WorksPhysics and Astronomy
Reconciling global and local spin models
Abstract
Magnetic interactions based on first-principles calculations can be casted either into global spin models, where the interactions are configuration-independent, or into local spin models, where the effective parameters depend explicitly on the underlying spin configuration. Although these two perspectives often yield seemingly inconsistent results, we demonstrate that they can be reconciled by restricting the analysis to transverse spin fluctuations. We develop a projection scheme that systematically removes spurious longitudinal contributions from local interaction matrices, thereby establishing a rigorous connection between the projected local interactions and the underlying global multispin Hamiltonian. Applying our approach to the prototype systems (i) a Cr trimer on Au(111) (ii) Fe and Cr dimers on Cu(001), we show that the projected local interaction parameters quantitatively reproduce those obtained from the global four-spin model. This work provides a unified framework to interpret local and global spin models on equal footing and resolves long-standing ambiguities in the identification of effective interactions such as the Dzyaloshinskii-Moriya interaction (DMI) in a non-collinear framework.
Introduction
Effective spin models provide a powerful framework for bridging first-principles electronic-structure calculations with the collective phenomena in magnetic materials. Since the early formulation of the Heisenberg Hamiltonian [1], such models have been indispensable in describing magnetic ordering, excitation spectra and finite temperature magnetism in solid magnets, for a comprehensive review see Ref. [2]. Evaluation of exchange interactions using Green function-based methods [3] has played a crucial role in the advancement of the study of bulk magnets and magnetic systems in reduced dimensions, including thin films and interfaces [4], as well as complex magnetic textures such as domain walls [5, 6] and magnetic skyrmions [7, 8, 9].
To capture the richness of competing interactions in these systems, extensions of the bilinear Heisenberg model have been proposed. In particular, multispin Hamiltonians including four-spin interactions [10, 11, 12] provide a configuration-independent description that naturally accounts for the stabilization of complex non-collinear states. Such global spin models have been successfully applied to a wide range of magnetic systems, from chiral magnets to atomic clusters at the nanoscale [7, 9, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20].
In the past decade, local spin models based on infinitesimal rotations [21, 22, 23] were developed, in which effective interactions explicitly depend on the underlying spin configuration. Although these models have provided useful insight into complex non-collinear states, they also led to surprising results, such as the emergence of Dzyaloshinskii-Moriya-like terms in the absence of spin-orbit coupling [24, 25]. This has sparked an active debate regarding the consistency between local and global formulations and the possibility of unifying them within a coherent theoretical framework [26, 27, 28, 29].
In this work, we address this issue by establishing a direct connection between global and local spin models. Within a non-relativistic multiple scattering framework, in Section II we outline a method that enables the evaluation of two-spin and four-spin interactions in a global spin model and then in Section III we describe the ab initio method to calculate local exchange interactions in terms of infinitesimal rotations, whereby we introduce two formalisms providing identical two-spin variation of the energy. Applying these schemes to the prototypical case of a Cr trimer on Au(111), in Section IV we show that the two alternative formalisms yield remarkably different local magnetic interactions. In Section V, we present local interactions based on the previously introduced two-spin and four-spin global interactions and demonstrate that the local interactions are not uniquely defined due to longitudinal terms in the two-spin variation of the energy. In Section VI we use a projection method that systematically removes spurious longitudinal contributions from local interaction matrices and give evidence that the projected local interactions are well-defined, i.e., do not depend on the local formalism introduced before. Moreover, the projected local interactions are reproduced by the global four-spin Hamiltonian with high accuracy. In Section VII we extend the study for Fe and Cr dimers on Cu(001), then in § 0.8 we summarize by concluding that local magnetic interactions straightforwardly descend from global multi-spin interactions, thereby giving no additional insight to the physics of magnetic systems.
Isotropic global spin model from first principles
The most general isotropic spin Hamiltonian containing two-spin and four-spin interactions can be written as
where the parameters and can be evaluated using multiple scattering theory based on the expansion of the grand potential in the spin variables up to the fourth order. The grand potential is given by Lloyd’s formula [30, 31] [4],
where is the grand potential of the free electrons, is the Fermi energy, and boldface quantities denote matrices in site, angular momentum and spin indices. The scattering path operator (SPO) matrix is defined as
where collects the single-site scattering matrices and is the matrix of the free space structure constants. It should be mentioned that we considered only the single-particle energy of the electron system. This approach is based on the magnetic local force theorem which is strictly valid only to first order in the change of the total energy at a stationary state.[3] Importantly, the derivation of the global spin-model parameters is based on the same theoretical foundations as the torque method presented in the following section for determining the parameters of a local spin model.
To obtain a systematic expansion in terms of spin variables, we decompose the -matrix at site in angular-momentum and spin-space as
where in the first term is the -matrix of the non-magnetic reference system, is the two-by-two unit matrix and the second term stands for the perturbation due to spin polarization, with being the vector of the Pauli matrices and we dropped the energy argument of the corresponding matrices. Note that (4) is valid only within a non-relativistic theory and the single underline refers to matrices with angular momentum indices only, while double underlines denote matrices with spin and angular momentum indices.
Defining the so-called structural Green’s function matrix of the non-magnetic reference system,
we can express the logarithm of the SPO matrix as
Performing the trace in spin space in the third term of the above equation,
while keeping only terms up to only, we end up with (1) (see Appendix A), where the two-spin and four-spin exchange interactions are defined as
It should be emphasized that the sums in (1) include all site indices. Using such a spin model in Monte Carlo or spin dynamics simulations demands to rearrange the Hamiltonian in terms of multisite interactions to avoid the coincidence of the sites in the summations. One can do this by systematically studying all possible index contractions in the Hamiltonian. The multisite Hamiltonian then reads
where the first, second, third, and fourth terms comprise the effective bilinear two-site exchange interactions, the two-site four-spin (biquadratic) interactions, the three-site four-spin and the four-site four-spin interactions, respectively. All the summations in (10) are taken over mutually distinct indices, i.e., none of the different indices coincide. The multisite interaction parameters can be expressed in terms of the two-spin and four-spin integrals as,
and
Note that the overall minus sign in the first term of (10) implies that positive/negative bilinear two-site interactions refer to ferromagnetic/antiferromagnetic pairwise couplings.
Ab initio theory of local exchange interactions
The method of infinitesimal rotations [32] [3] has been extended to non-collinear reference states by Szilva et al. [21] and later on used to explore the local DMI by Cardias et al. [24, 25]. Here we give a short recap of the method within the multiple scattering formalism. There are two ways to implement infinitesimal rotations: either in terms of the single-site scattering matrices and the structural resolvents or in terms of the inverse of the -matrices and the SPO matrix as done by Szilva et al. [21]. As what follows, the formalism based on the single-site scattering matrices and the structural resolvents will be presented and in Appendix B it will be shown how the results can be transferred to the formalism of Szilva et al. [21].
The interacting part of the grand potential for two-spin rotations is defined as,
where and denote variations due to the rotation of the spin at sites and , respectively, while stands for the variation by simultaneous rotations of the two spins. Based on (2), the single-spin variation of the grand potential can be calculated as
where the single-spin variation of the single-site scattering matrix is defined as
giving nonzero contribution only at site . For infinitesimal rotation of the spin vector we have , termed to as transverse fluctuation. In (15),
is the structural Green’s function matrix of the system. For the two-spin variation one can derive a similar expression,
Inserting Eqs. (15) and (18) into Eq. (14) and expanding the logarithm as in (7) to first order in the variation of the -matrix we obtain,
We note that we arrived at an expression similar to (8), the Green’s function matrix of the non-magnetic reference system being replaced by that of the magnetic system in the chosen spin-configuration. Since the Green’s function is now spin dependent it can be decomposed in spin space using the Pauli matrices,
Substituting this expression into (20) and evaluating the trace in spin space we arrive at expressions, where quantities similar to the tensorial exchange interactions [4] can be recognized. The two-spin variation of the grand potential separates into three contributions based on the number of Pauli matrices appearing in the spin trace,
where
The interaction parameters are defined as
where
and
The above expressions will be referenced as the formalism of the local magnetic interactions. Similar equations can be obtained for the local exchange interactions if we switch to the formalism used by Antropov et al. [32], Szilva et al. [21] and Cardias et al. [24], with the and matrices being replaced by the and matrices, respectively. Hence, we will reference it as the formalism of local magnetic interactions. The details can be found in Appendix B. From the derivation presented there, it appears that the formula (20) and the formula (B4) are identical for infinitesimal rotations of the spin vectors. However, there is no guarantee that the three contributions to are identical in the two local formalisms. This will be illustrated by numerical calculations in the next section.
It should also be emphasized that relativistic effects like spin-orbit coupling (SOC) were excluded from the theory, but we obtained interactions, like the Dzyaloshinskii–Moriya interaction or exchange anisotropy, known to be related to the SOC. The origin of the DMI in the absence of SOC was explained in Ref. [26] demonstrating that the two-spin variation of the isotropic four-spin multisite Hamiltonian in (10) leads to local interactions which appear in the same form as (24) and (25).
Local interactions for a Cr trimer on Au(111)
For our numerical considerations we choose an equilateral Cr trimer on the (111) surface of Au, used in many previous studies [25, 26, 33, 34] and we compare the numerical results obtained for the and types of local interactions implemented within the embedded cluster multiple scattering Green’s-function framework [35].
The Cr trimer on the surface of Au(111), in short Cr/Au(111), is considered in a non-relaxed geometry, where the Au and the Cr atoms occupy the positions in an ideal face-centered cubic lattice, with lattice constant 287.4 pm. The Cr atoms are sitting above the hollow positions of the top Au layer forming a right equilateral triangle. The host is calculated with 16 energy points on a semicircular path on the upper complex half plane with 2160 points in the irreducible wedge of the 2D Brillouin zone at the Fermi level and an angular momentum cut-off . The local interactions are calculated along a path in the configuration space illustrated in Figure 1, where was changed in 10° steps using the self-consistent potential obtained for each value of . We have verified that the magnitudes of the magnetic moments remain constant within a relative accuracy of 1% along the path considered, providing strong support for the rigid-moment approximation employed throughout our work.

Figure 1. The spin configuration of the Cr trimer on the surface of Au(111) used for the calculation of the local interactions.
The two-spin variation of the grand potential (22), henceforth referred to as the energy of the spin system, and its three contributions are shown in Figure 2 along the chosen path as calculated from both the and formalisms. Note that is normalized to unit variation of the angle , i.e., is presented in the figure. As predicted in the previous section, the total variations (red curves) agree perfectly for the two formalisms. The decompositions of the two-spin variation of the energy into the three contributions are, however, significantly

Figure 2. The two-spin variation of the energy for Cr/Au(111) along the path shown in Figure 1, by varying the angle of the spin vectors: (a) formalism and (b) formalism.
different in the two types of calculations. For example, around , provides the largest contribution to in the formalism, while and approach zero. In contrast, in the formalism, dominates , whereas and are suppressed.
The different contributions to the energy variation can be further analyzed in terms of the interaction parameters. The interactions in (26), (27) and (29), or their analogues in the formalism can be organized into a matrix, similarly as one introduces the exchange tensor for the relativistic two-spin interactions [4]. Then the number of independent interaction parameters can be reduced based on symmetry arguments in the same way as in Ref. [36]. The simplest matrix structure is assigned with the interaction between the two sites at the lower edge of the triangle labeled by 13 (the bond along the axis),
The interaction matrices related to the other two pairs can be obtained by rotation of this matrix, which means that the local interactions are described by the six parameters in Eq. (30). We should note that there exists a distinct set of these parameters for each spin configuration considered. The interaction parameters for Cr/Au(111) are plotted in Figure 3. It is evident that both the energy contributions and the interaction parameters differ between the two formalisms. Our calculations then suggest that neither the decomposition of the energy variations nor the corresponding interactions like local DMI can uniquely be defined. For the collinear case () the local DMI and the matrix elements of vanish, and only the isotropic interaction differs from zero. Interestingly, in this limit the value of is the same for the formalism and , representing the well-known Heisenberg exchange coupling invented by Liechtenstein et al. [3].

Figure 3. The local interaction parameters for the Cr/Au(111) as specified in the matrix (30) for the formalism (left panels) and for the formalism (right panels). The data labeled as ‘Cardias’ in panel (e) is copied from Ref. [24].
Despite the ambiguity of the local interaction parameters, a comparison can be made between parameters that are calculated within the same formalism. Figure 3(e) compares the components of the local DM vector we calculated in the formalism and those by Cardias et al. in Ref. [24]. Apparently, there is an almost perfect agreement between the parallel components, and the perpendicular components are also in the same range of magnitude, contrary to the results shown in Figure 2 of Ref. [26].
Local interactions from the global spin model
In the previous section we concluded that the method of infinitesimal rotations as described in Section III doesn’t provide unique local interactions. In order to get a deeper understanding why this ambiguity happens, we make use of the global isotropic four-spin model introduced in Sec. II and analytically derive the two-spin variation of the energy [29].
The multisite Hamiltonian defined in Eq. (10) applied to an equilateral trimer reads as
that means we have only three independent parameters, , , and . The two-spin variation of this Hamiltonian with respect to spins and can be expressed as
where and . This can be written as
with the local interaction matrix,
Substituting the spin configuration of the Cr trimer indicated in Figure 1,
into expression (34), the explicit form of the interaction can be given as a function of . The local interaction matrix between spins 1 and 3 then reads as
Note that the above matrix matches the form of in Eq. (30) dictated by symmetry with the following parameters:
The local interactions obtained from the expressions (37)–(42) are presented in Figure 4. Apparently, the energy range of these parameters compares well to that of the parameters calculated from the local model, see left panels of Figure 3. This can be readily understood, as the two calculations rely on closely related formalisms. Nevertheless, apart from the isotropic parameter , the interactions obtained from the local model show a more complicated –dependence than those from the global model. An even more serious problem arises from the fact that adding longitudinal terms, proportional to or , that leave invariant, can arbitrarily modify the local interaction parameters. An example is presented in Appendix C.

Figure 4. The local interaction parameters for the Cr/Au(111) from the global four-spin model as defined in Eqs. (37)–(42). The parameters entering these relations were calculated based on expressions (11) and (12): meV, meV and meV.
Projection to the transverse subspace
In this section, we demonstrate that all issues raised above are resolved by systematically removing the longitudinal contributions from the two-spin energy variation (33). Let’s define the basis vectors of the local frame of reference at site as
and the matrix elements of with respect to these basis vectors,
with . The interaction matrix can then be expressed as
where denotes the dyadic product of the vectors and .
The variation of the spin vector by infinitesimal rotation can be written as
The two-spin transverse variation of the energy is then given by
from which the transverse second derivatives of the energy can be inferred,
Since in Eq. (49) only transverse matrix elements of are present, a unique definition of the local interaction matrix can be achieved by projecting it to the transverse subspace,
This matrix can be used in the two-spin variation of the energy (33) when only transverse variations (infinitesimal rotations) of the spin vectors (48) are considered, while the longitudinal components of the matrix (47) are irrelevant.
For the pair 1–3, the local projected interaction matrix can be written in the form defined in Eq. (30) with the ‘isotropic’ parameter,
the ‘DM-like’ parameters,
and the ‘anisotropy’ parameters,
By comparing the above projected interactions with the original ones in Eq. (37)–(42), one can see that the projection introduces higher-order harmonics in .
The transverse two-spin derivatives of the energy, Eqs. (50) and (51) can also be evaluated with the local magnetic interactions, see Eqs. (26)–(29). In Figure 5 we show the calculated transverse two-spin derivatives for pair 1–3 along the chosen path in the configuration space using the and the formalisms, as well as the global four-spin model shown in (34). From Figure 5 one can infer that the transverse second derivatives based on the two local models almost perfectly agree (in fact, the difference is below 1 eV). They are also in very good agreement with those obtained from the global four-spin model, with a difference of about 1 meV which still refers to a relative accuracy of 1 %. This small difference likely arises due to higher-order multispin interactions incorporated in the local models, but not present in the global spin model (1).

Figure 5. The transverse two-spin derivatives of the energy for the Cr pair 1–3 in Cr/Au(111) along the path sketched in Figure 1 as calculated using the two different local spin models and the global spin model.
Finally, in this section, we calculate the local interaction matrices projected onto the transverse subspace [Eq. (52)] for both the local and global spin models. As demonstrated in Figure 5, the transverse second derivatives do not depend on which local formalism is used, therefore, the matrix elements of are also independent on the type of the formalism. In Figure 6, we compare the projected interaction parameters calculated from the local spin model with the analytic expressions obtained from the global isotropic four-spin model. Apparently, for each of the six parameters, the two curves agree with high accuracy, which is not surprising given the excellent agreement found for the corresponding transverse derivatives (see Figure 5).
Figure 6. The local projected interaction parameters for Cr/Au(111) calculated from the local spin model in the formalism (blue open squares) and from the global four-spin model by using the analytic expressions (53)–(58) (orange solid hexagons). Dashed and dotted lines serve as guide for the eyes.
Fe and Cr dimers on Cu(001)
So far, using the textbook example of a Cr trimer on Au(111), we demonstrated in detail how the two formalisms for describing local interactions are related and how local, configuration-dependent interactions can be expressed in terms of global, configuration-independent parameters. To prove the robustness of the method, we also considered dimers of Fe and Cr on the (001) surface of Cu, as studied by Carcia et al. [37]. The dimer is placed in nearest-neighbor positions along the (100) direction on the surface, with no structural relaxation performed. For the configuration path we choose a symmetric tilting of the spins in the plane defined by the bond of the dimer with a relative angle as illustrated in Figure 7(a).

Figure 7. (a) Illustration of the spin-configuration used for the dimers on Cu(001). The spin vectors are confined to the plane. (b) and (c) Calculated local DMI parameters for Fe/Cu(001) and Cr/Cu(001), respectively. Note that, by symmetry, the local DM vector is parallel to the M–M bond (M = Fe or Cr).
In Ref. [37], the local DMI for dimers, including Fe and Cr on Cu(001), was reported both in the presence and absence of SOC using what we refer to as the formalism. In Figure 7(b) and (Figure 7c) we compare our results with the ones shown in Figure 3 of Ref. [37] for the component of the DMI vector in the absence of SOC. It should be noted that the two atoms of the dimer are related by mirror symmetry, implying that the (non-relativistic) local DM vector is parallel to the line connecting them.
The results obtained from the formalism show good qualitative agreement with those of Ref. [37]: for the Fe dimer our values are approximately twice as large, while for the Cr dimer we found a scaling factor of about compared to those in Ref. [37]. However the results from the formalism significantly differ both qualitatively and quantitatively. This again confirms the ambiguity in the calculation of the local DMI.
Next, we attempted to establish the relationship between the local model parameters and those of the global spin Hamiltonian. The three site term in Eq. (31) obviously vanishes, while we extend the model with a six-spin term:
where and are the previously introduced bilinear and biquadratic interactions, respectively, and is the so called bicubic interaction parameter. The two-spin variation of this Hamiltonian can easily be calculated to obtain the dimer analogue of Eq. (34) as
From the interaction in Eq. (60) and the parametric configuration space path the local derivatives can be constructed and the same projection procedure can be applied as shown in Eq. (52).
The perturbative method introduced in section II is, in principle, suitable to calculate the parameters , and , in particular, for the Fe dimer we found a very slow convergence for the multi-spin integrals. In this section, we therefore employ a different strategy to determine the global parameters by fitting the analytic form of the two-spin energy variation. The transverse derivatives along the path indicated in Figure 7(a) can be expressed as
where the two non-zero curves can be used for the fit. To include information from both transverse derivatives, we perform simultaneous least square fits of the global parameters. The result of this procedure are collected in Tab. I. Based on the estimated errors, the fit is more accurate for Cr/Cu(001), with an absolute error of approximately 0.1 meV. In contrast, the fit for Fe/Cu(001) exhibits larger uncertainties, with absolute errors of 4 meV and 6 meV for the parameters and , respectively. In both cases we got a large bilinear term with opposite sign suggesting an anti-ferromagnetic ground state for the Fe dimer and a ferromagnetic one for the Cr dimer, in line with the observation in Ref. [37]. The biquadratic interaction is by one magnitude smaller for both cases, but significantly larger for the Fe dimer. The bicubic interaction is relevant only for the Fe dimer.
| Fe/Cu(001) | 244(4) | 84(4) | −43(6) |
| Cr/Cu(001) | −283.1(0.1) | 35.5(0.1) | 0.9(0.2) |
Table 1. The global interaction parameters measured in meV for Cr and Fe dimers on Cu(001) obtained from fitting the two-spin variation of the energy. The numbers in the parentheses represent the absolute errors of the parameters.
action is relevant only for the Fe dimer. Figure 8 shows the non-vanishing transverse derivatives obtained from the and formalisms, together with the corresponding analytic curves based on the fitted parameters in Tab. I. Consistent with both the theoretical framework introduced above and the results for Cr/Au(111), the two local formalisms yield identical transverse derivatives. Moreover, the global model, Eq. (59), reproduces the second derivatives remarkably well. Owing to the very small fitting errors in the multispin parameters, the agreement is almost perfect for the Cr dimer, as was also found for the Cr trimer on Au(111). For the Fe dimer, the larger discrepancy between the local and global models in the two-spin energy derivatives may indicate that higher-order terms omitted from the global model become significant.

Figure 8. The transverse two-spin derivatives of the energy for (a) Fe/Cu(001) and (b) Cr/Cu(001) along the path sketched in Figure 7(a) as calculated using the two different local spin models and the fitted global spin model.
Discussion and conclusions
Based on non-relativistic density functional theory, we established the relationship between the local and the global spin models, resolving the ambiguities raised in the literature [26, 27, 28]. In order to do so we started with the derivation of the global isotropic two- and four-spin interactions using a multiple scattering Green’s function technique. Then we derived the local two-spin variation of the energy within the same microscopic model in two ways, which we referred to as the and the formalism, and the local interaction parameters introduced by Cardias in Refs. [24, 25]. In the collinear case the interpretation of the local derivatives is well established. For the relativistic case, Udvardi et al. [4] describe how one can construct the interaction matrix based on the second derivatives of the energy, requiring at least three linearly independent collinear configurations. In the non-relativistic case, this approach simplifies to a single calculation in any collinear state since the magnetic energy surface is isotropic and a unique Heisenberg-like interaction is obtained [3]. On the contrary, by taking non-collinear states as reference, even in the non-relativistic case non-Heisenberg-like interactions appear. These interactions are the DMI and the anisotropy in a non-collinear framework, see Eqs. (25) and (24). Obviously, such interactions cannot happen in a global spin model and we refer to them as local interactions displaying a pronounced configuration dependence.
In Ref. [26] it was already shown that the configuration dependence of the local interactions can be related to global four-spin interactions, however, this relationship was not directly shown in terms of calculating the four-spin and the local interactions on the same footing. In the present work, we showed results based on the implementation of the interaction parameters for an equilateral Cr trimer on Au(111) as a prototype system extended with the study of Fe and Cr dimers on Cu(001). The results for the local models showed that the interaction parameters significantly depend on the formalism used for the calculation.
The analysis of the transition between the two formalisms highlighted an important detail also emphasized by Udvardi et al. [4], namely, that in a classical spin model the derivation of the interaction parameters should exclusively rely on transversal derivatives of the energy. By deriving local interactions from a global model we explicitly showed that longitudinal terms in the two-spin variation of the energy make the local interactions ambiguous.
In order to eliminate the longitudinal components from the interaction matrix we introduced projections to the transversal subspace in the local frame of reference. In terms of the global four-spin model containing three isotropic parameters (, and ), the projection was performed analytically and we obtained local interaction parameters along a chosen path in the configuration space. These parameters showed an almost perfect agreement with numerical data derived from either of the and projected local spin models. Thus, we unambiguously demonstrated the direct relationship between the global and the local spin-model parameters.
In Ref. [37], Cardias et al. attempted to establish a quantitative relation between the (non-projected) local DMI and spin currents for magnetic dimers on non-magnetic substrates. They found that, to first order in the Dyson series, the time-reversal-invariant, spin-dependent part of the Green’s function is proportional to , which they associated with the spin current induced by the non-collinear spin structure. Since this contribution to the Green’s function enters the expression for the local DMI, they concluded that the non-relativistic local DMI originates predominantly from spin currents.
Based on the global spin Hamiltonian given in Eq. (1) of the manuscript, the local DMI interaction parameters can be expressed as
see also Eq. (6) of Ref. [26]. This result can directly be compared to that of Ref. [37]. Importantly, however, the local DMI emerges naturally from the same theoretical framework used to derive the global spin Hamiltonian, without the need to invoke any additional microscopic interpretation of its origin.
We have to emphasize that a local spin model describes the transverse two-spin variations of the energy in the vicinity of a non-collinear ground state spin configuration. These energy variations can be fully described in terms of four transverse second derivatives of the energy. Reconstructing a local interaction matrix from these four parameters does not provide any additional physical insight. Instead, the transverse energy derivatives can be directly employed to calculate the adiabatic spin-wave spectrum of non-collinear magnetic systems, as demonstrated for the 120° Néel structure of a Cr monolayer on Au(111) [38].
Data availability
The data that support the findings of this study, including the raw values of global and local magnetic interaction parameters, are openly available at [39].
Acknowledgments
We acknowledge fruitful discussions with Stefan Blügel, Manuel dos Santos Dias, Samir Lounis, Attila Szilva, Levente Rózsa and László Udvardi. This work was supported by the Ministry of Culture and Innovation and the National Research, Development and Innovation Office under No. K142652 and ADVANCED 149745. Bendegúz Nyári acknowledge funding by the Hungarian Academy of Sciences via the János Bolyai Research Grant (Grant No. BO/00893/25). We acknowledge the Digital Government Development and Project Management Ltd. for awarding us access to the Komondor HPC facility based in Hungary.
The evaluation of the two-spin and four-spin interactions
Here we give some more details of the derivation of expressions (8) and (9). By introducing the non-magnetic reference system of (4), (3) can be split as
from which one can decompose the grand potential (2) as
The first term is the constant term already present in (2), the second one is a contribution of the non-magnetic reference system, the third one is an onsite anisotropy and the last one is the interaction term. Applying the Taylor series (7), the interaction term takes the following form,
It can be shown that the terms corresponding to odd orders of vanish due to time reversal symmetry, so up to fourth order in we have two terms only,
Separating the trace over the spin and the orbital quantum number the two spin contribution is
where we used the identity,
The definition of the two-spin integrals can easily be read off from (A5),
Similar derivation applies for the four-spin interactions:
where we used
The four-spin interaction parameter is then given by
The formalism of the local interactions
Let us introduce the -matrix as the inverse of the -matrix,
The infinitesimal variation of the -matrix can be expressed in terms of the infinitesimal variation of the -matrix,
Within the non-relativistic theory, (16) and the corresponding variation for the matrices apply. It can be shown that this is compatible with (75) for infinitesimal rotations of the spin vectors .
Substituting (75) into (20) and using the relationship between the SPO matrices and the structural Green’s function matrices for ,
one can derive the equation for the two-spin variation of the grand potential in terms of the - and the -matrices,
This equation is formally identical to (20), the - and -matrices being replaced by the - and -matrices, respectively. The interaction parameters can be derived in the same manner as before yielding
where and
with , being referenced as the formalism of the local magnetic interactions.
The ambiguity of the local interactions
The two-spin variation of the energy in the local spin models are separated into three terms (23)–(25). In (32), we deduced the local variation of the energy from the global spin model, but it requires further manipulation to make the mentioned separation visible. This can be achieved by adding terms to the variation of the energy proportional to and , thereby leaving the variation unchanged. First we rewrite (32) as
which can further be rewritten as,
where we used the vector identity . From (C2) one can easily read off the local ‘isotropic’, ‘anisotropy’, and ‘DM’ contributions to the energy variations. Substituting the parametric spin vectors (35) into (C2), we obtain the local interaction parameters,
Note that the local DM interactions (C4) and (C5) are identical with those derived in Ref. [26] and it has to be emphasized that the above expressions are different as compared to (37)–(42). Clearly enough, the reason for this is the inclusion of additional longitudinal terms in the energy variation.
References
References
- [7]S. Heinze, K. Von Bergmann, M. Menzel, J. Brede, A. Kubetzka, R. Wiesendanger, G. Bihlmayer, and S. Blügel, Spontaneous atomic-scale magnetic skyrmion lattice in two dimensions, nature physics 7, 713 (2011).
- [8]A. Fert, N. Reyren, and V. Cros, Magnetic skyrmions: advances in physics and potential applications, Nat. Rev. Mater. 2, 17031 (2017).DOI
- [9]Y. Tokura and N. Kanazawa, Magnetic skyrmion materials, Chemical Reviews 121, 2857 (2020).DOI
- [10]S. Mankovsky, S. Polesya, and H. Ebert, Extension of the standard Heisenberg Hamiltonian to multispin exchange interactions, Physical Review B 101, 174401 (2020).DOI
- [11]S. Brinker, M. d. S. Dias, and S. Lounis, The chiral biquadratic pair interaction, New journal of physics 21, 083015 (2019).
- [12]M. Gutzeit, A. Kubetzka, S. Haldar, H. Pralow, M. A. Goerzen, R. Wiesendanger, S. Heinze, and K. von Bergmann, Nano-scale collinear multi-Q states driven by higher-order interactions, Nature Communications 13, 5764 (2022).DOI
- [13]S. Hayami, R. Ozawa, and Y. Motome, Effective bilinear-biquadratic model for noncoplanar ordering in itinerant magnets, Physical Review B 95, 224424 (2017).DOI
- [14]N. Romming, H. Pralow, A. Kubetzka, M. Hoffmann, S. von Malottki, S. Meyer, B. Dupé, R. Wiesendanger, K. von Bergmann, and S. Heinze, Competition of Dzyaloshinskii-Moriya and higher-order exchange interactions in Rh/Fe atomic bilayers on Ir (111), Physical review letters 120, 207201 (2018).DOI
- [15]A. Krönlein, M. Schmitt, M. Hoffmann, J. Kemmer, N. Seubert, M. Vogt, J. Küspers, M. Böhme, B. Alonazi, J. Kügel, et al., Magnetic ground state stabilized by three-site interactions: Fe/Rh (111), Physical review letters 120, 207202 (2018).DOI
- [16]S. Paul, S. Haldar, S. Von Malottki, and S. Heinze, Role of higher-order exchange interactions for skyrmion stability, Nature communications 11, 4576 (2020).
- [17]N. D. Khanh, T. Nakajima, X. Yu, S. Gao, K. Shibata, M. Hirschberger, Y. Yamasaki, H. Sagayama, H. Nakao, L. Peng, et al., Nanometric square skyrmion lattice in a centrosymmetric tetragonal magnet, Nature Nanotechnology 15, 444 (2020).DOI
- [18]M. Hoffmann and S. Blügel, Systematic derivation of realistic spin models for beyond-heisenberg solids, Physical Review B 101, 024418 (2020).DOI
- [19]S. Grytsiuk, J.-P. Hanke, M. Hoffmann, J. Bouaziz, O. Gomonay, G. Bihlmayer, S. Lounis, Y. Mokrousov, and S. Blügel, Topological-chiral magnetic interactions driven by emergent orbital magnetism, Nature communications 11, 511 (2020).DOI
- [20]S. Brinker, M. dos Santos Dias, and S. Lounis, Prospecting chiral multisite interactions in prototypical magnetic systems, Physical review research 2, 033240 (2020).
- [21]A. Szilva, M. Costa, A. Bergman, L. Szunyogh, L. Nordström, and O. Eriksson, Interatomic Exchange Interactions for Finite-Temperature Magnetism and Nonequilibrium Spin Dynamics, Phys. Rev. Lett. 111, 127204 (2013).pubmed.ncbi.nlm.nih.gov/24093297
- [22]Y. O. Kvashnin, R. Cardias, A. Szilva, I. Di Marco, M. I. Katsnelson, A. I. Lichtenstein, L. Nordström, A. B. Klautau, and O. Eriksson, Microscopic Origin of Heisenberg and Non-Heisenberg Exchange Interactions in Ferromagnetic bcc Fe, Phys. Rev. Lett. 116, 217202 (2016).DOI
- [23]D. C. M. Rodrigues, A. Szilva, A. B. Klautau, A. Bergman, O. Eriksson, and C. Etz, Finite-temperature interatomic exchange and magnon softening in Fe over-layers on Ir(001), Phys. Rev. B 94, 044113 (2016).
- [24]R. Cardias, A. Bergman, A. Szilva, Y. O. Kvashnin, J. Fransson, A. B. Klautau, O. Eriksson, and L. Nordström, Dzyaloshinskii–Moriya interaction in absence of spin-orbit coupling, arXiv preprint arXiv:2003.04680 (2020).
- [25]R. Cardias, A. Szilva, M. Bezerra-Neto, M. Ribeiro, A. Bergman, Y. O. Kvashnin, J. Fransson, A. B. Klautau, O. Eriksson, and L. Nordström, First-principles Dzyaloshinskii–Moriya interaction in a non-collinear framework, Scientific Reports 10, 20339 (2020).DOI
- [26]M. dos Santos Dias, S. Brinker, A. Lászlóffy, B. Nyári, S. Blügel, L. Szunyogh, and S. Lounis, Proper and improper chiral magnetic interactions, Physical Review B 103, L140408 (2021).
- [27]R. Cardias, A. Szilva, A. Bergman, Y. Kvashnin, J. Fransson, S. Streib, A. Delin, M. I. Katsnelson, D. Thonig, A. B. Klautau, et al., Comment on “Proper and improper chiral magnetic interactions”, Physical Review B 105, 026401 (2022).DOI
- [28]M. dos Santos Dias, S. Brinker, A. Lászlóffy, B. Nyári, S. Blügel, L. Szunyogh, and S. Lounis, Reply to “Comment on ‘Comment on “Proper and improper chiral magnetic interactions”’, Physical Review B 105, 026402 (2022).
- [29]A. Lászlóffy, Study of magnetic nanoparticles at the atomic scale based on spin models from first principles, Ph.D. thesis, Budapest University of Technology and Economics, Budapest, Hungary (2022).
- [30]P. Lloyd, Wave propagation through an assembly of spheres: II. The density of single-particle eigenstates, Proceedings of the Physical Society 90, 207 (1967).DOI
- [31]S. Lounis, Multiple-scattering approach for multi-spin chiral magnetic interactions: application to the one- and two-dimensional rashba electron gas, New Journal of Physics 22, 103003 (2020).DOI
- [32]V. Antropov, M. Katsnelson, and A. Liechtenstein, Exchange interactions in magnets, Physica B: Condensed Matter 237-238, 336 (1997), proceedings of the Yamada Conference XLV, the International Conference on the Physics of Transition Metals.DOI
- [33]A. Antal, B. Lazarovits, L. Udvardi, L. Szunyogh, B. Újfalussy, and P. Weinberger, First-principles calculations of spin interactions and the magnetic ground states of Cr trimers on Au (111), Physical Review B 77, 174429 (2008).
- [34]G. Martínez-Carracedo, L. Oroszlány, A. García-Fuente, B. Nyári, L. Udvardi, L. Szunyogh, and J. Ferrer, Relativistic magnetic interactions from nonorthogonal basis sets, Physical Review B 108, 214418 (2023).
- [35]B. Lazarovits, L. Szunyogh, and P. Weinberger, Fully relativistic calculation of magnetic properties of Fe, Co, and Ni adclusters on Ag (100), Physical Review B 65, 104441 (2002).DOI
- [36]B. Nyári, A. Deák, and L. Szunyogh, Weak ferromagnetism in hexagonal Mn 3 Z alloys (Z= Sn, Ge, Ga), Physical Review B 100, 144412 (2019).DOI
- [37]R. Cardias, J. d. S. Silva, A. Bergman, A. Szilva, Y. O. Kvashnin, J. Fransson, A. B. Klautau, O. Eriksson, A. Delin, and L. Nordström, Unraveling the connection between high-order magnetic interactions and local-to- global spin Hamiltonian in noncollinear magnetic dimers, Physical Review B 108, 224408 (2023).DOI
- [38]L. Balogh, L. Udvardi, and L. Szunyogh, Magnetic anisotropy and chirality of frustrated cr nanostructures on au(111), Journal of Physics: Condensed Matter 26, 436001 (2014).DOI
- [39]B. Nyári, A. Lászlóffy, and L. Szunyogh, Dataset for “Reconciling global and local spin models”, 10.5281/zenodo.1849938 (2026).arxiv.org/abs/2611.18207
