Abstract

The Focused Ion Beam (FIB) tool has been successfully used as both a means to prepare site-specific TEM foils for subsequent analysis by TEM, as well as a stand-alone instrument for micromachining of materials. TEM foil preparation with FIB technique has drastically changed traditional TEM specimen preparation because it allows site-specific foil preparation. FIB consists of cutting electron transparent foils through Ga-ion milling on a bulk sample. Optical microscopy together with a micromanipulator is used to transfer the foil from the specimen to a TEM grid coated with a holey carbon support film. No further carbon coating is required. This novel technology offers significant benefits in precision and speed of preparing site-specific TEM foils from inclusions in minerals, grain boundaries, microfossils, thin films on substrates and various coatings. FIB cut TEM foils can be investigated with modern TEM techniques such as various diffraction techniques, analytical electron microscopy (AEM) including line scan and elemental mapping, electron energy-loss spectroscopy (EELS) and high-resolution electron microscopy (HREM). With the FIB instrument (FEI FIB200) operated since fall 2002 at GFZ Potsdam we have prepared and investigated several hundreds of high quality TEM foils from silicates, carbonates, metals alloys, ceramic materials and diamond.

Introduction

In the past twenty years, several technical improvements in transmission electron microscopy (TEM) and analytical electron microscopy (AEM) have pushed the capability of investigating solid-state materials to an atomic scale. Nevertheless, since the invention of TEM by E. Ruska (1934) in the thirties of the last century until in the early nineties it mostly remained a dream for an electron microscopist to investigate precisely that part of a specimen he is actually interested. All specimen preparation techniques applied so far - such as argon ion milling, electrochemical polishing and different replica techniques - resulted in electron transparent foils, which were not site-specific and displayed varying thickness. However, in the late eighties and the early nineties a novel technique - focused ion beam (FIB)- permitting site-specific specimen preparation of TEM foils was reported (Kirk et al., 1989; Young et al., 1990; Basile et al., 1992; Overwijk et al., 1993). Focused gallium ion beam devices, originally developed at the University of Chicago and at the Oregon Graduate Institute in the mid-1970's and operated in the semiconductor industry, were used to cut thin foils from precisely the location of interest in the sample. The authors also published a procedure to lift the foil out of the excavation site of the specimen and place it onto a TEM copper grid, the so-called lift-out technique, which later was improved by Giannuzzi et al. (1997). In the following years FIB technique was preferentially applied to TEM foil preparation in the semiconductor industry and gradually more and more in materials science.

Recently, first FIB applications in geosciences were reported (Kuhlman et al., 2000; Wirth, 2000, 2001, 2002, 2003; Dobrzhinetskaya & Green, 2001; Dobrzhinetskaya et al., 2001; Heaney et al., 2001; Dobrzhinetskaya et al., 2002; Lee et al., 2003; Dobrzhinetskaya et al., 2003; Graham et al., 2004). The tremendous benefit of site-specific TEM foil preparation for TEM investigations in geosciences is illustrated by a few examples. Submicron sized mineral inclusions or melt inclusions in minerals (e.g. inclusions in diamond) give us an insight into the state of the Earth's mantle or the lower crust during the formation of the host mineral. TEM investigation of foils containing such inclusions reveals information about the chemical composition, the structure and the paragenesis of the inclusions. The major problem with conventional TEM specimen preparation is the difference in hardness of multiphase materials, which results in preferential thinning of the weaker phase. Similar problems arise with TEM foil preparation of interfaces by conventional argon ion milling. The interface region deviates in chemical composition and bonding from the bulk structure of the two adjacent grains, and therefore it will be preferentially thinned. This problem is overcome by FIB specimen preparation (Heinemann et al., 2003; Seydoux-Guillaume et al., 2003). TEM foil preparation of inclusions in meteorites and that of dust particles is extremely challenging, however, it is tremendously facilitated using FIB (Vicenzi & Heaney, 1999; Lee et al., 2003; Graham et al., 2004).

Starting in fall 2002 the first FIB device (FEI, FIB200) used in geosciences was operated at the GeoForschungsZentrum (GFZ) Potsdam. The present paper reports the unique capabilities of FIB based on our own experience with different materials, especially silicates. We have cut site-specific TEM foils to solve special problems in geosciences such as identifying submicron sized mineral and melt inclusions in host minerals, structure of interfaces, characterization of very small samples from high pressure experiments (microstructure, chemical composition of involved phases), microfossils (relicts of carbon from cell walls) and special problems in applied mineralogy (coatings or thin films on substrates; micromachining of diamond). A brief introduction into the technical details of the methods is followed by selected examples of successful application of FIB in combination with TEM investigations. The advantages and disadvantages of the method are discussed in the final section of the paper. It is the aim of the present paper to show the new and extraordinary possibilities offered by FIB in combination with TEM to the geosciences community.

Description of the method

Electron transparent foils suitable for TEM investigations such as conventional bright- and dark-field imaging, electron diffraction, electron energy-loss spectroscopy (EELS), high-resolution electron microscopy (HREM) and analytical electron microscopy (AEM) can be successfully produced using FIB technique. Suitable starting materials for FIB specimen preparation are uncovered thin sections of rocks or minerals, small polished pieces of rock, powder material embedded in epoxy resin, metals, alloys or ceramic devices. The shape of the specimen is not crucial, and even from spheres TEM foils can be milled. Semiconductors or insulators need a thin coating with conducting material (e.g. carbon, gold) prior to FIB milling. A FIB prepared TEM-ready foil usually is typically 10–20 μm wide, 5–15 μm high and about 100–200 nm thick. The final thickness can be less than 50 nm. It depends on the material properties and last but not least on the operator's skills. TEM ready foils can be prepared in a fully automated process, which takes about 4 hours for a TEM-ready foil with the dimensions 15 × 10 × 0.200 μm.

Sputtering of the target material with Ga-ions usually accelerated to 30 kV is the basic physical process behind the FIB technique. The principles of FIB milling and detailed descriptions are given elsewhere (Overwijk et al., 1993; Heaney et al., 2001; Roberts et al., 2001; Orloff et al., 2002; Lee et al., 2003). The ion-sample interactions between the Ga-ions and the target are discussed in Prenitzer et al. (2003).

The present paper focuses on application of FIB in geosciences and starts with a brief summary of the major FIB milling steps. During the fully automated preparation sequence beam shift and sample drift would pose problems, especially during the final milling steps when the foil thickness is only a few hundred nanometres. Therefore, the system must compensate for beam shift and/or sample drift during milling using reference markers on the sample (e.g. crosses), which are cut into the surface near the area of interest (Fig. 1a). Next and obligatory step is the deposition of a thin (1-2 μm) protection layer of platinum onto the area of interest. Platinum deposition occurs by decomposing an organic Pt-gas in the Ga-ion beam, while the beam is scanned over the preselected pattern. The platinum layer protects the foil from abrasion by the Ga-ion beam, in particular during the late stage of the preparation sequence when the foil thickness is less than 500 nm. At the beginning of the milling process two larger trenches are milled one in front and another one behind the foil with a high ion current (2700 pA) (Fig. 1b, c). Beam size and ion current are defined by the diameter of the selected aperture. The dimensions of the milling pattern (trench size and depth) depend on the selected foil dimensions, and are optimized by the computer software in order to minimize the total milling time. The actual progress of the sputtering process is monitored acquiring secondary electron (SE) images of the sample. The interaction of Ga-ions with atoms and electrons of the target material creates secondary electrons that are used for imaging. Milling is continued with reduced ion beam size (reduced beam current, 350 pA) until the foil thickness is approximately 500 nm. At that stage the specimen is tilted 45° with respect to the ion beam and the foil can be cut free at its base and at both sides, leaving only a narrow (ca. 1 μm) strip of material to fix the foil in its position (Fig. 1d). Subsequently, the specimen is tilted back to its original position and fine polishing with the ion beam (70 pA) nearly parallel to the foil occurs until the preselected final thickness is reached. Without precaution, the final cross-section of the foil would be wedge shaped because of the Gaussian profile of the ion beam. However, the beam profile is compensated by tilting the specimen during the fine milling procedure at about 1 degree (negative, positive) with respect to the incident beam. Finally, only the stabilization strips on both sides need to be cut and the foil is ready for lift-out (Fig. 2). Figure 2 shows that the Pt-strip is nearly removed in the centre of the foil. This is the result of slight bending of the foil towards the beam during final polishing. In some cases, bending is so strong that the beam will cut through the foil. If bending of the foil is noticed during the milling process the foil can be cut free at one side, thus allowing the foil to relax and milling will be continued. The specimen is removed from the FIB device, and under an optical microscope, the foil is lifted-out of the excavation site by means of a sharp tip of a glass fibre attached to a manipulator. Only electrostatic forces keep the foil at the tip of the glass fibre. Finally, the foil is placed onto the membrane of a TEM copper grid. It is necessary to press the foil with the glass fibre softly onto the carbon film to make sure that the adhesive forces keep it in its position, and it will not fall off during handling or inserting the specimen into the TEM. The adhesion of the foil to the membrane is strong enough to allow shipping it in an appropriate container. In our TEM applications, we prefer holey carbon membranes on a 3 mm copper TEM grid to fix the foil to the grid. Sometimes carbon is an interesting element in the sample, and only a perforated carbon film allows interpretable measurement of carbon in the sample by EELS and/or EDX analysis.

Presently, the milling software offered by the manufacturer is optimized for silicon material only. However, from our experience with FIB that program can be applied with very good results to most silicates, many ceramics and even metals and alloys. For most materials, the depth of the foil is always somewhat larger than the preselected depth using the silicon program. However, it is not applicable for diamond. Because of the strong binding of the carbon atoms in diamond, the sputtering rate of carbon is significantly reduced. That disadvantage is compensated using an additional device, which is called selective carbon mill (SCM ® FEI Company). Heating Mg- hydrate sulphate in a crucible releases water vapour. With the SCM device active a needle is inserted into the specimen chamber with its tip close to the selected milling area. Water vapour is released into the chamber, decomposed by the ion beam, and finally carbon is oxidized. SCM enhances milling of diamond significantly (factor 20).

Application of FIB and results

Mineral and melt inclusions

Inclusions in diamond

TEM foils from diamonds are prepared with FIB from single crystals, microdiamond inclusions hosted e.g. in garnet, which can be observed in thin sections without cover glass, and from fragments of crushed diamond. Figure 3a shows a diamond fragment from a stone collected from the kimberlite pipe Molodost (Yakutia). The fragment contains several small inclusions, and one of them is displayed in Fig. 3b. A TEM foil was cut with FIB from the lower part of the inclusion to get a complete cross section of the inclusion together with the host diamond (Fig. 3c). The investigation of the TEM foil (Fig. 3d) identified the inclusion by diffraction analysis and chemical composition as an omphacite single crystal, which is slightly corroded by a small amount of melt.

Submicrometer sized inclusions in diamond from the Diavik mine (Canada, NWT) have been investigated by TEM, analytical electron microscopy (AEM) and electron energy-loss spectroscopy (EELS). TEM foils were cut with FIB from polished crystals using the selective carbon mill (SMC) device. The small electron absorption of diamond allows the use of rather thick TEM foils (300 nm). Larger foil thickness has the advantage that some small inclusions might still be closed when studied by TEM. Such inclusions typically contain solid inclusions, gas and /or fluids. Diffraction analysis (selected area diffraction, convergent beam electron diffraction CBED) of individual inclusions reveals structural information of single phases. The chemical composition is determined by energy dispersive X-ray analysis (EDX) or by EELS. Figure 4 shows an example of a particular inclusion that contains two solid phases (phlogopite and carbonate) plus fluid or gas component. Diffraction patterns of the carbonate phase prove its aragonite structure. The major benefit with FIB-prepared TEM foils studying inclusions in diamond is the unique possibility to investigate the mineral assemblages as nature has created them, in contrast to the usual method of burning or crushing diamonds and investigating the residuum.

Melt inclusions in a host mineral

Optical microscopy of orthopyroxene (opx) mega crystals reveals several small inclusions (10-20 μm in diameter), which are always associated with healed cracks. Origin of that sample is the kimberlite pipe Slyudyanka, Yakutia. Opx mega crystals from that location have formed under conditions of the graphite-pyrope-facies. Microprobe analyses of these inclusions cannot be interpreted. FIB prepared TEM foils of these inclusions immediately show why microprobe data cannot be interpreted. Already during FIB milling, the acquired SE-images of the inclusion cross-section exhibit patchy or flame-like bright and dark contrast (Fig. 5), thus suggesting the presence of at least two different phases. It should be mentioned that orientation contrast can also produce changes in brightness in the SE-image. AEM measurements of the phase with dark contrast in the TEM bright field image prove that it is a Fe-Ni-Cu-sulphide phase. The presence of copper in sulphide needs to be documented by EEL spectra exhibiting the Cu-L3,2 edge. The copper Kα line in the EDX spectrum, which is nearly always visible in the EDX spectrum, might be due to copper in the TEM grid (secondary fluorescence) and cannot be attributed to a particular phase. The X-Ray elemental map in Fig. 6 illustrates that the dark phase in the STEM image is mainly composed of Fe, Ni and S, whereas the bright contrasted areas are composed of Mg, Si and lower concentrations of Fe. A line scan across a section of the bright area using Mg-Kα, Si-Kα and Fe-Kα intensities shows rather homogeneous distribution of these elements (Fig. 7). Chemical analysis together with lattice spacing derived from diffraction patterns and high-resolution images (not shown here) identified antigorite to be the bright phase. The chemical composition of antigorite is slightly variable thus indicating the presence of relicts of the original melt between individual antigorite grains (MgO: 39.9 - 40.6 wt.%; Al2O3: 1.4 - 2.2 wt.%; SiO2: 42.3 - 43.3 wt.%; FeO: 13.1 - 13.2 wt.%; NiO: 0.5 - 0.8 wt.%). The results of TEM investigation of the inclusions in opx are interpreted as mingling of sulphide melt with silicate melt prior to its enclosure in opx. During uplift and cooling antigorite has crystallized from silicate melt. It is assumed water was added and ion exchange occurred along cracks in opx, which are now healed.

Figures 6 and 7 demonstrate another great benefit of FIB prepared TEM foils. In contrast to conventional wedge shaped argon ion milled samples the uniform thickness of the TEM foils allows an unambiguous interpretation of elemental maps and line scans. Varying X-ray intensities according to varying brightness of the different colours in the map or in the line scan are only related to concentration variations of the elements displayed and not to thickness variations in the foil.

The investigation of melt inclusions in host minerals of mantle xenoliths has great scientific potential for studying mantle processes. This has been demonstrated recently by the discovery of nanocrystalline diamond in melt inclusions in opx and cpx in mantle xenoliths in Hawaiian lava (Wirth & Rocholl, 2003).

Grain and phase boundaries

Forsterite bicrystals

A series of symmetric grain boundaries were synthesized in forsterite bicrystals by direct bonding with tilt axis ao and tilt angles in the range of 9°–21° (Heinemann et al., 2001, 2003). Two oriented and polished synthetic forsterite single crystal plates were contacted at room temperature and annealed at 400°C in vacuum for one week. High-resolution transmission electron microscope (HREM) images from the interface show that the grain boundary structure already developed below 400°C and does not change during further annealing at 1650°C. Figure 8 exhibits a Fourier filtered image of a typical 11° tilt boundary with the geometrically necessary dislocations indicated. The specimen was oriented in the FIB in such a way that the grain boundary plane is normal to the surface. A TEM foil cut normal to the surface of the bicrystal and normal to the grain boundary plane will contain the grain boundary. Grain boundary structure was derived from HREM images. However, HREM images need very thin specimen. The usual foil thickness of about 100–150 run is not suitable for HREM imaging. The foil thickness near the interface can be reduced significantly by polishing an additional narrow window (1-2 M wide and 2–3 μm deep) into that part of the foil that contains the interface plane. With a narrow Gaion beam (11 pA) foil thickness is gradually reduced in this particular area, whereas the rest of the foil remains thicker thus stabilizing the foil during lift-out and handling.

Olivine-serpentine-olivine phase boundary

Olivine grain boundaries from a sheared lherzolite from the kimberlite pipe Udachnaya, (Yakutia) display nothing peculiar along the interfaces under the optical microscope. The grains are recrystallized with an average grain size of about 5–10 μm; no subgrain boundaries are visible. The interfaces, viewed edge on, appear as thin, straight dark contrast. Several TEM foils were cut with FIB across the olivine grain boundaries from a thin section without cover glass. TEM investigation of the foils shows that the dark contrast of the grain boundary under the optical microscope actually is a complex olivine-serpentine-olivine phase boundary (Fig. 9). The TEM bright field image exhibits that one of the olivine - serpentine interfaces is porous. Olivine and serpentine are in contact only in some places. This part of the olivine-serpentine phase boundary is dominated by voids and channels, which are open for fluid transport thus enabling serpentinization of the olivine. This interpretation of the image is allowed because FIB preparation of interfaces avoids preferential thinning of the interface region. The other serpentine-olivine interface is straight without any voids. Argon ion milling of the same material would have preferentially thinned the serpentine at the interface and thus produced artefacts.

Thin silicate films (enstatite+ forsterite) on a quartz substrate

Grain boundary diffusion in enstatite-rich pyroxene was investigated in pulsed-laser deposited thin films (Dohmen et al., 2002) by the rim growth method. The starting material consisted of polycrystalline layers of isotopically doped (18O, 29Si) olivine and pyroxene on a polished quartz surface. Annealing experiments were carried out at 1000°C and 1200°C (Milke et al., 2003). The thickness of the layers and their microstructure were determined from FIB-cut TEM foils from the staring material as well as from the annealed samples. In such case the TEM-foils need to be only about 20–15 μm wide and 5 μm deep to get representative cross-sections of the layers and the substrate, thus reducing the preparation time significantly less than two hours.

Characterization of very small samples from high-pressure experiments

Chemical analysis of FIB cut foils with analytical electron microscopy (AEM)

One characteristic feature of high-pressure experiments (e.g. synthesis of high-pressure minerals, mineral reactions, phase transformations) is the extremely small experimental setup, especially with diamond anvil cells (DAC). Usually, samples and reaction products are so small that X-ray analysis and spectroscopic techniques (IR, Raman) are the only methods to investigate the bulk specimen. Chemical composition of individual phases cannot be measured by electron microprobe analysis (EMP) due to the small grain size, which is mostly much smaller than the diameter of the electron beam and the excitation volume of the electron beam interacting with the specimen. Conventional TEM specimen preparation by argon ion milling is hindered by the small size of the sample and mostly by multiphase material with different hardness of the individual phases. Here, FIB offers a unique opportunity to cut TEM foils from the areas of interest without destruction of the rest of the sample, thus permitting subsequent application of X-ray and/or spectroscopic methods. TEM foils can be cut from multianvil experiments as well as from DAC experiments. With FIB-cut TEM foils, the chemical composition of individual phases can be determined by analytical electron microscopy (AEM) and electron energy-loss spectroscopy (EELS). Structural data obtained by diffraction techniques from the same phases (selected area diffraction, convergent beam electron diffraction) complete the characterization of individual phases. However, not only chemical and structural data can be derived from FIB cut foils but also microstructural data such as grain size, dislocations and stacking faults.

IR- and Raman spectroscopy: determination of sample orientation

Detailed interpretation of IR-spectra often requires the knowledge of the orientation of the specimen during the measurement with respect to the beam. The orientation of the grain can be derived from a FIB cut TEM foil. It is easy to cut a TEM-foil from an embedded crystal parallel to the requested direction (even if the crystal is embedded in epoxy resin) and reconstruct the orientation of the foil with respect to its former position in the sample. The orientation relationship of the sample with respect to the beam (IR, Raman) can be derived from diffraction pattern or from high-resolution TEM images. This technique was successfully applied to a single crystal of super hydrous B with a grain size of about 120 micrometer, which was IR-spectroscopically investigated previously.

Microstructure

Microstructural TEM investigation of samples from phase transformation experiments is useful understanding the mechanisms of a particular phase transformation. TEM foils can be cut across the interface separating two phases and subsequently behind the transformation front in required distances. TEM foils from different locations allow the detailed investigation of the microstructural evolution of the stable phase beyond the transformation boundary. In case of coesite → quartz phase transformation, it is demonstrated that at the transformation front palisade-like quartz grains with either high dislocation density or Brazilian twins were created. The stored internal stress (high dislocation density) is released by recrystallization of the palisade-like arranged quartz grains. Recrystallization results in a quartz grain fabric, which is characterized by nearly dislocation-free grains of similar grain size and 120° angles of the grain edges at triple junctions (Wirth & Stöckhert, 1993, 1995; Langenhorst & Porier, 2002; Lathe et al., submitted).

FIB based TEM specimen preparation of fossils and microfossils

FIB application in preparation of TEM foils from fossils is demonstrated by a recent TEM investigation of the internal structure of a deep-sea sponge (monorhapis chuni). That animal is connected to the seafloor by a thin needle-like structure (typically several mm in diameter). Early investigations nearly 100 years ago found that the major constituent of the needle is opal. Cross sections of the needles reveal under the optical microscope a radial alternating pattern of bright silica-rich (20-30 gin thick) and dark, organic material rich layers (several hundred nm thick). The SE image in Fig. 10 displays a cross-section of such a needle etched with HE A TEM foil was cut that it contains both the bright silica-rich material and the dark material (insert in Fig. 10). However, TEM investigation revealed that HF etching produced artefacts. Therefore, we started from a polished cross-section without previous etching of the surface. The unetched surface did not display the radial pattern in SE images. In order to make the radial pattern visible again, we used the in-situ etching capability of FIB. Etching is performed by inserting XeF2 gas by means of a needle into the FIB close to the observed area. Scanning the surface of the sample with the Ga-ion beam and acquiring SE images, the ion beam decomposes the XeF2 and the surface is slowly etched (Fig. 11). The major advantage of this technique is the visual control of the etching procedure.

The TEM investigation of the FIB cut foil exhibits the nature of the radial pattern in the cross section. The bright parts in the SE images consist of amorphous, water rich silica (opal), whereas the thinner dark sections in the SE images are mainly composed of organic material. Figure 12 shows a TEM foil, which was cut from the outermost part of the needle. The SE image displays the foil still resting in the excavation site. The slightly etched surface of the specimen is visible. In the TEM bright-field image opal appears as dark grey uniform contrast, whereas the organic enriched thinner and slightly curved structures appear in bright contrast due to the lower mass absorption contrast of the organic material. The dark grey, net-like structure is part of the perforated carbon film the foil is placed on. C-elemental map using EELS (three window technique) demonstrates the presence of carbon in the thin layers. The structure of the needle can be described as a natural composite of organic and inorganic material. Strong bending of the fibre structure due to water current on the seafloor might cause cracks and finally failure of the structure, thus killing the animal. The composite structure avoids such catastrophic failure of the needle because the thin organic layers hinder crack propagation. The composite structure of the needle allows nature to minimize the diameter of the needle at maximum strength.

Applied mineralogy

Coatings on jet engine turbine blades

Jet engine turbine blades located in the hot section of an engine are exposed to high temperatures and corrosive gases during operation. To prevent catastrophic failure of the engine during operation in an aircraft the surface of the Ni-based super alloys of the turbine blades needs to be protected by coatings, so called thermal barrier coatings (TBC). The TBC layer, which is yttrium partially stabilized ZrO2 (Y-PSZ), is connected to the super alloy by an additional bond-coat layer (Ni-Cr-Co-Al-Y). Between TBC and bond-coat layer a so-called mixed zone of thermally grown oxides has formed already during deposition of the TBC layer. Figure 13a shows a cross section through a part of such a turbine blade (bright part of the optical micrograph). A close-up of the bond-coat layer, the TGO and the thermal barrier coat is presented in the SE-image of Fig. 13b. The position of the TEM foil to be cut is indicated by the small Pt-bar, which has been deposited prior to cutting the foil. The SE-image in Fig. 13c shows the TEM foil tilted at 45° after cutting the foil free at its base and at both sides. The different contrast in the left part of the image is due to orientation contrast of the grains of the bond-coat layer. Orientation contrast imaging (SE images) can be used in the FIB together with available software for grain size measurements and grain size statistics. The TGO layer is characterized by dark contrast in the SE image. Fractography of TBC coatings from jet engine turbine blades has shown that failure is closely related to the flaw size distribution and the particular phase assemblages of the thermally grown oxide (TGO) layer. The TGO layer comprises a coarse-grained corundum bottom layer close to the bond-coat followed by a finely dispersed mixed zone consisting of alumina and zirconia adjacent to the thermal barrier coating (TBC). The dynamic pattern of coexisting phases during nucleation and thickening of the TGO layer obtained via FIB cut foils by TEM investigation is supplemented by thermodynamic and constitutional considerations. The microstructural model so derived emphasizes that the microstructural approach by FIB and TEM as pursed in this study is mandatory for TBC lifetime assessment (Fritscher et al., 2004)

Micromachining of diamond

Recently the FIB instrument at the GFZ was operated to machine diamond anvils for use in a hydrothermal diamond anvil cell (HDAC). The HDAC was designed and constructed specifically for observing on fluid and fluid plus solid samples at high temperatures and pressures. Bassett et al. (2000) and Anderson et al. (2002) modified the HDAC by laser drilling a sample chamber and grooves into the diamond anvil to enhance the transmission of X-rays entering and exiting the fluid sample chamber. Precise milling of diamond with a FIB is useful for removing the exact amount of diamond to optimize X-ray transmission without compromising the strength of the anvil. It is also important in X-ray spectroscopic experiments with the HDAC to observe phase behaviour within the sample chamber as temperature and pressure is changed (Anderson & Wirth, 2003). However, the optical quality of a laser ablated sample chamber is poor due to the irregular surface of the floor of the chamber. This problem is overcome by milling a sample chamber with a focused ion beam. Figure 14 shows a FIB milled pit (114 × 108 × 65 μm) that was cut within 114 hours (FEI FIB200, Ga-source, 30 kV, 11500 pA). The smooth floor produced by FIB milling provides excellent observation of the contained sample using a transmitted light microscope. However, milling 114 hours at maximum beam current to get a relatively small pit is not acceptable. The milling time is significantly reduced by applying the above described selected carbon mill device (SCM).

Discussion

Artefacts and problems caused by FIB milling

Ga-ion implantation and damage of the surface layer

During the milling process and every time an SE-image is acquired the ion beam is scanned over a selected area and interacts with the sample. There are several mechanisms of ion beam/sample interaction. There is energy transfer from the ions to the target atoms by momentum transfer from nucleus to nucleus. Additionally, there is interaction between the electrons of the ion and the electrons of the target. The ion beam/material interactions are discussed in detail in Prenitzer et al. (2003). Finally, these interactions are responsible for the implantation of Ga-ions into the crystal lattice of the target material thus leading to a local disturbance of the lattice. Exceeding a certain amount of implanted ions the long-range order of the lattice gets lost and an amorphous state is approached. The near-surface amorphization of Si-material by the ion beam, which is especially sensitive to ion beam induced damage, was investigated and described by Engelmann et al. (2002). The study showed that operating the FIB using standard parameters (30 kV acceleration voltage, 350 pA beam current, sample tilt ± 1°) a near-surface amorphization in silicon of 17 run thickness at each side results. Because of the amorphization layer, it is useless to thin the sample to a final thickness of 50 nm or less. The amorphous layer would be 34 nm and only 16 nm of crystalline material would remain. However, FIB applied to metals using standard conditions causes only very thin amorphous layers - e.g. copper: ca. 1 nm (Engelmann et al., 2002). The nearsurface amorphization of silicates is in between that of pure Si and that of metals and depends on crystal structure. Polishing the sample to the final thickness with a 70 or 11pA beam instead of 350 pA, which is a standard parameter, can further reduce the surface amorphization layer.

Amorphization of the sample surface by ion implantation causes problems with X-ray analysis and high-resolution electron microscopy (HREM). The energy resolution of the solid-state detector usually is about 130 eV That means, the Na Kα-line (1.041 eV) interferes with the Ga Lα-line (1.098 eV) and the Ga-Kα-line (9.241 keV) overlaps with Ir-Lα (9.174 keV). However, in that case other Ir lines can be used for analysis purposes.

The amorphous layer affects HREM imaging by contributing noise to the image. This effect can be minimized using filtering methods such as Fourier (Gatan Digital Micrograph software package).

Redeposition of sputtered material

Another problem with FIB cut TEM foils is redeposition of sputtered material and gallium onto the foil surface. This might affect EDX-analysis as well as HREM imaging. Redeposition of material especially is a problem if the depth of the trench is large compared to its width. During sputtering material from deeper parts of the foil by the Gaion beam material can be redeposited in upper parts of the foil. This effect can be minimized by enlarging the trenches in front and in the back of the foil.

At a certain step in the automated cutting process, when a foil thickness of about 500 nm is reached, the foil is cut free at its base and at both sides as mentioned above. In many cases the cut at the base is healed again during the final polishing procedure by redeposition. of material, thus welding the foil to its base. Again, it needs to be cut free when the polishing process is accomplished. This additional cutting will redeposit material, especially gallium, in the lower parts of the foil and makes some additional polishing necessary.

Bending of the foil during milling

Bending of the foil mostly occurs in the very late stage of the final polishing process. In some rare cases, bending is observed already at foil thickness of 500 nm resulting in a sudden bulging of the foil. Principally, there are two reasons for bending. One is slight heating of the foil by the ion beam causing thermal expansion. After cutting the foil free at its base and at both sides, the foil is only connected to the matrix by two thin platinum strips (see Fig. 1d). The heat transport from the foil to the matrix is reduced by the thin Pt-strips, and consequently the foil will expand with respect to the surrounding matrix. If bending of the foil is expected to occur or has already taken place, one side of the foil can be cut free completely, thus releasing the stress, and the foil will straighten again and polishing can be continued.

Another reason for bending is local residual stress in the foil, which is observed in strongly deformed material. In such case, it might be useful to cut the foil free on one side to allow stress relaxation and foil preparation can be continued automatically or in a manual mode of operation.

The author would like to thank W. Heinrich for substantially supporting the idea of operating a FIB at the GFZ. Thanks to L. Dobrzhinetskaya for drawing the authors attention to FIB technique. Constructive reviews by L. Dobrzhientskaya, J. Orloff and an anonymous person are gratefully acknowledged.