1 Introduction

Titanium aluminide (TiAl) alloys based on the ordered γ-TiAl phase have drawn much attention in recent years, due to their great potential in high temperature applications [1, 2]. The main characteristics distinguishing this intermetallic material from other construction materials are high specific yield strength, low density (3.9–4.2 g/cm3) high corrosion resistance and good creep properties at high temperatures [3,4,5]. The areas of application includes isothermally forged turbine blades for aircraft engines, as TiAl blades offer reduced weight and similar strength compared to the conventional nickel-based blades [6]. For this purpose, TNM-B1 is one of the most commonly used titanium aluminides with a chemical composition of Ti–(42–45)Al–(3–5)Nb–(0.1–2)Mo–(0.1–0.2)B (in at.%). However, the application possibilities are still limited, mainly due to the difficult production conditions. The limited workability leads to excessive wear of the expensive molybdenum-based tools, which in turn frequently leads to die fractures [7, 8]. As already shown by several studies, a suitable heat treatment can be a first step towards improving hot forming behaviour and microstructural development of titanium aluminides [9,10,11]. By using an appropriate heat treatment, optimal deformation properties can be achieved, allowing forging at either lower temperatures or higher speeds without damage. In this work, different heat treatments with the aim of improving the deformation behaviour of TNM-B1 have been analyzed and compared.

2 Materials and methods

The alloy TNM-B1 investigated in this work has a nominal composition of Ti–43.5Al–4Nb–1Mo–0.1B (in at.%). The alloy is a titanium aluminide based on the ordered γ-TiAl phase. It was produced by GfE Metalle und Materialien GmbH (Nürnberg, Germany) using a VAR skull melter and cast by a centrifugal casting process [2, 12]. Afterwards the material was hot isostatically pressed (HIPed) to close the casting porosities using the parameters 1200 °C, 200 MPa and 4 h. The produced ingot had a height of 150 mm and a diameter of 49 mm. Literature reveals that the Al-content has a big influence on the transition temperatures [2]. A modification in Al-content of 0.5 at.% leads to a shift in the γ-solvus temperature of about 30 °C. Table 1 shows the exact composition certified by GfE. A slight deviation in the Al-content of 0.2% with respect to the nominal value can be observed.

Table 1 Chemical composition of alloy TNM-B1 certified by GfE
Full size table

In order to find a suitable heat treatment (HT), three slices with a thickness of 10 mm were cut from the original cast and HIPed ingot. To avoid chemical impurities which may arise during production, all slices were cut from the same end of the ingot. Each slice was then cut into several cubes (approximately 10 × 10 × 10 mm). All heat treatments were performed in an industrial furnace from Nabertherm GmbH (Lilienthal, Germany) under atmospheric conditions. The heat treatment schedule is illustrated in Table 2.

Table 2 Considered heat treatment procedures for Ti–43.7Al–4Nb–1Mo–0.1B (in at.%)
Full size table

The first part of the heat treatment consisted of solution annealing. For this purpose, the temperatures were selected from various fields of the TNM-B1 phase diagram [13, 14]. The first temperature of 1230 °C is below the γ-solvus, the second of 1300 °C between the γ-solvus and the β-transus and the third of 1400 °C above the β-transus. All solution heat treatments were done for 1 h followed by air cooling (AC). Subsequently, the samples were heated once more to annealing temperatures of 900 °C, 1000 °C or 1100 °C for 5 h or 10 h followed by AC. The combination of solution heat treatment at 1400 °C and 10 h annealing time was not considered.

The microstructure of the samples was studied using a TESCAN Mira II XMH (Brno, Czech Republic) scanning electron microscope (SEM). All images were made in back-scattered electron (BSE) mode with an acceleration voltage of 15 kV. To ensure adequate image quality the sample surface was ground with a RotoPol-22 from Struers (Berlin, Germany). SiC grinding paper with a gradation ranging from 240 to 2500 was used. After grinding the samples were polished in a mixed solution of distilled water and colloidal silica, from LECO Corporation (St. Joseph, Missouri, USA) by using a VibroMet 2 from Buehler (Uzwil, Switzerland) for 72 h.

SEM images with an area of 252.8 µm2 were used for the microstructure analysis. The images were taken in the middle of the samples. For the analysis of the phase content and the grain size, the software Atlas from TESCAN was used. Since only the β/β0 phase can be determined exactly by the optical evaluation, only these values were used for the evaluation. To analyse the lamellar colonies, the lamellar colonies were colored using a single color. Subsequently a software called Gwyddion 2.42 was used to determine the lamellar content.

The phase transition temperatures were verified using high-temperature differential scanning calorimetry (DSC) with the device 404F3 from Netzsch (Selb, Germany) and a heating rate of 40 K/min from RT to 800 °C and 10 K/min from 800 to 1450 °C. The contact surface of samples with 5 mm in diameter and approximately 1 mm in height was made planar by means of grinding. The measurements were conducted in an Ar-atmosphere using a gas flow of 40 ml/min. A temperature accuracy of about ± 5 °C below and ± 7 °C above 1300 °C was obtained. To ensure reproducibility, the experiments were repeated five times for each condition.

The original cast and HIPed ingot (h = 150 mm, Ø = 49 mm) was first cut (h = 95 mm, Ø = 49 mm). Smaller cylinders (h = 95 mm, Ø = 7 mm) were produced from the ingot by electrical discharge machining (EDM) by Erocontur GmbH (Müncheberg, Germany). Both the HIPed and the heat treated cylinders were cut and turned into test samples (h = 8 mm, Ø = 5 mm). Subsequently, hot compression tests were performed with a DIL805A/D/T quenching and deformation dilatometer from TA Instruments (New Castle, Delaware, USA) former Bähr Thermoanalyse GmbH. Molybdenum plates with a thickness of around 1 mm were placed between the specimen and the Si3N4-punches inside the dilatometer. The tests were conducted in an Ar-atmosphere to protect the samples from oxidation. The cylindrical samples were isothermally deformed at a constant strain rate of 0.01 s−1 and 0.001 s−1 at the constant temperature of 1150 °C. Samples and molybdenum plates were first heated by induction using a heating rate of 10 K/s, and held at the testing temperature for 3 min to obtain a homogeneous microstructure and quasi-isothermal conditions prior to deformation. Finally, the samples were deformed to a total strain of 0.8 and cooled using argon gas with a cooling rate of 150 K/s. The microstructures after forming have been recorded in the middle of the sample.

3 Results

3.1 Heat treatment study

According to [11], the solidification path from the liquid to the stable solid for alloy TNM-B1 is: L → L + β → β → α + β → α + β + γ → α + β + β0 + γ → α + β0 + γ → α + α2 + β0 + γ → α2 + β0 + γ. Figure 1a shows the initial material TNM-B1 produced by GfE, which was cast and HIPed. A typical three-phase microstructure can be observed, consisting of γ-TiAl (dark), α2-Ti3Al (gray) and β/β0-TiAl (bright). The fine and homogeneous structure due to the solidification after casting was further refined by the subsequent HIP process. The microstructure includes 57% α2/γ-colonies, 18% β/β0 phase, 25% globular γ grains, and is classified as near lamellar [3]. The lamellar colonies mainly consisted of γ-TiAl and α2-Ti3Al. Occasionally some β0-TiAl can be observed within the colonies. This can be attributed to the HIP process, in which α is converted to γ + β. In addition, this can lead to an increase in thickness of the γ-lamellae. Furthermore, it can be assumed that ω-domains are embedded into the β/β0 phase [15]. The α2 phase has a D019 structure and the highest strength and lowest deformability within the TNM-B1 system. The γ phase has a L10 structure with higher strength than the β/β0 phase, but higher deformability than the α2 phase. The β0 phase has a B2 structure and the disordered β phase a A2 structure. At room temperature the β and the β0 phases are brittle and exhibits low deformability. However, at hot forging temperatures the both phases become more deformable and are the softest phases in the TNM-B1 system [16]. This occurs in the β phase partly due to sufficient independent slip systems being available at high temperatures [17]. For the β0 phase however, the Burgers vector is much larger than for the β phase. This causes the β0 phase to be comparatively less deformable than the β phase. Therefore, the parameter of interest for improved formability is not only the volumetric fraction of β/β0 phase, but also the dispersion and individual fractions of the two. Also the β/β0 phase carries a much higher strain during hot deformation than the nominal value, leading to highly misoriented subgrains and dynamic recrystallization (DRX).

Fig. 1
figure 1

Microstructure of alloy Ti–43.7Al–4Nb–1Mo–0.1B (in at.%): a initial material (cast + HIPed), annealing of 1 h at b 1230 °C, c 1300 °C and d 1400 °C

Full size image

To produce a microstructure for TNM-B1 with higher formability than the cast and HIPed one, several two-step heat treatments were considered. The first step of each heat treatment was a solution annealing in three different areas of the phase diagram. For annealing below the γ-solvus 1230 °C (α + β/β0 + γ field), above the γ-solvus 1300 °C (α + β field) and 1400 °C (α + β field) for the area near the β-transus boundary were chosen.

In the following, the microstructures are compared on the basis of the three parameters that have already been shown to have a positive effect on the forming behavior of TiAl [18, 19]. These parameters are high β/β0 phase fractions, high cellular phase fractions and low proportions of lamellar colonies.

Figure 1b shows the microstructure of alloy Ti–43.7Al–4Nb–1Mo–0.1B (in at.%) after annealing for 1 h at 1230 °C followed by AC. The microstructure consists of α2/γ-colonies with a significant amount of α and a few γ platelets with a lamellar content of 91.4%. Between these colonies is a globular structure consisting of β/β0 phase with a degree of γ grains at the colony boundaries. The proportion of the β/β0 phase has decreased from 18.0 to 7.4% compared to the cast and HIPed material. Figure 1c displays the microstructure after annealing for 1 h at 1300 °C followed by AC. Compared to Fig. 1b, solution annealing took place in the α + β phase region. Since this is above the γ-solvus, the γ content is very low and can be attributed to the cooling process. The microstructure consists of 92.3% lamellae. The β/β0 content is 5.5%. Figure 1d shows the microstructure of alloy Ti–43.7Al–4Nb–1Mo–0.1B (in at.%) after annealing for 1 h at 1400 °C followed by AC. Similarities to 1c are clearly recognizable, which is due to the fact that both annealing temperatures are above the γ-solvus. Such a solution treatment above the γ-solvus is also known as stabilization treatment [15]. Therefore Fig. 1c and 1d show very fine lamellae within the colonies. The lamellar content of the microstructure presented in Fig. 1d is 96.2% and the average colony size is much larger than in Fig. 1b and c due to the increased solution annealing temperature. The proportion of the β/β0 phase is 2.9%. Furthermore, no γ platelets can be seen in the β/β0 phase of Fig. 1d. In contrast, in Fig. 1c the needle-shaped γ platelets are distributed throughout the β/β0 phase. Thus, the character of the γ platelets seems to be changed. Further research is required for verification. It can be assumed that the platelet formation is considerably limited by the lower β content of the microstructure in Fig. 1d.

Figure 2 displays the microstructures after 5 h for the second heat treatments. The microstructures for the annealing temperatures 900 °C, 1000 °C and 1100 °C are shown from left to right. The different temperatures for the first stage of the heat treatment are shown from top to bottom.

Fig. 2
figure 2

Microstructure of alloy Ti–43.7Al–4Nb–1Mo–0.1B (in at.%) after both steps of heat treatment: X1) 1 h/1230 °C + 5 h/900 °C, X2) 1 h/1230 °C + 5 h/1000 °C, X3) 1 h/1230 °C + 5 h/1100 °C, X7) 1 h/1300 °C + 5 h/900 °C, X8) 1 h/1300 °C + 5 h/1000 °C, X9) 1 h/1300 °C + 5 h/1100 °C, X13) 1 h/1400 °C + 5 h/900 °C, X14) 1 h/1400 °C + 5 h/1000 °C, X15) 1 h/1400 °C + 5 h/1100 °C

Full size image

After the second heat treatment, more γ platelets are formed within the α grains, due to the longer heat treatment (Fig. 2a). Within the β/β0 phase, first parts of the cellular reaction are recognizable, which resembles a disordered accumulation of γ platelets [20]. This reaction starts at the α2/γ-colony boundaries according to (α2 + γ)lam → (α2 + β + γ)cell [11]. A detailed view is shown in Fig. 3a. The γ fraction of the cellular reaction consists of a higher concentration of platelets than the remaining γ fraction within the β0 phase. The isolated γ platelets form during cooling inside the β/β0 phase. In addition, globular γ. grains are located at the grain boundaries of the α grains, which precipitate during cooling from the α grains (in 2c from the α2/γ lamellae). The globular fraction in Fig. 2a is 14.1% and increases with increasing annealing temperature, thus the fraction in Fig. 2c reaches 44.6%. The higher annealing temperature of the microstructure distributed in Fig. 2c has ensured complete formation of the lamellae inside the α grains, resulting in α2/γ-colonies. Such a microstructure can be defined as duplex one according to [3].

Fig. 3
figure 3

Characteristics of the cellular reaction in Ti–43.7Al–4Nb–1Mo–0.1B (in at.%) after heat treatment: X1) 1 h/1230 °C + 5 h/900 °C, X7) 1 h/1300 °C + 5 h/900 °C, X8) 1 h/1300 °C + 5 h/1000 °C

Full size image

The microstructure distributed in Fig. 2d is similar to the one of 1c. Due to the longer heat treatment in the second step, the average α2/γ-colony size is decreased. The percentage of colonies is almost identical with 83.2%. The globular content of the γ phase is completely disappeared. Within the β/β0 phase (5.5%) the percentage of γ platelets is increased (right side of Fig. 3b). In Fig. 2d the cellular reaction can also be observed (left side of Fig. 3b). The y platelets and the cellular reaction can be distinguished by the strongly different γ content. In condition Fig. 2e these effects are more pronounced. The majority of the microstructure is formed by the cellular structure. There are different preferential directions within the cellular structure. This is shown in detail in Fig. 3c. Furthermore, the β/β0 content within the microstructure of Fig. 2e is increased to 14.0% and the lamella content is decreased to 17.7%. In the microstructure of Fig. 2f the lamella content drops to 10.4%. In addition, the thickness of the lamellas is increased. Within the cellular structure, the distance between the γ platelets and the lens shaped β/β0 portion are increased as well. The β/β0 content remains nearly identical to the one in Fig. 2e with 21.2%.

Figure 2g shows a nearly lamellar structure [3]. Compared to Fig. 1d, the colony size is decreased. The percentage of lamella colonies (84.4%) does not differ from the ones in Fig. 2a and d. Furthermore, the β/β0 content is with 4.5% the lowest of the tested heat treatments. This is probably a result of secondary heat treatments with an annealing temperature of 900 °C, and should be considered for future research. At many α2/γ-colony boundaries a weak approach of the cellular reaction can be recognized. In the microstructure Fig. 2h, the β/β0 content rises to 8.1%. At the same time, the lamellar portion decreases to 59.5%. This value is closer to the value in Fig. 2b than to the value in Fig. 2e. Thus, the content of the structure due to cellular reaction is significantly lower in Fig. 2h than in both Fig. 2e and i. Despite the higher β/β0 content of 18.9%, the proportion of lamella colonies (29.1%) is considerably higher than after the heat treatment with lower temperature in the first stage (Fig. 2f). However, cooling from the β region (1400 °C) results in a higher grain refinement compared to cooling from 1230 and 1300 °C. Table 3 shows the values determined for all heat treatments under consideration.

Table 3 Overview of the β/β0 and the lamellar content (in vol.%) in the air-cooled heat treated samples
Full size table

In general, it can be observed that the proportion of lamella colonies is lowest after solution annealing at 1300 °C (X7, X8, X9). At the same time, the cellular content is highest. On the other hand, the β/β0 content is further increased with increasing annealing temperature. The heat treatment X9 (1 h/1300 °C + 5 h/1100 °C) (ref. Table 2) leads to a microstructure suitable for hot forming, as it has the comparatively lowest lamella content and a relatively high β/β0 content. In addition, the content of the resulting structure due to the cellular reaction is the highest compared to the other heat treatments.

The microstructures after heat treatments X5–X6 and X10–X12 (ref. Table 2) are not explained in detail as the results hardly differ from those with 5 h annealing time. With regard to a potential industrial application of the final heat treatment, all treatments with a solution temperature of 1400 °C (X13, X14, X15) were excluded from further investigations. It can be assumed that this temperature is too high to achieve an economic benefit. Furthermore, the heat treatment routes X2 and X8 were excluded because results between X1 and X3 as well as X7 and X9 can be assumed. Thus, the four heat treatments X1, X3, X7 and X9 were selected for further investigation.

3.2 Hot-compression tests

In this investigation, the constant forming temperature of 1150 °C was purposely chosen to be below the industrial forging window for cast and HIPed TNM-B1 (1230 to 1250 °C). This was done in order to better assess and compare the deformation and damage behaviors of the heat treated material states, as the differences are expected to be more pronounced at this lower temperature. Furthermore, this was done to assess if the industrial forging window could be expanded to include 1150 °C using one of the studied heat treatments. The measured flow curves from the hot-compression tests for (a) T = 1150 °C/ \(\dot{\varepsilon } = 0.01 {\text{s}}^{ - 1}\) and (b) T = 1150 °C/ \(\dot{\varepsilon } = 0.001 {\text{s}}^{ - 1}\) are depicted in Fig. 4. The samples were deformed to a strain of 0.8. The maximum flow stress decreased with decreasing strain rate. All flow curves show a pronounced stress peak at strains between 0.02 and 0.04 before decreasing due to softening. The initial sharp increase in flow stress can be attributed to work hardening caused by rapid dislocation multiplication [21]. When the defect density reaches a critical value, dynamic recrystallization (DRX) begins, at the so-called critical strain. In general, the flow behavior of the material is governed by the competition of work hardening and work softening due to dynamic recovery (DRV) [22], phase rotation [23], deformation twinning [24], and platelet bending/kinking [25]. The peak stress marks where the hardening and softening processes are in balance. The peak stress values of the cast and HIPed material are (a) 164.5 MPa for \(\dot{\varepsilon } = 0.01 {\text{s}}^{ - 1}\) and (b) 80.2 MPa for \(\dot{\varepsilon } = 0.001 {\text{s}}^{ - 1}\). After the heat treatment X3 these values change to (a) 151.7 MPa and (b) 77.3 MPa, respectively. The values for the peak stress after heat treatment X9 are the lowest for both deformation velocities with (a) 146.6 MPa and (b) 64.4 MPa, respectively. This results in a peak stress reduction of more than 10% for heat treatment X9 compared to the cast and HIPed, at the tested conditions.

Fig. 4
figure 4

Flow curves of alloy Ti–43.7Al–4Nb–1Mo–0.1B (in at.%) at a T = 1150 °C/ \(\dot{\varepsilon } = 0.01 {\text{s}}^{ - 1}\) and b T = 1150 °C/ \(\dot{\varepsilon } = 0.001 {\text{s}}^{ - 1}\): starting material (cast + HIP), X1) 1 h/1230 °C + 5 h/900 °C, X3) 1 h/1230 °C + 5 h/1100 °C, X7) 1 h/1300 °C + 5 h/900 °C, X9) 1 h/1300 °C + 5 h/1100 °C

Full size image

3.3 Dynamic differential calorimetry

In order to ensure the operational capability of heat treatment X9, DSC measurements of the samples processed using heat treatment X9 and the cast and HIPed material were carried out. Table 4 displays the eutectoid temperature \({\text{T}}_{\text{eu}}\), the β0 → β transition temperature \({\text{T}}_{{\upbeta_{0} }}\), the γ-solvus temperature \({\text{T}}_{{\upgamma{\text{solv}}}}\) and the β-transus temperature \({\text{T}}_{\upbeta}\). The values represent the average value of 5 measurements. The results correlate well with the phase transition temperatures measured for the cast and HIPed samples.

Table 4 Phase transition temperatures [°C] of Ti–43.7Al–4Nb–1Mo–0.1B determined by means of DSC
Full size table

Subsequently, the reproducibility of heat treatment X9 was assessed. For this purpose, heat treatment X9 was independently performed four times Each sample was examined by means of SEM. The results are shown in Table 5. There was only one deviation in the measurement of the lamellar portion. Measurement 3 showed a lamellar proportion of 17.4%. This is slightly higher compared to the other measurements.

Table 5 Microstructure parameters of all separately performed heat treatments according to X9
Full size table

4 Conclusions

A set of two-step heat treatments was developed, and the resulting microstructures were analyzed with the aim of improving the hot workability of TNM-B1. For this purpose, three solution annealing temperatures (1230 °C, 1300 °C and 1400 °C) and three annealing temperatures (900 °C, 1000 °C and 1100 °C) were chosen. For the annealing temperatures, two different processing times (5 h and 10 h) were considered. However, no significant difference between the two could be observed. Therefore, 5 h was chosen for this investigation. The combination of 1 h/1300 °C + 5 h/1100 °C was shown to be the most suitable due to the resulting high cellular reaction fraction, the high β/β0 fraction and the low amount of lamellar colonies, leading to beneficial deformation behavior. Hot compression tests at the temperature 1150 °C and the strain rates 0.01 s−1 and 0.001 s−1 to a total strain of 0.8 were conducted. The heat treatments X9 (1 h/1300 °C + 5 h/1100 °C) and X3 (1 h/1230 °C + 5 h/1100 °C) resulted in about 10% lower flow stress, when compared with the cast and HIPed material. Finally, DSC measurements showed no large deviations in phase transition temperatures for the heat treatment X9.