Preparation of TEM samples of metal–oxide interface by the focused ion beam technique
Paul Scherrer Institut, Laboratory for Materials Behaviour, 5232 Villigen-PSI Switzerland *Swiss Federal Institute for Materials Testing and Research: EMPA, Überlandstrasse 129, CH-8600 Dübendorf, Switzerland
Paul Scherrer Institut, Laboratory for Materials Behaviour, 5232 Villigen-PSI Switzerland *Swiss Federal Institute for Materials Testing and Research: EMPA, Überlandstrasse 129, CH-8600 Dübendorf, Switzerland
This paper describes a procedure to prepare metal–oxide interfaces for transmission electron microscopy by the focused ion beam technique. The advantage of this procedure is to allow the observation of metal–oxide interfaces of irradiated samples with a homogeneous thickness without the need to have an instrument inside laboratories that are specialized for the manipulation of irradiated materials. A transmission electron microscopy sample is prepared by this method and analysed.
Observation of metal–oxide interfaces by transmission electron microscopy (TEM) requires delicate specimen preparation procedures. The metal–oxide interface can be described as a plane (but not a monolayer) that should be thinned to electron transparency, usually in a direction perpendicular to the interface for subsequent analysis. Thinning of the interface of an oxide is more demanding than that of a multiphase or composite material, because in the latter cases the material contains several interfaces and any thinned region will comprise at least one sufficiently thin interface. Therefore, the thinning process is not limited to a single interface region. This is not the case for a metal–oxide interface and the success rate for sample preparation is much lower, due to the fact that only a very specific plane (in the transverse section) has to be electron transparent. This point is illustrated by the schematic drawing in Fig. 1.
In the case of the metal–oxide interface of irradiated materials, the handling becomes an extra limitation. In addition, the need to perform experiments in a specialized laboratory means that the possibilities are limited by the laboratory equipment available.
The choice of technique for each material depends on the type of analysis required, as well as the limitations (characteristics and geometry) of the material. In the case of zirconium alloys, the oxidation of the material induces large internal stresses due to the difference of volume of the oxide with respect to the metal (Pilling-Bedworth ratio of 1.5). These stresses make sample preparation more difficult as the thin region could easily shatter if the stresses are not homogeneously distributed in the final sample.
One of the techniques that has shown advantages for the preparation of metal-oxide specimens in the case of these alloys is the FIB family of instruments (Abolhassani et al., 2001, 2002, 2003). Briefly, the method consists of scanning an ion beam over the surface of a sample, in a similar way as the electron beam is scanned in a scanning electron microscope. The ion beam is focused to a very small diameter that allows cutting the sample from the bulk material.
Three techniques are known for TEM sample preparation by FIB, namely ‘lift-out’, ‘prethinned’ and ‘block lift-out’. In the lift-out technique, a bulk specimen is used as a starting material; in the ‘prethinned’ technique a TEM sample is prethinned by other techniques and the region of interest is further thinned by FIB and finally, ‘block lift-out’ is a combination of the two previous techniques and is used when the carbon film is not desired. In this technique, the sample is thinned in the same way as in the lift-out technique and, at a thickness of 1 µm, is removed from the bulk, fixed to a TEM grid by Pt deposition and remounted in the FIB in the perpendicular direction for final thinning. In this study the ‘lift out’ technique was used.
When an instrument with a ‘dual beam’ can be used, i.e. where an electron beam is also available for the imaging of the sample, the region of interest is selected without causing damage to the specimen (in the case of imaging with an ion beam, the material will be slowly eroded prior to milling, removing the fine details present at the surface). The advantage of the associated electron beam, besides the possibility of accurately selecting the region before cutting, is to observe the thin film during the cutting and thinning of the sample.
The use of FIB for the preparation of irradiated samples is more complicated due to the contamination caused by these materials. In addition, such machines are very often not available in a specialized laboratory (also called ‘hot-laboratory’) and if so the transport of irradiated materials to such a laboratory is very complex.
However, if the specimen is thoroughly decontaminated in a specialized laboratory and has a radioactivity level that is below the exemption limit (for handling radioactive materials), the use of a FIB that is installed outside a specialized laboratory becomes possible for the preparation of TEM samples.
The aim of the present study was to describe the conditions necessary for the preparation of such a sample, and to show the details of the specimen preparation steps and the precautions to be taken. The results of the TEM analysis of a sample prepared by the above technique are presented.
Materials and methods
A category of zirconium-based cladding material irradiated in a Swiss pressurized water reactor was used for this study (Garzarolli et al., 1996). The material was irradiated in the reactor for four cycles (in this case each cycle was 11 months). Standard uranium oxide fuel was used in the fuel rod. It had shown a maximum oxide thickness in the order of 59 µm in the reactor and at the elevation studied the oxide thickness was 25 µm. In order to prepare a segment that could be used for FIB sample preparation, a small ring was cut from the cladding tube and the fuel was removed from it in the hot laboratory in a closed environment known as ‘hot-cell’. The ring was then cut into small segments in the same laboratory (Fig. 2a–c). Each segment was decontaminated from the inner wall by mechanical grinding in order to avoid the presence of any residual fuel in the cladding segment for further manipulation. The sample was then ultrasonically cleaned in ethanol to remove any loose contamination. Care was taken to avoid any acid cleaning which might attack the oxide or the metal and modify the chemical and structural properties of the specimen.
The segments were then transferred from the hot-cells to a standard metallography laboratory inside the specialized laboratories and two such segments were glued together, using an appropriate adhesive (M-Bond, Vishay Electronics GmbH, Germany), from their oxide faces as shown in Fig. 2(d). After re-dimensioning to the required size, they were mounted into a special TEM holder (also known as a Barna holder in reference to the author of a technique for TEM sample preparation by ion milling). Figure 2(e) shows the two segments after being mounted in the TEM holder. The sample so prepared was then ground from both sides until the sum of activity of the different isotopes fell below that of the lowest permitted activity for the hardest isotope. In the case of the sample studied here, this corresponded to the value for 60Co which is 3000 Bq (Swiss Legislation on Radiological Protection, 2001). The sum of activity of all isotopes in the specimen discussed here was below 800 Bq. The sample was then perfectly safe to be used outside the specialized laboratory.
The safety precautions were respected although the specimen had an activity that was far below the exemption limit for handling and could be handled without special precautions.
Once the specimen was in the FIB laboratory it was mounted on a holder using a silver paint and introduced to the FIB chamber.
TEM sample preparation by focused ion beam
In this experiment an FEI Strata DB 235 dual beam® (DB) FIB workstation was used (FEI, Eindhoven, The Netherlands). The DB workstation has an electron column with a field emission gun and is equipped with a secondary electron detector and through lens detector. The resolution of the electron beam is specified to be 3 nm at voltages from 1 to 30 kV, however, for each voltage a different working distance (2 mm for 1 kV to 7 mm for 30 kV) gives the optimum resolution. The ion (Ga) column is mounted at an angle of 52° with respect to the electron column. The ion beam current can be adjusted using 12 apertures from 1 to 20000 pA at 30 kV. The resolution of the ion beam is specified by the manufacturer to be 7 nm for 30 keV and 1 pA. Scan and pattern generators are digital.
The TEM specimens were prepared by milling an electron-transparent sample (about 20 µm long, 5 µm wide and 100 nm thick) out of a bulk sample. The preparation procedure consists of three main steps, described in the following sections.
(a) Selection of the area of interest and preliminary-thinning. The electron beam was used to examine the metal–oxide interface carefully and to select an intact region, i.e. a clean region with few cracks and also free from oxide spalling due to sample handling and sample preparation. Prior to ion milling, a platinum protection layer was deposited over the area of interest to protect the specimen surface from ion beam damage. Figure 3 indicates the area selected and the Pt coating. Subsequently, stair-step trenches were milled on both sides of the selected region. Furthermore, trenches were milled on the two extremities of the future sample to allow for stress release later on during the specimen preparation (Fig. 4). The remaining material was then thinned to a thickness of about 0.6 µm.
(b) Final thinning. At this stage, the specimen was cut at the bottom and partially on the sides. To do so, the specimen was tilted 52° with respect to the ion beam and a beam current of 300 pA was used for cutting. This operation at this early stage releases the stresses and reduces the damage to the thinned specimen (Fig. 5). These trenches and cuts on both sides of the specimen will absorb some of the compression in the thin foil. Thus, the tendency of the foil to bend during further thinning will reduce and the sample will not shatter. Further thinning to about 100 nm was performed in several steps using ion beam currents from 300 to 30 pA and a grazing ion beam incidence of 1.5°.
As the image contrast in TEM is dependent on the nature of the element observed and its atomic number (the atomic number of zirconium being equal to 40), in order to have sufficient contrast for the analysis of the structure, very thin regions are necessary for appropriate TEM analysis. To obtain such regions, the thinning of the sample can be performed according to two configurations: a wedge form can be given to the top or to the bottom of the sample. Both options were tested in this study. The sample was tilted less than 3° with respect to the ion beam and the thinning was continued at the top, or at the bottom. It was expected that having the thinnest region at the bottom would be advantageous, as it would cause less ion beam damage. However, it was observed that the bottom part had more damage from the ion beam and the thinned part at the bottom could not be used for imaging. This point needs further examination.
At the surface of the sample, an amorphous layer is usually formed due to ion beam damage. The thickness of this layer was reduced in a last step using a low energy ion beam. The typical operating parameters of the FIB at each stage are presented in Table 1.
Table 1. Typical focused ion beam parameters used for the different stages of TEM sample preparation for metal–oxide interface of zirconium based materials.
Acceleration voltage (kV)
Probe current (pA)
Making a stair-step trench
Removing amorphous layer
(c) Cutting and removal (lift-off) of the specimen. After the final polishing, the lamella was completely cut off at the sides. Figure 6 represents the thin foil after it was totally separated from the bulk material. The bulk specimen was then taken out of the FIB instrument, and the lamella was placed on a standard TEM grid with a thin carbon film, under an ex situ optical microscope. A micromanipulator arm with a glass rod attached was used to extract the sample and place it onto the TEM grid.
Internal stresses in the specimen
One of the main problems with this type of material in the process of thinning is the risk of shattering of the oxide due to its brittle nature and to the stresses present in the material as mentioned above. Our experience from the ion milling had shown that when the oxide layer is thin it can very easily break off. The oxide being under compression and the metal under a slight tension (Pilling-Bedworth parameter of 1.5), the foil could bend around an axis parallel to the interface, but also around an arbitrary axis in the plane of the foil. These deformations could lead to the shattering of the oxide.
To avoid this phenomenon, a type of frame was created to support the thin part. For this, the thin film was precut after reaching a thickness of about 0.6 µm. Only the central region of the lamella was thinned to electron transparency. The remaining structure was thick enough to resist bending.
A JEOL2010 (JEOL, Japan) operating at 200 keV (with a LaB6 filament) equipped with a Link-ISIS energy dispersive X-ray spectrometer (EDS) (Oxford Instruments, U.K.) was used for conventional TEM observations.
In order to show the characteristics of a TEM sample prepared by this technique, the advantages of FIB samples and the possible artefacts encountered, TEM images are shown in this section. In this study, two specimens were prepared by the above technique from the same material and both were used to examine the microstructure of the interface and the distribution of different elements by energy dispersive X-ray microanalysis. At the thin edges of the samples, at the metal–oxide interface, an attempt was made to perform high resolution TEM. However, only one of the specimens was sufficiently thin for high resolution TEM. The damage by Ga ions and the thickness of the sample were at the origin of this problem. The reasons for avoiding further thinning of the material were the extremely delicate nature of the sample due to internal stresses from metal-oxide phases and the presence of cracks in the oxide. The TEM analysis of the samples is presented below. Further analyses of the structure of these materials are presented elsewhere (Abolhassani et al., 2005).
Figure 7 represents a dark field contrast of a specimen showing the interface. Several aspects of the microstructure of the material at the oxidation interface are revealed. The following points are considered to be the advantages of this sample preparation method:
(a) The limits of the metal–oxide interface can be clearly detected.
(b) The cracks in the oxide region parallel to the interface can be observed. It is clearly concluded that these cracks are inherent to the material, as can be observed in Fig. 5. When the sample is being cut, the E-beam image of the specimen reveals cracks in the oxide. Figure 8 further indicates these cracks in the lamella during cutting. This point has been verified by examining different areas of the sample (not shown here).
(c) The microstructure of the material in the vicinity of the interface is observed and can be analysed. As shown in Fig. 9, the metal side of the interface shows a large density of radiation damage and black dots are observed in this sample. The metal side of the interface shows α-Zr crystal (hcp) structure and the density of c-dislocations is high; the oxide has a monoclinic structure (Abolhassani et al., 2005).
(d) The grain size of the oxide is measured near the interface at higher magnifications to be in the order of 20–40 nm and the shape of grains is mostly equi-axed (Fig. 9).
(e) In this specimen, the microstructure of the metal in the vicinity of the interface has been revealed to contain a large number of platelets which are expected to be hydrides. The bright field contrast and the diffraction pattern of the hydride phase can be observed in Fig. 10. The nature of hydrides in this material has been examined in the second sample. The analysis of the structure of these platelets corresponds to either δ-hydride or ɛ-hydride. It is possible that these materials would oxidize in the vicinity of the interface. This point should be examined by studying several different materials (Abolhassani et al., 2005).
(f) The distribution of alloying elements can be examined from the EDS maps of different constituents of the material (Fig. 11). As the thickness of the thin foil is homogeneous, such analysis can be performed without ambiguity. As the irradiation and oxidation cause a change of distribution of the alloying elements, it is useful to have an accurate evaluation of these changes (Abolhassani & Bart, 2004). This is not possible in the case of standard argon ion-milled specimens, as samples do not show homogeneous thicknesses (Fig. 12).
(g) The shape of the interface can be examined from the large area of observation; the present material has a corrugated interface (Fig. 11). The oxygen map confirms this point. A precipitate was observed in the oxide side of the interface, it contains both iron and chromium. Precipitates in these alloys provide an anodic protection for aqueous corrosion (Isobe et al., 1996). Such precipitates are not observed very often in the oxide after four cycles of irradiation; as a large number of them dissolve under irradiation (Garzarolli et al., 1996; Abolhassani et al., 2000).
(h) A series of point analyses along the interface showed oxygen dissolution in the metal side (Fig. 13). The zirconium-oxygen phase diagram predicts the maximum solubility of oxygen to be in the range of 28.6 at.% at the temperature range of up to 500 °C. However, the results of our observations show slightly higher oxygen concentrations in the range of 36 at.%. This point is to be further examined in order to find the origin of the discrepancy.
The extended TEM analysis of this irradiated material was only possible because the thickness of the foil was uniform. Therefore, microstructural and chemical analysis could be performed with a large accuracy. Such observations make the quantitative analysis of the metal–oxide interface feasible (Abolhassani & Bart, 2004; Abolhassani et al., 2005).
Damage due to Ga ion bombardment. The preparation of TEM thin foils can induce different types of artefact depending upon the sample preparation technique that is used; this is also the case for FIB sample preparation. In the FIB method, the sample is scanned with Ga ions during ion milling or imaging with the ion beam. Therefore, Ga can be implanted to a certain extent in the specimen surface. This point has to be taken into account in analytical microscopy, although it does not generally affect the quality of the TEM images made in bright field or dark field modes. In some cases however, certain areas of the sample can be coated with Ga and examination of the structure would not be possible in such cases. Figure 14 shows a region with Ga deposited on the surface of one of the specimens prepared in this study.
Amorphization of the structure is another artefact of FIB specimen preparation due to ion bombardment. The thickness of the amorphous layer produced depends on the nature of the material, the beam incidence and the acceleration voltage (Engelmann et al., 2002).
To avoid these artefacts, generally the imaging time with the ion beam should be kept as short as possible, and the beam current should be low. A reduction of the amorphization depth at each side of the lamella (and therefore a better contrast in TEM analysis) can be achieved with the following final procedure: the ion-beam acceleration voltage is reduced from 30 to 7 kV. A aperture for 300 pA ion-beam current is used which results in about 60 pA at 7 kV. The lamella is tilted several degrees out of the ion beam axis (4 to 6°). A fast scan time over the selected specimen area is applied for about 6–8 s, until a contrast change from light to dark becomes visible (Table 1).
Mass of material removed from the sample. For experiments dealing with irradiated material the contamination caused in the FIB chamber can be of concern. This point can be checked by performing a small calculation based on the results of a cutting experiment. The amount of material milled from the specimen can be calculated from the dimensions of the region that is milled during preparation. From the SEM images the average volume milled in the material is of the order of 3.7 × 10−9 cm3. This volume corresponds approximately to a mass of 24 × 10−9 g. The amount of material removed is therefore negligible. Also the mass of final specimen is in the range of 0.1 ng, so it can be handled without any risk.
Two aspects of the TEM observations are of interest, the possibility of observing a large region in a single thin foil and of examining the transition of the metal to the oxide at the interface. One of the interesting results in this study is the presence of an oxygen profile in the metal side of the interface. Although this concentration gradient has been predicted, and is expected from the diffusion calculations, it is very difficult to obtain such results with other experimental methods, as the concentration gradient drops rapidly with distance and other techniques can not provide such a high spatial resolution in order to analyse the vicinity of the oxide.
The high oxygen concentration in the metal side of the present sample implies that this material should oxidize faster than a material with low oxygen concentration in the metal side, as the oxidation rate has been shown to depend on the oxygen content of the metal. There is a possibility that the metal surface of the thin foil oxidizes after sample preparation, however, such an oxidation layer should lead to a homogeneous high oxygen signal as the depth of penetration of the oxygen should be the same over the whole sample. If we assume a thin oxide layer on the metal side of the interface, it could mean that there is an oxygen signal from the bulk material over which an additional surface oxygen signal must be counted.
The intensity of the oxygen signal from the oxide is strongly dependent upon the thickness of the foil. This point was examined on a ZrO2 powder and it was observed that the relative intensity of the oxygen signal decreases with increased thickness. The same argument is valid for the metal side of the interface. This parameter could induce an error of estimation by underestimating the oxygen content of the material if the thickness is incorrectly measured. In the case of the present sample, the signal from the oxide can be used as a control. The fact that the oxide signal is in agreement with the ZrO2 structure implies that the thickness of the material was correctly estimated. However, the information can only be considered as semiquantitative and a very accurate measurement of the thickness and the sum of several readings at each point are required for a high compositional accuracy. The results are nevertheless a very useful indication of the composition of the interface.
In comparison, in the case of an ion-milled sample where the sample thickness varies rapidly, it is not possible to perform such a point analysis along the interface and each point measurement should be corrected by the thickness measured at the same point.
The use of FIB for the preparation of the metal–oxide interface of irradiated cladding material is possible by making a sufficiently low activity sample inside a specialized laboratory that will subsequently be transferred to a FIB instrument for the preparation of samples.
The structural and chemical analysis of a sample by TEM is possible and the uniform thickness of the sample together with the small amount of surface roughness allows an accurate chemical analysis. As can be observed, the advantage of such specimen preparation, besides the uniformity of the thickness of the film, is in the accuracy of selection of the region to be observed by TEM and the control of the quality of the specimen during the preparation procedure.
The artefact caused by ion bombardment can be detected; in this family of alloys usually no Ga is added, therefore the deposition of Ga can be checked. However, care must be taken to consider the presence of such features and artefacts in order to avoid misinterpretation of the results.
The TEM observations of the metal–oxide interface of the material studied indicated the presence of hydrides in the vicinity of the interface. The size of oxide grains in this region was of the order of 20–40 nm. The oxide showed cracks parallel to the interface. These cracks were also observed during specimen preparation by FIB (Fig. 6) and it is concluded that they are inherent to the material. The interface in this material is corrugated and at the region observed no plane interfaces are observed.
The authors wish to thank Dr K. Ospina and Dr T. Lüthi for the organization of the transport of the prepared sample. Dr A. Hermann is acknowledged for supplying the irradiated material. Mr T. Rebac and Mr R. Restani are thanked for their assistance in decontamination and mounting of the specimen.