The Effect of Phase Separation on the Mechanical Behavior of the Co–Cr–Cu–Fe–Ni High-Entropy Alloy

Phase separation phenomena in high-entropy alloys (HEAs) have attracted much attention since their discovery, but little attention has been given to the dynamics of the deformation mechanism of this kind of HEA during uniaxial tension, which limits their widespread and practical utility. In this work, molecular dynamics simulation was used to study the effect of phase separation on the mechanical properties of an HEA under uniaxial tensile loading. Moreover, the associated deformation behavior of the Co–Cr–Cu–Fe–Ni HEA was investigated at the nanoscale. Models with Cu-rich grain boundaries or grains were constructed. The results showed that Cu-rich grain boundaries or grains lowered the strength of the Co–Cr–Cu–Fe–Ni HEA, and Cu-rich grain boundaries significantly reduced ductility. This change of mechanical properties was closely associated with a deformation behavior. Furthermore, the deformation behavior was affected by the critical resolved shear stress of Cu-rich and Cu-depleted regions and the uneven stress distribution caused by phase separation. In addition, dislocation slipping and grain boundary sliding were the main mechanisms of plastic deformation in the Co–Cr–Cu–Fe–Ni HEA.


Introduction
High-entropy alloys (HEAs) contain multiple principal (or major) elements, commonly five or more, in equimolar or near-equimolar ratios, differently from conventional alloys. While the configurational entropy of an alloy increases with the increasing number of alloying elements, multiple principal elements of an HEA lead to high entropy of mixing that thus suppresses the formation of intermetallic compounds [1][2][3]. Therefore, stable solid solution phases, such as face-centered cubic (FCC) phase, body-centered cubic (BCC) phase, and hexagonal close-packed (HCP) phase can be obtained [4,5]. Meanwhile, HEAs have been shown to exhibit outstanding mechanical and chemical properties, such as high strength and hardness [1,6], exceptional ductility, as well as excellent oxidation and corrosion resistances [7], and thus have become potential candidate materials for many challenging industrial applications. The design and application of HEAs require a comprehensive understanding of their properties and underlying deformation mechanisms. Co-Cr-Cu-Fe-Ni, with a simple FCC structure [8], is the earliest discovered quinary highentropy alloy [9,10] and also one of the most studied high-entropy alloy systems. Al, Nb, Ti, and Zr are specially added to improve the properties of the Co-Cr-Cu-Fe-Ni HEA [11][12][13][14]. Recently, the Co-Cr-Cu-Fe-Ni HEA system has received more and more attention because it undergoes the interesting liquid-phase separation (LPS) phenomenon [15][16][17][18][19].
LPS was studied by Nakagawa in Cu-Co and Cu-Fe binary systems for the first time [20]. Up to now, it has been observed in a variety of materials, including metals,  Figure 2 shows the tensile stress-strain curves of the random composition distribution (RCD) samples at room temperature. In the elastic stage, different grain orientations [45] would cause the elastic modulus of the 2D-RCD sample to deviate from the value of the 3D-RCD sample. The different grain orientations and 3D grain boundary lowered the yield strength of the 3D-RCD sample. At the plastic stage, it was seen that, compared with the 2D-RCD sample, the strain softening resistance of the 3D-RCD sample was better. This was due to the formation of multiple deformation paths in the 3D-RCD sample. The intersections (Figure 3b black arrows) of these plastic slip deformation paths inhibited atoms' The equilibrium state of the HEA samples could be achieved with the following three steps: first, the as-created Co-Cr-Cu-Fe-Ni HEA samples initially underwent energy minimization; second, the samples were relaxed for 2 ns at the hydrostatic pressure of 5 GPa and 300 K under the NPT (constant number of atoms (N), constant pressure (P), and constant temperature (T)) ensemble; finally, they were relaxed subsequently for 2 ns at the hydrostatic pressure of 0 GPa and 300 K under the NPT ensemble. After reaching equilibrium configurations, uniaxial tensile deformation along the X-direction was applied at 300 K and constant strain rate of 4.0 × 10 7 s −1 . In this study, the software OVITO [42] was employed to visualize or analyze the data generated by the MD simulations. The adaptive common-neighbor analysis (CNA) algorithm [43] was used to characterize the local atomic structural environment, in which the green atoms indicate the FCC structure, the blue atoms represent the BCC structure, the red atoms stand for stacking faults, and the white atoms suggest grain boundaries or dislocation cores. Meanwhile, an atomic strain modifier [44] was used to calculate the atomic level deformation gradient; displacement vectors modifier indicated the glide directions of atoms in certain close-packed planes. The local atomic strains were referenced to the relaxed samples before the tensile test. Figure 2 shows the tensile stress-strain curves of the random composition distribution (RCD) samples at room temperature. In the elastic stage, different grain orientations [45] would cause the elastic modulus of the 2D-RCD sample to deviate from the value of the 3D-RCD sample. The different grain orientations and 3D grain boundary lowered the yield strength of the 3D-RCD sample. At the plastic stage, it was seen that, compared with the 2D-RCD sample, the strain softening resistance of the 3D-RCD sample was better. This was due to the formation of multiple deformation paths in the 3D-RCD sample. The intersections (Figure 3b black arrows) of these plastic slip deformation paths inhibited atoms' sliding in certain close-packed planes and acted as an obstacle to dislocation motion [46], resulting in a slow drop and small fluctuation of the flow stress of the 3D-RCD sample. As expected, the relatively high loading rate of 4.0 × 10 7 s −1 due to the timescale limitations of MD, the ideal crystal structure, and the extreme purity of the sample led to considerably higher yield stress (in GPa) as compared to that observed in conventional tensile experiments (limited to few hundred MPa) [47,48]. Moreover, the frequent drops and rises of the stress-strain curves were caused by the competition between plastic deformation and strain hardening, which were observed experimentally [47]. sliding in certain close-packed planes and acted as an obstacle to dislocation motion [46], resulting in a slow drop and small fluctuation of the flow stress of the 3D-RCD sample. As expected, the relatively high loading rate of 4.0 × 10 7 s −1 due to the timescale limitations of MD, the ideal crystal structure, and the extreme purity of the sample led to considerably higher yield stress (in GPa) as compared to that observed in conventional tensile experiments (limited to few hundred MPa) [47,48]. Moreover, the frequent drops and rises of the stress-strain curves were caused by the competition between plastic deformation and strain hardening, which were observed experimentally [47].  To better understand the deformation behavior, the microstructural evolution of the 2D-RCD sample was examined. According to the microstructure evolution visualized by CNA and the atomic strain distribution of the 2D-RCD sample, presented in Figure 4, in the elastic stage (0-4.7%), the strain accumulated mainly in grain boundaries (Figure 4e). sliding in certain close-packed planes and acted as an obstacle to dislocation motion [46], resulting in a slow drop and small fluctuation of the flow stress of the 3D-RCD sample. As expected, the relatively high loading rate of 4.0 × 10 7 s −1 due to the timescale limitations of MD, the ideal crystal structure, and the extreme purity of the sample led to considerably higher yield stress (in GPa) as compared to that observed in conventional tensile experiments (limited to few hundred MPa) [47,48]. Moreover, the frequent drops and rises of the stress-strain curves were caused by the competition between plastic deformation and strain hardening, which were observed experimentally [47].  To better understand the deformation behavior, the microstructural evolution of the 2D-RCD sample was examined. According to the microstructure evolution visualized by CNA and the atomic strain distribution of the 2D-RCD sample, presented in Figure 4, in the elastic stage (0-4.7%), the strain accumulated mainly in grain boundaries (Figure 4e). To better understand the deformation behavior, the microstructural evolution of the 2D-RCD sample was examined. According to the microstructure evolution visualized by CNA and the atomic strain distribution of the 2D-RCD sample, presented in Figure 4, in the elastic stage (0-4.7%), the strain accumulated mainly in grain boundaries (Figure 4e).

Random Composition Distribution
In the plastic stage, due to the different orientation of each grain, the resolved shear stress along the slip direction of the grains, being in a favorable orientation, reached the critical value earlier, and plastic deformation occurred. The sample began to yield, which means that dislocations started to occur and continued along the slip planes, generating double-layer HCP structures in the form of intrinsic stacking fault (ISF) (Figure 4 black arrows). Dislocation slipping is accompanied by grain rotation, which is induced by grain boundary sliding [46]. When the strain increased from 4.8% to 5.1%, the in-plane rotation angles of grains G1 and G2 indicated in Figure 5b were approximately 2.17 • and 0.94 • , respectively. It indicated that dislocation slipping and grain boundary sliding dominated the primary plastic behavior. When the surrounding grains have different orientations, the dislocations cannot cross the grain boundaries and thus pile up at the grain boundaries [49]. The piling up of the dislocation resulted in high stress, which caused the formation of some slip systems in adjacent grains and other grains with favorable orientation (Figure 4c), thus accommodating the stress and restarting the dislocation of the original grains, as seen in Figure 4d. In this way, plastic deformation was transferred between adjacent grains and finally spread to the entire sample. In the plastic stage, due to the different orientation of each grain, the resolved shear stress along the slip direction of the grains, being in a favorable orientation, reached the critical value earlier, and plastic deformation occurred. The sample began to yield, which means that dislocations started to occur and continued along the slip planes, generating double-layer HCP structures in the form of intrinsic stacking fault (ISF) (Figure 4 black arrows). Dislocation slipping is accompanied by grain rotation, which is induced by grain boundary sliding [46]. When the strain increased from 4.8% to 5.1%, the in-plane rotation angles of grains G1 and G2 indicated in Figure 5b were approximately 2.17° and 0.94°, respectively. It indicated that dislocation slipping and grain boundary sliding dominated the primary plastic behavior. When the surrounding grains have different orientations, the dislocations cannot cross the grain boundaries and thus pile up at the grain boundaries [49]. The piling up of the dislocation resulted in high stress, which caused the formation of some slip systems in adjacent grains and other grains with favorable orientation ( Figure  4c), thus accommodating the stress and restarting the dislocation of the original grains, as seen in Figure 4d. In this way, plastic deformation was transferred between adjacent grains and finally spread to the entire sample.
To further investigate the characteristics of the slip systems in the grains, we selected the stacking faults from the 2D-RCD sample at a strain of 4.9%; only atoms of the stacking faults are displayed in Figure 5. From the snapshot, it is evident that the activated slip systems in the grains were <112>{111} type ( Figure 5b). The reasons why the slip system was <112>{111} rather than the common <110>{111} are as follows: on the one hand, direct slipping along the direction of <110> (Figure 5c-b1) resulted in significant collision with the adjacent atom A, causing large local lattice distortion and a significant increase in energy; on the other hand, the lattice distortion caused by the distribution of component atoms with different sizes [48] made slipping along the <110> direction more difficult. Moreover, lattice distortion also reduces stacking fault energy (SFE) [50], which makes it easier for a full dislocation to split into two partials with a wider stacking fault ribbon between them [51]. Therefore, the ideal slip path was along the <112> direction, arriving at position C first (Figure 5c-b2) and then getting to the adjacent position B (Figure 5c-b3). In a word, the <112>{111} system was more easily activated than the <110>{111} system.

Cu-Rich Grain Boundaries
In order to investigate the effect of composition segregation on the mechanical properties under uniaxial tensile loading and the associated deformation behavior of Co-Cr-Cu-Fe-Ni HEA, a series of MD models were constructed by substituting Cu atoms for Co (Cr, Cu, Fe, Ni) in specified regions (grain boundary or grain) and maintaining the desired equiatomic composition of the whole samples. Therefore, Cu-rich and Cu-depleted regions would separate in the samples. By means of the above method, samples with Curich grain boundaries (CRGB sample, Figure 6) and Cu-rich grains (CRG sample in Section 3.3) were obtained. Moreover, according to the experimental data [16], the Cu content in the Cu-rich regions was as high as 80%. To further investigate the characteristics of the slip systems in the grains, we selected the stacking faults from the 2D-RCD sample at a strain of 4.9%; only atoms of the stacking faults are displayed in Figure 5. From the snapshot, it is evident that the activated slip systems in the grains were <112>{111} type (Figure 5b). The reasons why the slip system was <112>{111} rather than the common <110>{111} are as follows: on the one hand, direct slipping along the direction of <110> (Figure 5c-b 1 ) resulted in significant collision with the adjacent atom A, causing large local lattice distortion and a significant increase in energy; on the other hand, the lattice distortion caused by the distribution of component atoms with different sizes [48] made slipping along the <110> direction more difficult. Moreover, lattice distortion also reduces stacking fault energy (SFE) [50], which makes it easier for a full dislocation to split into two partials with a wider stacking fault ribbon between them [51]. Therefore, the ideal slip path was along the <112> direction, arriving at position C first (Figure 5c-b 2 ) and then getting to the adjacent position B (Figure 5c-b 3 ). In a word, the <112>{111} system was more easily activated than the <110>{111} system.

Cu-Rich Grain Boundaries
In order to investigate the effect of composition segregation on the mechanical properties under uniaxial tensile loading and the associated deformation behavior of Co-Cr- Cu-Fe-Ni HEA, a series of MD models were constructed by substituting Cu atoms for Co (Cr, Cu, Fe, Ni) in specified regions (grain boundary or grain) and maintaining the desired equiatomic composition of the whole samples. Therefore, Cu-rich and Cu-depleted regions would separate in the samples. By means of the above method, samples with Cu-rich grain boundaries (CRGB sample, Figure 6) and Cu-rich grains (CRG sample in Section 3.3) were obtained. Moreover, according to the experimental data [16], the Cu content in the Cu-rich regions was as high as 80%.

Cu-Rich Grain Boundaries
In order to investigate the effect of composition segregation on the mechanical properties under uniaxial tensile loading and the associated deformation behavior of Co-Cr-Cu-Fe-Ni HEA, a series of MD models were constructed by substituting Cu atoms for Co (Cr, Cu, Fe, Ni) in specified regions (grain boundary or grain) and maintaining the desired equiatomic composition of the whole samples. Therefore, Cu-rich and Cu-depleted regions would separate in the samples. By means of the above method, samples with Curich grain boundaries (CRGB sample, Figure 6) and Cu-rich grains (CRG sample in Section 3.3) were obtained. Moreover, according to the experimental data [16], the Cu content in the Cu-rich regions was as high as 80%. For Cu-rich grain boundaries, the corresponding stress-strain curves of the 2D and 3D samples are shown in Figure 7. In Figure 7a, it can be observed that, compared with the 2D-RCD sample, the yield strength and plasticity of the 2D-CRGB sample were simultaneously decreased at the constant strain rate of 4.0 × 10 7 s −1 . Especially, regarding plasticity, when the strain reached 7.4%, intergranular fracture occurred (see Figure 8c). As for the 3D-CRGB sample, its strength and flow stress obviously decreased compared with those of the 3D-RCD sample (Figure 7b). For Cu-rich grain boundaries, the corresponding stress-strain curves of the 2D and 3D samples are shown in Figure 7. In Figure 7a, it can be observed that, compared with the 2D-RCD sample, the yield strength and plasticity of the 2D-CRGB sample were simultaneously decreased at the constant strain rate of 4.0 × 10 7 s −1 . Especially, regarding plasticity, when the strain reached 7.4%, intergranular fracture occurred (see Figure 8c). As for the 3D-CRGB sample, its strength and flow stress obviously decreased compared with those of the 3D-RCD sample (Figure 7b).  Figure 8 shows the microstructure evolution and the local atomic strain distribution of the 2D-CRGB sample. In the elastic stage (0-4.2%), the local atomic shear strain was concentrated on grain boundaries (Figure 8a,d), similar to the elastic deformation of the 2D-RCD sample. However, differently from the 2D-RCD sample, after yielding, a small dislocation slip within grains occurred (Figure 8b,e), and the local stress concentration was relaxed through grain boundary cracking in the 2D-CRGB sample. The reason for this phenomenon may be that, compared with Cu-depleted regions inside the grains, Cu-rich regions at the grain boundaries have a low strength. Furthermore, atoms in grain bound- phenomenon may be that, compared with Cu-depleted regions inside the grains, Cu-rich regions at the grain boundaries have a low strength. Furthermore, atoms in grain boundaries are arranged irregularly, which means they have larger potential energy than those in grains [52]. As a result, the grain boundaries cracked before plastic deformation in the grains could relax the local stress concentrations on the grain boundaries. The microstructure evolution of the 3D-CRGB sample was examined. The results showed that dislocation slip also occurred inside the grains of the 3D-CRGB sample, but the dislocation density was lower than that of the 3D-RCD sample (Figure 9 top panel). Moreover, the local strain was more seriously concentrated on grain boundaries in 3D-CRGB samples (Figure 9 bottom panel), similar to what observed for the 2D samples. This high concentration of strain on the grain boundaries was the reason for the decrease of flow stress in the 3D-CRGB sample. When the strain reached 10%, obvious intergranular  Figure 8 shows the microstructure evolution and the local atomic strain distribution of the 2D-CRGB sample. In the elastic stage (0-4.2%), the local atomic shear strain was concentrated on grain boundaries (Figure 8a,d), similar to the elastic deformation of the 2D-RCD sample. However, differently from the 2D-RCD sample, after yielding, a small dislocation slip within grains occurred (Figure 8b,e), and the local stress concentration was relaxed through grain boundary cracking in the 2D-CRGB sample. The reason for this phenomenon may be that, compared with Cu-depleted regions inside the grains, Curich regions at the grain boundaries have a low strength. Furthermore, atoms in grain boundaries are arranged irregularly, which means they have larger potential energy than those in grains [52]. As a result, the grain boundaries cracked before plastic deformation in the grains could relax the local stress concentrations on the grain boundaries.
The microstructure evolution of the 3D-CRGB sample was examined. The results showed that dislocation slip also occurred inside the grains of the 3D-CRGB sample, but the dislocation density was lower than that of the 3D-RCD sample (Figure 9 top panel). Moreover, the local strain was more seriously concentrated on grain boundaries in 3D-CRGB samples (Figure 9 bottom panel), similar to what observed for the 2D samples. This high concentration of strain on the grain boundaries was the reason for the decrease of flow stress in the 3D-CRGB sample. When the strain reached 10%, obvious intergranular fracture occurred in the 3D-CRGB sample (Figure 9 black arrow). The intergranular fracture started at a strain of 7.8%, which led to a sudden decrease in the flow stress of the 3D-CRGB sample. fracture occurred in the 3D-CRGB sample (Figure 9 black arrow). The intergranular fracture started at a strain of 7.8%, which led to a sudden decrease in the flow stress of the 3D-CRGB sample. Figure 9. Microstructure evolution (a-d) and local atomic strain distribution (e-h) for the 3D-CRGB and 3D-RCD samples at strains of 7% and 10%, respectively.

Cu-Rich Grains
As shown in Figure 10, a series of Cu-rich nanograin samples were constructed. Figure 11a shows the tensile stress-strain curves of the 2D-RCD and CRG samples. As shown in Figure 11a, compared with the 2D-RCD sample, the strength of the 2D-CRG samples was obviously reduced. In addition, the higher the number of Cu-rich grains, the more obvious the decrease of the strength, indicating that the Cu-rich regions distributed in the grains lowered the strength of the samples. This is consistent with previously reported experimental results [18]. As for the 3D-CRG samples, their strength also decreased compared with the 3D-RCD sample (Figure 11b).

Cu-Rich Grains
As shown in Figure 10, a series of Cu-rich nanograin samples were constructed. Figure 11a shows the tensile stress-strain curves of the 2D-RCD and CRG samples. As shown in Figure 11a, compared with the 2D-RCD sample, the strength of the 2D-CRG samples was obviously reduced. In addition, the higher the number of Cu-rich grains, the more obvious the decrease of the strength, indicating that the Cu-rich regions distributed in the grains lowered the strength of the samples. This is consistent with previously reported experimental results [18]. As for the 3D-CRG samples, their strength also decreased compared with the 3D-RCD sample (Figure 11b). ture started at a strain of 7.8%, which led to a sudden decrease in the flow stress of the 3D-CRGB sample. icrostructure evolution (a-d) and local atomic strain distribution (e-h) for the 3D-CRGB and 3D-RCD samples f 7% and 10%, respectively.

Cu-Rich Grains
As shown in Figure 10, a series of Cu-rich nanograin samples were constructed. Figure 11a shows the tensile stress-strain curves of the 2D-RCD and CRG samples. As shown in Figure 11a, compared with the 2D-RCD sample, the strength of the 2D-CRG samples was obviously reduced. In addition, the higher the number of Cu-rich grains, the more obvious the decrease of the strength, indicating that the Cu-rich regions distributed in the grains lowered the strength of the samples. This is consistent with previously reported experimental results [18]. As for the 3D-CRG samples, their strength also decreased compared with the 3D-RCD sample (Figure 11b).  We also investigated microstructure evolution and local atomic strain distribution for the 2D-CRG samples due to their tensile deformation, presented in Figure 12. Compared with the 2D-RCD sample, except for the 2D-CRG sample 1, the 2D-CRG samples 2-4 yielded from Cu-rich grain, leading to plastic deformation. With the increase of deformation, the strain seemed more concentrated on the Cu-rich regions ( Figure 12 right column), different from the local strain distribution of the 2D-RCD sample under the same strain. This was because, compared with the Cu-depleted grains, the strength of the Curich grains was lower. In other words, the yield tended to start from the Cu-rich grain first, which led to a decrease of the strength of the 2D-CRG samples 2-4. As for the CRG sample 1, the yield started from the Cu-depleted grain, which was caused by the uneven stress distribution. Therefore, compared with the 2D-CRG samples 2-4, the strength of the 2D-CRG sample 1 did not decrease significantly (Figure 11b). Detailed explanations are reported below. We also investigated microstructure evolution and local atomic strain distribution for the 2D-CRG samples due to their tensile deformation, presented in Figure 12. Compared with the 2D-RCD sample, except for the 2D-CRG sample 1, the 2D-CRG samples 2-4 yielded from Cu-rich grain, leading to plastic deformation. With the increase of deformation, the strain seemed more concentrated on the Cu-rich regions ( Figure 12 right column), different from the local strain distribution of the 2D-RCD sample under the same strain. This was because, compared with the Cu-depleted grains, the strength of the Cu-rich grains was lower. In other words, the yield tended to start from the Cu-rich grain first, which led to a decrease of the strength of the 2D-CRG samples 2-4. As for the CRG sample 1, the yield started from the Cu-depleted grain, which was caused by the uneven stress distribution. Therefore, compared with the 2D-CRG samples 2-4, the strength of the 2D-CRG sample 1 did not decrease significantly (Figure 11b). Detailed explanations are reported below.
In order to further reveal the origin of the deformation behavior of the 2D-CRG samples 1-4, Schmid's law was adopted. According to Schmid's law [53], slip systems in grains can only be activated if the resolved shear stresses along the slip directions in the applied stress field are greater than the critical resolved shear stress (CRSS). Therefore, under the condition that the slip system is known, the critical resolved shear stress τ y can be expressed as τ y = µ·σ y , where µ is the Schmid factor, and σ y is the yield strength (GPa).
To provide a clearer description as to the calculation process of CRSS, a schematic illustration is shown in Figure 13. First, atoms in a cube were selected from the inside of a grain (Cu-rich or Cu-depleted), and these selected atoms were tracked during the whole deformation process of the sample (Figure 13a). Cu-rich and Cu-depleted grains where the dislocation slip first occurred in the sample were selected. Second, the stress-strain curves (as shown in the top panel in Figure 14) of the selected cubes (the bottom panel in Figure 14) were calculated to determine the yield stresses of these Cu-rich and Cu-depleted grains. Finally, the slip directions and the slip planes of the slip systems inside the cubes were determined (Figure 13c) to calculate the Schmid factors. Therefore, CRSS could also be calculated. The detailed results are presented in Table 1. In order to further reveal the origin of the deformation behavior of the 2D-CRG samples 1-4, Schmid's law was adopted. According to Schmid's law [53], slip systems in grains can only be activated if the resolved shear stresses along the slip directions in the applied stress field are greater than the critical resolved shear stress (CRSS). Therefore, under the condition that the slip system is known, the critical resolved shear stress can be expressed as = • , where is the Schmid factor, and is the yield strength (GPa).
To provide a clearer description as to the calculation process of CRSS, a schematic illustration is shown in Figure 13. First, atoms in a cube were selected from the inside of a grain (Cu-rich or Cu-depleted), and these selected atoms were tracked during the whole deformation process of the sample (Figure 13a). Cu-rich and Cu-depleted grains where the dislocation slip first occurred in the sample were selected. Second, the stress-strain curves (as shown in the top panel in Figure 14) of the selected cubes (the bottom panel in Figure 14) were calculated to determine the yield stresses of these Cu-rich and Cu-depleted grains. Finally, the slip directions and the slip planes of the slip systems inside the 14, x FOR PEER REVIEW 11 of 15 cubes were determined (Figure 13c) to calculate the Schmid factors. Therefore, CRSS could also be calculated. The detailed results are presented in Table 1.     It can be clearly seen from Table 1 that the CRSS of the Cu-depleted grains w greater than that of the Cu-rich grains. It means that the slip systems in the Cu-rich gra were more easily activated. In other words, plastic deformations tended to occur in Cu-rich grains first. Therefore, all the 2D-CRG samples 2-4 yielded from Cu-rich gra ( Figure 12). According to Figure 14, the stress-strain curves of Cu-rich grains in the 2 CRG samples 2-4 reached the yield point first, while in the 2D-CRG sample 1, those Cu-depleted grains yielded first. This confirmed the yield phenomenon of the 2D-C samples 1-4 observed in Figure 12. In addition, it can be seen from the stress-strain cur in Figure 14 that, compared with the stresses received by the Cu-rich grains, the stres received by the Cu-depleted grains were much higher. This uneven stress distribut made it possible for the sample to yield from the Cu-depleted grain, as observed for 2D-CRG sample 1 (Figure 12). This is why, compared with the 2D-CRG samples 2-4, strength of the 2D-CRG sample 1 did not decrease significantly. The results in Table 1 a  It can be clearly seen from Table 1 that the CRSS of the Cu-depleted grains was greater than that of the Cu-rich grains. It means that the slip systems in the Cu-rich grains were more easily activated. In other words, plastic deformations tended to occur in the Cu-rich grains first. Therefore, all the 2D-CRG samples 2-4 yielded from Cu-rich grains ( Figure 12). According to Figure 14, the stress-strain curves of Cu-rich grains in the 2D-CRG samples 2-4 reached the yield point first, while in the 2D-CRG sample 1, those of Cu-depleted grains yielded first. This confirmed the yield phenomenon of the 2D-CRG samples 1-4 observed in Figure 12. In addition, it can be seen from the stress-strain curves in Figure 14 that, compared with the stresses received by the Cu-rich grains, the stresses received by the Cu-depleted grains were much higher. This uneven stress distribution made it possible for the sample to yield from the Cu-depleted grain, as observed for the 2D-CRG sample 1 ( Figure 12). This is why, compared with the 2D-CRG samples 2-4, the strength of the 2D-CRG sample 1 did not decrease significantly. The results in Table 1 also prove that the strength of the Cu-rich regions was much lower than that of the Cu-depleted regions, which made the macroscopic strength of the 2D-CRGB sample and 2D-CRG samples 1-4 decrease. This is consistent with the results of the tensile tests reported above (Figure 11). Figure 15 shows the atomic snapshots of the 3D-CRG samples 1-2. It was found that the 3D-CRG samples 1-2 yield from Cu-depleted grains, as observed for the 2D-CRG sample 1. Therefore, the decrease in strength of the 3D-CRG samples 1-2 was not as obvious as that of the 2D-CRG samples 2-4 ( Figure 11b). CRG samples 1-4 decrease. This is consistent with the results of the tensile above ( Figure 11). Figure 15 shows the atomic snapshots of the 3D-CRG s was found that the 3D-CRG samples 1-2 yield from Cu-depleted grains, a the 2D-CRG sample 1. Therefore, the decrease in strength of the 3D-CRG sa not as obvious as that of the 2D-CRG samples 2-4 ( Figure 11b).

Conclusions
In summary, we investigated the effect of Cu-rich grain boundarie grains on the mechanical properties of the Co-Cr-Cu-Fe-Ni HEA. The ass mation behavior was also studied by using MD simulation. Moreover, mechanism was presented.
Dislocation slipping and grain boundary sliding were the main deform nisms in the Co-Cr-Cu-Fe-Ni HEA. The internal slip system of the Co-HEA appeared to be mainly of the <112>{111} type. Cu-rich grain boundarie reduce the plasticity of the Co-Cr-Cu-Fe-Ni HEA, because they can lead to fracture of CRGB samples. At the same time, the strength of the alloy was sl by Cu-rich grain boundaries. Cu-rich grains can seriously reduce the stren Cr-Cu-Fe-Ni HEA. The larger the volume fraction of Cu-rich grains is, the the decrease of strength. The reason is that the resolved shear stress requir systems in Cu-rich grains is much smaller than that required in Cu-depl other words, plastic deformation tends to occur in Cu-rich grains. Furtherm stress distribution was detected, where the stress on Cu-depleted grains wa than that on Cu-rich grains. The mechanical properties of the Co-Cr-Cu-F obviously affected by phase separation, which provides a basis for improvi ical properties. This investigation can help to expand the practical applicati of alloy.

Conclusions
In summary, we investigated the effect of Cu-rich grain boundaries and Cu-rich grains on the mechanical properties of the Co-Cr-Cu-Fe-Ni HEA. The associated deformation behavior was also studied by using MD simulation. Moreover, a deformation mechanism was presented.
Dislocation slipping and grain boundary sliding were the main deformation mechanisms in the Co-Cr-Cu-Fe-Ni HEA. The internal slip system of the Co-Cr-Cu-Fe-Ni HEA appeared to be mainly of the <112>{111} type. Cu-rich grain boundaries can seriously reduce the plasticity of the Co-Cr-Cu-Fe-Ni HEA, because they can lead to intergranular fracture of CRGB samples. At the same time, the strength of the alloy was slightly reduced by Cu-rich grain boundaries. Cu-rich grains can seriously reduce the strength of the Co-Cr-Cu-Fe-Ni HEA. The larger the volume fraction of Cu-rich grains is, the more obvious the decrease of strength. The reason is that the resolved shear stress required to start slip systems in Cu-rich grains is much smaller than that required in Cu-depleted grains. In other words, plastic deformation tends to occur in Cu-rich grains. Furthermore, an uneven stress distribution was detected, where the stress on Cu-depleted grains was much higher than that on Cu-rich grains. The mechanical properties of the Co-Cr-Cu-Fe-Ni HEA are obviously affected by phase separation, which provides a basis for improving its mechanical properties. This investigation can help to expand the practical application of this kind of alloy.