Page 125 - Read Online
P. 125
Chen et al. J Mater Inf 2023;3:10 https://dx.doi.org/10.20517/jmi.2023.06 Page 5 of 19
Figure 2. Schematic of a data-driven approach for the design of eutectic high entropy alloys.
where C denotes the “super element” (e.g., the mixture of elements that may form a solid solution) and C
I II
the single element in a pseudo-binary system. Following Equation (2), a number of EHEA compositions
were discovered with the aid of CALPHAD [64-67,95-102] , as tabulated in Table 3. Here, x is the composition yet
to be determined via experiments or CALPHAD.
With CALPHAD, one can obtain the so-called pseudo-binary phase diagram as a function of x and
temperature, through which fully-eutectic compositions can be identified, if any, for C eutectic . For example,
[109]
Wu et al. identified the near-eutectic composition Al Co Cr Ni for the pseudo-binary (CoCrNi) 1-x
19.4
39.4
20.6
20.6
(AlNi) alloy, which is very close to the eutectic composition of Al Co Cr Ni verified experimentally.
x [110] 17.4 21.7 21.7 39.2
In light of the Scheil solidification theory, Yurchenko et al. also successfully found the Al Cr Nb Ti Zr
28 20 15 27 10
EHEA. However, we note that all these above-mentioned methods are semi-empirical since CALPHAD is
also based on the available database. Regardless of the difference in these methods, a more general method
is always desirable, which can be applied to a wide range of compositions. In addition to the above-
mentioned eutectic high entropy alloys, people also developed a number of eutectic refractory high entropy
[108]
alloys [105-107] , eutectic soldering high entropy alloys and eutectic high entropy ceramics [87,88] . Interestingly,
some of these eutectics could also be designed based on the aforementioned empirical rules. Therefore, we
also list them in Tables 2 and 3 for the sake of completeness.
DATA-DRIVEN METHODS FOR THE DESIGN OF EUTECTIC HIGH ENTROPY ALLOYS
Database
[69-78]
In recent years, ML has been widely used to accelerate the search for advanced alloys . As a data-driven
approach, the performance of ML models is highly dependent on the quantity and quality of data [111,112] .
Figure 2 illustrates the workflow of a typical ML approach to designing EHEAs. While EHEAs are
important and very useful, we note that only a limited number of EHEA compositions are located out of
vast compositional space with the ML approach [Table 4] [79-81,113] . While one can easily find the data of
binary and ternary eutectics from their corresponding phase diagrams, the data for EHEAs mainly comes
from the literature, including those found through the empirical methods illustrated in Figure 1 and/or the
results of CALPHAD calculations. Here, we note that the CALPHAD calculations are performed with