Preparation of Binary and Ternary Laves and µ-Phases in the Ta–Fe(–Al) System for Property Analysis at the Microscale

In the binary system, the Laves phase ranges from 27.7 to 40.9 at. pct Ta and the µ-phase from 46 to 60.5 at. pct Ta, see Figure 2.^[7] The nearly stoichiometric compositions were prepared for the TaFe₂ Laves phase with 33 at. pct Ta and the Ta₆Fe₇ µ-phase with 46 at. pct Ta. Furthermore, in the µ-phase, the Fe atoms of the triple layer can be replaced by Ta atoms, resulting in the Ta₇Fe₆ µ-phase with 54 at. pct Ta.^[18,19]

In addition, off-stoichiometric compositions are relevant for a thorough characterization of both phases, as a property change is to be expected due to the altered atomic size ratio. In order to cover the entire homogeneity range of the phases as far as possible, the Ta content for the Laves phase was reduced and increased by 5 at. pct, respectively. For the Ta₆Fe₇ and Ta₇Fe₆ µ-phase, the Ta content was increased by 4 at. pct in each case, allowing the off-stoichiometric composition between the two stoichiometric compositions as well as close to the phase boundary with a high Ta content to be represented in equal steps. This results in seven different compositions for the binary Ta–Fe system. All target compositions are indicated in Figure 2 and can be found in Table I, listing the content of the elements in both atomic and weight percentages.

In the ternary system, the Laves and µ-phase can have an Al content of up to 56 and 39 at. pct, respectively, see the solidus projection in Figure 2.^[8,9] Witusiewicz et al.^[9] calculated the liquidus and solidus projections for the Ta–Fe–Al system as well as various isothermal sections between 1700 °C and room temperature. Due to the use of an arc melter (Section II–C), where the sample is cooled to below 100 °C within a few seconds by the inert gas atmosphere and the water-cooled crucible plate, the solidus projection was chosen as a starting point for planning the sample compositions (cf. Figure 2). However, to really ensure that it is possible to prepare homogeneous Laves and µ-phase samples with the targeted stoichiometric compositions, the phase ranges of the liquidus and solidus projection as well as the isothermal section at 600 °C were superimposed, as shown in Figure 4.^[8,9] In this way, the solidification process according to the ternary phase diagram is considered during the sample synthesis, whereby the process sequence is clearly defined through the use of an arc melter: the input materials are melted together, i.e., they are in a liquid state and then solidify when the arc is switched off. At this point, due to the rapid cooling, which largely suppresses diffusion, it is important to be within the phase range in order to obtain the targeted phase.

Since the alloying of Al results in an additional degree of freedom that significantly increases the possibilities of sample compositions, the content of the larger Ta atoms in the ternary samples was kept constant at the stoichiometry and only the relative content of Fe and Al, that more likely share the smaller sites, was varied. The prototype µ-phase with 46 at. pct Ta, where the triple layer does not only consist of Ta atoms, is not realizable in this ternary system. Therefore, the µ-phase with 54 at. pct Ta was chosen as well as the nearly stoichiometric composition of the Laves phase with 33 at. pct Ta. With the Ta content remaining constant, half of the Fe was replaced by Al as starting compound, i.e., 33.5 at. pct each of Fe and Al for the Laves phase and 23 at. pct each of Fe and Al for the µ-phase, see Table I.

For the Laves phase, the Al content can be increased by more than half the Fe content. Under the premise of varying the composition as much as possible to investigate the influence of Al on the physical properties of the system, the content of 33.5 at. pct was increased as well as decreased by 11.5 at. pct. This results in Al contents of 22 and 45 at. pct, with 33 at. pct Ta and 45 and 22 at. pct Fe, respectively (cf. Table I).

Regarding the µ-phase, Figure 4 shows the following: with a constant Ta content of 54 at. pct, a further increase of the Al content would cause the phase region to be left. Therefore, at most, half of the Fe can be replaced by Al (23 at. pct each) and the Al content was reduced accordingly in 7 at. pct steps to 16 and 9 at. pct Al. The Fe content is consequently 30 and 37 at. pct, see Table I. Altogether, this results in six different compositions for the Ta–Fe–Al system, which are marked in the superimposed phase diagrams in Figure 4. Together with the seven different compositions in the binary system, three for the Laves and four for the µ-phase, this amounts to 13 different compositions for the Ta–Fe and Ta–Fe–Al systems, which are all highlighted in the phase diagrams in Figure 2.

Furthermore, the actual compositions of the samples determined by EDS are given in Table I. The column EDS Sum given in at. pct refers to the total result of EDS area measurements including possible second phases. The EDS PoI Line column, also given in at. pct, shows the composition of the phase of interest (PoI). For this purpose, the composition of the targeted phase was determined by line measurement within the performed area measurement. Further details on the performed EDS measurements are described in Section II–E.

When selecting the input materials, a high degree of purity is of decisive importance to avoid precipitates. C impurities in particular are common for the elements used and can lead to the formation of carbides, which significantly affects the microstructure, especially the homogeneity and average grain size. Therefore, not only the amount but also the type of impurities must be considered when selecting the input materials. To demonstrate this effect, more and less pure materials and their influence on the sample synthesis were tested. The ones that led to better results in terms of sample homogeneity and grain size are written in black in Table II and are indicated together with the purity levels of the individual elements given by the manufacturers as well as the performed processing steps. They have a purity of at least 4N (99.99 pct). Tested input materials that were found to be unsuitable are listed in gray font, with the problems encountered also indicated. However, the selection of these materials was not only based on the amount and type of impurities, but also on the piece size. For the sample synthesis, a small-scale laboratory arc melter was used (Section II–C) and a total sample weight of about 2.5 to 4 g, depending on the composition, was aimed for. This requires pieces of the input materials that are as small and easy to batch as possible. In order to obtain the most accurate sample composition possible, a scale with a readability of 0.1 mg was used (Mettler ME104T). With this scale, the minimum sample weight is 16 mg and the maximum measurement deviation at the factory is 0.2 mg. To determine the conditions prevailing at the measurement location, a certified calibration weight of 5 g was weighed 10 times and a measurement uncertainty of 0.055 mg on average was calculated from these measurements.

The used Fe has a purity of 99.99 pct (included in black in Table II). It was electrolytically deposited with an increased H and O content compared to other Fe. However, this content can be reduced by remelting it. Another tested Fe (included in gray in Table II) appeared to have the same purity on the provided analysis data sheet, but with the C content not being analyzed. Because of difficulties with the first synthesized samples, which exhibited inhomogeneity in terms of additional unwanted phases and a small grain size, a further analysis was performed measuring three Fe pieces, resulting in an average C content of 288 ppm. For the preparation of intermetallic phases, however, C should be avoided as far as possible, as undesired reactions lead to carbide formation and fine microstructures. In case of the Ta–Fe system, Ta carbides were formed among a number of other phases. By using the remelted electrolytically deposited low-C Fe, homogeneous samples with a considerable improvement in grain size could be realized (cf. Section III–C, EBSD maps after using different input materials).

The Ta used has a purity of more than 99.99 pct and does not need to be additionally treated before use (included in black in Table II). Semi-finished Ta with a purity of 99.95 pct as a starting material was also tested, but still showed white residues from laser processing, which are presumably oxides (included in gray in Table II). To remove these residues on the cutting edges, the semi-finished Ta was etched with 65 pct HNO₃ in water in a ratio of 1:1.5. Nevertheless, the initial purity of 99.95 pct could not be regained, as not all impurities were removable. Therefore, final synthesis was performed using the aforementioned Ta with a purity of more than 99.99 pct.

The Al used for the fabrication of the ternary samples has a purity of 99.999 pct. Since it had to be rolled for further processing, a two-step etching process was then carried out to remove the resulting contamination. For the first etching, KOH diluted in water in a ratio of 1:4 was used. The second etching step consisted of a mixture of HNO₃ and HCl in water in a ratio of 1:3:2.

Apart from the purity of the input materials, the work must of course also be clean in other regards in order to obtain homogeneous samples of the targeted compositions. To this end, all contact surfaces and objects must be cleaned before use and the tools used for mechanical comminution should be element specific.

A variety of synthesis techniques is suitable for the fabrication of Ta–Fe-based materials, ranging from powder-based solid-state sintering to large-scale induction melting. For the aims of this study, the following specific technological characteristics are most relevant: Liquid metallurgical techniques are preferable as they allow thorough homogenous intermixing of the constituting elements. The melt should ideally not be contained in refractory or graphite crucibles in order to avoid both chemical contaminations (possibly affecting phase stability regimes and deformation mechanisms) and unwanted inclusions (which could, e.g., limit the grain growth aimed for here at elevated annealing temperatures). Furthermore, sufficiently high melting temperatures must be achieved in order to completely liquify Ta (melting point of 3017 °C), and solidification should occur rapidly to avoid or at least limit segregation effects (cf. Figure 2), and thus, inhomogeneities of the sample. This is more readily achievable with smaller charge weights in the order of a few grams, which are sufficient for nanomechanical characterization methods and have the additional benefit of limiting alloy cost, although causing more effort to achieve precise chemical compositions.

These criteria can ideally be met by using a small-scale laboratory arc melter (Compact Arc Melter MAM-1, Edmund Bühler GmbH), in which sufficiently pure metals are melted on a water-cooled Cu-crucible plate under a protective gas atmosphere. In this study, charges between 2.5 and 4 g were prepared under an Ar atmosphere with a W electrode. The maximum arc current of the used arc melter is 200 A. Each sample was remelted for about 5 to 15 seconds and flipped at least four times to ensure full dissolution of Ta [cf. remaining Ta core in Figure 7(a)]. Solidification occurred in the order of 1 to 3 seconds once the arc was switched off. Al should only be added to already prepared Ta–Fe samples and subsequently remelted to avoid unnecessary overheating and evaporation.

Heat treatment or remelting are common techniques to make inhomogeneous or fine-grained samples usable for the following investigations by homogenization, recrystallization, and/or grain growth. The heat treatment of the samples was carried out using a vacuum retort furnace, at a pressure of approximately 10⁻⁵ mbar and a maximum temperature of 1130 °C, measured at the sample position. Since the melting points of the TCP phases considered are above 1750 °C, i.e., the heat treatment temperature is only less than two thirds of the melting temperature, long holding times of at least 500 hours were chosen. After heat treatment, the furnace was ventilated with Ar and the samples were then directly quenched in water.

Higher temperatures could be achieved using a high-frequency induction furnace (Ambrell EASYHeat 5060 LI, AMERITHERM Inc.) with a maximum power of 6 kW and a frequency of 150 to 400 kHz operated under an inert gas atmosphere, here Ar. This furnace allows both heat treatment by keeping the temperature close to the melting point and complete melting of the input materials for sample synthesis or remelting of already synthesized samples. As crucible material, Al₂O₃ as well as ZrO₂ were used, whereby ZrO₂ allows higher operating temperatures. For the Ta–Fe system, the furnace was heated to a maximum temperature of up to approximately 1900 °C at a frequency of approximately 300 kHz. Quenching of the samples is not possible with this furnace. By switching it off and increasing the Ar gas flow, cooling to room temperature is achieved within 10 minutes.

To achieve a microstructure of the synthesized phases that is suitable for subsequent investigations of crystal orientation and mechanical properties, microstructure characterization, in particular of sample homogeneity, phase and texture analysis, grain size, and surface quality, is indispensable. Details on shaping of the samples and metallographic preparation are given in the supplementary material. We identified grinding as the most suitable method for shaping and long polishing times as a key ingredient for high surface quality. The samples were examined using various analytical methods: Optical microscopy (Leica LEITZ DM RM) during and after metallographic preparation, the acquisition of secondary electron (SE) and backscattered electron (BSE) images, and more in-depth analysis using EDS and EBSD (Helios NanoLab 600i from FEI Inc. equipped with Octane Super A EDS detector and Hikari XP2 EBSD detector, both EDAX; software used for EDS quantification and EBSD indexing also from EDAX).

The SE and BSE images of the different samples were taken with an accelerating voltage of 5 kV and a beam current of 0.69 nA at a working distance of 4 and 5 mm, respectively. To analyze slip traces around the indents and on the deformed micropillars, the ultra-high resolution (UHR) mode was used with a working distance of 1.8 and 4 mm, respectively, at an accelerating voltage of 5 kV and a beam current of 0.34 nA.

For EDS and EBSD measurements, the acceleration voltage was increased to 20 kV and a beam current between 2.7 and 5.5 nA was chosen. Due to the detector alignments, EDS measurements were acquired at a working distance of 4 mm and EBSD or combined maps at 10–11.5 mm working distance with a 70° pre-tilted sample holder. The EBSD maps have a step size between 0.3 and 2.0 µm depending on the length scales of the microstructure to be analyzed. The EDS analysis was performed with a dwell time of 200 µs. For the EDS results given in Table I, area measurements with dimensions of 138 µm × 176 µm were performed. The line measurements have a length of 35 µm and 10 points each in X- and Y-direction. To evaluate the EDS measurements, the background was manually adjusted with the reference points kept constant within the phases to ensure the comparability of the different compositions. Due to the high atomic number of Ta of 73, the L line was selected, while the evaluation of Fe and Al was done based on K lines. The PeBaZAF base model was used for quantification. All EBSD maps are uncorrected. The grain orientations of the inverse pole figure (IPF) refer to the sample normal.

Room-temperature nanoindentation tests were performed using a load-controlled nanoindenter (iNano Nanoindenter, KLA Instruments) equipped with a diamond Berkovich indenter tip (Synton-MDP AG) and analyzed using the method defined by Oliver and Pharr.^[23,24] The experiments were carried out with a maximum load of 45 mN at a strain rate of 0.2 s⁻¹ resulting in a final depth between 350 and 400 nm depending on the tested sample.

For micropillar compression tests, the micropillars were prepared using a Dual-Beam FIB (Helios NanoLab 600i, FEI Inc.). They have an upper diameter of 2 µm, a height of around 5.25 µm, and a taper angle of approximately 3 deg. For the milling, an accelerating voltage of 30 kV was used and the beam current was decreased for each step starting from 21 nA to 80 pA. The deformation of the micropillars was done using a displacement-controlled picoindenter (Hysitron PI 89 SEM PicoIndenter, Bruker), in situ in an SEM (UHR SEM CLARA, TESCAN). A diamond flat punch with a diameter of 5 µm was used as indenter tip.

Inhomogeneous samples with small grain size were heat treated. Despite long holding times in the vacuum retort furnace of 504 and 672 hours at approximately 1130 °C and subsequent quenching, no significant change in the microstructure was observed. Due to the high melting points of the studied intermetallic phases of the Ta–Fe and Ta–Fe–Al systems with over 1750 °C, heat treatments accordingly need to be carried out at higher temperatures. To achieve such a temperature range, an induction furnace was used. Samples can be completely synthesized in an inert gas atmosphere or heat treated as a follow-up step after synthesis. The average grain size could, thus, be significantly increased to mm dimensions. However, inhomogeneities, like additional phases and inclusions, were not eliminated, but intensified by reactions with and possible contaminations from the crucible material. Eventually, the more significant influence on the improvement of the microstructure was found to be the change to purer input materials, as described in Section II–B. This made subsequent heat treatment obsolete and the TCP phases with sufficient grain size could be synthesized only by using the arc melter with an adjusted synthesis route.

The purity of the starting materials proved to have a major impact on the phase and grain formation. A comparison of EBSD maps is shown in Figure 8 and 9 for the standard route that is without any remelting, ordering, or dividing of the input materials, and the improved synthesis route, as described in Section III–B–1 for the binary and in Section III–B–2 for the ternary compositions, in addition to using higher-purity input materials. In Figure 8, the improvement for two Laves and µ-phase samples of the binary Ta–Fe system is illustrated as a combination of IPF and image quality (IQ) maps. The Laves phase sample in (a) with a Ta content of 33 at. pct shows slight inhomogeneities as well as a relatively fine grain structure. By changing the input materials towards the untreated Ta with higher purity and the remelted Fe with lower C content, and following the improved synthesis route for binary compositions described in Section III–B–1, homogeneous samples with much larger grains could be synthesized. The resulting Laves phase sample λ_b2 is shown in image (b).

The same applies to the µ-phase samples, where the improvement is even more pronounced. Here, the grain size could be increased from a few µm, as seen for the first µ-phase sample with 53 at. pct Ta in image (c), up to mm dimensions for the µ-phase sample µ_b3 with 54 at. pct Ta shown in (d).

Microstructural improvements were also achieved for the TCP phase synthesis of the ternary Ta–Fe–Al system, as shown in Figure 9. Since the binary samples were prepared first, the gained knowledge could be used directly for the synthesis of the ternary Laves and µ-phase samples. However, the addition of the Al also creates new challenges regarding the sample synthesis, e.g., because of the different melting points of the input materials. In the ternary case, all samples were synthesized with the remelted low-C Fe. The EBSD maps (Figure 9) reveal the effect of changing the Ta input material towards higher purity and selecting the improved synthesis route for ternary compositions, as described in Section III–B–2.

The ternary Laves phase sample λ_t2* with 30 at. pct Ta in Figure 9(a) already has a relatively large grain size, but also shows the presence of a second phase, which could be determined by EBSD as B2 FeAl with a cubic crystal structure. It formed due to the already low Ta content and an additionally remaining small but almost pure Ta core. Consequently, the formed Laves phase also deviates from the target composition (cf. EDS measurement results of sample λ_t2* in Table I). This issue of insufficient melting after the addition of a further element with a much lower melting point, has already been addressed in more detail in Section III–B–4. By changing the Ta input material, selecting the better synthesis route for ternary compositions as explained in Section III–B–2, and increasing the Ta content by 3 at. pct, the proportion of the second phase could be significantly decreased. The improved microstructure of the ternary Laves phase sample λ_t2 with 33 at. pct Ta can be seen in Figure 9(b).

The metallographic preparation of the TCP phases in the Ta–Fe(–Al) system, as described in the supplementary material (Section S1.2), proved to be suitable for the following investigations. Due to the long polishing times, the sample surface is processed without strong mechanical impact. In this way, breakouts due to the brittleness of the sample material, which lead to scratching of the surface, can be minimized. One examination method that requires careful sample preparation are EBSD measurements. In order to obtain high-quality maps, the recorded Kikuchi patterns must also have a high image quality, which allows a clear recognition and determination of the band structures. This is for the investigated TCP phases even more crucial due to the close crystallographic relationship between the two considered phases, with the Laves phase also contained as a building block in the µ-phase. Examples of captured patterns with indexed zone axes are shown in Figure 10. The pattern of the binary Laves phase is given in (a) and the one of the binary µ-phase in (b).

By combining EBSD with EDS on metallographically well-prepared surfaces, the microstructural analysis can be complemented by an analysis of crystal structure and orientation as well as local chemistry, and thus, enable a clear phase distinction. The distinction between crystallographically very different phases, e.g., the hexagonal Laves phase and any precipitated cubic B2 FeAl phase, is possible without difficulty using EBSD alone. In case of the two intermetallic phases under consideration, namely the Laves and µ-phase, a distinction based on the crystal structure is, however, much more difficult, due to the close relationship of these structures, with the Laves phase being a building block of the µ-phase. For this reason, the combination of EBSD with EDS is useful, since the two TCP phases differ significantly in their compositions. Figure 11 shows such a combined measurement of the binary µ-phase sample µ_b1 that has a targeted composition of 46 at. pct Ta. The measurement could confirm the presence of the Laves phase as a second phase. In image (a) the IPF of the Laves and µ-phase is given. For a better differentiation of these phases, the IPFs of only the µ-phase and only the Laves phase are shown in image (b) and (c), respectively. In addition, the color-coded phase map can be seen in image (d), with the µ-phase highlighted in red and the Laves phase in green. These EBSD maps agree with the EDS results, which, as seen in (e), show a higher Ta concentration in the matrix and a correspondingly higher Fe concentration, given in (f), in the second phase identified as Laves phase.

In addition, it is also of decisive importance that the phase files are as accurate as possible in order to distinguish the intermetallic phases. After selecting the reflectors listed in Table III together with a good pattern quality, it is possible to distinguish the Laves and µ-phase by EBSD alone.

For all samples listed in the upper part of Table I, optical microscopy images were taken, which can be seen in Figure 12. In images (a–c) of the first row, the binary Laves phases λ_b1, λ_b2, and λ_b3 are shown, which were weighed in at Ta fractions of 28, 33, and 38 at. pct, respectively. The samples of the binary µ-phases µ_b1, µ_b2, µ_b3, and µ_b4 can be seen in images (d–g) in the row below. They were weighed in at 46, 50, 54, and 58 at. pct Ta, respectively. Both the Laves and µ-phase samples are sorted by increasing Ta content. In the next two rows, the samples of the ternary Ta–Fe–Al system are given. Samples λ_t1, λ_t2, and λ_t3 of the ternary Laves phase are shown in images (h–j). They are sorted by increasing Al content of 22, 33.5, and 45 at. pct, respectively, while the Ta content remains constant at 33 at. pct. The ternary µ-phase samples are also sorted by increasing Al content at a constant Ta content of 54 at. pct. Images (k–m) show the samples µ_t1, µ_t2, and µ_t3 with targeted Al contents of 9, 16, and 23 at. pct, respectively. The undeclared percentage of the sample compositions is always Fe.

In addition, for a more detailed characterization of the samples, especially with respect to the presence of second phases, BSE images were taken, which can be seen in Figure 13. The arrangement of the samples of the two TCP phases of the Ta–Fe(–Al) system with different compositions is the same as in the optical micrographs in Figure 12. A detailed discussion of the observed inhomogeneity depending on the targeted composition is given in Section IV–A.

The actual compositions determined by EDS for all samples shown in Figure 13 and the target sample compositions for comparison can be found in Table I.

Despite the above-mentioned adjustments and improvements in the sample fabrication, some of the samples still show additional unwanted phases as inhomogeneities. The origin of these and their relation to the underlying phase equilibria and the experimental boundary conditions of arc melting as the core synthesis method will be discussed in the following.

Samples λ_b1, µ_b1, λ_t2 and µ_t3 present such cases at either end of the phase boundaries with respect to Ta content and a clear second phase visible in Figure 13(a), (d), (i) and (m), respectively. The targeted composition of sample λ_b1 (Figure 13(a)) is 28 at. pct Ta. According to the Ta–Fe phase diagram (Figure 2 on the left), the phase range of the binary Laves phase begins at 27.7 at. pct Ta and reaches this maximum value at a temperature of 1443 °C. At both higher and lower temperatures, the phase range becomes smaller. The presence of another Fe-rich phase is therefore likely and could indeed be confirmed by EDS analysis. Similarly, in the µ-phase sample µ_b1 with 46 at. pct Ta (closest to the stoichiometric composition of Ta₆Fe₇), the presence of a second phase is likely caused by the two-phase region (λ + µ) present at temperatures above 600 °C. As a result, the phase boundary of the µ-phase is reached only at low temperatures, where diffusion is already limited, impeding a full phase transformation. As expected, a combined EBSD and EDS measurement revealed that the second phase is the Laves phase (cf. Figure 11).

The effect of a slight deviation from the phase region, is clearly visible for the ternary Laves phase. Further reduction of the Ta content in sample λ_t2 from 33 to 30 at. pct (sample λ_t2* with small Ta core), results in an increased formation of the B2 FeAl phase with a dark contrast, as shown in Figure 13(i₁) and (i₂) (and cf. Figure 9). The ternary µ-phase sample µ_t3 (Figure 13(m)) also has a second phase present which is significantly brighter in the BSE image, indicating that this phase has a higher Ta content. Indeed, decreasing the Ta content from 54 to 52 at. pct Ta at constant Fe:Al ratio leads to a decrease in the proportion of the second phase (sample µ_t3*, Figure 14). Again, this composition is close to the phase boundary of the ternary µ-phase, as highlighted in the ternary phase diagram in Figure 2. In this case, the Ta₂Al σ-phase with dissolved Fe could be present. EDS confirms the higher Ta content indicated by the bright BSE contrast, however, the EBSD pattern quality of the precipitates is not sufficient to identify the crystal structure.

As expected, if temperature during solidification cannot be controlled carefully, it is difficult to obtain homogenous sample material close to the boundaries of the phase regions. In some systems, use of more elaborate casting and solidification equipment or employing long heat treatments at the highest possible temperature to encourage retrospective phase transformation may alleviate the problems encountered here. However, in some cases the formation of additional phases may not be avoided entirely even by these measures, particularly where the transformation occurs at temperatures much below the solidus temperature and transformation is sluggish and/or requires substantial compositional changes by diffusion. The Mo–Fe system^[30] is an example of such a system, where the R- and σ-phase form from the melt. The synthesis of samples with a large µ-phase fraction remains reasonably straightforward,^[31] with a transformation temperature of 1370 °C, not far from the solidus curve. But the synthesis of bulk Laves phase is in practice largely suppressed at the much lower transformation temperature of 977 °C and after passing two-phase regions involving α-Fe, the R- and µ-phase.^[30]

Despite the presence of second phases in samples with compositions close to the boundary of the phase field, a large fraction of the intended phase could be prepared for all cases. The compositions of these phases of interest determined by EDS (last column of Table I), deviate less than 1.5 at. pct from the intended compositions for the majority of the samples. A systematic investigation of the TCP phases of the binary Ta–Fe system with different compositions at increasing Ta content as well as the ternary Ta–Fe–Al system with varying Fe:Al ratio while keeping the Ta content as constant as possible has therefore proven to be achievable.

The synthesis and metallographic preparation methods presented in this work now make this kind of research possible for the Ta–Fe(–Al) system. To illustrate this, both nanoindentation and microcompression data are presented (Figure 15) in addition to the EBSD data already introduced above.

Figure 15 shows typical electron micrographs of a nanoindentation imprint in image (a), a deformed microcompression sample in (b), a nanoindentation load-displacement curve in (c) and an EBSD map on which a typical indentation array and locations for microcompression are indicated schematically in (d). In both micrographs, slip traces around the indent and on the micropillar, marked with small arrows, are clearly visible and the possible activated slip systems can be determined based on the local crystal orientation.^[32] The dominant formation of slip traces around the indentation on only one set of parallel planes confirms the prevalence of basal slip in the µ-phase, consistent with observations by Luo et al.^[19,33] on Nb–Co µ-phases with similar compositions in terms of the ratio of A and B atoms.

In addition to the characterization of mechanical properties and mechanisms, nanoindentation can also serve as a probe for the defect density at the surface after metallographic preparation. Where the stress field directly underneath the indenter samples a defect free volume, plastic deformation can only begin upon nucleation of dislocations. As dislocation nucleation requires higher stress than dislocation multiplication and motion, although the difference is much smaller in hard intermetallics than in soft metals, a so-called ‘pop-in’ occurs as the sample is loaded. Where this sudden displacement jump occurs at a stress level of the order of the theoretical shear strength of the material, it has been shown to be associated with dislocation nucleation in an originally defect free volume. Such a pop-in is indeed observed here in most indentation load-displacement curves, as indicated in Figure 15(c). The diamond indentation tip is of Berkovich shape overall (cf. Figure 15(a)) but is not infinitely sharp at its very end. At this point, it can normally be approximated by a sphere, giving rise to the elastic, Hertzian contact at the start of the indentation. A fit to the curve assuming a fully elastic Hertzian contact (Figure 15(c)) then allows an estimate of the maximum shear stress under the indenter, that is the point where dislocation nucleation is thought to take place and from where dislocations move very rapidly outwards to cause the displacement jump observed in the load-displacement signal. Here, the Youngs modulus of the sample is taken as 290 GPa, approximated from the indentation modulus, the Poisson ratio is assumed as 0.32, the tip radius as 275 nm and the surface as flat (sphere radius = infinity). These values provide a good fit to the data and at the load corresponding to the pop-in, the maximum shear stress in the volume under Hertzian contact conditions is 27 GPa.^[34] This is even slightly higher than the approximated ideal shear strength taken as \(\tau_{{{\text{theo}}}} = \frac{G}{2\pi } \), resulting in 17.5 GPa. The shear modulus \(G\) is thereby estimated to be \(G = \frac{E}{{2\left( {1 + \nu } \right)}} \), based on the Youngs modulus \(E\) and Poisson ratio \(\nu\), as used above. The deviation between the maximum shear stress and ideal shear strength values is very likely due to the approximate values used to fit the data as well as the assumptions of perfect Hertzian contact and isotropic behavior of the TCP phase both in terms of elasticity and availability of a suitable slip system with the maximum shear stress resolved on it. Overall, the pop-in is therefore taken as an indication of dislocation nucleation and thereby provides another source of evidence for the successful surface preparation in addition to the high-quality EBSD patterns acquired on the samples (cf. Figure 10).