An Advanced TiAl Alloy for High-Performance Racing Applications

Requirements and strict regulations for high-performance racing applications involve the use of new and innovative lightweight structural materials. Therefore, intermetallic γ-TiAl-based alloys enable new opportunities in the field due to their lower density compared to commonly used Ni-base superalloys. In this study, a β-solidifying TiAl alloy was examined toward its use as structural material for inlet and outlet valves. The nominal composition of the investigated TNM alloy is Ti–43.5Al–4Nb–1Mo–0.1B (in at%), which enables an excellent formability at elevated temperatures due to the presence of bcc β-phase. Different hot-extrusion tests on an industrial scale were conducted on the cast and hot isostatic pressed material to determine the ideal microstructure for the respective racing application. To simulate these operation conditions, hot tensile tests, as well as rotational bending tests, at room temperature were conducted. With a higher degree of deformation, an increasing strength and fatigue limit was obtained, as well as a significant increment of ductility. The fracture surfaces of the rotational bending test specimens were analyzed using scanning electron microscopy, revealing the relationship between crack initiation and microstructural constituents. The results of this study show that the mechanical performance of extruded TiAl material can be tailored via optimizing the degree of hot-extrusion.


Introduction
Engineering intermetallic γ-TiAl-based alloys are commonly used for high-performance applications, such as racing and aviation, due to their excellent specific mechanical and thermal properties [1][2][3][4]. In general, γ-TiAl alloys exhibit good oxidation resistance and microstructural stability during long-term thermal exposure up to service temperatures of 750 • C and, depending on their microstructure, a good creep behavior [5][6][7]. A major benefit is the low density of about 3900 kg/m 3 to 4200 kg/m 3 , being approximately half of the density of Ni-base superalloys [8]. Thus, γ-TiAl alloys are able to meet the aim of a higher efficiency of propulsion systems, which results in a reduction of fuel consumption and further, a decrease of CO 2 emissions [9]. Due to the reduced moving mass, the acceleration behavior of TiAl components is improved, yielding a significant advantage

Materials and Methods
The TNM material was processed by GfE Metalle und Materialien GmbH, Nuremberg, Germany, using double VAR followed by ISM [20]. For this purpose, the feedstock was melted in a water-cooled copper crucible where the melt is steered in a native skull of solidified TiAl until the whole ingot has liquefied. This allowed an excellent homogenization of the heavy alloying elements such as Nb and Mo [20,21]. Afterwards, the melt was centrifugal cast under vacuum in a rotating mold wheel. Subsequently, hot isostatic pressing (HIP) at 1200 • C and 200 MPa for 3 h allowed the closing of the remaining casting porosity, which also improved the chemical homogeneity. Further processing was carried out by MW Racing Märkisches Werk GmbH, Großbodungen, Germany, employing a single and multi-stage hot-extrusion process to obtain four different degrees of deformation ϕ. These ratios, including a ϕ of 0, 0.6, 1.4, and 1.9, are calculated according to Equation (1): where r 0 is the starting radius and r 1 represents the radius after deformation. This relation is displayed within the graphical abstract. In the following, the deformation ratios are given as positive numbers, although extrusion results in a reduction of the radius. The specific processing parameters cannot be disclosed, but [8,15,21] can be used as a guideline. To prevent oxidation, evaporation of Al, and heat loss of the rods during transfer and processing, a steel canning was used. A Mo foil acted as a diffusion barrier between the canning and the TNM ingot. After hot extrusion, a one-step heat treatment (HT) was conducted to achieve a stable microstructure close to thermodynamic equilibrium. Due to legal reasons, the HT parameters cannot be disclosed, however, the used temperature was above the service temperature. The specimens used in this study were electric discharge machined from the cast/HIPed, deformed, and HT material. The metallographic preparation, including grinding and electrolytic polishing, was performed as described in [22]. The microstructural characterization, as well as fracture surface analysis, was conducted by scanning electron microscopy (SEM) using an Evo50 from Zeiss, Oberkochen, Germany, at an acceleration voltage of 15 kV in secondary electron (SE) or backscattered electron (BSE) mode.
Rotating bending tests were conducted at room temperature (RT) with a Pun Z testing machine of the company Schenck, Darmstadt, Germany. To avoid uncontrolled heating of the sample during the experiment, the test frequency was limited to 100 Hz. The samples were polished to minimize residual stresses in the surface region and, therefore, avoid influences caused by the strong surface sensitivity of this testing method. The surface roughness of the tested specimens exhibited an R a value of 0.20 µm and a R z value of 1.62 µm.
Hardness tests according to Vickers HV10 were carried out using an M4C O25 G3M universal testing machine from Emco-Test, Kuchl, Austria. The values in this work correspond to the mean of at least 5 different measurements and the associated errors are given by the standard deviation. Hot and RT tensile tests were conducted at a starting strain rate of 2.4·10 −5 s −1 at GfE Fremat GmbH, Freiberg, Germany, at RT, 600 • C, 700 • C and 800 • C.

Microstructure
The influence of the microstructure of TiAl alloys on the mechanical properties has been extensively reviewed in the literature, e.g., see [8,18,[23][24][25]. It is evident that the grain size is one of the critical parameters for fatigue and tensile strength [26]. Using increasing deformation ratios, the grain size can be decreased significantly, which is mainly attributed to recrystallization leading to a refined and early texture-free microstructure [8,27,28].
In Figure 1a, the initial microstructure of the cast/HIPed and subsequently heat treated material is depicted. The TNM alloy shows coarse α 2 -Ti 3 Al/γ-TiAl colonies ((α 2 /γ) col ) surrounded by a seam of β o -TiAl phase and a high amount of globular γ, γ g , grains. These γ g grains exhibit a mean diameter (d) of 10.7 µm. Within the β o phase, small γ platelets, γ p , are visible, which form during the cooling sequence and the subsequent heat treatment [29]. Within the (α 2 /γ) col , secondary β o -phase precipitates during HIPing or subsequent HT to accomplish thermodynamic equilibrium [30,31]. The material condition shown in Figure 1a exhibits a mean hardness of 342 ± 3 HV10, representing a reasonable value for a cast/HIPed and HT microstructure [8].
In Figure 1b, the microstructure of the hot-extruded (ϕ = 0.6) material is shown, exhibiting a radial deformation degree of 0.6. The β o -phase appears elongated in loading direction, which indicates a good deformability of the disordered body-centered cubic (bcc) β-phase at deformation temperatures [15,21]. A major part of the microstructure consists of fine equiaxed globular α 2 -and γ-grains. However, a small number of (α 2 /γ) col remains and their lamellae are orientated in the extrusion direction [32]. Due to the applied deformation energy during the hot-extrusion, the disordered hexagonal α-phase and γ-laths of the colonies are subjected to a strong driving force for recrystallization [32]. Colonies arranged lengthwise or nearly lengthwise to the axis of the ingot are not deformed as much as their twisted counterparts [33,34]. Hence, the driving force for recrystallization is lower, reasoning the occurrence of single (α 2 /γ) col , which do not undergo recrystallization [33]. Similarly to this phenomenon, coarse γ-grains elongated in the deformation direction can be noticed. Within these grains, deformation twins are visible, which are marked by white arrows in Figure 1b. The mean size of these γ g grains is 6.9 µm and the hardness of this type of microstructure is 356 ± 3 HV10. Compared to the cast/HIPed and HT state, a higher hardness due to the smaller grain size is achieved [35,36]. In Figure 1c, the microstructure of the double-extruded sample with φ = 1.4 is shown. The increased degree of radial deformation leads to a finer and even more homogenous microstructure. Except for a few coarse γ (γcoarse) grains, the whole material is recrystallized [32]. Furthermore, elongated βo-grains are labeled as βo, long, which indicate the deformation direction. The average grain size of the globular γ-grains was reduced to 1.8 µm. In summary, persistent γcoarse-grains surrounded by a fine matrix of α2-and γ-grains, as well as elongated βo-grains, can be found. Their occurrence is explained by the growth and agglomeration of globular γ-grains during the HIP treatment [8,21,37]. During hot-deformation, these γ-grains are elongated and partly recrystallized [32,38]. Boron additions to the alloying system result in the formation of Ti-borides during solidification [39,40]. In the case of the TNM alloy, only monoborides have been detected. These borides act as heterogeneous nucleation sites during solidification and the solid-state transformation from β-to α-phase [41]. Hence, boron as a minor alloying element effectively reduces the grain size of the cast microstructure [39,40]. During hot deformation, these borides are fractured, as shown in Figure 1c, and act as nucleation sites for recrystallization and, thereby, facilitate the microstructural refinement [42,43].
In Figure 1d, the microstructure of the strongest deformed specimen with φ = 1.9 is depicted, which is very similar to the material with φ = 1.4, as shown in Figure 1c. These two conditions furthermore exhibit a comparable hardness, which amounts to 393 ± 2 and 386 ± 2 HV10 in case of φ = 1.4 and φ = 1.9, respectively. A mean grain size of the globular γ-grains of approximately 1.8 µm is observed Figure 1. SEM micrographs recorded in backscattered electron (BSE) mode of the microstructures of the different specimens investigated within this work. All specimens were heat treated to enable static recrystallization as well as to provide a thermally stable microstructure. (a) Cast/hot isostatic pressed (HIPed) and heat treated (HT) microstructure containing (α 2 /γ) col , γ g , and surrounding β o -phase; (b) the ϕ = 0.6 deformed specimen with partly recrystallized (α 2 /γ) col , as well as fine γ and α 2 grains. The arrows indicate twins within the γ g grains; (c) illustrates the microstructure of the ϕ = 1.4; and (d) the ϕ = 1.9 deformed samples. The latter two exhibit a similar appearance and an almost completely recrystallized microstructure. Solely isolated γ coarse -grains and the elongated β o -phase constitute deviations from the homogenous microstructure. The extrusion direction is indicated by white arrows in the right upper corner of the SEM images.
In Figure 1c, the microstructure of the double-extruded sample with ϕ = 1.4 is shown. The increased degree of radial deformation leads to a finer and even more homogenous microstructure. Except for a few coarse γ (γ coarse ) grains, the whole material is recrystallized [32]. Furthermore, elongated β o -grains are labeled as β o, long , which indicate the deformation direction. The average grain size of the globular γ-grains was reduced to 1.8 µm. In summary, persistent γ coarse -grains surrounded by a fine matrix of α 2 -and γ-grains, as well as elongated β o -grains, can be found. Their occurrence is explained by the growth and agglomeration of globular γ-grains during the HIP treatment [8,21,37]. During hot-deformation, these γ-grains are elongated and partly recrystallized [32,38]. Boron additions to the alloying system result in the formation of Ti-borides during solidification [39,40]. In the case of the TNM alloy, only monoborides have been detected. These borides act as heterogeneous nucleation sites during solidification and the solid-state transformation from βto α-phase [41]. Hence, boron as a minor alloying element effectively reduces the grain size of the cast microstructure [39,40]. During hot deformation, these borides are fractured, as shown in Figure 1c, and act as nucleation sites for recrystallization and, thereby, facilitate the microstructural refinement [42,43].
In Figure 1d, the microstructure of the strongest deformed specimen with ϕ = 1.9 is depicted, which is very similar to the material with ϕ = 1.4, as shown in Figure 1c. These two conditions furthermore exhibit a comparable hardness, which amounts to 393 ± 2 and 386 ± 2 HV10 in case of ϕ = 1.4 and ϕ = 1.9, respectively. A mean grain size of the globular γ-grains of approximately 1.8 µm is observed at ϕ = 1.9, indicating that using double extrusion no further refinement occurred by increasing ϕ from 1.4 to 1.9.

Room Temperature and Hot Tensile Tests
Tensile testing was conducted at RT, 600 • C, 700 • C, and 800 • C. The obtained results are summarized in Table 1. Figure 2a shows the yield strength (R p0.2 ) in dependence of temperature and deformation ratio of the differently hot-extruded conditions. From RT to 800 • C, the deformed specimens exhibit a clear reduction of the R p0.2 and ultimate tensile strength (UTS). The UTS of the undeformed condition behaves contrary to the measured R p0.2 , as shown in Table 1, and exhibits a moderate increase followed by a decrease at 700 • C and 800 • C. This effect is attributed to the higher elongation and, therefore, a higher UTS is obtained. The following decrease with increasing temperature of the UTS as well as R p0.2 in all conditions is ascribed to the higher mobility of thermally activated dislocations and the possibility of grain boundary sliding [44][45][46][47][48]. Table 1. Summary of the mechanical properties obtained for the TNM alloy at room temperature (RT), 600 • C, 700 • C, and 800 • C with varying ϕ, including the standard deviation of the respective data point. UTS: ultimate tensile strength. A clear tendency towards higher yield strength values with increasing deformation degree was observed, as shown in Figure 2a. However, at 800 • C the values of all four material conditions nearly coincide. Comparing the undeformed specimens (ϕ = 0) with the lowest extruded condition (ϕ = 0.6), an enhancement of around 180 MPa in R p0.2 at RT is evident. This gain of strength is mainly attributed to the smaller grain size within the microstructure, as shown in Figure 1b, due to continuous dynamic and static recrystallization processes during hot-extrusion and the following HT. In the course of this process, the grain size of the (α 2 /γ) col and γ g is effectively reduced and the strength increases through the well-known Hall-Petch mechanism [35,36,49]. Especially below the brittle-to-ductile transition temperature (between 600 • C and 700 • C), the grain size dominates the strength properties [50]. Materials 2020, 13, x FOR PEER REVIEW 6 of 14  Table 1. The error bars indicate the standard deviation of the respective data point, as shown in Table 1.
As already mentioned, a further increase in the degree of deformation leads to a mean grain size of 1.8 µm (φ = 1.4 and φ = 1.9), as shown in Figure 1c,d. This refinement results in a yield strength of more than 1100 MPa at RT, which is dominated by the already discussed mechanisms. The specimens with φ = 1.4 exhibit better mechanical behavior in terms of Rp0.2, UTS, and fracture strain (εF) compared to the φ = 1.9 deformed specimens.
The lower average tensile properties of the φ = 1.9 specimens may be reasoned by the increased diameter of the starting ingot, which leads to slower cooling conditions during solidification. The starting ingot diameter must be increased to achieve the desired degree of deformation while keeping the final dimension almost constant. Hence, the occurrence of segregations, as well as an unsuitable microstructure, is assumed to become more likely within this condition [51]. This may be the reason for the lower average tensile values of the φ = 1.9 compared to the φ = 1.4 specimens. Figure 2b shows the exemplary stress-strain (σ-ε) curves at temperatures from RT to 800 °C of the specimens with φ = 1.4. At RT, a maximum stress of 1221 MPa, as well as an εF value of 1.8%, could be obtained [7,9]. These high values are mainly attributed to the homogenous and fine-grained microstructure, as shown in Figure 1c. The cast/HIPed state exhibits at a temperature of 600 °C slightly higher εF values compared to RT tests [52]. This can be attributed to the enhanced possibility of dislocation motion and twin formation within the γ-phase leading to mitigation effects of local internal concentrations of stresses [15,45,48,52,53]. Hence, higher mechanical properties at RT are obtained when compared to the φ = 0 specimen. By increasing the temperature within the tensile tests from 600 °C to 700 °C, the brittle-to-ductile temperature is exceeded and a significant improvement of the fracture strain is observed. At a temperature of 800 °C, a maximum εF of about 87% occurred, which corresponds to a significant plastic deformability prior to failure [8,21]. However, the increasing fracture strain is accompanied by a decreasing Rp0.2 and, thus, restricts the service temperature of the alloy.

Rotating Bending Tests
To evaluate the effect of the deformation ratio on fatigue properties, rotating bending tests were conducted, and the results are also summarized in Table 1. With increasing φ, the fatigue limit, defined at 10 7 cycles in this study, could be significantly enhanced until the minimum grain size at φ = 1.4 was reached. At the highest degree of deformation, the same fatigue strength (SF) of 950 MPa was obtained, because of the similar grain size. The stress (S) applied to the sample, plotted against the number of cycles (N) to fatigue failure, commonly called the S-N curve, is shown in Figure 3. The  Table 1. The error bars indicate the standard deviation of the respective data point, as shown in Table 1.
As already mentioned, a further increase in the degree of deformation leads to a mean grain size of 1.8 µm (ϕ = 1.4 and ϕ = 1.9), as shown in Figure 1c,d. This refinement results in a yield strength of more than 1100 MPa at RT, which is dominated by the already discussed mechanisms. The specimens with ϕ = 1.4 exhibit better mechanical behavior in terms of R p0.2 , UTS, and fracture strain (ε F ) compared to the ϕ = 1.9 deformed specimens.
The lower average tensile properties of the ϕ = 1.9 specimens may be reasoned by the increased diameter of the starting ingot, which leads to slower cooling conditions during solidification. The starting ingot diameter must be increased to achieve the desired degree of deformation while keeping the final dimension almost constant. Hence, the occurrence of segregations, as well as an unsuitable microstructure, is assumed to become more likely within this condition [51]. This may be the reason for the lower average tensile values of the ϕ = 1.9 compared to the ϕ = 1.4 specimens. Figure 2b shows the exemplary stress-strain (σ-ε) curves at temperatures from RT to 800 • C of the specimens with ϕ = 1.4. At RT, a maximum stress of 1221 MPa, as well as an ε F value of 1.8%, could be obtained [7,9]. These high values are mainly attributed to the homogenous and fine-grained microstructure, as shown in Figure 1c. The cast/HIPed state exhibits at a temperature of 600 • C slightly higher ε F values compared to RT tests [52]. This can be attributed to the enhanced possibility of dislocation motion and twin formation within the γ-phase leading to mitigation effects of local internal concentrations of stresses [15,45,48,52,53]. Hence, higher mechanical properties at RT are obtained when compared to the ϕ = 0 specimen. By increasing the temperature within the tensile tests from 600 • C to 700 • C, the brittle-to-ductile temperature is exceeded and a significant improvement of the fracture strain is observed. At a temperature of 800 • C, a maximum ε F of about 87% occurred, which corresponds to a significant plastic deformability prior to failure [8,21]. However, the increasing fracture strain is accompanied by a decreasing R p0.2 and, thus, restricts the service temperature of the alloy.

Rotating Bending Tests
To evaluate the effect of the deformation ratio on fatigue properties, rotating bending tests were conducted, and the results are also summarized in Table 1. With increasing ϕ, the fatigue limit, defined at 10 7 cycles in this study, could be significantly enhanced until the minimum grain size at ϕ = 1.4 was reached. At the highest degree of deformation, the same fatigue strength (S F ) of 950 MPa was obtained, because of the similar grain size. The stress (S) applied to the sample, plotted against the number of cycles (N) to fatigue failure, commonly called the S-N curve, is shown in Figure 3. The slopes of the interpolated linear S-N curves are flat, which is in good accordance with [54]. Within each experimental series of the differently deformed conditions, a relatively low scattering is observed. The similar behavior of the two highest deformed conditions is attributed to their almost identical microstructure, as shown in Figure 1c,d. At RT, the γ-phase is prone to micro-plastic deformation through dislocation movement [47,[55][56][57]. The generated strain accumulation leads to crack initiation at the boundaries of the relatively soft and globular γ-grains [58,59]. Consequently, the coalescence of these early defects leads to the formation of a pre-crack at the grain boundaries, which propagates further by stable crack growth [56,58,60].
Materials 2020, 13, x FOR PEER REVIEW 7 of 14 slopes of the interpolated linear S-N curves are flat, which is in good accordance with [54]. Within each experimental series of the differently deformed conditions, a relatively low scattering is observed. The similar behavior of the two highest deformed conditions is attributed to their almost identical microstructure, as shown in Figure 1c,d. At RT, the γ-phase is prone to micro-plastic deformation through dislocation movement [47,[55][56][57]. The generated strain accumulation leads to crack initiation at the boundaries of the relatively soft and globular γ-grains [58,59]. Consequently, the coalescence of these early defects leads to the formation of a pre-crack at the grain boundaries, which propagates further by stable crack growth [56,58,60]. Taking into account the results of the tensile tests at RT and the determined fatigue limits, a proportionality factor ranging from 0.78 to 0.85, UTS, and SF can be determined. This is in good accordance with the work of Sastry and Lipsitt [61] for fatigue experiments up to 10 6 cycles, where the authors report a ratio of 0.8 at RT.

Fracture Analysis
The fractured surfaces of the specimens from rotating bending tests were analyzed with SEM using SE mode. Two different cases can be distinguished: The undeformed samples (cast/HIPed and HT) show smooth initial cracks within the occurring fracture mirror, as shown in Figure 4a. These cracks typically possess a dimension comparable to the size of the present (α2/γ)col and exhibit steps, which are exemplary marked by an arrow within Figure  4b. Thus, unfavorable orientated colonies were cleaved by interlamellar fracture [56,62]. Crack initiation was observed not to occur at surface defects, but at colonies located just below the sample surface. This is caused by the presence of compressive stress close to the specimens' surface, which stems from the mechanical preparation process. The cleavage of the lamellae mainly takes place at the α2/γ-interfaces [63]. Steps, as visible in Figure 4b, occur when the crack path continues to a neighboring α2/γ-interface [64,65]. Thus, it is assumed that these steps connect pre-existent cracks in front of the crack tip. During stable crack growth, the surrounding colonies largely fail in a translamellar manner at higher crack growth rates in case of a random orientation distribution [64]. As described by Leitner et al. [66], the occurrence of a brittle βo-film surrounding the colonies may accelerate the failure of the specimen. However, the presence of globular γ and βo grains along (α2/γ)col is considered to enhance the fracture behavior as discussed in [62]. Taking into account the results of the tensile tests at RT and the determined fatigue limits, a proportionality factor ranging from 0.78 to 0.85, UTS, and S F can be determined. This is in good accordance with the work of Sastry and Lipsitt [61] for fatigue experiments up to 10 6 cycles, where the authors report a ratio of 0.8 at RT.

Fracture Analysis
The fractured surfaces of the specimens from rotating bending tests were analyzed with SEM using SE mode. Two different cases can be distinguished: The undeformed samples (cast/HIPed and HT) show smooth initial cracks within the occurring fracture mirror, as shown in Figure 4a. These cracks typically possess a dimension comparable to the size of the present (α 2 /γ) col and exhibit steps, which are exemplary marked by an arrow within Figure 4b. Thus, unfavorable orientated colonies were cleaved by interlamellar fracture [56,62]. Crack initiation was observed not to occur at surface defects, but at colonies located just below the sample surface. This is caused by the presence of compressive stress close to the specimens' surface, which stems from the mechanical preparation process. The cleavage of the lamellae mainly takes place at the α 2 /γ-interfaces [63]. Steps, as visible in Figure 4b, occur when the crack path continues to a neighboring α 2 /γ-interface [64,65]. Thus, it is assumed that these steps connect pre-existent cracks in front of the crack tip. During stable crack growth, the surrounding colonies largely fail in a translamellar manner at higher crack growth rates in case of a random orientation distribution [64]. As described by Leitner et al. [66], the occurrence of a brittle β o -film surrounding the colonies may accelerate the failure of the specimen. However, the presence of globular γ and β o grains along (α 2 /γ) col is considered to enhance the fracture behavior as discussed in [62].
Materials 2020, 13, x FOR PEER REVIEW 8 of 14 Figure 4. Fracture mirrors including the critical crack length ac as well as crack initiation sites of selected, fractured rotating bending test specimens taken by SEM in secondary electron (SE) mode. A representative fracture surface of an undeformed specimen (a,b) appears to be rougher in contrast to the deformed specimen (c,d). It shows a cleaved (α2/γ)col as an initial crack, including a translamellar step, which is marked by an arrow in (b). In (c,d), a hot-extruded sample with a degree of deformation of φ = 1.9 is depicted. Here, coarse γ-grains with different angles to the image plane act as initial cracks and are highlighted in (d) by white arrows.
Crack propagation, including a step within an α2/γ-colony, is schematically depicted in Figure 5a. A red triangle symbolizes the initial crack below the surface at a γ-grain interface. Further crack growth through the coarse (α2/γ)col is illustrated by the red line at the interfaces within the colony. Hence, the crack reaches the surface at the prevailing tensile stress, which the rotating bending specimen is subjected to.
The fracture appearance of the hot-extruded samples, as shown in Figure 4c,d, exhibit smaller initial cracks. These cracks are surrounded by a typical fracture mirror and show almost the same size, as shown in Figure 4c, as the already mentioned coarse globular γ-phase within Figure 1b,d. Here, the relatively soft γ-phase is assumed to deform, which enables the formation of micro-cracks within the adjacent brittle βo-phase, as schematically shown in Figure 5b. The initial crack is symbolized by a red triangle and the growing crack is depicted by a red line. After crack initiation, the crack grows almost unconstrained through the near γcoarse-grains, since its propagation is hardly hindered by crack deflection.
Once the crack reaches a critical length of ac, the stress intensity at the crack tip exceeds the material's strength and the remaining cross-section fails by unstable crack propagation. Therefore, an assessment based on linear-elastic fracture mechanics according to Equation (2) can be conducted to . Fracture mirrors including the critical crack length a c as well as crack initiation sites of selected, fractured rotating bending test specimens taken by SEM in secondary electron (SE) mode. A representative fracture surface of an undeformed specimen (a,b) appears to be rougher in contrast to the deformed specimen (c,d). It shows a cleaved (α 2 /γ) col as an initial crack, including a translamellar step, which is marked by an arrow in (b). In (c,d), a hot-extruded sample with a degree of deformation of ϕ = 1.9 is depicted. Here, coarse γ-grains with different angles to the image plane act as initial cracks and are highlighted in (d) by white arrows.
Crack propagation, including a step within an α 2 /γ-colony, is schematically depicted in Figure 5a. A red triangle symbolizes the initial crack below the surface at a γ-grain interface. Further crack growth through the coarse (α 2 /γ) col is illustrated by the red line at the interfaces within the colony. Hence, the crack reaches the surface at the prevailing tensile stress, which the rotating bending specimen is subjected to.
The fracture appearance of the hot-extruded samples, as shown in Figure 4c,d, exhibit smaller initial cracks. These cracks are surrounded by a typical fracture mirror and show almost the same size, as shown in Figure 4c, as the already mentioned coarse globular γ-phase within Figure 1b,d. Here, the relatively soft γ-phase is assumed to deform, which enables the formation of micro-cracks within the adjacent brittle β o -phase, as schematically shown in Figure 5b. The initial crack is symbolized by a red triangle and the growing crack is depicted by a red line. After crack initiation, the crack grows almost unconstrained through the near γ coarse -grains, since its propagation is hardly hindered by crack deflection. coarser globular γ-grains and persistent (α2/γ)col present in this condition, as shown in Figure 1b. Here, the prevailing bimodal grain size distribution must be taken into account, including a fine-grained matrix as well as coarser microstructural constituents. Consequently, the material behaves like the φ = 1.4 and 1.9 hot-extruded materials in terms of fracture toughness. Therefore, crack nucleation and propagation are assumed to take place within the fine-grained matrix in the φ = 0.6 samples, which is comparable to case (ii), as previously discussed.

Conclusions
Within this study, the dependence of mechanical properties on microstructure in an engineering intermetallic γ-TiAl-based alloy was shown, against the background of obtaining optimized properties regarding high-performance racing applications, like inlet and outlet combustion valves. For this purpose, HIPed TNM ingots with a nominal composition of Ti-43.5Al-4Nb-1Mo-0.1B (in Figure 5. Schematic illustration of the two different predominant crack initiation and propagation modes. (a) Crack initiation (red triangle) at the interface between an α 2 -and γ-grain as well as the further crack growth within a cast/HIPed and heat treated microstructure. Subsequently, the crack splits a coarse (α 2 /γ) col , including inter-and translamellar propagation. (b) Crack growth within the hot-extruded material with a fine-grained microstructure. At RT, the crack initiation takes place below the specimen surface within the brittle elongated β o phase and propagates preferentially through the coarse γ grains.
Once the crack reaches a critical length of a c , the stress intensity at the crack tip exceeds the material's strength and the remaining cross-section fails by unstable crack propagation. Therefore, an assessment based on linear-elastic fracture mechanics according to Equation (2) can be conducted to allow an estimation of the critical conditional fracture toughness [67,68]: Here, the measured dimensions of the specimens' fracture mirrors from rotating bending tests were used to calculate the conditional critical fracture toughness K IC [67,68]. The geometrical constant Y is defined as 2/π and represents a penny-shaped inner crack under tensile loading. SEM investigations on the fractured specimens allowed the obtaining of the dimensions of the fracture mirrors, which correspond to a c , as shown in Figure 4a,c. Due to the experimental setup of rotational bending tests, the bending stress on the surfaces of the specimens is constant over the specimen's length. Therefore, the specimens are assumed to fail at the maximal bending stress σ. It should be pointed out that the determined K IC values are intended to be compared within this experimental study, but do not represent standardized K IC values [68]. Therefore, only conditional K IC values are discussed below. The mean K IC values, including a standard deviation from at least three specimens, are listed in Table 1. The results indicate a higher resistance to crack propagation of the undeformed microstructure, exhibiting a fracture toughness of about 9.2 ± 0.7 MPa √ m, in comparison to the extruded fine-grained microstructures which show lower values [23,65,66,69,70]. Within the deformed samples, the calculated fracture toughness is significantly lower (6.5 ± 0.5 MPa  m for similar TiAl alloys and material conditions. The decreasing fracture toughness of the hot-deformed specimens can be explained by the prevailing crack propagation mechanism. In the case of the fine-grained material conditions, the crack propagates easier through smaller globular grains than through a coarse-grained microstructure, where crack deflection and other toughening mechanisms prevail [70,72,73]. In this study, the specimens exhibiting a ϕ = 0.6 show a similar fracture toughness as the higher deformed samples despite the coarser globular γ-grains and persistent (α 2 /γ) col present in this condition, as shown in Figure 1b. Here, the prevailing bimodal grain size distribution must be taken into account, including a fine-grained matrix as well as coarser microstructural constituents. Consequently, the material behaves like the ϕ = 1.4 and 1.9 hot-extruded materials in terms of fracture toughness. Therefore, crack nucleation and propagation are assumed to take place within the fine-grained matrix in the ϕ = 0.6 samples, which is comparable to case (ii), as previously discussed.

Conclusions
Within this study, the dependence of mechanical properties on microstructure in an engineering intermetallic γ-TiAl-based alloy was shown, against the background of obtaining optimized properties regarding high-performance racing applications, like inlet and outlet combustion valves. For this purpose, HIPed TNM ingots with a nominal composition of Ti-43.5Al-4Nb-1Mo-0.1B (in at%) were hot-extruded to different degrees of deformation and subsequently heat treated. The adjusted microstructures were assessed using tensile tests, rotating bending tests, and microstructural analysis. The major findings can be summarized as follows: • With an increasing deformation ratio, a finer grain size can be obtained. This fact lasts until ϕ = 1.4, where a saturation level is reached with a grain size of about 1.8 µm. The characteristic α 2 /γ colonies disappear continuously and are replaced by globular α 2 -and γ-grains through dynamic and static recrystallization, surrounded by elongated β o -grains.

•
Employing RT tensile tests, a fracture strain of about 1.3% and a yield strength of 1121 MPa are determined for ϕ = 1.4. This condition exhibits a fine-grained and homogenous microstructure.
At 800 • C, the yield strength amounts to 404 MPa and a fracture strain of more than 87% can be obtained. Additionally, a decreasing grain size is observed, which results in a significantly lower brittle-to-ductile transition temperature.

•
A clear dependence between the fatigue limit and the degree of deformation is found. Caused by the smaller grain size as a result of the extrusion process, the material's fatigue limit at RT increases from 575 to 950 MPa. Here, large microstructural constituents are identified as initial defects within the microstructure. However, a high deformation degree reduces the size of these constituents, which is an effective way to increase the endurance limit. Comparing the ultimate tensile strength and the fatigue limit of the different deformed specimens allowed the determination of a proportionality factor in the range of 0.78 to 0.85.

•
Two different crack initiation modes are determined depending on the existing microstructure.
(i) Within the cast/HIPed and HT microstructure, the crack initiates at the boundaries of coarse (α 2 /γ) col near the sample surface and cleaves the colony in an interlamellar way along the α 2 /γ-interfaces. (ii) Initial defects are identified as elongated and globular γ-grains in deformed and heat treated material conditions. Consequently, to attain a homogeneous and defect-free microstructure is pivotal for the fatigue behavior. Based on the results of tensile and rotating bending tests, the process of hot-extrusion is a promising method for the production of material conditions required in high-performance racing applications. The high fatigue limit and strength of the deformed TNM alloy are beneficial for applications with fast cyclic loading conditions. The fracture behavior indicates a high sensitivity of the investigated alloy system on the microstructure and prevailing inner defects. Therefore, the effect of grain refinement and homogenization during the deformation process provides promising and balanced mechanical properties.