Microscopy of heat treated titanium alloy BT16

Material and heat treatment parameters

The material studied in the present paper is BT16 titanium alloy, produced in Russia for cold deformation processing of fasteners used in airplanes. The as received state was in the form of a 8 mm diameter rod, in hot rolled and annealed condition. The rod had a relatively homogeneous microstructure after hot rolling and annealing. The compositions of the BT16 alloy are 3Al, 4·6Mo, 4·9V, 0·01C, 0·17Fe, 0·04Si, 0·30Zr, 0·14O, 0·011N, 0·01H, ⩽0·30 others and Ti balance (wt-).20^–22 The alloy thermal history was proprietary to its manufacturer and was not declared. The microstructure of the as received alloy will be shown and statistically analysed later in the present paper.

The heat treatment of the rod material was carried out in a 4000W SX2-4-10 resistance furnace. The furnace temperature was calibrated using a standard NiCr–NiAl thermocouple, to an accuracy of ±5°C.

The heat treatment process involved the following steps. The alloy was inserted into the furnace preheated to the desired temperature and kept for a certain period, followed by different ways of cooling. Water quenching was used after solution treatment, which involved 2 h at 780, 800, 840, 860, 870 and 880°C. After annealing treatment, cooling involved furnace cooling at a rate of 2–4°C min⁻¹ to 550°C, followed by air cooling. Two groups of annealing treatment were used. The first group involved a constant annealing time of 2 h, at 550, 650, 700, 780 and 800°C. The second group involved a constant annealing temperature of 780°C, for 0·5, 1, 2, 3 and 4 h. Table 1 shows the heat treatment combinations.

OPEN IN VIEWER

Table 1

Heat treatment of BT16 alloy

Treatment	Temperature/°C	Isothermal time/h	Cooling process
Solution	780	2	Water quenching
	800	2	Water quenching
	840	2	Water quenching
	860	2	Water quenching
	870	2	Water quenching
	880	2	Water quenching
Annealing	550	2	Air cooling
	650	2	2–4°C min−1 to 550°C, air cooling
	700	2	2–4°C min−1 to 550°C, air cooling
	780	2	2–4°C min−1 to 550°C, air cooling
	800	2	2–4°C min−1 to 550°C, air cooling
	780	0·5	2–4°C min−1 to 550°C, air cooling
	780	1	2–4°C min−1 to 550°C, air cooling
	780	3	2–4°C min−1 to 550°C, air cooling
	780	4	2–4°C min−1 to 550°C, air cooling

Microstructural examination

After heat treating the specimens, the oxide layer on the surface was removed by grinding. Coarse and then fine grinding using sand papers followed, until the specimen surface was shiny. Then the specimens were electropolished. The electrolyte was 60 ml perchloric acid, 350 ml ethane-1,2-diol and 390 ml methanol. The process parameters were ∼30 V, ∼1 A, for ∼30 s.

The etching solution for the BT16 titanium alloy consisted of 100 ml H₂O, 4 ml HNO₃ and 2 ml HF. An XJP-3C optical microscope was used to examine alloy microstructure after heat treatment.

The procedure for transmission electron microscopy (TEM) and selected area electron diffraction of the as received specimens was as follows. Slices of 0·3 mm thickness were cut using an electrical discharge machine with molybdenum wires. This was mechanically thinned to 50 μm, and then ion beam thinned in vacuum using GL-6960 equipment, at a voltage of 6 kV, current of 0·4 mA and vacuum level of 10⁻⁵ Pa or better. Specimens of TEM were obtained in this way. The as received state of the BT16 titanium alloy was examined with a Hitachi H-800 TEM, using bright field imaging and selected area electron diffraction.

Procedures used for analysing microstructures

The statistical method for analysing microstructural parameters of α plates in the BT16 titanium alloy is as follows. In each heat treatment condition under examination, five metallographic photographs were taken at about ×1000 magnification, along the axial direction of the BT16 titanium alloy rod (Fig. 1). In each micrograph, a circle having a diameter of 50 μm was drawn, with the centre randomly selected (Fig. 2). The microstructural parameters of all the α plates intercepting the circumference are statistically collected, including length, width, and the angle between the length direction and the horizontal direction of the picture (Fig. 3). In each heat treatment condition, the number of α plates analysed is more than 800. The α grain size was statistically calculated by dividing the circle (Fig. 2) area by the number of α grains inside the circle. The latter value equalled the sum of α grains completely inside the circle and half of the α grains intersecting the circle.

OPEN IN VIEWER

Figure 1

Five photo locations for taking micrographs for statistical processing of microstructural parameters: 10 such 10 mm length samples were cut from as received 100 mm length rod

OPEN IN VIEWER

Figure 2

Method of choosing grains for statistical processing of microstructural parameters

OPEN IN VIEWER

Figure 3

Schematic diagram showing measurement of α plate microstructural parameters

As received microstructure

Figure 4 shows the typical microstructure of the as received BT16 titanium alloy rod, on a longitudinal section. The microstructure consists of a mixture of β phase (note: not transformed β phase) and fine plate shaped α phase. The average length of α plates is ∼4·4 μm, width ∼1 μm and aspect ratio ∼5·3. The aspect ratio is the ratio of the length to width. However, the average aspect ratio here (5·3) is the average of the aspect ratio values measured on individual plates, and not the ratio of the average length (4·4 μm) and the average width (1 μm). This is why it deviates from 4·4/1, but not very significantly. The β fraction is ∼33. This was measured on optical micrographs, and not using TEM. Scanning electron microscopy is unnecessary for this purpose. Five measurements were made, but no error data were kept after the experimental work. The error among the five measurements was small. The β fraction was determined by the alloy composition, hot working process and so on. In the present paper, the value would not be changed by heat treatment. Thirty-three per cent is an average value. The as received hardness is 30 HRC. The α+β two-phase state, which the BT16 alloy is in, is common in nearly all titanium alloys, a range of which is discussed in Ref. 12. The difference is the detailed α morphology, shown here for this particular alloy not previously well characterised.

OPEN IN VIEWER

Figure 4

Typical microstructure of BT16 titanium alloy for cold deformation

Selected area electron diffraction results show that in the as received state, only the hcp α phase and the bcc β phase are in the BT16 titanium alloy (Fig. 5). The β phase is distributed around the α phase.

OPEN IN VIEWER

Figure 5

a plate morphology microstructure and b electron diffraction pattern of BT16 alloy rods for cold deformation

Effects of solution treatment parameters on microstructure

In order to determine the phase transformation temperature of the BT16 titanium alloy, systematic solution treatment on the alloy was carried out. The results show that there are two types of effects of the solution temperature on the metallographic phase formation and morphology in the BT16 titanium alloy, depending on the solution treatment temperature range, at 880°C, or 870°C or lower. Differential scanning calorimetry was not used.

At a solution temperature of 880°C or higher, the quenched microstructure is mainly martensitic (Fig. 6). This is because, before quenching, the alloy was in the single β phase field. Quenching induces the β to α″+β_residual transformation. The high temperature β phase has transformed to coarse α″ phase.

OPEN IN VIEWER

Figure 6

Metallography of BT16 titanium alloy after 2 h at 880°C and water quenching

When the solution temperature was 870°C or lower, the quenched microstructure is a mixture of α″ martensite and α phase (Fig. 7). Because the alloy was quenched from the α+β two-phase field, there is also β to α″+β_residual transformation at quenching. However, as the transformation happened in the two-phase field, there was already a certain amount of α existing before quenching. Thus, the final microstructure is an α+α″+β_residual mixed structure. With increasing heating temperature and gradually decreasing the amount of α phase, the amount of α″ martensite continues to increase (Fig. 7).

OPEN IN VIEWER

Figure 7

Metallography of BT16 titanium alloy quenched from two-phase field of a 870°C and b 840°C: specimens were held at 870 or 840°C for 2 h before quenching; α phase appears as numerous small dots in a, as it is in shape of particles; its quantity increases in b

Comparing Figs. 6 and 7a, there is no α phase in the microstructure of BT16 titanium alloy quenched at 880°C, but there is a small amount of α phase particles in the alloy microstructure after quenching at 870°C. This is the standard way of measuring the phase transformation temperature of titanium alloys,12 which in this case is determined as 875±5°C for the BT16. Such a β transus temperature may be compared with β transus temperatures of other commercial and experimental titanium alloys, discussed in detail in Ref. 12.

X-ray diffraction cannot detect phases in small quantities, at the level shown in Fig. 7a. If the amount of a phase in the material is less than ∼2, its diffraction peaks cannot be revealed to statistically significant levels. Differential scanning calorimetry, on the other hand, is not a good tool to measure isothermal transformation temperatures. This is because this technique measures phase transformation during continuous heating or cooling. Under continuous heating or cooling, there is always over heating or under cooling, before a transformation happens. Therefore, phase transformation temperatures measured by differential scanning calorimetry are not transformation temperatures under isothermal conditions, but the normal heat treatment of titanium alloys is almost always conducted isothermally.

Statistical analysis of microstructure

In order to better understand the growth process of the α plates during annealing, and provide information for possible future numerical simulation of the deformation process of the BT16 titanium alloy, statistical analysis is required of the probability distribution of microstructural parameters of the α plates in the BT16 titanium alloy under different annealing conditions. At present, two types of models for describing the microstructural evolution of materials during heat treatment are widely used. The first type is the grain growth model and its derivatives. 23 23,24 The second type uses a statistical approach, without considering grain growth mechanisms. 25 25,26 The first type is mainly for describing the steady state grain growth process, but the BT16 titanium alloy has phase transformation during annealing, so only the second type can be used here to describe the α grain growth phenomenon in the BT16 titanium alloy. As this is a statistical approach as against a physical one, it can be used on two-phase alloys.

Statistical analysis shows that, in the as received state as well as after different annealing process, the probability distribution function for all microstructural parameters of the α plates in the BT16 titanium alloy follows three-parameter Weibull distribution27 where k and γ are characteristic parameters of Weibull density function, x is a microstructural parameter of the α plates and x₀ is its minimum probability parameter. This is after normalisation, i.e. dividing the parameter values by the average value of that parameter under the processing condition.

Figures 8 and 9 show the statistical probability distributions and their numerical fitting results of α plate length, width and aspect ratio in as received state and after annealing for 2 h at 780°C respectively. In the horizontal coordinate of these diagrams, the data was normalised, i.e. the plate width was divided by the average value of plate width. The column chart data was fitted by minimising the sum of the squares of the errors (the difference between experimental data and fitting curve), i.e. the method of least squares. A good match can be seen between the measurement data and the numerical fitting curves. In both the as received and the annealed states, all three microstructural parameters of the α plates have non-symmetrical distribution, with the peak position to the left. This confirms the tendency of non-symmetric probability distribution of microstructural parameters in the BT16 titanium alloy. This tendency appears to be the strongest in the case of aspect ratio.

OPEN IN VIEWER

Figure 8

Distributions of α plate in as received state of BT16 titanium alloy

OPEN IN VIEWER

Figure 9

Distributions of α plate in BT16 titanium alloy after annealing for 2 h at 780°C

The microstructural probability distribution in the BT16 titanium alloy shows that, in terms of numerical values, for most plates, their microstructural parameters are smaller than the average. A small fraction of plates has such values higher or even much higher than the average. Such large plates may have deformation incompatibility with other plates, affecting the deformability of the alloy.

Figure 10 shows the effect of annealing temperature on the probability distribution of microstructural parameters of the α plates in the BT16 titanium alloy. As shown in Fig. 10a and b, when annealed at the same temperature, the probability distribution is very similar, for the length and the width of the α plates.

OPEN IN VIEWER

Figure 10

Effect of annealing temperature on distribution of microstructural parameters of BT16 titanium alloy: annealing time is 2 h

A normally distributed α plate size would be beneficial to facilitating a good deformation response, as against a strongly skewed Weibull distribution. Deformation occurs by coordinated movement and shape change of α plates, and good coordination will be easier to achieve when all α plates have similar sizes. In a strongly skewed Weibull distribution, however, the deformation or the movement of a few very large α plates could introduce cracking, which does not help the process. It is unclear at this stage how the distribution becomes more normal/Gaussian with increasing heat treatment temperature and how the heat treatment has removed some of the outliers that skew the data. The finding itself as shown in the measurement data, however, is significant and the study has practical implications.

It may be noted that the method used to determine the α plate dimensions is not a standard testing method (e.g. ASTM or ISO method). It is a new method just provided by the authors.20

Though not used for describing the probability distribution of microstructural parameters of titanium alloys, Weibull distribution has been used in relation to fatigue life properties of titanium alloys. For example, it was found that a simple Weibull distribution describes very well the number to failure statistics in a near-α titanium alloy IMI 834.28 In a separate work, statistical characteristics in simulated life distributions were investigated using Weibull distribution function.29 Other properties, such hardness, are investigated with Weibull statistics. For example, it was found that the nanostructured coating and the conventional coating on titanium alloy substrates exhibit a bimodal distribution of microhardness values, which are analysed via Weibull statistics.30