Dynamic Behavior of Additively Manufactured FeCoCrNi High Entropy Alloy

Abstract

Additively manufactured face-centered-cubic high entropy alloys have a combination of high strength and good ductility, and are promising impact-resistant structural materials. However, the dynamic behavior of additively manufactured face-centered-cubic high entropy alloys is seldomly reported. In this study, FeCoCrNi high entropy alloy was fabricated, using the laser beam powder bed fusion technique, and dynamic tests were performed by means of a Split Hopkinson Pressure Bar. The high entropy alloy showed a more excellent combination of yield stress and toughness at high strain rates, than previously reported alloys. This was attributed to the dislocation cell structure of the additively manufactured FeCoCrNi HEA, which provided high local stress concentration, leading to the formation of microbands and deformation twins. The high entropy alloy showed higher strain rate sensitivity than the cast counterpart, at both quasi-static and strain rates over 3000 s⁻¹. Interestingly, the yield stress kept stable at a strain rate from 1000 s⁻¹ to 3000 s⁻¹, showed a steep decrease of strain rate sensitivity and a four-fold increase in activation volume, implying a transition in deformation mechanism to collective dislocation nucleation.

1. Introduction

High entropy alloys (HEAs), also known as multiple principal component alloys, have broken the boundaries of traditional alloy design strategy [1], and have given rise to many alloys with exceptional properties compared to conventional alloys, which contain only one principal element [2,3]. In particular, the NiCoCrFeMn high entropy alloys possess fracture toughness as high as 200 MPa·m^1/2 at 77 K [3]. Aside from quasi-static properties, Zhang et al. [4] investigated the dynamic tensile mechanism of the NiCoCrFe high entropy alloy and found that its strength and ductility were enhanced with increased strain rates, which could be attributed to significant strain rate sensitivity and extraordinary work-hardening capacity. Also, the CrMnFeCoNi high entropy alloy showed high energy absorption and strain hardening ability, due to various deformation mechanisms, including stacking faults, twins, phase transformation and amorphization [5]. Thus, HEAs are prospective materials in engineering applications, such as in the automobile and aerospace industries for anti-collision. However, the coarse-grained single-phase FCC alloys usually possess a low yield strength, which is worth further improvement in the context of energy saving. Traditional methods, including cold rolling and forging, introduce high dislocation density and grain refinement in HEAs, which significantly elevate their strength, but may deteriorate their ductility [6].

Recently, the laser beam powder bed fusion technique has proved an effective way to produce metallic alloys with both high strength and ductility [7,8]. Brif et al. [9] first produced equal-atomic FeCoCrNi HEA by laser beam powder bed fusion, and found that the alloy was of a single-phase solid-solution with outstanding mechanical properties comparable to stainless steels. Zhu et al. [10] investigated the dynamic behavior of the additively manufactured CoCrFeNiMn HEA, which possessed a hierarchical structure, including melt pools, columnar grains and sub-micron dislocation cell structures, and also possessed both high strength and good ductility [10]. The additivelymanufactured FeCoCrNi HEAs had better quasi-static mechanical properties than the cast HEAs and conventional alloys [11]. Thus, they are promising impact-resistant structural materials. However, their dynamic properties and deformation mechanisms are seldom studied. How the dislocation structures evolve and affect the dynamic properties at high strain rates are still unknown.

In the present study, the laser beam powder bed fusion technique was used to produce FeCoCrNi HEA. Dynamic compression experiments were subsequently performed on the as-fabricated FeCoCrNi HEA, as well as the cast specimens for comparison. The microstructural evolution was characterized by using electron backscatter diffraction and transmission electron microscopy in order to reveal the possible deformation mechanisms under dynamic loading.

2. Materials and Methods

All the raw materials used had purity larger than 99.9 wt.%. The pre-alloyed FeCoCrNi powder was produced in Central South University with a gas atomization instrument (PSI, UK). After sieving, the powder showed an average size of 18 μm (D₁₀ = 7.02 μm, D₅₀ = 14.2 μm, D₉₀ = 27.1 μm). The powder had a nearly equal atomic composition of Fe_25.14Co_24.95Cr_25.53Ni_24.38. A selective laser melting (SLM) machine (FS271M), with the power bed fusion technique, was used to fabricate the FeCoCrNi HEA. We noted that the laser power was set at 400 W, the scan speed at 1200 mm/s, the beam size at 120 μm, and the layer thickness at 30 μm. Argon was used as the protection atmosphere. The laser rotated 67° after melting of each layer. In order to release the fabrication introduced residual stress, the samples were annealed at 673 K for 3 h and cooled by water quenching. The production condition was chosen based on our group’s previous research, which provided a fully dense AMed HEA [11]. For comparison, vacuum arc melting was used to produce the coarse grain FeCoCrNi specimens. The ingot was flipped and melted 4 times to ensure chemical homogeneity, then cast into a copper mold with the dimensions of 16 × 16 × 100 mm³. After homogenization at 1423 K for 4 h, the ingot was subjected to 60% cold rolling. The rolled sheet was heat treated at 1273 K for 1 h to obtain a fully recrystallized microstructure. The specimens made by selective laser melting and casting were thereafter named As-SLM and As-Cast, respectively. Cylindrical specimens with 4 mm in diameter and 3 mm in height were cut by electrical discharge machining.

Strain rate jump tests were performed using MTS Landmark with the strain rate ranging from 10⁻⁴/s to 10⁻²/s. The dynamic mechanical behavior of As-SLM and As-Cast was tested using a Split Hopkinson Pressure Bar (SHPB) instrument at strain rates varying from 1000/s to 10,000/s at room temperature. To reduce the friction between bars and the specimen, molybdenum sulfide was used as a lubricant. At each strain rate, three SHPB tests were conducted to ensure data reliability. The dynamic strain rate $\dot{\epsilon }$, true strain ε and true stress σ can be derived from the following equations [12]:

$$\dot{\epsilon }=-\frac{2C_0}{L_0}{\epsilon }_r\left(t\right)$$

(1)

$$\epsilon =-\frac{2C_0}{L_0}\underset{0}{\overset{t}{{\displaystyle \int }}}{\epsilon }_t\left(t\right)dt$$

(2)

$$\sigma =\frac{A_b}{A}E{\epsilon }_t\left(t\right)$$

(3)

where, C₀ is the elastic longitudinal wave propagation velocity in the bars, L₀ is the length of the specimens, E is the elastic modulus of the bars, A and A_b are the cross-section areas of the specimens and bars, ε_r and ε_t are the recorded strains derived from the incident and transmit pulses, respectively.

Strain rate sensitivity (m) and apparent activation volume (V) can be derived from the stress-strain curves using the following equations [13]:

$$m=\frac{\partial ln\sigma }{\partial ln\dot{\epsilon }}$$

(4)

$$V=\sqrt{3}{\mathrm{k}}_{B}T\frac{\partial ln\dot{\epsilon }}{\partial \sigma }$$

(5)

where, k_B is the Boltzmann constant and T is the temperature.

The phase structures were analyzed with an X-ray diffractometer (XRD, Rigaku D/MAX-2550, Rigaku, Tokyo, Japan). Back scattered electron (BSE) images were taken by using Tescan MARA4 (Tescan, Brno, Czech Republic) at 5 KV. Electron backscatter diffraction (EBSD, was performed on a FEI Helios Nanolab G3 UC instrument (FEI, Brno, Czech Republic) at 15 kV with a step size of 0.3 μm. Transmission electron microscopy (TEM) was performed on a FEI Tecnai G2F20 instrument (FEI, Brno, Czech Republic) at 200 kV. Specimens for EBSD and TEM tests were ground by silicon carbide papers, then twinjet polished using the Struers A3 instrument (Struers, Copenhagen, Denmark). The electrolyte was composed of 10% perchloric acid and 90% ethanol, the working temperature was 243 K and the applied voltage was 30 V.

3. Results

3.1. Microstructures

Initial microstructures of As-Cast and As-SLM specimens are presented in Figure 1. XRD results indicated that the As-Cast and As-SLM specimens were both of single FCC phase (Figure 1a). The inverse pole figure (IPF) of the As-Cast and As-SLM specimens are shown in Figure 1b,e, respectively. The As-Cast and As-SLM specimens both displayed equiaxial grain structure and similar average grain size, ~37 μm for As-Cast and ~42 μm for As-SLM (Figure 1d). The enlarged IPF maps showed distinct differences between the two specimens. In the As-Cast specimen, the color was uniform in each grain (Figure 1c). While in the As-SLM specimen, obvious color contrast was presented, which indicated the As-SLM specimens might have a strong lattice distortion (Figure 1d). Further TEM observation revealed that the As-Cast specimen had a very low dislocation density and a few annealing twins, with a thickness of several microns (Figure 1g). In contrast, a cellular dislocation structure was detected in the As-SLM specimens, with an average cell size around 500 nm and a wall thickness of 20–80 nm.

3.2. Mechanical Properties

The representative true stress-strain curves of the As-Cast and As-SLM specimens compressed at different strain rates are summarized in Figure 2. All the specimens had no shear traces on their outer surfaces, from which it could be inferred that both As-Cast and As-SLM specimens had excellent resistance to adiabatic shearing in the strain rate range.

At high strain rate loading, the yield stress for the As-Cast specimen increased rapidly with strain rate, from 370 MPa at 1000/s, 410 MPa at 3000/s, 428 MPa at 5000/s to 525 MPa at 10,000/s (Figure 2a). The As-Cast specimen strain hardened significantly with further deformation at each of the strain rates. This trend was also found in other HEAs with low stacking faults [4,14], which was attributed to deformation twinning [4] or the dynamic Hall-Petch effect [14] (i.e., the formation of twins effectively reduces the dislocation mean free path). By comparison, the As-SLM specimens showed very different behavior compared to the As-Cast ones. From Figure 2b, it is evident that the yield stress remained nearly constant (~650 MPa) when the strain rate increased from 1000/s to 3000/s, but increased rapidly to 960 MPa at 5000/s and 1070 MPa at 10,000/s. Furthermore, the strain hardening rate of the As-SLM specimens hardly increased as the strain rate rose from 1000/s to 5000/s. At the high strain rate of 10,000/s, it appeared that there was a significant increase in the strain hardening rate. However, after the initial strain hardening, strain softening could be observed in the As-SLM specimens, causing a significant drop in stress. After the instantaneous strain softening, the flow stress was witnessed to increase again in the As-SLM specimen till unloading. In addition, strain rate jump tests were conducted at the strain rate of 10⁻⁴/s, 10⁻³/s and 10⁻²/s for both As-Cast and As-SLM specimens for further comparison. As seen in Figure 2c, the stress variation in the As-SLM specimen was more significant than that of the As-Cast one, indicative of pronounced thermal activation in the As-SLM specimen.

The strain rate sensitivity factor m was extrapolated from the dual logarithmic plot of yield stress versus strain rate (Figure 2d). Evidently, the dynamic behavior of the As-Cast specimens was characterized by a constant m, 0.1423, the magnitude of which was nearly one order of magnitude higher than that at the quasi-static strain rate (0.0157), which might be due to dislocation viscous dragging, particularly at the high strain rate (i.e., >1000/s). On the other hand, the dynamic behavior of the As-SLM specimens could be categorized as having two stages over the range of the investigated strain rates, with a very low m (0.007) for the strain rate varying from 1000/s to 3000/s and a high m (0.41) for strain rates larger than 3000/s.

3.3. Deformed Microstructure

After dynamic compression, the As-Cast and As-SLM specimens were subjected to SEM observation (Figure 3). Band-like structures were detected in the 1000/s deformed As-Cast specimens (Figure 3a), which were arrested at grain boundaries and showed different orientations between neighboring grains. With further increase in the strain rate, the density of the band-like structures increased at 3000/s (Figure 3b) and then decreased at 5000/s (Figure 3c). Similar microstructure evolution was seen in the As-SLM specimens (Figure 3d–f).

The IPF maps of the deformed As-Cast and As-SLM specimens are presented in Figure 4a–c, respectively. Other than the initial annealing micro-twins, several narrow deformation twins were observed in the As-Cast specimen, deformed at 1000/s (Figure 4a). In Figure 4b, the As-Cast specimen deformed at 3000/s showed many deformation twins. This showed that twinning was an important deformation mechanism for the As-Cast specimen under dynamic loading. In contrast, only a few twins were observed in the As-SLM specimen until the strain rate reached 5000/s. Therefore, the major deformation mechanism for the As-SLM specimens was still dislocation movements at strain rates below 5000/s. The IPF map of the As-SLM specimen deformed at 10,000/s is not shown because of the low confidence index caused by severe plastic flows.

TEM examination was performed to further study the dynamically deformed microstructures. In the As-Cast specimens deformed at 1000/s, massive planar dislocations were observed (Figure 5a), indicative of easy planar slips. Several deformation twins, with a width around 10 nm. were also observed (Figure 5b). After being deformed at 5000/s, the dislocation density became much higher than that at 1000/s (Figure 5c). Intersection of the deformation twins were also observed (Figure 5d).

By comparison, in the As-SLM specimens deformed at 1000/s, dislocation cells were still preserved, and the cell wall width increased from originally 20–80 nm to 70–250 nm, which implied that the dislocation cells could be effective obstacles to dislocation gliding and serve as “dislocation sinks” at 1000/s (Figure 6a). Microbands were formed with an average spacing of 150 nm (Figure 6b). After being deformed at 3000/s, more dislocations were arrested by the dislocation cells (Figure 6c) with abundant microbands having an average spacing of 40 nm (Figure 6d).

As the strain rate rose to 5000/s, the dislocation cell walls were seemingly partially “dissolved” (Figure 7a) and microbands with a spacing of 100 nm were witnessed within some small grains. However, in the coarse grain regions, no microbands were observed. Interestingly, the overall density of microbands at 5000/s was lower than those at 1000/s and 3000/s (Figure 7b). A few deformation twins were observed with thicknesses ranging from 30 to 150 nm (Figure 7c).

While the dislocation cells were elongated in the specimens deformed at 10,000/s, the cellular structure seemed to be more intact than that at 5000/s. Many nanotwins could be observed. As seen in Figure 8b, the density of microbands increased, and the average spacing was around 70 nm. The high-resolution TEM (HRTEM) image and the corresponding diffraction pattern in Figure 8c show a 9R structure at the intersection of the twin and dislocation cell wall. The 9R structure formed at the dislocation cell interface, leaving a common twin boundary until it was blocked by the dislocation cell at the other end. Interaction of primary twins and secondary twins was observed (Figure 8d).

4. Discussion

4.1. Strain Rate Sensitivity

Strain rate sensitivity is one of the important indicators of dynamic deformation and can be used to characterize different microscopic deformation mechanisms. Dislocation movement always precedes other deformation mechanisms, so strain rate sensitivity is related to microstructure and depends on the type of barriers. In principle, these barriers can be classified as long-range barriers or short-range barriers, depending on the thermal activation energy needed to overcome them. However, with an increasing strain rate, the time left for barrier crossing becomes shorter, compromising the effect of thermal energy, which, in turn, increases the yield stress. It is generally believed that grain boundaries and precipitation are the long-range barriers, while dislocation forests are short-range ones. For example, in the Al_0.3CoCrFeNi high entropy alloys, grain refinement increases the fraction of grain boundaries and, therefore, causes strain rate sensitivity to decrease [15].

The As-Cast specimens showed a more or less constant strain rate sensitivity, which was consistent with the reported dynamic behavior of other HEAs [16]. However, the dislocation cell structures formed due to SLM were a weaker hindrance to dislocation motion than grain boundaries [7,17], the effect of which on strength depends mainly on dislocation densities, not on the cell size [17]. Thus, the dislocation cells behave more like dislocation forests in plasticity strengthening. In FCC-structured metals and alloys, the strain rate sensitivity is mainly affected by the dislocation forest [18]. With the increase of dislocation density, the activation volume of dislocation motion decreases and the strain rate sensitivity increases [18]. The initial dislocation density of the As-SLM specimens was much higher than that of the As-Cast specimens (Figure 1d, h). This high dislocation density reduced the activation volume from 65 b³ of the As-Cast to 25 b³ of the As-SLM specimens. Thus, the As-SLM specimens possessed a higher m value than that of the As-Cast specimens under quasi-static strain rate. Similar results can be found in 316 L stainless steels produced by laser beam powder bed fusion [19].

In contrast, the As-SLM specimens showed a very low m value between 1000/s and 3000/s. At this strain rate range, the V was 100 b³, 4 folds higher than that at the quasi-static strain rate. It means that a larger number of atoms was involved at dislocation nucleation. This high V value could be associated with collective dislocation nucleation [20]. A similar behavior was reported in the high strain rate compression of additively manufactured copper micropillars [13], which was attributed to change of the dislocation nucleation mechanism from Franck-Read sources to collective dislocation nucleation [13], particularly at an ultra-high stress level [21]. In the present As-SLM specimens, although the yield stress was around 650 MPa at 1000/s to 3000/s, lower than the reported 950 MPa for copper, this collective dislocation nucleation mechanism was considered highly plausible, because of the lower energy state of the dislocation cells than that of traditional grain boundaries. As the strain rate increased over 3000/s, the m increased significantly and the V decreased to several b³. The steep change of m and V values indicated that the controlled mechanism changed from thermal activation to dislocation dragging [22].

4.2. Microstructural Evolution

In As-Cast specimens, the deformation mechanism changed from dislocation slip at 1000/s to twinning at a higher strain rate. The increase of strain rate reduced the time of the thermal activation process, leading to higher external stress for dislocation slip. Meanwhile, the critical twin formation stress was not sensitive to strain rate, which made them more prevalent under high strain rate deformation. The microstructural evolution agreed with the reported literature [23]. In contrast to the As-Cast specimens, deformation twins started to occur in As-SLM specimens until 5000/s. This was due to the fact that the critical stress of twinning formation increased with grain refinement [24]. The relation between twinning formation stress σ_T and grain size d is as follows [25]:

$${\sigma }_T=M\frac{\gamma }{b_p}+\frac{k_T}{\sqrt{d}}$$

(6)

where M is the Taylor factor, with a value of 3.06, γ is the stacking fault energy, with a value of 32.5 mJ/m² [26], b_p is the Burgers vector of partial dislocation with a value of 0.146, calculated from the Burgers vector of full dislocation [26], k_T is the Hall-Petch constant for twins, which is 414 MPa (1.5 times the Hall-Petch constant k_S, 276 MPa μm^1/2 [27], for dislocation slip). The As-Cast and As-SLM specimens had the same composition, and, thus, the same stacking fault energy. Although the contribution of dislocation cell size to the strength was lower than that of traditional grain boundaries, the dislocation cells played a similar role as the grain boundaries in affecting the deformation mechanism. For the As-Cast specimens, the grain size was 30 μm, and the critical stress of twin formation was calculated to be 718 MPa. If the dislocation cells were taken as equivalent to “grains”, the critical stress of twin formation could be estimated as around 1214 MPa. An obvious upturn of flow stress was observed around 1200 MPa on the flow stress curve obtained at 5000/s, quite close to the calculated critical twinning stress. So Equation (6) is applicable in the present situation.

With respect to dislocation cell structure evolution, we proposed that it was mainly due to a dislocation annihilation process. Under dynamic loading, dislocations are generated to accommodate the applied strain if other plasticity mechanisms (i.e., twinning) are not operative. Therefore, if the generated dislocations had opposite signs to those dislcations constituting the cell structures, and were absorbed in the cells, plastic deformation would decrease the dislocations. This might rationalize the dissolution of dislocation cells at the strain rate of 5000/s. However, we noticed that, at 10,000/s, a large amount of deformation twins was generated. This transition of dynamic deformation mechanisms from dislocation movements to twinning helped preserve the dislocation cells, as seen in TEM results.

Microbands have been reported as an additional strengthening mechanism in HEAs at both quasi-static and dynamic loadings [28]. Those HEAs usually possess a low stacking fault energy, having a strong tendency to form planar dislocation arrays. With further deformation, these planar dislocation arrays would produce the localized microband structure. Precipitates also seem to remarkably increase the occurrence of microbands. The L12 phase strengthened FeCoNi-based HEA deformed at quasi-static strain rate exhibited 1.5 GPa strengths and 50% ductility, and microbands were the main deformation mechanism [29]. Li et al. investigated the dynamic behavior of Mo-doped CoCrNi HEAs, and found that the precipitates created high local stress concentration, and facilitated the formation of microbands [30]. In an early work, Huang et al. [31] observed the deformation behavior of various metals and alloys, and concluded that the microband formation was controlled by local stress concentration and independent of strain rate or strain. In the present work, the dislocation cell structures contributed to both higher stress level and a larger percentage of interfaces, leading to easier microband formation in the As-SLM specimens than in the As-Cast specimens. In the As-Cast specimens, the early formation of deformation twins relieved the stress concentration, leading to a lower driving force for microband formation. Since the formation of microbands needs high local stress concentration, the evolution of microband density is directly related to the local stress concentration which is affected by the stability of cell walls. The microband density increased from 1000/s to 3000/s due to the competition between small annihilation rate and relatively large dislocation pile-up. Then it decreased rapidly to 5000/s with the partially missing cell walls. At 10,000/s, the microband density increased again with the high local stress concentration contributed by the deformation twins and stable cell walls.

4.3. Dynamic Behavior

For the As-Cast specimens, with strain rate increases, higher density of twins formed, and the mean-free-path of dislocations reduced. Thus, the stronger dynamic Hall-Petch effect led to a higher work hardening rate when the strain rate increased.

In the As-SLM specimens, the initial work hardening rate decreased with strain rate from 1000/s to 5000/s, and increased again at 10,000/s. The stability of the cell walls became weaker with strain rate increasing from 1000/s to 5000/s, as discussed above. The less intact cell walls lost the ability to impede dislocation movement, thus leading to the reduction in the work hardening rate. At the strain rate of 10,000/s, the stress soon increased to the critical twinning stress, abundant deformation twins were formed and contributed to the ultra-high work hardening rate, both through dynamic Hall-Petch effect and stabilizing of the cell walls.

The following stress drop and upturning would be caused by the evolution of microbands. The formation of a microband is believed to be a softening process [32], since it creates an easy slip channel for dislocation movement. In comparation with other microband forming materials, the stress drop was much more obvious in the As-SLM specimens. This might be due to the high strain rate and the small-sized dislocation cell structure where other deformation mechanisms, like dislocation movement and twinning, were impeded. The amount of stress drop was consistent with the microband density, which increased slightly from 1000/s to 3000/s, then decreased to 5000/s and increased again at 10,000/s. Since the microbands were local microstructures and were stopped by the grain boundaries, they did not lead to the catastrophic failure of the specimens. After the microbands were fully formed, the microbands continued to strengthen the As-SLM specimens in a Hall-Petch relation [33], then their contribution to work hardening was positive related to their density and spacing.

The dynamic compressive properties of the As-SLM HEA, other HEAs, Al alloys, Ti alloys and stainless steels are compared in Figure 9. The product of maximum stress and strain was used to represent the dynamic energy absorption. The As-SLM HEA in this study had excellent comprehensive dynamic mechanical properties.

5. Conclusions

The dynamic behavior of As-SLM and As-Cast FeCoCrNi HEA, over a wide range of strain rates. was studied, and the conclusions are as follows:

(1)

The As-SLM FeCoCrNi HEA showed a more excellent combination of toughness and yield stress than previously reported alloys due to the high density of deformation microbands and twins.

(2)

The dislocation cell structures contributed to high local stress concentration and facilitated the formation of microbands at strain rates of 1000/s and 3000/s. With further increase of strain rate, the deformation mechanism changed to twins.

(3)

The AS-SLM HEA showed a low strain rate sensitivity between strain rates of 1000/s and 3000/s due to the collective dislocation nucleation mechanism. At higher strain rates, the strain rate sensitivity increased rapidly.