Abstract
The Skyrmeparticle, the skyrmion, was introduced over half a century ago in the context of dense nuclear matter. But with skyrmions being mathematical objects—special types of topological solitons—they can emerge in much broader contexts. Recently skyrmions were observed in helimagnets, forming nanoscale spintextures. Extending over length scales much larger than the interatomic spacing, they behave as large, classical objects, yet deep inside they are of quantum nature. Penetrating into their microscopic roots requires a multiscale approach, spanning the full quantum to classical domain. Here, we achieve this for the first time in the skyrmionic Mott insulator Cu_{2}OSeO_{3}. We show that its magnetic building blocks are strongly fluctuating Cu_{4} tetrahedra, spawning a continuum theory that culminates in 51 nm large skyrmions, in striking agreement with experiment. One of the further predictions that ensues is the temperaturedependent decay of skyrmions into halfskyrmions.
Introduction
Skyrmionic spin textures in magnetic materials correspond to magnetic topological solitons. They were first observed in the noncentrosymmetric B20 helimagnets MnSi^{1,2}, FeGe^{3} and Fe_{0.5}Co_{0.5}Si (ref. 4). These skyrmionic textures are encountered also in a completely different branch of physics: in the theoretical description of nuclear matter^{5,6}. In this setting the skyrmions are of course not related to magnetic degrees of freedom, but rather to particles emerging from cold hadron vector fields at densities a few times that of ordinary nuclear matter. This is the density range relevant for compact astronomical objects such as neutron stars^{7,8,9}. The perhaps perplexing connection between these two seemingly disparate fields of physics is borne out of the underlying mathematical structures^{7,8,9,10,11,12,13,14}. The physical phenomena in the two different settings are both governed by an emerging set of differential equations with topological solitonic solutions: the skyrmions found first by Skyrme in the 1960s (ref. 5).
In this context, we investigate the formation and microscopic origin of the observed magnetic skyrmions in helimagnets. These skyrmions are large objects compared with the atomic length scale: they are about three orders of magnitude larger in size than the interatomic lattice spacing. Understanding the origin of these nanometrescale skyrmions therefore requires a multiscale approach. In the abovementioned B20 helimagnets, such is however not viable because all these materials are metallic. The metallicity causes lowenergy, delocalized electronic and magnetic degrees of freedom to mix so that they intrinsically involve multiple energy and spatial scales, which renders a multiscale approach presently intractable.
This is very different in the recently discovered skyrmionic material Cu_{2}OSeO_{3}, a large bandgap Mott insulator. The band gap enforces a natural separation between electronic and magnetic energy scales, which is advantageous for theoretical studies, as it allows for an accurate multiscale approach. Cu_{2}OSeO_{3} is actually the first example of an insulating material displaying the chiral helimagnetism that is desired for skyrmion formation while sharing the noncentrosymmetric cubic space group P2_{1}3 of the metallic B20 phases, but with a unit cell that is much more complex, containing 16 Cu atoms. Due to the presence of a magnetoelectric coupling^{15,16}, its skyrmions can be manipulated by an electric field^{17,18,19}, which is in principle very energy efficient as this avoids losses due to joule heating.
Here we present the results of a quantitative multiscale study of Cu_{2}OSeO_{3}, from the atomic up to the mesoscopic length scale. We find that the chiral helimagnetism and skyrmion formation in Cu_{2}OSeO_{3} relies on the presence of Cu_{4} tetrahedral building blocks that have a strong quantummechanical character. These effective entities spawn a continuum theory that results in skyrmions with a diameter of 51 nm, in quantitative agreement with reported experiments^{15}, and lead to a series of key experimental ramifications. Most notably, a skyrmion fractionalization in a specific temperature and magnetic field range, the existence of weakly dispersive magnetic excitations at high energies, and the presence of a weak antiferromagnetic indentation of the primary ferrimagnetic order parameter, which turns out to be responsible for the sign of the magnetic chirality. The quantum nature of the tetrahedral building blocks is also responsible for the reduction of the local spin expectation values and the low effective energy scale (which sets, for example, the value of the ordering temperature T_{C}), compared with the microscopic exchange energy scales.
Results
Roadmap of multiscale approach
The key aspects of helimagnetism and skyrmion formation in chiral magnetic systems such as Cu_{2}OSeO_{3} are shown in Fig. 1 and the main steps of our multiscale approach are outlined in Fig. 2. We first evaluate the Heisenberg exchange and DzyaloshinskiiMoriya (DM) interactions at the atomic level, using ab initio density functional theory (DFT) calculations (Fig. 2a). The resulting parameters reveal a striking separation of exchange energy scales which split the system into a network of magnetically ‘weak’ and ‘strong’ tetrahedra (Supplementary Note 1). The latter are the basic building blocks of helimagnetism in Cu_{2}OSeO_{3}: each ‘strong’ tetrahedron behaves as a spin1 triplet (apart from a finite quintet admixture, see Methods section), protected by a large energy gap of Δ≃275 K (Fig. 2c). Integrating out the dominant energy scale leads to an effective model on the socalled trillium lattice (Fig. 2d), marking a close analogy to MnSi and FeGe. The description at the mesoscopic scale is finally achieved by a longwavelength expansion of the effective model, which delivers the two basic parameters that control the skyrmionic physics (Fig. 1) in Cu_{2}OSeO_{3}: the exchange stiffness and the twisting parameter . Supplementing the free energy with further anisotropies and entropic effects, and using a minimal amount of experimental information, we arrive at a quantitative Dzyaloshinskii model, which in turn gives access to the magnetic phase diagram and other experimentally accessible observables. The ensuing notable predictions are the presence of a fractionalized skyrmion phase (Fig. 1c), the presence of a small AFM canting, which is adiabatically tied to the twisting of the total magnetization (Fig. 1e,f), and the important role of the canting in controlling the sign of the magnetic chirality. The quantummechanical nature of the ‘strong’ tetrahedra also has an additional number of profound ramifications that will be discussed in detail.
Description at the atomic level
To elucidate the quantum origin of the skyrmion textures in Cu_{2}OSeO_{3}, we proceed with a calculation of magnetic interactions at the atomic level. In undoped cuprates with Cu^{2+} atoms, such as Cu_{2}OSeO_{3}, the localized S=1/2 magnetic degrees of freedom originate from a single hole in the 3d electronic shell. In the crystal structure of Cu_{2}OSeO_{3}, the Cu^{2+} atoms make up a threedimensional network of cornersharing tetrahedra (Fig. 2b) with two inequivalent Cu sites, Cu(1) and Cu(2) featuring a different local environment: Cu(1)O_{5} bipyramids and distorted Cu(2)O_{4} plaquettes, respectively^{20,21}. Each tetrahedron contains Cu(1) and Cu(2) in a ratio of 1:3. The resulting net of magnetic Cu ions in Cu_{2}OSeO_{3} thus has a structure that is rather different from the previously mentioned metallic B20 helimagnets such as MnSi, where the magnetic Mn ions form a threedimensional cornersharing net of triangles, commonly referred to as the trillium lattice. The more complex crystal structure of Cu_{2}OSeO_{3} leads to five inequivalent superexchange coupling constants J_{ij} between neighbouring S=1/2 copper spins i and j (see Supplementary Fig. 1) and also five different Dzyaloshinskii–Moriya (DM) vectors D_{ij} in the microscopic magnetic Hamiltonian
where S_{i} denotes the quantummechanical spin operator at the Cu site i. We have determined these coupling constants (Table 1) by means of an extended set of ab initio densityfunctionalbased electronic structure calculations. As our numerical estimates are substantially higher than the values reported in the literature^{22}, they were crosschecked by calculating the magnetic ordering temperature T_{C} and the T dependence of magnetization and magnetic susceptibility by means of largescale Quantum Monte Carlo (QMC) simulations (Fig. 3). As these simulations agree very well with the measurements, they inspire further confidence in the accuracy of the DFTbased values.
We emphasize that the calculated Js between Cu(1) and Cu(2) are antiferromagnetic (AFM) and between Cu(2) and Cu(2) ferromagnetic (FM), and so the resulting spin model is not frustrated. The skyrmionic physics stems from the DM interactions, which impair the homogeneity of the FM order parameter and cause a chiral doubletwisting of the spins^{10,11,12}.
Identification of the basic magnetic building block
The exchange integrals (Table 1) reveal a pronounced hierarchy of magnetic energy scales in Cu_{2}OSeO_{3}. Specifically, there is a striking difference between two groups of exchange couplings, splitting the system into two kinds of Cu_{4} tetrahedra: one with strong (J_{S}~130–170 K) and the other with weak (J_{W}~30–50 K) exchange couplings (see Supplementary Fig. 2). Most of the DM terms are much smaller than the isotropic exchange: D_{ij}≪J_{ij}.
The four S=1/2 spins of a strong Cu_{4} tetrahedron can couple together to form either two total singlets (that is, with total spin S_{t}=0), three triplets (S_{t}=1) or a quintet state (S_{t}=2). In a tetrahedron with three AFM and three FM exchange couplings, the ground state (GS) is a total S_{t}=1 triplet, see Fig. 2b,c. The triplet GS is separated from the other spinmultiplets by a large energy gap of ~275 K. An important point is that the Cu_{4} tetrahedron triplet wavefunction is not the classical (tensor product) state ↑↑↑⇓› (where the double arrow labels the Cu(1) site in the tetrahedron) but rather a coherent quantum superposition of four classical states
with M labelling the three orthogonal triplet states with M=−1, 0, 1 (for brevity, only the M=1 wavefunction is given above). Although these are not the exact tetrahedron basis states due to the presence of a finite tripletquintet mixing (see methods section), this representation is qualitatively correct. This effective spin wavefunction is consistent with the experimental observation of a locally ferrimagnetic order parameter^{20,21}. The quantum fluctuations ingrained into these triplet wavefunctions, however, give rise to a substantial reduction of the local moments (Table 2), pinpointing the origin of the small moments observed experimentally^{20}. As opposed to transversal spin fluctuations arising from spin waves (which are small^{23}, on account of the high dimensionality, the ferrimagnetic nature of the order parameter and the absence of frustration), these local quantum fluctuations are longitudinal in character and hence directly affect the effective magnitude of the spin. This picture is confirmed by a lattice QMC simulation for the full model of Cu S=1/2 spins (Fig. 3b).
Not only are the effective triplets responsible for the spin length reduction, they also have a decisive role for several other aspects of direct experimental interest, such as the ordering temperature, the skyrmion diameter and even the sign of the magnetic chirality. The quantummechanical nature of the effective triplets is therefore utterly essential, in as much as a simple classical treatment delivers a quantitatively and even qualitatively wrong picture, as becomes clear below.
Effective description
The above description establishes Cu_{4} tetrahedra carrying magnetic triplets as building blocks in Cu_{2}OSeO_{3} at the next step of the multiscale approach. Within this abstraction, each of these tetrahedra can be contracted to a single lattice point which, by symmetry, can be chosen to coincide with the position of the corresponding Cu(1) site. The ensuing framework is shaped by cornershared triangles that together constitute a trillium lattice, which is precisely the same lattice that is formed by the Mn atoms in the B20 structure of MnSi and the Fe atoms in FeGe. This establishes a very close analogy between Mott insulating Cu_{2}OSeO_{3} and these wellknown metallic helimagnets, despite the fundamental differences in electronic structure. However in Cu_{2}OSeO_{3}, the effective triplet interactions can be derived relying on rigorous microscopic results. At this point both the weaker superexchange couplings J_{w} and the DM interaction D_{ij} become crucial. A straightforward perturbative calculation reveals that their net effect is a weak FM interaction between nearestneighbour (NN) and nextnearestneighbour (NNN) tetrahedral entities, with effective exchange couplings (Table 2, without spinmixing)
respectively, reflecting the tendency of the system towards FM ordering. This drastic reduction of the energy scale in the effective model is caused by the renormalization of the local spin moments and the sizable quantum fluctuations inside the magnetically strong tetrahedra. Experimentally, the most direct consequence of this energy scale reduction is the value of T_{C}: while a singlesite meanfield decoupling in the original spin1/2 model gives T_{C}=160 K, the corresponding decoupling in the effective spin1 model gives T_{C}=−4(J_{1}+J_{2})≃65 K, in much better agreement with the experimental value of 60 K (refs 20, 21). Details of these decouplings are provided in Supplementary Methods and Supplementary Fig. 3.
Besides the exchange interactions, also a DM coupling between NN and NNN entities emerges. Crucially, in the GS of a single strong tetrahedron, all diagonal matrix elements of the DM couplings within the tetrahedron vanish by symmetry. The twisting mechanism that causes chiral helimagnetism in Cu_{2}OSeO_{3}, therefore originates from the effective DM couplings between the strong Cu_{4} tetrahedra: these will be the root cause for skyrmions to emerge. Specifically, we find (Table 2, wthout spinmixing),
for a representative NN and NNN bond, respectively, while the DM vectors for the remaining bonds follow by symmetry (Table 3).
Description at the mesoscopic level
Having established the effective trillium lattice model of Cu_{2}OSeO_{3}, we now proceed to the longwavelength magnetic continuum theory that allows us to describe the helimagnetism of Cu_{2}OSeO_{3} on the mesoscopic scale. To this end, we follow Gnezdilov et al.^{24} and rewrite the four tetrahedral entities in each unit cell in terms of four coarsegrained fields, the total spin per site (in the effective model) and three AFM fields , μ=1–3. The energy in the continuum limit then becomes (see Methods section), in units of K:
where j≡8(J_{1}+J_{2}), a=8.91113 Å is the lattice constant,
and y=0.88557 is a structural parameter. The first term in equation (5) is the exchange energy of the FM ground state. The second line is the standard expression for the elastic and the twisting portion of the energy in cubic crystals. The last two lines are the crosscoupling terms between the FM mode F and the three AFM canting modes L_{μ}. Importantly, this coupling is particularly strong since ξ∝J_{1,2}. Minimizing equation (5) with respect to L_{μ} gives
So the AFM modes follow adiabatically the twisting of the F field, leading to a characteristic AFM indentation of the doubletwisted skyrmionic textures (Fig. 1e,f). This AFM superstructure will be discussed further below in relation to experiments.
Integrating out the AFM canting modes gives the effective energy density for the total magnetization alone:
where χ_{F}=3j+2κ^{2}/j, ′=−ξ^{2}/(2j), and ′=+κξ/j. So we find that both and are renormalized by the antiferromagnetic canting. Importantly, ξ^{2}/j∝J_{1,2}, so we expect a large renormalization for the elastic constant . Indeed, we find (Table 2, with spinmixing) ′=−11.19 K, almost two and a half times smaller than =−27.10 K. Intuitively, this reduction can be understood by the AFM nature of the canting modes which disrupts the dominant FM correlations. Turning to the twisting parameter, we find (Table 2, with spinmixing) =1.27 K and ′=−2.46 K. So not only is ′ larger than , but it also has the opposite sign. This means that the canting modes change the magnetic handedness of the system, an important piece of information which can be accessed experimentally (see Discussion).
Finally, with the characteristic diameter Λ′ of the doubletwisted skyrmion structures being 2π/Q′ (this distance coincides with the helix period in the helical phase^{15}), with wavenumber , the calculated magnetic constants result in a helix period of (Table 2, with spinmixing) Λ′≃50.89 nm, which is in striking agreement with the experimentally measured^{15} value Λ_{exp}≃50 nm. Thus, these numbers give a very satisfactory estimate of the effective chiral DM couplings given that they are describing a weak spinrelativistic exchange effect, which is greatly modified by the complicated internal canting of the spin structure in Cu_{2}OSeO_{3}.
Dzyaloshinskii model
To calculate the magnetic phase diagram we need to go one step further and include entropic effects and further anisotropies beyond the DM interactions. In the following, we cast the theory in terms of the total magnetization per volume M(r)=M_{sat}(0)F, where M_{sat}(0) is the magnetization density in the fieldaligned, ferrimagnetic (1/2 plateau) state at T=0. At the atomic level, M_{sat}(0)=4gμ_{B}/a^{3}, where g is the electronic spectroscopic factor of the Cu^{2+} ions, while experimentally^{16}, the corresponding magnetization is 0.531μ_{B} per Cu atom, which amounts to M_{sat}(0)=111.348 kA m^{−1}.
The phenomenological free energy density can be decomposed into three contributions,
The first contribution f_{e} contains the last two terms of (8) and the Zeeman energy,
where H is the internal magnetic field, =−k_{B}′/aM_{sat}(0)^{2} and =k_{B}_{exp}/(aM_{sat}(0))^{2}, where we readjust the DM parameter so that the experimental helix period Λ_{exp}=4π/ is reproduced. The second term of equation (9) holds the anisotropic couplings beyond the DM contributions. For a cubic crystal, the leading terms,
describe the exchange anisotropy ′ and the cubic magnetocrystalline anisotropies _{1} and _{2}. Finally, the third term of equation (9) collects the leading homogeneous free energy terms, which can be written as the usual Landau expansion for the order parameter field M,
where is the Curie temperature in the absence of DM couplings or anisotropies. The parameters , α, β, ′, K_{1} and _{2} can be estimated with a minimum of empirical additional information from experiment in conjunction with QMC data, following the procedure outlined in the Methods section. The resulting estimates are listed in Table 4.
From these parameters we can derive two key quantities of basic experimental interest. The first is the temperature range of the socalled precursor region^{12,25,26,27},
where two main effects take place. First, the doubletwisted core of skyrmions is energetically favourable and is stabilized at , while the helical ground state sets in only at , and second, the chiral twisting is accompanied by strong longitudinal modulations of the magnetization M. Hence, anomalous skyrmionic spintextures can be expected within this temperature range. The model parameters yield an estimate ΔT≃1.04 K, in good agreement with the interval of about 2 K in which the socalled ‘Aphases’ appear under magnetic field in Cu_{2}OSeO_{3} crystals^{28,29}. Given the crude representation of the thermodynamic potential term in the continuum theory by the Landau expansion, this value gives only the order or magnitude of the expected effects.
The second key quantity of interest is the characteristic field strength of the Dzyaloshinskii model,
which sets the scale for the unwinding of the chirally twisted helical state into the fieldaligned state. With coefficients from Table 4, a value H_{D}≃0.34 T is found. In turn, this gives the critical field for the closing of the cone state by a continuous spinflip, H_{c2}=H_{D}/2≃0.17 T. Again, these values are in good agreement with experiment (see Methods section).
Magnetization process and phase diagram
With all this in place, we can fully determine the equilibrium twodimensional solutions using micromagnetic simulations (see Methods section)^{12,25}. Figure 4 illustrates the magnetization processes for two different directions of the applied field for constant temperaturedepending magnetization . The corresponding characteristic plateau fields H_{p} for the transition between the conical helix and the fieldaligned state are shown in Fig. 5a. For fields along [100], the continuous transition from the conical helix state into the fieldenforced ferrimagnetic state is found at H_{c2}≃80 mT at 50 K, in very good agreement with experimental data.
The hightemperature region of the phase diagram is magnified in Fig. 5b to focus on the precursor region close to T_{C}, where skyrmionic cores are energetically more favourable compared with onedimensional helix solutions^{12,25}. Here we find two competing skyrmionic phases (Fig. 5c–f). The first one (Fig. 5c,e) is the standard fielddriven ‘−π’skyrmion phase of Fig. 1b with the radial skyrmions ordered in a hexagonal lattice^{10,11}. The second (Fig. 5d,f) is the ‘π/2’skyrmion state of Fig. 1c, which actually is the stable state at zero and low fields, because the fractionalization of skyrmions into halfskyrmions yields a higher packing density of the energetically advantageous skyrmionic cores.
It should be noted that the stability regions of condensed skyrmion phases may be underestimated compared with conical and helicoidal onedimensional states, as the parallel packing of skyrmionic strings may eventually lower its energy by undergoing a further twisting in the third spatial direction, amounting to a nontrivial conformation of the strings^{25}. Furthermore, thermal fluctuations may enhance the stability of skyrmionic phases as they may acquire large configurational entropic contributions compared with the ideal crystalline longrange ordered states, similar to liquids of polymer strings or condensed vortexmatter phases in typeII superconductors.
Discussion
The splitting of defects or of solitonic entities emerges in different research fields^{13,14,30,31,32}. Yet, for the Dzyaloshinskii models of chiral magnets, a detailed theoretical understanding of the full range of possible textures is still lacking. The main reason is that the splitting of solitons and defect formation relies on a delicate balance between coreenergies of solitons and their mutual interactions. In this regard, the emergence of the halfskyrmion phase in the vicinity of the ‘−π’skyrmion lattice of Cu_{2}OSeO_{3} opens a new venue to study the properties of textures with split topological units in experiment. The occurrence of this phase at low fields near the ordering transition should be traceable by thermodynamic measurements revealing further phase transitions in very narrow regions of the precursor range ΔT. Indeed, experimental indications for the presence of more than one mesophase in the precursor region has been reported from macroscopic thermodynamic measurements in the related chiral compounds with B20structure^{26,33,34}, and in Cu_{2}OSeO_{3} itself^{28,35}. Whether one of the phases observed is related to the halfskyrmion phase remains experimentally open, however, and requires further dedicated investigations.
To this end, we note that halfskyrmionic phases feature a sizeable longitudinal modulation of the order parameter, and contain defects like hedgehogs or narrow line or wall defects, where the magnetic order parameter M passes through zero. Hence, these textures are expected to display wide distributions of the quasistatic or ordered local magnetic moments, which could be discernible by local probes, such as μSR and NMR, and used as a signature of the halfskyrmion phase. The defected arrangement of partial skyrmions may also be detected by examining the higher harmonics in the diffraction data. While the finite extension of the radial solitonic cores requires the harmonic contents of a lattice arrangement to display all higher harmonic modes, the dense arrangement of skyrmionic cores and defects in a condensed phase of partial skyrmions will exacerbate these anharmonic effects.
Besides the interplay of skyrmion and halfskyrmion phases, the preceding analysis gives several additional important predictions. The first prediction hinges on the observation that the essential magnetic building blocks being weakly coupled, S_{t}=1 tetrahedra has profound ramifications for the magnetic excitation spectrum. It implies that the discrete spinmultiplet excitation spectrum of the four strongly coupled tetrahedron spins (consisting of two total singlets, two triplets and a quintet, see Fig. 2c) is largely preserved in the translationally invariant system. Therefore, besides the canonical Goldstone mode, a distinct set of weakly dispersive highenergy magnetic excitation modes should appear, where the highest mode goes up to an energy of ~450 K. The detailed form of this structure has been reported by Romhányi et al.^{23}, using a fully quantummechanical multiboson expansion. Quite remarkably, this structure was firmly established very recently by Ozerov et al.^{36}, using highfield electron spin resonance experiments with a terahertz free electron laser source. The characteristic dispersions associated with the hierarchy of magnetic interactions in Cu_{2}OSeO_{3} can be further probed by inelastic neutron scattering.
The second experimental consequence of the conceptual framework for skyrmion formation in Cu_{2}OSeO_{3} that we have established here is a canting of the four magnetic sublattices in the trillium lattice. This is similar in spirit to MnSi^{37}, but in Cu_{2}OSeO_{3} the canting must lead to a weakly antiferromagnetic modification of the primary ferrimagnetic order parameter that is related to the longrange ordering of the S_{t}=1 tetrahedron spins. This weak antiferromagnetism in Cu_{2}OSeO_{3} is the dual counterpart of the weak ferromagnetism that is present in chiral acentric bipartite antiferromagnets^{38}. The presence of this antiferromagnetic superstructure implies that in neutron (or resonant xray) diffraction experiments there be an additional set of corresponding magnetic Bragg peaks. Establishing their presence experimentally will prove that the skyrmions in Cu_{2}OSeO_{3}, which have a typical length scale of ~50 nm, carry an extra antiferromagnetic spinmodulation on a length scale more than two orders of magnitude smaller (Fig. 1e,f).
We have also shown that the canting has a strong influence on the primary order parameter, on account of a sizeable crosscoupling mechanism. Besides the renormalization of the helix period, this coupling actually reverses the magnetic handedness of the system, a key aspect that can be confirmed experimentally with polarized neutrons^{39,40,41,42,43,44}. Indeed, while preparing the final version of our manuscript, we became aware of a new experimental study by Dyadkin et al.^{45}, which confirms our theoretical prediction for the sign of the magnetic handedness of Cu_{2}OSeO_{3}.
Methods
Choice of enantiomer
All our calculations were done for the enantiomer defined by the following crystallographic positions of the Cu sites (in units of the lattice constant a=8.91113 Å):
where y=0.88557, a=0.13479, b=0.12096 and c=0.87267. The first four are the Cu(1) sites (Wyckoff positions 4a) and the remaining ones are the Cu(2) sites (general 12b positions). In the convention used in MnSi crystals (with Mn atoms corresponding to Cu(1) atoms in Cu_{2}OSeO_{3}), this is the righthanded enantiomer. The positive value of the wavevector Q′ (Table 2) then means that in Cu_{2}OSeO_{3}, the magnetic structure displays the same chirality with the crystal structure, in agreement with the new experimental study of Dyadkin et al.^{45}
Ab initio calculations
DFT calculations were performed using fplo version 9.07 (ref. 46) based on an extended set of local orbitals and VASP version 5.2 (refs 47, 48) employing the projectoraugmented wave method. For DFT calculations, we adopt the experimental structural data from Table 1 of ref. 20, without performing additional relaxations of unit cell parameters and atomic coordinates. For the exchange and correlation potential, we used the GGA (PBE96) parameterization. The kmeshes were checked for convergence. Relevant couplings (Supplementary Table 1) were evaluated by mapping the GGA band structure (Supplementary Fig. 4) onto the Cucentered Wannier functions. Isotropic (anisotropic) magnetic exchange couplings were evaluated using scalarrelativistic (full relativistic) spinpolarized DFT+U total energy calculations, adopting U_{d}=10.5 eV and 9.5 eV for Cu(1) and Cu(2), respectively. Further details on the computational method are provided in Supplementary Methods.
QMC simulations
QMC simulations of thermodynamical behaviour were done using the code loop from the software package ALPS version 1.3 (ref. 49). Simulations were performed on finite lattices of 8,192 spins S=1/2 using periodic boundary conditions, with 40,000 sweeps for thermalization and 400,000 sweeps after thermalization.
Tripletquintet spinmixing
We first summarize the main spectral features of an isolated ‘strong’ tetrahedron. This is described by the spin Hamiltonian (modulo an overall constant),
where S_{123} is the total spin quantum number of the Cu(2) sites, S_{4} is the Cu(1) spin, and S=S_{123}+S_{4} is the total spin. Thus the spectrum can be labelled by the good quantum numbers S_{12}, S_{123}, S, M›, where M is the total magnetization. The eigenspectrum of _{0} (Fig. 2c) consists of a ground state triplet 1, 3/2, 1, M›, two excited singlets 0, 1/2, 0, 0› and 1, 1/2, 0, 0› at energy (measured from from the ground state), a quintet 1, 3/2, 2, M› at and two triplets 0, 1/2, 1, M› and 1, 1/2, 1, M› at .
The spinmixing effect amounts to a finite admixture of the quintet state into the triplet ground state and originates in the weak couplings between the ‘strong’ tetrahedra. To understand this effect, we treat the weak intertetrahedral couplings at the meanfield level which, as shown in ref. 23, is an excellent approximation for the ground state. The resulting ‘tetrahedral meanfield’ (TMF) theory amounts to decoupling the bare S=1/2 model into a sum of independent TMF Hamiltonians with local mean fields acting on the four spins,
where ζ and η are selfconsistent meanfield parameters,
Given that and do not commute with the total spin S, the mean fields admix states with different S. The admixed states must however share the same M and S_{123}, which remain good quantum numbers. It follows that the triplet ground state is admixed only with the quintet state. In the M=1 manifold {φ_{1}› ≡ 1, 3/2, 1, 1›,φ_{2}› ≡ 1, 3/2, 2, 1›} we have, apart from an overall constant:
Solving selfconsistently for the ground state of this matrix leads to the local spin lengths and .
Effective model
As mentioned above, the effective description of Cu_{2}OSeO_{3} relies on replacing the magnetically ‘strong’ tetrahedra with effective quantummechanical entities, which by symmetry can be placed on the corresponding Cu(1) sites, forming a trillium lattice. For the moment, we disregard the spinmixing effect, that is, we regard each entity as a faithful spin1 object (we shall return to this below).
The firstorder coupling between the effective S_{t}=1 entities can be found using equivalent operators
where the spin lengths l_{1,2} can be found from (2): l_{1}=1/4 and l_{2}=5/12. A close inspection of the structure reveals that there are two types of bonds between effective spins and (Fig. 6), one where the tetrahedra t_{1} and t_{2} are connected via two weak Cu(1)Cu(2) couplings, and , and another where they are connected via a weak Cu(2)Cu(2) coupling . In the trillium lattice, and are, respectively, NN and NNN sites, and the corresponding couplings from firstorder perturbation theory are
To check the adequacy of firstorder perturbation theory, we have pushed the expansion up to second order. We found that the secondorder corrections are , , an order of magnitude weaker than the firstorder contributions, and so we can safely disregard them for all practical purposes. For completeness we should also note that in second order there appears an effective biquadratic exchange term of the form . The corresponding NN and NNN coefficients are also extremely small (, ) and can be safely disregarded.
Next, we treat the DM couplings, beginning with the ones within the strong tetrahedra, namely and . It turns out that all diagonal matrix elements of these couplings within the ground state triplet manifold vanish by symmetry and so, to lowest order, and do not have any role at all (this remains true with spinmixing present as well). In other words, the doubletwisting mechanism responsible for the appearance of the helical and the skyrmion lattice phase originates in the DM couplings between the strong tetrahedra. The firstorder effect of the latter can be treated in exactly the same way as the exchange couplings. Again there are only two independent DM couplings, one for NN and another for NNN bonds of the trillium lattice. By symmetry, we can write all DM couplings in terms of the two vectors of equation (4), see Table 3.
To find the latter, we use Table 3 and proceed as follows. The bond {ρ_{1}, ρ_{8}} maps to the bond {ρ_{1}, ρ_{3}+(1,0,0)} in the effective lattice. So, the firstorder contribution from to the effective DM vector is . Similarly, the bond {ρ_{4}, ρ_{12}} maps to {ρ_{4}, ρ_{2}}, and thus the contribution from to the effective DM vector is . Altogether,
Finally, the two Cu(2) sites {ρ_{5}, ρ_{12}} map to {ρ_{1}, ρ_{2}+(1,1,0)}, up to a primitive translation. Hence, equals , that is,
Let us now incorporate the groundstate spinmixing effect discussed above. Due to this spinmixing, the operators S_{t} do not satisfy the commutation relations of spins, but they can still be thought of as vector operators of length one. The firstorder perturbation theory in the ground state of equation (19) amounts to simply replacing the local spin lengths and in the above firstorder expressions. Table 2 lists the resulting values of various parameters of the model with and without spinmixing.
Continuum limit of the effective model
Here we briefly outline the basic steps and assumptions leading to equation (5). Our approach follows the standard route of Taylor expanding the magnetic energy in the longwavelength limit. Namely,
for an effective entity at R+r_{i}, where R labels the Bravais positions in the trillium lattice. In the next step, we express S_{i}(R) in terms of F(R) and L_{μ}(R) (see above), and collect all terms resulting from the 24 exchange and 24 DM anisotropy terms per unit cell, which requires performing a number of lattice summations (see Supplementary Methods, Sec. 4). Next, given that the ferromagnetic mode F is the dominant one, we disregard quadratic terms in L_{μ}, as well as terms of the type a D_{ij}L_{μ}∂F or a^{2}D_{ij}L_{μ}∂^{2}F. The latter scale, respectively, as and since , see equation (7). Essentially, this amounts to collecting all energy terms up to . Disregarding unimportant boundary terms then leads to equation (5), which is invariant under the symmetry group P2_{1}3 of the crystal, and in particular under the threefold rotation around ρ_{1}, which takes (x, y, z) → (z, x, y) in spin space, and (ρ_{2}, ρ_{3}, ρ_{4}) → (ρ_{3}, ρ_{4}, ρ_{2}) (see coordinates above), or its equivalent form (F, L_{1}, L_{2}, L_{3}) → (F, L_{2}, L_{3}, L_{1}).
The present approach is different from the one followed in ref. 50 for the model on MnSi (which has the same lattice with our effective model). We first remark that in the crystal structure of Cu_{2}OSeO_{3}, the quantity x≡1−y=0.11443<1/8, while in MnSi x=0.138>1/8. This means that the second (third) neighbour bonds in Cu_{2}OSeO_{3} correspond to third (second) neighbour bonds in MnSi (at x=1/8, the two bond types have the same distance). So to apply the results of ref. 50 to our case, we must replace and in the expressions of ref. 50. With these replacements, the resulting expressions for and are identical with the first two lines of equation (6), but with the important difference that 1−y is replaced by a quantity that depends nonlinearly on the exchange couplings, see ref. 50 for further details. This leads to very different numerical values for and and a helix period Λ≃1514, nm, which is too large compared with the experimental Λ_{exp}=50 nm (ref. 15).
Parameters of the Dzyaloshinskii model
The parameters , α and β (Table 4) were extracted by fitting the meanfield equation of state for the net magnetization M=M in the vicinity of the ordering temperature (where the effect of f_{a} can be disregarded, see below),
to the QMC data, Fig. 3, in appropriate ranges 40 K<T<60 K and H<5 T. For the anisotropic parameters of f_{a}, the empirical data from experiment show sizeable and positive _{1}, and negligible _{2} and ′. This can be seen from the reported^{16} anisotropic behaviour of the spinflip field H_{c2}, which corresponds to the unwinding of the chirally twisted helical state into the fieldaligned state: At temperatures below ~(2/3), H_{c2} becomes anisotropic, with the [100]direction as easy axis and the [111]direction as hard axis. The splitting between hard and easy axis directions, , reaches ~0.04 T at low temperatures^{16}. This behaviour can be qualitatively understood by noting that the effect of _{1} sets in when the magnetization tends towards saturation, and if the exchange anisotropy ′ is relatively much smaller (the latter has been checked independently by ab initio calculations at the atomic level). That the value of _{1} is sizeable can be also seen by the remarkably high fields H_{c1} ~0.02–0.035 T (ref. 16), corresponding to the helix reorientation transition from the helicoidal states or a multidomain helix state into the conical helix state, that is, the flopping of the helix in field direction. As the ratio H_{c2}/H_{c1} is not small, also the ratio between anisotropy and DMcoupling is expected to be sizeable. One source of _{1} comes from the magnetoelectric effect but, as shown in the Supplementary Methods (Sec. 3) based on the experimental magnetoelectric data^{16}, this contribution is too small to account for the experimental data. Here, the numerical value of _{1} (Table 4) has been adjusted to reproduce the field H_{c2} in the (100) field direction.
The quality of these numbers can be appreciated by a comparison with experiment. Magnetization data on single crystals show that H_{c2} does not depend significantly on the applied field direction for T>35 K (refs 16, 29, 35). In this range, we can also use the isotropically averaged magnetization data on polycrystalline samples^{20}, which give H_{c2}≃73 mT at T=50 K. At the same temperature, single crystal magnetization data give estimates H_{c2}≃68 and 70 mT for fields in [100] and [−554] direction, respectively, after the demagnetization corrections^{16}, while magnetization data on another single crystal gives H_{c2}≃60, 58 and 56 mT for field directions [100], [110] and [111], respectively^{35}. Other data by Adams et al.^{29} report higher values (ref. 29), while Seki et al.^{15} report . Some of the deviations may be traced back to demagnetization corrections, but the model is seen to be in reasonable agreement with the experimental facts.
Micromagnetic calculations
The determination of equilibrium states for the completed Dzyaloshinskii continuum model requires energy minimization of the free energy density functional of equation (9). The calculation of modulated states including the cubic anisotropy with coefficient K_{1} as given in Table 4 has been performed by numerical micromagnetic calculations using finite difference discretization with adjustable grid spacings and periodic boundary conditions. Energy minimization used simulated annealing for the search of initial solutions and a relaxation and grid refinement. In this way, twodimensional equilibrium solutions have been calculated (In Fig. 5, the ‘−π’Skyrmion lattice is the solution for H=7 mT and T=58.1 K, while the ‘π/2’lattice is the solution for H=4 mT and T=58.7 K.) These solutions are constrained approximations to threedimensional textures composed of skyrmionic strings, as they represent states with homogeneous spin structure in the third spatial direction.
Additional information
How to cite this article: Janson, O. et al. The quantum nature of skyrmions and halfskyrmions in Cu_{2}OSeO_{3}. Nat. Commun. 5:5376 doi: 10.1038/ncomms6376 (2014).
References
 1
Mühlbauer, S. et al. Skyrmion lattice in a chiral magnet. Science 323, 915–919 (2009).
 2
Tonomura, A. et al. Realspace observation of skyrmion lattice in helimagnet MnSi thin samples. Nano Lett. 12, 1673–1677 (2012).
 3
Yu, X. Z. et al. Skyrmion flow near room temperature in an ultralow current density. Nat. Commun. 3, 988 (2012).
 4
Yu, X. Z. et al. Realspace observation of a twodimensional skyrmion crystal. Nature 465, 901–904 (2010).
 5
Skyrme, T. H. R. A unified field theory of mesons and baryons. Nucl. Phys. 31, 556–569 (1962).
 6
Witten, E. Global aspects of current algebra. Nucl. Phys. B 223, 422–432 (1983).
 7
Ouyed, R. & Butler, M. Skyrmion stars. Astrophys. J. 522, 453 (1999).
 8
Luckock, H. & Moss, I. Black holes have skyrmion hair. Phys. Lett. B 176, 341–345 (1986).
 9
Lee, H. K. & Rho, M. inThe Multifaceted Skyrmion eds Brown G. E., Rho M. Ch 6, 147World Scientific (2010).
 10
Bogdanov, A. N. & Yablonsky, D. A. On the theory of inhomogeneous states of magnetic substances in the region of phasetransitions induced by a magneticfield the mixed state of antiferromagnets. Zh. Eksp. Teor. Fiz. 95, 178–182 (1989).
 11
Bogdanov, A. & Hubert, A. Thermodynamically stable magnetic vortex states in magnetic crystals. J. Magn. Magn. Mater. 138, 255–269 (1994).
 12
Rößler, U. K., Bogdanov, A. N. & Pfleiderer, C. Spontaneous skyrmion ground states in magnetic metals. Nature 442, 797–801 (2006).
 13
Castillejo, L., Jones, P., Jackson, A., Verbaarschot, J. & Jackson, A. Dense skyrmion systems. Nucl. Phys. A 501, 801–812 (1989).
 14
Kugler, M. & Shtrikman, S. Skyrmion crystals and their symmetries. Phys. Rev. D 40, 3421 (1989).
 15
Seki, S., Yu, X. Z., Ishiwata, S. & Tokura, Y. Observation of skyrmions in a multiferroic material. Science 336, 198–201 (2012).
 16
Belesi, M. et al. Magnetoelectric effects in single crystals of the cubic ferrimagnetic helimagnet Cu2OSeO3 . Phys. Rev. B 85, 224413 (2012).
 17
White, J. S. et al. Electric field control of the skyrmion lattice in Cu2OSeO3 . J. Phys.: Condens. Matter 24, 432201 (2012).
 18
Lin, S.Z., Reichhardt, C., Batista, C. D. & Saxena, A. Driven skyrmions and dynamical transitions in chiral magnets. Phys. Rev. Lett. 110, 207202 (2013).
 19
Fert, A., Cros, V. & Sampaio, J. Skyrmions on the track. Nat. Nanotechnol. 8, 152–156 (2013).
 20
Bos, J.W. G., Colin, C. V. & Palstra, T. T. M. Magnetoelectric coupling in the cubic ferrimagnet Cu2OSeO3 . Phys. Rev. B 78, 094416 (2008).
 21
Belesi, M. et al. Ferrimagnetism of the magnetoelectric compound Cu2OSeO3 probed by ^{77}Se NMR. Phys. Rev. B 82, 094422 (2010).
 22
Yang, J. H. et al. Strong DzyaloshinskiiMoriya interaction and origin of ferroelectricityin Cu2OSeO3 . Phys. Rev. Lett. 109, 107203 (2012).
 23
Romhanyi, J., van den Brink, J. & Rousochatzakis, I. Entangled tetrahedron ground state and excitations of the magnetoelectric skyrmion material Cu2OSeO3. Phys. Rev. B (Rapid), in press. Preprint at http://arxiv.org/abs/1403.4081 (2014).
 24
Gnezdilov, V. P. et al. Magnetoelectricity in the ferromagnetic Cu2OSeO3: symmetry analysis and Raman scattering study. Low Temp. Phys. 36, 550 (2010).
 25
Rößler, U. K., Leonov, A. A. & Bogdanov, A. N. Chiral skyrmionic matter in noncentrosymmetric magnets. J. Phys.: Conf. Ser. 303, 012105 (2011).
 26
Wilhelm, H. et al. Precursor phenomena at the magnetic ordering of the cubic helimagnet FeGe. Phys. Rev. Lett. 107, 127203 (2011).
 27
Wilhelm, H. et al. Confinement of chiral magnetic modulations in the precursor region of FeGe. J. Phys.: Condens. Matter 24, 294204 (2012).
 28
Seki, S. et al. Formation and rotation of skyrmion crystal in the chirallattice insulator Cu2OSeO3 . Phys. Rev. B 85, 220406(R) (2012).
 29
Adams, T. et al. Longwavelength helimagnetic order and skyrmion lattice phase in Cu2OSeO3 . Phys. Rev. Lett. 108, 237204 (2012).
 30
Kutka, R., Trebin, H. R. & Kiemes, M. The topological theory of semidefects in ordered media. J. Phys. (France) 50, 861–885 (1989).
 31
Morinari, T. Halfskyrmion picture of a singleholedoped CuO2 plane. Phys. Rev. B 72, 104502 (2005).
 32
Hen, I. & Karliner, M. Hexagonal structure of baby skyrmion lattices. Phys. Rev. D 77, 054009 (2008).
 33
Pappas, C. et al. Chiral paramagnetic skyrmionlike phase in MnSi. Phys. Rev. Lett. 102, 197202 (2009).
 34
Grigoriev, S. V. et al. Chiral criticality in the doped helimagnets Mn1−yFeySi. Phys. Rev. B 83, 224411 (2011).
 35
Živković, I., Pajić, D., Ivek, T. & Berger, H. Twostep transition in a magnetoelectric ferrimagnet Cu2OSeO3 . Phys. Rev. B 85, 224402 (2012).
 36
Ozerov, M. et al. Establishing the fundamental magnetic interactions in the chiral skyrmionic Mott insulator Cu2OSeO3 by terahertz electron spin resonance. Phys. Rev. Lett. 113, 157205 (2014).
 37
Dmitrienko, V. E. & Chizhikov, V. A. Weak antiferromagnetic ordering induced by DzyaloshinskiiMoriya interaction and pure magnetic reflections in MnSitype crystals. Phys. Rev. Lett. 108, 187203 (2012).
 38
Bogdanov, A. N., Rößler, U. K., Wolf, M. & Muller, K.H. Magnetic structures and reorientation transitions in noncentrosymmetric uniaxial antiferromagnets. Phys. Rev. B 66, 214410 (2002).
 39
Tanaka, M., Takayoshi, H., Ishida, M. & Endoh, Y. Crystal chirality and helicity of the helical spin density wave in MnSi. I. Convergentbeam electron diffraction. J. Phys. Soc. Jpn 54, 2970–2974 (1985).
 40
Ishida, M., Endoh, Y., Mitsuda, S., Ishikawa, Y. & Tanaka, M. Crystal chirality and helicity of the helical spin density wave in MnSi. II. Polarized neutron diffraction. J. Phys. Soc. Jpn 54, 2975–2982 (1985).
 41
Dyadkin, V. A. et al. Control of chirality of transitionmetal monosilicides by the Czochralski method. Phys. Rev. B 84, 014435 (2011).
 42
Grigoriev, S. V., Chetverikov, Y. O., Lott, D. & Schreyer., A. Field induced chirality in the helix structure of Dy/Y multilayer films and experimental evidence for DzyaloshinskiiMoriya interaction on the interfaces. Phys. Rev. Lett. 100, 197203 (2008).
 43
Grigoriev, S. V. et al. Crystal handedness and spin helix chirality in Fe1−xCoxSi. Phys. Rev. Lett. 102, 037204 (2009).
 44
Grigoriev, S. V. et al. Interplay between crystalline chirality and magnetic structure in Mn1−xFexSi. Phys. Rev. B 81, 012408 (2010).
 45
Dyadkin, V. et al. Chirality of structure and magnetism in the magnetoelectric compound Cu2OSeO3 . Phys. Rev. B 89, 140409(R) (2014).
 46
Koepernik, K. & Eschrig, H. Fullpotential nonorthogonal localorbital minimumbasis bandstructure scheme. Phys. Rev. B 59, 1743–1757 (1999).
 47
Kresse, G. & Furthmuller, J. Efficient iterative schemes for ab initio totalenergy calculations using a planewave basis set. Phys. Rev. B 54, 11169–11186 (1996).
 48
Kresse, G. & Furthmuller, J. Efficiency of abinitio total energy calculations for metals and semiconductors using a planewave basis set. Comput. Mater. Sci. 6, 15–50 (1996).
 49
Albuquerque, A. et al. The ALPS project release 1.3: opensource software for strongly correlated systems. J. Magn. Magn. Mater. 310, 1187–1193 (2007).
 50
Chizhikov, V. A. & Dmitrienko, V. E. Multishell contribution to the DzyaloshinskiiMoriya spiraling in MnSitype crystals. Phys. Rev. B 88, 214402 (2013).
 51
Belesi, M. et al. Magnetic properties of the magnetoelectric compound Cu2OSeO3: Magnetization and ^{77}Se NMR study. J. Phys.: Conf. Ser. 303, 012069 (2011).
Acknowledgements
We acknowledge fruitful discussions with J.Ph. Ansermet, A.N. Bogdanov, V.A. Chizhikov, V.E. Dmitrienko and Y. Onose. I.R. was supported by the Deutsche Forschungsgemeinschaft (DFG) under the EmmyNoether program. O.J. and A.A.T. were partly supported by the Mobilitas program of the ESF, grant numbers MJD447 and MTT77, respectively.
Author information
Affiliations
Contributions
I.R., O.J., U.K.R., J.v.d.B. and H.R. devised the project. The DFT calculations were carried out by O.J. and A.A.T. The QMC calculations were performed by O.J. The effective model was worked out by I.R. The continuum expansion of the effective model was done by I.R. and U.K.R. The micromagnetic calculations were carried out by U.K.R. and A.A.L. The manuscript and the supplement were written by J.v.d.B., I.R., O.J. and U.K.R. All the authors contributed to the analysis and interpretation of the results.
Corresponding authors
Ethics declarations
Competing interests
The authors declare no competing financial interests.
Supplementary information
Supplementary Information
Supplementary Figures 14, Supplementary Table 1, Supplementary Note 1, Supplementary Methods and Supplementary References (PDF 3962 kb)
Rights and permissions
About this article
Cite this article
Janson, O., Rousochatzakis, I., Tsirlin, A. et al. The quantum nature of skyrmions and halfskyrmions in Cu_{2}OSeO_{3}. Nat Commun 5, 5376 (2014). https://doi.org/10.1038/ncomms6376
Received:
Accepted:
Published:
Further reading

Energyefficient ultrafast nucleation of single and multiple antiferromagnetic skyrmions using inplane spin polarized current
Scientific Reports (2021)

Melting of a skyrmion lattice to a skyrmion liquid via a hexatic phase
Nature Nanotechnology (2020)

Multiple lowtemperature skyrmionic states in a bulk chiral magnet
npj Quantum Materials (2019)

Linearly polarized GHz magnetization dynamics of spin helix modes in the ferrimagnetic insulator Cu2OSeO3
Scientific Reports (2017)

Dramatic pressuredriven enhancement of bulk skyrmion stability
Scientific Reports (2016)
Comments
By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.