3. Tailoring of Thermophysical Properties of New TRIP/TWIP Steel Alloys to Optimize Gas Atomization

Inert gas atomization is one of the variants of a metal powder production. Its main advantages are the sphericity, surface cleanness and controlled chemical composition of the final powder. Among disadvantages are relative complexity of the equipment and high cost of the obtained metal powders. Currently, inert gas atomization technology is the primary source of high-quality powders for the emerging metal 3d-printing industry. The principle of the gas atomization is based on the disintegration of the liquid metal stream by the gas jet/jets which is well described in the literature [11, 12]. In the present project the inert gas atomization unit VIGA-1B (ALD-Vacuum Technologies) was used. Principal scheme of the atomizer is given in Fig. 3.13. Sample of metal is inductively heated and melted in a 1 litre tundish located in the upper chamber of the atomizer. Upper chamber can provide vacuum, argon or nitrogen atmosphere. When the liquid is heated to a target temperature, the stopper rod is lifted and metal can flow thru a zirconia nozzle out of the tundish. A stream of metal meets super-sonic close-coupled confined gas jet, streaming out of the steel ring nozzle. Sprayed liquid droplets rapidly solidify while falling into the bottom of the atomization vessel. After cooling of the unit, a fine powder is collected from the powder can (trap).

For the atomization of metals argon gas was used in VIGA-1B. Gas was supplied from the liquid storage tank and was re-gasified before injection into the ring nozzle. Gas pressure was 26 bar. In all the experiments the gas flow rate was kept constant and was equal to the technical maximum for the atomization unit.

The main target was to investigate the effect of thermophysical properties of the liquid alloys on the inert gas atomization. One of the easiest characteristics for modification of the molten steel is surface tension. To date, a vast amount of literature data is available on the surface-active elements, which can be applied for steel. Sulphur is known for its strong reducing effect on a surface tension of iron and iron-based alloys [10, 15]. Therefore, a series of alloys based on 16-7-9 steel was prepared with variable sulphur content in the range of 114–984 mass ppm S (Table 3.6). The surface tension of the alloyed steels was measured with the application of MBP-technique. Figure 3.11 shows that surface tension temperature curves lay substantially lower with the growth of sulphur content in the alloy. As a result, the surface tension of the 16-7-9 alloy at the target atomization temperature of 1600 °C dropped from 1250 to 1050 mN m⁻¹ what was a 16% reduction.

Further these sulphur-alloyed steels together with 304 steel (which has much higher surface tension due to lower Mn-content) were atomized at a constant gas pressure and mass flow rate. Analysis of the median particle size (d₅₀) of the obtained powders (Fig. 3.14) showed a clear and strong dependence of d₅₀ on the surface tension of the atomized metal.

Thus, sulphur alloying has shown a remarkable effect on the atomization. Sulphur itself is known as one of the least desirable trump element for most of the steel applications. With this regard, it was decided to check the possibility of d₅₀ reduction with other surface active elements. As was given above, selenium showed further tremendous reduction of the surface tension of steel alloy with already low surface tension (Figs. 3.14 and 3.15a).

Results of atomizations of Se-alloyed steels showed the same trend on the reduction of median particle size due to a surface tension reduction (Fig. 3.15b). However, data points had higher scatter compared to S-alloyed powders.

It is known from literature, that atomization process is affected by the viscosity of the atomized liquid. In order to investigate this phenomenon, it was necessary to alloy the atomized steels with viscosity modifier. One of the studies accomplished in the framework of CRC 799 has indicated improvement of flowability of the phosphorus-alloyed steel [16]. Moreover, addition of phosphorus did not affect the surface tension—another parameter influencing atomization. Therefore, phosphorus was an ideal alloying element to study the modification of viscosity while other melt parameters being fixed.

However, viscosity research required a high-precision viscosity measurement device. High precision was required in order to address even minor changes of such low-viscous liquids as molten Fe-alloys. It is known that steels have a viscosity only slightly higher than water. Beside that, it should be measured at temperatures well above 1450 °C. Additionally, alloys of interest contained 7% of Mn, what posed additional challenge to selection of materials for measurement instrument. All the above mentioned topics illustrates that the development of the viscosity measurement device was not a trivial task.

An entirely new device capable of low viscosity measurements was developed and patented [17]. Figure 3.16 shows the scheme of the unit and its’ measurement head. Vibrating finger viscometer has a unique resonance oscillating head with the amplitude controlled by a laser micrometer. The amplitude is provided by the neodymium magnet excited by a field coil. As soon as a vibrating finger, attached to the oscillating head, touches the liquid the current in the field coil rumps-up in order to keep stable amplitude. For the measurement of viscosity, the vibrating finger is immersed into the liquid on 20 mm depth. The square root of the product of density and viscosity is a function of relative current change in this case:

In order to define the viscosity of the measurement a cell is calibrated at room temperature. Calibration is accomplished with the reference liquids of known viscosity (silicon oils) in the range of 0.7–200 mPa s. As a result, the calibration line is obtained (Fig. 3.17). As can be seen from (3.5), for the estimation of viscosity, the density of the liquid is required.

An essential aspect of a new cell development was the selection of the material for vibrating finger. The requirements to this material were as follows: easy machinability in order to create complex geometry (screw-thread), high-temperature corrosion resistance to withstand contact with liquid metals and alloying elements, relatively low cost. All these requirements were met in a boron nitride ceramic which was further used as a material for vibrating finger.

The new measurement cell was tested on the metals with vast reference data on their viscosity. The vibrating finger was immersed into liquid gold, silver and tin. Density data on these metals were taken from the previous study [20]. Viscosity values obtained by the newly developed unit are in a very good agreement with the literature values. A viscosity of the liquid gold measured with a vibrating finger viscometer was ~ 0% higher than previously reported in literature and decreased from 5.6 mPa s at 1100 °C (1373 K) to 4.0 mPa s at 1400 °C (1673 K). A viscosity of liquid silver was 10–20% lower than in available literature and decreased from 3.4 mPa s at 1000 °C (1273 K) to 2.3 mPa s 1400 °C (1673 K). Liquid tin showed viscosity ~20% lower than reported in literature and reducing from 1.3 mPa s at 600 °C (873 K) to 1.1 mPa s at 800 °C (1073 K). This research has proven the ability of new viscometer to measure the viscosity of low reactive metals at elevated temperatures.

After a successful measurement of noble metals’ viscosity, the next step was to apply the vibrating finger viscometer for the study of the viscosity of the liquid steel alloys. Research cell was used to investigate the effect of nickel content on the viscosity of the Cr–Mn–Ni steel alloys in the temperatures range of 1500–1600 °C [18]. The measurements were accomplished without severe erosion of vibrating finger surface and geometry changes. As can be seen from Figs. 3.18 and 3.19, the viscosities of CrMnNi steels decrease with temperature and significantly reduce at 20% of Ni.

In order to investigate the effect of melts’ viscosity on the atomization process, a series of phosphorus alloyed steels were prepared (Table 3.5). It is known from literature that phosphorus significantly reduces the viscosity of liquid iron and iron-based alloys [21]. Application of a vibrating finger viscometer has confirmed a decrease of viscosity due to additions of phosphorus. Increase of phosphorus content from 0.022 to 0.2% led to a reduction of viscosity from 5.0 mPa s to 3.2 mPa s measured at 1600 °C [8].

Atomization of the phosphorus alloyed steels showed a tremendous effect of viscosity on the size of the powder particles [22]. As it can be seen in Fig. 3.20, there is a linear dependence of the median particle size on the dynamic viscosity at 1600 °C. Decrease of viscosity of 36% (from 5.0 to 3.2 mPa s) led to reduction of d₅₀ of 34% (from 39.5 to 26 μm).

One of the further targets was to develop a more precise density measurement cell based on the Archimedean principle. Why was the Archimedean principle selected to measure density? There are two features of the density measurement with the Archimedean principle to be selected:

There are several versions of the direct Archimedean principle methods to measure the density of liquids. In a classical single-sinker method one utilizes one body of a known volume which is immersed in the liquid. The body is suspended on a wire or attached to a thin rod. Presence of the rod has an almost negligible effect on the error of the measurement when single sinker method is applied for measurement of low surface tension liquid, e.g. water or oils. However, when the metals are measured, the effect of a surface tension on the rod has to be accounted for. Surface tension effect depends on the wetting angle between the liquid and the rod material and also on the capillary meniscus direction. These two factors are the primary source of errors for the estimation of liquid metals density with the use of the single-sinker method.

In order to overcome this effect of surface tension, a variety of modifications are proposed in literature [24]. They are based on the idea to use two bodies with different volume but the same rod diameter. When bodies are immersed, the effect of surface tension on their rods is supposed to be the same. However, the weight change difference between the two different volumes indicates the density of the measured liquids (3.6). Therefore, the two-sinker method was selected for the measurement of the liquid iron alloys density:where W₁ is the weight of the big sinker immersed in liquid, W₀ is the weight of big sinker before immersion; w₁ is the weight of small sinker after immersion in liquid; w₀ is the weight of small sinker before immersion; V₁ is a volume of a big sinker, V₂ is a volume of a small sinker.

An additional measurement cell was added to existing MBP and vibrating finger viscometer facility in such a way that all three measurement units are utilizing the same induction furnace, positioning system and computer hardware. Scheme of the new unit is shown in the Fig. 3.21a. As it can be seen on the scheme, measurement of a density is accomplished via a high-precision balance (1) with the sinker assembly attached to its bottom hook. The assembly consists of molybdenum rod (2) connected to the ceramic rod with the sinker at the end (5). Additional weight (3) can be loaded on the assembly to keep it in vertical direction. The furnace has an extra thermocouple which measures temperature near the melt surface (10). The thermal field in the crucible was checked by immersion of a shielded thermocouple in the molten Ni and also in solid Mo. It was estimated that temperature in the melt was 8–10 K lower than the one measured by the side thermocouple. Therefore, side thermocouple is used as a sensor of melt temperature during experiments.

The procedure of measurement consists of:

Step 4 was added after the preliminary test in liquid lead that showed significant improvement of accuracy due to reducing of a surface tension effect on the ceramic rod (Fig. 3.21b). All the sinkers used in this measurement cell before experiments were immersed into deoxidized water in order to estimate their volume (volume difference of pairs) at room temperature.

Copper, silver, and tin metals were selected for the first application of the measurement cell. The materials were melted in ZrO₂ crucibles (yttria-stabilized) and measured with ZrO₂ (MgO-stabilised) sinkers. As can be seen in the Fig. 3.26a, density of copper measured by Archimedean principle decreased from 8.06 g/cm³ at 1090 °C to 7.65 g/cm³ at 1520 °C. Measured density deviates not more than 1% from values recommended in literature [26]. Density of silver had slightly higher deviation from recommended values [27] (Fig. 3.22b). Measured density decreased from 9.44 g/cm³ at 978 °C to 8.94 g/cm³ at 1364 °C. As for liquid tin, obtained density (Fig. 3.23) deviated up to 3% from recommended values [26]. Possible source of the error in the measurement of tin was oxidation of the surface level at temperatures below 700 °C.

In general, the measurement cell showed a relatively good accuracy and is considered for further study of iron-based alloys. It was decided to test a two-sinker method on the investigation of the density of steel alloys 15NC10.15 and 19NC15.15.

Along the duration of the CRC 799 inew steels 15NC10.15 and 19NC15.15 (Table 3.7) were developed. Exceptional tensile strength features of these steels are provided by high nitrogen content. At the same no extensive information on the effect of nitrogen alloying of the melt on the atomization process was found in the literature. Therefore, it was decided to conduct a series of atomization of the new nitrogen-alloyed steels [28]. In this study two parameters were selected as variables—nitrogen content and the temperature of the atomized melt. In order to broaden the range of soluble nitrogen content in the steel 19NC15.15 it was alloyed with 0.65% of vanadium and was named as 19NC15.15+V. Atomizations were also accomplished at 1600 and 1650 °C for steels 15NC10.15 and 19NC15.15.

The amount of dissolved nitrogen in the liquid alloy was provided by adjusting argon/nitrogen ratio in the atmosphere in the upper vessel of VIGA-1B. A sampling at 1550 °C was done before atomization was launched. Unfortunately, not all samplings were successful. However, with the exception of 19NC30.15+V results showed relatively low nitrogen loss directly during atomization—in a range of 80–130 ppm. As for 19NC15.15+V, the significant difference between the sample and the powder nitrogen composition could be explained with the sampling procedure described in reference [28].

Series of atomization with variable nitrogen content did not reveal an apparent effect on the median particle size d₅₀. At the same time, another factor showed a strong influence on particles size. During the atomizations a frozen crown was formed on the face of the tundish nozzle (Fig. 3.24). The size of this crown was a defining factor for the metal flow rate. During big crown formation, the metal flow rate was reduced what led to lower spraying rate (higher gas/metal ratio). It is known that in industrial practice the gas/metal ratio is the main parameter to control powder size [11]. The results of these experiments were in very good agreement with the typical square root dependence of d₅₀ from gas/metal ratio (see Fig. 3.25).

Series of atomization conducted at 1650 °C showed a significant shift of the value d₅₀ below the square root trendline of 1600 °C, thus, illustrating the effect of temperature on refining the size of the particles. These results are in a good agreement with the existing literature on this topic [29].

The samples of two powders with maximum and minimum nitrogen content were analyzed with the use of magnetic saturation method. The results showed that powder with 765 ppm on N had 54% of ferrite, while sample with 2384 ppm N had only 17% of ferrite phase. Further analysis with electron backscatter diffraction of SEM confirmed a prevailing of austenite phase in high N-containing alloy (Figs. 3.26 and 3.27). Furthermore, particles smaller than 20 µm had more ferrite phase than bigger ones (20 < d < 200 µm). This indicated control of both cooling rate and nitrogen upon the austenite stabilization.

For measurement of the particle surface temperature a 2-color thermographic camera unit was developed [30]. The setup consists of two commercial CCD-cameras with installed dielectric optical filters, optical beam splitter and two lenses as object lenses. The components were selected with the requirements for being operational in the selected filter wavelengths. The maximal spectral emissive power of the studied liquid metal droplets at 1873 K is expected to be close to a wavelength of 1.5 µm. CCD camera sensors have their peak sensitivity typically at 0.5 µm. Their sensitivity at 0.9 µm is only a tenth of the maximum value. Therefore, the combined effect is that the recorded signal, i.e. intensity times sensitivity is low. In order to handle this shortcoming longer exposure times are used. This means that not the surface temperature of individual droplets is recorded, but a temporal average of the droplets that cross the area imaged onto each pixel. The cameras have a resolution of 1600 × 1200 pixels of 4.4 µm pixel size. Maximum frame rate is 16 fps. Images were obtained in an 8-bit greyscale. The optical dielectric bandpass filters have central transmission wavelengths at ʎ₁ = (850 ± 8) nm and ʎ₂ = (900 ± 8) nm. The set-up of the two-color pyrometer is shown in Fig. 3.28. The image was focused with the objective “Rollei-HTF Distagon 2.8/35”. The images were focused onto the CCD sensors through the beam splitter cube. It splitted the optical path for the two camera sensors.

Two-color pyrometer was calibrated using a high-temperature furnace and a sample of silicon carbide in a temperature range of 1273–1873 K. The temperature of the sample was measured with a type B thermocouple assuring ±1.5 K absolute measurement error. After calibration a particular correction term was obtained.

The newly developed two-color pyrometer was applied for the measurement of the metal stream temperature distribution at atomization in the VIGA-1B unit described previously in Fig. 3.13. The liquid steel (16-7-6 alloy) was heated ca. 200 K over its liquidus line to 1873 K. When the target temperature was reached, the atomization was launched. Figure 3.29 shows the results of measurements of the time-averaged temperature of the surface of the droplets. Due to the small depth of the field features of the camera, the pictures represent cross-sections of the spraying cone in the first half-second after the metal spraying initiation. The 0 mm point in the figure refers to the line of sight, which was 40 mm below the ceramic nozzle face. At 0.0 s the gas jet was not launched and only the metal stream is visible. At 0.1 s the disintegration of the metal flow began due to the gas stream. After 0.3 s the cone of sprayed metal was formed and later remained relatively stable. The primary breakup point was situated about 20 mm below the line of sight. Below that level a hollow cone of atomized melt with the length of approximately 50 mm was formed. At lower levels the hollow cone disintegrated on ligaments and droplets. Temperature iso-lines indicate that between 60 and 80 mm below the line of sight (nozzle metal) material was cooled down to its liquidus line. It corresponds to 100–120 mm distance from the nozzle face. This distance is about 30 times the diameter of the nozzle and is in a good agreement with literature data [31].

For the measurement of the velocity of the particles during metal atomization a special Particle Image Velocimetry (PIV) was applied (Fig. 3.30) [30]. The PIV measurement method is a non-contact optical method for measuring particle velocities in a 2D observation area within a particle flow. This area is defined by a thin light sheet generated by the passage of a laser pulse through a cylindrical lens. Two pulse laser cavities emit two laser pulses with an adjustable temporal separation. The pulse duration is on the order of 5 ns. The scattered light of the illuminated particles is picked up by a double-frame PIV camera arranged perpendicular to the light cross-section. Due to the time interval, the local positions of the particles differ in the double images. Special PIV evaluation software detects these positional differences using correlation algorithms within the small windows in the image area, which defines a uniform grid structure. From the position difference at the correlation maximum and the time interval, a particle velocity vector is then calculated and assigned to the respective window center point. The result is a 2D velocity distribution in the observation area. PIV measurements in steel atomizers differ from those in continuous flows due to other boundary conditions.

Usually tracer particles are supplied to the fluids to be examined, which emit scattered radiation when passing the laser light section detected by the camera. Through this controlled particle addition (seeding), the number density of the determined velocity vectors can be significantly increased that aim to measure the flow velocity of the gas. In the present case of steel atomization, the aim is to truly measure the velocity of ligaments and droplets that constitute the particles. The use of a narrow band filter, whose wavelength coincides with that of the PIV laser, effectively blocks the thermal radiation of the melt.

Due to its large particle sizes, the melt in the vicinity of the ceramic nozzle scatters light strongly which leads to over-exposure of adjacent pixels on the camera sensor (blooming effect) and images with overexposed areas. Therefore, the correlation evaluation of the double images gives the wrong velocity data in this region.

Based on numerical simulations [32], it is known that gas velocities in the range of 400–500 m/s occur near the atomizer nozzle. The particle velocities in the axial range were calculated to be up to 250 m/s. The measurement of such high speeds requires a correspondingly small time interval between the double pulses of the laser in the order of 2–3 µs. The entire measuring system must be triggered exactly in time in order to minimize the measuring error component of the time interval.

Measurements of the particle velocities were conducted for atomization of two steels: X5CrNi18-10 and X3CrNiMo13-4 in VIGA 1B unit. Both steel samples were atomized at the same conditions. Atomization was initiated at ca. 200 K superheating above the liquidus line at 1600 °C. Evolution in time of the measured particles velocities in the first 0.5 s after atomization launch is seen in the Figs. 3.31 and 3.32. The plots show a section through atomization cone. It can be seen that after 60 mm below the ceramic nozzle the metal stream breaks up and forms a hollow cone confirming the results published in [14]. The high speeds in the upper third of the observation window are due to the blooming effect. In the area of the interaction of the melt and atomizing gas, the measured particle velocities range between 120 and 210 m/s. The white dashed line indicates the area where the solidification temperatures are reached. The particle velocities of both melts in the solidification area are approximately 120 m/s. Obtained velocities are in a good quantitative agreement with numerically determined data, e.g. as [33]. Comparison of the averaged particles’ velocities of two steels revealed that in case of X3CrNiMo13-4 the velocities were smaller than for X5CrNi18-10.

Based on the results obtained in the previous chapters it became clear that there are three fundamental problems in the existing inert gas atomization techniques. The first problem is the very thick melt stream coming out of the nozzle. This is necessary to avoid freezing of the nozzle [34]. The result shows that the atomizing inert gas first breaks the stream into larger ligaments and sheets that have to be broken up in further steps downstream of the nozzle. The second problem is extremely low temperature of the gas when it contacts with the melt. The argon has expanded from an initial temperature around or slightly above room temperature to a velocity higher than Mach 1 up to 3 depending on the particular geometry. The result shows that the gas has a temperature around −50 °C for Mach 1 and down to −150 °C for higher Mach numbers. In principle, the high velocity of the gas is necessary for atomization but when it is that cold, it is counterproductive. This is the reason for the very high superheat of the melt needed for the VIGA set-up (200 K). Finally, the break-up of a thick stream in the centre of the flow with a surrounding gas can be changed. A thin layer of melt on the inner wall of a high-velocity nozzle for the gas is the closest the melt can be exposed to the gas. This is only possible if the gas is hot enough otherwise the nozzle would freeze. This rationale led to the development of a new set-up which has been patented [35]. The idea has been implemented first for the atomization of tin as a proof of principle. Figure 3.33a shows schematically the tin atomization set-up. Figure 3.33b shows the resulting particle size distribution measured by laser scattering. The peak of the distribution centers at 10 µm for the first experiments. This factor is 2–3 lower than the VIGA results. Of course, the one result is for tin, and the other is for steel. However, when comparing with literature results for tin using similar set-ups as the VIGA the factor 3 is still valid [35].