ARTICLE

https://doi.org/10.1038/s41467-021-24523-9 OPEN

High-throughput design of high-performance

lightweight high-entropy alloys
Rui Feng1,2, Chuan Zhang 3 ✉, Michael C. Gao 4,5 ✉, Zongrui Pei 4,6, Fan Zhang3, Yan Chen2, Dong Ma7,
Ke An 2, Jonathan D. Poplawsky 8, Lizhi Ouyang9, Yang Ren 10, Jeffrey A. Hawk4, Michael Widom 11 &
Peter K. Liaw1 ✉
1234567890():,;

Developing affordable and light high-temperature materials alternative to Ni-base superalloys

has significantly increased the efforts in designing advanced ferritic superalloys. However,
currently developed ferritic superalloys still exhibit low high-temperature strengths, which
limits their usage. Here we use a CALPHAD-based high-throughput computational method to
design light, strong, and low-cost high-entropy alloys for elevated-temperature applications.
Through the high-throughput screening, precipitation-strengthened lightweight high-entropy
alloys are discovered from thousands of initial compositions, which exhibit enhanced
strengths compared to other counterparts at room and elevated temperatures. The experi-
mental and theoretical understanding of both successful and failed cases in their strength-
ening mechanisms and order-disorder transitions further improves the accuracy of the
thermodynamic database of the discovered alloy system. This study shows that integrating
high-throughput screening, multiscale modeling, and experimental validation proves to be
efficient and useful in accelerating the discovery of advanced precipitation-strengthened
structural materials tuned by the high-entropy alloy concept.

1 Department of Materials Science and Engineering, The University of Tennessee, Knoxville, TN, USA. 2 Neutron Scattering Division, Oak Ridge National

Laboratory, Oak Ridge, TN, USA. 3 Computherm, LLC, Middleton, WI, USA. 4 National Energy Technology Laboratory, Albany, OR, USA. 5 Leidos Research
Support Team, Albany, OR, USA. 6 ORISE, 100 ORAU Way, Oak Ridge, TN, USA. 7 Neutron Science Platform, Songshan Lake Materials Laboratory, Dongguan,
Guangdong, China. 8 Center for Nanophases Materials Sciences, Oak Ridge National Laboratory, Oak Ridge, TN, USA. 9 Department of Physics and
Mathematics, Tennessee State University, Nashville, TN, USA. 10 Advanced Photon Source, Argonne National Laboratory, Argonne, IL, USA. 11 Department of
Physics, Carnegie Mellon University, Pittsburgh, PA, USA. ✉email: [email protected]; [email protected]; [email protected]

NATURE COMMUNICATIONS | (2021)12:4329 | https://doi.org/10.1038/s41467-021-24523-9 | www.nature.com/naturecommunications 1

ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-021-24523-9

T
he fields of aerospace and fossil energy have benefited from performance LWHEAs for elevated-temperature applications.
decades of the development of Ni-base superalloys. How- Since materials in use at high temperatures approach equilibrium
ever, to obtain further efficiency gains and environmental states, knowledge of stable phase equilibrium at fabrication and
friendliness, developing new materials with inexpensive, light, working temperatures are very important. Miracle and Senkov
and strong characteristics is required. Numerous efforts can be et al.21,22 have explored the development of multi-principal ele-
found in the development of novel ferritic superalloys containing a ment alloys for structural applications and have summarized a
disordered body-centered-cubic (BCC) matrix with ordered B2 few criteria as guidelines for the design of high-temperature
and/or L21 precipitates due to their good creep resistance at ele- structural materials:
vated temperatures and relatively lower densities1,2. However,
these emerging precipitation-strengthened ferritic alloys still face 1. No first-order phase transformation in the temperature
some drawbacks, such as the low high-temperature strength and range of the application is allowed to ensure structural
not lightweight, which limit their applications. To overcome these stability. Therefore, all phase transformations, if present,
issues, new alloy-design strategies are required. Recently, the must be above the operating temperature, Tuse (about
concept of high-entropy alloys (HEAs) or multi-principal-element 0.8Tm, where the temperatures are in Celsius degree).
alloys (MPEAs) has revolutionized the traditional alloy-design 2. Good ductility and fracture toughness are required. There-
strategy, using multi-principal components (≥5) instead of one or fore, the microstructure must include a ductile solid-
two key components 3–10. Due to the existence of many kinds of solution primary phase as the matrix and a precipitation-
elements with different atomic sizes, atoms in HEAs tend to strengthening phase that can dissolve at high temperatures
deviate from their ideal lattice sites and give rise to severe local and then re-precipitate above Tuse but below Tm.
5,6
lattice distortion , which could impede dislocation motion, Our previous research on the Al–Cr–Fe–Mn–Ti HEAs18
leading to the pronounced strengthening effect11,12. To pursue reported the formation of high-density coherent L21 precipitates
higher strengths, the formation of coherent intermetallic pre- in a disordered BCC solid-solution matrix. Therefore, the L21
cipitates while maintaining a medium- to high-entropy matrix has phase is chosen as the target strengthening-precipitate phase. The
been attempted during the design of HEAs10,13–15. Moreover, the optimal molar fraction of the L21 phase is not known. Hence, we
low lattice misfit that can decrease the nucleation barrier for set it to be 0.05–0.5 at the application temperatures as a
precipitation, and thus stabilize precipitates with a high number reasonable approximation. The specific criteria used for the
density is likely found in HEAs10,16. These effects have demon- present HTCs are
strated the concept of HEAs being a new avenue for developing
new-type precipitation-strengthened lightweight and low-cost (I) Tm > 1250 °C and f(BCC) = 1 at Tm;
materials for high-temperature applications 17,18. In the develop- (II) f(BCC) + f(L21) = 1 and 0.05 < f(L21) < 0.5 at 0.8Tm;
ment of such materials, critical goal-oriented design strategies are (III) f(BCC) + f(L21) = 1 and 0.05 < f(L21) < 0.5 at 0.5Tm.
the prerequisite, including the inexpensive raw materials, low Here f(BCC) and f(L21) refer to the mole fractions of the BCC
density, high melting temperature (Tm), good oxidation resistance, and L21 phases, respectively. The first criterion ensures that the
great creep resistance, and high strength with acceptable ductility. primary phase is the disordered BCC solid solution. The second
Thus, a rational selection of chemical compositions is of very and third criteria ensure that the L21 phase is the precipitation-
significance to take into account these essential factors. Our pre- strengthening phase and no phase transformation in the
vious work discovered potential L21 precipitation-strengthened temperature range for target applications (0.5Tm < T < 0.8Tm).
lightweight HEAs (LWHEAs) in an Al–Cr–Fe–Mn–Ti system, but The HTC flowchart and results are summarized in Fig. 1. In
the formation of the brittle C14 Laves phase deteriorates Fig. 1b–e, Area 1 only meets criterion I. Area 2 meets criteria I
their properties because the dissimilarity of crystal structures and and II, and Area 3 meets all three criteria. Using a step size of 5
different thermal expansion coefficients between the C14 Laves atomic percent (at.%) for each component in the composition
and BCC-base phases increase the tendency to crack18. Hence, range from 0 to 50 at.%, the solidification HTCs for a total of
exploring suitable chemical compositions without the formation of 3246 alloys are calculated, using the lever-rule model. From this
detrimental intermetallic phases is crucial for developing the first-round HTC, the melting temperature (Tm) of each alloy as
desired materials. However, the vastness of the compositional well as its phase stability at Tm can be obtained. Two criteria are
space poses a huge challenge for efficiently screening out suitable used to screen our target alloys: Tm > 1250 °C and a single BCC
compositions. The trial-and-error experimental method is phase at Tm. One third of these alloys (1168) are found to meet
obviously not suitable, since it is extremely costly and time- the criterion (Area 1). In the second round, the 0D point HTCs
consuming to experimentally screen the proper alloys in the vast are carried out for the 1168 alloys at 0.8Tm, and the criterion of f
composition space. This trend is even true for a quinary system, (BCC) + f(L21) = 1 and 0.05 < f(L21) < 0.5 is as the filter. A total
not to mention a higher-order (n ≥ 5) multicomponent system, of 44 alloys are identified and shown as Area 2 in Fig. 1b–e. In the
like the Al–Cr–Fe–Mn–Ti system. Fortunately, the pace of dis- third round, the 0D point HTCs are performed for these 44 alloys
covering promising HEAs can, in principle, be accelerated by the at 0.5Tm and the same criterion of (BCC) + f(L21) = 1 and 0.05 <
development of efficient computational screening methods and f(L21) < 0.5 is used as the filter. Only eight alloys (density, ρ ≤
tools19,20. In the present work, the CALculation of PHAse Dia- 6.5 g/cm3) survive out of the over 3000 compositions eventually.
grams (CALPHAD)-based high-throughput computational tool is The compositions that meet all three criteria (I, II, and III) are
employed to efficiently explore the Al–Cr–Fe–Mn–Ti system for shown as Area 3 in Fig. 1b–e. Detailed compositions and the
discovering new-type precipitation-strengthened HEAs. The sub- calculated equilibrium phase of all identified alloys are listed in
sequent experimental and theoretical studies on the discovered Supplementary Table 1 and Supplementary Fig. 1, respectively.
lightweight HEAs provide insights into developing high-
performance HEAs via the high-throughput alloy-design method.
Microstructural and mechanical validations on the discovered
LWHEAs. Systematic experimental investigations on the identi-
Results fied eight LWHEAs were then carried out not only to validate the
CALPHAD-based high-throughput calculation (HTC) screen- CALPHAD-based HTC predictions, but also to obtain essential
ing results. The goal of the current work is to design high- data (microstructures and mechanical properties) and valuable

2 NATURE COMMUNICATIONS | (2021)12:4329 | https://doi.org/10.1038/s41467-021-24523-9 | www.nature.com/naturecommunications

NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-021-24523-9 ARTICLE

a b c
Flowchart of HTC

d e

Fig. 1 High-throughput screening optimal alloy compositions in the Al–Cr–Fe–Mn–Ti system. a Flowchart of the current HTC. b Al–Cr projection. c Al–Fe
projection. d Al–Mn projection. e Al–Ti projection.

insights for improving the accuracy of high-throughput predic- and volume fraction (43.6 volume percent, vol.%) (~44 vol.% by
tions further. the Rietveld refinement, Supplementary Fig. 5) in Alloy 1 were
The synchrotron x-ray diffraction (XRD) patterns (Fig. 2a) also determined by the APT analysis. Note that the number
demonstrate that all the eight alloys in the as-cast states are density is highly underestimated because precipitates are touching
composed of L21 and possible BCC phases without other or overlapping so that they cannot be delineated from each
intermetallic phases. All the eight alloys have equiaxed grains other, using isosurfaces. In contrast to Alloy 1, the chemical
(Fig. 2b, c and Supplementary Fig. 2), indicative of a single-phase distribution of constituent elements in Alloy 7 is uniform (Fig. 2i).
structure. The morphologies of the L21 phase in these LWHEAs Together with the TEM observations (Fig. 2e), it suggests
were characterized by the transmission-electron microscopy that Alloy 7 consists of ordered L21 APDs separated by APBs,
(TEM) dark-field (DF) images (Fig. 2d, e and Supplementary and no chemical segregation is found between the APDs and the
Fig. 3). Interestingly, two different L21 morphologies are present APBs.
among the eight alloys. The high-density nanoscaled L21 To evaluate the mechanical properties of the identified eight
precipitates (10–30 nm) that are uniformly distributed within alloys in a high-throughput manner, we performed compression
the BCC matrix are found in Alloy 1 (Al20Cr5Fe50Mn20Ti5) and tests on the eight alloys at room temperature (RT) (Fig. 3a). One
Alloy 8 (Al15Cr5Fe50Mn25Ti5) (Fig. 2d and Supplementary Fig. 3), can see that the yield strengths of these alloys vary in a wide range
while the typical ordered L21 antiphase domains (APDs) with a from 500 to 1642 MPa (Supplementary Table 2). The high-
thin film of a disordered BCC phase on the antiphase domain temperature performances of Alloys 1 and 8 were also evaluated
boundaries (APBs) are observed in Alloys 2–7 (Fig. 2e and and compared to other representative alloys (Fig. 3b, Supple-
Supplementary Fig. 3). In addition to the as-cast states, we also mentary Fig. 6, and Supplementary Table 3). Alloys 1 and 8
performed the TEM characterizations on the representative exhibit high yield strengths below 700 °C (1072 ± 25 MPa at 600 °
Alloys 1, 2, and 7 after homogenization at 1200 °C (Supplemen- C for Alloy 1; 1032 ± 10 MPa at 600 °C for Alloy 8). As the
tary Fig. 4). For Alloy 1, its annealed microstructure is very temperature increases to 700 °C, the yield strengths of Alloys 1
similar to that in the as-cast state (Fig. 3d), but with a growth of and 8 are still kept at 607 ± 22 MPa and 643 ± 16 MPa,
the L21 precipitates to ~39 nm (Supplementary Fig. 4a). For respectively, though an obvious softening occurs like many other
Alloys 2 and 7, the sizes of APD are greater than those in the as- precipitation-strengthened ferritic BCC alloys23–28. However, the
cast condition (Supplementary Fig. 4b, c), indicating that the L21 newly designed L21 precipitation-strengthened LWHEAs outper-
phase is stable at 1200 °C. To further demonstrate the L21 phase’s form traditional precipitation-strengthened BCC alloys, such as
stability at intermediate temperatures, we also annealed the as- BCC + L21, BCC + D03, and BCC + B2 alloys, in terms of yield
cast Alloy 7 at 700 °C for 2 hours. As shown in Fig. 2g, the sizes of strengths and specific yield strengths (Fig. 3b and Supplementary
L21 APD grew greatly, suggesting that the alloys with an L21 APD Fig. 7). Moreover, the softening resistance in these alloys above
morphology are more like a single L21 phase below its melting 700 °C could be further improved by the alloy-design wisdom in
point. Ni-base superalloys, such as the addition of minor refractory
The chemical compositions and morphologies of the consti- elements, e.g., W, Ta, and Mo29.
tuent phases in the representative Alloys 1 and 7 were determined Based on the above experimental analyses, we can categorize
by atom-probe tomography (APT) (Fig. 2h, i). In Alloy 1, the L21 the eight alloys into two groups. G1: Alloys 1 and 8 with the BCC
phase is rich in Al and Ti, while the BCC phase contains mostly matrix + nanoscale L21 precipitates (Fig. 2 and Supplementary
Cr, Mn, and Fe (Fig. 2h and Table 1). A similar compositional Fig. 3), in line with our thermodynamic prediction. G2: Alloys
distribution in the L21 and BCC phases is also observed in Alloy 8 2–7 with the L21 as the major phase (Fig. 2 and Supplementary
(Table 1). Moreover, the L21 number density (2.76 × 1022/m3) Fig. 3), in which the L21 phase fractions are too high and deviate

NATURE COMMUNICATIONS | (2021)12:4329 | https://doi.org/10.1038/s41467-021-24523-9 | www.nature.com/naturecommunications 3

ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-021-24523-9

a d f

Alloy 1

e Alloy 7
Annealed at 700 C, 2 h
g 110

APB

b Alloy 1 c Alloy 7
h i

Alloy 7
Alloy 1
Fig. 2 Microstructural information of the discovered LWHEAs. a Synchrotron XRD patterns of Alloys 1–8 in their as-cast states. b, c Backscattered
electron (BSE) images of the representative Alloys 1 (Al20Cr5Fe50Mn20Ti5) and 7 (Al30Cr5Fe50Mn10Ti5), representatively. d, e The TEM DF images taken
by the unique (111) reflection of the L21 phase and the corresponding selected-area diffraction (SAED) patterns along the [100] and [110] zone axes of
Alloys 1 and 7 in their as-cast states, respectively. The arrows indicate the features of precipitates, antiphase domain (APD), and antiphase boundary
(APB). f High-resolution TEM (HRTEM) image of Alloy 1 taken along the [110] zone axis, showing the coherent interfaces between the BCC matrix and the
L21 nanoscale precipitates. The next two images are the fast Fourier Transform (FFT) patterns of the BCC matrix and L21 precipitates of Alloy 1. g The TEM
DF images taken by the unique (111) reflection of the L21 phase of the Alloy 7 after annealing at 700 °C for 2 h, showing the greatly grown APDs. h, i APT
atomic maps of Alloys 1 and 7, respectively.

from our thermodynamic prediction. For easy understanding, our attributed to the precipitation strengthening, besides the effect of
discussion in the following part will base on these two groups. atomic-level complexity in HEAs.

Strengthening behavior. The different mechanical responses Order-disorder transition behavior. The morphology of the L21
between G1 (Alloys 1 and 8) and G2 (Alloys 2–7) are attributed to phase strongly affects the mechanical performance of the high-
the precipitation strengthening and the intrinsic-deformation throughput-designed LWHEAs, which is related to the order-
behavior of L21, as dispersed particles or a single phase, respec- disorder transition behavior. Thus, we performed the in situ
tively, considering L21’s different morphologies and volume neutron scattering on the representative Alloy 1 (Al20Cr5-
fractions (e.g., Alloy 1: precipitates, 44 vol.%; Alloy 7: APDs, ~85 Fe50Mn20Ti5) in G1 and Alloy 7 (Al30Cr5Fe50Mn10Ti5) in G2 to
investigate this behavior. As shown in Fig. 4a for Alloy 1, the
vol.%). The precipitation-strengthening contribution in Alloy 1 is
unique L21 peaks (highlighted by the green color) are observed
quantitatively calculated (Supplementary Note 1), which is
only at 900 °C and below, indicating that L21 is stable up to
mainly attributed to the ordering strengthening (~904 MPa),
900 °C, below which both the BCC and L21 phases coexist. The
close to the yield strength differences between Alloys 1 and 2 relative intensity of superstructure peaks for the L21 phase
(~874 MPa), and between Alloys 1 and 7 (~860 MPa). The decreases upon heating, which is due to the reduced fraction of
precipitation-strengthening effect in Alloy 1, but not in Alloy 7, is the L21 phase. For Alloy 7, the order peaks of the L21 phase
also demonstrated by the in situ neutron-diffraction results. persist to 1050 °C, suggesting that L21 exists even up to 1050 °C
Figure 3c, d exhibit the lattice strain versus applied compressive (Fig. 4b and Supplementary Fig. 8).
stress curves corresponding to the loading direction in both We further conducted the metropolis Monte Carlo (MC)
Alloys 1 and 7. One can notice that Alloy 1 displays a clear load- calculations to understand the order-disorder transitions in
transfer behavior upon the BCC phase yielding (around 1000 Alloys 1 and 7 during solidification, using much larger supercells
MPa), i.e., the hard L21 phase bears a larger load with greater of 2000 atoms at 107 MC steps within the nearest-neighbor
lattice strains (L21 {222} and {420}), while the soft BCC phase interaction approximation. With the concentrations as the only
undertakes a lower load with lower lattice strains (BCC {110} and variables for a given system, this method gives a more clear and
{211}), which is a reflection of the precipitation strengthening. In definitive picture of the influence of constitutions on the order-
contrast, no load transfer was observed in Alloy 7 (Fig. 3d). The disorder transition. The output of the MC calculations allows us
synchronized response of various hkl-specific lattice strains versus to calculate the short-range ordering (SRO) and long-range
applied stress indicates that Alloy 7 behaves like a single L21 ordering (LRO) parameters for both Alloys 1 and 7 between 0 and
phase. Therefore, the high yield strengths of Alloys 1 and 8 are 2500 K. Define ξ iμ ¼ 1when a site, i, is occupied by an atom of a

4 NATURE COMMUNICATIONS | (2021)12:4329 | https://doi.org/10.1038/s41467-021-24523-9 | www.nature.com/naturecommunications

NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-021-24523-9 ARTICLE

Fig. 3 Mechanical responses of the discovered LWHEAs. a Compressive stress-strain plots of Alloys 1–8 at RT. b Comparison of yield strengths as a
function of temperature between Alloys 1, 8, and other counterpart materials (all these BCC-base alloys are in their as-cast states)23–27. c, d Lattice strain
as a function of applied stress along the loading direction in Alloys 1 and 7 at RT, respectively, measured by in situ neutron diffraction. The error bars are
obtained from the uncertainties of the single-peak fitting on hkl diffraction peaks.

Table 1 Chemical compositions of the BCC and L21 phases in Alloys 1 and 8, measured by APT.

Alloy 1 (Al20Cr5Fe50Mn20Ti5)
Phase Al (at.%) Cr (at.%) Fe (at.%) Mn (at.%) Ti (at.%)
BCC 16.63 ± 0.09 6.70 ± 0.05 53.10 ± 0.10 22.79 ± 0.09 0.79 ± 0.02
L21 23.67 ± 0.14 3.21 ± 0.06 49.82 ± 0.14 11.84 ± 0.11 11.46 ± 0.10
Alloy 8 (Al15Cr5Fe50Mn25Ti5)
Phase Al (at.%) Cr (at.%) Fe (at.%) Mn (at.%) Ti (at.%)
BCC 13.31 ± 0.04 6.04 ± 0.01 53.09 ± 0.05 26.35 ± 0.04 1.21 ± 0.01
L21 22.14 ± 0.09 2.62 ± 0.02 49.49 ± 0.10 10.40 ± 0.06 15.32 ± 0.07

type, μ, and ξ iμ ¼ 0 otherwise. A configuration is uniquely The L21 structure has three different sublattice sites (Wyckoff
sites): 8c, 4a, and 4b, in which 4a and 4b can be considered as one
specified by the set, fξ iμ g. Note the identities, μ ξ iμ ¼ 1 and
site and 8c as the other site. Therefore, for simplification, we can
<ξ iμ > ¼ cμ . The frequency of the nearest-neighbor pairs of treat an L21 structure as a B2 structure to calculate the LRO
species, μ and ν, isyijμν ¼ <ξ iμ ξ jν >. We define parameter. The LRO can be expressed as32
β
yαi  yi
wijμν  yijμν =cμ ¼ <ξ ijμν ξ ijνν >=cμ ð1Þ ηi ¼ β
ð2Þ
yαi þ yi
β
where yαi and yi are the site occupancy of the element, i, on
As the conditional probability that a site, j, has species, ν, given sublattices, α and β, respectively. The value of ηi falls within the
ij
that i has species, μ. Our wμν is related to the usual Warren- range of −1 to 1, as shown in Fig. 4c, d. When ηi ¼ 0, the
ij ij
Cowley order parameter30,31 as αuv ¼ 1  wμν =cμ . element, i, is randomly distributed on both α and β sublattices.

NATURE COMMUNICATIONS | (2021)12:4329 | https://doi.org/10.1038/s41467-021-24523-9 | www.nature.com/naturecommunications 5

ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-021-24523-9

Fig. 4 Order-disorder transitions of Alloys 1 and 7. a, b In situ neutron-scattering patterns of Alloys 1 (Al20Cr5Fe50Mn20Ti5) and 7 (Al30Cr5Fe50Mn10Ti5)
at elevated temperatures, respectively. The label of order means that the reflections are exclusive to the L21 structure, while the label of fundamental
means that the reflections are common to both the ordered L21 and disordered BCC structures. c, d The MC-calculated LRO parameters as a function of
temperature in Alloys 1 and 7, respectively. e, f The MC-calculated SRO parameters as a function of temperature in Alloys 1 and 7, respectively. The black
arrows indicate the change of the most abundant, Fe, as the neighbor of all elements.

When ηi equals 1 or −1, the element, i, can be exclusively found simulation with 250 atoms at 106 MC steps especially for this
on the α or β sublattice. system between −273 and 2727 °C (0–3000 K) (Supplementary
The order-disorder transition can be directly reflected by the Figs. 10, 11). The ordered L21 structure starts to become less
LRO parameters (Fig. 4c, d). In both Alloys 1 and 7, as the ordered at ~ 1227 °C (1500 K). Both the LRO and SRO disappear
temperature decreases, Al, Cr, and Ti tend to segregate to one at about 2327 °C (2600 K). Through comparing the LRO and SRO
sublattice, while Fe and Mn primarily occupy the other. As between the Fe2AlTi-type L21 and Alloys 1 and 7, we can find that
indicated by the red arrows, the ordering transition temperature, the ordering temperature in Alloy 7 with a higher Al content is
Tc, of Alloy 1 [1607 °C (1880 K)] is much lower than that of Alloy closer to that of the Fe2AlTi-type L21 structure, implying that the
7 [1997 °C (2270 K), far beyond its melting temperature], Fe2AlTi-type L21 can survive at higher temperatures in Alloy 7
suggesting that the ordered L21 phase is less easily retained in than Alloy 1, in line with the in situ neutron-scattering results
the solid state in Alloy 1 than in Alloy 733. This prediction is (Fig. 4a, b).
consistent with the experimental observations (Fig. 4a, b), i.e., Ab initio molecular dynamics (AIMD) simulations were also
Alloy 1 is composed of L21 and BCC phases from RT to 900 °C, performed on Alloy 1 in its molten state to support the MC-
while only the ordered L21 phase still exists in Alloy 7 from RT to calculation results. Figure 5a–e shows the selected partial PDF of
1050 °C. The inset subfigures give the snapshots at −173 °C Alloy 1 at T = 1600 °C to the first nearest neighbor. The
(100 K), which directly show the L21-phase formation. Besides the intensities of pair correlations of the Al–Al and Ti–Ti bonds
LRO, the SRO parameters reveal more details of the order- are significantly lower than the other pair correlations, indicating
disorder transition (Fig. 4e, f). Since Fe is the most abundant the strong tendency for Al and Ti atoms to bond with other
elements. On the other hand, the Al–Fe pair correlation has the
element in this system, we consider its changes as the fingerprint
highest intensity, and the Cr–Fe pair correlation has the second-
of the ordering transition. In both alloys, the temperature-
highest intensity, compared to other pair correlations. The
dependent SRO parameters of Fe atoms as the nearest neighbors
persistent presence of stronger Al–Fe and Cr–Fe pair correlations
of Al and Ti increase upon cooling. In contrast, the Cr and Mn
in the Al–Cr–Fe–Mn–Ti alloys suggests that Al–Fe and Cr–Fe
have the opposite trend (Fig. 4e, f). This trend is particularly pair interactions may promote the formation of L21 and BCC
obvious for Ti that has only Fe as its neighbors around and below phases, respectively, as evidenced by the chemical compositions
727 °C (1000 K). Moreover, the SRO parameters also demonstrate of both phases (Table 1).
that the formation of the L21 phase is favored by Fe–Al and Fe–Ti
pairs and that of the BCC phase by Cr–Fe pairs, in agreement
with the chemical compositions of L21 and BCC phases (Fig. 2h Reason for the formation of different L21 morphologies. The
and Table 1). The reliability of the MC 2000-atom superstructures different L21 morphologies exhibited between G1 (nanoscale
for both Alloys 1 and 7 are verified by the fitted neutron- precipitates) and G2 (APDs) are strongly correlated to their dif-
scattering pair-distribution function (PDF) results (Supplemen- ferent chemical compositions (Supplementary Table 1), which
tary Fig. 9) and by comparison with the result of first-principles directly affect their order-disorder transitions, namely, the
simulations (Supplementary Table 4). atomic-site occupancy of the constituent elements. The atomic-
To better show the ordering transition due to the formation site occupancy of the formed multicomponent L21 structure is
of the Fe2AlTi-type L21 phase, we performed another MC revealed as follows. The L21 (X2YZ) structure (Prototype,

6 NATURE COMMUNICATIONS | (2021)12:4329 | https://doi.org/10.1038/s41467-021-24523-9 | www.nature.com/naturecommunications

NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-021-24523-9 ARTICLE

a b c
Al-Fe Fe-Al
Cr-Fe

d e f
Neutron

Fig. 5 First-principles calculations on Alloy 1. a–e AIMD-simulated partial PDFs of Alloy 1 (Al20Cr5Fe50Mn20Ti5) at T = 1600 °C, and f Neutron-
weighted PDF.

Table 2 The nearest-neighbor interaction parameters, Atomic site occupancy in the multicomponent L21 structure
Jij , in eV.

Jij Al Cr Fe Mn Ti
Al 0 −0.0224 −0.0942 −0.0970 −0.1280
Cr −0.0224 0 0.0210 −0.0163 −0.0043
Fe −0.0942 0.0210 0 0.0094 −0.1181
Mn −0.0970 −0.0163 0.0094 0 −0.0930
Ti −0.1280 −0.0043 −0.1181 −0.0930 0

Fe2AlTi; Pearson type, cF16) has three different sublattice sites:

X @ 8c(1/4, 1/4, 1/4), Y @ 4a(0, 0, 0), and Z @ 4b(1/2, 1/2, 1/2).
According to the calculated LRO parameters (Fig. 4c, d), Fe and
Mn prefer to occupy one sublattice, while Al, Ti, and Cr tend to
segregate into another sublattice. In the L21 structure, the sub-
lattices of Y and Z can be considered equally in terms of site, but APD Precipitates
L21
differ in terms of composition. Taking account of the strong Meet Deviate
atomic pairs of Al–Fe, Fe–Ti, Fe–Cr, and Ti–Al (Fig. 5), we can L21 stoichiometry L21 stoichiometry
assert that Y(4a) and Z(4b) are filled with Al and (Ti, Cr),
respectively. However, if the content of Al is more than 25 at.%, Fig. 6 Schematic of the atomic-site occupancy and morphology evolution
the extra Al atoms can also occupy the site of Z(4b). In the L21 of the L21 structure. A schematic showing the atomic-site occupancy in the
structure of the present Al–Cr–Fe–Mn–Ti system, Fe atoms are multicomponent L21 structure and the morphology evolution of the L21
always in the site of X(8c) because the content of Fe is not beyond phase that depends on the chemical compositions of discovered LWHEAs.
50 at.%. As for Mn, its site occupancy depends on Fe contents,
that is, if the Fe content is less than 50 at.%, Mn can occupy the
unfilled sites of X(8c) by Fe because Mn–Al and Mn–Ti have BCC matrix of Alloys 1 and 8 (Table 1). From the known atomic-
comparable interaction energies with Fe–Al and Fe–Ti, respec- site occupancy, the different L21 morphologies between G1 and
tively (Table 2). Once the Fe content equals 50 at.%, the sites of X G2 can be understood from the L21 stoichiometric composition,
(8c) are completely by Fe atoms, and thus, Mn can occupy the i.e., ð50ÞXFeðMnÞ ð25ÞYAl ð25ÞZTi;Cr ðAl;MnÞ . For G2 (Alloys 2–7), their
unfilled sites at Z(4b) by Ti and Cr. Similar atomic occupancy of chemical compositions can exactly match the perfect L21 stoi-
Mn at Z(4b) can be found in a Fe2AlMn L21 structure34. Note chiometry, resulting in the formation of the single-phase L21
that the solubility of Mn at sites of Z(4b) should not exceed 15 at. phase in the form of APD. In contrast, for G1 (Alloys 1 and 8),
% because the Mn content in L21 is always less than 15 at.% their chemical compositions deviate far away from the perfect L21
(Table 1 and Supplementary Table 1). When the content of Mn is stoichiometry, leading to the formation of nanoscaled L21 pre-
more than 15 at.%, the extra Mn goes to the BCC matrix. Cr tends cipitates within the BCC matrix. The atomic-site occupancy in
to segregate with Fe and Mn together in the disordered BCC the multicomponent L21 structure and the composition-
matrix, as evidenced by the observed chemical composition in the dependent phase-evolution behavior are schematically presented

NATURE COMMUNICATIONS | (2021)12:4329 | https://doi.org/10.1038/s41467-021-24523-9 | www.nature.com/naturecommunications 7

ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-021-24523-9

in Fig. 6, which can be further integrated with our HTC frame to precipitates is required during alloy design. However, the current
accelerate the discovery of advanced precipitation-strengthened high-throughput predictions are valid in the composition range of
LWHEAs. Al < 25 at.% and Mn > 15 at.% with the unchanged contents of
other elements. Especially, the HTC can become more efficient
when it is coupled with the deciphered atomic occupancy of the
Discussion multicomponent L21 structure and its related phase-evolution
The concept of HEAs provides us an unprecedented degree of rule. Therefore, the high-throughput design and the fundamental
freedom in the design of advanced alloys with promising prop- understanding of the discovered LWHEAs will accelerate the pace
erties. For the rapid exploration of the vast compositional space of discovering promising HEAs, especially for multicomponent
and investigation of the composition and temperature effects on systems.
the microstructure, advanced HEA-design strategies and efficient
tools are necessary. At present, a few advanced HEA-design
strategies were reported by combining the physicochemical cri- Methods
The CALPHAD-based HTCs. The CALPHAD approach38–40 is currently the only
teria, CALPHAD, data mining, and multi-objective optimization method that can be used to obtain multicomponent phase diagrams with enough
algorithm21,22,35–37, demonstrating that the CALPHAD method accuracy for practical applications without the need of the heavy experimental
has played an important role in the design of advanced HEAs. work41. This argument has made the CALPHAD method a key building block in
There are two essential requirements to carry out calculations, the two largest driving forces in materials engineering today—ICME (Integrated
computational materials engineering)42 and MGI (Materials Genome Initiative)43.
using the CALPHAD method: computational software and The application of the CALPHAD method requires both the thermodynamic
database. database to provide the Gibbs energies (as a function of pressure, temperature, and
The computational software could significantly affect the effi- composition) for the individual phases and the computational software to calculate
ciency of HTC since massive calculations are carried out. In the the equilibrium state by an energy-minimization procedure. Over the past three
decades, the development of consistent multicomponent thermodynamic databases
example of the Al–Cr–Fe–Mn–Ti quinary system, it takes only a has grown steadily, and several commercial software, such as PandatTM44, Thermo-
few hours on a desktop computer to finish the HTCs (~3 ´ 103 ) Calc45 and FactSage46, has become available. Although the CALPHAD approach
within the composition range of 5–50 at.% for each element and has been well accepted in the design and development of advanced
finally screen out eight potential alloys. However, if a step size of 1 materials21,47–51, its full potential has not been fully released due to the low effi-
at.% for this quinary system is used, the computational time will ciency to explore the entire composition and temperature space of a multi-
component system. Here we need to emphasize that this trend is due to the
easily increase to a few thousands of hours, posing a challenge for a limitation of the computational tools, but not the CALPHAD method.
single personal computer. The computational time will be scaled up In order to meet the increasing demand for massive calculations in the field of
substantially with higher-order systems (the number of components HEAs, an HTC tool in the frame of CALPHAD is developed and implemented in
> 5). Fortunately, the current HTC module can be easily converted PandatTM 44. This CALPHAD-based HTC tool enables thousands of calculations in
a defined compositional space to be performed automatically. Alloy compositions
to parallel calculations for efficiently utilizing the available com- that satisfy user-defined criteria can then be identified through mining the
putational resource, which is the target that we are working on. In thousands of simulated results. Here we would like to emphasize that the current
addition, the CALPHAD-developed datasets are multi-dimensional HTC module is a combinatorial tool for both the high-throughput calculation and
and very large. Thus, the organization of the large volume of data high-throughput screening. Its speed and efficiency are markedly improved and
significantly advantageous over the conventional CALPHAD method.
will be especially critical for efficiently retrieving the requested
subsets of the large datasets. At present, each calculation is stored
separately in a workspace and organized by its calculation identity Sample preparation. To verify the reliability of the current HTC method, eight
different alloys identified from thousands of compositions were fabricated via the
(ID). The calculation ID is usually the alloy composition, and the arc-melting method with 99.9 weight percent (wt%) purity constituent elements.
detailed information of each calculation is represented by the pre- The nominal chemical compositions of these alloys are listed in Supplementary
defined features, such as the composition, temperature, phase Table 1. To ensure chemical homogeneity, the ingot was melted at least six times
fraction, etc. Note that we can generally include all kinds of ther- before drop-casting.
mochemical and thermophysical properties in an individual CAL-
PHAD calculation. The users can easily access these results via their Microstructural characterizations. Scanning-electron (SEM), APT, and TEM
calculation IDs and efficiently retrieve the requested datasets were used for microstructural characterizations. Synchrotron XRD experiment was
performed on the 11-ID-C beamline at the Advanced Photon Source (APS),
through the customized features. Here, we would like to emphasize Argonne National Laboratory (ANL). A beam energy of 111 keV (wavelength of
that any algorithms that are able to be ascribed to these features can 0.1173 Å) with a beam size of 0.5 × 0.5 mm was used. The samples for the SEM and
be employed as the screening criteria for the CALPHAD-calculated XRD observations were polished, using the vibration polishing method after finer
datasets. Moreover, the current HTC tool allows users to customize grinding with a 1200 grit SiC paper. The TEM specimen was prepared by a Ga+
focus ion beam (FIB) with a final milling voltage of 5 keV. The APT specimens
the outputs of different types of CALPHAD calculations for the were fabricated, employing the method described by Thompson et al.52. Equilateral
possible combination with other datasets for machine learning and triangular prism wedges were lifted out, using a Kleindiek MM3a, mounted on Si
the usage of other data mining tools. microtip array posts, sharpened using a 30 kV Ga+ ion beam, and cleaned using a
On the other hand, the accuracy of the CALPHAD method 2 kV ion beam. The APT experiment was run, employing a CAMECA LEAP
4000XHR in a laser mode with 60 pJ, 10-ps-laser pulsed, a 30K-base temperature,
depends on the reliability of the thermodynamic database. The 0.5% detection rate, and a pulse repetition rate that allowed for all ions to be
microstructure discrepancy between the thermodynamic predic- collected in the mass spectrum. The APT results were reconstructed and analyzed
tion and experimental observation indicates that our current using the CAMECA’s interactive visualization and analysis software (IVAS 3.8).
thermodynamic database needs further improvement on the
stability of the L21 phase. In addition to the current experimental Mechanical tests. Room and elevated-temperature compression tests of these
results, long-period annealing on these identified alloys and eight alloys were carried out at the strain rate of 2 × 10−4 s−1, using the hydraulic
quantitative characterization will be necessary in order to Mechanical Testing System (MTS). The sample size for the compression tests is 3
mm in diameter and 6 mm in length. Each test was repeated three times.
improve our current thermodynamic database.
Our results on the discovered LWHEAs demonstrate the sen-
sitivity of the nanoscaled L21 precipitates-strengthened BCC Neutron-scattering measurements. In situ neutron-scattering experiments were
performed on selected discovered alloys at the Spallation Neutron Source (SNS),
microstructure to alloys’ chemical compositions. Therefore, the Oak Ridge National Laboratory (ORNL), to understand the deformation
scrutiny of the alloys’ chemical compositions that are related to mechanisms and the order-disorder transition behavior. In situ neutron-diffraction
the critical ordering temperature and morphology of L21 experiments were conducted on Alloys 1 and 7 during their uniaxial compressive

8 NATURE COMMUNICATIONS | (2021)12:4329 | https://doi.org/10.1038/s41467-021-24523-9 | www.nature.com/naturecommunications

NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-021-24523-9 ARTICLE

deformation on the VULCAN Engineering Materials Diffractometer, SNS, ORNL. References

The incident neutron beam with a beam size of 5 mm × 6 mm hits the cylindrical 1. Song, G. et al. Ferritic alloys with extreme creep resistance via coherent
samples with a size of ϕ6 mm × 12 mm. To obtain a better d-spacing resolution, the hierarchical precipitates. Sci. Rep. 5, 16327 (2015).
high-resolution (HR) mode was chosen. After the measurements, the data were 2. Sun, Z. et al. New design aspects of creep-resistant NiAl-strengthened ferritic
analyzed by the single-peak fitting method, using the event-based data reduction alloys. Scr. Mater. 68, 384–388 (2013).
software, VULCAN Data Reduction and Interactive Visualization softwarE 3. Yeh, J. W. et al. Nanostructured high-entropy alloys with multiple principal
(VDRIVE)53. The PDFs of Alloys 1 and 7 were measured on the Nanoscale- elements: novel alloy design concepts and outcomes. Adv. Eng. Mater. 6,
Ordered Materials Diffractometer (NOMAD) instrument at SNS, ORNL54. The 299–303 (2004).
total neutron-scattering data were collected from RT to 1050 °C/1080 °C, using the 4. Cantor, B., Chang, I., Knight, P. & Vincent, A. Microstructural development in
ILL furnace installed on the NOMAD instrument. The samples were put into equiatomic multicomponent alloys. Mater. Sci. Eng. A 375, 213–218 (2004).
vanadium cans, and then heated at a ramp rate of 10 °C/min. under a vacuum 5. Zhang, Y. et al. Microstructures and properties of high-entropy alloys. Prog.
environment (~10−6 mbar). At selected temperatures, the samples were held for Mater. Sci. 61, 1–93 (2014).
24 min. for data collection. The PDF, G(r), was obtained by a Fourier transfor- 6. Miracle, D. B. & Senkov, O. N. A critical review of high entropy alloys and
mation of the structure function, S(Q), using a Qmax of 22 Å−1 55, related concepts. Acta Mater. 122, 448–511 (2017).
Z
2 7. Diao, H. Y., Feng, R., Dahmen, K. A. & Liaw, P. K. Fundamental deformation
GðrÞ ¼ Q½SðQÞ  1 sinðQrÞdQ ð3Þ
π behavior in high-entropy alloys: An overview. Curr. Opin. Solid State Mater.
where Q is the scattering vector. The measured PDFs were fitted with the MC- Sci. 21, 252–266 (2017).
calculated 2000 atoms structure, using the PDFgui software56. 8. George, E. P., Raabe, D. & Ritchie, R. O. High-entropy alloys. Nat. Rev. Mater.
4, 515–534 (2019).
9. Lei, Z. et al. Enhanced strength and ductility in a high-entropy alloy via
AIMD simulations. The atomic structure was predicted from AIMD simulations via ordered oxygen complexes. Nature 563, 546–550 (2018).
the Vienna Ab Initio Simulation Package (VASP)57,58 in a canonical ensemble, i.e., a 10. Yang, T. et al. Multicomponent intermetallic nanoparticles and superb
constant mole, volume, and temperature. Newton’s equations of motion were mechanical behaviors of complex alloys. Science 362, 933–937 (2018).
integrated, using the Verlet algorithm59 with a time step of 1 fs, and the
11. Sohn, S. S. et al. Ultrastrong medium-entropy single-phase alloys designed via
atomic-configuration relaxation and temperature were controlled by a Nose
severe lattice distortion. Adv. Mater. 31, 1807142 (2019).
thermostat60. Projector augmented-wave (PAW) potentials61 and the revised
12. Ma, E. Unusual dislocation behavior in high-entropy alloys. Scr. Mater. 181,
Perdew–Burke–Ernzerhof (PBE)62 gradient approximation to the exchange-
127–133 (2020).
correlation functional were used. We applied spin polarization, a single k-point, and
an enhanced energy cutoff of 350 eV within a cubic supercell of 200 atoms. The liquid 13. Antonov, S., Detrois, M. & Tin, S. Design of novel precipitate-strengthened
densities were determined by adjusting the cell volume so that the average pressure Al–Co–Cr–Fe–Nb–Ni high-entropy superalloys. Metall. Mater. Trans. A 49,
vanished in equilibrium at T = 1600 °C, yielding an atomic volume of 12.72 Å3/atom. 305–320 (2018).
Subsequently, hybrid Monte Carlo/molecular dynamics simulations63 were performed 14. Detrois, M. et al. Design and thermomechanical properties of a γʹ precipitate-
at this volume to facilitate sampling of the equilibrium ensemble. strengthened Ni-based superalloy with high entropy γ matrix. J. Alloys
The atomic structure in the liquid state can reveal the useful information about Compounds 792, 550–560 (2019).
the preferred interatomic bonding that may impact the formation of the disordered 15. Miracle, D. B., Tsai, M.-H., Senkov, O. N., Soni, V. & Banerjee, R. Refractory
solid solution during solidification18,64. The partial PDF gives the information high entropy superalloys (RSAs). Scr. Mater. 187, 445–452 (2020).
about the probability of such bond formation by measuring the intensity of near- 16. Liang, Y.-J. et al. High-content ductile coherent nanoprecipitates achieve
neighbor pairs against the total random distribution, and partial PDF [g ab ðrÞ] was ultrastrong high-entropy alloys. Nat. Commun. 9, 4063 (2018).
calculated, using 17. Feng, R. et al. Design of light-weight high-entropy alloys. Entropy 18, 333
(2016).
V 1 N N
g ab ðrÞ ¼ ∑ a ∑ b hδðjr ij j  rÞi ð4Þ 18. Feng, R. et al. Phase stability and transformation in a light-weight high-
N a N b 4πr 2 i¼1 j¼1 entropy alloy. Acta Mater. 146, 280–293 (2018).
where V is the volume of the supercell, Na and Nb are the numbers of elements, 19. Rickman, J. M. et al. Materials informatics for the screening of multi-principal
a and b, jr ij jis the distance between elements, a and b, and the bracket, < >, denotes elements and high-entropy alloys. Nat. Commun. 10, 2618 (2019).
the time average of different configurations. 20. Coury, F. G., Clarke, K. D., Kiminami, C. S., Kaufman, M. J. & Clarke, A. J.
High throughput discovery and design of strong multicomponent metallic
MC simulations. A lattice-gas model can describe the order-disorder phase solid solutions. Sci. Rep. 8, 8600 (2018).
transitions for the same and fixed lattice. In principle, sufficiently large lattices with 21. Miracle, D. B. et al. Exploration and development of high entropy alloys for
sufficiently accurate interactions between the lattice points can accurately describe structural applications. Entropy 16, 494–525 (2014).
the real phase transitions that do not involve only changes of lattice types. The 22. Senkov, O. N., Miller, J. D., Miracle, D. B. & Woodward, C. Accelerated
serious restrictions of lattice-gas models include the effects of lattice distortion and exploration of multi-principal element alloys with solid solution phases. Nat.
lattice vibration. Still, such models are found to successfully predict the order- Commun. 6, 6529 (2015).
disorder phase transitions in a number of conventional and high-entropy alloys 23. Krein, R., Palm, M. & Heilmaier, M. Characterization of microstructures,
(HEAs)3,33,65–68. mechanical properties, and oxidation behavior of coherent A2+L21 Fe–Al–Ti.
In this study, a lattice-gas model was constructed to describe the order-disorder J. Mater. Res. 24, 3412–3421 (2009).
phase transitions in a BCC lattice. We performed Metropolis MC simulations, 24. Stallybrass, C. & Sauthoff, G. Ferritic Fe–Al–Ni–Cr alloys with coherent
using the nearest-neighbor interaction model for simplicity. In the nearest- precipitates for high-temperature applications. Mater. Sci. Eng. A 387-389,
neighbor interaction model, only the atomic interactions of the nearest neighbors 985–990 (2004).
between atoms are considered. The interaction parameters,J ij , are computed by 25. Zhou, Y., Jin, X., Zhang, L., Du, X. & Li, B. A hierarchical nanostructured
Fe34Cr34Ni14Al14Co4 high-entropy alloy with good compressive mechanical
J ij ¼ ΔH ij =z ð5Þ
properties. Mater. Sci. Eng. A 716, 235–239 (2018).
where ΔH ij is the formation energy per atom of the solute pair, i, j, and z is the 26. Shaysultanov, D. G. et al. Novel Fe36Mn21Cr18Ni15Al10 high entropy alloy with
number of the nearest-neighbor bonds per atom. In the BCC structure, z = 8/2 = bcc/B2 dual-phase structure. J. Alloys Compounds 705, 756–763 (2017).
4. Although each atom has 8 nearest-neighbor bonds, two atoms share them. 27. Morris, D. G. & Gunther, S. Room and high temperature mechanical
Hence, the number has to be divided by 2. The formation energies are taken from behaviour of a Fe3Al-based alloy with α-α″ microstructure. Acta Mater. 45,
the Aflow database of Curtarolo et al.69. The J ij parameters used in this study are 811–822 (1997).
tabulated in Table 2. 28. Michalcová, A. et al. Laser additive manufacturing of iron aluminides
Applying numbers listed in Table 2, we performed three Metropolis MC strengthened by ordering, borides or coherent Heusler phase. Mater. Des. 116,
simulations. Two are for Alloys 1 and 7 using 2000 atoms with 107 MC steps, the 481–494 (2017).
third for the L21 phase using 250 atoms with 106 MC steps. 29. Reed, R. C. The Superalloys: Fundamentals and Applications (Cambridge
University Press, 2008).
30. Cowley, J. M. An approximate theory of order in alloys. Phys. Rev. 77,
Data availability 669–675 (1950).
The data that support the findings of this study are available from the corresponding 31. Cowley, J. M. X-Ray measurement of order in single crystals of Cu3Au. J. Appl.
author upon reasonable request. Phys. 21, 24–30 (1950).
32. Warren, B. E. X-ray Diffraction (Courier Corporation, 1990).
Received: 14 November 2018; Accepted: 16 June 2021; 33. Santodonato, L. J., Liaw, P. K., Unocic, R. R., Bei, H. & Morris, J. R. Predictive
multiphase evolution in Al-containing high-entropy alloys. Nat. Commun. 9,
4520 (2018).

NATURE COMMUNICATIONS | (2021)12:4329 | https://doi.org/10.1038/s41467-021-24523-9 | www.nature.com/naturecommunications 9

ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-021-24523-9

34. Wittmann, M., Baker, I. & Munroe, P. R. The structure and mechanical 67. Eisenbach, M., Pei, Z. & Liu, X. First-principles study of order-disorder
properties of Fe2AlMn single crystals. Philos. Mag. 84, 3169–3194 (2004). transitions in multicomponent solid-solution alloys. J. Phys. Condens. Matter
35. Abu-Odeh, A. et al. Efficient exploration of the high entropy alloy 31, 273002 (2019).
composition-phase space. Acta Mater. 152, 41–57 (2018). 68. Pei, Z., Li, R., Gao, M. C. & Stocks, G. M. Statistics of the NiCoCr medium-
36. Tancret, F., Toda-Caraballo, I., Menou, E., Rivera & Díaz-Del-Castillo, P. E. J. entropy alloy: Novel aspects of an old puzzle. npj Comput. Mater. 6, 122
Designing high entropy alloys employing thermodynamics and Gaussian (2020).
process statistical analysis. Mater. Des. 115, 486–497 (2017). 69. Curtarolo, S. et al. AFLOWLIB. ORG: a distributed materials properties
37. Senkov, O. N., Miller, J. D., Miracle, D. B. & Woodward, C. Accelerated repository from high-throughput ab initio calculations. Comput. Mater. Sci.
exploration of multi-principal element alloys for structural applications. 58, 227–235 (2012).
CALPHAD 50, 32–48 (2015).
38. Hillert, M. Phase Transformations (ASM, 1968).
39. Kaufman, L. & Bernstein, H. Computer Calculation of Phase Diagrams Acknowledgements
(Academic Press, 1970). P.K.L. and R.F. very much appreciate the supports from (1) the National Science
40. Chang, Y. A. et al. Phase diagram calculation: past, present and future. Prog. Foundation (DMR-1611180 and 1809640) with program directors, Drs. J. Yang, G.
Mater. Sci. 49, 313–345 (2004). Shiflet, and D. Farkas and (2) the US Army Research Office (W911NF-13-1-0438 and
41. Kattner, U. R. The CALPHAD method and its role in material and process W911NF-19–2-0049) with program managers, Drs. M.P. Bakas, S.N. Mathaudhu, and D.
development. Tecnol. Met. Mater. Min. 13, 3–15 (2016). M. Stepp. M.C.G. acknowledges the support of the U.S. Department of Energy (DOE)’s
42. National Research Council, Committee on Integrated Computational Fossil Energy Cross-Cutting Technologies Program at the National Energy Technology
Materials Engineering. Integrated Computational Materials Engineering: A Laboratory (NETL) under the RSS contract, 89243318CFE000003. M.W. was supported
Transformational Discipline for Improved Competitiveness and National through the DOE grant, SC-0014506, for the development of simulation methods. APT
Security (National Academic Press, 2008). was conducted at the Oak Ridge National Laboratory (ORNL)’s Center for Nanophase
43. Office of Science and Technology Policy. Materials Genome Initiative for Materials Sciences (CNMS), which is a U.S. DOE Office of Science User Facility. The
Global Competitiveness (Office of Science and Technology Policy, 2011). neutron-scattering work was carried out at the Spallation Neutron Source (SNS), which is
44. PandatTM. Thermodynamic Calculations and Kinetic Simulations the U.S. DOE user facility at ORNL, sponsored by the Scientific User Facilities Division,
(CompuTherm LLC, 2021). Office of Basic Energy Sciences. This research used resources of the Advanced Photon
45. Thermo-Calc. Thermodynamic Equilibrium Calculations (Thermo-Calc Source (APS), a U.S. DOE Office of Science User Facility operated for the DOE Office of
Software, 2021). Science by the Argonne National Laboratory (ANL) under Contract No. DE-AC02-
46. FactSage. Thermochemical Software (FactSage, 2021). 06CH11357. The authors thank James Burns for his assistance in performing the APT
47. Zhang, C., Zhang, F., Chen, S. & Cao, W. Computational thermodynamics sample preparation and running the APT experiments.
aided high-entropy alloy design. JOM 64, 839–845 (2012).
48. Zhang, C. & Gao, M. C. in: High-Entropy Alloys—Fundamentals and Author contributions
Applications (eds Gao, M. C., Yeh, J.-W., Liaw, P. K. & Zhang, Y.) (Springer R.F. conducted and analyzed the experiments. C.Z. and F.Z. performed the CALPHAD
International Publishing, 2016). calculations. M.C.G., L.O, and M.W. carried out and analyzed the AIMD simulations.
49. Gao, M. C. & Alman, D. E. Searching for next single-phase high-entropy alloy Z.P. and M.W. performed the MC simulations. R.F., Y.C., D.M., and K.A. conducted and
compositions. Entropy 15, 4504–4519 (2013). analyzed the neutron-scattering experiments. J.D.P. performed the APT experiments and
50. Zhang, F. et al. An understanding of high entropy alloys from phase diagram analysis. R.F. and Y.R. carried out the synchrotron X-ray diffraction experiments. R.F.
calculations. CALPHAD 45, 1–10 (2014). and C.Z. led the writing of the paper with contributions from all authors. P.K.L., C.Z.,
51. Zhang, C. et al. Understanding phase stability of Al–Co–Cr–Fe–Ni high M.C.G., J.A.H., and F.Z. initiated and supervised the project. All authors contributed to
entropy alloys. Mater. Des. 109, 425–433 (2016). the discussions.
52. Thompson, K. et al. In situ site-specific specimen preparation for atom probe
tomography. Ultramicroscopy 107, 131–139 (2007).
53. An, K. VDRIVE—Data Reduction and Interactive Visualization Software for Competing interests
Event Mode Neutron Diffraction ORNL Report, ORNL-TM-2012-2621 (Oak The authors declare no competing interests.
Ridge National Laboratory, 2012).
54. Neuefeind, J., Feygenson, M., Carruth, J., Hoffmann, R. & Chipley, K. K. The
nanoscale ordered materials diffractometer NOMAD at the spallation neutron
Additional information
Supplementary information The online version contains supplementary material
source SNS. Nucl. Instrum. Methods Phys. Res. 287, 68–75 (2012).
available at https://doi.org/10.1038/s41467-021-24523-9.
55. Egami, T. & Billinge, S. J. Underneath the Bragg Peaks: Structural Analysis of
Complex Materials (Elsevier, 2003).
Correspondence and requests for materials should be addressed to C.Z., M.C.G. or P.K.L.
56. Farrow, C. L. et al. PDFfit2 and PDFgui: computer programs for studying
nanostructure in crystals. J. Phys. Condens. Matter 19, 335219 (2007). Peer review information Nature Communications thanks KC Harikumar, Levente Vitos
57. Kresse, G. & Hafner, J. Ab initio molecular dynamics for liquid metals. Phys. and the other, anonymous, reviewer(s) for their contribution to the peer review of
Rev. B 47, 558–561 (1993). this work.
58. Kresse, G. & Furthmüller, J. Efficient iterative schemes for ab initio total-energy
calculations using a plane-wave basis set. Phys. Rev. B 54, 11169–11186 (1996). Reprints and permission information is available at http://www.nature.com/reprints
59. Verlet, L. Computer “Experiments” on Classical Fluids. I. Thermodynamical
Properties of Lennard-Jones Molecules. Phys. Rev. B 159, 98–103 (1967).
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in
60. Nosé, S. A unified formulation of the constant temperature molecular
published maps and institutional affiliations.
dynamics methods. J. Chem. Phys. 81, 511 (1984).
61. Blöchl, P. E. Projector augmented-wave method. Phys. Rev. B. 50,
17953–17979 (1994).
62. Perdew, J. P. et al. Restoring the density-gradient expansion for exchange in Open Access This article is licensed under a Creative Commons
solids and surfaces. Phys. Rev. Lett. 100, 136406 (2008). Attribution 4.0 International License, which permits use, sharing,
63. Widom, M., Huhn, W. P., Maiti, S. & Steurer, W. Hybrid Monte Carlo/ adaptation, distribution and reproduction in any medium or format, as long as you give
molecular dynamics simulation of a refractory metal high entropy alloy. appropriate credit to the original author(s) and the source, provide a link to the Creative
Metall. Mater. Trans. A 45, 196–200 (2014). Commons license, and indicate if changes were made. The images or other third party
64. Ganesh, P. & Widom, M. Ab initio simulations of geometrical frustration in material in this article are included in the article’s Creative Commons license, unless
supercooled liquid Fe and Fe-based metallic glass. Phys. Rev. B 77, 014205 indicated otherwise in a credit line to the material. If material is not included in the
(2008). article’s Creative Commons license and your intended use is not permitted by statutory
65. Santodonato, L. J. et al. Deviation from high-entropy configurations in the regulation or exceeds the permitted use, you will need to obtain permission directly from
atomic distributions of a multi-principal-element alloy. Nat. Commun. 6, 5964 the copyright holder. To view a copy of this license, visit http://creativecommons.org/
(2015). licenses/by/4.0/.
66. Pei, Z., Eisenbach, M., Mu, S. & Stocks, G. M. Error controlling of the
combined Cluster-Expansion and Wang–Landau Monte-Carlo method and its
application to FeCo. Comput. Phys. Commun. 235, 95–101 (2019). © The Author(s) 2021

10 NATURE COMMUNICATIONS | (2021)12:4329 | https://doi.org/10.1038/s41467-021-24523-9 | www.nature.com/naturecommunications