Abstract Inconel 738 is one of the widely used nickel-based superalloys in high-temperature applications

Inconel 738 is one of the widely used nickel-based superalloys in high-temperature applications, especially in land-based and aerospace gas turbine engines. This paper reports the feasibility of direct laser metal deposition (LMD) of Inconel 738. Cracks evolved during deposition at the substrate/deposit interface and within the deposit along high angle grain boundary for scanning speed of 6 and 12 mm/s due to the intense residual stress and incipient melting. Results showed liquation cracking due to low melting crack boundary ?? and significant micro-segregation of Al and Ti along the crack boundaries. By maximizing the energy density and by reducing the scanning speed to 3 mm/s, crack-free single wall specimens were successfully manufactured. Microstructural evolution of primary, secondary, grain boundary ??, MC carbides and MB2 borides in the as-deposited and heat-treat specimens are discussed. Mechanical properties and microstructural development were investigated using tensile testing and scanning electron microscopy. Energy dispersive spectroscopy confirmed significant micro-segregation on various elements along the inter dendrite and grain boundaries. X-ray diffraction validated the presence of the observed carbides and borides in the as-deposited and heat-treated samples.

Keywords: Additive Manufacturing; Inconel 738; Laser Metal Deposition; Microstructure; Gamma/gamma-prime superalloy

1. Introduction
Inconel 738 is a nickel-based precipitation hardening superalloy, which is one of the materials being introduced as a blade material used in high-pressure stage of land-based and aero gas turbines. Inconel 738 found elevated temperature applications up to 980°C due to its superior high-temperature mechanical property such as creep rupture strength and improved hot corrosion resistance 1. Like many other nickel-based superalloys, IN 738 is primarily hardened by the ordered, coherent precipitation of gamma prime (??) consisting of Ni3(Al,Ti) intermetallic compound in the disordered (?) matrix. Solid solution strengthening is derived from Cr, Mo, W and Co, and grain boundary strengthening by carbides and grain boundary ?? 2-5. Ni-based superalloys have the potential to form several different precipitates like ??, grain boundary ??, MC carbides and M23C6 precipitates which have the greatest effect on stress rupture life and ductility. The mechanical properties of the alloy are strongly dependent on the size and volume fraction of the ?? precipitates. The ever-growing demand for producing more advanced superalloys which can withstand increasingly higher temperature and stresses for better efficiency can be met by increasing the volume fraction of ?? between 30% to 80% 6-8. It is well known that precipitation strengthened nickel-based superalloys containing a very large amount of ?? suffer from relatively poor claddability, an issue which includes solidification cracking and liquation cracking 9, 10. The role of ?? precipitates in promoting susceptibility to microfissuring in superalloys has generally been reported in welding to be through their rapid-precipitation behavior during cooling from cladding temperatures, which induces significant shrinkage stresses. The resultant strengthened grains resist stress relaxation thereby causing cladding stresses to be concentrated across grain boundaries increasing the driving force of cracking. Another factor of extensive penetration and wetting of grain boundary due to constitutional liquation of ?? precipitates is well agreed upon 8, 11, 12. Casting 11, 13, powder metallurgy 14 and directional solidification 15, 16 are some of the processes used to fabricate IN 738. Modern processing technique which implements rapid prototyping or additive manufacturing (AM) shows an upscale demand for fabricating high-temperature components made of nickel-based superalloys. The AM offers the flexibility of manufacturing complex near-net-shape components without the need of traditional machining. Quick production of components of complex geometries produced by AM saves cost and time compared to conventional casting.
In the present study, samples processed using direct laser metal deposition (LMD) were used to explain the mechanism of hot cracking of IN 738 superalloy. High resolution scanning electron microscopy and energy dispersive X-ray spectroscopy techniques are used to identify the evolution of hot cracking by comparing with the models developed for conventional manufacturing methods. By optimizing the processing parameters, hot cracking and liquation that embrittles grain boundaries and cracking due to cladding stresses have been prevented. A scanning strategy has been developed to manufacture a prototype turbine blade made out of IN 738 by LMD without preheating the substrate.
2. Experimental
2.1. Laser metal deposition (LMD) of Inconel 738
Laser metal deposition (LMD) process starts with creating a CAD model. The CAD model is imported as an STL file into Skeinforge software in which the scanning raster and build height is specified and an output tool path is obtained. Typically, the build height is 1/3rd or 1/4th of the laser beam diameter. A 6-axis robot (ABB IRB 1410 M2004) is used to navigate the tool path of the laser deposition, which is controlled by a robot controller (IRC 5 M-2004). The coaxial nozzle is mounted to the robot, and a 1.2 KW diode laser (LDM 1200-40) is used as a source of heat generation. The laser beam used in this study has a circular spot size of 2 mm diameter. Inert gas (Ar) is used as a medium to transport the metal powder and provides a protective environment to prevent oxidation. The metal powder is delivered such that the powder stream converges at the same point with the focused laser beam as schematically shown in figure 1(a).
2.2. Materials
Commercially available gas atomized IN 738 metal powder (Carpenter Powder Products) was used in the present investigation. Figure 2 shows the powder morphology, and table 1 lists the chemical composition of the as-received IN 738-powder. The powder particles exhibit mostly spherical morphology with a few irregular shaped particles and smaller satellite particles stuck to their surface. The average powder particle size was found to be 75 ?m and 90% of the particles were within the 60 – 120 ?m range.
2.3. Sample manufacturing
Experiments were conducted using three different parameters, as shown in table 2, to demonstrate the effects of cracking and the process of optimizing the LMD parameter. Single wall samples of 100 mm × 15 mm × 2 mm (thickness of 2 mm) were deposited on 1020 steel rolled plate. Back and forth scanning strategy was employed indicated in figure 3(b) where the scan direction of each layer changed by 180° with the direction of the previously deposited layer. In all three samples, the laser power in the first two layers was 1000 W with 8 gm/min powder flow rate to ensure a complete metallurgical bond is made between the substrate and the deposit. The laser power was gradually reduced from 1000 W to 800 W at 50 W step after every layer in sample A from table 2 and was kept at 800 W until the end of the deposition. A similar principle was followed for samples B and C. The processing parameter of sample C was used to deposit a turbine blade using IN 738 superalloy, as shown in figure 3(c). The scanning strategy implemented for the turbine blade took the trajectory of clockwise in one layer to counterclockwise the following layer. The repetition of clockwise and counter-clockwise scanning raster ensured a successful crack free build.
2.4. Microstructural characterization and mechanical testing
Samples were cut on the Y-Z plane along the build direction for microscopy, and X-Z plane cut samples were used for XRD analysis. The samples used for high-resolution imaging were electrolytically etched in a solution of 12 ml H3PO4 + 40 ml HNO3 + 48 ml H2SO4 at 6 V for 5 – 6 s. This etching technique was particularly used to reveal ?? phase in the deposit. Microstructural characterization, elemental distribution, and phase constitution of IN 738 have been studied using optical microscopy (Olympus BX51), scanning electron microscopy (JOEL-7600 FE SEM), and X-Ray diffraction (BRUKER D8 XRD) techniques. XRD was performed using CuK? radiation (0.15425 nm) at 40 KV and 40 mA. Diffraction profile was collected from 25°- 105° with a sampling interval of 0.01°. Simulation of solidification segregation was performed using the commercial Thermocalc software. Heat treatment of the as-deposited IN 738 samples was performed using a tube furnace (OTF – 1200X by MTI). Heat treatment was carried out to enhance the mechanical response and study the microstructural evolution and elemental distribution of the as-deposited samples. Samples were solution treated (SHT) at 1120 oC for 2 h and air-cooled followed by precipitation aging at 850 ?C for 24 h followed by air cooling. The size distribution and volume fraction of ?? and carbide particles were measured using Image J. Mechanical properties were studied by performing tension test on a universal tensile testing machine (MTS 810). Tension tests were conducted at an extension rate of 0.01 mm/s for as-deposited, and heat-treated samples.

3. Results and discussion
3.1. Optical microscopy (OM)
Optical micrograph of the longitudinal section (X-Z plane) is shown in figure 4a. A low magnification SEM image of the transverse section (Y-Z plane) is shown in figure 4b. Note that this deposit was made with processing parameters C, as shown in table 2. The clad is porosity free, with no cracks in the entire deposit or debonding at the interface. The deposit consists mostly of columnar dendrites, which grew epitaxially from the substrate along the deposition direction that is the Z-axis. The deposit and substrate act as the heat sink during solidification of the melt pool. This phenomenon promotes the directional growth of the grains counter to the heat flux direction and forms a columnar structure. The change in the direction of the dendrite growth in figure 4a is due to the cooling direction of the melt pool changing with the laser scanning direction as the heat flux direction is in near proximity of the secondary dendrites of the previously deposited layer. Dinda et al. reported secondary dendrites are mostly perpendicular to the primary dendrites, therefore, during LMD the secondary dendrites of the previously deposited layer act as a growth site for the primary dendrites of the newly deposited layer 17. This change in growth direction was noticed across few regions of the deposit.
3.2 Effect of LMD parameters on the cracking susceptibility
During LMD, laser power, scanning speed, and powder flow rate play an important role in determining the quality of the deposit. No macroscopic cracks were observed in sample A but debonding of the single wall deposit from the substrate on the two ends of the wall was noted. An I-beam design was introduced in the first two layers to ensure a good metallurgical bond of the single wall deposit, followed by the regular back and forth scanning strategy till the end of the deposition in sample B. Debonding of the deposit with the substrate was eliminated, but macro-cracks are observed in sample B, as shown in figure 3a. The cracks propagated along the build direction with a crack length of 10 – 13 mm. Samples A and B were processed with a scanning speed of 12 mm/s, and 6 mm/s, respectively. During LMD the deposited material undergoes rapid cooling when the injected metal powders are rapidly heated and melted as the focused laser beam passes the point. During deposition at higher scanning speeds in samples A and B, the thermal gradient between the top of the deposit and the heat-affected zone (HAZ) is high, which produce high tensile stress in the previously deposited layer (HAZ). As a result, cracks are initiated in the deposit at higher scanning speed. On the other hand, sample C was deposited with low laser power (400 watts) and very low scanning speed (3 mm/s), which developed relatively low tensile stress in the HAZ. Consequently, sample C was successfully deposited with no relevant cracks or defects. During selective laser melting and electron beam melting, discussion on preheating of the substrate which reduces the possibility of cracking has been highlighted 18. Egbewande et al. showed that by increasing the welding speed, cracking susceptibility reduced 19. Whereas, during LMD by reducing the scanning speed crack-free deposit is observed in sample C.
3.3 Cracking mechanisms in LMD of IN 738
Cracking is a direct result of the high amount of Al and Ti and the complex rapid heating and cooling history generated during LMD of IN 738. The possible cracking mechanisms are solidification cracking, liquation cracking, and cracking due to oxidation products. Figure 5a shows a grain boundary crack along the build direction of sample B. Dendrite growth in different directions indicated by the red and yellow arrows are the two different grains with a crack along the grain boundary. A high-resolution SEM image of the crack along the grain boundary is shown in figures 5b and 5c. The two sides of the crack boundaries are decorated with bright white particles. EDS analysis confirmed the behavior of oxide dispersion rich in aluminum. Heavy precipitation of the fine crack boundary ?? particles with a size ranging between 142 – 218 nm with a mean size of 185 ± 19 nm was observed along the crack site. Figures 6b and 6c exhibit the path of grain boundary ?? precipitates and as it approaches the crack it transforms to a region consisting of heavy precipitation of the crack boundary ?? particles, as shown in figures 6(c-f). White arrows indicate the grain boundary ?? and the orange arrows indicate the crack boundary ??. Depending on the size of the ?? precipitate, micro-segregation behavior was noticed in the distribution of Al and Ti contents in the primary, secondary, grain boundary, and crack boundary ?? precipitates, as shown in table 3. Crack boundary ?? precipitates has the lowest Al and Ti content compared to primary, secondary and grain boundary ??. The crack along the grain boundary is decorated with aluminum-rich oxide product. Grain boundary is a path of rapid diffusion and grain boundary oxidation rate is higher. Consequently, grain boundary penetration is deeper than the surface oxidation 20. Several mechanisms have been proposed for cracking due to oxidation products. Dynamic embrittlement involving the migration of elemental oxygen ahead of the crack front 21 and stress assisted grain boundary oxidation 22. Viskari et al. suggested the oxidation of secondary ?? precipitate lead to the formation of Al-rich oxides. Oxides are formed on the surfaces between 700 and 1000 °C and an internal oxidation zone has been shown to penetrate the ? grain boundaries and ?? interfaces 23, 24. In the current study, oxidation assisted cracking during deposition occur preferentially around the crack boundary ?? precipitates and at high angle grain boundaries.
Solidification cracking may occur in the interdendritic regions of the same grain (intragranular) and between dendrite arms of different grains (intergranular). Pre-deposited areas experience thermal cycles above the liquidus or solidus temperature due to melting, remelting, partial melting, cyclic annealing, etc. This initiates resolidification by the primary reaction of liquid (L) ? ? during rapid cooling rates leading to non-equilibrium solidification. This leads to enrichment of Al and Ti in the interdendritic or intergranular regions. As the temperature cools down to the eutectic temperature during the final stages of solidification the content of Al and Ti exceeds the critical value in these regions leading to the following eutectic reaction, L ? ? + ??. Figures 6(c-f) exhibit that the crack boundary ?? has a unique mean particle size of 185 nm compared to the rest of the deposit as observed surrounding the cracked regions. These low melting crack boundary ? + ?? resulted in a small cohesive force between dendrites in the intragranular and intergranular regions. At this point, thermal and shrinkage stresses are not accommodated at the eutectic temperature due to the low ductility of the material. Cracking is a result of the competition between internal stress-strain and the ductility of the material 25. Thus, solidification cracks occur when the accumulated strain during rapid cooling overcomes the low ductility of the material at the eutectic temperature.
Liquiation cracking occurs during partial remelting and annealing of the pre-deposited regions between the solidus and liquidus temperature range. Due to rapid heating during subsequent layer deposition, complete dissolution of ?? to the ? matrix is not possible due to insufficient time and the large volume fraction of ?? in IN 738. Due to non-equilibrium processing, the volume fraction of ?? deviates from the equilibrium processed condition. Upon reaching the eutectic temperature during subsequent layer deposition, the retained crack boundary ?? reacts with the ? matrix to produce low melting point eutectic liquid film observed in figure 6c at the crack initiation site. These liquation sites are observed to be spread along the grain boundaries as resolidified products during cooling. Regions where liquation cracks are observed in the deposit is the point of low ductility with sufficient accumulation of strain resulting in cracks along the grain boundary. The angle made between the dendrites from the two grains was observed to be greater than 15° suggesting a high angle grain boundary (HAGB) in figure 5a. Note that the HAGB are sensitive to cracking. Sidhu et al. 26 reported the formation of secondary solidification constituents formed from the interdendritic liquid such as MC carbide, M3B2-boride, and ?- ?? eutectic causing liquation cracks during welding of IN 738.
To get an approximation of the micro-segregation of the elements/phases observed during solidification, Thermo-Calc software was used for non-equilibrium (infinite diffusion coefficients of solute redistribution in the liquid phase and no diffusion in the solid phases) and equilibrium (infinitely slow cooling) solidification. Thermodynamic calculations were simulated based on the observed microstructure and the phases that were not observed are excluded. Non-equilibrium solidification simulation result, as shown in figure 10a, displays that the solidification begins with ? phase at 1340 °C. At 78 % liquid, M(Ti,Ta,Nb)C carbide started to form around 1330 °C. At 1175 °C with 6% liquid remaining, the formation of ?? begins. In the final stages of solidification, liquid is enriched with small fraction of Mo, Cr and boron resulting M2(Mo, Cr)B phase that exists until 950 °C. The solidification range is given by the difference between the liquidus (1340 °C) and the solidus (950 °C) temperature which in Scheil-Gulliver (non-equilibrium) calculation is about 390 °C, and 100 °C under equilibrium condition.