Characterization of the Interface Between α and β Titanium Alloys in the Diffusion Couple

The LM observation revealed the single-phase microstructure of the examined titanium alloys (Figure 2). The microstructure of α-Ti alloy (Figure 2(a)) was composed of recrystallized equiaxed α grains—typical for hot-rolled titanium sheets. In the case of deformed β-Ti alloy (Figure 2(b)), elongated grains are observed. Such preferential orientation of the grain orientation is caused by the mechanical treatment of β alloy, namely, by the cold rolling process.

The annealing process at 850 °C led to grain growth resulting from further static recrystallization in both coupled materials (Figure 3(a))—especially in β-Ti. Clear α/β interface was detected between the α-Ti and β-Ti alloys (Figure 3(b)). Its location was confirmed by the linear distribution of alloying elements using the SEM/EDS method (Figure 4). Near the α/β interface, the diffusion-affected zones (DAZs) were easily observed. The zone in α-Ti was marked as DAZ-1, while on the β-Ti side, it was labeled as DAZ-2. The aforementioned DAZs were characterized by different microstructure when compared with the regions located further from the α/β interface. Moreover, within the DAZ-2 (in β-Ti), formation of the pores aligned in the direction parallel to the α/β interface was observed. The pores formed the so-called Frenkel plane (Figure 3(b)). Such porosity observed in diffusion couples is related to the Kirkendall effect.[21] It is generally accepted that cavities are formed by the precipitation of supersaturated vacancies produced by unequal diffusion currents of the Kirkendall effect—from β-Ti, containing alloying elements, to commercial pure titanium (α-Ti) in the described case. Control of voiding in the alloys is an important practical problem of materials science. As a rule, the voiding should be suppressed since it leads to failure of final parts, e.g., microelectronic circuits.[25,26] The voids are generated from the beginning of the diffusion process at the original interface between the α/β interface, and during the diffusion process, they grow and move with the extremum of diffusion flux of the most mobile component toward the β-Ti side. The diffusion process in β-Ti can be observed; the measured element concentrations change with the measuring distance in β-Ti. Meanwhile, in α-Ti, the Ti concentration is almost constant. This means that α-Ti is consumed during the diffusion process (Figure 4).

When two materials with different chemical composition in contact with each other are exposed to elevated temperature for a certain time, the diffusion processes will surely occur due to the difference in chemical potential.[18,19] The diffusion is undoubtedly connected with the mass transport in the solid state. It is widely known that the activation energy for diffusion through the grain boundaries is smaller as compared with the activation energy for volume diffusion.[19] Therefore, the conclusion is that the mass transport through the grain boundaries will occur much earlier than that through the volume of the grain. Using SEM/BSE methods, the grooving of the α/β interface was confirmed in the examined diffusion couple (Figure 5).

The grooving phenomenon in the examined couple seems to be determined by grain boundaries in both materials, namely, in β-Ti (marked as GB-β (Figure 5)) and α-Ti (marked as GB-α (Figure 5)). It is worth mentioning that both α and β phases penetrate the α/β interface (the red and blue arrows, respectively, in Figure 5). Moreover, the SEM/EDS measurement of the chemical composition on the GB-β revealed slightly lower content of Ti. It is commonly observed that the alloying elements tend to segregate at grain boundaries such as in the present case. Thus, due to the reactive diffusion process, the boundary between the α/β phase moves into the α-Ti side (α-Ti is consumed). Due to the assumed presence of alloying elements at grain boundaries, a different diffusion flux at grain boundaries and grain volume is observed and, as a consequence, the grooving is observed.

Apart from grooving the α/β interface, microstructural changes in the DAZ-2 (β-Ti) were also observed (Figure 6(a)). As shown in a magnified view of the DAZ-2 region (Figure 6(b)), apart from pores, formation of needlelike precipitations took place (Figure 6(b)).

Martensitic β → α’(α″) transformation in titanium alloys is usually related to the high cooling rate from the β phase or α + β regions or to the plastic deformation (deformation-induced martensitic transformation). Even if cooling conditions after annealing favored martensite formation, it is more probable in the case of α-Ti[3] (if during annealing the temperature could be within or above α + β → β transformation) than in β-Ti. The orientation of precipitates is typical for Widmanstätten microstructure:[3] α laths could be formed as a result of Ti diffusion from α-Ti alloy. In fact, parallel laths are detected also in DAZ-1 (α-Ti) (Figure 6(b)). It is interesting that their orientation is different in neighboring α grains (Figure 7). Lath formation in DAZs, both in α-Ti and β-Ti, caused by diffusion of alloying elements seems to be an interesting phenomenon and needs to be explained in further investigation. Deformation-induced martensitic transformation due to the stress generated by the diffusion process or the grain growth in both dissimilar titanium alloys should also be taken into consideration.

The EBSD studies in the region of the α/β interface (marked with an arrow in Figure 8(a)) indicated that the lamellar zone was formed in near-β alloy composed of α phase plates (Figures 8(b) and (c)). The phase distribution map (Figure 8(c)) indicates that the reaction front moved toward α-Ti (the α-Ti phase is consumed). This observation is supported by the left side of the mentioned map. No interdiffusion zone is observed in the α-Ti side. Meanwhile, in the β-Ti side, an interdiffusion zone is observed. This interdiffusion zone is characterized by the number of differentially oriented grains (Figure 8(b)). The EBSD orientation map (Figure 8(b)) confirmed the grooving of the α/β interface by the α grain, marked with the frame in Figure 8(a), which keeps its orientation in the β phase matrix. The region of α laths in the β phase matrix, among the α/β interface, gradually vanishes in the bulk of the β grain.

Additionally, TEM observations were carried out to learn more details about the indicated phase constituents. A low-magnification image (Figure 9(a)) revealed differently oriented needle-shaped phases. Their interior exhibits nonhomogenous amplitude contrast, which can be interpreted as the presence of an internal strain. The diffraction pattern taken from this region (Figure 9(b)) clearly indicates spots from both α and β phases.

The results of nanohardness measurement indicate a large difference between the hardness of examined α-Ti and β-Ti alloys—about 210 and 300 HV, respectively (points 8 and 1 in Figure 10). Moreover, hardness in both DAZs is higher. In the case of β-Ti alloy, it results from the presence of α lamellae, confirmed by all microscopic methods used; the measured hardness of about 350 HV is typical for two-phase titanium alloys having lamellar microstructure.[36] It is worth mentioning that hardness in this region decreases toward the α/β interface (Figure 10), which can be associated with a diffusion of alloying elements to α-Ti alloy (Figure 4(b)) and pores aligned in the Frenkel plane. The alloying elements diffusing from the β-Ti alloy harden α-Ti alloy in its DAZ. Except for a quite thin (less than 10 µm) layer of laths (Figure 8), the microstructure of the DAZ in the α-Ti alloy is single phase. Therefore, it can be assumed that the observed hardening was caused by a solid solution strengthening mechanism.

For a better description of the experimental results, the model predicting the void growth during the annealing time was introduced to allow extrapolation of the process. The voids are generated during the annealing process and moved with the extremum flux of the most mobile component (Ti in the present case). Thus, the entire model can be simplified to solve one equation and the radii can be calculated when the vacancy diffusion coefficient and mean free path of vacancies are known. To correlate the experimental observations with the model, a measurement of the average void size was performed for the images taken by the LM, as shown in Figure 11. The average size of the voids was determined based on 45 measurements of randomly chosen locations in the cross section of the diffusion couple after the annealing process. The average diameter of the voids was measured to be 12.3 µm.

To calculate the void radius, the kinetic equation can be used as follows[25]:where R is the the void radius; L_V is the the mean free path of vacancies; D_V is the vacancy diffusion coefficient; and N_V and \( N_{\text{V}}^{\text{eq}} \) are the vacancy molar fraction and equilibrium vacancy molar fraction \( \left( {N_{\text{V}}^{\text{eq}} \, = \,0.0002} \right) \), respectively. The strength of sinks can be characterized by the mean free path of vacancies \( L_{\text{V}} \approx \surd \left( {D_{\text{V}} \tau_{\text{V}} } \right) \). The mean migration length describes the distance covered by a vacancy. It can be indicated that the high value of the L_V parameter results in a higher average void radius. In the case of the interdiffusion system between two components, the vacancy diffusion coefficient can be determined as[26]where \( D_{i}^{\text{I}} \) is the intrinsic diffusion coefficient of the ith component.

Equation [1] was solved numerically to determine the radius of the void. It should be noted that when the initial stage for void formation is neglected, this equation can be reduced to the following form:and solved by the simple Euler scheme:where R^min is the minimum void’s radius and t is the time of the experiment (duration in seconds).

The data used in the calculations are as follows:The calculated radius of voids was in good agreement with the experimentally determined values (Table I). Potentially, the model can be extended to simulate the movement of the boundary between α and β phases and to calculate the phase concentration in the interdiffusion zone. However, only a simplified model was presented in this article. It seems that such an approach can be used as a criterion in the material selection for the DB process of dissimilar titanium alloys. In the case of DB technology, where a compressive load is applied, some deformation mechanisms should be taken into consideration as well.[34,35]