Scripta Materialia 252 (2024) 116240

Contents lists available at ScienceDirect

Scripta Materialia
journal homepage: [Link]/scripta-materialia

Accelerated discovery of refractory high-entropy alloys for

strength-ductility co-optimization: An exploration in NbTaZrHfMo system
by machine learning
Cheng Wen a, b , Haicheng Shen a, c , Yuwan Tian a, c, * , Gongqi Lou a , Nanchuan Wang a ,
Yanjing Su d
a
School of Mechanical Engineering, Guangdong Ocean University, Zhanjiang, 524088, China
b
Zhanjiang Key Laboratory of Ocean Engineering and Equipment Corrosion and Protection, Zhanjiang, 524088, China
c
Guangdong Provincial Ocean Equipment and Manufacturing Engineering Technology Research Center, Zhanjiang, 524088, China
d
Beijing Advanced Innovation Center for Materials Genome Engineering, University of Science and Technology Beijing, Beijing, 100083, China

A R T I C L E I N F O A B S T R A C T

Keywords: High temperature (HT) strength and room temperature (RT) ductility trade-offs are always unavoidable in the
Refractory high-entropy alloys design of refractory high entropy alloys (RHEAs). Exploring the vast chemistry space to find optimally strong and
Strength-ductility co-optimization ductile RHEAs remains a highly challenging task. Herein, we formulate a machine learning-based design strategy
Machine learning
fusing uncertainty estimation and clustering analysis to explore a special alloy system (NbTaZrHfMo) for
Alloys design
NbTaZrHfMo
collaborative optimization of HT strength and RT ductility. Four non-equimolar alloys with a superior combi­
nation of HT strength, RT ductility and HT specific yield strength were discovered and synthetized. The influence
of elements on mechanical properties are analyzed and an optimal composition range is identified based on
model prediction. This work provides a general design approach enabling the concurrent optimization of con­
flicting properties with a small data-trained machine learning model, thereby producing a recipe to accelerate
the discovery of desired RHEAs and other materials within a vast search space.

Refractory high entropy alloys (RHEAs) are an emerging type of high entropy alloys (HEAs). Senkov et al. [7] further noted that there is a
temperature (HT) material that has drawn much attention due to their proportional trend between the VEC of RHEAs and their RT ductility, but
ability to retain high strength at 1000 ℃ and above [1-2]. By employing there is no clear boundary value separating toughness from brittleness.
refractory elements with high melting points (Tm) from Group IVB-VIB, VEC alone may not sufficiently characterize ductility in RHEAs. Never­
several RHEAs have already demonstrated HT strength that outperform theless, the consensus among scholars is that reducing VEC is beneficial
Ni-base superalloys. Typically, the NbTaMoWV alloy exhibits yield to their ductility design. Concerning the strength improvement,
stress up to 477 MPa, even at 1600 ℃ [3]. However, most RHEAs suffer increasing Mo content significantly contributes to the strength of RHEAs
from room temperature (RT) brittleness with compressive fracture strain [8], especially at HT [9-10]. Recently, Hu et al. [11] further revealed
(FSRT) below 10 % and thus are not machinable. As an exception, the that Mo-containing RHEAs have in general higher HT strength than
equimolar NbTaZrHfTi alloy exhibits outstanding deformability with those without Mo. The addition of W and Ta can also improve the
FSRT over 50%, but its weak temperature resistance is not competitive temperature resistance of RHEAs, but will seriously deteriorate their RT
compared to other RHEAs or superalloys [4]. Strength-ductility-driven ductility [12-15].
composition design has been a key issue in the development of RHEAs Exploring the vast chemistry space to find optimally strong and
since their proposal. Sheikh et al. [5] first proposed a valence electron ductile RHEAs remains a highly challenging task, given a palette con­
concentration (VEC) rule that there exists a threshold VEC of 4.5 below taining more than ten elements. Several efforts have been attempted to
which intrinsic ductility can be achieved in RHEAs. Mak et al. [6] found accelerate the discovery of alloys with desired properties. Maresca et al.
that VEC = 5 is an approximate borderline. Consequently, the use of VEC [16] performed a computationally guided search in an
< 5 is proposed for rapid screening of the ductile compositions of high Al-Cr-Hf-Mo-Nb-Ta-Ti-V-W-Zr HEA system to identify alloy

* Corresponding author.
E-mail address: tianyuwan90@[Link] (Y. Tian).

[Link]
Received 8 April 2024; Received in revised form 13 June 2024; Accepted 14 June 2024
Available online 24 June 2024
1359-6462/© 2024 Acta Materialia Inc. Published by Elsevier Ltd. All rights are reserved, including those for text and data mining, AI training, and similar
technologies.
C. Wen et al. Scripta Materialia 252 (2024) 116240

compositions with potentially high HT strength. Menou et al. [17-18] alloy in NbTaZrHfMo composition system, we rank all the alloys
proposed a computation-based strategy to design light and strong using non-dominated sorting based on these EI values to identify
single-phase HEAs by searching for alloys in the the Pareto front for the virtual space. The EI values of YS1000℃ and
Al-Co-Cr-Cu-Fe-Hf-Mn-Mo-Nb-Ni-Ru-Ta-Ti-V-W-Zr composition space FSRT for RHEAs are shown in the 2D Pareto plot in Fig. 1(c), where the
using an evolutionary algorithm. Similar work has been undertaken for boundary points define an EI-based Pareto front (EIPF) representing the
the composition search of 5-element HEAs with hardness improvement optimal trade-off between two objectives. All candidate compositions on
in a 16-element palette [19]. Actually, searching the broad swaths of the Pareto front (PF) are equally favorable and expected to provide
compositions within a given alloy system is likely to be more effective flexibility in the selection of potential alloys with strength-ductility
for finding the target alloy with the desired properties, avoiding blind improvement. In this study, clustering analysis on the EIPF is utilized
and time-consuming searches on many ingredient continents. Recently, to determine the potential alloys viable for experimental synthesis.
efficient composition design of RHEAs have been reported in some Specifically, we use the K-means algorithm and the “elbow method”
special chemistry systems such as Nb-Ta-Ti-Hf-Mo [20], W-Ta-V-Ti-Cr [26] to select a reasonable number of clusters centers (K) by calculating
[21], and W-Mo-V-Ta-Nb-Al [22]. the sum of squares within groups (WGSS) as K varies. As shown in Fig. 1
Here we focus on a not-yet systematically-explored alloys system (e), the WGSS decreases very slowly after K = 4, thus we consider four
NbTaZrHfMo, derived from the two typical RHEAs of NbTaMoWV and clusters as a good fit of the data. Another two widely-used methods give
NbTaZrHfTi. In this chemical space, Nb and Ta have a VEC value of 5 the same K value recommendation (see details in Supplementary Ma­
that is the threshold suggestion for toughness design in RHEAs [6], Zr terial). Thus, we cluster the data points in the EIPF using K = 4, and four
and Hf with low VEC value (4) should be beneficial to ductility corresponding alloys are accordingly identified, as marked in Fig. 1(c).
improvement. Mo addition may be unfavorable due to its higher VEC Table 1 provides the composition information of the newly-designed
value (6), but is expected to provide considerable (HT) strengthening of RHEAs. In Fig. 1(d), the model-predicted YS1000℃ and FSRT of RHEAs
RHEAs. We formulate a strategy based on machine learning (ML) by in the EIPF are plotted together with that of the original data, in which
integrating an uncertainty estimation for alloy search and clustering the error bars represent prediction uncertainty. These candidate alloys
analysis for alloy selection, as shown in Fig. 1(a), systematically exhibit decent properites combination, with 450 MPa ~ 1200 MPa
exploring the NbTaHfZrMo composition space to accelerate the dis­ YS1000℃ and 10% ~ 40% FSRT. According to the error bars, the pre­
covery of RHEAs with synergistic improvements in strength and diction deviation shows an increasing trend with the increase in both HT
ductility, i.e., the yield strength at HT (YS1000℃) and fracture strain at strength and RT ductility. It means that the model-prediction uncer­
RT (FSRT). tainty increases as the composition exploration moves to the high-value
Two datasets of YS1000℃ and FSRT are first assembled from the region of alloy’s properties.
literature and our lab. All the training data used for ML modeling are for The four designed RHEAs are then synthesized by vacuum arc-
alloys fabricated by vacuum arc-melting only and their property is melting and phase-detected by XRD before and after compression
measured in the as-cast form, in order to minimize the influence of deformation at 1000 ℃. As shown in Fig. 2(a) and (b), the XRD patterns
processing method. Accordingly, composition is regarded as the primary suggest that the phase structure is the same before and after hot defor­
factor influencing the properties of RHEAs. These data are then used to mation for alloys A1 and A2, with a disordered BCC solid solution. The
train ML models using support vector machine algorithm, which is al­ as-cast alloys A3 and A4 show a BCC structure with an intermetallic (IM)
ways used for modeling small data. Using the elemental content as input, phase, while IM is not detected for alloy A3 after HT deformation, and
the ML model is evaluated by a 5-fold cross-validation method. From some diffraction peaks of IM phase disappear for alloy A4. From alloy Al
Fig. 1(b), the root mean square error (RMSE) of the model prediction on to A4, the diffraction peak shifts to the right as Mo content increases
YS1000℃ is 138 MPa, and the correlation coefficient rcv between the could be identified, indicating a reduction of the lattice parameter. This
experimental strength and the predicted values approaches 0.9. The data may be ascribed to the replacement of Zr or Hf with the larger atomic
points distributed along the diagonal line indicates that the model is well radius of Mo.
fitted overall. However, there appears to be an obvious underestimation Fig. 2(c) and (d) provide the compressive stress-strain curves of the
of yield strength within the high-value range. For these alloys with designed RHEAs at RT and 1000 ℃. Fig. 3(a) illustrates a comparison
YS1000℃ over 800 MPa, the predicted RMSE is 218 MPa, far beyond the between the properties of these alloys to those in the original data. Alloy
average value, while RMSE of the remaining alloys is only 110 MPa. The A1 shows excellent RT ductility with FSRT over 50%, which is compa­
same situation occurs in the prediction on RT ductility (seen in Fig. S1). rable to the typical NbTaZrHfTi alloy, while its YS1000℃ value is nearly
Such model prediction is actually unfavorable for guiding RHEAs design, 1.5 times that of the latter. The YS1000℃ value of alloy A4 exceeds 1000
considering the goal of high strength and high ductility. MPa beyond most of the reported alloys. At a similar high-level of
Actually, for the small data-trained ML model, uncertainty consid­ YS1000℃, the FSRT of alloy A4 is improved by more than 3 times from
eration of the target prediction is essential, especially when applied to 4.3% (Al0.025Ti0.2V0.075Zr0.1Cr0.2Nb0.2Mo0.2) [27] to 17.1%. When
design materials in a large search space [23]. The expected improve­ compared to another typical NbTaMoWV alloy, alloy A4 shows a 26%
ment (EI) utility indicator balances exploitation (choosing the material increase in HT strength, and the significant improvement in FSRT re­
with the high predicted target) and exploration (using the predicted lieves the alloy of RT brittleness, making the alloy more mechanically
uncertainties to study regions of the searchspace where the model is less workable. It is worth emphasizing that the RT yield strength (YSRT) of
accurate), as shown in Fig. 1(a). The EI indicator has been demonstrated alloy A2 is 1501 MPa and the FSRT is 30.5%. In a recent study [7],
as efficient for accelerating materials discovery for single-objective Senkov et al. pointed out that RHEAs with a minimum of 30% FSRT may
optimization [24-25]. We therefore utilize EI indicators instead of possess commercially relevant levels of tensile ductility (~4%), which
directly using the ML-predicted YS1000℃ and FSRT to guide alloy dis­ are correlated with the compressive YSRT of ≤ 1500 MPa. The A2 alloy is
covery. The statistical method of ‘bootstrap’ sampling is used to evaluate such RHEAs that is currently the most likely to achieve the objective of
the properties-prediction uncertainty of each alloy in the NbTaZrHfMo extending the use temperature of structural alloys. Additionally, we can
system (the molar content of each element is restricted to 5~35 at.% to notice that the newly-designed RHEAs belong to the
produce 553,401 alloys). Then, the EI values of strength and ductility for (NbTa)0.48(ZrHf)xMo0.52-x subsystem, and from A1 to A4, the HT
a given alloy can be calculated according to the prediction distribution. strength increases gradually as the RT ductility decreases with the
The details of EI calculation could be found in our previous work [24]. addition of Mo and a certain reduction in Zr and Hf. According to the
Unlike the work of single-property optimization using an EI indica­ VEC rule, the increase of Zr and Hf lowers VEC and benefits alloys
tor, the balance between antagonistic objectives must be considered in ductility. This implies that the potential relationship between alloy
multi-objectives design. After obtaining the two EI values for each components and mechanical properties is identified by the model by

2
C. Wen et al. Scripta Materialia 252 (2024) 116240

Fig. 1. The ML-driven alloys design. (a) The workflow including four steps; (b) The prediction value of YS1000℃ for the alloys by ML model; (c) The EIPF of all the
alloys in the NbTaZrHfMo system and composition identification by clustering alloys in the EIPF; (d) ML-predicted YS1000℃ and FSRT of the alloys in the EIPF with
uncertainty estimation; (e) Determination of the number of cluster centers (K) by “elbow method”.

3
C. Wen et al. Scripta Materialia 252 (2024) 116240

Table 1
The composition, phase structure and mechanical properties of the designed RHEAs.
No. Designed RHEAs Phase (as-cast) YSRT, MPa YS1000℃, MPa YS1200℃, MPa FSRT, %

A1 Nb0.29Ta0.18Zr0.24Hf0.2Mo0.09 BCC 1217 714 304 >50

A2 Nb0.26Ta0.19Zr0.19Hf0.16Mo0.2 BCC 1501 974 669 30.5
A3 Nb0.31Ta0.17Zr0.15Hf0.14Mo0.23 BCC+IM 1487 979 710 20.1
A4 Nb0.27Ta0.21Zr0.13Hf0.13Mo0.26 BCC+IM 1547 1061 940 17.2
vs. NbTaZrHfMo BCC 1524 927 694 16
vs. NbTaZrHfTi BCC 929 295 92 >50
vs. NbTaMoWV BCC 1246 842 735 1.7

Fig. 2. The phase structure and mechanical properties of the newly-designed RHEAs. (a) XRD paterns of the alloys as-cast (a) and after deformatin at 1000℃ (b); The
compressive stress-strain curves of the newly-designed alloys at RT (c) and 1000℃ (d).

learning from the original data information. This is very important rapidly above 1000 ℃, typically like RHEAs containing Al [27,30] and
especially when we want to design alloys in a vast space by adjusting the Si [30]. According to Fig. 3(b), as a comparison, the YSRT of
ratios of multiple elements. AlMo0.5NbTa0.5TiZr can reach up to 2000 MPa, but it drops precipi­
To further explore the temperature-resistance performance of the tously to 250 MPa at 1200 ℃, indicating its low temperature resistance.
newly-designed RHEAs, compressive deformation tests are conducted at Meanwhile, the FSRT below 10% indicates its poor RT workability.
1200 ℃. Fig. 3(b) compares their YS1200℃ values to that of typical Among the designed RHEAs, alloy A4 exhibits the best HT strength with
RHEAs and shows significant improvement in softening resistance. YS1200℃ reaching 940 MPa, and its yield strength loss is below 12% from
Particularly, alloys A2 and A3 show a high yield stress at 1200 ℃, and 1000 ℃ to 1200 ℃, which is superior to almost all other RHEAs. A
the simultaneously-generated higher RT ductility make them more much-delayed reduction behavior in strength indicates its substantial
attractive than typical RHEAs (for example, NbTaMoWVTi [29] with resistance to softening at HT. In contrast to the typical NbMoTaWV
FSRT below 14%). The IM phase cannot be detected for alloy A3 after HT alloy, the YS1200℃ value of alloy A4 increases by 27.9%. The FSRT
deformation, indicating the possible dissolution of secondary phases at exceeding 17% also indicates it has much better RT workability than
1000 ℃ and greater, as previously reported for multi-phase RHEAs [30]. NbMoTaWV. When compared to the equimolar NbTaZrHfMo, the
However, the strength loss is only 34.2% from RT to 1000 ℃ and is designed alloys A2 and A3 show a similar temperature-bearing level but
52.3% from RT to 1200 ℃. It can be inferred that the SSS effect [28] is with a notable increase ratio of FSRT at 25.6% and 90.6 %, and the yield
important for alloy A3 at HT and thus can contribute to its strength. strength of alloy A4 at 1000 ◦ C and 1200 ◦ C increase by 14.5% and
RHEAs with extremely high YSRT values (>1600 MPa) are not attractive 35.4%, respectively. With the further intuitive comparison of
HT structural materials as they are brittle at RT and always lose strength YS1000℃/YSRT and YS1200℃/YSRT shown in Fig. 3(c) and Fig. 3(d), the

4
C. Wen et al. Scripta Materialia 252 (2024) 116240

Fig. 3. The properties comparision between the newly-designed alloys and the reported RHEAs. (a) The HT strength and RT ductility; (b) The temperature-resistence
in strength; (c) The ratio of yield stress at 1000 ℃ to that at RT (YS1000℃/YSRT); (d) The ratio of yield stress at 1200 ℃ to that at RT (YS1200℃/YSRT); (e) The specific
yield strength at 1000 ℃ (YS1000℃/ρ) and (f) the specific yield strength at 1200 ℃ (YS1200℃/ρ) .

newly-designed alloys exhibit better strength retention than most of the Al/Ti/V-containing RHEAs naturally show low density and thus produce
reported RHEAs, simultaneously exhibiting superior RT ductility. high YS1000℃/ρ, their very poor RT ductility also leads to little
Density is another important concern for alloys used in flight appli­ machineability, making them unattractive for various applications.
cations, and alloys with high specific strength (strength-to-density ratio) Regarding YS1200℃/ρ, our designed alloys exhibit almost absolute su­
are preferable for reducing weight and fuel consumption. In fact, the periority to the reported RHEAs, except for NbTaTiHfMo, by substituting
existing RHEAs always suffer from high density due to the addition of Ti with Zr to lower alloy density. The FSRT-YS1200℃/ρ PF moves forward
heavy elements like W, Ta and Hf. Therefore, we compare the specific obviously, indicating a great attraction and potential of the new RHEAs
yield strength (SYS) at 1000 ℃ (YS1000℃/ρ) and 1200 ℃ (YS1200℃/ρ) of to be used as HT structural material. By exploring the special
the newly-synthesized alloys with the reported RHEAs. As shown in NbTaZrHfMo alloy system from an equimolar composition to non-
Fig. 3(e), the new alloys form a new PF of the YS1000℃/ρ and FSRT. For equimolar compositions based on our ML approach, further extensive
those RHEAs exhibiting high RT ductility (over 50% FSRT), alloy A1 and significant co-improvement of RT ductility, HT strength and HT
shows a conspicuous improvement of YS1000℃/ρ. Alloy A2 dominates specific yield strength can be realized. The microstructure of the newly-
such RHEAs with medium ductility with 25~35 % FSRT. Although the designed RHEAs present typical dendrite-interdendrite morphology (see

5
C. Wen et al. Scripta Materialia 252 (2024) 116240

in Supplementary Material). In summary, we formulate a data-driven strategy combining ML-

Even if the ML models may lead to an underestimated prediction of model prediction of alloy properties, EI-based uncertainty utility and
alloy properties within the high-value region, it also can yield reason­ clustering-based selection of alloy composition to systematically explore
able trends, providing us an effective tool to understand RHEAs and the NbTaZrHfMo chemical space, an alloy system determined by
thereby accelerating composition screening. Therefore, we use the ML- experience-based intuition, to accelerate the design of RHEAs with a
model to predict YS1000℃ and FSRT for all alloys in the NbTaZrHfMo synergistic improvement in HT strength and RT ductility. This approach
system to reveal the element influence on alloy properties. As plotted in is demonstrated as efficient, and four non-equimolar alloys with a su­
Fig. 4, the desired properties region is marked by a dashed box repre­ perior combination of HT strength, RT ductility and HT specific yield
senting a co-optimization for HT strength (temperature-resistance) and strength are discovered and synthetized, which are expected to serve as
RT ductility (workability) of RHEAs. Among the five elements, Mo potential alternative materials for HT structural applications. Addi­
shows the most significant impact on alloys’ mechanical properties, and tionally, the influence of elements on alloy properties are also analyzed
Mo content tends to decrease from the region with high FSRT to that with in this system, based on ML model prediction, to help us to understand
high YS1000℃. The preferable Mo content is approximately 6~28 at.% in the composition design of RHEAs. This work provides a general design
such alloy system, which actually covers the range of the newly- approach enabling concurrent optimization of conflicting properties
designed alloys (9~27 at.%). It is worth noting that excessively high with small data-trained machine learning. It also offers a recipe to
or low Mo content is detrimental to multi-properties optimization in rapidly optimize other special materials systems within a vast search
RHEAs design. The increase of Zr and Hf are beneficial to alloy ductility, space and multi-property requirements.
and there is a difference in the favorable range (Hf, 12~20 at.%, Zr,
15~28 at.%) between them even if they have the same low VEC value of CRediT authorship contribution statement
4. The trend of a lower content of Hf is advantageous given its higher
cost when considering the practical application of RHEAs. The effect of Cheng Wen: Writing – original draft, Methodology, Investigation,
Ta content on alloy properties is not obvious in such composition sys­ Conceptualization. Haicheng Shen: Validation, Software, Data cura­
tem. The favorable content is a narrow range approximately 16~25 at. tion. Yuwan Tian: Writing – review & editing, Formal analysis,
%, and excessively high content is always undesirable for both HT Conceptualization. Gongqi Lou: Software. Nanchuan Wang: Valida­
strength and RT ductility. From the heatmap highlighting Nb, it seems tion, Software, Data curation. Yanjing Su: Resources,
that a medium-to-high content of Nb is beneficial for alloy design with a Conceptualization.
strength-ductility balance. It also shows that Nb content below 10 at.%
will significantly reduce the RT ductility and HT strength of RHEAs.

Fig.4. The heatmap of properties distribution predicted by ML model for NbTaZrHfMo composition system, revealing the ML-identified relationship between the five
elements and alloys properties.

6
C. Wen et al. Scripta Materialia 252 (2024) 116240

Declaration of competing interest [13] Q. Li, H. Zhang, D. Li, Z. Chen, Z. Qi, The effect of configurational entropy on
mechanical properties of single BCC structural refractory high-entropy alloys
systems, Int. J. Refract. Met. Hard Mater. 93 (2020) 105370, [Link]
The authors declare that they have no known competing financial 10.1016/[Link].2020.105370.
interests or personal relationships that could have appeared to influence [14] V.T. Nguyena, M. Qian, Z. Shi, T. Song, L. Huang, J. Zou, Compositional design of
the work reported in this paper. strong and ductile (tensile) Ti-Zr-Nb-Ta mediumentropy alloys (MEAs) using the
atomic mismatch approach, Mater. Sci. Eng., A 742 (2019) 762–772, [Link]
org/10.1016/[Link].2018.11.054.
Acknowledgements [15] M. Wang, Z. Ma, Z. Xu, X. Cheng, Microstructures and mechanical properties of
HfNbTaTiZrW and HfNbTaTiZrMoW refractory high-entropy alloys, J. Alloys
Compd. 803 (2019) 778–785, [Link]
This work was supported by the National Natural Science Foundation [16] F. Maresca F, W.A. Curtin, Mechanistic origin of high strength in refractory BCC
of China (Grant No. 52203293) and the National Key Research and high entropy alloys up to 1900K, Acta Mater 182 (2020) 235–249, [Link]
Development Program of China (Grant No. 2021YFB3802100). 10.1016/[Link].2019.10.015.
[17] E. Menou, I. Toda-Caraballo, P.E.J. Rivera-Díaz-del-Castillo, C. Pineau, E. Bertrand,
G. Ramstein, F. Tancret, Evolutionary design of strong and stable high entropy
Supplementary materials alloys using multi-objective optimisation based on physical models, statistics and
thermodynamics, Mater. Des. 143 (2018) 185–195, [Link]
matdes.2018.01.045.
Supplementary material associated with this article can be found, in [18] E. Menou, E. Tancret, I. Toda-Caraballo I, G. Ramstein, P. Castany, E. Bertrand,
the online version, at doi:10.1016/[Link].2024.116240. N. Gautier, P.E.J. Rivera Díaz-Del-Castillo, Computational design of light and
strong high entropy alloys (HEA): Obtainment of an extremely high specific solid
solution hardening, Scr. Mater. 156 (2018) 120–123, [Link]
References
scriptamat.2018.07.024.
[19] J.M. Rickman, H.M. Chan, M.P. Harmer, J.A. Smeltzer, C.J. Marvel, A. Roy,
[1] J. Chen, X. Zhou, W. Wang, B. Liu, Y. Lv, W. Yang, D. Xu, Y. Liu, A review on G. Balasubramanian, Materials informatics for the screening of multi-principal
fundamental of high entropy alloys with promising high-temperature properties, elements and high-entropy alloys, Nat. Commun. 10 (2019) 2618, [Link]
J. Alloys Compd. 760 (2018) 15–30, [Link] 10.1038/s41467-019-10533-1.
jallcom.2018.05.067. [20] Y. Rao, C. Baruffi, A.D. Luca, C. Leinenbach, W.A. Curtin, Theory-guided design of
[2] O.N. Senkov, D.B. Miracle, K.L. Chaput, J.P. Couzinie, Development and high-strength, high-melting point, ductile, low-density, single-phase BCC high
exploration of refractory high entropy alloys-A review, J. Mater. Res. 19 (2018) entropy alloys, Acta Mater. 237 (2022) 118132, [Link]
3092–3128, [Link] actamat.2022.118132.
[3] O.N. Senkov, G.B. Wilks, D.B. Miracle, C.P. Chuang, P.K. Liaw, Refractory high- [21] Il.H. Kim, H.S. Oh, K.S. Lee, E.S. Park, Optimization of conflicting properties via
entropy alloys, Intermetallics 18 (2010) 1758–1765, [Link] engineering compositional complexity in refractory high entropy alloys, Scr.
intermet.2010.05.014. Mater. 199 (2021) 113839, [Link]
[4] O.N. Senkov, J.M. Scott, S.V. Senkov, F. Meisenkothen, D.B. Miracle, C. [22] B. Vela, C. Acemi, P. Singh, T. Kirk, W. Trehern, E. Norris, D.D. Johnson,
F. Woodward, Microstructure and elevated temperature properties of a refractory I. Karaman, R. Arr´oyave, High-throughput exploration of the WMoVTaNbAl
TaNbHfZrTi alloy, J. Mater. Sci. 47 (2012) 4062–4074, [Link] refractory multi-principal-element alloys under multiple property constraints, Acta
s10853-012-6260-2. Mater. 248 (2023) 118784, [Link]
[5] S. Sheikh, S. Shafeie, Q. Hu, J. Ahlström, Alloy design for intrinsically ductile [23] Y. Tian, R.H. Yuan, D.Z. Xue, Y. Zhou, T. Lookman, Role of uncertainty estimation
refractory high-entropy alloys, J. Appl. Phys. 120 (2016) 164902, [Link] in accelerating materials development via active learning, J. Appl. Phys. 128
10.1063/1.4966659. (2020) 014103, [Link]
[6] E. Mak, B. Yin, W.A. Curtin, A ductility criterion for bcc high entropy alloys, [24] C. Wen, Y. Zhang, C. Wang, D. Xue, Y. Bai, S. Antonov, L Dai, T. Lookman, Y. Su,
J. Mech. Phys. Solids 152 (2021) 104389, [Link] Machine learning assisted design of high entropy alloys with desired property, Acta
jmps.2021.104389. Mater. 170 (2019) 109–117, [Link]
[7] O.N. Senkov, D.B. Miracle, S.I. Rao, Correlations to improve room temperature [25] D. Xue, P.V. Balachandran, J. Hogden, J. Theiler, D. Xue, T. Lookman, Accelerated
ductility of refractory complex concentrated alloys, Mater. Sci. Eng., A 820 (2021) search for materials with targeted properties by adaptive design, Nat. Commun. 7
141512, [Link] (2016) 11241, [Link]
[8] C.C. Juan, K.K. Tseng, W.L. Hsu, M.H. Tsai, C.W. Tsai, C.M. Lin, S.K. Chen, S.J. Lin, [26] G. Tzortzis, A. Likas, The MinMax k-Means clustering algorithm, Pattern Recogn.
J.W. Yeh, Solution strengthening of ductile refractory HfMoxNbTaTiZr high- 47 (2014) 2505–2516, [Link]
entropy alloys, Mater. Lett. 175 (2016) 284–287, [Link] [27] O.A. Waseem, H.J. Ryu, Combinatorial development of the low-density high-
matlet.2016.03.133. entropy alloy Al10Cr20Mo20Nb20Ti20Zr10 having gigapascal strength at 1000 ℃,
[9] P. Singh, A. Sharma, A.V. Smirnov, M.S. Diallo, P.K. Ray, G. Balasubramanian, D. J. Alloys Compd. 845 (2020) 155700 [Link]
D. Johnson, Design of high-strength refractory complex solid-solution alloys, npj jallcom.2020.155700.
Comput. Mater. 4 (2018) 16, [Link] [28] C. Wen, C.X. Wang, Y. Zhang, S. Antonov, D. Xue, T. Lookman, Y. Su, Modeling
[10] K.K. Tseng, C.C. Juan, S. Tso, H.C. Chen, C.W Tsai, J.W. Yeh, Effects of Mo, Nb, Ta, solid solution strengthening in high entropy alloys using machine learning, Acta
Ti, and Zr on Mechanical Properties of Equiatomic Hf-Mo-Nb-Ta-Ti-Zr Alloys, Mater. 212 (2021) 116917, [Link]
Entropy 21 (2018) 15, [Link] [29] Z.D. Han, N. Chen, S.F. Zhao, L.W. Fan, G.N. Yang, Y. Shao, K.F. Yao, Effect of Ti
[11] Q. Hua, S. Guo, J.L. Guo, F.F. Luo, J.W. Wang, Effect of Mo on high-temperature additions on mechanical properties of NbMoTaW and VNbMoTaW refractory high
strength of refractory complex concentrated alloys: A perspective of entropy alloys, Intermetallics 84 (2017) 153–157, [Link]
electronegativity difference, J. Alloys Compd. 906 (2022) 164186, [Link] intermet.2017.01.007.
10.1016/[Link].2022.164186. [30] O.N. Senkov, S. Gorsse, D.B. Miracle, High temperature strength of refractory
[12] Y. Yuan, Y. Wu, Z. Yang, X. Liang, Z. Lei, H. Huang, H. Wang, X. Liu, K. An, W. Wu, complex concentrated alloys, Acta Mater. 175 (2019) 394–405, [Link]
Z. Lu, Formation, structure and properties of biocompatible TiZrHfNbTa high- 10.1016/[Link].2019.06.032.
entropy alloys, Mater. Res. Lett. 7 (2019) 225–231, [Link]
21663831.2019.1584592.