Effects of Different Wire Drawing Routes on Grain Boundary Character Distribution, Microtexture, δ-Phase Precipitation, Grain Size and Room Temperature Mechanical Behavior of Alloy 718

aUniversidade Federal do Rio de Janeiro, Instituto Alberto Luiz Coimbra de Pós-Graduação e Pesquisa de Engenharia, Programa de Engenharia Metalúrgica e de Materiais, Av. Horácio Macedo, 2030, Cidade Universitária, 21941-972, Rio de Janeiro, RJ, Brasil bInstituto Militar de Engenharia-IME, Departamento de Ciências dos Materiais, Praça General Tibúrcio, Urca, 22290-270, Rio de Janeiro, RJ, Brasil cPrograma de Pós-Graduação em Engenharia de Minas, Metalúrgica e Materiais, Universidade Federal do Rio Grande do Sul, Av. Paulo Gama, 110, Farroupilha, 90040-060, Porto Alegre, RS, Brasil. dCentro Federal de Educação Tecnológica Celso Suckow da Fonseca, Departamento de Engenharia Mecânica, Avenida Maracanã, 298/299, Maracanã, 20271-110, Rio de Janeiro, RJ, Brasil


Introduction
The nickel-based superalloy 718 is commonly used in the oil and gas, nuclear and aerospace industries in structural applications that combine a wide range of temperatures with mechanical solicitations (high yield strength, toughness and creep resistances), as well as corrosive environments [1][2][3][4] . This alloy has a face centered cubic (FCC) matrix (γ) with medium to low stacking fault energy (SFE) 5 . The relatively high strength of this material is owed to the precipitation of the metastable coherent phase γ" (BCT -Ni 3 Nb -DO 22 ) and the stable coherent γ' (FCC -Ni 3 (Al,Ti) -L1 2 ) in the γ matrix 6 . Besides these phases, the precipitation of the stable δ-phase (Orthorhombic -Ni 3 Nb -DO a ) may occur 7 , which can precipitate following the dissolution of the metastable γ" phase or directly from the γ matrix [7][8][9] . The influence of the δ-phase on the mechanical properties of the nickel-based superalloy 718 is ambiguous. According to Singh et al. 10 , the δ-phase precipitates at grain boundaries during the recrystallization process, which limits the grain boundaries mobility and controls alloy's grain size. In contrast, Medeiros et al. 11 related the loss in ductility due to the presence of the δ-phase since these particles work as preferential sites for voids and cracks during mechanical solicitations.
Alloy 718 is often susceptible to intergranular failures of different nature, such as oxygen assisted intergranular cracking (OAIC) and hydrogen embrittlement 12,13 . Hence, some authors propose the deliberated manipulation of grain boundary character distribution (GBCD) in order to mitigate those phenomena by breaking the connectivity of random high angle boundaries, more prone to crack propagation 14,15 . That manipulation is directly related to Σ3 twin boundary mobility and interaction through a mechanism that is known as "Σ3 regeneration model", where twins interact, generating new Σ9 and Σ27 boundaries segments 16 . Increase the length *e-mail: luizmauricio@metalmat.ufrj.br fraction of Σ3 n boundaries (Ʃ3 + Ʃ9 + Ʃ27), in replacement of random HAB, is achieved by subjecting low stacking fault energy (SFE) materials to iterative thermomechanical processing (TMP), alternating successive cycles of deformation followed by solution annealing 8,15 . So far, there is still some uncertainty in the literature regarding the mechanism accountable to multiple twinning: recrystallization, grain growth or grain boundary migration induced by deformation 17,18 .
After a literature review, it was evidenced that most knowledge about GBCD in alloy 718 is built based on multiple cycles of cold rolling followed by solution annealing 8,11,19,20 . However, alloy 718 can be fabricated as different geometries and by different thermomechanical process as wire drawing or extrusion. Wires of alloy 718 can be used, for instance, to fabricate high temperature coil springs. Therefore, assessing alloy´s response to different formation mechanisms in regard to GBCD evolution is needed.
Other materials, such as high purity Cu, have already had their GBCD evolution studied after being subjected to different deformation modes like wire drawing. For example, Baudin et al. 21 observed the higher in length fraction of Σ3 n for a low cross-sectional area reduction. Opposingly, Matsushita et al. 22 have observed the most relevant increase in Σ3 n for a greater area reduction. Yet, those controversies evidence how little understanding is available regarding the influence of wire drawing iterative thermomechanical processing on material's GBCD.
Thus, the goal of the present study is to investigate the applicability of GBCD manipulation via wire drawing, subjecting alloy 718 to iterative processing routes analogous to the ones already established in the literature, by means of cold rolling. The deformation mode related to drawing and its effect on GBCD as well as on crystallographic orientation were studied together with the influence of δ-phase on grain size and consequently on resulting mechanical properties. Alloy 718 wires were subjected to two different iterative thermomechanical processing, alternating cycles of drawing followed by solution annealing both at sub and super δ-solvus temperatures.

Material and processing
The as-received material was supplied as a wire rod of alloy 718 produced by hot rolling with 6.7mm diameter, with chemical composition as presented in Table 1.
Two distinct thermomechanical routes were applied encompassing three steps of wire drawing with area reductions of 24.4%, 17.7% and 15.4%, respectively, followed by solution annealing after each cycle at two different temperatures, by that designated as routes A and B. The commercial drawing dies diameters and geometries are listed in Table 2, according to manufacturer specifications.
The δ-phase solvus temperature was determined through thermodynamic calculation, using the Thermo-calc  2017b software and the TCNI8 database 23 and established as 1030°C. Therefore, the solution annealing temperatures were defined based on the calculated δ-phase solvus temperature data. The as-received material was initially solution heat treated above δ-solvus temperature, at 1050°C. Route A consisted of solution annealing at sub δ-solvus temperature of 975°C, for 15 minutes, after each step of wire drawing. Route B consisted of a step of solution annealing at super δ-solvus temperature of 1050°C, for 30 minutes, after the two first steps of wire drawing. The final step of wire drawing was followed by solution annealing at 975ºC for 75 minutes. This final step under δ−phase solvus temperature was made in order to inhibit grain growth and, consequently, maintain a fine grain size distribution. Ultimately, samples from both routes were subjected to aging heat treatments in order to promote the precipitation of the hardening precipitates γ' and γ", comprising 285 minutes at 760ºC, followed by controlled cooling at rate 0.67ºC /min up to 650ºC, being held for 75 minutes before air cooling. The thermomechanical processing routes A and B are depicted in Figure 1.

Numerical simulation
Numerical simulation sought to compute the resulting effective strain distribution along wire longitudinal and transversal directions, during each wire drawing step. Finite Element Method (FEM) were performed by DEFORM  commercial software package. The asymmetric model is shown in Figure 2, which was created from three-dimensional numerical simulation. Three reductions steps were measured, as described previously in Figure 1.
The first and the second were 3mm long, configuring a longer length in the die's calibration zone. The last reduction was 2.5 mm long. The considered speed of each drawing pass was 100 mm/s. For these analyzes, the same mechanical properties were assumed for the material between drawing passes. It implies that the deformation hardening effect was not taken into consideration from one pass to next, which is adequate since annealing heat treatment was performed between each deformation pass. This work is based on a model that assumes that the wire is a deformable element, therefore it was considered as a tetrahedral volumetric mesh. While the wire drawing die surface was considered as a rigid element, therefore, a triangular mesh was used. FEM was based on materials and processing parameters, which are listed in Table 3. The as-received material was subjected to uniaxial tensile testing, with strain rate of 7.06 *10 -4 s -1 , and the stress versus strain dataset was uploaded in the software database in order to validate the power law. The isotropic hardening rule and the yield function type Von Mises were considered during deformation in this work.

Scanning electron microscopy
Quantification of δ-phase was based on five different fields of image, randomly distributed on sample surface from the center to half the wire radius (1/2R). Vega 3LMU Tescan scanning electron microscope (SEM) was used in the backscattered electron (BSE) mode for atomic contrast, with 20KV high voltage and 2000x nominal magnification. The solution treated sample was analyzed for δ-phase quantification, after grinding up to 1200 mesh, polishing with diamond paste up to 1μm and no etching. SEM/BSE images were processed for δ particles area measurements in Image J, an open source software, through binarization of the original images.

Electron backscattered diffraction
The aged samples were prepared for electron backscattered diffraction (EBSD) imaging through grinding with 1200 mesh sandpaper and chemical electropolishing using double jet technique in a Struers Tenupol 5 polishing apparatus with a solution of 9% perchloric acid (HClO 4 ) and 91% ethanol, at -20 o C temperature, 20.5kV and 45 seconds of exposure time. Electropolishing resulted in a restricted 2mm diameter polished area in each sample, where two distinct fields were   EBSD scanned for each TMP route: in the center and in half of the wire radius (1/2R). EBSD scanning was performed in two different instruments: route A samples were scanned in a CAMSCAN 3200 LV SEM, with EBSD Oxford camera and HKL system; route B samples were scanned in a QUANTA 250 SEM equipped with field emission gun (FEG) and Bruker detectors controlled by CrystAlign software. Both routes were investigated with a step size of 1 μm.
Scans datasets were processed using MTEX 5.1.1, that encompasses quantitative texture analysis 24 . From the EBSD data it was possible to analyze the grain size through the linear intercept method 25 , the identification and classification of CSL boundaries, which was based on Brandon's criteria to define limits to coincidence lattice sites 26 , and the inverse pole figure to microtexture analysis. EBSD data was also used to generate kernel average misorientation (KAM) maps by MTEX. KAM consists in measurements of the average misorientation angle around a pixel with respect to a set of neighboring points, and was used in order to reveal strain distribution after deformation through routes A and B.

Tensile testing
Tensile tests were performed with a 98kN load cell in a EMIC DL10000 machine. The strain rate was 7.06*10 -4 s -1 and the specimens were machined in compliance with ASTM E8/ E8M -16a standard.

Numerical simulation
The effective strain distribution in the symmetric half of the wire, obtained from computational simulation, is shown in Figure 3, where dimensionless strain is shown in the color scale on the left, and the X, Y and Z axes are shown on the bottom of the figure. The columns (a), (b) and (c) represents 1 st , 2 nd and 3 rd steps of drawing. It is possible to observe the heterogeneity on the strain distribution in the wire cross-section, along Z axis. The three steps of drawing presented a deformation gradient from the surface to center of the wire, being the larger deformation adjacent to the surface. Hasani et al. 27 studied the drawing process of a Cooper wire, and observed a significant difference in microhardness values between surface and wire center. That discrepancy could be associated with processing parameters such as die tool geometry, the chosen lubricant and wire area reduction. Understanding the strain distribution originated during drawing process is important to comprehend its effect on microstructural related phenomena, such as: recrystallization, boundary migration mechanism and δ-phase precipitation in alloy 718 28 and, in this regard, the GBCD. Therefore, it is expected that the aforementioned phenomena respond distinctively along the wire cross-section region. In view of this, the present study analyzed two regions of the cross section of the wire: center and 1/2 of radius of wire. Table 4 presents the area fraction for δ-phase for each of the processing routes. Comparing routes A and B, it can be observed that route A presented an overall higher area fraction of δ-phase, which can be associated with the cold deformation cycle prior to a solution annealing step at a sub δ-phase solvus temperature, which leads to an increased precipitation kinetics in every drawing step of route A 28 . Route B, on the other hand, involves intermediate steps of solution annealing over the δ-phase solvus temperature, despite the longer sub δ-phase solvus solution annealing in the final step, which is associated with the smaller amount of deformation, as can be seen in Figure 1. In addition, δ-phase presented different morphologies when comparing route A and route B resultant microstructure. The binarized SEM images of the central area of the wires, evidencing δ-phase in black, are presented in Figure 4. For route A (Figure 4(a)), which involved δ precipitation during the intermediate steps, δ morphology was more fragmented, induced by migration of recrystallized grains, that cause a morphology change from acicular to spheroidal 29 , both at center and at mid-radius. The aligned distribution followed previously deformed grains, as δ tends to precipitate preferentially at grain boundaries. Furthermore, the precipitated δ particles were fragmented in the subsequent drawing steps. Mei et al. 29 reported that, as the degree of prior deformation increases, due to higher dislocation density within the grain, δ-phase particles tends to be more uniformly distributed during the heat treatment. Jeong et al. 30 pointed out that cold drawing would favor the round-shaped particles, as well as a shorter annealing time.

δ-phase area fraction analysis
On the other hand, for route B (Figure 4(b)), where δ precipitation occurred only during the heat treatment after the ulterior drawing cycle, resulting particles were predominantly acicular and formed at recrystallized grain boundaries. The lack of previously fragmented particles and the longer final annealing time would favor the precipitation and growth of the needle-like particles, as observed by Jeong et al. 30 .
Other important aspect to be accounted for is the predominant precipitation of δ-phase closer to the surface than to the wire's center, which occurred for both routes, A and B. It is an evident result of the heterogeneity of strain distribution during drawing processing, which is increased towards the surface. Data is summarized in Table 4.
A consequence of the more intense δ-phase precipitation in the intermediate steps of drawing in route A is a considerably reduction in the average grain size obtained after processing, as can be seen in Table 4. Some authors discuss about that the phenomenon 8,11 and it is attributed to the pinning effect of δ-phase, restricting grain boundary movement. For route B, the intermediate steps were heat treated over the δ-phase solvus temperature, providing rec pinning effect to grain boundary movement. Only after the last deformation step the heat treatment was conducted under the δ solvus temperature. This resulted in a bigger grain size and less δ precipitation. Figure 5 shows the grain boundary character distribution maps, where Ʃ3 boundaries are shown in red, Ʃ9 in green, Ʃ27 in blue and random high angle boundaries (Ʃ>29) are shown in black. The total Σ3 n length fraction for each route, both in the center of the wire and in the 1/2R position, are presented in Table 4. The smaller grain size of route A was accompanied by a lower proportion of Ʃ3 n boundaries. Route B presented 20% more Ʃ3 n boundaries, both for center and 1/2R positions. Randle 14 states that one of the main conditions for the Σ3 n regeneration process to take place is grain boundary mobility. Therefore, route B, which induced δ precipitation only after the ulterior drawing step, allowed higher grain boundary mobility, as no δ particles were introduced in the microstructure after the intermediate annealing heat treatments. It made the Ʃ3 n regeneration process more effective. Nevertheless, Araújo et al. 8 studied iterative processing of alloy 718, comprising cold rolling and solution annealing at 975° for 70 minutes, and they reported an increase in Σ3 n length fraction concomitant with grain refining due to δ-phase precipitation, which was not observed in the present investigation. Figure 6 presents the KAM maps of samples from route A at the wire center (a) and at the 1/2R position (b), and from route B at the center (c) and at 1/2R position (d). Comparing Table 4. Summary of the results regarding δ-phase area fraction (%), average grain size (mm), and Ʃ3 n (Ʃ3 + Ʃ9 + Ʃ27) length fraction (%) calculated for both routes, A and B, both to the center and to the 1/2 R position.    both routes, it can be stated that the strain distribution and, consequently, the degree of recrystallization are heterogeneous for the sample from route A, showing lower degrees of strain concentrations inside the recrystallized grains (which appear in darker blue in Figure 5 (a) and (b)). Route A was annealed at 975°C for 15 minutes between every reduction pass, temperature that favors precipitation of δ-phase. During wire drawing, matrix deforms and dislocations accumulate around the precipitates cyclically, promoting the onset of recrystallization and acting as preferential nucleation sites for more δ. This phenomena annihilates de potential nucleation sites for twins, according to Prithiv et al. 31 .

Mesotexture analysis
The lower Ʃ3 n fraction found in route A is in accordance to that statement. Furthermore, according to Neng-Yong et al. 32 , the complete recrystallization of superalloy 718, subjected to a 30% reduction cold rolling followed by annealing heat treatment at 980°C, is achieved after 30 minutes. Analogous behavior was observed in the present study, which indicates that 15 minutes of annealing time is sufficient only for partial recrystallization, as seen in Figure 6. Nevertheless, the opposite behavior regarding strain distribution was observed on route B, where KAM maps show a very homogenous strain distribution either in the wire center and closer to the surface. It is due to a higher annealing temperature and time between passes and in the final step for route B.
Additionally, as can be seen in Table 4, route A, in particular, presented an increase of around 8% in the length fraction of Σ3 n closer to the wire 1/2R in comparison to the center, even though in the 1/2R position the amount of δ-phase was around 25% higher than in the center. For route B, no significant difference in Σ3 n length fraction between wire center and surface was observed, which can be attributed to the longer solution annealing time after the final step of deformation (975 o C for 75 minutes), leading to complete recrystallization along the entire wire cross section and enough time to the onset and progress of boundary mobility.

Microtexture analyses
The present investigation evaluated microtexture in alloy 718 based on inverse pole figures (IPF) using as reference the <100>//Drawing direction (DD), generated from EBSD mappings data, presented in Figure 7. Both routes, A and B, have shown stronger intensity along the fiber <110> while fibers <111> and <100> are weaker. Samples from route A and B exhibited the <110> fiber stronger at 1/2R and at the center of the wire, however, at the 1/2R of the route B sample the maximum intensity texture was deviated from the ideal <110> position, as it can be seen in Figure 7d. Recrystallization texture was relatively weak, with maximum intensity of 1.8 for the <110> fiber.
Deformation texture in low stacking fault energy (0-60 mJ/m 2 ) FCC materials processed by wire drawing is widely discussed in literature and characterized as concomitant presence of <100> and <111> fibers 33 . The intensity of each fiber depends on the plastic deformation the material is subjected to, so that for deformations below 90%, <111> fiber is stronger 34 . On the other hand, dynamic recovery occurs and leads to a stronger intensity at fiber <100> for more than 90% of deformation 32 . Other complex fibers are mentioned in the literature, with limited explanations on how they also appear in material texture, though. Chen et al. 35 and Ma et al. 33 , subjected Copper and Silver to deformation below 10% reduction to investigate the effect of deformation on the evolution of texture in those materials, respectively. They observed that resulting deformation texture was characterized by complex fibers whereas <111> and <100> fibers were present but weaker. GVK et al. 36 verified the presence of <110> fiber associated with deformations inferior than 50% due to rotation of grains with <111> crystallographic orientation toward <110>. Whitis 37 investigated René DT88 superalloy subjected to deformation degree inferior than 10% and observed intense deformation texture at fibers <111> and <110>, and after recrystallization, <110> fiber became the most intense fiber with 1.8 intensity, which was the case observed in this work Recrystallization texture after drawing in FCC superalloys is still not well established. Although, from the above, it is possible to establish a correlation between the literature and the recrystallization texture obtained in the present investigation for alloy 718. Last drawing step promoted 15.4% of deformation, similar to Whitis' study 37 . Therefore, such amount of deformation tends to form complex fibers as a result of deformation texture in FCC materials after drawing, including <110> fiber. Following the recrystallization, configuration of the fibers did not change and it became the material's final texture. It is likely that <110> fiber is stable for a lower deformation amount and that fact would explain the presence of this fiber in alloy 718. However, further studies are needed since different fiber orientations, other than <100> and <111>, in FCC materials, are little addressed in the literature.

Mechanical properties
Tensile tests results are presented in Table 5. For the solution annealed samples, route A produced the higher values of yield (YS) and ultimate tensile strength (UTS). As in the solution annealed condition there are no γ´ and γ´´ precipitates, the main hardening mechanism is grain refining 38 . In contrast, there is a decrease in ductility of samples from route A, probably related to increased volume fraction of δ-phase precipitated in the intermediate steps of drawing 11 .
In the aged condition, yield and tensile strength increases drastically, while ductility decreases when compared to solution annealed specimen. There is a slight increase in the yield and tensile strength in route B over route A, which is associated with the lower δ-phase precipitation in route B, which derives into the matrix more niobium available for precipitation of γ´´ and, consequently, gives rise to improved mechanical properties 7,39 . On the other hand, there is no significant difference in ductility for routes A and B in the aged condition.

Conclusions
Based on the results it is possible to conclude: • The processing with no precipitation of δ−phase in the intermediate steps (route B) was more effective to induce Ʃ3 regeneration mechanism and, consequently, a higher fraction of Ʃ3 n boundaries. However, this improved boundary mobility lead to a higher grain size than the route with δ precipitation in the intermediate steps;

•
The combination of deformation and the sub-solvus δ heat treatment during the intermediate steps promoted relevant δ precipitation, pinning grain boundary movement and resulting in refined grains; • The wire drawing simulation indicated a deformation gradient from the center to the surface of the wire which was corroborated by the microstructural changes of the cross section of the wire; • The deformation gradient was associated with variation in the δ precipitation when comparing center and 1/2R positions. However, regarding the GBCD, only a slight variation was evidenced for route A; • It was shown that the deformation during the process has influence on the final texture of the material leading to the formation of <110> fiber texture, which was deviated from the common fibers <100> and <111> for wire drawn FCC materials. The route A, annealed for 15 minutes, presented greater difference in intensity of texture between 1/2R and the center of the wire which is associated with the partial recrystallization associated with the low heat treatment time; • The intermediate δ precipitation and consequent grain refinement of route A resulted in a higher strength for the solution annealed condition. However, despite the smaller grain size, the elongation was inferior to route B, being related to the δ precipitation and its role on cavitation and crack propagation. For the aged condition, route B presented the higher strength in comparison to route A. This was related to the fact that the lower precipitation of δ made more Nb available for γ´´ precipitation.