Static coarsening behaviour of lamellar microstructure in selective laser melted Ti-6Al-4V

1. Introduction

Titanium (Ti) alloys are broadly used in different applications from chemical to aerospace due to the lightweight, high strength, outstanding high temperature mechanical property, and superior corrosion resistance [1,2]. Ti-6Al-4V is the most widely used Ti alloy combining excellent properties and productivity [1,3].

Selective laser melting (SLM) is a powder bed fusion additive manufacturing process [4,5], in which laser beam travels across the powder bed at a speed of 10²-10³ mm/s and forms a small melt pool with width at orders of 10-10² μm [6]. The high laser power, fast laser scan speed and small laser spot size lead to high heating and cooling rates. Experiment and numerical studies reported extremely high cooling rate between 10⁶ and 10⁷ K/s in SLM [6,7]. Therefore, as-selective laser melted (as-SLMed) Ti-6Al-4V is characterized by a fine martensitic (α') microstructure with high residual stress [[8], [9], [10], [11], [12], [13], [14]]. Such a microstructure and high residual stress contribute to high strength but low ductility. As a result, a post-SLM heat treatment is generally required to achieve a desirable combination of mechanical properties [[15], [16], [17], [18]].

The applied post-SLM heat treatment temperature and holding time influence the microstructure evolution and resultant mechanical properties. For heat treated SLMed Ti-6Al-4V with a lamellar microstructure, it is well documented that the thickness of α lamellae has a direct impact on yield strength through Hall-Petch relationship in both tensile [13,19,20] and compression tests [11]. Hence, it is important to understand the lamellae coarsening behaviour of SLMed Ti-6Al-4V since the growth of α lamellae may result in a degradation of strength.

Previous studies investigated lamellae coarsening behaviour of Ti alloys extensively. Stefansson and Semiatin [21] revealed that lamellae coarsening behaviour can be described by a modified Lifshitz, Slyozov, and Wagner (LSW theory [22]) in static heat treatment of Ti-6Al-4V after compression [21,23]:

in which dt is the average lamellar width, Vm is the molar volume, γ is the α/β interface energy, Cβ and D are the solute equilibrium concentration (atomic fraction) and diffusion coefficient in β matrix respectively, R is the gas constant, T is the absolute temperature in Kelvin, K is constant of proportionality, t is time, and n is coarsening coefficient. When n is equal to 0.5, coarsening process is controlled by interface reaction; when n is equal to 0.33, coarsening process is controlled by bulk diffusion [24]; when n is equal to 0.2, diffusion is through dislocation pipes [25]. For work conducted on wrought Ti-6Al-4V with a lamellar microstructure, the coarsening coefficients were found to be 0.44 and 0.45 for heat treatment at 900 °C and 950 °C, respectively [21]. Another study [26] showed coarsening coefficients at 0.26 and 0.11 for 760 °C and 843 °C, respectively. As a result, it was concluded that coarsening process is dominated by interface reaction at higher temperatures (>850 °C) and bulk diffusion at lower temperatures (≤850 °C) [21].

However, Eq. (1) has not considered the initial average lamellae width d₀. In this study, another spatio-temporal power law is used to describe lamellae coarsening phenomena [[27], [28], [29]]:

$d_t^{1/n}$-$d_0^{1/n}$=K't (2)

By using Eq. (2), Xu et al. reported similar results in another α + β Ti alloy Ti-5Al-2Sn-2Zr-4Mo-4Cr (Ti-17) [28], the controlling mechanism changed from interface reaction at 860 °C to bulk diffusion at 820 °C.

All the above previous works investigated coarsening behaviour in wrought and/or cast Ti alloys. In this work, the coarsening behaviour is studied for SLMed Ti alloy and a larger range of temperatures are considered from 700 °C to 950 °C. In addition, this work seeks to establish a quantitative relation between the microstructure evolution and post-SLM heat treatment in α + β phase field to control the microstructure and property of SLMed Ti-6Al-4V. This will benefit the optimization of post-SLM heat treatment practice for Ti-6Al-4V in providing insights into microstructure predictions and obtaining desirable strength.

2. Experimental methods

Gas atomized pre-alloyed Ti-6Al-4V powder with a chemical composition (wt%) of 6.14Al-4.13V-0.008C-0.03N-0.001H-0.09Fe-bal.Ti was provided by Falcon Tech Co., Ltd. The SLM process was carried out in an EOS M280 machine at a 99.99% argon atmosphere. More information regarding the Ti-6Al-4V powder and SLM fabrication process can be found in a previous study [13]. According to ThermalCal calculation, Ti-6Al-4V used in this study has a β transus temperature at approximately 1005 °C. It is slightly higher than that of normal Ti-6Al-4V at 995 °C [11,15] due to the relatively high content of Al at 6.14 wt% and O at 1645 ppm. Sub-transus (β transus) post-SLM heat treatments were carried out at 700 °C, 800 °C, 900 °C, and 950 °C for various holding time to investigate the coarsening behaviour in α + β phase field. In order to mimic the actual heat treatment used in SLMed parts and avoid re-introducing residual stress, Furnace cool (FC) was applied to all conditions by leaving samples inside the furnace after the furnace was switched off. The FC rate was determined at the range of 2-3 °C/min based on thermocouple measurement.

Microstructure characterizations was conducted by SEM on horizontal cross sections of SLMed samples. All samples were ground down to 4000-P SiC abrasive sand paper, and then polished on a MD-Chem polishing plate by a mixture of H₂O₂ and oxide polishing suspension (OP-S) with a volume ratio at 1:5. An FEI Quanta 3D FEG microscope was employed to obtain the back-scattered electron (BSE) images. Based on BSE images, lamellar thickness (10 image analyses for each sample condition) measurements were carried out by ImageJ software.

Micro-hardness testing was carried out on 800 °C and 900 °C heat treated samples to investigate the relationship between lamellar thickness and hardness value. A Struers Duramin-A300 hardness tester was employed, and the loading condition was 2 kg for 10 s. For each heat treatment condition, hardness was measured by pressing ten indents in the centre of horizontally built cylinders.

3. Results and discussion

3.1. Microstructure evolution and coarsening mechanism in post-SLM heat treatment

Fig. 1 shows the as-built microstructure on horizontal cross-section of SLMed Ti-6Al-4V. The as-SLMed Ti-6Al-4V had a martensitic microstructure. Based on image analysis, the average lamellar thickness¹(1 The as-SLMed Ti-6Al-4V has an acicular martensitic microstructure, and the thickness of acicular martensite is regarded as initial lamellar width (d₀).) is approximately at 460 nm, which is slightly higher than the value reported at 240 nm for as-SLMed Ti-6Al-4V in Ref. [30]. This might be related to the different laser scan strategy used, a continuous laser exposure was used in this work and point scan was applied in [30]. Fig. 1(b) reveals that twins prevail in the as-deposited microstructure, and high magnification inset illustrates that martensite lamellae contains a high density of twins. This has been reported in previous studies [13,14,31], which is induced by the rapid cooling rate between 10⁶ and 10⁷ K/s in SLM [6,7]. During SLM, the rapid cooling and thermal cycling result in martensitic transformation and thermal stress, which lead to the activation of twinning [1,31,32].

View original graphic| Download| PPT slide

Fig. 1.   BSE images of as-SLMed Ti‒6Al‒4V at different magnifications (a, b); white arrow in (b) illustrates martensite lamellae containing twins, and associated inset is a high magnification image of twins inside lamellae.

After SLM process, samples were heat treated at various temperatures to facilitate martensite decomposition (α'→α + β) and the microstructure evolution. Fig. 2 shows the microstructure of SLMed samples heat treated at 700 °C for various holding time. A small volume of β particles appeared in the sample annealed at 700 °C for 1 h, indicating martensite decomposition started. This is similar to what has been observed in other studies [33] that martensite transforms to α + β phases in post-SLM heat treatment and/or HIP at temperatures below β transus. Another phenomenon observed as shown in Fig. 2 is that the dominion β phase morphology changes from particles to lath after a prolonged annealed time of 10 h. According to Fig. 2(a-c), twins are still observed in 1 h, 2 h, and 4 h annealed conditions at 700 °C. This is consistent with the observation reported in Ref. [13], which shows twins inside lamellae after a short holding time when heat treatment temperature is below 800 °C. After detailed BSE investigation, twins were not observed in sample annealed at 700 °C for 10 h. The growth of lamellae was slow at 700 °C, the average lamellar width was approximately 1 μm after 10 h.

View original graphic| Download| PPT slide

Fig. 2.   BSE images of post-SLM heat treatments at 700 °C for: (a) 1 h, (b) 2 h, (c) 4 h, and (d) 10 h. White arrows are twins inside lamella, and associated insets are high magnification BSE images. In all images, dark area is α phase and bright area is β phase.

At 800 °C (Fig. 3), the coarsening rate of lamellae was slightly faster than that at 700 °C. For a short anneal time of 0.5 h (Fig. 3(a)), β phase appeared as a mixture of particles and lamellae. While for anneal time longer than 1 h, β phase mainly appeared as lamellae form rather than particles. In contrast to the twins retained in samples after 4 h anneal at 700 °C, twins were not observed in sample after 2 h anneal at 800 °C (Fig. 3(c, d)). According to a previous study [13], they are retained twins in martensite formed during the cooling stage of SLM process. This indicated martensite decomposition rate is accelerated by increased temperatures as would be expected.

View original graphic| Download| PPT slide

Fig. 3.   BSE images of post-SLM heat treatments at 800 °C for: (a) 0.5 h, (b) 1 h, (c) 2 h, (d) 4 h, (e) 6 h, and (f) 12 h. White arrows are twins inside lamella, and associated insets are high magnification BSE images. In all images, dark area is α phase and bright area is β phase.

Compared to β particles observed in heat treated samples at 700 °C and 800 °C, β phase had a lamellar morphology at high temperatures (900 °C and 950 °C) even for a short holding time of 0.5 h as shown in Figs. 4(a) and 5 (a). Another difference between low and high temperatures was that twins were not present in samples heat treated at 900 °C (Fig. 4) and 950 °C (Fig. 5). The lamellae coarsening rate was higher at 900 °C than those at 700 °C and 800 °C. After 12 h heat treatment at 900 °C, the average lamellar width was approximately 3 μm. In addition, some equiaxed α grains were observed at prior β grain boundaries after 12 h heat treatment at 900 °C as indicated by white arrows in Fig. 4(f).

View original graphic| Download| PPT slide

Fig. 4.   BSE images of post-SLM heat treatments at 900 °C for: (a) 0.5 h, (b) 1 h, (c) 2 h, (d) 4 h, (e) 6 h, and (f) 12 h. White arrows are equiaxed α grains. In all images, dark area is α phase and bright area is β phase.

View original graphic| Download| PPT slide

Fig. 5.   BSE images of post-SLM heat treatments at 950 °C for: (a) 0.5 h, (b) 1 h, (c) 2 h, (d) 4 h. White arrows are equiaxed α grains. In all images, dark area is α phase and bright area is β phase.

When a heat treatment of 950 °C (which is close to the β transus of Ti-6Al-4V at 1005 °C based on ThermalCal calculation) was applied to SLMed Ti-6Al-4V, the lamellar width increased significantly with time as can be seen in Fig. 5. The lamellae growth rate was the highest at 950 °C among all heat treatments investigated in this study as would be anticipated. In addition, equiaxed α grains, denoted by the white arrows in Fig. 5(b) and (c), appeared in 1 h and 2 h heat treated conditions at 950 °C, and became more prevalent in 4 h situation (Fig. 5(d)). As this work focused on lamellae coarsening behavior, heat treatment time longer than 4 h was not investigated at 950 °C.

For thermomechanical processed Ti-6Al-4V, the globularization models has been extensively documented in literature [21,24]. There are generally two stages consisting i) boundary splitting in the initial stage, ii) lamellar termination migration after prolonged heat treatments at high temperatures (>14 h at 900 °C for thermomechanical processed Ti-6Al-4V [21]). The globularization phenomenon in laser additive manufactured Ti-6Al-4V has been reported and discussed in recently published works [34,35]. In thermomechanical processed Ti-6Al-4V with a bi-modal microstructure, deformation is required to store the energy (dislocations) for the following recrystallization. In SLM process, the rapid cooling rate at 10⁶-10⁷ K/s [6,7] results in a non-equilibrium martensitic microstructure containing a high density of dislocations and twins [13,32]. When SLMed Ti-6Al-4V is heat treated at high temperatures of 900 °C and 950 °C, the sub-transus heat treatment promotes dislocation slip, climb, and formation of polygonised dislocations [34]. This leads to sub-grain boundary formation within lamellae, following boundary splitting with β phase diffusion into boundaries, and final globularization. Therefore, the globularization process is mainly controlled by boundary splitting in the current study as only short time heat treatments investigated.

Fig. 6 displays the measured lamellar width in samples heat treated from 700 °C to 950 °C at different holding time. It revealed that lamellar thickness increased monotonically with heat treatment time for all four temperatures. At each temperature, the coarsening rate decreased with time based on the factor of the reduced slope of tangents. Thus, it can be observed that the static coarsening rate was faster at higher temperatures. The 700 °C and 800 °C conditions showed a similar lamellae growth trend with a low coarsening rate. Heat treatment at 950 °C provided the fastest lamellae growth in SLMed Ti-6Al-4V.

View original graphic| Download| PPT slide

Fig. 6.   Measured lamellar width of post-SLM heat treated Ti‒6Al‒4V at various temperatures with different holding times. Solid symbols were data points measured in this work, and the four hollow symbols at 700 °C/2 h, 800 °C/2 h, 800 °C/6 h, and 800 °C/12 h were data from a previous study [13].

Based on Eq. (2), the dt1/n-d01/n curves were plotted versus heat treating time at various temperatures for different n values. Fig. 7 reveals the lamellae coarsening kinetics for 700 °C and 800 °C heat treated conditions. These curves change from concave to convex shapes when increasing the coarsening coefficient n from 0.2 to 0.5. A linear relationship between dt1/n-d01/n and heat treatment time was obtained when n is approximately at 0.33.

View original graphic| Download| PPT slide

Fig. 7.   dt1/n-d01/n vs. heat treatment time for temperature at 700 °C and 800 °C; coarsening coefficient n = 0.2, 0.25, 0.33, and 0.5.

Fig. 8, Fig. 9 display the dt1/n-d01/n curves for 900 °C and 950 °C heat treated samples, respectively. For both temperatures, a similar concave to convex curves transition was observed with the increase in n, and a linear relationship is to be located in between. For 900 °C, the linear relationship between dt1/n-d01/n and heat treatment time will be obtained when n is at 0.37, which is between 0.33 and 0.4 according to Fig. 8(b-d). The coarsening coefficient is at 0.44 for 950 °C, which is between 0.4 and 0.5 (Fig. 9(b-d)).

View original graphic| Download| PPT slide

Fig. 8.   dt1/n-d01/n vs. heat treatment time for temperature at 900 °C; coarsening coefficient n = 0.2, 0.33, 0.37 and 0.4.

View original graphic| Download| PPT slide

Fig. 9.   dt1/n-d01/n vs. heat treatment time for temperature at 950 °C; coarsening coefficient n = 0.2, 0.4, 0.44 and 0.5.

As shown in Fig. 7, Fig. 8, Fig. 9, the coarsening coefficient for SLMed Ti-6Al-4V was found to be approximately 0.33, 0.33-0.4, 0.4-0.5 at 700-800 °C, 900 °C, and 950 °C, respectively. According to the above analysis, the controlling mechanism of coarsening behavior varied with heat treatment temperatures. Based on LSW theory described in the introduction, the coarsening process is controlled by bulk diffusion when annealing at low temperatures of 700-800 °C. Once heat treatment temperature increased to 900 °C and 950 °C, lamellae coarsening mechanism should be a combination of bulk diffusion and interface reaction. Specifically, bulk diffusion is still the dominant at 900 °C, as the coarsening coefficient is between 0.33 and 0.4. However coarsening coefficient is increased to a range from 0.4 to 0.5, which indicates the dominant contribution is interface reaction for the combined coarsening mechanism at 950 °C.

A previous study [26] reported that a reduced coarsening coefficient was observed with temperature due to the increased volume fraction of growth inhibitor - β phase. In another α + β Ti alloy work [25], the coarsening coefficient was found to be independent to the volume fraction of α and β phases, which increased with temperature at a constant volume fraction of α and β [25]. Base on the current work, although the high temperature stable phase β has an increased equilibrium volume fraction with temperature, the coarsening coefficient still increases with temperature. By considering the result in this work and the above previous studies, the coarsening mechanism is concluded to be bulk diffusion at 700-800 °C annealing, and a combination of bulk diffusion and interface reaction coarsening mechanism is determined for 900 °C (the dominion is bulk diffusion) and 950 °C (the dominion is interface reaction).

3.2. Hardness

As shown in Fig. 10a, the as-SLMed condition had the highest hardness at 381 HV. This is similar to the average hardness reported in Ref. [36] at 395 HV for SLMed Ti‒6Al‒4V, but higher than the value reported in Ref. [5] at 342 HV. This is related to the different HCP-Ti width; the as deposited microstructure has martensite width at 0.46 μm in this work, which is much lower than the lamellar thickness of 1.5 μm in Ref. [5]. Once the sample is heat treated, hardness went down with time. The 900 °C heat treated samples had a lower hardness than those annealed at 800 °C. For both temperatures, hardness values decreased with time. Fig. 10(b) reveals the relationship between the hardness and lamellar width (d). For heat treated samples at 800 °C and 900 °C, hardness had a linear relationship with d-1/2, which indicated that the hardness was governed by the lamellar width after heat treatment. However, the as-SLMed sample had a higher hardness (381 HV) than the extrapolation of the linear relationship (367 HV). This means in as-SLMed Ti-6Al-4V there is another contribution to hardness other than lamellar width strengthening. It is widely recognized by previous works that as-SLMed Ti-6Al-4V has a fine martensitic (α') microstructure [[8], [9], [10], [11], [12], [13], [14]]. α' martensite has a moderate hardening effect as interstitial elements introduce a small elastic distortion of the hexagonal lattice [1,37]. Therefore, the 4% higher hardness in as-SLMed sample is likely to be related to martensitic transformation and accompanying lattice distortion induced by the rapid cooling in SLM process. After martensite decomposition is facilitated by heat treatments at 800 °C and 900 °C, the hardness is mainly governed by lamellar width.

View original graphic| Download| PPT slide

Fig. 10.   (a) Hardness of As-SLMed and post-SLM heat treated samples at 800 °C and 900 °C with different holding time, (b) a Hall‒Petch relationship between hardness of post-SLM heat treated samples and measured α lath width in Fig. 6.

4. Conclusions

Post-SLM heat treatments were performed on as-SLMed martensitic Ti‒6Al‒4V samples to study the lamellae coarsening behaviour and hardness of heat treated samples. The following conclusions were drawn from this work.

(1) Globularized grains are observed in SLMed Ti-6Al-4V samples heat treated at 900 and 950 °C. The mechanism of globularization is boundary splitting for the short time sub-transus heat treatments of 900-950 °C. In addition, the required time to form equiaxed grains decreases with increasing temperature.

(2) Martensite decomposition rate increases with heat treatment temperature. A decomposed α + β microstructure can be obtained in less than 0.5 h at heat treatment temperature above 900 °C.

(3) The measured hardness of heat treated SLMed Ti‒6Al‒4V and lamellar width obey the Hall-Petch relationship. The higher hardness in as-SLMed sample than that predicted by the Hall-Petch relationship is attributed to the martensite hardening effect.

(4) The lamellae coarsening rate increases with temperature but decreases with holding time at each temperature.

(5) For SLMed Ti‒6Al‒4V, the static coarsening behaviour can be predicted by LSW theory. The coarsening mechanism of lamellar microstructure is bulk diffusion when annealing at 700-800 °C. The coarsening mechanism changes to a combined bulk diffusion and interface reaction when heat treatment temperature is increased to 900 °C and 950 °C. Although both 900 °C and 950 °C have the combined mechanism, bulk diffusion is dominant at 900 °C and interface reaction is dominant at 950 °C.

Acknowledgements

This research is funded by Monash Centre for Additive Manufacturing (MCAM) and Australia Research CouncilIH130100008 “Industrial Transformation Research Hub for Transforming Australia's Manufacturing Industry through High Value Additive Manufacturing”, and the National Natural Science Foundation of China (No. 51701124). The authors acknowledge use of facilities within the Monash Centre for Electron Microscopy. This research used equipment funded by Australian Research Council (No. LE0882821).

The authors have declared that no competing interests exist.