Experimental and Numerical Results for Diffusion Bonded Joints (Experimental and Applied Mechanics)

ABSTRACT

Commercially pure Fe and Ni have been diffusion bonded. Pure Cu (99.999%) and Au-20Sn eutectic alloys have been used to bond Fe while Cu has been used to bond Ni. Fe bonded using Cu at 1071°C for 10 h showed ~20 ^m thick residual Cu in the bond centerline. However, at 1100°C bonding temperature for 10 h the residual Cu disappeared and the Cu content of the joint centerline was ~ 7wt.%. Although Sn forms intermetallics with both Fe and Au, no intermetallics were found when Au-20Sn eutectic was used to bond Fe at 600-650°C for 10h. Ni bonded at 1071°C for 10h contained 40 wt.% Cu in the joint centerline with the Cu content decreasing gradually with increasing distance from the bondline. Cu content in the bond centerline decreased to 35wt.% at 1100°C for 10 h. The concentration profiles of Cu in a Ni-Cu diffusion couple were simulated with DICTRA/Thermocalc. The simulated profiles were comparable to the experimental profiles. The ultimate tensile strengths obtained for Fe-Cu system were 245 and 276 MPa at 1085±1 and 1090±1°C for 10 h, respectively.

INTRODUCTION

To enhance overall efficiency, the operating temperatures of aero-engines and power generation turbines are continuously increasing. Higher operating temperatures result in increased effects from creep, fatigue and corrosion causing rapid degradation of the components. This damage requires the repair of these components in order to extend their total life and to keep the material costs to a minimum. Two material families that are often used in hot sections of aero-engines and power generation turbines are nickel-based superalloys and stainless steels. Since these alloys are exposed to severe operating conditions, they require resistance to high temperature creep, sufficient tensile strength and corrosion resistance [1]. For example, stainless steels are generally used in process plants, petrochemical industries, aero-engine hot section components, pump and valve shafts, steam generators, expansion joints, super-heaters, re-heaters, etc. because of their high strength, ductility, resistance to creep and resistance to corrosion at elevated temperatures [2].

The most widely used repair process for metals and alloys is the fusion welding process, which typically involves relatively large scale melting at the joint line and the introduction of a filler metal. Depending on the material chemistry, borides and silicides can be formed during fusion welding. These phases are brittle and detrimentally affect the mechanical properties of the joints [3-5]. Also, weldability of nickel-based superalloys largely depends on the Ti and Al content. Precipitation-hardened nickel-based superalloys containing high Al and Ti are susceptible to heat affected zone (HAZ) cracking during welding and post-weld heat treatment [6-8]. Different stainless steels also show intergranular cracks in the HAZ during fusion welding. Fusion welding disrupts the ferrite-to-austenite ratio in duplex stainless steels [9]. In contrast to fusion welding, diffusion bonding is a joining process in which bonding is primarily due to atomic diffusion between base metals and filler materials (if any). Melting may occur, as in the case of transient liquid phase (TLP) bonding in which a filler metal or alloy with a relatively low melting temperature is used. Joints may also remain solid throughout the diffusion bonding process if a phase with a higher melting temperature forms before reaching the bonding temperature. Acceptable joint morphology depends on proper control of the thermodynamics and kinetics of the materials involved in the joining process. Transient liquid phase bonding (TLP), or diffusion brazing, has become a preferred method of joining for both nickel-based superalloys and stainless steels [10]. Optimum joint quality requires proper control of parameters which includes temperature, time, filler alloy composition and thickness.

To simulate diffusion-controlled transformations, two steps are required: (1) calculation of the thermodynamic quantities, and (2) modeling of the kinetics of the transformation. Thermocalc, developed at the Royal Institute of Technology in Stockholm, can predict the correct equilibrium state in multi-component alloy systems [11]. DICTRA (Diffusion Controlled TRAnsformation), can simulate the diffusion controlled transformation in multi-component systems [12]. DICTRA is a finite difference code and uses a Newton-Raphson iteration technique to solve the multi-component diffusion equation. In solving the diffusion equation, DICTRA uses kinetic databases and recalls thermodynamic quantities from Thermocalc. The accuracy of DICTRA simulations depends on the accuracy of the thermodynamic and kinetic data and can be verified by comparison with experimental results.

In the present investigation, commercially pure iron and nickel are diffusion bonded using different interlayers. The ultimate goal of the work is to optimize bonding conditions, including temperature, bonding time, interlayer type and thickness, with respect to the resulting joint strength. Joint microstructures and chemical compositions are reported. The concentration profiles of the diffusing species in a diffusion couple are simulated using DICTRA/Thermocalc software and compared with experimental results.

EXPERIMENTAL METHOD

Initial studies of the diffusion bonding joint morphology were performed using base metal rods with a diameter of 6.35 mm, 8 mm long. After cutting, the samples were polished to a 1200 grit finish and cleaned in an ultrasonic bath using isopropyl alcohol. The interlayer foils were cut the same diameter as the base metal rods. Immediately after cleaning, the samples were placed in a jig and put in a tube furnace. The tube was repeatedly evacuated and purged with argon gas. All bonding was done under vacuum. The bonding jig was made of Kovar to minimize the thermal expansion mismatch. Following morphology studies, the tensile specimens were fabricated. The specimens had a gauge length of 59±0.5 mm and diameter of 9.0±0.1 mm. These specimens were also polished to a 1200 grit finish and cleaned in ultrasonic bath. The sample geometries are shown in Fig. 1.

Fig. 1 (a) Schematic of diffusion bond geometry, (b) diffusion bonded sample and (c) tensile specimen before and after diffusion bond

Cu interlayer was used for bonding both commercially pure Fe and Ni while Au-20Sn interlayer was used for Fe only. Cu foil has been selected because it does not form any intermetallics with either Fe or Ni and it has a melting temperature less than either of the base metals. The thicknesses were 100 and 25 ^m for Cu interlayers and 100 ^m for Au-20Sn interlayers. The bonding temperatures were 1050°C to 1100°C for Cu interlayers. For Au-20Sn interlayers, 600-650°C temperatures were used. Following bonding, the morphology samples were mounted in epoxy and polished perpendicular to the bondline to prepare for metallography. The morphology of the bonded samples was examined using a Hitachi S-3400N scanning electron microscope and the composition and different phases in the bond area were determined by energy dispersive spectrometry (EDS). The strength of the tensile specimens were measured using a Shimadzu AG-IS 50kN universal testing machine. Samples were tested under a cross-head speed of 1mm/min.

For the numerical simulation, only half of the joint was used modeled due to symmetry. The integration points were assumed to be linearly distributed throughout the Fe-Cu diffusion couple. The input parameters were bonding temperature, bonding time, heating and cooling rate, compositions and phases of the diffusion couple. To initiate the calculation, the Cu region was set with a ‘seed’ of 0.001 wt.% Ni in it and Ni region was set with a ‘seed’ of 0.1 wt.% Cu in it.

RESULTS AND DISCUSSION

Fig. 2 (a) SEM images of diffusion bonded Fe using Cu interlayer (25 |um) at 1071°C for 10h, (b) magnified image of (a) and (c) Concentration profile of Cu from the bond centerline to the interior of base metal

Initially, 1050°C and 100 |m C foil was used for bonding Fe, but the interdiffusion was negligible. The temperature was then increased to 1071°C and the foil thickness was decreased to 25 |m. Fig. 2 shows (a) SEM images of diffusion bonded Fe using Cu interlayer (25 |m) at 1071°C for 10h, (b) magnified image of (a) and (c) concentration profile of Cu from the bond centerline to the interior of base metal. A significant amount of Cu remained at the centerline of the joint. The thickness of the residual Cu was ~ 20|m (the foil thickness was 25 |m). The concentration profile indicates that the Cu concentration suddenly dropped just after interface and gradually decreased after that. The Fe content in the residual Cu was ~3 wt.%. According to the Fe-Cu phase diagram (Figure 3) y-Fe can dissolve ~ 7.2 wt.% Cu at 1100°C and s-Cu can dissolve ~ 4.15 wt.% Fe at the same temperature [13]. So, to enhance the inter-diffusion, the bonding temperature was increased to 1100°C.

Fig. 3 Phase diagram of Fe-Cu generated using Thermocalc software [11]

Fig. 4 (a) SEM images of diffusion bonded Fe using Cu interlayer (25 ^m) at 1100°C for 10h and (b) the concentration profile of Cu taken from the bond centerline to the interior of the base metal

Fig. 4 shows (a) SEM images of diffusion bonded Fe using Cu interlayer (25 ^m thick) at 1100°C for 10h and (b) the concentration profile of Cu taken from the bond centerline to the interior of the base metal. The bond centerline contains only ~ 7.4 wt.% Cu. In diffusion bonding it is necessary for the interlayer to diffuse into the base metal as completely as possible so that the bond strength remains close to that of the base metal. However, some micro-pores appeared in the joint area. This might be due to the fact that y-Fe gets supersaturated at 7.2 wt.% Cu while e-Cu gets supersaturated at 4.15 wt.% Fe at 1100°C; this may create an atomic flux imbalance. Investigations into this phenomenon are ongoing.

Fig. 5 shows SEM images of diffusion bonded Fe using Au-20Sn interlayer (100 ^m thick) (a) at 600°C for 10h, (b) magnified image of (a) and (c) at 650°C for 10h. After holding at 600°C for 10 h, almost all interlayer has been diffused into the base metal. The thickness of the residual interlayer was less than a micrometer with a composition of 40-45 wt.% Au and 19-21 wt.% Sn. The joint area away from the residual interlayer contains a maximum of 4 wt.% Au and 0.2 wt.% Sn and decreased gradually. No intermetallics were found in the joint area. After increasing the bonding temperature to 650°C, the residual interlayer disappeared. Fast diffusion occurred at such a low temperature because the Au-20Sn eutectic alloy has a melting temperature of 277°C.

Fig. 5 SEM images of diffusion bonded Fe using Au-20Sn interlayer (100 ^m) (a) at 600°C for 10h, (b) magnified image of (a) and (c) at 650°C for 10h

Fig. 6 shows (a) SEM image of diffusion bonded Ni using Cu interlayer (25^m thick) at 1071°C for 10 h and (b) measured and predicted (DICTRA) concentration profiles of Cu from the bond centerline to the interior of base metal. The microstructures of the base metal and the joint area were not distinguishable because Cu and Ni form a complete solid solution in any composition range. The composition of the bond centerline was ~ 40 wt.% Cu and the Cu content decreased gradually away from the bond centerline. DICTRA was able to capture the diffusion behavior of the system quite well.

Fig. 6 (a) SEM image of diffusion bonded Ni using Cu interlayer (25^m) at 1071°C for 10 h and (b) concentration profile of Cu from the bond centerline to the interior of base metal

Fig. 7 (a) SEM image of diffusion bonded Ni using Cu interlayer (25|m) at1100°C for 10 h and (b) concentration profile of Cu from the bond centerline to the interior of base metal

Fig. 7 shows (a) SEM image of diffusion bonded Ni using Cu interlayer (25|m thick) at1100°C for 10 h and (b) measured and predicted concentration profiles of Cu from the bond centerline to the interior of base metal. At the 1100°C bonding temperature, the Cu content at the bondline decreased to ~ 35 wt.% and gradually decreased with distance from the centerline, as found in the previous condition. Cu also diffused farther from the bondline than before. In Fig. 6 and Fig. 7, the centerline of the joint is shown with a broken line. The DICTRA-simulated profile agreed well with the experimental profile for this condition.

The average tensile strength for commercially pure Fe is 495 MPa and for annealed Cu is 240 MPa. The Fe-Cu joint bonded at 1085±1°C for 10 h had a measured strength of 245 MPa and the strength increased to 276 MPa when bonded at 1090±1°C. Under both of these bonding conditions, residual Cu remained in the joint area. The measured joint strength was slightly higher than that of pure Cu. The strength increased beyond Cu strength because of solid solution effect as the residual Cu contained small percentage of Fe (~3 wt.%). The strength of Fe-Cu system is yet to be optimized while the strengths of other systems (Fe-(Au-20Sn), Ni-Cu) have yet to be determined.

CONCLUSION

Fe was diffusion bonded using Cu and Au-Sn interlayers without forming any intermetallics. However, residual Cu of approximately 20 |m thick was found in the joint centerline at 1071°C for 10 h. At 1100°C for 10 h of bonding, the Cu content in the bond centerline was ~7.15 wt.%, but some micro-pores appeared in the joint area which could potentially decrease the strength of the joint. Au-20Sn eutectic interlayers allowed bonding Fe at 650°C for 10 h without having any residual interlayer material at centerline. Ni bonded at 1071°C for 10 h contained 40 wt.% Cu and at 1100°C for 10 h contained 35 wt.% Cu at the centerline. The simulated profiles using DICTRA/Thermocalc software for Ni-Cu system agreed well with the profiles obtained experimentally. The strengths found for Fe-Cu joint were 245 and 276 MPa at 1085±1 and 1090±1°C, respectively.