
2
This invites a detailed benchmark study to validate different
approaches, however, so far this has only been performed for
the ultrathin film system of a Co monolayer on Pt(111) [54].
Such comparative studies are crucial to understanding the ori-
gin of skyrmion stability, particularly for the newly discovered
2D magnets.
Here, using ab initio calculations, we compare systemati-
cally three current state-of-the-art approaches to extract mag-
netic interaction parameters in Fe3GeTe2(FGT) based het-
erostructures, namely (i) by the Green’s function method [55],
(ii) by using the generalized Bloch theorem (gBT) [43, 54, 56,
57], or (iii) by using the supercell approach [49]. First we
study the shell-resolved exchange interaction between the Fe
atoms of the different layers in a freestanding FGT monolayer.
We find that the approaches (i) and (ii) are in good qualita-
tive agreement. We then focus on the structural and magnetic
properties of the 2D vdW heterostructure of an FGT mono-
layer deposited on germanene under strain, stacking, and elec-
tric field. We find that a small compressive strain (γ) of about
3% can significantly enhance the DMI in FGT heterostruc-
tures by more than 400% compared to the value without strain.
The variation of DMI is mainly due to the geometrical change
of the FGT monolayer. Such a large DMI is comparable to
that in state-of-the-art FM/HM heterostructures, which have
been demonstrated as prototypical multilayer systems to host
individual skyrmions even at room temperature. Furthermore,
the DMI can be substantially modified via different stacking
geometry due to the hybridization effect at the interface.
Upon applying an electric field the strength of the DMI
varies almost linearly and can even change sign when a strong
electric field (E>1 V/ ˚
A) is applied. The exchange constants
are also considerably modified due to an electric field while
the effect on the magnetocrystalline anisotropy energy (MAE)
is small. However, the MAE is dramatically reduced to 25%
of its original value at a compressive strain of γ=−3%. In
connection with the exchange frustration in FGT/germanene
these large changes of DMI and MAE open the possibility of
zero-field magnetic skyrmions [58]. For the DMI in FGT/Ge,
the three theoretical approaches are in good qualitative agree-
ment, and we also discuss quantitative comparison in detail.
The paper is organized as follows. In Sec. II, we describe
the three theoretical approaches used to obtain the relevant
spin-spin interaction parameters by mapping the ab initio DFT
calculations onto an extended Heisenberg model. In Sec. III,
we examine the Heisenberg exchange for free-standing FGT
followed by the DMI and MAE for FGT heterostuctures. Dif-
ferent theoretical approaches are carefully benchmarked. We
further investigate the effects of biaxial strain, stacking con-
figuration as well electric field on the magnetic interactions in
FGT heterostuctures. Finally, we summarize our main con-
clusions in Sec. IV.
II. Methods and computational details
In order to describe the magnetic properties of FGT het-
erostructures, we use the extended Heisenberg model for the
spins of Fe atoms in the hexagonal structure:
H=−X
ij
Jij (mi·mj)−X
ij
Dij ·(mi×mj)
−X
i
Ki(mz
i)2(1)
where miand mjare normalized magnetic moments at po-
sitions Riand Rirespectively. The three magnetic interaction
terms correspond to the Heisenberg isotropic exchange, the
DMI, and the MAE, respectively, and they are characterized
by the parameters Jij ,Dij , and Kiin the related terms. Note,
that by using Eq. (1) it is assumed that the magnetic moments
are constant.
During the past decade, in order to obtain the parame-
ters very accurately in Eq. (1), several approaches based on
density functional theory (DFT) have been developed, which
is frequently named ab initio atomistic spin model. In this
work, we apply three different approaches for the calcula-
tion of magnetic interactions in FGT vdW heterostructures:
(i) The Green’s function method [55, 59] (also known as the
Liechtenstein formula) employing infinitesimal rotations; (ii)
The generalized Bloch theorem (gBT) [56] which allows cal-
culating the total energy of spin-spirals of any wave vec-
tor qin magnetic nanostructures [57]; (iii) The supercell
approach [49] which is straightforward but computationally
heavy due to the comparison of total energies in a supercell
geometry. We performed DFT calculations using two com-
munity ab initio codes which differ in their choice of basis
set: The QuantumATK (QATK) package [60] uses an ex-
pansion of electronic states in a linear combination of atomic
orbitals (LCAO) while the FLEUR code [61] is based on
the full-potential linearized augmented plane wave (FLAPW)
formalism. The former is computationally very efficient,
while the latter ranks amongst the most accurate implemen-
tations of DFT. In the following, we denote the three differ-
ent approaches as LCAO-Green, FLAPW-gBT, and LCAO-
supercell for simplicity. Apart from the methods presented
above, there are also other approaches widely used in the
community for calculations of spin-spin interactions, e.g., the
four-state method [62] and the machine learning approach
[63].
II.1. The Green’s function method: LCAO-Green
Throughout this paper, vectors are denoted with bold char-
acters while matrices are represented by bold plus single un-
derline (e.g., Gij ). Moreover, Lrepresent orbital index while
σ= (σx, σy, σz)is spin index.
The variation of total energy due to the spin interactions in
Eq. 1, we obtain the following variation with respect to the
miand mj:
δEij =−2Jij (δmi·δmj)
−2δmiJani
ij δmj
−2Dij ·(δmi×δmj)
(2)
where the first term represents the isotropic exchange
(i.e., Heinsenberg), the second term is the the symmetric