1 Introduction

The γ-TiAl-based intermetallic is characterized by a high-specific strength retained through wide temperature range, what makes it a perspective substitute for nickel-based superalloys. The most desirable are castings with a nearly lamellar microstructure (NLγ), i.e. being characterized by colonies of mixture of γ-TiAl and α2-Ti3Al platelets separated by groups of the equiaxed grains of γ-TiAl phase.[1,2] In that form the material slightly gained in ductility, even as the creep got worse. Therefore, at first the TiAl-based alloys found only a few niche application including the low pressure part of gas turbines or valves in sport combustion engines.[2] A wider use of these alloys during the 1970s through 1990s were barred both by their high both brittleness (limiting even its machining) and susceptibility to oxidation at higher temperatures.

The excessive growth of scale on the γ-TiAl/α2-Ti3Al alloys exposed to oxidizing atmosphere at high temperature has been originally ascribed to an inability to form a tight alumina layer at their surface.[3] Then, it was found out, that a combination of small amount of alloying additions such as Nb, Cr, Mo, Si or W helps to attain not only improvement in ductility, but also restrict scale growth.[4] The last three elements as well as C are also rising the creep strength of these alloys.[5] The data accumulated at that time opened field for development of the first generation of these alloys known as 47–2–2, i.e. composed of Ti–(46.5–48)Al–2Cr–2Nb (at. pct).[6] A second generation of these alloys was connected with introduction of subsequent alloying additions, i.e. Ti–(45–48)Al–(1–3)X–(2–5)Y–(< 1)Z, where X = (Cr, Mn, V), Y = (Nb, Ta, W, Mo), Z = (B, C, Y), in which Cr, Mn and V contributed to improved RT ductility, B lowered grain size, while all the others improved oxidation resistance.[2,6] Finally, a third generation of TiAl-based alloys was developed accordingly to formula Ti–(42–48)Al–(0–10)X–(0–3)Y–(0–1)Z–(0–0.5 RE), where X = (Cr, Mn, Nb, and Ta), Y = (Mo, W, Hf, and Zr), Z = (C, B, and Si) and RE refers to rare earth elements.[6,7] The Ti–48Al–2Cr–2Nb alloy (US. Patent/1989) turned out to be the most successful one as from 2007 it has been used for production of GEnx engines powering the B787 planes. Others, like multicomponent G4 Ti–47Al–1Re–1W–0.2Si (at. pct)[6] or RNT650 Ti–48Al–2Nb–0.7Cr–0.3Si (at. pct) alloy planned for turbocharger spinner are still at a testing stage.[8]

Problems encountered during production of parts using even the latest generation of TiAl-based alloys with conventional mould casting were solved only partially, because the high brittleness of these materials has persisted. This situation stimulated interest in new fabrication methods. Experiments with centrifugal casting process showed that it is suitable for production of low-pressure turbine blades.[6] The other process worked out by Daido Steel (LEVICAST™) allowed to produce turbine wheel castings retaining their high strength up to 850 °C, i.e. temperature of the exhaust gases driving the turbochargers.[9] Application of the additive manufacturing (AM) realized with Electron Beam Powder Bed Fusion (EB-PBF) tested at Fraunhofer IFAM Dresden helped to strongly decrease the grain size of parts made of TiAl base alloys.[10] The new casting processes or additive fabrication methods can strongly modify their characteristic lamellar microstructure toward a duplex one, what affects diffusion processes in sub-surface areas and ask for experimental re-examination of the present view on their oxidation.

The early models of oxidation in air of the TiAl-based alloys, like that of Kekare et al.[11] were adopted from that developed for binary TiAl-treated in oxygen,[12] i.e. they were taking into account only reactions between Al, Ti and O. It was presumed that growth of the scale starts via a nucleation of intermixed Al2O3 and TiO2 crystallites. Faster growth rate of TiO2 as compared with Al2O3 would cause, that the alumina nucleus in-between TiO2 crystallites could grow only toward the core of the oxidized material. It would result in formation of TiO2 layer on top, a mixed Al2O3 and TiO2 layer below and Al2O3 at the bottom. It was also stated that up to 800 °C the scale growth is controlled by O inwarddiffusion through the scale, while below that temperature by outward-diffusion of Ti.[13] Eventually, it should lead to development of a TiO2 surface layer characterized by dense upper part and highly porous bottom. This view of oxidation of TiAl-based alloys was put forward using data acquired with methods of limited spatial resolution, like XRD, OM and SEM/EDS. Such approach directed investigation toward samples treated at higher range of anticipated operating temperatures foreseen for these alloys as well as for longer exposure times, as in both cases thicker scales were produced. All the above caused that especially the beginning of oxidation of the TiAl alloys in these models have relatively week experimental backing.

The successive oxidation models of TiAl (50:50) alloy relying on methods characterized by higher spatial resolution such as TEM observations (samples treated in air at 900 °C to 1000 °C) divided this process into: simultaneous nucleation of Al2O3 and TiO2 (stage I), preferential growth of Al2O3 capping the scale, what slows down upward diffusion of Al to much higher extent than that of Ti resulting in formation of TiO2 layer spreading over Al2O3 layer (stage II), thickening of the TiO2 layer accompanied by inward growth of porous Al2O3/TiO2 mixture accompanied by nucleation of metastable nitrides (stage III).[14] Similar investigation performed on Ti48Al8Cr2Ag alloy (oxidized in air at 900 °C) generally repeated above findings, even as problem of scale porosity was skipped.[15] Other methods like X-ray and Auger electron spectroscopy XPS/AES measurements of oxidized TiAl alloy (650 °C/air) indicated somewhat different sequence separated into following events: adsorption and absorption of oxygen species (stage I), nucleation and growth of ultra-thin alumina layer (stage II), depletion of sub-surface areas with Al promoting Ti diffusion toward the surface and enabling formation of TiO2 (stage III).[6] The differences between all these models regarding especially the early stages of oxidation might have arisen, among the others, due to much lower temperature during treatment of TiAl alloy studied using XPS/AES, as compared with that investigated with TEM.

A phase composition of scale formed on the TiAl based alloys was documented in numerous works, but due to differences in setting up of each experiment (including alloys chemical composition, oxidizing environments and treatment temperatures) comparison of these data is a problem. Even so, most of the XRD investigation of these scales were pointing toward presence of the R-TiO2 (rutile) and α-Al2O3 either as exclusive or at least dominating phases.[3,12,13,16] The similar studies using TEM selected area electron diffraction and TEM/EDS confirmed that the R-TiO2 occupies the surface, while a mixture of both these phases forms the rest of the scale.[11,14,15] Some of the works suggested also formation of minor amounts of metastable γ-Al2O3, κ(κ′)-Al2O3, Ti2O3 and X-phase,[13,16,17] as well as TiN and Ti2AlN[14,15,16] at the start of the oxidation process. It was suggested that the TiN and other nitrides (AlON, Ti2AlN) continuously forms at the scale/TiAl alloy interface only to be oxidized by inward-diffusing oxygen.[14,16] Aside of them, small amount of Z-Ti5Al3O2 phase was detected at the scale bottom.[18,19] The multi-phase fine-crystalline character of the scale formed on TiAl based alloys causes that within a wide temperature range (600 °C to 1200 °C) only presence of its two major components, i.e. TiO2 (rutile) at the surface and α-Al2O3 below it could be taken as confirmed. However, their distribution in relation to scale depth as well as occurrence of other phases, needs verification in future works.

Therefore, the aim of the present work was to compare the microstructure, chemical and phase composition of the oxide scale formed on samples both from the mould cast (MC) and fabricated using the Electron Beam Powder Bed Fusion (EB-PBF) of third generation of the TiAl-base alloy, i.e. RNT650 Ti-48Al-2Nb-0.7Cr-0.3Si (at. pct) being dedicated for additive manufacturing (AM) fabrication method. The samples were treated at 650 °C in air and in steam atmosphere, what allowed to study the relatively early stages of their oxidation process. The microstructure observations and accompanying analysis were performed with both standard TEM (Tecnai/FEG) and probe corrected Titan—class TEM equipped with large active area EDS detector.

2 Experimental Procedure

The RNT650 (Ti–48Al–2Nb–0.7Cr–0.3Si in at. pct) was prepared both by cold graphite mould casting (MC) and electron beam powder bed melting EB-PBF methods. The chemical composition of the powder, EB-PBF specimen and reference MC alloy were given in Table I.

Table I Chemical Compositiona,b of Powder, EB-PBF and MC Materials (in Wt Pct)
Full size table

The rods (ϕ-13 mm) produced using both of these methods were sliced with a diamond saw to 3 mm discs and next fine grinded (till their roughness/Ra was diminished down to 0.18 ± 0.03 µm and to 0.13 ± 0.03 µm for CM and EB-PBF, respectively). The dry-air oxidation process was conducted in an open furnace Carbolite tube furnace (type TZF 12/65/550 with alumina tube inside; hot zone ~ 10 cm) for 1000 hours at 650 °C. The test was divided into 4 cycles each lasting 250 hours. The samples were cleaned in iso propanol using ultrasonic bath for 15 minutes at 40 °C, placed on the alumina holder and inserted into the furnace hot zone (calibrated with thermocouple K). Next, both ends of the alumina tube was sealed with ceramic plugs with electrical connections for the thermocouples monitoring temperature inside the rig. The steam oxidation process was carried out in a close loop system. A di-ionised water was used to ensure a steady-state conditions (as well as diminish conductivity of water). The water was constantly (through-out the whole experiment) pumped and purged by oxygen-free nitrogen (OFN) gas to minimise oxygen level in it, i.e. keep constant pO2 derived only on oxygen from H2O. During the test a constant steam flow of 2.833 mL/min was maintained. The surface of the samples was prepared in the same way as for dry-air oxidation test. Next, they were kept at 200 °C for 2 hours inside the rig and purged by OFN gas to remove residual air and moisture. Then, the temperature in the rig was raised at a rate of 5 °C/min to reach the required temperature. The other details of alloys oxidation treatment including the scheme of the rig used for it were given elsewhere.[19] The samples produced from mould cast (MC) material were denoted with a letter A, while that by EB-PBF were denoted by B (Table II). The number following the letter refer to the oxidation processes, i.e. the air treated ones were ascribed with number four and the steam oxidized with number eight.

Table II Notation of Samples Prepared from MC and EB-PBF RNT650 TiAl Alloys
Full size table

The microstructure investigations were perform with Tecnai FEG (200 kV) and probe-corrected Themis (200 kV) transmission electron microscopes (TEM) equipped with energy-dispersive spectroscopy (EDS) system produced by ThermoFisher. The former microscope was used for microstructure observations in bright field (BF) and high-resolution (HREM) mode, as well as for filtering phase information with selected area electron diffractions. The latter microscope was used for acquisition information on local chemical composition by acquiring the Z (atomic number) sensitive images in scanning transmission mode (STEM) with the help of high angle annular dark field (HAADF) detector. These investigations were backed with maps built of 500 × 500 pixels (0.5 nm electron probe) and profiles (each step being a sum of the intensity of a line of 100 pixels/~ 200 nm) documenting a distribution of the alloying elements within the scales. The above information was filtered from the background using “net counts” calculated using Velox software, what allowed to differentiate the oxygen and nitrogen signals. The thin foils for TEM observations were prepared with the focused ion beam method (FIB) using Scios 2 Dual Beam (e−/Ga+) also produced by Thermo Fisher. They were cut in the form of rectangular lamellas of ~ 12 × ~ 6 μm and up to 100 nm thickness. The scales were protected from the beam firstly with an evaporated thick layer of carbon and secondly with a platinum bar on top.

3 Results

The low-magnification overview cross-section TEM images of the microstructure of near surface areas of the cast or EB-PBF RNT650 alloy oxidized for 1000 hours at 650 °C in air indicate that this treatment results in formation of similar type of scale, i.e. thin nano-crystalline layer of comparable surface morphology (Figures 1(a), (b)). The microstructure of the core of MC material is dominated by presence of sets of heavily dislocated parallel plates, while within the EB-PBF core the colonies of platelets are smaller and arrange in random way with respect to one another. The same materials exposed to superheated steam at a corresponding time–temperature frame developed not only three to five times thicker scale, but also of higher roughness due outgrowth of nano-whiskers at their surfaces (Figures 1(c), (d)). In both cases the sub-scale area of both alloys showed very strong microstructure coarsening. It was especially striking in case of the EB-PBF substrate, in which the small colonies of platelets were substituted with layers of equiaxed grains.

Fig. 1
figure 1

TEM/BF over-view images of microstructure of scales formed on: (a) MC/air oxidized (A4), (b) EB-PBF/air oxidized (B4), (c) MC/steam oxidized (A8) and (d) EB-PBF/steam oxidized (B8)

Full size image

The high-magnification images of the upper part of areas in Figure 1 clearly shows, that the scale formed on the MC alloy is also significantly thicker than the one on the EB-PBF (Figures 2(a), (b)). Additionally, the former carries a string of porosity close to its centre, while the latter is nearly porosity-free. The surface of both scales show presence of stubby rods or platelets sticking out from them at various angles contributing to their roughness. The scale/substrate interface is slightly wavering and of diffused character due to interconnected net of nano-voids. The same type of observation of steam-treated materials helped to established that scale formed on both of them is characterized by presence of both of a string of larger (20 to 50 nm) porosity near their centre probably denoting the pre-treatment position of the surface and a zone with high density of nano-voids (< 10 nm) at their bottom (Figures 2(c), (d)). Additionally, their surfaces are decorated by very thin whiskers, which in case of MC alloy are of the length comparable with the scale thickness.

Fig. 2
figure 2

TEM/BF images presenting microstructure details of scales formed on: (a) MC/air oxidized (A4), (b) EB-PBF/air oxidized (B4), (c) MC/steam oxidized (A8) and (d) EB-PBF/steam oxidized (B8). Rings denote areas used for acquiring selected area electron diffractions

Full size image

The selected area electron diffraction patterns (SA EDP) acquired from the upper, and middle part of the scale (from areas marked in Figure 2 as X and Y, respectively) consisted only of randomly distributed fine spots arranged in rings confirming their nano-crystalline character and absence of any crystallographic texture (Figures 3(a), (b)). Integrating their intensity along the rings radii indicated, that the upper part of scale contains approximately equal fractions of R-TiO2 rutile and α-Al2O3 (Figure 3(a)), while the middle part is built nearly exclusively of the former phase (Figure 3(b)). The SA EDP obtained from the lower part of the scale was dominated by the signal from the substrate represented by set of stronger spots belonging to [110] γ-TiAl zone axis (possibly superimposed with the TiN). Simultaneously, the numerous fine spots in the background gave a best match with α-Al2O3 phase.

Fig. 3
figure 3

SAED patterns acquired from: top (X), middle (Y) and bottom (Z) of scale (see markings in Fig. 2) accompanied by graphs presenting spots intensities integrated along radii of respective diffractions

Full size image

The HREM observations of the scale surface helped to establish that lattice spacing within the platelets and whiskers correspond to that of the R-TiO2 rutile phase (Figure 4). The long axis of the rods was found to be parallel to {110} planes.

Fig. 4
figure 4

HREM images of rutile TiO2 nano-rods with amorphous coating (a) and without it (b) grown on the surface of scale formed over MC alloy oxidized in steam atmosphere (A8)

Full size image

The first compact layer adjacent to the surface was built of comparable amounts of relatively equiaxed nano-crystallites (< 50 nm) of R-TiO2 rutile and α-Al2O3 (Figure 5). Both type of crystallites were highly defected, but rutile ones also showed presence of numerous planar defects. The next two layers were dominated either with R-TiO2 (middle one) or α-Al2O3 (bottom one) crystallites of similar characteristic as it concerns their size, shape and presence of defects. The scale/substrate transition zone turn out to be highly corrugated mostly due to preferential oxidation/dissolution of α2-Ti3Al platelets. The latter process caused formation of amorphous alumina pockets separating sticking out fingers of the γ-TiAl phase (Figure 6). These amorphous areas were characterized by presence of thinned out parts or even voids, as well as nucleus of crystallites of α-Al2O3 phase. Tips of γ-TiAl platelets adjacent to the interface were highly defective, as was documented by their frequently changing orientation caused probably by nucleation and growth of TiN-type nitrides within that area.

Fig. 5
figure 5

HREM images of R-TiO2 (a) and α-Al2O3 (b) crystallites grown in the top layer of scale formed over MC alloy oxidized in steam atmosphere (A8)

Full size image
Fig. 6
figure 6

HREM images of substrate/scale interface formed between EB-PBF alloy oxidized in air (B4) presenting it at [112]γ-TiAl (a) and [110]γ-TiAl orientations

Full size image

The scanning transmission (STEM) observations of the investigated scales performed in the dark field (HAADF) mode confirmed that they were highly porous (Figure 7). In their bottom part it took form of a spongy matter (documenting itself by numerous black dots over grey background), while in the centre they carried larger voids of the size comparable with the neighbouring crystallites (only the scale formed over EB-PBF treated in dry air was relatively dense in that area). Simultaneously, these observations revealed that the scales are of layered structure, i.e. aside of the very surface layer of a mixed contrast, the rest consist of lighter/darker/lighter layers. This stratification is especially well visible in the least porous scale formed on EB-PBF alloy oxidized in air (Figure 7(b)).

Fig. 7
figure 7

STEM/HAADF images presenting microstructure details of scales on: (a) MC/air oxidized (A4), (b) EB-PBF/air oxidized (B4), (c) MC/steam oxidized (A8) and (d) EB-PBF/steam oxidized (B8)

Full size image

The EDS/TEM maps presenting local chemical composition at the near surface area of samples cut from MC and EB-PBF alloys after high temperature treatment in dry air helped to establish that even as the scale is rich in oxygen, though below it a high nitrogen content prevails (Figure 8, O and N maps; column I and II). Comparing a concentration distribution of main alloying elements of the RNT650 alloy, within the zone of high oxygen content it is possible to differentiate a layer of mixed content of both Al and Ti, followed by a Ti rich layer and backed with a one dominated by Al (Figure 8, Al and Ti maps, column I and II). The above relation holds also in case of scales formed at steam atmosphere on both type of alloys, independently of the fact that due to their high roughness and porosity they layered structure is strongly distorted (Figure 8, O, N, Al and Ti maps; column III and IV). The distribution of minor alloying elements, i.e. Nb, Cr and Si turned out sensitive both to base material processing route as well as to oxidizing treatments. In the core of the MC material Cr and Si is located mostly in the fine precipitates at the inter-laths boundaries, while some tendency toward Si diffusion to the scale very surface was also noted (Figure 8, Cr and Si maps, column I). The EB-PBF alloy treated in dry air showed similar tendency as it concerns the Si distribution, even as the precipitates rich in this alloying elements were much coarser. However, the situation differs in case of Cr, which in this case out diffuses from the substrate and segregate into Ti and N rich layer at the scale bottom (Figure 8, Cr and Si maps, column II).

Fig. 8
figure 8

EDS/TEM maps with distribution of O, N, Ti, Al, Nb, Cr and Si within scale and adjacent substrate. Columns from left to right represent: (a) MC/air oxidized (A4), (b) EB-PBF/air oxidized (B4), (c) MC/steam oxidized (A8) and (d) EB-PBF/steam oxidized (B8) alloy

Full size image

The steam-treated samples of investigated alloys generally repeated the trends observed for dry air oxidized ones, but the accumulation of Si in the upper part of the respective scales already rich in titanium was much higher (Figure 8, Cr and Si maps, column III and IV). The latter treatment also causes precipitation of very fine Cr rich precipitates in the scale formed over both alloys. Niobium, forming the largest fraction of minority alloying elements show least tendency toward out diffusion from the substrate toward the scale, as only small enrichment in it could be found in EB-PBF material after any of the two applied treatments (Figure 8, Nb maps, column I to IV).

The concentration profiles of alloying elements (intensity at each step obtained after summing signal from 100 pixels strung along the coating surface of the maps in Figure 8, i.e. each time averaging information from ~ 200 nm) helped to remove much of the noise from the 2D images. They confirmed that the scale formed on the MC and air-treated materials consist of three separate oxide layers i.e. of a mixture of TiO2 and Al2O3 oxides, backed by TiO2 (practically free of Al2O3) and finally Al2O3 dominated layer, as well as Ti- and N-enriched zone (possibly TiN) within the substrate (Figure 9a). In case of EB-PBF material subjected to the same oxidizing treatment the presence of Al2O3 at the scale surface is even more evident, but the layers below the mixed one show presence—aside of oxygen—only prevailing amounts of Ti and Al, respectively (Figure 9(b)). The steam-treated alloys repeated this tendency only partially, as the last layer bearing nitrogen was additionally intermixed with oxygen which presence extends even deeper than that of the nitrogen (Figures 9(c) and (d)). The latter might be a result of preferential oxidation of α2-Ti3Al causing formation of amorphous alumina pockets extending toward the core. The built concentration profiles confirmed also presence of substantial amounts of silicon in the upper part of the scales formed over steam treated alloys, while the averaged content of the other minority alloying elements was found negligible. The least certain seems the chemical composition of the very surface as for the air treated alloys the profiles seem to indicate the preference of aluminium and oxygen rich phase, while for the steam treated ones rather to titanium and oxygen rich phase. Such a tendency reversal would be of high importance, but again, strong roughness and a fine scale of the analysed features causes that signal from that areas are very weak and noisy, i.e. least reliable of the above investigation.

Fig. 9
figure 9

EDS/TEM profiles with distribution of O, N, Ti, Al, Nb, Cr and Si across the scales: (a) MC/air oxidized (A4), (b) EB-PBF/air oxidized (B4), (c) MC/and steam oxidized (A8) and (d) EB-PBF/steam oxidized (B8) alloy. HAADF represent Z-sensitive intensity

Full size image

The high-magnification EDS/TEM mappings presenting distribution of Ti, Al and Si only within the scales developed after air and steam treatment of the MC alloy proved that in this oxygen permeated areas the titanium stubby rods or small platelets (in air treated materials) and extended very fine nano-rods/whiskers (in steam-treated ones) are usually covered (both at the side and top) with an Al2O3 and barely discernible amounts of a silicon oxide (Figures 10, 11). They also showed that the aluminium-oxygen veins pass from the surface of the scale through the titanium oxide rich layer, being especially well visible in Figure 10 (compare distribution of Ti and Al). Simultaneously, it should be noted that some of the nano-rods are totally void of alumina and silicon oxide cover (compare the O, Al and Si distribution in Figure 10).

Fig. 10
figure 10

EDS/TEM magnified maps presenting distribution of Ti, Al, Si, Nb, Cr, O and N within scale formed on air oxidized EB-PBF (4B) alloy. Broken yellow line delineates extent of oxidation affected zone (Color figure online)

Full size image
Fig. 11
figure 11

EDS/TEM magnified maps presenting distribution of Ti, Al, Si, Nb, Cr, O and N within scale formed on MC/steam oxidized (8A) alloy. On Ti mapping full arrows point to TiO2 rods with Al2O3(SiO2) coating, while empty ones toward same type of rods without it

Full size image

4 Discussion

The oxidation in dry air of the MC as well as the EB-PBF RNT alloy caused formation of scale characterized by presence of similar stubby nano-rods or nano-platelets covered with amorphous alumina (carrying traces of SiO2) at the surface, while the rest of it was built mostly of roughly equiaxed TiO2 rutile and α-Al2O3 nano-crystallites. The scale formed on the EB-PBF alloy differed from the MC one only by being practically porosity free and of slightly lower thickness. It was the Z (atomic number) sensitive STEM/HAADF observations backed with EDS profiles, which helped to show that in both alloy the scale was built as follows: (1st) the surface was covered with an ultra-thin layer of amorphous Al2O3 (overlaid with traces of amorphous SiO2) covering TiO2 stubby nano-rods/platelets, (2nd) below it resided a mixture of Al2O3 and TiO2 oxides, (3rd) backed with a layer filled with TiO2 crystallites broken with occasional Al2O3 channels, (4th) and finally by a dense layer of α-Al2O3 backed with highly porous amorphous Al2O3 penetrating the substrate along the α2-Ti3Al platelets (i.e. like in the scheme presented in Figure 12 for MC and EB-PBF/air oxidized alloys). The above findings disagree with the previous works, which all claimed that the scale top part is formed by a continuous TiO2 layer, while its middle and bottom part by a comparable fractions of Al2O3 and TiO2 oxides with some admixture of TiN or Ti2AlN.[7,11,12,13,14,15,21] The reason, why the surface alumina was up till now overlooked, could be at least threefold, i.e. its amorphous character, ultra-small thickness or lack of spatial/energy resolution of TEM/EDS systems of their times.

Fig. 12
figure 12

Scheme presenting main phases of layered structure grown over MC and EB-PBF alloys oxidized in dry air and steam at 650 °C for 1000 h

Full size image

The oxidation in steam atmosphere of the same alloys at the same time–temperature condition, may at first, i.e., using low-magnification TEM/BF microstructure observation seem very different from that resulting due to oxidation in the dry air. Namely the scales formed at such condition are much thicker (especially the one grown over the MC sample), their surfaces are overgrown with whiskers and their bottom boundary is of irregular character. However, the STEM/HAADF high-magnification observations of the rest of the scale and especially the TEM/EDS mappings of the distribution of the oxygen, nitrogen and the alloying elements shows close similarities, with the foremost being that they all are built of the same sequence of four layers (i.e. like in the scheme for MC and EB-PBF oxidized alloys presented in Figure 12). Additionally, most of the TiO2 whiskers at their surface are coated with ultra-thin Al2O3 layer bearing traces of SiO2 over it. Among the differences one should point that in such atmosphere the second layer built of a mixture of Al2O3 and TiO2 oxides carries also pockets filled with of amorphous SiO2. Furthermore, the steam treatment caused formation of extensive oxidation affected zone consisting of remnants of aluminium depleted γ-TiAl ingrown with TiN and strings of alumina particles growing toward the core along the α2-Ti3Al platelets. The observation that oxidizing atmosphere carrying admixture of the water vapour accelerates scale growth agrees with previous findings concerning oxidation of the TiAl or Ti47Al1CrSi alloy in such atmosphere at comparable temperature—time frame (700 °C to 750 °C/1000 hours).[22,23] Interestingly, the scale formed on the steam-treated EB-PBF alloy was markedly thinner, than the one formed at the same condition on the MC alloy, i.e. like in the case of the air oxidized samples. This situation was mostly caused by less-extensive penetration of the substrate by oxygen, which in MC alloy was accelerated by preferential reaction with sets of elongated α2-Ti3Al platelets (practically removed from EB-PBF alloy).\

The alloying of the TiAl with small additions of Cr (~ 0.7 at. pct) and Si (~ 0.3 at. pct) caused that in the MC alloy most of it formed numerous fine precipitates at the γ-TiAl/α2-Ti3Al lamellas interfaces, while in EB-PBF alloy the chromium-silicon precipitates were less numerous, coarser and located at the grain boundaries. The much larger Nb addition (~ 2 at. pct) was practically all dissolved within the matrix. During oxidation of these alloys the Si turned out to be the most active under both oxidizing environments, i.e. by diffusing to the scale surface. The switching from dry air to humid conditions increased Si mobility up to the point, that its presence in the form of silica pockets started to be easily discerned also in the scale sub-surface layer occupied by mixture of TiO2 and Al2O3 crystallites. The reason of this high activity of this element during oxidation of this alloy is probably that the adsorbed H2O particle dissociate to H+ and OH− particles.[28] Then, it is capable of faster than O inward diffusion into scale/substrate interface, at which hydrogen bind with titanium opening way for reaction of hydroxyl group with silicon forming silicon hydroxides.[27] The volatile Si(OH)x compound up-diffuses through the porous scale and encountering downward diffusing oxygen transforms to amorphous silicon filling-in all the porosity on the way as well as forming extremely thin deposit at the surface. On the other hand the Nb and Cr took more passive role as they just tend to accumulate below the scale. It could had been driven by upward diffusion from the core as well as being ejected from the TiAl/Ti3Al areas reacted with inward diffusing oxygen and transformed to alumina (with Ti out-diffusing toward the scale surface). In the steam oxidized alloys in the upper part of the scales presence of a small number of very fine Cr rich particles were also noted. The Nb accumulating close to the interface with the scale blocking exchange of material between the scale and the core has been among underlying assumptions concerning development of all three generations of TiAl-based alloys.[4,6,16] However, the high Si mobility during oxidation processes of γ-TiAl based alloys has been uncovered only within the present experiment. It should be noted, that pin-pointing the changes in Nb, Cr and Si distribution in the scale was possible only due to high sensitivity of the TEM/EDS used presently, as within the discussed area their actual amounts were either small (Nb) or extremely small (Cr, Si) for such measurements, i.e. only slightly above the detection limits, what could be best assessed at the attached distribution profiles in Figure 6.

Analysis of the O and N distribution using TEM/SEM EDS systems are based on measurements of the intensity of their X-ray K-α lines, which are partly overlapped with the soft radiation Ti-Lα line. That is why, taking into account their raw intensity would make such analysis a problem, but using net-intensity (cutting off background simultaneously removes side arms/tails of the neighbouring Gaussian peaks representing characteristic X-ray lines) allows to differentiate oxygen and nitrogen signal. The above approach helped to prove that within the investigated scales—accordingly to expectations—various oxides occupies the scale, while the nitrides forms a thin layer beneath it (Figure 5, 7, 8). The development of the latter at the scale bottom was previously documented on similarly treated TiAlNbMoB (TNM) alloy.[21] Actually, in one of the work the formation TiN crystallites at the surface during initial stages of air-oxidizing of TiAl alloy was presumed to hinder formation of a continuous protective alumina layer and later on to block diffusion between scale and the substrate.[12] The above idea seems questionable, as the present HREM observations as well as EDS mappings showed that even as there exist nitride layer, though it is a discontinuous one, i.e. penetrated by alumina channels expanding on account of Ti3Al platelets. The preferential reaction of oxygen with the Ti3Al over the γ-TiAl during high temperature exposure found in this experiment stays in opposition to the observation perform by Qu et al.,[24] which worked however with TiAl having different sets of alloying additions and oxidized their material at higher temperature. The ineffectiveness of nitride layer as a diffusion barrier in MC alloy in steam environment is also visibly by the formation of thick oxidizing affected zone developed below the scale. This zone is to large extent filled with chains of alumina particles penetrating nitrogen rich/discontinuous nitride layer proving that there is no problem for O to penetrate down there. However, in EB-PBF alloy the nitride phases form a thick continuous sub-scale layer which indeed could be effective in slowing down an ongoing scale development.

The phase analysis of the investigated scale performed with the selected area electron diffraction (ED) technique helped to establish that the upper part of scale was built with comparable amounts of R-TiO2 and α-Al2O3 phases, while its middle and bottom part were filled mostly the former and the latter phase, respectively. The scales comprised also a significant share of amorphous alumina and silica phases, which were identified at the very surface of the scale as well as within it, i.e. the silica intermixed with crystalline α-Al2O3 and R-TiO2 in the scale upper part and amorphous Al2O3 as admixture to crystalline α-Al2O3 one at its bottom. Lack of solubility of these amorphous phases documented itself at the oxide scale surface (Figures 7, 8) and therefore it should be accepted for the remaining part of the scale. The previous reports based on XRD data[11,13,16,20,24] as well as TEM investigations[12,14,15,18,24] also pointed toward R-TiO2 and α-Al2O3 as dominating components of the scales formed during oxidation of TiAl-based alloys, but only present one allow to point toward location of two amorphous phases, namely SiO2 and Al2O3.

The porosity plays an important role in scale growth as it opens easy diffusion paths to all involved elements as well as may intensify its spallation processes. The oxidation in dry air of MC alloy resulted in formation of strings of larger porosity of up to 50 nm size located between the layers dominated by rutile and alumina, i.e. in the middle of the scale. It would comply with the Taniguchi et al. model,[12] as long as the original surface was located between these two layers. The scale formed on the EB-PBF alloy after the same treatment was practically porosity free. It means that a refined grain size and therefore presence of more numerous grain boundaries in the substrate supplied enough of titanium not only to continue the scale growth but also to build up the voids with more rutile. The oxidation of the same alloys at presence of water vapour caused formation of string of voids of similar size at the corresponding areas (rutile layer easily distinguished by lighter shade than that of neighbouring alumina one as shown in Figure 7). In this case there was practically no difference between both type of CM or EB-PBF alloys, what was probably due presence of much thicker oxidation affected zone formed below the scale.

Other type of porosity of much finer size (< 10 nm) and high density was formed at the scale/substrate (or scale/oxidation affected zone in case of steam oxidized materials) boundary. Presence of these defects turns the material in this area in to the spongy type. Their formation is equally important as the larger porosity in the middle of the scale, because they also may link up and take over as an easy crack propagation path. The tendency to develop parallel fissures at exactly the same places within the scale formed over TiAl layers was documented also in other papers.[3,6,9,14,15,19,20] During the late oxidation stages presence of such defects may cause extensive spallation and therefore development of TiAl alloys should include requirement of minimization of the porosity within the scale forming over them.

The dry air oxidation process of both MC and EB-PBF alloys has to be controlled by the same parameters as in both cases it resulted in formation of the same sequence of layers dominated by R-TiO2 + α-Al2O3/R-TiO2 (+ minor α-Al2O3)/α-Al2O3 (+ minor am- Al2O3) (see Figure 12). Additionally, referring to the models on the air/oxygen supported scale growth over TiAl alloy,[11,12] one may point to the TiO2 (+ minor α-Al2O3)/α-Al2O3 boundary as the position of the original surface. At the start of oxidation process, the proximity of the substrate being an easily accessible source of titanium and faster kinetics of rutile growth decided that the scale surface was soon capped with R-TiO2 blocking expansion of α-Al2O3 crystallites as well as leaving behind some relatively large (~ 50 nm) closed porosity. This process is controlled by inward diffusion of the oxygen, which reacts with the α2-Ti3Al/γ-TiAl producing amorphous Al2O3 and freeing Ti, which is out diffusing into the surface. At later oxidation stages the influx of nitrogen to scale /substrate boundary reaches the level at which it could also react with the α2-Ti3Al/γ-TiAl forming TiN, what makes more aluminium available for up diffusion to the surface. The latter development might explain the presence of thin coating of amorphous alumina over the surface of R-TiO2 platelets and rods. The diffusivity of Si turned out on a par with that of Al and Ti as its presence was also noted even at the very surface of the scale. The other alloying additions, i.e. the Nb and Cr were found to accumulate below the scale forming a nitride layer, as documented in the other study.[25]

Switching the oxidizing environment to a steam resulted in evident thickening of the scale inner layers, as well as development of much thicker oxidation affected zone after the same time/temperature treatment. However, the most pronounced change was that of the surface morphology, i.e. the small TiO2 platelets/stubby rods covering it after dry air oxidation was substituted with whiskers after steam treatment. It is significant that while some of them were carrying an amorphous alumina coating with small admixture of silica and chromia, though quite a number of them was free of any coating. The former whiskers nucleated probably below alumina native oxide, which was next supportive for development of alumina thin coating over fast growing R-TiO2 whiskers. The above mechanism may operate only in case as the growth of the whiskers proceed at their anchoring place, like it was documented for reaction between ZrO2 powder mixed with carbon nano-tubs resulting in growth of ZrC whiskers.[26] The steam treatment of the EB-PBF alloys resulted also in higher mobility of the minor alloying additions as compared with the MC one, as evidenced by development of both fine chromium oxide and silicon oxide particles in the upper part of the scale.

Analysis of distribution of the alloying additions within the scale developed during dry air treatment of the MC and EB-PBF alloys indicated, that the silicon released from the oxidation affected area of the substrate is fast upward-diffusing into the scale surface. Simultaneously, the niobium and chromium accumulate in the substrate close to the boundary with the scale. Changing the oxidizing environment into that with water vapour accelerated all involved processes, increased the oxidation affected zone and freed even more silicon. It agrees with other findings that admission of water vapour is able to strongly accelerate oxidation processes both in TiAl and other alloys.[27,28] In the effect the outward-diffusing silicon not only covered the scale surface with ultra-thin silica coating but also filled most of the voids formed within the upper-most layer of the scale. Under this treatment also chromium out-diffused into the scale (occasionally even to its surface) forming sparsely distributed very fine particles. Only the niobium stays out of the scale, i.e. showing a tendency to accumulate right below it.

5 Conclusions

The present experiment allowed to compare the effect of replacing the mould casting (MC) with the electron beam melting (EB-PBF) of the TiAl–2Nb–0.7Cr–0.3Si (in at. pct) alloy on the microstructure of the scale formed in dry air and in the presence of water vapour. This alloy seems to be among a few in this field, which also could find some practical application being well adjusted for additive manufacture processing. In near future, TiAl alloys may find their place in high-temperature application on a par with NiAl superalloys, but right now the most important is the knowledge of their response to service at medium–high temperatures, like that (600 °C to 800 °C) in this experiment. The performed observations and measurements allowed to draw a following conclusions:

  1. 1.

    Oxidation of both MC and EB-PBF TiAl–2Nb–0.7Cr–0.3Si alloys in dry air as well as in the presence of water vapour leads to development of three-layers scale, within which the major fractions are as follows: R-TiO2 + α-Al2O3/R-TiO2/α-Al2O3. Adding the water vapour not only promoted the scale formation, but also changed the surface morphology of the rutile plates from rod or plate like into whiskery ones.

  2. 2.

    The more refine microstructure and resulting higher diffusivity of the EB-PBF than that in the MC alloy caused that the scale formed under dry air oxidation was more compact and free of larger porosity. Changing oxidizing atmosphere to steam one caused that scale over both MC and EB-PBF alloys were characterized by presence of surface whiskers and high porosity, as well as much more extensive oxidation affected zone.

  3. 3.

    The most mobile of the alloying additions in scale development turned out the silicon, which was able to diffuse up to the very surface of the scale formed even in dry air. The presence of water vapour not only increased mobility of silicon significantly rising its presence in the upper part of the scale, but also stimulated that of chromium forming throughout the scale dispersed fine particles enriched in this element.

  4. 4.

    The substrate adjoining to the scale formed using dry air showed evident enrichment in nitrogen, niobium and chromium. Oxidation of this alloy in steam atmosphere caused both in MC and EB-PBF alloy development of extensive oxidation affected zones comprising a number of phases.