Notice: Wiley Online Library will be unavailable on Saturday 27th February from 09:00-14:00 GMT / 04:00-09:00 EST / 17:00-22:00 SGT for essential maintenance. Apologies for the inconvenience.
The microanalysis of interfaces in fibre-reinforced composite materials is dependent on the successful preparation of specimens suitable for transmission electron microscope (TEM) inspection. Ideal samples should possess large amounts of structurally intact and uniform thin area in the fibre/matrix interface regions of the samples. Because fibre/matrix interfaces in this class of materials are often designed to fail under mechanical stress, conventionally prepared samples are prone to interfacial failure and differential thinning, both of which preclude detailed TEM microanalysis. These effects were seen in a conventionally dimpled and ion-beam-thinned specimen prepared from a continuous fibre reinforced ceramic composite composed of CaWO4-coated Nextel 610TM fibres in an alumina matrix. The dimpled specimen showed large amounts of interfacial failure, with only thick regions of the specimen left intact. To overcome these limitations, a focused ion beam (FIB) technique was applied to this same material. The superiority of the FIB-produced sample is evident in both the morphology and scanning transmission electron microscopy analyses of the sample.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
In the design of continuous fibre-reinforced ceramic composites (CFCCs), it is widely recognized that the fibre/matrix interface plays a critical role in the mechanical performance of the materials system. In this vein, the placement of an interphase material between the fibre and matrix phases has received considerable attention (Davis et al., 1993; Morgan & Marshall, 1993; Faber, 1997; Marshall et al., 1997). The purpose of the interphase coating is to enhance toughness of the composite by influencing crack trajectories at the mechanically critical fibre/coating interface.
This interest in coated-fibre composites has been accompanied by the need for accurate microanalysis of these CFCC systems. Because both morphological and chemical incompatibilities will be most pronounced at the interfaces between the dissimilar fibre and coating materials, particular attention must be paid to this region (Cooper & Hall, 1993; Davis et al., 1993; Morgan & Marshall, 1993; Marshall et al., 1997; Reig et al., 1997a, b). Consequently, the microanalysis has been driven below the micrometre scale, into the length scale of nanometres. Thus, the examination of the fibre/coating interfaces in CFCCs falls squarely into the realm of transmission electron microscope (TEM) and scanning transmission electron microscopy (STEM) analyses.
The challenges involved in producing samples suitable for TEM analysis are not trivial, however. This is especially true of CFCC systems. Because the constituent phases in these materials systems tend to have very different hardnesses, conventional methods of sample preparation involving mechanical grinding and polishing typically produce large amounts of differential thinning. The problem of differential thinning is also present in ion milling procedures typically used in conjunction with mechanical grinding and polishing (Howitt, 1984; Lee et al., 1987; Dawson et al., 1990; Cinibulk et al., 1996).
Differential thinning aside, the very mechanical design of CFCC systems poses serious challenges to TEM sample preparation as well. Because the fibre/coating interfaces in these systems are typically engineered to be structurally weak, conventional sample preparation involving mechanical thinning techniques can cause widespread structural failure. Although observation of such debonded fibre/coating interfaces in the TEM is a good indication of crack deflection necessary for proper composite behaviour, the debonding obviously precludes meaningful microanalysis of the interfacial region in its as-processed state.
Of the mechanical routes of TEM sample preparation of CFCC specimens, the most successful by far has been the ‘wedge polishing’ technique (Bradley et al., 1987; Klepeis et al., 1987; Dawson et al., 1990; Berger & Bunsell, 1993; Cheng, 1996; Cinibulk et al., 1996). This technique, especially as applied by Cinibulk et al. (1996) produces relatively large amounts of thin area and good sample yields as well. However, this technique is better suited to the preparation of samples with random arrangements of fibres rather than directionally orientated fibres found in most composite samples. That is, the ‘wedge polishing’ technique is better suited to sample preparation of coated fibres that are free of a binding matrix than to preparation of TEM samples from a full-scale composite body.
The issues involved in FIB sample preparation of the composite systems under study here, however, are different from those of microelectronic materials. Microelectronic FIB sample preparation places premium importance on the site specificity of the thin area produced. The thin area in these samples is typically limited to a single device of interest, however, so the amount of thin area produced is not an overriding concern. In comparison, sample preparation of composite materials places the opposite priority on the criteria of site-specificity and extent of thin area. The specific location of the thin area is not a concern in fibre-reinforced composite systems where fibre/matrix interfaces are ubiquitous. The major criterion for successful composite sample preparation is the production of a large amount of uniform and structurally intact thin area across fibre/matrix interface regions. As noted above, mutually satisfying these criteria is not trivial in composite systems where the interface regions are often: (1) composed of materials with very different sputtering rates; and (2) designed to mechanically fail.
In this study, FIB is used to prepare TEM samples from a full-scale composite composed of Nextel 610TM fibres with a CaWO4 coating in an alumina matrix. This composite system represents a ‘weakly bonded interface’ approach to composite design, where the fibre/coating interfaces are intended to mechanically fail under stress. Such a system is very challenging for TEM sample preparation, and provides an illuminating view of the feasibility of FIB sample preparation for mechanically delicate samples.
To illustrate the challenges involved in TEM sample preparation of these composite materials, as well as the advantages of FIB techniques, a comparison between mechanically dimpled and FIB-prepared TEM specimens will be made. Finally, a brief demonstration of the suitability of FIB-prepared specimens for TEM microanalysis will be shown as well.
2. Materials and methods
2.1. Conventional sample preparation
The unidirectional composite sample was obtained from McDermott Technologies (Lynchburg, VA), and specifics regarding the composite fabrication and processing are given elsewhere (Goettler et al., 1997a, b). The as-received composite was a bar measuring approximately 100 mm × 20 mm × 2 mm, with the long axis corresponding to the fibre direction. Because of the rather thin nature of this composite (2 mm), making mechanically robust assemblages with the standard TEM sample diameter of 3 mm was necessary. The procedure used to do this is described below and illustrated in Fig. 1.
Strips approximately 15 mm long were cut from the composite along the fibre axis using a low speed diamond saw. Three strips were then ground down in width and thickness to fit tightly within a 3 mm diameter copper tube. An epoxy (Epo-Tek® 353ND; Epoxy Technology, Billerica, MA) was then placed in the tube and air bubbles removed from the tube by a slight vacuum. The epoxy was then cured in an evacuated oven. This effectively produced a 3 mm diameter cylinder of orientated (fibre direction coincident with cylinder axis) composite with mechanical robustness suitable for subsequent TEM sample preparation.
The copper tube was then sectioned normal to its axis with a slow speed diamond saw, producing slices approximately 1.2 mm thick. The resulting discs were then ground to approximately 80 μm thickness using SiC papers, with care taken to ensure the ground faces were parallel. Both sides of a single composite disc were then polished to a 0.1 μm finish using diamond lapping films. This disc was then dimpled to approximately 10 μm thickness using water-based diamond suspensions. Final thinning to electron transparency was performed using an Ar+ ion beam operated at 3 kV and 15° incidence to the sample surface. To minimize ion bombardment damage, the specimen was kept on a liquid nitrogen cooled stage during the ion milling.
2.2. FIB sample preparation
2.2.1. FIB instrumentation
FIB instrumentation is a hybrid of SEM, ion milling and computer aided design (CAD) systems. The FIB is conceptually very similar to SEM, with the gallium ion beam used instead of an electron beam. The beam is focused and rastered across the sample to achieve material removal through sputtering. This process also creates a secondary electron signal from the sample, which is collected by a detector to produce an image of the probed area. Thus, as the ion beam is milling the sample, an electron image is formed simultaneously. Site-selective milling is accomplished through a CAD-like computer interface control of the beam rastering. By drawing boxes over those areas of the sample that are to be milled, the user selectively thins different areas of the specimen. During the thinning process, the electron image of the sample area being milled can be monitored to assess the progress. It should be noted that, although this simultaneous link between milling and monitoring caused by the gallium beam is highly useful, the corollary is that imaging of the sample only occurs at the expense of some milling of the sample. Thus, examination of sample regions beyond the actual areas being milled (i.e. viewing the sample at large) should be kept to a minimum in order to limit extraneous beam damage.
2.2.2. FIB procedure
FIB sample preparation followed the route illustrated in Fig. 2. First, a 80 μm-thick disc from the same epoxy-impregnated copper tube used in the conventionally prepared samples was obtained. The disc was then thinned to ~30 μm thickness with SiC grinding papers, followed by polishing with 0.1 μm diamond lapping films.
The leading edge of both the epoxy resin and the copper tube was then removed from the disc with a scalpel to facilitate FIB processing, as shown in Fig. 2. The modified discs were thinned using a Hitachi FB-2000 A FIB instrument, with the gallium ion beam accelerated to 30 kV for the sample preparation.
In order to produce TEM samples cross-sectional with respect to the fibre axis of the composite, the general technique illustrated in Fig. 2 was employed. As shown, the gallium ion beam was directed normal to the fibre axis. This produced the electron-transparent area in an edge-on fashion. For TEM viewing, the sample was then reoriented such that the TEM electron beam was normal to the thin section, producing a cross-sectional view of the fibres.
A unique feature of the Hitachi FB-2000 A FIB instrument used here is the dual FIB- and TEM-compatible sample holder that was used. This specimen holder allows the sample rotation between FIB and TEM viewing shown in Fig. 2 to be performed without sample removal. This not only facilitates checking of the sample thickness, but also helps maintain accurate positioning of the sample between the two instruments.
FIB milling of the sample proceeded until the thin window was electron-transparent, approximately 100 nm thick. The area of the thin window typically measured ~10 μm × 15 μm. A SEM image of a completed FIB sample is shown in Fig. 3.
3.1. General morphology
To compare the dimpled vs. FIB-prepared specimens, the samples were viewed in a Hitachi HF-2000 TEM with a cold field emission gun and 200 kV acceleration. The overall amount of thin area, thickness of the thin area, intactness of the fibre/coating interfaces and general structural integrity were examined when determining the sample quality. A representative series of TEM bright-field images of the dimpled sample is shown in Fig. 4. In comparison, a TEM bright-field image from the FIB-prepared sample is shown in Fig. 5.
3.2. Analytical suitability
The use of TEM analysis on CFCC systems is motivated by the chemical microanalysis capabilities of both TEM and STEM instruments. Microanalysis in S/TEM instruments permits chemical information about the fibre/coating interfaces in these materials to be gathered with nanometre-scale spatial resolution. This level of spatial resolution can be vital to the accurate assessment of chemical and morphological phenomena at the fibre/coating interfaces.
To determine the suitability of the dimpled and FIB-processed samples to this kind of microanalysis, the STEM capabilities of the Hitachi HF-2000 were used. Annular dark-field (ADF) images of the samples were obtained using a Gatan 679 ADF detector and energy-dispersive spectroscopy (EDS) line profiles across fibre/coating interfaces were performed with the aid of an Oxford PentafetTM (Oxford Instruments, Eynsham, U.K.) spectrometer. The STEM capabilities of this microscope were accessed through ESVisionTM software from EmiSpec, Inc (Tempe, AZ) which controlled the ADF imaging, probe positioning and recording of the EDS signal.
A qualitative interpretation of the EDS linescan data was performed by plotting the ratios of integrated intensities of characteristic EDS peaks in the spectra. By plotting ratios of Ca–K : Al–K and W–L : Al–K integrated EDS peak intensities, the chemical abruptness of the fibre/coating interfaces was revealed. A plot of these integrated intensity ratios and corresponding ADF images indicating the location of the linescans across the sample are shown in Figs 6 and 7 for the dimpled and FIB-processed TEM samples, respectively.
Because the ultimate goal of TEM examination of CFCC materials is to yield information about various chemical and morphological aspects of the fibre/coating interfaces, ideal TEM samples of CFCC materials should possess three qualities:
1 Mechanically intact interfaces
2 Large amounts of thin area to either side of the interfaces
3 Uniform thin area with no interfacial grooving.
These three factors ultimately govern the usefulness of analytical data that will be gathered via the TEM. Evaluation of the two sample preparation techniques follows along these lines.
4.1. Dimpled sample
4.1.1. General morphology
Representative micrographs of the dimpled sample are shown in Fig. 4. These figures depict the problems associated with conventional mechanical sample preparation of CFCC materials. In Fig. 4(a), the extreme thickness of the prepared area is apparent, as well as the problems associated with matrix porosity. Figure 4(b) illustrates the problem of differential thinning at the critical fibre/coating interface region. Although the coating is thin, the fibre possesses very little thin area, and the surrounding matrix suffers from extensive areas that have ‘fallen out’. Conversely, Fig. 4(c) shows a region with better fibre thinning, but the surrounding coating and matrix regions have been left thick.
Also noticeable in Fig. 4 is ubiquitous debonding at the fibre/coating interfaces. Although this large amount of interfacial debonding has encouraging implications for composite behaviour, it also precludes detailed analysis of the fibre/coating interface. The dimpled samples fail to meet the three critical criteria outlined above on all three counts, making them unsuitable for detailed microanalysis.
4.1.2. STEM microanalysis
Because the interfaces were only left intact in thick areas, the STEM analysis was compromised. The thickness of the sample caused problems for both the imaging and analytical capabilities of the STEM. The electrons diffracted by the sample form the ADF image. With thick samples, multiple scattering of the diffracted electrons results in ADF images that have poor contrast and spatial resolution, as seen in Fig. 6.
The sample thickness also negates the atomic number contrast that is available in ADF imaging. The CaWO4 coating has a higher average atomic number than the Al2O3 fibre, and should ideally appear brighter in the ADF image. The observed contrast in Fig. 6 is the opposite of this, however, because extreme thickness of the sample precludes Z-contrast.
The Ca–K : Al–K and W–L : Al–K integrated intensity ratios shown in Fig. 6 are representative of the intact interfaces throughout the sample. The profiles do not show a clean transition from the coating to fibre phases. Moreover, the intensity ratios do not reach an apparent steady value on either side of the interface, which could be misinterpreted as diffusion between the phases. This is largely due to sample thickness effects.
The results of the STEM analysis are not initially promising and point to sample limitations. The prevalence of interfacial debonding seen in the TEM images seems to indicate that the coating is performing as intended, with ubiquitous failure at the fibre/coating interface. However, the lack of thin area in the sample makes meaningful STEM imaging and EDS analysis difficult at best.
4.2. FIB-produced sample
4.2.1. General morphology
Some of the advantages of sample preparation via FIB are immediately apparent from the SEM image of the FIB-prepared sample shown in Fig. 3. The large amount of thin area produced by the FIB is easily seen. The figure shows an area approximately 40 μm × 15 μm that has been thinned to electron transparency. This large amount of thin area is also localized in one area, allowing TEM analysis to be performed at several locations of the fibre/coating interface. The TEM bright-field image of the thin section in Fig. 5 confirms that the entire FIB window is electron-transparent. Moreover, the thin area generated is much thinner than in the dimpled sample, evidenced by the appearance of grain structure in the fibre, coating and matrix phases of the sample.
The SEM image also indicates that the fibre/coating and coating/matrix interfaces are intact, supported by the TEM image shown in Fig. 5. This is the result of several factors that distinguish FIB from other methods of sample preparation. First, FIB does not use abrasive methods to remove material. Therefore, much less mechanical stress is imposed on the sample than in conventional grinding, dimpling and polishing. These methods of sample preparation not only require the use of abrasive media for grinding and polishing, but also the associated use of pressure to achieve material removal. The obvious tendency for a properly designed CFCC material under these conditions is to promote failure at the fibre interfaces, as seen in the case of the dimpled and ion-milled samples. The gallium ion beam used in FIB does not use rough abrasives or applied pressure to achieve material removal. In comparison to the mechanical sample preparation required before conventional ion beam thinning becomes effective, FIB is much more mechanically benign.
The FIB sample also does not create thin area throughout the entire sample. Rather, the thin area is concentrated in the milled region only. Because the majority of the bulk sample is left untouched by the ion beam, FIB samples are very structurally robust. The unthinned regions provide mechanical support, alleviating stresses that might cause interfacial failure in the thin foil section. Also, these unthinned areas are relatively thick and therefore less likely to suffer from failure induced by sample preparation. The unthinned sections help to keep the interfaces intact by not introducing matrix cracks into the thin region.
The TEM image in Fig. 5 also indicates that the thin area produced by FIB is remarkably uniform in thickness over wide areas, with no evidence of interfacial grooving. Because the gallium ion beam impinges on the sample at a low incidence angle (almost tangential to the foil surface), there is little preferential milling of the interfaces. Typically, the angle of incidence of the gallium beam is a few degrees in TEM sample preparation. This low-angle beam orientation controls the milling process so that material removal is rate limited by the material that sputters most slowly. In effect, the slower milling phases shield the faster ones, encouraging the creation of flat and uniform surfaces.
Based on initial inspection of the overall morphology of this sample, it appears that the criteria for high quality TEM samples (widespread and uniform thin area across structurally intact interfaces) have been met by FIB-processing methods.
4.2.3. STEM microanalysis
Following the procedure used for the dimpled specimen, ADF images and EDS line profiles across fibre/coating interfaces were performed on the FIB-processed sample. Figure 7 shows representative results from this analysis. The thin nature of the FIB-created sample is evidenced by the appearance of grain structure in the ADF image. This is in stark contrast to the ADF image obtained from the dimpled specimen, and demonstrates the quality of FIB-processed sample.
The EDS microanalysis further supports the superiority of the FIB processing. The transition from CaWO4 to Al2O3 shown by the EDS ratios in Fig. 7 occurs over a width of less than 15 nm. Moreover, the Ca–K : Al–K and W–L : Al–K ratios reach steady values on both sides of the interface, indicating that the electron beam is probing homophase regions to either side of the interface. This is due to the uniform thin area present in the interfacial region. The superiority of FIB-prepared over the dimpled specimen is evident in both the cleanliness and narrow width of the transition.
Modern STEM instrumentation can easily produce probe diameters < 5 nm. For non-model systems this ultimately limits the resolution of microanalysis to the quality of the specimens used. The results shown here indicate that in the case of the Nextel 610TM/CaWO4 CFCC, the FIB specimen is suitable for more detailed microanalysis, whereas the dimpled specimen does not warrant such further investigation.
4.3. FIB-induced artefacts
4.3.1. Ion bombardment damage
Despite the fact that the FIB produces large amounts of uniform thin area with no interfacial grooving and structurally intact interfaces, the use of the 30 kV gallium ion beam does introduce artefacts of its own (Ishitani & Yaguchi, 1996; Walker & Broom, 1997).
The gallium ion beam used for milling is quite aggressive in its attack on the sample. This is necessary for removing large amounts of material in a reasonable amount of time. The effect of ion bombardment damage is most pronounced at the front edge of the thin section. As this region experiences the most bombardment and the highest angle of ion beam incidence (close to 90°), the first few nanometres of the leading edge of the sample are typically highly damaged. To minimize the intrusion of this damaged edge on the sample area of interest, a protective layer (typically tungsten or SiO2) is often deposited on the leading edge of the sample before milling begins.
The ionic bombardment that produces sputtering also leaves a thin layer of damaged material on the sample surfaces milled by FIB (Ishitani & Yaguchi, 1996; Walker & Broom, 1997; Tanaka et al., 1998). This limits the minimum sample thickness that can be produced while leaving some undamaged material for observation. This damage layer also contains some gallium implantation, which may obscure chemical analysis. In extreme cases, this damage layer can be seen in the TEM as black spots, indicative of gallium precipitation (Tanaka et al., 1998).
To assess the severity of FIB-induced damage on the samples here, Monte Carlo Stopping and Range of Ions in Matter (SRIM) routines written by Ziegler (1998) were used. Materials parameters including the average atomic numbers, densities and molecular weights of both alumina and CaWO4 were used as input parameters. To closely model actual FIB operation, the simulated gallium ion beam was accelerated to 30 kV and directed at a 1° angle of incidence to the sample surface. The results of the simulations were fully developed after 1500 ion trajectories for both materials.
The results of the SRIM simulations are shown in Fig. 8. The gallium ion trajectories are shown in black. The grey regions show the trajectories of displaced atoms in the sample itself, known as the ‘cascade region’. As seen, the cascade region extends beyond the range of gallium implantation. The simulations in Fig. 8 show that the damage layer extends further into CaWO4 than into alumina, and the overwhelming majority of the damage is limited to a depth of ~20 nm. In practice, the damage layer thickness may be less than this value, since well-characterized binding energies of the alumina and CaWO4 phases were not available for use in the SRIM simulations. Thus, for a TEM sample 100 nm thick, at least 60 nm (100 nm, less 20 nm per FIB-processed side) of material remains undamaged by FIB processing.
The extent of gallium implantation in the sample is much less severe, covering a maximum depth of approximately 10 nm on each face of the sample. Again, this maximum depth occurs in the CaWO4 phase. The presence of this gallium can obscure elemental analysis by EDS and electron energy-loss spectroscopy (EELS). The constituent elements present in the Nextel 610TM/CaWO4 composite do not overlap with gallium in either the EDS or EELS spectra, however, so this effect is negligible.
Given the quality of the samples generated by FIB as opposed to conventional sample preparation techniques, the advantages to FIB use easily outweigh potential downsides. The simulated damage layer extends a maximum of ~40 nm (20 nm per side) into the thickness of the sample used here. Assuming a total sample thickness of 100 nm, this still leaves an abundant amount of pristine material to probe. The simulated gallium implantation extends a maximum of ~30 nm into the thickness of the sample, again leaving plenty of pristine material to probe. Finally, the presence of gallium in the sample should not complicate the chemical analysis because characteristic gallium peaks in EDS and EELS do not overlap with the peaks of any of the elements intrinsic to the composite material (aluminium, calcium, tungsten and oxygen).
Because of the discrepancy between the thickness of the milled and unmilled portions in a FIB sample, specimen-induced fluorescence can significantly influence the EDS signal obtained (Longo et al., 1998; Saito et al., 1998). The fluorescence effect is most pronounced in FIB specimens where the thinned window has very small dimensions of length and width as compared with the thickness of the unmilled bulk slab of material.
Saito et al. (1998) have demonstrated that the specimen-induced fluorescence is caused by forward scattered electrons that impinge on the unmilled areas of the sample surrounding the thin window region. This specimen-induced fluorescence is manifested as extremely high EDS collection dead times and a strong presence of matrix character in the EDS spectra. Neither of these effects was seen in the FIB specimen produced here.
These artefacts were circumvented in the sample here because of the extremely large amount of thin area produced, even by FIB standards. As seen in Fig. 5, the thin area extends far in both the length and width directions, measuring at least 25 μm × 10 μm. As the EDS analysis was limited to the leading 2 μm of the thin area, the effects of forward scattering electrons striking the unmilled bulk of the material slab were avoided.
More importantly, the quality of the samples produced by FIB allows for microanalysis by EELS. EELS does not suffer from effects of either forward-scattered or secondary electron-induced fluorescence like EDS (Colliex, 1985; Williams, 1987; Williams & Carter, 1996), making artefacts caused by the FIB specimen geometry a moot point. The use of EELS analysis on the FIB-produced sample shown here will be the topic of a future publication.
The results shown here demonstrate that FIB-processing is superior to mechanical dimpling for producing TEM samples of CFCC material systems that are suitable for detailed microanalysis, which has implications for the utility of FIB sample preparation of mechanically delicate materials in general. Both morphological and chemical S/TEM characterization of the critical fibre/coating interfaces in coated-fibre CFCCs is dependent on obtaining samples with large amounts of uniform thin area that retain structurally intact interfaces. In stark contrast to mechanical dimpling routines, FIB processing is extremely effective in meeting these criteria. The ‘suitability to task’ of FIB-produced samples in achieving highly spatially resolved morphological and chemical analyses has been demonstrated. Even in light of possible sample preparation artefacts, the use of FIB over sample preparation routines mechanical in nature is compelling.
The authors would like to thank Richard W. Goettler at McDermott Technologies, Inc. for supplying the composite material used in this study. Thanks are also due to Sankar Sambasivan of ACTG/Northwestern University for his contributions to this research.