Creep Behavior of Diffusion-Welded Alloy 617

: Plate-type heat exchangers are anticipated to be used in the next-generation nuclear industry, and solid-state diffusion welding is a critical technology for building plate-type heat exchangers with high integrity. In this study, we manufactured a diffusion weldment and evaluated its creep behavior. Microscopic analysis revealed that Al-rich oxides were developed along the interface, signiﬁcantly impeding grain-boundary movement across the interface. Oxide-containing planar grain boundaries resulted in premature brittle fracture at the interface with less than 9% creep strain under all test conditions. The time to rupture and time to 1% creep strain of the diffusion weldment were less than those of the as-received alloy, while the slopes in double-logarithmic plots were almost identical for both alloys. In a Larson–Miller parameter study, the stress to rupture of the diffusion weldment reached 95.59% of that of the as-received alloy, whereas the stress to 1% creep strain steeply decreased in the low-stress range.


Introduction
Plate-type heat exchangers have attracted much attention for application in the nextgeneration nuclear industry because they can realize high heat-transfer efficiency between primary and secondary systems [1][2][3]. Solid-state joining methods, such as brazing, transient liquid phase bonding, and diffusion welding, are essential processes for manufacturing such plate-type heat exchangers.
In brazing and transient liquid phase bonding, a homogenization heat treatment may result in microstructure and mechanical properties similar to those of as-received alloy. However, secondary brittle intermetallic phases are inevitably created near the interface because of the presence of melting-point depressants, such as boron, phosphorous, and/or silicon, in an interlayer. Such secondary phases degrade mechanical properties with negligible permanent deformation at the interface.

Materials and Methods
We used Alloy 617 (Ni-Cr-Co-Mo), which is a solid-solution-strengthened Ni-based superalloy that retains superior mechanical properties and oxidation/corrosion resistance at high temperatures. A sheet with dimensions of 3048 (L) mm × 1219 (W) mm × 1.60 (T) mm was supplied by the Special Metals Company. Table 1 presents the chemical composition of the sheet (Heat No. XX4516UK), which meets the ASTM: B168-19 specification. This alloy contains Cr and Al for oxidation/corrosion resistance as well as Co and Mo for solid-solution hardening. It is widely known that the grain boundary of this alloy is slightly covered with Cr-rich M 23 C 6 carbide, and Ti-rich MC carbide is evenly spread in the matrix. The supplied sheet was sectioned using the water-jet cutting method. Sixty sheets with dimensions of 200 (L) mm × 200 (W) mm × 1.60 (T) mm were prepared for diffusion welding. The rolling direction of the sheet was not considered. The number and dimensions of the sheets employed in this study were determined on the basis of the procedure and performance qualification listed in Section IX QW-185 of the American Society of Mechanical Engineers (ASME). Both sides of all sheets were mechanically ground with up to #1200 SiC grit paper to remove surface contamination, ultrasonically cleaned, and then dried in the air before being placed into diffusion-welding equipment.
Diffusion welding was conducted at 1150 • C for 2 h (4 h in total, including homogenization of the temperature in the diffusion-welding equipment) with a compressive uniaxial pressure of 14.7 MPa. A high vacuum condition (~10 −6 Torr) was employed to impede the formation of tenacious oxides on the surface as much as possible. The filler metal was not applied to preserve the similarity of joints. The stack was subjected to post-weld heat treatment (PWHT) at 1150 • C for 8 h in a high-vacuum (~10 −6 Torr) furnace for further atomic diffusion near the interfaces. The stack was cooled to room temperature using inert high-purity nitrogen gas. These parameters produced a stack (hereinafter, diffusion weldment) with dimensions of 200 (L) mm × 200 (W) mm × 91.05 (T) mm.
For scanning electron microscopy (SEM) analysis, mechanically sectioned specimens were dipped into a chemical etchant (HNO 3 :HCl = 1:3, vol.%) for approximately 40 s to reveal the grain boundary and interface of the diffusion weldment. The focused ion beam (FIB; FEI NOVA200) method was used to extract a specimen at an interface for transmission electron microscopy (TEM; JEM-ARM200F, JEOL Ltd., Tokyo, Japan) analysis.
Cylindrical bar-type specimens with a gauge diameter of 6.0 mm and a gauge length of 30.0 mm were produced for stress-rupture tests (Figure 1), which were conducted at high temperatures of 800, 850, 900, and 950 • C according to the ASTM: E139-11 (2018) standard. Interfaces in the direction perpendicular to tensile loading were located in the gauge section. The distance between interfaces approximately corresponds to the thickness of the sheet (~1.60 mm). The circular ridges were machined at both ends of the gauge section to attach an extensometer.   Figure 2 shows the test apparatuses for the stress-rupture test. Both dead weight and lever type machines were employed in this study; a lever arm ratio of 20:1 was applied. The creep strain was determined by the elongation measured by the extensometer equipped at the circular ridges. Relative movement of the circular ridges during the stress-rupture test was displayed in a linear variable differential transducer (LVDT) attached at the bottom side of the furnace, and then the record was automatically transferred to a personal computer. A three-zone split furnace was used. A K-type thermocouple was used to monitor the temperature in the gauge section. The temperature was maintained within ±2 • C during the stress-rupture test. We produced the pull rod and jig with Ni-based superalloy to prevent severe oxidation and thermal degradation.  near the interface (an interface is marked with arrows in each micrograph), implying that the diffusion-welding parameter used in this study is applicable to this alloy. Further, the interface can be easily recognized. Secondary precipitates are observed along the interface, and simultaneously, grain-boundary movement across the interface is limited. Figure 3 shows cross-sectional SEM and TEM micrographs of the region near an interface. No defects (detectable cracks, incomplete bonds, or porosity) were observed near the interface (an interface is marked with arrows in each micrograph), implying that the diffusion-welding parameter used in this study is applicable to this alloy. Further, the interface can be easily recognized. Secondary precipitates are observed along the interface, and simultaneously, grain-boundary movement across the interface is limited. Following the authors' previous TEM analysis using energy-dispersive X-ray spectroscopy (EDS) and electron energy-loss spectroscopy (EELS), the secondary precipitates formed along the interface were identified as Cr-rich carbides and Al-rich oxides [36,39]; as indicated in the authors' previous work, the black dots in Figure 3b are Al-rich oxides. The Al-rich oxides stem from the reaction between Al in the matrix and oxygen in the diffusion-welding equipment. Although the diffusion-welding process was conducted under a high-vacuum condition (~10 −6 Torr), the oxygen level in the diffusion-welding equipment was high enough to produce Al-rich oxides on the surfaces at high temperatures. The discrete Al-rich oxides along the interface restricted grain-boundary movement, leaving a planar grain boundary. Some researchers have emphasized that the evolution of stable surface oxides (film or discrete type) is a governing factor limiting metalto-metal joining [34,36,38,39,[50][51][52][53][54].

Creep Behavior
Stress-rupture tests were conducted at 800, 850, 900, and 950 °C in ambient air under various stress levels to characterize the creep behavior of the diffusion weldment. A fixed load was applied to the specimen in each stress-rupture test. The stress levels and test temperatures applied in this study reflect the anticipated operating conditions of hightemperature reactor systems [55]. Figure 4 shows curves of creep strain versus time, where Figure 4a-d shows the results of stress-rupture tests performed at 800, 850, 900, and 950 °C, respectively. Table 2 summarizes the time to rupture (trup), strain at rupture (εrup), and time to 1% creep strain (t1%). Following the authors' previous TEM analysis using energy-dispersive X-ray spectroscopy (EDS) and electron energy-loss spectroscopy (EELS), the secondary precipitates formed along the interface were identified as Cr-rich carbides and Al-rich oxides [36,39]; as indicated in the authors' previous work, the black dots in Figure 3b are Al-rich oxides. The Al-rich oxides stem from the reaction between Al in the matrix and oxygen in the diffusion-welding equipment. Although the diffusion-welding process was conducted under a high-vacuum condition (~10 −6 Torr), the oxygen level in the diffusion-welding equipment was high enough to produce Al-rich oxides on the surfaces at high temperatures. The discrete Al-rich oxides along the interface restricted grain-boundary movement, leaving a planar grain boundary. Some researchers have emphasized that the evolution of stable surface oxides (film or discrete type) is a governing factor limiting metal-to-metal joining [34,36,38,39,[50][51][52][53][54].

Creep Behavior
Stress-rupture tests were conducted at 800, 850, 900, and 950 • C in ambient air under various stress levels to characterize the creep behavior of the diffusion weldment. A fixed load was applied to the specimen in each stress-rupture test. The stress levels and test temperatures applied in this study reflect the anticipated operating conditions of high-temperature reactor systems [55].  Table 2 summarizes the time to rupture (t rup ), strain at rupture (ε rup ), and time to 1% creep strain (t 1% ).  At 800 °C, creep curves were drawn for applied stresses of 60, 70, and 120 MPa (Figure 4a). A test conducted at 100 MPa showed abnormal features in the creep curve, which  At 800 • C, creep curves were drawn for applied stresses of 60, 70, and 120 MPa (Figure 4a). A test conducted at 100 MPa showed abnormal features in the creep curve, which could be attributed to the jig material; hence, the data for 100 MPa were discarded (Table 2). Further, a creep curve with t rup = 4661 h was successfully obtained for applied stress of 60 MPa. Primary and secondary creep stages were rarely observed. The tertiary creep stage commenced early at a low creep strain. At this temperature, the strains at rupture were low (<2.07%). At 850 • C, one creep curve was drawn for applied stress of 40 MPa (Figure 4b). In this case, t rup was 3985 h, and ε rup was 5.10%. Again, primary and secondary creep stages were hardly observed. The specimen was fractured in the tertiary creep stage but showed little permanent deformation. The creep curves at 900 • C are shown in Figure 4c It is well known that as-received Alloy 617 specimens are fractured in the ductile mode in short-term (high-stress level) tests, and the fracture mode changes to brittle in long-term (low-stress level) tests. Further, ε rup usually decreases as t rup increases. In contrast, the ε rup of the diffusion weldment did not exhibit a clear relation with t rup . In all cases, limited creep strains (<9%) with fractures near the interface were observed in the diffusion weldment. Figure 5a,b shows plots of log stress versus log time to rupture and log stress versus log time to 1% creep strain, respectively. Both plots clearly show linearity, implying that the following power-law function is applicable for the diffusion weldment: where σ is the stress (MPa), A (MPa·h −m ) is a material constant, and m is an exponent. A is the intercept, and m is the slope in the double-logarithmic plot. Here, t can be the time to rupture, time to tertiary creep stage (t ter ), or time to 1% creep strain. In this study, we plotted the time to rupture and time to 1% creep strain because the time to tertiary creep stage was not ascertained in some of the stress-rupture tests ( Figure 4).
rupture, time to tertiary creep stage (tter), or time to 1% creep strain. In this study, we plotted the time to rupture and time to 1% creep strain because the time to tertiary creep stage was not ascertained in some of the stress-rupture tests (Figure 4). The values calculated using Equation (1) are listed in   The absolute inverse of the slope (|1/m|) is usually considered an indicator of the creep deformation and fracture mechanism. The absolute inverse values of the slope of the double-logarithmic plot for the time to rupture (|1/m rup |) were in the range of 4.639-5.172. A value of~5 is typical in pure metals and Class I solid-solution alloys [56], implying that dislocation climb governs the creep mechanism for the diffusion weldment. The absolute inverse values of the slope of the double-logarithmic plot for the time to 1% creep strain (|1/m 1% |) were in the range of 5.904-6.476. These values are slightly greater than those for the time-to-rupture plot. This is thought to result from the formation of Al-rich oxides along the interface. The Al-rich oxides can act as pinning obstacles at early creep stages with high efficiency; however, the Al-rich oxides become aligned at intervals of~1.60 mm (the thickness of the sheet for diffusion welding). As the Al-rich oxides at the interfaces account for only a small fraction of the total amount, the overall creep mechanism at the early creep stages continues to be dislocation climb. Table 3. The intercept and slope in the power-law function (Equation (1)).

Fractography
All specimens failed at an interface because the oxide-containing interfaces were much harder and less ductile than that of the matrix. Figure 6 shows the representative SEM morphology of the crept specimens, where Figure 6a,b show the fracture surfaces tested under 25 and 50 MPa at 900 • C, respectively. Necking at the interface is rarely observed without significant plastic deformation, regardless of the stress levels. Furthermore, the fracture surfaces are rather flat and perpendicular to the loading direction. The flat zone shows traces of mechanical grinding, which was performed during the surface preparation step for the diffusion-welding process. This is another indication of insufficient grainboundary movement during the diffusion welding and post-weld heat treatment.
At higher magnification, two types of fracture morphologies were observed in the micrographs. They are classified into a brittle zone (denoted by B in blue) and ductile zone (denoted by D in yellow). The brittle zone occupies most of the fracture surface. Nonetheless, in the ductile zone, clear evidence of plastic deformation during the stressrupture test is observed. Predominantly brittle fracture surfaces were previously reported for diffusion weldments. For example, Mylavarapu et al. reported the immediate failure of applying a load to diffusion weldment [1]; they reported brittle failure with a flat surface, even with the aid of a Ni-foil interlayer between Alloy 617 plates.

Life Estimation
The Larson-Miller parameter is a time-temperature parameter (TTP) that has been widely used to predict long-term creep life (or stress) from short-term stress-rupture test data [57]. The Larson-Miller relation is expressed as follows: where LMP is the Larson-Miller parameter calculated from Equation (2); T is the temperature in kelvin; t is the time to rupture, time to the tertiary creep stage, or time to 1% creep strain in h depending on the purpose of the life estimation; C is a material constant. As pointed out by many researchers [58][59][60], determining C is critical in life estimation. A unique C value of 16.73 was used for estimating LMP (t rup ) following data analysis of creep and creep-rupture at Idaho National Laboratory [60]. The same value was adopted for predicting LMP (t 1% ).

Life Estimation
The Larson-Miller parameter is a time-temperature parameter (TTP) that has been widely used to predict long-term creep life (or stress) from short-term stress-rupture test data [57]. The Larson-Miller relation is expressed as follows: where LMP is the Larson-Miller parameter calculated from Equation (2); T is the temperature in kelvin; t is the time to rupture, time to the tertiary creep stage, or time to 1% creep strain in h depending on the purpose of the life estimation; C is a material constant. As pointed out by many researchers [58][59][60], determining C is critical in life estimation. A unique C value of 16.73 was used for estimating LMP (trup) following data analysis of creep and creep-rupture at Idaho National Laboratory [60]. The same value was adopted for predicting LMP (t1%). LMP is also expressed as a linear function of stress: where B is the intercept, and n is the slope of the plot of log stress versus LMP (trup and t1%). LMP is also expressed as a linear function of stress: where B is the intercept, and n is the slope of the plot of log stress versus LMP (t rup and t 1% ). Figure 7 shows the Larson-Miller relation of the diffusion weldment, where Figure 7a,b plot log stress versus LMP (t rup and t 1% , respectively). Data points and the corresponding regression for the as-received alloy are drawn for comparison. The data points for the as-received alloy were gathered from the authors' previous reports [61][62][63]. Table 4 lists the intercepts and slopes of both the as-received and diffusion-welded alloys. Figure 7 shows the Larson-Miller relation of the diffusion weldment, where Figure  7a,b plot log stress versus LMP (trup and t1%, respectively). Data points and the corresponding regression for the as-received alloy are drawn for comparison. The data points for the as-received alloy were gathered from the authors' previous reports [61][62][63]. Table 4 lists the intercepts and slopes of both the as-received and diffusion-welded alloys.     (3)). C in Equation (2)  For LMP (t rup ), the intercepts (B rup ) are 557,109 and 532,562 for the as-received and diffusion-welded Alloy 617, respectively. The intercept for the diffusion weldment reached 95.59% of that for the as-received alloy, and similar observations were made for the slopes (n rup ) of the as-received and diffusion-welded alloys, implying that both alloys have the same creep deformation and fracture mechanism.
In contrast, LMP (t 1% ) shows anomalous behavior. The intercepts (B 1% ) were 143,516 and 628,072 for the as-received and diffusion-welded alloy, respectively. The intercept of the diffusion weldment was much higher than the practically achievable limit; moreover, the slope (n 1% ) of the diffusion weldment was lower than that of the as-received alloy. In other words, two straight lines intersect at an LMP (t 1% ) value of 23,347 and a stress value of 23.01 MPa (Figure 7b). This implies that in the low-LMP (t 1% ) range, the diffusion weldment requires higher stress than the as-received alloy to induce 1% creep strain, whereas the situation is the opposite in the high-LMP (t 1% ) range. Such phenomena seem to result from the presence of Al-rich oxides at the interfaces. As discussed above, Al-rich oxides placed in rows perpendicular to the loading direction can significantly obstruct creep deformation during the early creep stages. However, when the creep specimen enters the secondary creep stage, the oxide-containing interfaces cannot extend significantly. Therefore, in the high-LMP (t 1% ) range, stress values much lower than expected are inevitable for the diffusion weldment.

Conclusions
Diffusion welding with Alloy 617 was investigated for application to compact heat exchangers in the next-generation nuclear industry. The following conclusions can be drawn from the microscopic analysis and experimental stress-rupture tests.
(1) Al-rich oxides formed along the interface are strongly associated with planar grain boundaries at the interface. Such secondary precipitates inhibit the grain-boundary movement across the interface with high efficiency.
(2) The diffusion weldment showed limited creep strains (<9%) at high temperatures and crept specimens exhibited premature rupture at the interface. A ductile zone was partially observed, but a brittle zone occupied most of the fracture surface.
(3) The stress exponents obtained from double-logarithmic plots of log stress versus log time to rupture and log stress versus log time to 1% creep strain follow a power-law relation. An absolute inverse of the slope of~5 in the plots indicates that dislocation climb is the dominant creep mechanism in the diffusion weldment.
(4) For LMP (t rup ), the intercept value of the diffusion weldment reached~95.59% of that of the as-received alloy; a similar trend was observed for the slope. In contrast, LMP (t 1% ) showed anomalous behavior. For LMP (t 1% ) less than 23,347, the stress level of the diffusion weldment was higher than that of the as-received alloy, whereas, for LMP (t 1% ) greater than 23,347, the opposite was the case.