Ti-17 (Ti-5Al-2Sn-2Zr-4Mo-4Cr) is a beta-rich alpha-beta (α + β) titanium alloy exhibiting high strength, good fracture toughness and excellent creep resistance, which is used for heavy-section forgings of gas turbine engine components, such as fan or compressor discs, at elevated temperature in aerospace industry [1,2]. Due to addition of 4 wt% of the easy-segregating alloying element Cr, difficulties in controlling the chemical homogeneities on the macro- and micro-scales have been well known during melting process [3,4]. A typical melt related defect containing a high content of Cr, which has been termed as beta-fleck (β-fleck) frequently occurs in Ti-17 alloy. Experience has shown that, once formed, β-flecks are very difficult to eliminate. The defects cause an extremely detrimental influence on the mechanical properties, particularly on low cycle fatigue and fracture toughness [[5], [6], [7]] and may impose uncontrollable risks on safety operation of components. Therefore, well understanding and precisely predicting the formation of β-flecks are necessary to eliminate this kind of defect and improve the manufacturing processes.

Segregated areas with similar shapes and ranges are referred as channel segregation, also called A-segregation, freckles, or chimneys in industrial castings, which have long been a subject of investigation in the fields of solidification studies [[8], [9], [10], [11], [12]]. The occurrence and evolution of these segregation behaviors are thought to be induced by such forces as the density driven force from the natural thermosolutal convection, the solid movement force of the grain settling or flotation, or the shrinkage force of the volume contraction during solidification [13,14]. For titanium alloys, it’s generally agreed that β-flecks are caused by micro-segregation, which contain a higher content of beta stabilizing elements than in the bulk, usually iron and/or chromium, and appear in the range from hundreds of micrometers to several millimeters in ingots [15,16]. Mitchell et al. [15] proposed that the formation mechanism of β-flecks is the same as freckles which are caused by a fluid flow resulting from liquid density differences. Yuan et al. [17] used phase field method to study equiaxed dendritic growth during vacuum arc remelting (VAR) and thought that the segregated solute trapped by settling grains causes the micro-segregation. Although much research work has been done to investigate the micro-segregation, the conclusions are still controversial and few experimental evidences have been published on the segregation behaviors of β-flecks in titanium alloys.

In order to clarify the possible causes of β-flecks, the directional solidification method is an effect way to study the solute distribution and microstructure evolution during solidification process. Based on unidirectional thermal field and simplified concentration field, solute distribution can be examined with the evolution of the solid-liquid interface [[18], [19], [20], [21]]. Planar front can be stabilized at low growth rates. This near equilibrium process can clearly reveal the distribution rule of each solute component [[22], [23], [24]]. By increasing growth rate, cellular or dendritic growth can be observed to investigate industrial solidification process approximately [25,26]. The aim of this article is to: (i) illustrate the segregation behavior of alloying elements in Ti-17 alloy; (ii) establish the relationship between solid-liquid interface and solidification rate under directional solidification condition; (iii) predict the possible onset of β-flecks formation with the theoretical model in the microstructural level and try to make clear the formation mechanisms of β-flecks in Ti-17 alloy.

Feed and seed rods for directional solidification were prepared by industrial Ti-17 alloy. Each rod was machined to a cylindrical shape with a diameter of 9 mm and a length of 100 mm. Chemical composition is listed in Table 1.

Fig. 1 shows the schematic of optical floating zone technique. Crystal growth was performed in the optical floating zone furnace. Four xenon lamps were installed as infrared radiation sources. The high temperature gradients required to stabilize the molten zone were ensured by four ellipsoidal reflectors focusing the energy of four lamps on the common focus point. The temperature of the molten zone was precisely controlled by adjusting the current intensity through the lamps. In this study, the temperature of the molten zone can be estimated to be around 1700 °C (T_L+20-30 °C) and the temperature gradient was about 2 × 10⁴ K/m. The crystal growth was carried out in an enclosed quartz tube, where high-purity argon gas was filled and the controlled gas pressure was applied to prevent oxidation and inclusion. Growth experiments started on the polycrystalline seed rod. The feed rod and the growing crystal rotated at 15 rpm in the opposite direction to ensure an efficient mixing and uniform temperature distribution in the molten zone. Six different withdraw rates (3 mm/h, 6 mm/h, 15 mm/h, 30 mm/h, 60 mm/h and 150 mm/h) were applied and we successfully obtained the high-quality crystals with a height of 15 mm.

After solidification, all samples were heat treated at a special scheme of 910 °C/0.5 h/FC+875 °C/1 h/WQ. The β transformation temperature (T_β) of Ti-17 is about 895 °C determined by the metallographic method. The former 910 °C/0.5 h/FC is to obtain a uniform Widmanstätten type microstructure, and the latter 875 °C/1 h/WQ is for the convenience of detecting β-flecks using optical microscope (OM) method, which is a normal operation for observation of β-flecks in Ti-17 alloy [15].

For micro/macrostructure observation and composition analysis, all samples were cut along the longitudinal section and prepared by the standard mechanical polishing operation. They were subsequently etched in a mixture solution of HF: HNO₃: H₂O with a ratio of 1:3:40 for micro/macrostructure observation. The macrostructure was examined on a STEMI2000 microscope. The microstructure was examined on a scanning electron microscope (SEM). The composition analysis of alloy elements was analyzed using energy dispersive spectroscopy (EDS) on a ZEISS MERLIN Compact SEM. The energy and capture time of the EDS was 20 keV and 60 s respectively. The elemental distribution of all samples was determined by EDS in a 100 μm × 80 μm tetragonal area in the longitudinal section.

Fig. 2 shows two typical macrostructures of the same sample, which corresponds to the as-solidified condition and the heat-treated condition respectively. No clear trace of elemental segregation can be seen in Fig. 2(a). However, macrostructure heterogeneity is found in Fig. 2(b), which is characterized by "four zones and two lines". Four zones refer to base zone (BZ), heat affected zone (HAZ), uniform growth zone (UGZ) and rapid separated zone (RSZ). Two lines refer to growth start line (GSL) and growth finish line (GFL). BZ is part of the seed rod which is unaffected or less affected by the heat produced during directional solidification process. HAZ is between BZ and UGZ, exhibiting a gradient macrostructure similar to HAZ found in a fusion weld [27]. UGZ is characterized by a uniform cylinder since the withdrawal rate is constant. In this region, the macrostructure is relatively uniform and the withdrawal rate of heating spot relative to the sample is equal to the solidification rate. On the top, RSZ exists in a convex parabolic contour and exhibits the darkest contrast. As for "two lines", GSL corresponds to the interface between UGZ and HAZ, while GFL corresponds to the interface between UGZ and RSZ. It should be noted that, when the feed rod and the seed rod separate rapidly, the molten zone is located in RSZ and the solid-liquid interface during solidification process can be observed around GFL. In this article, the emphasis is placed on the solid-liquid interface and its effect on RSZ and UGZ.

Fig. 3 shows macrostructures of all directionally solidified samples at various withdrawal rates from 3 mm/h to 150 mm/h after the heat treatment. It can be seen that at R < 15 mm/h, a single crystal forms along the longitudinal section in UGZ. RSZ and UGZ can be clearly distinguished by their distinct contrasts. At R = 30 mm/h, columnar grains growing upwards is found in UGZ and dendrites begin to develop in RSZ. At R > 30 mm/h, dendritic structure extends from RSZ to UGZ, and the size of coarse columnar grains decreases.

From Fig. 3, it can be seen that the size and distribution of the dark areas are closely related to the withdrawal rate. At R ≤ 15 mm/h, the dark area is integrate and homogeneous. Its shape corresponds well with the morphology of RSZ. With the increase of R, the integrity of the dark areas begins to collapse and gradually transforms into a dendritic structure. At R ≥ 60 mm/h, GFL separating UGZ and RSZ becomes unclear and dendritic structure begins to extend to UGZ. When R is up to 150 mm/h, no clear differences can be found between the macrostructures of UGZ and RSZ. It can be concluded that the effect of withdrawal rate on macrostructure is mainly embodied in the variation of the dark areas, which are mainly located in RSZ at low R and extending to UGZ when dendritic structure forms with increasing R.

Fig. 4(a) and 4(b) show the microstructures inside and outside the dark areas respectively. At high magnification, no α phase exists in the dark area. Many α plates embedded in β phase matrix are found outside the dark area, showing light contrast at low magnification.

The microstructures around GFL are shown in Fig. 5. At high magnification, a smooth planar interface between RSZ and UGZ can be observed at R = 3 mm/h (Fig. 5(a)). Instability of the planar interface begins occur at R = 6 mm/h (Fig. 5(b)). At R = 15 mm/h, cellular structure with a height of 500 μm begins to develop, and GFL is not integrate and exhibits a complex shape (Fig. 5(c)). At R = 30 mm/h, the height of the cellular region increases to about 2 mm (Fig. 5(d)). At R = 60 mm/h, a transition from cellular structure to dendritic structure occurs (Fig. 5(e)). At R = 150 mm/h, fully developed dendritic structure is found but second dendrite arms are not clear (Fig. 5(f)).

3.3.1. Intercellular or interdendritic composition

Relationship between elemental content (Cr, Zr and Mo) and withdrawal rate is given in Fig. 6. The sites for composition analysis are in the dark contrast areas in Fig. 5(d)-5(f) near the solid-liquid interface. From Fig. 6, we can see that the maximum content of Cr and Zr emerges at R = 30 mm/h with a cellular structure. With the increase of R, Cr and Zr contents gradually decrease to a nearly constant value. The fluctuation range of Zr content is smaller than that of Cr. However, the Mo exhibits an opposite trend with Cr. Minimum content of Mo occurs at R = 30 mm/h, exhibiting a quick increase from R = 30 mm/h to 60 mm/h and then gradually reaching a nearly constant value.

3.3.2. Elemental distribution along the longitudinal section

Fig. 7 shows two elemental distribution profiles from BZ to RSZ along the longitudinal section, corresponding to R = 3 mm/h and 150 mm/h respectively. At R = 3 mm/h, from BZ to GFL, Al keeps constant and then follows a drop above GFL. Mo increases quickly to the maximum at GSL, then follows a decrease trend in UGZ and a sharp drop at GFL. Cr exhibits a totally opposite trend with Mo and apparently segregates in RSZ. Zr exhibits a similar trend with Cr. Sn exhibits a similar trend with Al. The segregation phenomena of Cr and Mo fits well with the Ti-Cr, Ti-Mo binary phase diagrams [28]. The segregation of Zr in RSZ is in accordance with the enrichment of Zr in β phase in near α titanium alloys [29]. At R = 150 mm/h, from BZ to GFL, Al keeps constant and begins to fluctuate around GFL. Mo exhibits similar trend with that at R = 3 mm/h in a more zigzag way but the decreasing trend is not apparent. Cr fluctuates in a wide range above GSL and the mean level above GFL is a little higher. Both Zr and Sn exhibit a similar trend with Al. All elements exhibit large fluctuations of data points around GFL.

It can be seen from Fig. 7 that the withdrawal rate also has an apparent influence on the profile of elemental distribution curve. Fig. 8 shows the Cr, Mo and Zr profiles along the longitudinal section at different withdrawal rates. Each sample at different withdrawal rates shows a unique trend. As for Cr, when R ≤ 30 mm/h, the profiles are similar and can be roughly divided into three apparent stages: initial increasing stage (I), middle steady stage (II) and last rapid pull-up stage (III) (Fig. 8(a)). Stage I and II are located in UGZ and stage III is located in RSZ. The variation of Cr in stage I can be described as follows[30]:

Here, k is the partition coefficient of Cr as 0.67 in Ti-17 [28]. D is the liquid diffusion coefficient of Cr as 5 × 10^-9 m²/s [31]. C₀ is the initial concentration as 4 wt%. C_s is the concentration in solid phase. z is the distance from GSL. Fig. 9 shows the measured and calculated Cr profiles in stage I at the withdrawal rates of 3 mm/h to 15 mm/h. At R ≤ 6 mm/h, the experimental value fits well with the calculated value according to the Eq. (1). At R = 15 mm/h, the predicted value departs far away from the experimental value. It can be noted in Fig. 8(a) that at R = 15 mm/h, stage I is shortened, which means the increase of Cr content in this stage is more rapid and hence leading to lower concentration of Cr in stage III. With the increase of R, the peak value of Cr in stage III lowers. At R = 30 mm/h, the profile of Cr in UGZ is still similar to those at R < 30 mm/h. At R ≥ 60 mm/h, three stages are no longer apparent and data points of Cr seem to fluctuate around a nearly horizontal line. At 150 mm/h, the Cr profile changes into a wavy zigzag type and near the GFL, the fluctuating amplitude of data points enlarges.

The profiles of Mo are just opposite to their counterparts of Cr at R ≤ 30 mm/h (Fig. 8(b)). In UGZ, the content of Mo exhibits a slow decrease trend in an oscillating mode. Near GFL, a quick decrease is seen. At R ≥ 60 mm/h, Mo content begins to fluctuate around a horizontal line.

The profiles of Zr are similar to its counterparts of Cr (Fig. 8(c)). In UGZ, the profiles of Zr exhibits a slow increasing trend. Near GFL, a quick increase is seen.

At R ≥ 60 mm/h, both Mo and Zr contents seem to fluctuate around a horizontal line, but the fluctuating amplitude of Zr is narrower than Mo counterpart. Both of them are much narrower than Cr.

In directional solidification, the situation where the heat flow is opposite to the growth direction is usually referred to as the constrained growth. The proceeding rate of the isotherms constrains the front of solidification to grow at a given rate [32,33]. When the constant temperature gradient is greater than zero (G > 0), R is the main factor determining the morphology of the solid-liquid interface. Because of the rapid separation of the feed rod and the seed rod, the contour of the solid-liquid interface can be preserved as shown in Fig. 5. Based on the results, a schematic of the evolution of solid-liquid interface is illustrated in Fig. 10. The morphology of solid-liquid interface changes in the sequence of planar, cellular and dendritic types with the increase of R. Based on the thermophysical property values of Cr in Ti-17, there are two critical rates, i.e. planar to cellular transition rate (R_c1) and cellular to dendritic transition rate (R_c2). On the basis of the MS criteria of interfacial stability [30,34], R_c1 can be simplified as Eq. (2),

where m is the liquidus slope as 8.4 [28] and ΔC_o is the mass concentration difference of liquid and solid as ΔC_o=$C_{ l }^{*} - C_{ s }^{*}$=C_o(1-k)/k. From Eq. (2), R_c1 is estimated to be about 22 mm/h. According to the theory of Kurz-Fisher [35], the approximate cellular to dendritic transition rate (R_c2) can be expressed as Eq. (3).

Based on the Eq. (3), R_c2 is estimated to be about 32 mm/h. In this study, at R = 6 and 15 mm/h, morphological instability of planar interface can be observed, as an initial cellular interface in Fig. 5(c). From R = 30-60 mm/h, the cellular to dendritic transition is observed, as a cellular interface in Fig. 5(d) and a dendritic interface in Fig. 5(e). Calculations fit well with the experimental results.

At R ≥ 60 mm/h, dendritic growth prevails. From Fig. 5(e) and 5(f), with the increase of R, primary dendrite arm spacing (λ) decreases from about 120 μm to 100 μm. The λ is still far larger than those tested in industrial castings and reported in literatures [36,37]. Hence it can be inferred that at R = 60 and 150 mm/h, the formation of dendritic structure is still in its early stage and the large λ results in relatively large interdendritic region with high solute enrichment in the mushy zone [24,38]. At R = 150 mm/h, the interdendritic region evolves into a channel-like structure with not well-developed second dendrite arms as shown in Fig. 5(f). Growing dendrite would bring more solute into the interdendritic region and the accumulated solute cannot effectively diffuse into the liquid without additional driving forces [12]. The enrichment of solute would result in the restraining growth of second dendrite arms and channel-like structure forms in the final microstructure.

Ti-17 is a typical multi-component alloy. Except the easy-segregating element Cr, the Mo and Zr also show a high tendency of redistribution at the solid-liquid interface. When R is relatively low, as the solid-liquid interface evolving from planar to cellular structure, solute elements with k=C_s/C_l<1 as Cr and Zr concentrate in liquid near the solid-liquid interface, while elements with k=C_s/C_l>1 as Mo concentrate in solid phase [28]. With the advance of solid-liquid interface and the increase of solid phase, the tendency of enrichment of k<1 elements and deficiency of k<1 elements in liquid phase become more pronounced, leading to an increasing trend of k<1 elements and a decreasing trend of k>1 elements from GSL to UGZ until reaching a stable stage as C_s=C₀. With the increase of R, for element Cr, the profiles of elemental distributions from GSL across UGZ to RSZ transits from a smooth increasing→ stable→ sharp increasing trend to a zigzag profile when cellular or dendritic solidification prevails. The zigzag profile is corresponding to the fact that the cellular or dendritic growth is not straight upwards and the data collected in cell arms or dendrite arms exhibit higher Mo and lower Cr and Zr. And the data collected in the intercellular or interdendritic regions exhibits lower Mo but higher Cr and Zr. Hence a zigzag profile of elemental distribution is naturally expected. It can be deduced that at higher R (R > 150 mm/h), the small λ can be expected, which may result in a weaker interdendritic segregation and a more smooth elemental distribution profile is anticipated.

The microstructural morphology is determined by elemental distribution. A seemingly homogeneous microstructure was found under as-solidified condition at R = 150 mm/h (Fig. 2). After the heat treatment at 910 °C/0.5 h/FC+875 °C/1 h/WQ, microstructure heterogeneity is revealed. From Fig. 6 and Fig. 8, we can see that the variation of elemental distribution between cell arms and intercellular regions (or dendrite arms and interdendritic regions) was not prominent. Since elemental variation may affect the local T_β, this effect can be used to clearly detect the correlation between elemental variation and microstructure heterogeneity by proper tailoring of α phase. At β stabilizing element rich areas, T_β is lower than that at β stabilizing element lean areas. Selecting proper heat treatment temperature causes a significant difference of amount of α phase in the two areas (Fig. 4). Areas with few α phase exhibit dark contrast while areas with large amount of α phase exhibit light contrast. Hence relationship between weak elemental segregation and metallurgical features is clearly seen. In other words, microstructure variations induced by weak elemental segregation can be manifested by proper heat treatment.

β-flecks have been found in many titanium alloys with high content of Cr or Fe such as Ti-1023 and Ti-17. Due to the differences of melting methods, solidification conditions and thermomechanical processes, β-flecks show different shapes and sizes in castings or forgings [6,15]. The β-fleck is considered as a segregation defect, arising during solidification process. By tailoring α phase amount after the certain heat treatment, all β-flecks are characterized by dark contrast under OM [15], similar to the dark areas in this study. It can be concluded that enrichment of β stabilizing elements with k<1 is the main criteria for determination of β-flecks. For Ti-17, β-flecks show an increase compared with the normal areas of 1.0-1.5 wt% Cr and 0.5 wt% Zr but a decrease of 0.5 wt% Mo and 0.2 wt% Al [5]. Obviously, the dark areas found in this study are in accordance with all characteristics of β-flecks regardless of shape and size. Since the macrostructure and distribution of dark areas is determined by melting process and solidification conditions, we can conclude that the dark areas found in the present study are β-flecks in nature. Thus, based on the above study for the dark areas, the formation mechanism of β-flecks can be reasonably understood and inferred.

A 220 mm diameter Ti-17 ingot was produced by VAR process without electromagnetic stirring, and then heat treated at the same scheme of 875 °C/1 h/WQ. Two kinds of typical β-flecks were found in the ingot as shown in Fig. 11. Morphology and composition of the β-flecks are similar to the dark areas in this study. According to the shape and size, β-flecks in titanium alloys can be roughly summarized into four types: (i) Large size ones induced by macro-segregation as indicated in Fig. 3 at R ≤ 60 mm/h which is characterized by distinct composition difference between UGZ and RSZ; (ii) Channel-like ones induced by interdendritic segregation as indicated in Fig. 3 at R > 60 mm/h which is characterized by apparent dendritic structure, which is similar to those found at the top of industrial ingots melted by VAR process, (Fig. 11(a)); (iii) irregular spot-like β-flecks located in equiaxed zones in the central region or the columnar-equiaxed transition (CET) region of industrial ingots (Fig. 11(b)); (iv) spherical or ellipsoidal β-flecks which are frequently found in semi- and final forging products of Ti-17 or Ti-1023 alloys [6,39]. Formation mechanisms of each kind of β-flecks can be summarized as follows.

(i) Large size β-flecks induced by macro-segregation emerge only at R < 60 mm/h as illustrated in this study. Extremely low R promotes sufficient diffusion of solute elements through solid-liquid interface under a near thermodynamic equilibrium condition, resulting in a planar solid-liquid interface. As shown in Fig. 8, when R is less than 60 mm/h, significant solute segregation exists in the final solidified liquid. And at two sides of the interface, solid solution with different contrast are apparently seen in Fig. 3. The macro-segregation behavior results in the large size β-flecks in solidified samples. This phenomenon is hardly seen in real industrial ingots since in practical melting process, the cooling rate is much higher than 60 mm/h and such macro-segregation is not possible to occur.

(ii) Channel-like β-flecks produced by interdendritic segregation emerge at moderate R around 150 mm/h as illustrated in this study. When the solid-liquid interface evolves into dendritic structure, large λ accommodates a high volume of liquid enriched by k<1 elements in interdendritic regions. The interdendritic segregation behavior leads to the channel-like β-flecks. β-flecks with similar morphology are found on the top of VAR ingots near the macro-shrinkage cavity (Fig. 11(a)). This kind of β-flecks is also hardly found in industrial products due to removal of the riser before mechanical deformation imposed to ingots.

(iii) Irregular spot-like β-flecks are widely found in equiaxed zones of industrial VAR ingots of certain titanium alloys. During solidification process, temperature gradient in the central region of the VAR ingot is shallow. Crystals nucleate and grow uniformly in this region and large equiaxed crystals usually form at relatively low solidification rate. During this process, elements with k<1 keep enriching in liquid and crystals tend to pile up due to sinking of crystals owing to higher density of crystals than matrix liquid [40]. At the end of solidification, the residual liquid left in the interstices of the piled crystals transit to irregular spot-like β-flecks. Such β-flecks are supposed to occur during solidification with a deep metal liquid pool when relatively high melting rates are used [41]. Additionally, the irregular spot-like β-flecks cannot emerge during the directional solidification process. Further experimental results of the β-flecks in VAR ingots are needed to clarify the formation mechanism.

Spherical or ellipsoidal β-flecks are often detected in semi- or final products of titanium alloys like Ti-1023 or Ti-17, which experience the forging deformation from ingots to final products and hence cannot be regarded as one type of β-flecks.

It should be noted that in VAR melting practice of titanium alloys, electromagnetic stirring is usually employed to homogenize microstructure and composition. Under this circumstance, the solidification behavior of VAR metal pool is much more complicated. Distribution and shape of β-flecks is inevitably affected by electromagnetic stirring. The relevant investigations will be presented in another article.

Ti-17 alloys were melted and directionally solidified at a constant temperature gradient (G = 2 × 10⁴ K/m) at different withdrawal rates (R = 3-150 mm/h). The aim of the present investigation is to make clear the formation mechanism of β-flecks in a controllable manner. The conclusions are as follows:

(1)For directionally solidified Ti-17 alloy, macrostructure characterized by “four zones and two lines” can be only observed after the heat treatment. The effects of various solidification rates on microstructure heterogeneity and elemental segregation are visualized on the evolution of “four zones and two lines”.

(2)Solidification rate increasing from 3 mm/h to 150 mm/h, shape of solid-liquid interface transits from planar through cellular to dendritic structure. Critical rate at which the planar interface collapses completely is about 15 mm/h and transition rate from cell to dendrite is about 30 mm/h. The critical rates can be well predicted based on the solidification theory.

(3)Elemental segregation is clearly visualized by areas with dark and light contrasts in macrostructure. Dark contrast areas are rich of Cr, Zr but lean of Mo. With the increase of R from 3 mm/h to 150 mm/h, macro-segregation transits to local segregation.

(4)β-flecks found in Ti-17 titanium alloy are induced by segregation of alloying elements during solidification process. Areas enriched by β stabilizing elements with k<1 and exhibiting dark contrast at low magnification under OM are the main characteristic and criteria for determination of β-flecks. On this basis, three types of β-flecks are proposed and their formation mechanisms are discussed.