed metals. Although there is no simple technique for avoiding these problems, the measurement of thickness or porosity by other techniques can give some indication of whether the measurements obtained from metallography are consistent. A material-specific preparation technique can then be developed for the coating material and process combination of interest.

When dealing with very thin coatings, more detail can be revealed by using a taper section. A taper angle of 5.7°, for example, will give a 10* magnification in the plane of the section. In this case, the sample is mounted in the resin with one edge supported by a spacer, the height of which is chosen to give the required tilt angle, and then prepared in the same way as a cross section (Ref 15). Figure 4 shows a taper section of a PVD titanium nitride coating on stainless steel with a 400 nm (16 pin.) titanium interlayer that is not visible in the polished cross section, but which can be easily resolved in the taper section. Because of nonuniform polishing, the initially planar interface has become irregular in the taper section. This is almost impossible to avoid in practice.

Fig. 4 Micrographs of a PVD titanium nitride coating on stainless steel. (a) Cross section. (b) Taper section showing the presence of a 400 nm (16 pin.) titanium interlayer

Fig. 4 Micrographs of a PVD titanium nitride coating on stainless steel. (a) Cross section. (b) Taper section showing the presence of a 400 nm (16 pin.) titanium interlayer

The grain structure of the substrate and coating and, sometimes, even the presence of the coating itself can be revealed by etching with an appropriate reagent, as is done for bulk materials. Because it is often difficult to choose an etch that will attack both coating and substrate, a multistage etching process may be needed. It is important to determine whether an etch used for the coating will badly attack the substrate, because that could lead to the observation that there is interfacial porosity, when this is really just an artifact of etch selection. Figure 5 shows the structure of a PVD Fe-Cr-Al-Y coating on stainless steel after 100 h of oxidation at 1000 °C (1830 °F). The structures of both coating and substrate have been revealed by electrolytic etching in CrO3/H2O. Image analysis can be used to determine grain size and shape, along with porosity, in a quantitative manner (Ref 16).

Fig. 5 Polished cross section of a Fe-Cr-Al-Y coating on austenitic stainless steel after 100 h at 1000 °C (1830 °F), electrolytically etched in CrO3/H2O to reveal grain structure of coating and substrate

X-ray Diffraction. In most cases, diffraction analysis is used to identify the structure of a deposited film from a list of known possibilities, rather than to identify the structure of a completely unknown or new substance. The analysis is thus essentially limited to the comparison of an experimentally observed diffraction pattern with patterns of substances, the structures of which are already known. This is made considerably easier if the composition of the coating can be determined, because the number of possible structures is then limited. In a typical x-ray diffraction experiment, a monochromatic beam of x-rays, wavelength X, hits the coated surface from which it is diffracted. The intensity of the diffracted beam is recorded as a function of the diffraction angle, 9. The intensity of the beam will be nonzero only at those diffraction angles at which the Bragg condition is satisfied (Ref 17):

where dhkl is the spacing of the lattice planes with Miller indices hkl. All elements or compounds consist of atomic arrays with a unique combination of geometry and spacing. The smallest repeatable unit that completely defines the structure is called the unit cell, and the spacing of these unit cells is defined by one or more lattice parameters (depending on crystal structure). The position of the diffraction maxima gives information about the size and shape of the unit cell, whereas the width of the maxima can be used to evaluate the size, orientation, and strain in grains of polycrystalline materials.

The intensity of the diffracted beam depends on several factors:

• Structure of the material

• Volume of irradiated material

• Diffraction geometry

• Sample alignment

For a more detailed discussion of the x-ray diffraction technique, readers are referred to Ref 17.

Once an experimental diffraction pattern has been obtained, the position and intensities of the peaks can be easily measured. It is then possible to calculate the intensities and positions for all candidate structures and compare these to the measurements to determine the structure of the film. However, a much easier way is to rely on powder diffraction standards (Ref 18), which represent a collection of experimentally determined relative intensity values for a large number of substances. These are available in book, microfilm, or card form, and can also be retrieved from disks or data bases. A typical card, for example, contains information on the structure of the material as well as the relative density data (Fig. 6). Information is available for both stoichiometric and nonstoichiometric compounds.











Titanium Nitride


Dla. Cutoff CoLL

l/l-j Ditfracto meter

Ref. Beadle and VerSnyder, Trans. ASM 45, 397 (1953)

75 100 55 25 15

Dla. Cutoff CoLL

l/l-j Ditfracto meter

Ref. Beadle and VerSnyder, Trans. ASM 45, 397 (1953)

2V Dx5.4mp


Color: Vu How

Coritalra trace of Zr dA

75 100 55 25 15

a 12

20 20


400 331 420 422 333,511

Fig. 6 Powder diffraction standard card for stoichiometric titanium nitride

The data on peak positions for titanium nitride (from Fig. 6) are represented as vertical lines in Fig. 7, which also shows an experimental diffraction pattern obtained for a titanium nitride film produced by sputter ion plating. It is clear that the observed peaks line up with the standard peak positions, but the measured pattern is slightly shifted to higher angles because of the presence of residual stress within the coating. The intensities, however, differ considerably from the values of the card shown in Fig. 6 because of the preferred orientation of the film. To fully assess the preferred orientation, it is necessary to compare the intensity of the measured line with that of the standard. A texture coefficient, T*, can be defined as:

In this equation, h k T are the Miller indices of the plane of interest, I is the measured intensity, and I0 is the intensity of the standard, which is randomly oriented. A value of unity corresponds to a random texture, and values greater than this show an increasing level of preferred orientation. For the data presented in Fig. 7, a {111} preferred orientation is observed, because ^*{iii> = 4.

Fig. 7 Measured x-ray diffraction pattern for sputter ion plated titanium nitride. Vertical lines represent data peak positions for titanium nitride from Fig. 6. Film shows a [111] preferred orientation. C/s, counts per second

Once the phase composition of the coating has been identified, lattice parameter measurements can be undertaken. This is particularly important if measurements of residual stress are to be made or if there is a need to assess either the stoichiometry of a material or the composition of a solid solution. Because the accuracy of measurement increases as the Bragg angle increases, high angle reflections of sufficient intensity should be used for stress measurements. For the accurate measurement of peak position, it is necessary to fit a function to the peak profile (Ref 19). It is often assumed that a Gaussian profile provides a good description of peak shape, but, in many cases, a Pearson 7 function will provide a better fit. If the film is under stress, then measurements of the lattice parameter from a single reflection can be misleading, because of anisotropy constants. It is therefore important to use an extrapolation function, such as the Nelson-Riley or cos9 cot9 function, to improve reliability (Fig. 8). In this operation, the lattice parameter determined for a number of reflections is plotted against cos9cot9, and a line through the points is extrapolated to 9= 0° to determine a0. The deviations of some reflections from the line that are due to pseudo macrostresses can be revealed by this approach (Ref 19), which has the added advantage of minimizing alignment errors.



SS 0.424 n




SS 0.424 n



o As d




{422) o" (400)

<3t1) -

(200) (111)

16.69 E es









Fig. 8 Variation of lattice parameter with cos6cot6 for sputtered titanium nitride in order to extrapolate a lattice parameter, a0, corrected for measurement errors

The width of a diffraction line is determined by the grain size and strain in a material, as well as by instrumental factors. A typical diffractometer will have an angular resolution of approximately 0.01°, which means that line broadening can be observed if the grain size D is lower than approximately 1 pm (40 pin.). This is easily achievable by various coating techniques, and x-ray diffraction can be used to give a measure of grain size for these technologies, averaged over the penetration depth of the x-ray (Ref 20).

If pis the full width at half-maximum intensity of the diffraction peak (for the case of a Cauchy peak shape, a2 stripped and corrected for instrumental broadening), then (Ref 19):

where s is the average strain in the material. Thus, by plotting PcosBA. against sinBA., which is the so-called Hall-Williamson plot (Ref 21), it is possible to determine both D and s. Figure 9 compares the Hall-Williamson plots for titanium nitride coatings produced by several deposition technologies. The broadening is dominated by strain, because the intercept is close to zero, implying a grain size greater than 1 pm (40 pin.).

Fig. 9 Hall-Williamson plots for physical and chemical vapor deposited titanium nitride, showing increased strain in the physical vapor deposited films

Electron Microscopies. Perhaps the most useful techniques for characterizing thin films are scanning electron microscopy and transmission electron microscopy, because they can be used to investigate morphology, crystal structure, grain structure, and porosity. Both techniques depend on good sample preparation. Samples are relatively easy to produce for the former technique, but they are time-consuming and often difficult to prepare for the latter technique.

Scanning Electron Microscope (SEM). Numerous signals are available for SEM imaging (Ref 22). When coupled with energy-dispersive x-ray spectroscopy (EDS), the SEM can be used to obtain a wide range of information about surface topography, composition, crystallography, and electronic properties (Table 2). For example, the SEM can be used to assess the grain size, packing density, and porosity of a coating if a fracture cross section is imaged. The chamber of many SEMs is large enough that many components can be imaged nondestructively, and the surface topography induced by surface treatment, together with damage to the coating (after deposition or after service) can be determined.

Table 2 Scanning electron microscope imaging modes


Detected emission

Best resolution

Information available


Secondary electrons

5-15 nm (0.2-0.6 pin.)



Primary electrons (back scattered)

50-250 nm (2-10 pin.)

Topographic, atomic number


Specimen current

>1 pm (>40 pin.)

Topographic, defects in some materials


X-ray fluorescence

>1 pm (>40 pin.)

Chemical analysis (Z >5)



>1 pm (>40 pin.)

For some materials, quantitative chemical analysis or defect imaging


Electron channeling

0.05° from 1 pm (40

Lattice parameter selected area and defect density


In a typical SEM, an electron beam from a tip or filament is accelerated by the anode (typically, by 5 to 30 kV) and focused by two magnetic lenses to a fine spot on the sample. At the same time, the beam is scanned in the x and y direction. The electrons that are emitted from the sample are then either collected by a back-scattered electron detector, or, more generally, by a detector for secondary electrons. After amplification, the signal is fed to a cathode-ray tube, which is deflected in the same manner as the microscope electron beam. The magnification of the image is fixed by the x-y deflection coils.

The primary electron beam generates a tear-drop-shape interaction volume with the specimen. Elastically reflected electrons emerge from the surface with energies close to those of the primary beam. The proportion of back-scattered electrons depends on the material. As the atomic number of the element increases, the back-scattering coefficient increases and such atomic number contrast can be used to distinguish between phases in coatings and substrates. However, back-scattered electron emission is highly directional, and strong topographic contrast is visible for rough surfaces, depending on the position of the detector.

Secondary electrons leave from the near-surface region of the sample when energies are less than approximately 50 eV and show no such directional effect, because they are collected by a positively biased detector from all points on the sample. Their emission intensity is not strongly materials dependent. The secondary electron image clearly shows topographic features, such as cracks in electroplated coatings (Fig. 3). However, it should be realized that these images should not be interpreted like reflected light micrographs, because shadows are not generally caused by topography effects.

The primary electron beam can excite core electrons in the sample, which relax by the emission of an x-ray, as is characteristic of the excited atom. Such x-rays typically escape from 0.5 pm or greater depths. X-ray analysis can be achieved by energy-dispersive or wavelength-dispersive spectroscopies (EDS or WDS, respectively). The EDS system generally uses a lithium-doped silicon detector in conjunction with a beryllium window for analysis of elements with atomic number, Z, greater than 9 (i.e., neon and heavier atoms), although the range can be reduced to Z = 5 (boron) with a windowless detector. The WDS method uses a range of crystals with different lattice parameters that can be rotated to focus x-rays from elements where Z > 6 into a detector. The EDS system is faster than WDS but has poorer energy resolution. Both EDS and WDS can be quantitative methods of evaluating elemental composition if care is taken in correcting for instrumental, chemical, and physical factors. This is a standard feature in most commercial analysis systems.

Fracture sections provide a very good means for assessing the structure of thin PVD coatings. Figure 2 shows fracture sections through PVD tungsten coatings deposited under unbiased conditions (zone 1), with an applied bias voltage of -150 V (zone T). The columnar structure is clearly visible for both coatings, but the columnar packing density increases with applied substrate bias (i.e., energy of ion bombardment). Because each column consists of an array of grains, a true measure of grain size is not generally possible using the SEM, but an indication of relative grain size may be possible from the column dimensions. The coatings are deposited onto a ferritic steel substrate that was cut through from the back side with a saw until it reached approximately 1 mm (0.04 in.) below the coating. The sample was then immersed in liquid nitrogen to cool below the substrate ductile-brittle transition temperature. After 10 min, the sample was removed and half placed in a vise. Then, the free end was struck with a hammer to fracture the sample in two. The coated side was struck so that the section was not put into compression during fracture, in order to avoid excessive coating damage. The cracks started in the saw cut and ran through the coating without damaging it. Similar results can be achieved for coatings on brittle substrates, such as silicon wafers, by using a diamond scribe to initiate fracture on the back of the substrate.

Insulating substrates or coatings will become charged in the electron beam, which quickly reduces image quality and occasionally damages the material. Charging can be avoided by depositing a thin carbon or gold film on the specimen surface. However, this will reduce the accuracy of any subsequent microanalysis and should only be used if necessary. Alternatively, a low accelerating voltage can be used for the primary beam, because when voltages are less than 3 kV, the sum of the secondary and back-scattered electrons generated can equal the number of primary electrons, and no charging occurs. Such low-voltage SEMs have the added advantage of better resolution, but their poorer signal-to-noise ratio means that some image processing may be necessary to obtain the best results.

Relatively poor resolution can also be encountered with magnetic coatings or substrates. For high-resolution work, it is recommended that a nonmagnetic substrate, such as silicon or stainless steel, be used wherever possible.

Transmission electron microscope (TEM) studies require very thin specimens (5-500 nm, or 0.2-20 pin.), depending on the material. These specimens of foil must be prepared from the bulk of the coating or substrate. With such foils, it is possible to visualize structures on the nanometer level, and when specialized techniques are used, it is even possible to see the presence of single-atom columns and to locate dislocations in the crystal (Ref 23). Grain boundaries and interfaces can also be studied using this technique. When the crystallites range in size from micrometers to nanometers, the TEM can be used for electron diffraction studies to identify phase composition, as well (Ref 24).

The preparation of TEM samples is a skilled operation, and considerable effect is often required to develop the technique for a specific coating-substrate system. Plan-view TEM samples are by far the easiest to produce and consist of sections of the coating parallel to the interface (Fig. 10). In general, plan-view sample preparation takes place in two stages. The first is mechanical thinning, in which a parallel slice of material is cut from the coating and machined to a 3 mm (0.12 in.) disk using a punch or ultrasonic drill. The disk is then polished to approximately 0.1 to 0.2 mm (0.04 to 0.08 in.) in thickness. Finally, the center of the disk is further reduced in thickness by dimpling, that is, grinding with a spherical abrasive tool, which leaves a depression in the center that can be 20 pm (800 pin.) thick.

Fig. 10 Plan-view transmission electron microscope images of sputtered titanium nitride coatings. (a) Bright-field image. (b) Dark-field image obtained by putting an aperture over two bright {200} diffraction spots. (c) Corresponding diffraction pattern

The second stage involves final thinning to perforation, which occurs by either electropolishing, using a fine jet of etchant, or ion beam milling, using an argon ion beam of approximately 3 kV, which is directed at the surface at a low angle (5-20°). For metals and semiconductors, the electropolishing route is preferred, whereas for ceramic materials, ion beam milling is usually necessary. Depending on whether the area of interest in the coating is near the substrate interface or further out in the film, thinning of the sample will occur on the film side or the substrate side, respectively, before final perforation is achieved by thinning from both sides. Using this approach, a section can be produced from any level within the coating, but it is difficult to ascertain how far from the interface any plan-view sample lies.

Another method used to produce foils is to deposit a sufficiently thin coating onto a soluble substrate, such as aluminum (soluble in sodium hydroxide) or sodium chloride (soluble in water). After dissolution of the substrate, the film is transferred to a copper grid for imaging. However, in some cases, the film will collapse when the substrate is removed, leaving a fine powder, which can be captured on a copper grid in order to study individual grains (or columns, for PVD coatings) in more detail, but which gives no information about columnar packing.

More information can be obtained from cross-sectional TEM samples, which are considerably more difficult to produce. Usually, the sample is cut in two and joined, coating side to coating side, with a good epoxy adhesive. The assembly is then cut into thin slices and prepared in a similar manner to plan-view samples. However, the differential thinning rates of coating, substrate, and adhesive mean that it is often difficult to prevent the sample from collapsing during thinning. A number of workers have used this technique to study the interfacial regions of hard coatings in some detail (Ref 25, 26).

A TEM consists of an electron gun (typically, 100 to 400 kV for analytical configurations) and an assembly of lenses, all enclosed in a column evacuated to about 1.3 * 10-3 Pa (10-5 torr). The optical arrangement is similar to that of a light microscope, but additional stages of magnification are used. Condenser lenses collimate the electron beam, which passes through the specimen, and an objective lens is then used to form a first image in the object plane of the first projector lens. This image is magnified about 40 times. A small area from this image is then projected as an intermediate image, magnified about 40 times by the first projector lens. A small area of this image is then projected onto a fluorescent screen or photographic plate by another projector lens. The image is formed from the intensity distribution of the electrons leaving the bottom surface of the specimen. A total magnification ranging from 10,000 to 1,000,000 is achievable.

In addition to imaging, the TEM can be used for diffraction patterns. The diffraction pattern is formed at the back focal plane of the objective lens, and can be imaged by adjusting the projector lens excitation, which is an automatic function for most modern TEMs. The resolution of the TEM increases with electron-beam energy, and is approximately 0.19 nm (0.008 pin.) at 400 kV.

Contrast in the images is generated in two different ways. For amplitude-contrast imaging, an aperture is used to select either the primary transmitted electron beam (bright field) or one of the diffracted beams (dark field). The aperture is placed in the back focal plane of the objective lens, and the beam is focused to form an image with a further lens. Dark-field imaging is particularly useful when several phases are present. The formation of a dark-field image using a diffracted beam from one of the several phases present highlights the locations in the sample where that phase can be found. Microcracks, grain boundaries, stacking faults, and other defects can also be identified using dark-field imaging techniques (Ref 27).

For phase-contrast imaging, the primary and diffracted beams are allowed to interfere with one another. Local phase shifts of the electron waves are created by voids, defects, heavier atoms, and other phenomena. Phase contrast is the principle underlying lattice imaging (Ref 23). In addition, phase-contrast imaging under defocus conditions is used to identify voids or density variations (Ref 14, 28).

The combination of selected-area diffraction with imaging makes the TEM a very powerful tool. Phase identification and morphological characteristics can be achieved on the same areas of a sample. State-of-the-art analytical TEMs can focus the beam to a spot that is only several nanometers in diameter. The use of an x-ray detector on the column allows the analysis of chemical composition with nanometer-range spatial resolution. However, it should be realized that only a very small volume of material is sampled, because of the high magnifications used and the thin samples needed for imaging. Many microstructural variations occur on a larger scale. Therefore, x-ray diffraction is a more appropriate analytical technique for the study of large-scale structural features such as these, because it samples larger volumes of the sample.

Porosimetry. Porosity can be measured by the mercury intrusion porosimetry (MIP) technique, tests that use corrosive gases to decorate defects, density bottle methods, or optical or electron micrographs. These methods measure different types of porosity, namely open porosity for MIP and gas methods, closed porosity for the density bottle methods and total porosity (including sample preparation artifacts) for the microscopy measurements. No technique is particularly suitable when the porosity levels are low or the pores very small. In these cases, only electron micrographs can be used, and these sample a rather small volume of material.

MIP is very suitable for measuring the intersplat porosity for plasma-sprayed coatings. The determination of pore size distribution by MIP is based on the physical principle that mercury will not penetrate fine pores until sufficient pressure is applied to force its entry. The relationship between the applied pressure, P, and the pore radius into which mercury will intrude, r, is:

where y is the surface tension of mercury and B is the contact angle between mercury and the pore wall. Varying the externally applied pressure results in changes in the intruded volume that can be related to pore-size distribution. The method assumes an ideal model in which the specimen pore structure is represented by a labyrinth of interconnected cylindrical pores of sequentially diminishing size. Most natural materials diverge from this. Completely closed pores will not be assessed, and pores accessed through a narrow neck will have their size underestimated. However, as a method of quality-control testing or to compare materials produced by different processes, this is a valuable technique. It cannot be used for materials with which mercury forms an amalgam.

The visibility of porosity can be enhanced by filling pores with a colored mount or metal. For instance, if a plasma-sprayed alumina coating is deposited onto a metal substrate, then copper can be electroplated into the pores, which will make them extremely easy to identify in polished cross sections. This minimizes the effects of grain pullout on the accuracy of porosity measurements.

0 0

Post a comment