Tuning magnetocrystalline anisotropy of Fe3 Sn by alloying

Olga Yu. Vekilova∗

Department of Physics and Astronomy,
Uppsala University, Box 516, 75121 Uppsala, Sweden
arXiv:1803.08292v1 [cond-mat.mtrl-sci] 22 Mar 2018

Bahar Fayyazi, Konstantin P. Skokov, and Oliver Gutfleisch

Materials Science, TU Darmstadt, Alarich-Weiss-Str. 16, 64287 Darmstadt, Germany

Cristina Echevarria-Bonet and José Manuel Barandiarán

BCMaterials, UPV/EHU Science Park, 48940 Leioa, Spain

Alexander Kovacs, Johann Fischbacher, and Thomas Schrefl

Department for Integrated Sensor Systems, Danube University Krems,
Viktor Kaplan Str. 2/E, 2700 Wiener Neustadt, Austria

Olle Eriksson
Department of Physics and Astronomy,
Uppsala University, Box 516, 75121 Uppsala, Sweden and
School of Science and Technology, Örebro University, SE-701 82 Örebro, Sweden

Heike C. Herper
Department of Physics and Astronomy,
Uppsala University, Box 516, 75121 Uppsala, Sweden

(Dated: March 23, 2018)

1
 Abstract
The electronic structure, magnetic properties and phase formation of hexagonal ferromagnetic
Fe3 Sn-based alloys have been studied from first principles and by experiment. The pristine Fe3 Sn
compound is known to fulfill all the requirements for a good permanent magnet, except for the
magnetocrystalline anisotropy energy (MAE). The latter is large, but planar, i.e. the easy mag-
netization axis is not along the hexagonal c direction, whereas a good permanent magnet requires
the MAE to be uniaxial. Here we consider Fe3 Sn0.75 M0.25 , where M= Si, P, Ga, Ge, As, Se, In,
Sb, Te and Bi, and show how different dopants on the Sn sublattice affect the MAE and can alter
it from planar to uniaxial. The stability of the doped Fe3 Sn phases is elucidated theoretically
via the calculations of their formation enthalpies. A micromagnetic model is developed in order
to estimate the energy density product (BH)max and coercive field µ0 Hc of a potential magnet
made of Fe3 Sn0.75 Sb0.25 , the most promising candidate from theoretical studies. The phase sta-
bility and magnetic properties of the Fe3 Sn compound doped with Sb and Mn has been checked
experimentally on the samples synthesised using the reactive crucible melting technique as well as
by solid state reaction. The Fe3 Sn-Sb compound is found to be stable when alloyed with Mn. It is
shown that even small structural changes, such as a change of the c/a ratio or volume, that can be
induced by, e.g., alloying with Mn, can influence anisotropy and reverse it from planar to uniaxial
and back.

PACS numbers: 75.50.Ww, 75.30.Gw, 75.20.En

∗
olga.vekilova@physics.uu.se

2
I. INTRODUCTION

Strong permanent magnets creating high magnetic field are of ultimate importance for
many technological applications, from magnetic resonance imaging to magnetic hard disk
drives in information storage. They are also used in a number of green energy applications,
like motors for hybrid and electric cars and direct-drive wind turbines [1]. The strongest
known permanent magnets typically contain rare earth elements [2]. For the past few decades
the demand for such magnets has substantially increased. As the cost of the rare-earth based
materials is high, the search for magnets, that are cheaper and contain smaller amounts of
rare earth elements, has become an important field of research [1, 3, 4].

Ferromagnets with rather high Curie temperature (TC ) above 400 K and high saturation
magnetization as well as high magnetocrystalline anisotropy energy (MAE) are considered
as good candidates for permanent magnet applications. Furthermore, for such materials
the axis corresponding to the longest lattice constant, i.e. the c axis in most hexagonal
structures, should be the unique easy magnetization direction [1, 3, 4]. These properties can
be found in particular in Fe-rich materials with non-cubic uniaxial crystal structures. The
hexagonal Fe3 Sn compound satisfies these conditions to a large degree. Among five existing
intermetallic compounds containing Fe and Sn the Fe3 Sn phase is one of the most attractive
ones, due to the highest concentration of iron and therefore highest magnetic moment. The
other advantages of the rare-earth free Fe3 Sn system are its relatively low price and rather
high TC of about 743 K [5]. However, as it has recently been shown both experimentally
and theoretically, the magnetocrystalline anisotropy of Fe3 Sn is planar, which is undesirable
for a permanent magnet [6]. It has also been suggested by Sales et. al that alloying with
Sb might change the anisotropy to uniaxial [6]. Motivated by this research we studied
the influence of different dopants on the MAE of the Fe3 Sn compound. The electronic
structure and magnetic properties, as well as the effect of the hexagonal c/a ratio on the
magnetocrystalline anisotropy in the Fe3 Sn compound and its alloys were addressed from
first principles by means of highly accurate full-potential linear muffin-tin orbital (LMTO)
method implemented in the RSPt code. The stability of the doped phases were elucidated
by the calculation of formation enthalpies by means of the VASP code.

Theoretical calculations were combined with an experimental study in order to verify

the phase stability as well as the magnetic properties of Fe3 Sn compounds. Combinatorial

3
materials science and high-throughput screening methods, such as Reactive Crucible Melting
(RCM) technique, offer an efficient strategy for the discovery of new materials with promising
properties. Using the knowledge of the required synthesis conditions for the formation of
the Fe3 Sn phase [7], we performed the RCM method in order to search for the (Fe)3 (SnM)
phase. To get more insight into the magnetic properties of the (Fe)3 (SnM) phase, additional
experiments were performed using the Solid State Reaction (SSR) method, which was used
for the preparation of the desirable compound from the mixture of the starting elements
by means of atomic diffusion. For the most promising system a micromagnetic model was
developed to calculate the magnetic induction as a function of the internal field.
The paper is organized as follows. Section II A describes theoretical methods used for the
first-principles calculations. Section II B provides the experimental details. In sections III A,
III B, III C and III D the results of theoretical simulations as well as experimental results are
given. Section III E addresses the results of the micromagnetic simulations. Conclusions are
given in section IV.

II. METHODS

A. Theory

The high temperature phase of Fe3 Sn has a hexagonal crystal structure with space group
P63 /mmc (#194) and contains 8 atoms per unit cell (see Fig. 1). For calculations of
phase stability Vienna Ab Initio Simulation Package (VASP) [8–10] was used within the
projector augmented wave (PAW) method [11]. The electronic exchange and correlation
effects were treated by the generalized gradient approximation (GGA) in the Perdew, Burke,
and Ernzerhof (PBE) form [12] in all the used methods. A 64 atom supercell of Fe3 Sn
comprising of 46 Fe and 16 Sn atoms was considered. The plane-wave energy cut-off was set
to 350 eV. The converged k-point mesh was found to be 8 × 8 × 8 k-points. The obtained
magnetic moment on iron was ∼ 2.4µB per atom, in agreement with previous studies [6].
The Sn atoms of the Fe3 Sn compound were partially replaced by alloying elements. In or-
der to examine the phase stability of considered doped systems for each impurity, i.e. M=Sb,
Ga, Ge, and Hf, we calculated 3 different distributions of impurity atoms, namely ordered,
random (i.e. mimicking a disordered alloy), and phase separated (i.e. mimicking clusteriza-

4
FIG. 1. (Color online) 1 × 1 × 2 Fe3 Sn hexagonal cell with one impurity atom on the tin sublattice.
Iron atoms are shown with brown spheres, Sn atoms with grey and M (M=Si, P, Ga, Ge, As, In,
Sb, Te and Bi) impurity atom is shown with the green sphere.

tion of dopants) (See Supplemental Materials). We used the special quasi-random structure
(SQS) technique [13–15] to generate the corresponding supercells. For the estimation of
phase stability of ternary compounds the following equation was used:

∆H = HFe3 Snx M1−x − xHFe3 Sn − (1 − x)HFe3 M (1)

where HFe3 Snx M1−x is the enthalpy of a ternary compound and 1 − x is the concentration of
M, an impurity element on the Sn sublattice.
The 1 × 1 × 2 supercell of Fe3 Sn comprising of 12 iron and 4 tin atoms, one of which
is further substituted by an impurity (see Fig. 1), was used for the calculation of the
magnetic properties with help of the full-potential linear muffin-tin orbital (FP-LMTO)
method implemented in the RSPt code [16, 17]. We performed integration over the Brillouin
zone, using the tetrahedron method, with Blöchl’s correction [18]. The k-point convergence
of the MAE for the chosen supercell size was found when increasing the Monkhorst-Pack
mesh [19] to 24 × 24 × 24, that was further used in all calculations.
The following impurities were considered in Fe3 Sn0.75 M0.25 compound: M=Si, P, Ga, Ge,
As, Se, In, Sb, Te, and Bi. With one substitutional impurity atom in the considered supercell,
the dopant concentration is fixed to 6.25 at.%. The equilibrium lattice parameter and
hexagonal c/a ratio were calculated for every structure using the VASP code. The effective
exchange interaction parameters (Jij ) were obtained using Lichtenstein et al. method [20,
21], as implemented in RSPt [22]. In this technique the energy of the system is mapped

5
FIG. 2. (Color online) Enthalpy of formation (∆H) of the ternary Fe3 Sn0.75 M0.25 compound as
a function of the concentration of dopant M, calculated in comparison with the Fe3 Sn binary on
the left-hand side and Fe3 M (M=Ga, Ge and Hf) binary compound or a mixture of Fe and the M
element on the right-hand side of the dashed vertical line.

onto a classical Heisenberg model with the following Hamiltonian:

1X
Ĥ = − Jij ~ei · ~ej , (2)
2 i6=j

where ~ei denotes the unit vector along the magnetic moment at the site i. The exchange
parameter between sites i and j is defined in the following way:
" #
T X ˆ i (iωn )Ĝ↑ (iωn )∆
ˆ j (iωn )Ĝ↓ (iωn ) ,
Jij = Tr ∆ ij ij (3)
4 n

where T is the temperature, ∆ is the on-site exchange potential, Gij is an inter-site Greens
function and iωn is the n-th fermonic Matsubara frequency [22].

B. Experiment

Screening through a large variety of material compositions and possible temperature

stability ranges using traditional equilibrium alloy approach, which is a one-alloy-at-a-time
practice, is a laborious task. Combinatorial materials science and high-throughput screening
methods enable the synthesis of large number of samples at once and therefore speed up
the discovery of the new materials. Furthermore, using high-throughput characterization

6
methods, their properties can be efficiently determined. Several phase diagrams of material
systems have been constructed using different combinatorial approaches [23, 24], e.g. thin
film deposition [25] and bulk high-throughput techniques, such as reactive diffusion method
[26], and reactive crucible melting [7, 27–31].
The RCM method is firstly introduced in Ref. [30] as a tool to search for hard magnetic
phases. The method is based on diffusion processes driven by the formation of concentration
gradient between the crucible material and other elements which are filled into it. Production
procedure and working principle of the RCM method is fully described in Ref. [7], where the
method is applied to the Fe-Sn binary system and all five known intermetallic compounds of
the system were synthesized in the reactive crucibles. The Fe3 Sn (3:1) phase was obtained in
RCM by quenching of the crucibles after annealing at high temperatures (1023 K-1098 K).
To explore the hexagonal Fe3 Sn phase in the quest for uniaxial anisotropy, Fe-Snx M1−x
crucibles with M = Sb, Si, Ga, Ge, Pb, In, Bi and 0.5 < x < 0.75 were synthesized. To
further extend our search, quaternary Fe-Mn, Sn, Sb crucibles were additionally produced.
The crucibles were made of 99.95 % pure Fe and they were filled with about 1 g of the rest
elements of the examined system. The filling elements with the purity > 99.9 were added
in the form of crushed pieces. The samples were annealed at three selected temperatures of
1013 K, 1043 K, and 1073 K for one week and subsequently quenched. For details we refer
to [7].
For high-throughput characterization, the microstructure of the formed phases was stud-
ied by Philips XL30 FEG scanning electron microscopy (SEM) in back-scattered electron
(BSE) contrast mode and their chemical compositions were determined using energy disper-
sive x-ray (EDX) spectroscopy. In addition, a Zeiss Axio Imager.D2m magneto-optical Kerr
effect (MOKE) microscopy was used to display the magnetic domain structure of the formed
phases which may give a clear hint for identification of the phases with uniaxial anisotropy.
Several Fey Mn3−y Snx Sb1−x samples (y=3 and x=1; y=2.5, 2.25, 2, 1.5 and x=0.75; and
y=1.5 and x = 0.9) were prepared by solid state reaction (SSR) [32] through the following
procedure. First, stoichiometric amounts of powders of the starting elements (99.9+% purity,
particles of less than 50 microns in size) were handmilled with an agate mortar and pestle
and then compacted into pellets using pressures up to 0.5 GPa at room temperature (RT).
Pellets were then encapsulated in a quartz ampoule, heated to 1073 K for 48 hours in vacuum
and then quenched into ice water.

7
 This process was repeated twice in order to homogenize the composition in the sample.
X-Ray Diffraction (XRD) measurements were performed on a Philips X’Pert Pro diffrac-
tometer, in the Bragg-Brentano geometry, using Cu Kα radiation (λ = 1.5418 Å). The
samples were placed on a spinner to avoid a possible preferential crystalline orientation.
XRD patterns were analyzed by Rietveld refinements, through the FullProf Suite [33], us-
ing a Thompson-Cox-Hastings pseudo-Voigt function to describe the profile of the peaks.
Temperature dependent magnetization was measured up to 823 K, using H=0.01 T, in a
vibrating sample magnetometer EZ7-VSM from Microsense.

III. RESULTS

A. Phase stability

The enthalpy of formation was calculated for ternary Fe3 Sn0.75 M0.25 alloys, where M=Sb,
Ga, Ge and Hf for different concentrations of dopants from 6.25, 12.5, 18.75 and up to 25 at.
% (see Fig. 2). The energies of these systems were examined and compared with the energies
of the binary Fe3 Sn compound on the left-hand side and binary Fe3 M (or the mixture of pure
Fe and M if the binary phase was not energetically favorable) on the right-hand side of Fig.
2 (see also Eq. 1). For the sake of simplicity, whether this particular phase was compared
with the binary phase or with a mixture of pure elements, is specified on the right hand side
of Fig. 2 for each considered structure. A positive slope corresponds to an unstable ternary
compound, while the negative one corresponds to a stable structure (See Fig.2).
Investigating the phase diagrams of the Fe-Ga and Fe-Ge binaries [34, 35] one can see
that in the region up to 10-20 at. % of dopant elements at temperatures up to 1500 K
the Fe3 M phase decomposes into a mixture of α-Fe and pure Ga or Ge phases [34, 35]. At
higher concentration of dopants the phase is, however, stable. For this reason, enthalpy of
Fe3 Sn0.75 M0.25 (M=Ga, Ge) was calculated relative to both the mixture of pure elements as
well as the Fe3 M (M=Ga, Ge) phase, see Fig. 2. The dashed vertical lines show the area
around 15 at. % of dopant concentration where the mixture of these phases might exist
depending on temperature and the dopant. At a concentration of dopants lower than ∼ 15
at. %, when compared to the mixture of pure elements on the right-hand side, the enthalpy
is negative indicating stability of the structure (see lower red and blue lines in Fig. 2).

8
However, when compared with the binary phase Fe3 M, the stability of the ternary structure
should be estimated from the upper positive curves for both Ga and Ge dopants (see upper
red and blue lines in Fig. 2).
In the considered Fe3 Sn0.75 M0.25 alloy the concentration of dopants is equal to 6.25 at %,
which means the structures could be considered as stable based on this theoretical estima-
tion. In contrast, checking the phase diagram of Fe-Sb one can see that the binary Fe3 Sb
phase cannot exist. Furthermore, when the formation enthalpy of Fe3 Sn0.75 Sb0.25 phase was
calculated in comparison with the mixture of pure elements on the right hand side (see the
green curve in Fig. 2) it appeared that the phase is unstable. However, as it is shown in
section III B, this phase has shown promising magnetic properties, so in order to stabilize it
experimentally Mn was added to the Fe sublattice, see sections III C and III D.
Similarly, for the Hf dopant even comparing with the mixture of elements (see the black
line in Fig. 2), the enthalpy curvature is positive indicating thermodynamic instability of
this phase. For that reason Hf was further excluded from consideration and calculation of
the magnetic properties.

B. Magnetocrystalline anisotropy

We calculated the saturation magnetic moment of Fe3 Sn0.75 M0.25 system. For the pristine
Fe3 Sn it is equal to µ0 Ms = 1.49 T, or Ms =1.19 MA/m, in very good agreement with
previous theoretical and experimental estimations [6]. Due to the low concentration of the
impurities, 6.25 at. %, in the doped system, the obtained value of saturation magnetization
is close to the one of the undoped system. For instance, for the case of Sb impurity it is
equal to 1.51 T, or 1.2 MA/m.
Magnetocrystalline anisotropy of the Fe3 Sn0.75 M0.25 system, where M=Si, P, Ga, Ge,
As, Se, In, Sb, Te and Bi was calculated from first principles electronic structure theory.
The resulting MAE values are shown in Fig. 3 for the dopants grouped according to their
positions in the Periodic Table. For the pristine Fe3 Sn the resulting anisotropy was found to
be equal to -1.5 MJ/m3 (see Fig. 3). In magnitude, this is one of the largest values among all
the considered structures, however, the minus sign indicates the easy plane of magnetization.
These results are in good agreement with other theoretical and experimental estimations (-
1.59 MJ/m3 and -1.8 MJ/m3 respectively [6]).

9
FIG. 3. (Color online) Calculated values of magnetocrystalline anisotropy energy (MAE), K1 ,
and easy magnetization direction for different impurities for Fe3 Sn alloy grouped according to the
positions of dopants in the Periodic Table. A negative value of K1 indicates the in-plane easy
magnetization axis, while the positive one shown with the filled blue rectangulars corresponds to
the desirable uniaxial magnetocrystalline anisotropy.

As one can see in Fig. 3, the largest magnitude of the MAE (albeit favoring in-plane
easy axis), of all the considered structures with dopants, was obtained for the Fe3 Sn0.75 In0.25
compound. An uniaxial anisotropy was found for M=As, Sb and Te. However, the MAE
for the Te doping is very small. The anisotropy of Fe3 Sn0.75 Sb0.25 is in agreement with the
existing theoretical data of 0.5 MJ/m3 [6]. As it was mentioned in Sec. III A, in order to
stabilize the Fe3 Sn0.75 Sb0.25 system Mn was added on the Fe subblatice. The anisotropy of
Fe1.5 Mn1.5 Sn0.75 Sb0.25 system is equal to -1.49 MJ/m3 , which is very close to the value for the
undoped Fe3 Sn system. Adding Sb to the pristine Fe3 Sn allows for the change of anisotropy
from planar to uniaxial, while adding Mn for stabilization reverts it back to planar.

The obtained MAE data were plotted as a function of the number of valence electrons per
formula unit in the Fe3 Sn0.75 M0.25 system for doping atoms from three rows of the Periodic
Table, see Fig. 4. Positive numbers correspond to a uniaxial MAE. As one can see, a
similar tendency is shown for the anisotropy in all considered rows. Starting from the lowest
values, corresponding to high but planar anisotropy, for the dopants from the Group IIIA
(Ga and In) the MAE moves towards positive values, with an increase of the number of
valence electrons in the system. The maximal value of MAE, which is positive for As and Sb
dopants, is obtained for Group VA elements. This tendency is shown for all the considered

10
FIG. 4. (Color online) The magnetocrystalline anisotropy energy (MAE) in Fe3 Sn0.75 M0.25 com-
pound as function of the number of valence electrons, N, of the system for three rows: Si-P (shown
with the green line), Ga-Ge-As (shown with the red line) and In-Sn-Sb-Te (shown with the blue
line). Doping from Group IIIA, Group IVA, Group VA and Group VIA, correspond to N=27.75,
28.25 and 28.5, respectively.The area of interest, where the easy magnetization axis is uniaxial,
is shown with the gradient background. The horizontal dashed line shows the magnitude of the
Fe3 Sn0.75 In0.25 anisotropy, as its number of valence electrons is substantially lower.

rows of elements, notifying that Group VA dopants seem to be the most promising for the
consideration, specifically, As and Sb. Two other elements of this group, P and Bi, do not
flip anisotropy to uniaxial (see Fig. 3). Anisotropy of Fe1.5 Mn1.5 Sn0.75 Sb0.25 is shown with
the horizontal dashed line. The number of valence electrons in the unit cell of this system
is 26.75, which is out of the scale of Fig. 3. The above described tendency together with
the result of the Mn addition indicates that alloying with other elements around Sn in the
Periodic Table, not considered in this search, can hardly bring new candidates able to flip
the easy magnetization axis to uniaxial. However, more studies can be of interest.

The dependence of the magnetic properties on the number of valence electrons in dopants
was further illustrated by the calculation of the Heisenberg exchange parameters Jij s between
the atoms (see Fig. 5). The Fe3 Sn system, as well as the Fe3 Sn0.75 M0.25 alloys, where M=Sb,
As and In, were shown to be strongly ferromagnetic as indicated by large and positive J0
(defined as a sum of all Jij ’s). In Fig. 5 the systems with planar anisotropy, the pristine
Fe3 Sn (a) and Fe3 Sn0.75 In0.25 (d), which has the largest planar anisotropy, are shown in

11
FIG. 5. (Color online) Inter-site exchange parameters Jij between the iron atoms i and j separated
by the distance Rij of pristine Fe3 Sn compound (a), as well as of Fe3 Sn0.75 M0.25 (M=Sb (b), As (c)
and In (d)). Systems with planar MAE are shown with the blue lines, while systems with uniaxial
MAE are shown with red lines. a stands for the lattice constant.

comparison with the systems with significant uniaxial MAE, namely, Fe3 Sn0.75 Sb0.25 (b) and
Fe3 Sn0.75 As0.25 (c).

Notice that the value of the nearest neighbor interaction, corresponding to the shortest
inter-atomic distance, differs slightly among the considered systems. The largest nearest-
neighbor first exchange interaction value was obtained for the Fe3 Sn0.75 Sb0.25 system (∼
1.98 mRy) with the highest positive value of MAE. A plot of the nearest neighbor exchange

12
FIG. 6. (Color online) Dependence of the nearest neighbor exchange parameter on the magne-
tocrystalline anisotropy of Fe3 Sn and Fe3 Sn0.75 M0.25 (M=Sb, As and In). The area below zero
corresponds to the planar anisotropy while the area above zero represents the uniaxial anisotropy.

interaction as function of the MAE is shown in Fig. 6. The second highest nearest-neighbor
exchange was obtained in Fe3 Sn0.75 As0.25 (∼ 1.88 mRy). The JN N value for the system with
largest negative anisotropy, Fe3 Sn0.75 In0.25 is the smallest of all (∼1.57 mRy). The JN N
value for pristine binary system Fe3 Sn has a value of ∼ 1.75 mRy, which is higher than that
of the system with In, but lower than the one out the systems with uniaxial anisotropy.
This clearly indicates that in the presently investigated alloys, there is a certain correlation
between the strength of the nearest neighbor couplings Jij s and MAE of the system.

Taking into account that atomic relaxations substantially influence the MAE, we also
studied the effect of the hexagonal lattice ratio, c/a, on the value of MAE. Fig. 7 shows the
change of the MAE when the c/a ratio increases from 1.4 to 1.9 for Fe3 Sn0.75 M0.25 (M= Sn,
Ge, As and Sb) systems at a fixed equilibrium volume. The averaged equilibrium c/a ratio
(∼1.58) is shown with the vertical dashed line. As the dopant concentration is low (6.25 at
%) the deviations from 1.58 of the c/a ratio for different dopants are rather minor.

For all considered dopants the MAE behavior with the c/a increase is basically linear,
illustrating that when the crystal is stretched, the easy axis can be switched from planar
to uniaxial for all the structures. For instance, for Ge-doped Fe3 Sn a change of the c/a
ratio at the fixed volume leads to the change of the anisotropy value from rather high with

13
FIG. 7. (Color online) Dependence of the magnetocrystalline anisotropy on the hexagonal c/a ratio
for M=Sn, Ge, As and Sb impurities. The averaged equilibrium c/a ratio about 1.58 is shown with
the dashed line.

planar easy magnetization direction, ∼ -2.7 MJ/m3 , to almost ∼0.7 MJ/m3 with uniaxial
easy magnetization direction. This strain induced change is largest for Ge doping, out of all
considered dopants. It is worth noting that changing MAE to uniaxial requires a variation
of c/a from ∼1.58 to 1.8, corresponding to 14 %. In the experiment it is hardly possible
to change c/a of such systems by more than few percents without structural changes, i.e.
without altering the hexagonal structure.

It is important to mention that for all the structures there is a peak of the MAE vs. c/a
curve (Fig. 7) in the region near the equilibrium c/a. In some cases, like for M=Sb and
As it can lead to the switching of the easy magnetization direction, however, this region
is rather narrow and a small change in the c/a ratio (like 0.5% from the equilibrium) can
switch the axis from planar to uniaxial and back. The highest peak is for Sb- doped Fe3 Sn;
it corresponds to the highest anisotropy value and uniaxial easy magnetization direction.

For Ge-doped structure, as well as for the binary Fe3 Sn slight modifications of the c/a
near the equilibrium are not resulting in the MAE peak sufficiently high to switch the easy
magnetization direction. It is interesting to notice, that for these two structures the peak is
not exactly at the equilibrium c/a ratio but shifted, meaning that the small change of the
c/a ratio due to, for instance, alloying or heating, can slightly increase anisotropy for some
structures. Therefore we underline that the MAE of Fe3 Sn with any of the dopants turns
out to be very sensitive to the change of c/a.

14
 C. Reactive Crucible Melting technique

The effect of substituting Sn by its neighboring elements was experimentally studied by

high-throughput RCM method and the production of Fe-Snx M1−x crucibles, M= Sb, Si, Ga,
Ge, Pb, In, Bi and 0.5 < x < 0.75. The diffusion zone of the crucibles annealed at the
selected temperatures in the temperature interval from 1013 K to 1073 K were carefully
screened by the EDX analysis. None of the Fe3 (Sn,M) structures was stable and therefore
not formed in any of the crucibles (for details on the results see Supplemental Material).
Our theoretical calculations predict the change of anisotropy from planar to uniaxial
in Fe3 Sn system by partial substitution of Sn by Sb. However, the formation enthalpy
plot shows that Fe-Snx Sb1−x is unstable in the addressed concentration range (see Fig. 2).
Besides that, it was experimentally found that Sb destabilized the formation of 3:1 structure
and instead the Fe3 (Sn,Sb)2 phase was formed. This phase is not desirable for permanent
magnet applications due to the small concentration of Fe and therefore low magnetization
as well as its negligible MAE [6, 7]. For stabilization of Fe3 (Sn,Sb) phase, substitution of Fe
by Mn (with one less electron) was suggested to be effective to compensate an extra electron
of Sb in comparison to Sn. Accordingly, the Fe-Mn0.75 Sn0.5 Sb0.5 and Fe-Mn1.5 Sn0.75 Sb0.25
crucibles were synthesized and annealed in the temperature range of 1013 K to 1073 K.
It was found that Mn and Sb substitution preserved the hexagonal Ni3 Sn structure of the
parent Fe3 Sn phase and (Fe,Mn)3 (Sn,Sb) phase formed in the quaternary reactive crucibles.
Fig. 8 shows the microstructure forming in the diffusion zone of the Fe-Mn0.75 Sn0.75 Sb0.25
reactive crucible annealed at 1073 K for 5 days and subsequently quenched. A thin layer
of (Fe0.6 Mn0.4 )3 (Sn0.75 Sb0.25 ) phase is formed and bordered by Fe crucible. In addition,
(Fe0.5 Mn0.5 )3 (Sn0.75 Sb0.25 )2 phase is formed in large quantity on top of the 3:1 phase. The
phase stability of the Sb-Mn doped Fe3 Sn compound is confirmed by RCM. However, to
measure magnetic properties, such as Ms , TC , and MAE; a single-phase material has to be
synthesized. That was done by the solid state reaction (SSR).

D. Solid State Reaction

Several Fey Mn3−y Snx Sb1−x samples (y=3 and x=1; y= 2.25, 2, 1.5 and x=0.75; and
y=1.5 and x = 0.9) with different concentrations of Mn and Sb were prepared by solid state

15
FIG. 8. BSE image of the Fe- Mn0.75 Sn0.75 Sb0.25 crucible annealed at 1013 K. Narrow layer of
(Fe0.6 Mn0.4 )3 (Sn0.75 Sb0.25 ) is formed. However 3:2 phase is the dominant forming phase in the
crucible.

FIG. 9. (Color online) XRD patterns of the Fey Mn3−y Snx Sb1−x samples.

reaction. XRD patterns of the produced samples by two subsequent SSRs are shown in Fig.
9. For all the samples, the 3:1 phase was found though only for the parent Fe3 Sn alloy we
found a single phase. Rietveld refinements were performed on all the XRD patterns and the
results are shown in Table I.
The lattice parameters of the alloys (see Table I and Figure 10) follow an increasing trend
for higher values of the Mn concentration. The dependence of the lattice parameters on the
concentration of Sb was not studied as only two points were available.
Temperature dependent magnetization curves M(T) were measured for two selected sam-
ples with the highest amount of the 3:1 phase, namely, Fe3 Sn, and with equal concentration
of Fe and Mn, Fe1.5 Mn1.5 S0.75 Sb0.25 . The M(T) curves are shown in Fig. 11. The calculation

16
 Sample a(Å) c(Å)

Fe3 Sn 5.4621(5) 4.3490(6)

Fe2.25 Mn0.75 Sn0.75 Sb0.25 5.4858(5) 4.3721(6)
Fe2 Mn1 Sn0.75 Sb0.25 5.5000(4) 4.3829(6)
Fe1.5 Mn1.5 Sn0.75 Sb0.25 5.5338(1) 4.4270(2)
Fe1.5 Mn1.5 Sn0.9 Sb0.1 5.5545(3) 4.4453(4)

TABLE I. Lattice parameters of the Fey Mn3−y Snx Sb1−x alloys, obtained by Rietveld refinements
of the XRD patterns

FIG. 10. (Color online) Evolution of the lattice parameters and volume, obtained by Rietveld
refinements of the XRD patterns, of the Fey Mn3−y Snx Sb1−x alloys. These values are also shown
in Table 1.

of the Curie temperatures were performed by the derivative of the M(T). The obtained Curie
temperature of Fe3 Sn system, TC =748 K, is rather high and in very good agreement with
the existing experimental data (TC =743 K [5] and TC =725 K [6]), while that of the diluted
sample, 393 K, is much lower than that of the parent composition Fe3 Sn. Similar values
were found previously in literature for Fe1.5 Mn1.5 Sn0.9 Sb0.1 and Fe1.5 Mn1.5 Sn0.85 Sb0.15 , with
TC = 405 K [6].

In order to examine how the addition of Mn to the Fe3 (SnSb) compound affects the

17
FIG. 11. (Color online) M(T) curves for selected compositions within the Fey Mn3−y Snx Sb1−x al-
loys. The abrupt drop corresponds to the Curie temperature, determined precisely by the derivative
of the curve.

anisotropy, the domain structure of the formed 3:1 phase was investigated. For the fer-
romagnets with uniaxial anisotropy observation of characteristic domain patterns, namely
stripe or branched domains, is expected. However, our Kerr analysis on the formed Mn/Sb
substituted 3:1 phase shows non-uniaxial domain structure. This observation agrees with
the negative theoretical value of anisotropy for the Fe1.5 Mn1.5 Sn0.75 Sb0.25 system.

E. Micromagnetic simulations

A micromagnetic model was developed in order to estimate the energy density product
(BH)max and coercive field µ0 Hc of a potential magnet made of Fe3 Sn0.75 Sb0.25 . Using the
software tool Neper [36] we created a synthetic microstructure based on Voronoi tessellation
as shown in Figure 12.
The granular grain structure consisted of 27 grains with an average grain diameter of
50 nm. Each grains spontaneous magnetization, anisotropy energy density were taken from
ab-initio calculations (µ0 Ms,grain = 1.52 T, Ku,grain = 0.33 MJ/m3 ). The exchange stiffness
constant was assumed to be Aex,grain = 10 pJ/m. The grains easy axes were randomized
within a cone angle of 5 with respect to positive z-axis.
Between the grains we assumed a 4 nm thick iron-rich ferromagnetic grain boundary phase

18
FIG. 12. (Color online) Synthetic microstructure used for the simulation of the demagnetization
curve of a nanocrystalline Fe3 Sn0.75 Sb0.25 magnet.

(Ku,gb = 0). Its spontaneos magnetization was assumed to be Ms,gb =0.81 T. Accordingly,
the exchange stiffness constant was reduced to Aex,gb = 3.7 pJ/m.

The magnetization value of the grain boundary phase was obtained by numerical opti-
mization. We used the numerical optimization framework Dakota [37] and maximized the
energy density product using µ0 Ms,gb as a free parameter. The volume fraction of the grain
boundary phase was 10 percent. The model was discretized into a uniform mesh with an
edge length of 2 nm. A finite element energy minimization code [38] was used to compute
the static hysteresis properties of the proposed model.

In order to compute the demagnetization curve we calculated the equilibrium states for
a subsequently decreasing external field Hext . The field step µ0 ∆Hext = −1 mT. In order
to compute the B(H) loop and the expected energy density product we corrected the loop
which was obtained for a magnet with cubic shape with the macroscopic demagnetization
factor N = 1/3.

Figure 13 shows the computed B(H) curve and Fig. 14 illustrates the equilibrium mag-
netic states before and after the first switching event, respectively. The computed coercive
field is µ0 Hc = −0.49 T and the computed energy density product is 290 kJ/m3 . This
is about 3/4 of the maximum energy density product reported for commercially available
Nd-Fe-B magnets [2, 39, 40].

19
FIG. 13. (Color online) Computed B(H) curve (magnetic induction as function of the internal
field) for Fe3 Sn0.75 Sb0.25 . Remanence, energy density product, and coercive field are marked with
a rectangle, diamond, and circle respectively.

FIG. 14. (Color online) During magnetization reversal domain walls become pinned near the grain
boundaries.

IV. DISCUSSION AND CONCLUSIONS

Electronic structure and magnetic properties of the hexagonal Fe3 Sn compound doped
with 6.25 at % of Si, P, Ga, Ge, As, Se, In, Sb, Te, and Bi were studied theoretically from
first principles and experimentally. Our calculations show that at low concentrations of
some dopants, such as Ga and Ge, the considered phases are stable. In contrast to that, in
the case of the Sb dopant the phase was shown to be unstable against decomposition into
the mixture of pure elements. Our experimental study using RCM method supports this
theoretical prediction that the addition of Sb into Fe-Sn system destabilizes the formation
3:1, however further doping of Mn into the Fe sublattice stabilizes the structure. The SSR
technique confirmed stabilization of the 3:1 phases with different concentrations of Mn.

20
 Theoretical simulations predict that doping with As, Sb and Te can change the easy
magnetization direction to uniaxial. However, the change of the c/a ratio also substantially
influences the MAE. In the case of Sb and As the peak in the MAE vs c/a dependence leads
to the change of the easy magnetization axis from planar to uniaxial in the region close to
the equilibrium c/a. For the Sb and As dopants, the region where anisotropy is uniaxial,
is rather narrow and even a small change of c/a, for instance due to alloying with small
amounts of Mn, can lead to the switch of the easy magnetization direction.
As follows from our calculations, the change of anisotropy due to addition of Mn to
Fe3 Sn0.75 Sb0.25 system turns uniaxial anisotropy back to planar. This estimation indicates
that the predicted uniaxial anisotropy can hardly be observed experimentally. On one
side the presence of Mn stabilizes the Sb-doped alloy, but on the other side it basically
flips the anisotropy back to the planar value of an undoped system. The experimentally
stabilized (FeMn)3 SnSb phase shows non-uniaxial domain structure in nice agreement with
the theoretical prediction.
We studied all the dopants around Sn in the Periodic Table that can occupy the Sn
sublattice and can likely be mixed with Fe3 Sn. From the MAE data grouped in the way
as these dopants are placed in the Periodic Table of elements one can see the tendency to
increase the value of MAE from the III to the V group with the increase of the number
of valence electrons. However, in the VI group there is a slight decrease for Te and Se,
respectively. Thus the Group V seems to be the most promising out of all the considered
groups and the Sb addition gives the largest uniaxial anisotropy. Looking at the vertical
distribution in the Periodic Table, it appears that MAE increases from P to As and to
Sb, but then for Bi there is a clear drop. We conclude that the most preferable choice of
dopants should be the column with As and Sb, and we do not expect that other dopants
can substantially improve the desired properties of the Fe3 Sn compound.
Further, the micromagnetic simulations allowed us to estimate the magnetic induction of
the most promising system, Fe3 Sn0.75 Sb0.25 , as a function of the internal field. The computed
coercive field is equal to -0.49 T. The calculated density energy product, 290 kJ/m3 , is at
the level of best known to date magnets.
We investigated the Fe3 Sn0.75 M0.25 system by different theoretical approaches as well
as experimentally, in order to get a wider view on the magnetic properties of the system.
Certain dopants, like Sb, can turn MAE of the hexagonal Fe3 Sn uniaxial. Furthermore, other

21
magnetic properties of this system, such as saturation magnetization of 1.51 T and energy
density product compatible with the values of the best known magnets are impressively
high. However, such a turn of MAE is very sensitive to lattice deformations and the value
of the magnetocrystalline anisotropy is reduced compared to the one of the parent phase.
Further, the hexagonal phase becomes unstable with respect to decomposition. Addition of
Mn allowed us to stabilize the system with dopants experimentally, however, the anisotropy
turned back to planar. Therefore further search for better stabilizers or their combinations
(the so-called co-doping) might be considered in order to find the compromise between the
stability and uniaxial MAE.

ACKNOWLEDGMENTS

Authors acknowledge support from NOVAMAG project, under Grant Agreement No.
686056, EU Horizon 2020 Framework Programme. The computations were performed on
resources provided by the Swedish National Infrastructure for Computing (SNIC) at PDC
and NSC centers. O.E. acknowledges support from STandUPP, eSSENCE, the Swedish
Research Council and the KAW foundation (grants 2012.0031 and 2013.0020). O.G. ac-
knowledges the Hessen LOEWE Response programme. Authors are thankful to Yaroslav
Kvashnin for useful discussions.

[1] R. McCallum, L. Lewis, R. Skomski, M. Kramer, and I. Anderson, Annual Review of Materials
Research 44, 451 (2014), http://dx.doi.org/10.1146/annurev-matsci-070813-113457.
[2] O. Gutfleisch, M. A. Willard, E. Brck, C. H. Chen, S. G. Sankar, and J. P. Liu, Advanced
Materials 23, 821 (2011).
[3] M. D. Kuzmin, K. P. Skokov, H. Jian, I. Radulov, and O. Gutfleisch, Journal of Physics:
Condensed Matter 26, 064205 (2014).
[4] K. Skokov and O. Gutfleisch, Scripta Materialia (2018),
https://doi.org/10.1016/j.scriptamat.2018.01.032.
[5] G. Trumpy, E. Both, C. Djéga-Mariadassou, and P. Lecocq, Phys. Rev. B 2, 3477 (1970).

22
[6] B. C. Sales, B. Saparov, M. A. McGuire, D. J. Singh, and D. S. Parker, Scientific Reports 4,
7024 EP (2014).
[7] B. Fayyazi, K. P. Skokov, T. Faske, D. Y. Karpenkov, W. Donner, and O. Gutfleisch, Acta
Materialia 141, 434 (2017).
[8] G. Kresse and J. Hafner, Phys. Rev. B 48, 13115 (1993).
[9] G. Kresse and J. Furthmüller, Phys. Rev. B 54, 11169 (1996).
[10] G. Kresse and J. Furthmüller, Computational Materials Science 6, 15 (1996).
[11] P. E. Blöchl, Phys. Rev. B 50, 17953 (1994).
[12] J. P. Perdew, K. Burke, and M. Ernzerhof, Phys. Rev. Lett. 77, 3865 (1996).
[13] A. Zunger, S.-H. Wei, L. G. Ferreira, and J. E. Bernard, Phys. Rev. Lett. 65, 353 (1990).
[14] I. A. Abrikosov, S. I. Simak, B. Johansson, A. V. Ruban, and H. L. Skriver, Phys. Rev. B
56, 9319 (1997).
[15] A. V. Ruban, S. I. Simak, S. Shallcross, and H. L. Skriver, Phys. Rev. B 67, 214302 (2003).
[16] J. M. Wills and B. R. Cooper, Phys. Rev. B 36, 3809 (1987).
[17] J. M. Wills, M. Alouani, P. Andersson, A. Delin, O. Eriksson, and O. Grechnyev, Full-
Potential Electronic Structure Method, Springer series in solid state science, Vol. 167 (Springer,
Berlin, Germany, 2010).
[18] P. E. Blöchl, O. Jepsen, and O. K. Andersen, Phys. Rev. B 49, 16223 (1994).
[19] H. J. Monkhorst and J. D. Pack, Phys. Rev. B 13, 5188 (1976).
[20] M. I. Katsnelson and A. I. Lichtenstein, Phys. Rev. B 61, 8906 (2000).
[21] A. Liechtenstein, M. Katsnelson, V. Antropov, and V. Gubanov, Journal of Magnetism and
Magnetic Materials 67, 65 (1987).
[22] Y. O. Kvashnin, O. Grånäs, I. Di Marco, M. I. Katsnelson, A. I. Lichtenstein, and O. Eriksson,
Phys. Rev. B 91, 125133 (2015).
[23] K. Rajan, Annual Review of Materials Research 38, 299 (2008).
[24] W. Maier, K. Stöwe, and S. Sieg, Angewandte Chemie International Edition 46, 6016 (2007).
[25] K. Kennedy, T. Stefansky, G. Davy, V. F. Zackay, and E. R. Parker, Journal of Applied
Physics 36, 3808 (1965), http://dx.doi.org/10.1063/1.1713952.
[26] J.-C. Zhao, Journal of Materials Research 16, 15651578 (2001).
[27] D. Goll, R. Loeffler, J. Herbst, R. Karimi, U. Pflanz, R. Stein, and G. Schneider, IEEE
Transactions on Magnetics 50, 1 (2014).

23
[28] D. Goll, R. Loeffler, J. Herbst, C. Frey, S. Goeb, T. Grubesa, D. Hohs, A. Kopp, U. Pflanz,
R. Stein, and G. Schneider, Physica Status Solidi Rapid Research Letters 9, 603 (2015).
[29] S. Ener, J. Kroder, K. P. Skokov, and O. Gutfleisch, Journal of Alloys and Compounds 683,
198 (2016).
[30] A. Lüdtke, B. Stahl, I. R. Harris, and G. S. Schneider, Proceedings of the 16th International
Workshop on Rare-Earth Magnets and their Applications, Japan Institute of Metals (2000).
[31] F. Gross, Search for new permanent magnetic phases by reaction crucible analysis and devel-
opment of high throughput methods, PhD thesis (The University of Birmingham, 2004).
[32] E. A. M. Lesley E. Smart, Solid Sate Chemistry: An Introduction, 3rd ed. (Taylor and Francis,
Boca Raton London New York Singapore, 2005).
[33] J. Rodrguez-Carvajal, Physica B: Condensed Matter 192, 55 (1993).
[34] B. Predel, “Fe-Ga (iron-gallium),” in Dy-Er ... Fr-Mo, edited by O. Madelung (Springer Berlin
Heidelberg, Berlin, Heidelberg, 1995) pp. 1–4.
[35] B. Predel, “Fe-Ge (iron-germanium),” in Dy-Er ... Fr-Mo, edited by O. Madelung (Springer
Berlin Heidelberg, Berlin, Heidelberg, 1995) pp. 1–5.
[36] R. Quey, P. Dawson, and F. Barbe, Computer Methods in Applied Mechanics and Engineering
200, 1729 (2011).
[37] B. M. Adams, W. Bohnhoff, K. Dalbey, J. Eddy, M. Eldred, D. Gay, K. Haskell, P. D. Hough,
and L. Swiler, Sandia National Laboratories, Tech. Rep. SAND2010-2183 (2009).
[38] J. Fischbacher, A. Kovacs, H. Oezelt, T. Schrefl, L. Exl, J. Fidler, D. Suess, N. Sakuma,
M. Yano, A. Kato, et al., AIP Advances 7, 045310 (2017).
[39] S. Bance, F. Bittner, T. G. Woodcock, L. Schultz, and T. Schrefl, ArXiv e-prints (2017),
arXiv:1703.10710 [cond-mat.mtrl-sci].
[40] J. Coey, Scripta Materialia 67, 524 (2012), viewpoint Set No. 51: Magnetic Materials for
Energy.

24
 SUPPLEMENTAL MATERIAL FOR “TUNING MAGNETOCRYSTALLINE
ANISOTROPY OF Fe3 Sn BY ALLOYING”

Theoretical calculations

Figure S1 shows the calculated density of states for the Fe3 Sn and Fe3 Sn0.75 Sb0.25 com-
pounds, respectively. Figure S2 shows the three tested distributions of M impurity atoms
on the Sn sublattice in the Fe3 Sn0.75 M0.25 compound. For the sake of simplicity only the
results for 50 at. % of impurities is shown in all the cases. As one can see in the case of the
ordered distribution, (a), impurity atoms are placed in layers mixed with the layers of the
Sn atoms. In the case of the random alloy, (b) atoms are distributed in an SQS-like manner
with the short-range order close to 0 for a few first coordination shells. For the case of the
separated distribution of atoms, (c), the cell is split into two parts and one is fully occupied
with the M atoms, while the other one is occupied exclusively with the Sn atoms. According
to our estimation, for the low concentration of dopants, the ordered distribution is the most
energetically preferable.

FIG. S1. Density of states of the Fe3 Sn and Fe3 Sn0.75 Sb0.25 compounds without spin-orbit coupling
taken into account.

25
 RCM measurements

Using the RCM method, Fe-SnM crucibles with M= Sb, Ga, Ge, Si, In, Bi and Pb were
synthesized. The crucibles were annealed in the temperature range of 1013 K to 1073 K for 5
days and subsequently quenched. Theoretical calculations predict that doping of Sb changes
the planar anisotropy in the Fe3 Sn compound to uniaxial. Our experimental study using
RCM shows a destabilization of the 3:1 phase in Fe-Sn1−x Sbx , 0.15 < x < 0.5 crucibles. The
3:1 phase did not form in any of the synthesized samples, however, instead the 3:2 phase was
stabilized. Figure S3 shows the BSE image of the diffusion zone in Fe-Sn0.85 Sb0.15 reactive
crucible annealed at 1073 K. The only intermetallic compound formed in the crucible is the
3:2 phase.
In order to check whether substituting Ge and Ga in the Fe3 Sn compound preserves the
parent structure, Fe-SnGe(Ga) crucibles were synthesized. The diffusion zone of the crucibles
annealed at a specific temperature of 1013 K are shown in figure S4 for Fe-Sn0.5 Ga0.5 and
in figure S5 for Fe-Sn0.5 Ge0.5 . The observed microstructure in Fe-SnGa crucible contains a
solid solution Fe1−x Gax region with x up to 20 at. %. The remaining Ga has formed the
Fe62 Ga35 Sn3 phase (composition found by EDX) in the diffusion zone and the Sn content is
left mainly unreacted. The crucibles annealed at the higher temperatures (not shown here)
contained no intermetallic compounds and only solid solution Fe1−x Gax was found. In the
Ge containing crucible, Sn remained totally unreacted. On the Fe-rich side of the crucible, a
solid solution Fe1−x Gex region with x up to 17 atomic percent was formed. Moreover, distinct
layers of Fe3 Ge and Fe2 Ge intermetallic compounds were formed on top of the solid solution
region. A very similar microstructure was observed for the crucibles annealed at higher
temperatures up to 1073 K. Considering the above mentioned conditions, the stabilization
of 3:1 compound was not possible.
Further experiments have been performed in order to stabilize the Fe3 (SnM) compound
with M= Si, In, Bi, Pb and Sn:M varied from 2:1 to 1:1. Our RCM study shows that
no reaction has been occurred for the 3 latter substitutions after annealing at the selected
temperature range. Addition of Si to the Fe-Sn crucible mimics the microstructure forming
in the Fe-Sn binary crucibles [S1] whereas Si totally dissolves in Fe and does not diffuse to
any of the formed Fe-Sn binary intermetallic compounds.
Figures S7 and S6 show the diffusion zone of the Fe-Sn0.5 Si0.5 crucibles annealed at 1073 K

26
FIG. S2. Structure of the Fe3 Sn0.75 M0.25 system. Green spheres (impurity atoms) are distributed
on the Sn sublattice, shown with the grey spheres (host atoms) in (a) ordered, (b) random or (c)
separated ways.

and 1043 K respectively. In the crucible annealed at 1073 K, the high temperature phase
Fe5 Sn3 was formed whereas at lower temperature (1043 K), FeSn, Fe3 Sn2 and Fe3 Sn phases
were formed. Although the 3:1 structure is formed in the crucibles, however, according to
the EDX quantification, Si did not diffuse into 3:1 structure.

27
FIG. S3. BSE image of the Fe-Sn0.85 Sb0.25 crucible annealed at 1073 K. Substitution of Sb led to
formation of Fe3 (SnSb)2 and deformation of the desirable Fe3 Sn phase.

FIG. S4. BSE image of the diffusion zone in the Fe-Sn0.5 Ga0.5 crucible annealed at 1073 K. Sn
remained almost unreacted and the only ternary intermetallic compound Fe62 Ga35 Sn3 is formed.

28
FIG. S5. BSE image of the diffusion zone in the Fe-Sn0.5 Ge0.5 crucible annealed at 1073 K. Sn
remained almost unreacted and no ternary intermetallic compounds are formed.

FIG. S6. BSE image of the diffusion zone in the Fe-Sn0.5 Si0.5 crucibles annealed at 1073 K. The
binary Fe3 Sn phase is not formed in the crucible annealed at 1073 K.

29
FIG. S7. BSE image of the diffusion zone in the Fe-Sn0.5 Si0.5 crucibles annealed at 1043 K. The
binary Fe3 Sn phase is formed in the crucible annealed at 1043 K. However, Si totally dissolved in
Fe in a separate region and did not diffuse into the 3:1 structure.

30
[S1] B. Fayyazi, K.P. Skokov, T. Faske, D.Yu. Karpenkov, W. Donner, and O. Gutfleisch, Acta
Materialia 141, 434 (2017)

31