PHYSICAL REVIEW B 85, 094429 (2012)

Chiral modulations and reorientation effects in MnSi thin films

E. A. Karhu,1 U. K. Rößler,2 A. N. Bogdanov,2 S. Kahwaji,1 B. J. Kirby,3 H. Fritzsche,4 M. D. Robertson,5

C. F. Majkrzak,3 and T. L. Monchesky1,*
1
Department of Physics and Atmospheric Science, Dalhousie University, Halifax, Nova Scotia, Canada B3H 3J5
2
Institut für Festkörper- und Werkstoffforschung, IFW Dresden, Postfach 270016, D-01171 Dresden, Germany
3
Center for Neutron Research, NIST, Gaithersburg Maryland 20899, USA
4
National Research Council Canada, Canadian Neutron Beam Centre, Chalk River, Ontario, Canada K0J 1J0
5
Department of Physics, Acadia University, Wolfville, Nova Scotia, Canada B4P 2R6
(Received 1 November 2011; published 27 March 2012)
We present an experimental and theoretical investigation of the influence of a uniaxial magnetocrystalline
anisotropy on the magnetic textures that are formed in a chiral magnetic system. We show that the epitaxially
induced tensile stress in MnSi thin films grown on Si(111) creates an easy-plane uniaxial anisotropy. The
magnetoelastic shear stress coefficient is derived from SQUID magnetometry measurements in combination with
transmission electron microscopy and x-ray diffraction data. Density functional calculations of the magnetoelastic
coefficient support the conclusion that the uniaxial anisotropy originates from the magnetoelastic coupling.
Theoretical calculations based on a Dzyaloshinskii model that includes an easy-plane anisotropy predict a variety
of modulations to the magnetic order that are not observed in bulk MnSi crystals. Evidence for these states is
found in the magnetic hysteresis and polarized neutron reflectometry measurements.

DOI: 10.1103/PhysRevB.85.094429 PACS number(s): 75.25.−j, 75.30.−m, 75.70.Ak

I. INTRODUCTION and the region of regular chiral modulations where a number

of complex multidimensional states and numerous physical
Manganese silicide (MnSi) belongs to a group of magnetic
anomalies have been observed.19–27 These unconventional
noncentrosymmeteric crystals with space group P 21 3 (B20-
textures (so-called precursor states) arise near the TC as a result
type structures), called cubic helimagnets.1,2 In these materi-
of a softening of the magnetization.13,28 They are characterized
als, antisymmetric Dzyaloshinskii-Moriya (DM) interactions
by a strong coupling between angular and longitudinal mod-
induce long-range helical modulations with a fixed sense of the
ulations and differ fundamentally from regular chiral phases
magnetization rotation.1–3 Recent theoretical work shows that
with angular modulations and a magnetization with a fixed
in magnetic materials with intrinsic or induced chirality, DM
magnitude.13,26–28 In this paper, we limit our investigation of
interactions provide a unique mechanism to stabilize two- and
MnSi thin films to the region of regular modulations, below
three-dimensional modulations of the magnetic order (chiral
this precursor regime.
skyrmions).4–7 Recently, skyrmionic states have been observed
According to theoretical analysis, magnetic modulations
in thin layers of cubic helimagnets8,9 and in ultrathin magnetic
in the form of helices, distorted by an applied magnetic
films10 with surface- or interface-induced DM interactions.11
field transverse to their propagation direction, here called
Chiral skyrmions have also been observed in thin layers
helicoids, and two-dimensional modulations, skyrmion lat-
of MnSi.12 Similar skyrmionic states are predicted to exist
in other classes of noncentrosymmetric materials, e.g., in tices, also arise in cubic helimagnets as regular solutions
chiral liquid crystals, ferroelectrics, multiferroics, and in of the corresponding micromagnetic equations.1,6,29 In bulk
confined systems (e.g., nanolayers of magnetic metals)6,13,14 samples, they usually only exist as metastable states, while
and are expected to find applications in emerging spintronic the cone phase corresponds to the global energy minimum
technologies.10,15 over practically the whole region where chiral modulations
MnSi is an interesting and convenient material to investigate occur (for details, see Ref. 29). However, the cone phase
chiral magnetism as well as to explore this class of materials can be suppressed by intrinsic cubic anisotropies or (more
in spintronic devices. In bulk MnSi at zero magnetic field, effectively) by induced uniaxial distortions.29 Thus, a surface-
the competition between the DM and the direct exchange induced uniaxial anisotropy that has an easy axis can stabilize
interactions produces helical magnetic order below a Curie hexagonal skyrmion lattices in magnetic nanostructures with
temperature of TC = 29.5 K with a wavelength of 2π/Q = the skyrmion axis along the anisotropy axis. This effect
18 nm. The pitch vector Q points along 111 and the magnetic explains the experimental observations of skyrmion lattices
moments lie perpendicular to it.16 If an external magnetic field in thin layers of cubic helimagnets (Fe,Co)Si and FeGe, as
is applied in a direction other than 111, Q will rotate in the reported in Refs. 8 and 9. An important issue to address, from
direction of the magnetic field and become parallel to it at a both a fundamental and an applications point of view, is the
field μ0 HC1  0.1 T, as shown theoretically by Plumer and response of the helical order in thin films to both surface-
Walker,17 and confirmed more recently with small-angle neu- induced magnetic couplings and to an external magnetic field.
tron scattering (SANS).18 Above HC1 , the magnetic moments One of the important differences between the results conducted
form a conical phase, which collapses into a ferromagnetic on bulk MnSi and those conducted on thin films is the presence
state at μ0 HC2 = 0.6 T. In cubic helimagnets, there is a of strain, which is expected to produce a uniaxial anisotropy
narrow temperature interval between the paramagnetic phase through magnetoelastic coupling. Generally, both strain and

1098-0121/2012/85(9)/094429(12) 094429-1 ©2012 American Physical Society

E. A. KARHU et al. PHYSICAL REVIEW B 85, 094429 (2012)

surface effects offer control of the magnetic anisotropies over magnetization M(r) and its gradients1,2
a wide range of energies, which should influence the chiral c
magnetic textures in thin layers of cubic helimagnets. w(M) = Ms2 (∇m)2 + bMs2 m · (∇ × m)
2
In contrast to bulk crystals, there has been relatively little
1
work on the magnetic properties of MnSi thin films.30–33 We + Ku (m · n̂)2 − μ0 H · M − μ0 Hd · M , (1)
have shown that the helical magnetic order is preserved in 2
MnSi(111) films grown on Si(111), which makes this material where m = M/Ms is a unit vector along the direction of
attractive for spintronics studies of helical magnets. We cal- the magnetization M (Ms = |M|). The magnetic stiffness
culate the magnetic phase diagram for the phenomenological associated with the exchange interaction c is related to the spin-
Dzyaloshinskii model1 with easy-plane uniaxial anisotropy wave stiffness A by c = AS/(Ms2 a 3 ), where S = 0.8 is the
in applied magnetic fields transverse to the axis in order to spin per unit cell in units of h̄ and a = 0.4558 nm is the lattice
predict the stable modulations of the magnetic order that can be constant. The Dzyaloshinskii-Moriya constant D is related to
observed in the chiral magnetic thin films. These calculations the b coefficient describing the Lifshitz invariant through b =
display existence regions for stable helicoids and skyrmion DS/(Ms2 a 3 ). Aside from these two major exchange couplings
lattices with tubular skyrmion strings running along the applied that are present in isotropic or cubic chiral magnets,1 this
field in a direction perpendicular to the anisotropy axis. Dzyaloshinskii model only considers Zeeman energy (H is the
In this paper, we show that it is possible to induce an easy- applied magnetic field) and demagnetization energy with stray
plane-type magnetic anisotropy in epitaxial MnSi(111) thin field Hd .34
films with a positive and isotropic in-plane strain. Although The cubic anisotropy and the anisotropic exchange are
Si has a lattice constant of 0.5431 nm that is 16% larger than omitted in functional (1) because these terms are small. Adding
that of MnSi, a = 0.4558 nm, the Si(111) surface unit cell these terms recovers the model used by Bak and Jensen to
is only 3.1% larger than that of MnSi(111) for an epitaxial explain the magnetic structure and magnetization processes
relationship of [110]MnSi [112]Si. The noncentrosymmetric in bulk MnSi and related chiral cubic helimagnets.2 Enhanced
MnSi crystal structure results in inversion domains in films uniaxial anisotropy can arise in magnetic nanolayers as a result
grown on Si(111) (Ref. 30) that result in both left- and of elastic strains induced by a lattice mismatch between the
right-handed helical magnetic domains,31 which explains the magnetic layer and the overlying or underlying layers.35,36 In
glassy magnetic behavior observed in these films. Polarized addition, anisotropies may be introduced by symmetry break-
neutron reflectometry (PNR) and superconducting quantum ing at the layers boundaries, and due to physical and chemical
interference device magnetometry (SQUID) results show that modifications at the surfaces and interfaces.37,38 Depending
the pitch vector is along the film normal, and that the helical on material and geometrical parameters, the surface-induced
wavelength 2π/Q = 13.9 ± 0.1 nm is smaller than measured anisotropy can vary from cases when it is strictly confined to
in bulk material.31 Here, we present a study of the response the surfaces or interfaces to cases when it smoothly extends
of MnSi thin films to a magnetic field that is transverse to into the depth of magnetic layers.39 An inhomogeneous
Q displaying interesting reorientation of the helical magnetic distribution of the induced anisotropy across the thickness
order, which is significantly different than in bulk crystals. of the film may stabilize specific spatially inhomogeneous
SQUID magnetometry studies of the magnetic reorientation magnetic states, so-called twisted phases.39,40 However, their
are complemented by PNR measurements to obtain a depth existence regions are restricted to specific relations between
profile of the magnetization in the films. To understand material and external parameters.39,40 In this paper, we neglect
the origin of these differences, we measure the uniaxial effects imposed by spatial inhomogeneity of the induced
magnetocrystalline anisotropy in the films as a function of anisotropy and describe the induced uniaxial anisotropy by
thickness by SQUID magnetometry. A comparison of the an energy contribution with constant Ku [Eq. (1)]. We show
anisotropy with the strain determined by x-ray diffraction and in this paper that the MnSi films have an induced anisotropy
transmission electron microscopy (TEM) gives the magnetoe- that is of easy-plane type. For this case, Ku > 0 in functional
lastic coupling coefficients. These results are compared to a (1), and the ground state of the system is a single harmonic
fully relativistic electron-structure calculation using density helix with a wave number Q = b/c that propagates along n̂.
functional theory (DFT) of magnetocrystalline anisotropy in An applied magnetic field perpendicular to the film surface
rhombohedrally distorted MnSi bulk. transforms the helix into the conical phase [Fig. 1(a)] with
analytical solutions
⊥
II. MAGNETIC PHASE DIAGRAM OF CHIRAL CUBIC ψ(ξ ) = ξ/LD , cos(θ ) = H /HC2 , (2)
HELIMAGNETS WITH INDUCED UNIAXIAL
where θ is the polar angle of the magnetization with respect
ANISOTROPY
to n̂ and ψ is the azimuthal angle, ξ is a spatial variable
To model the magnetic states in epitaxial MnSi films, along the propagation direction, and LD = 2π c/b = 2π/Q is
we add an additional uniaxial magnetic anisotropy to the the wavelength of the modulation. The out-of-plane saturation
⊥
phenomenological free energy for a noncentrosymmetric cubic field HC2 can be expressed as
ferromagnet, which is appropriate as long as the films are 2K0 2Ku
⊥
strained homogeneously and are not too thin, so that specific μ0 HC2 = + + μ0 Ms ,
surface-induced effects can be neglected. This phenomeno- Ms Ms
 (3)
logical model can be written as a functional W = w(M)dV b2 Ms2 1 AQ2
K0 = = Ms ,
with the energy density w written in terms of a slowly varying 2c 2 gμB

094429-2
CHIRAL MODULATIONS AND REORIENTATION EFFECTS . . . PHYSICAL REVIEW B 85, 094429 (2012)

FIG. 1. (Color online) Modulated states in a thin layer of a cubic

helimagnet with in-plane magnetization: (a) a conical phase that
arises in the applied field perpendicular to the film surface, (b)
a helicoid phase, which is a helix distorted by the applied field
perpendicular to the propagation direction, (c) elliptically distorted
conical phases formed in a strong in-plane magnetic field, and (d) a
hexagonal skyrmion lattice with elliptical distortions imposed by a
uniaxial anisotropy. The unit vector n̂ indicates the direction of the
uniaxial anisotropy.

where K0 is the “effective stiffness” of the conical phase.

The conical phase is characterized by the constant modulation
period LD and a linear magnetization curve. Transversally
distorted helices (helicoids) [Fig. 1(b)] gradually increase their
periods with increasing magnetic fields. At a certain critical
field μ0 Hh = π 2 K0 /(16Ms ), helicoids transform into a set
of isolated 360◦ domain walls separating domains with the
FIG. 2. (Color online) Low anisotropy range of the magnetic
magnetization along the applied field direction.1
phase diagram in reduced variables for uniaxial anisotropy k =
One-dimensional modulations propagating along the in-
Ku /K0 and applied magnetic field h = H /HD for model (1) with
plane field also arise as regular solutions for the model an in-plane magnetic field [HD = 2K0 /Ms is the saturation field
presented in Eq. (1). In such modulations, the uniax- for a bulk cubic helimagnet with Ku = 0 and K0 is an “effective
ial anisotropy causes elliptical distortions of the “cones” stiffness” (Eq. (3))]. The regions of global stability for helical
[Fig. 1(c)]. The equilibrium parameters for these distorted and skyrmionic states are indicated by different colors. Solid lines
cones have no analytical solutions. However, for Ku  K0 designate the first-order transitions: H1 represents the transition

and H near the in-plane saturation field HC2 , we can use between helicoid and distorted cone, and HS1 and HS2 are transitions
Eqs. (2) to derive the following result: from a distorted skyrmion lattice to helicoid and distorted cones,
||
respectively. The dashed line indicates the critical field HC2 [Eq. (4)]

μ0 HC2 = (2K0 − Ku − Km ) Ms−1 , for the distorted cone phase. Hh = π K0 /(16μ0 Ms ) = 0.617HD is
2

LD the transition between the helicoid and the saturated states. Triangles
Km = μ0 Ms2 [1 − exp (−2π d/LD )] , (4) and squares show experimental values of critical fields HC2 ||
and Hα
4π d
for MnSi films of different thickness d.
where Km (d/LD ) is the stray field contribution and d is the
film thickness. phase diagram (Fig. 2) demonstrates that H ⊥ n̂ can induce
The equations that minimize the energy functional (1) with skyrmion phases in noncentrosymmetric magnets with an
an in-plane magnetic field also include solutions for two- easy-plane-type uniaxial anisotropy when Ku is above a
dimensional modulations in the form of hexagonal skyrmion threshold value. However, the helicoids propagating along
lattices with elliptical distortions caused by the uniaxial the axis of the uniaxial anisotropy retain thermodynamical
anisotropy [Fig. 1(d)]. We have studied the magnetic phase stability over a wide range of magnetic fields and transform by
diagram that results from model (1) for a bulk system with a first-order process, either into the distorted conical phase
periodic boundary conditions by using numerical methods, at a critical field H1 (Ku /K0 ) (for Ku < Ku∗ = 0.12K0 ) or
as in previous work.6,13 We have calculated stable solutions into the skyrmion lattice. In the magnetic phase diagram
rigorously of one- and two-dimensional modulated states (Fig. 2), critical lines HS1 (Ku /K0 ) and HS2 (Ku /K0 ) separate
for H ⊥ n̂ by direct energy minimization and testing their the skyrmionic states from the regions with one-dimensional
stability. By comparing the energies of these solutions, we modulations.
determine the thermodynamically stable states and the first- The results of the model apply to the strained MnSi films
order transition lines between them. The resulting magnetic investigated in this work. However, the neglected effects of

094429-3
E. A. KARHU et al. PHYSICAL REVIEW B 85, 094429 (2012)

cubic anisotropy and additional surface-induced anisotropies

can modify the details and quantitative evaluation of the (a)
magnetic phase diagram. For thin films in particular, the
isotropy of the chiral Dzyaloshinskii model could be impaired
and inhomogeneous competing surface anisotropies, e.g., with
easy-axis character, may become important.

III. MAGNETIC CHARACTERIZATION

We performed SQUID measurements of MnSi(111) thin
films in order to determine the energy scales that govern
the spin reorientation transition. The films were grown on
Si(111) by molecular beam epitaxy (MBE), and capped with
a protective 20-nm-thick amorphous Si cap, as described in (b)
Ref. 31. The M-H loops were measured in an applied magnetic
field oriented along the out-of-plane [111] and the in-plane
[110] directions at a temperature of T = 5 K. The linear
diamagnetic susceptibility of the substrate was determined
from the least-squares fits to M between μ0 H = 3 and 5 T and
subtracted from the data. Six representative curves are shown
in Fig. 3. The out-of-plane curves are qualitatively similar
to bulk for all film thicknesses.41 M increases approximately
⊥
linearly with field up to a field HC2 , and reaches a saturation
magnetization Ms = 163 ± 3 kA/m, which corresponds to
a moment μ = 0.416 ± 0.003 μB /Mn as compared to 0.39
μB /Mn in bulk. The knee in the M-H loops, which is (c)
determined from the minimum in d 2 M/dH 2 , is used as a
⊥
measure of HC2 and is plotted as a function of thickness in
Fig. 4.
For in-plane magnetic fields, the shape of the M-H
loops is qualitatively different than bulk. The first difference
is the existence of a remanent magnetization. Since the
magnetization lies in plane with the pitch vector along the
film normal, there are uncompensated moments when the film
thickness is a noninteger multiple of the helical wavelength.
Oscillations in the remanent magnetization as a function
of film thickness provide a measure of this wavelength.31
Second, we find first-order-like jumps in the magnetization
in Figs. 3(b) and 3(c), labeled Hα . For some thicknesses, there
is an additional kink in the magnetization below Hα , which is
labeled Hβ in Fig. 3(c). The values for the in-plane saturation FIG. 3. SQUID M-H curves of MnSi thin films with an in-plane
 ⊥
fields HC2 were determined in the same way as for HC2 , field applied along [110] (filled points) and out-of-plane field along
with the exception of the two lowest film thicknesses where [111] (open points) measured at T = 5 K. The film thicknesses are
rounding of the M-H curves prevented a clear determination d = 11.6 nm (a), 17.6 nm (b), and 26.7 nm (c).
of the minimum in the second derivative. In these two cases,
the knee was determined by extrapolation of the low-field M
values of the M-H curve up to Ms , and the second derivates A = 0.45 ± 0.01 meV nm2 as compared to the bulk value
were calculated from the decreasing-H branch of the hysteresis A  0.50 meV nm2 .42,43 This is consistent with the fact that
loops since they provided sharper transitions. TC is nearly constant between d = 10 and 40 nm. Given that
The effective stiffness of the conical phase can be derived Q = D/A, we then infer that D = 0.203 ± 0.005 meV nm
⊥  is larger than the bulk value, 0.18 meV nm. However,
from the measured saturation fields HC2 and HC2 :
Grigoriev et al. argue that the Dzyaloshinskii interaction is
 
Ms ⊥  2Km best characterized by D/(μa) since they find it to vary about
K0 = μ0 HC2 + 2μ0 HC2 − μ0 Ms + . (5) a value of 1.15 ± 0.01 meV/μB for both the Mn1−x Fex Si and
6 Ms
Fex Co1−x Si B20 alloys.44,45 We obtain a value D/(μa) =
Since we have found that the helical wavelength 2π/Q = 1.06 ± 0.03 for the films that is in good agreement with this
13.9 ± 0.1 nm is independent of d for a film thickness value. The 22% reduction in the wavelength of the helix in the
range of at least 7 to 40 nm,31 K0 provides a measure thin films can therefore be understood by the combination of a
of the spin-wave stiffness. Figure 5(b) shows that A is decrease in the spin-wave stiffness, and the increase in D due
approximately independent of thickness with a mean value to the increase in the magnetic moment of the film.

094429-4
CHIRAL MODULATIONS AND REORIENTATION EFFECTS . . . PHYSICAL REVIEW B 85, 094429 (2012)

(a)

(b)

FIG. 4. The in-plane saturation field (filled squares) and the out- (c)
of-plane saturation field (open circles) determined from M-H curves
measured at T = 5 K, like those shown in Fig. 3

The magnetocrystalline anisotropy, which is the second

energy scale important to understand the magnetic phase
⊥
diagram, is also determined from the saturation fields HC2

and HC2 ,
 
Ms ⊥  Km
Ku = μ0 HC2 − μ0 HC2 − μ0 Ms − , (6)
3 Ms
FIG. 6. (Color online) (a) In-plane elastic strain measured from
and is plotted in Fig. 5(a). The anisotropy is found to be plane-view TEM selected area diffraction patterns. (b) Out-of-
positive, which shows that it is of easy-plane type. plane strain determined from x-ray diffraction measurements of the
Given the epitaxially induced strain in the films, this MnSi(111) peak. In both sets of measurements, the Si substrate was
uniaxial anisotropy would be expected to originate, at least used as an internal calibration standard. (c) Ratio of the out-of-plane
in part, from magnetoelastic coupling. Furthermore a surface strain to twice the in-plane strain. The three samples indicated by
open red squares contain MnSi1.7 precipitates in excess of 20%. All
error bars are ±1σ .

anisotropy of the form Kus /d cannot fit the data in Fig. 5(a).
From the magnetoelastic free energy, expressed in terms of the
components of the directional cosines mi and the strains ij ,
defined with respect to the cubic axes of MnSi,46
 
wme = B1 xx m2x + yy m2y + zz m2z
+ B2 ( xy mx my + yz my mz + xz mx mz ), (7)
one can show that a trigonal distortion of the (111) planes pro-
duces an isotropic magnetoelastic stress energy contribution
B1 . However, the magnetoelastic shear stress contribution to
the free energy of the film is anisotropic:

wme = 12 B2,eff ( ⊥ −  )(m · n̂)2 . (8)

If we attribute the uniaxial magnetocrystalline anisotropy to
this contribution, we can obtain an estimate of an effective
magnetoelastic shear stress B2,eff . To this end, we measured the
in-plane strain  from TEM selected area diffraction patterns
and the out-of-plane strain ⊥ by using x-ray diffraction curves
from the MnSi(111) planes, both of which were performed
FIG. 5. The uniaxial magnetocrystalline anisotropy (a) and the at room temperature. The results are shown in Fig. 6. The
⊥ 
exchange stiffness (b), calculated from the HC2 and HC2 in Fig. 4, as strains are higher in the MBE samples compared to the solid-
a function of MnSi film thickness phase epitaxy (SPE) samples.30 We attribute the higher residual

094429-5
E. A. KARHU et al. PHYSICAL REVIEW B 85, 094429 (2012)

0.02

0.01

FIG. 8. (Color online) The in-plane fields Hα and Hβ , as shown

in Fig. 3, determined from the peak in dM/dH of the SQUID
measurements with the magnetic field along [110]. The filled (open)
circles show Hα calculated from the decreasing-H (increasing-H )
branch of the M-H curve measured at T = 5 K. The red squares
show Hβ for decreasing H (filled squares) and increasing H (open
squares).

symmetry of the crystal. In the case of a trigonal distortion of

FIG. 7. (Color online) (a) The volume strain as a function of a MnSi(111) film, nonlinear magnetoelastic theory53 gives a
three times the shear strain. The red line shows the polynomial fit contribution proportional to the shear strain, with coefficient
to the data over the range cover by the B2,eff data. (b) The effective D1 , as well as to the volume strain, parametrized by D2 :
magnetoelastic shear stress, where the red line is a fit to the data.
The fit in (a) and (b) gives B2 = 1.6 ± 1.0 MJ/m3 , D1 = −78 ± 20 B2,eff = B2 + D1 ( ⊥ −  ) + D2 ( ⊥ + 2  ), (9)
MJ/m3 , and D2 = −460 ± 120 MJ/m3 . All error bars represent ±1σ .
where B2 is the linear magnetoelastic shear stress. Since
the volume and shear strains are related through the elastic
strains in the films to the shorter annealing times during MBE constants, we fit the volume strain in Fig. 7(a) to c1 ( ⊥ −
growth. On the other hand, the ratio ⊥ /2  , which is related  ) + c2 ( ⊥ −  )2 over a range spanned by the data in
to the ratio of the bulk modulus to the shear modulus c44 ,30 Fig. 7(b) to enable a fit to the B2,eff with a single variable.
is the same for both MBE and SPE grown samples, which The parameters c1 and c2 together with the fit in Fig. 7(b)
implies that the elastic constants for both sets of samples are the give B2 = 1.6 ± 1.0 MJ/m3 , D1 = −78 ± 20 MJ/m3 , and
same, despite the presence of a small concentration of MnSi1.7 D2 = −460 ± 120 MJ/m3 .
precipitates in MBE grown samples. A large concentration To date, there are no values for B2 for MnSi provided in the
of precipitates, however, does affect the elastic properties of literature. However, the magnetoelastic stress B1 equals 2B2
the films. The strain in three of the samples that contained in an isotropic medium.54 By analysis of the magnetostriction
more than 20% MnSi1.7 , as determined from plane-view TEM data given in Ref. 55, and by correcting the scale in the plot of
images, has relaxed more than the other samples. The ⊥ /2  is the low-field magnetostriction data, as pointed out in Ref. 56,
high in these samples, which suggests the precipitates do soften we obtain B1  1 MJ/m3 for bulk MnSi at low temperature.
the c44 of the films when in large concentrations. Despite these Therefore, B2 = 1.6 may overestimate the magnetoelastic
differences, the magnetic moments of the MnSi films remain shear stress in the films. However, the reasonable fit shows
unaffected by the presence of the MnSi1.7 . that the uniaxial anisotropy can be explained in terms of a
We made estimates of the effective magnetoelastic shear magnetoelastic effect.
stress B2,eff ( ) = 2Ku /( ⊥ −  ), shown in Fig. 7(b). Here, The measurement of the uniaxial anisotropy provides a
we assume that the magnetocrystalline surface anisotropy is basis for attempting to understand the reorientation transition
negligible. This is a reasonable assumption since the surface in MnSi thin films. Unlike the out-of-plane measurements, the
anisotropies Kus are in the range (0.2 − 1) mJ/m2 for Fe, M-H curves measured with an in-plane magnetic field show
Co, and Ni,35,37,47 and would be expected to be considerably significant departures from the characteristic behavior of bulk
smaller for MnSi given the small Ms = 163 kA/m and that MnSi, as described by the in-plane transition fields Hα and Hβ
Ku scales as M 2 .48–50 shown in Fig. 3. We plot the values for these fields obtained
The magnetoelastic shear stress was found to vary with from peaks in dM/dH (see Fig. 8). The first-order-like step at
thickness, as found for other thin-film systems. Wedler et al. Hα only appears above a thickness d = 12 nm, which is close
measured the nonlinear B2,eff in Fe/MgO(100), and introduced in magnitude to 2π/Q. For film thicknesses close in value to
an  -dependent contribution to B2,eff .51 However, as pointed 4π/Q, a second transition Hβ appears. These two transitions
out in Ref. 52, this ansatz does not properly account for the are more difficult to explain with existing models. Plumer

094429-6
CHIRAL MODULATIONS AND REORIENTATION EFFECTS . . . PHYSICAL REVIEW B 85, 094429 (2012)

(a)

(b)

FIG. 10. (Color online) XRR measurements of the 20 nm

Si/26.7 nm MnSi/Si(111) sample shown in Fig. 9. The solid line
is a fit to the data, which gives the x-ray SLD shown in the inset.

direct measure of the pitch of the helix.31 Samples were field

(c) cooled with μ0 H = 0.8 T applied along the [110] direction
to a temperature of 6 K, and then four sets of measurements
were performed in fields of 1 mT, 0.3 T, 0.5 T, 0.6 T, and
0.8 T. In this more conventional geometry, the scattering cross
sections R(++) and R(−−) are sensitive to both chemical
and magnetic structure. Evidence for helical order would
normally appear as a Bragg-type peak in the R(+−) or R(−+)
channels, which are sensitive to the in-plane component of
(d) the magnetization that is perpendicular to the field direction.
However, the presence of left- and right-handed magnetic
domains largely cancel one another, evidenced by the small
spin-flip (SF) signal, as can be seen in Fig. 9.
In order to extract the magnetic scattering length, we infer
the nuclear scattering length density (SLD) profile from the x-
ray SLD profile determined from a fit to the x-ray reflectometry
(XRR) data in Fig. 10. The x-ray SLD is uniform across the
thickness of the silicide film and agrees well with the value
expected for MnSi, which indicates that the film is single
phase without any secondary phase, in agreement with TEM
FIG. 9. (Color online) PNR with ±1σ error bars of 20 nm measurements. We use the nuclear SLD to then fit R(++) and
Si/26.7 nm MnSi/Si(111) measured at T = 6 K. The magnetic field R(−−) in Fig. 9 and obtain the magnetic SLD profile displayed
applied along the in-plane [110] direction is (a) 1 mT, (b) 0.3 T, (c) in each of the insets. A uniform magnetic SLD profile that
0.5 T, and (d) 0.8 T. The fits to the PNR data, shown by the solid corresponds to a moment of m = 0.38μB /Mn fits the 0.8-T
lines, yield the magnetic moment profiles given in each of the insets. data well. A conical phase with Q||[110] is consistent with such
an SLD profile since the component of the magnetization in
and Walker modeled the reorientation of the helical order in the field direction in this orientation is constant throughout the
bulk MnSi where small departures from linear M-H curves film. The cone angle inferred from PNR measurements, θ =
are observed below the reorientation transition HC1  0.1 T.17 cos−1 (m/msat ) = 25◦ , is in reasonable agreement with θ =

However, their solutions to a Landau-type free energy are not cos−1 (H /HC2 ) = 28◦ obtained from SQUID measurements.
⊥
able to account for the large values of Hα relative the HC2 . In an applied field of 1 mT oriented in the in-plane [110]
direction, the pitch vector Q is pointing in the out-of-plane
[111] direction. A sinusoidal magnetic SLD profile with a
IV. MAGNETIC DEPTH PROFILES
period of 2π/Q = 13.9 ± 0.01 nm provides a good fit to
In order to better understand the magnetic structure during the data. This is entirely consistent with the out-of-plane
reorientation, we performed a series of PNR measurements measurements.31 The small magnetic moment of 0.20 μB /Mn
on a d = 26.7-nm-thick film with the aim of determining that is obtained from the fit is accounted for by the presence
a depth profile of the magnetization for a range of applied of domains. The SLD profile for the 0.3 T field in Fig. 9(b)
magnetic fields. The results of the measurements performed shows distortions to the sinusoidal SLD, which become greater
on the NG-1 reflectometer at the NIST Center for Neutron at 0.5 T. The evolution and reorientation of the helical order,
research are shown in Fig. 9. We present measurements however, shows clear departures from bulk behavior. The PNR
with the magnetic field and neutron polarization along the data provide an in-plane average of the component of the
in-plane [110] direction in Fig. 9, in contrast to our previous magnetization in the direction of the magnetic field. The fact
measurements where an out-of-plane orientation enabled a that the depth dependence of the magnetization is described

094429-7
E. A. KARHU et al. PHYSICAL REVIEW B 85, 094429 (2012)

60 250 by the layered model rules out a second-order-like transition

Energy
(a) E0 1 -1 - E111 where the pitch vector rotates in the direction of the magnetic
200

magnetic anisotropy energy ( kJ / m )

50 field. Furthermore, the magnetization can be seen to be larger

3
150 in the field direction than in the antiparallel direction. This
difference contributes to the net magnetization, and is in
ΔE ( meV / unit cell )

40 100
agreement with M(H ) plotted in Fig. 3(c).
50
30
0
V. THEORETICAL CONSIDERATIONS AND DISCUSSION
20 -50
In the following two sections, the experimental information
10
-100 is discussed using additional insights from theoretical calcula-
-150
tions. The first section reports ab initio calculations to assess
0
the induced uniaxial anisotropy in strained cubic helimagnets.
ε_|_ ε|| ε_|_ - ε|| The results corroborate the basic magnetic model of a chiral
0.05 10 x (ε_|_ + 2 ε|| ) magnet with a homogeneous easy-plane uniaxial anisotropy
strains

0 [Eq. (1)] to describe the properties of the epitaxial MnSi

films. In the second part, the experimental observations on the
-0.05
(b) field-driven evolution of the modulated states are discussed.

(c)
1.3 -0.123 A. Comparison with electronic-structure calculations
In order to ascertain the character and origin of the
1.2 uniaxial anisotropy in strained epitaxial MnSi films, we have
spin moment ( μB / f.u. )

-0.124
performed electronic-structure calculations. The calculations
1.1 for rhombohedrally distorted MnSi provide theoretical esti-
mates for the magnetoelastic coupling coefficient B2 and the
1
induced uniaxial anisotropy, which is a bulk effect present
total
-0.125 in homogeneously strained films. Density functional theory
0.9
Mn 1a calculations of the electronic structure are useful to understand
0.8
Mn 3b and evaluate the properties of d-electron-based metallic mag-
Si 1a
-0.126
netic systems.57 In particular, trends in magnetic properties
Si 3b
of metallic nanostructures can be understood from such
calculations, even though the exact quantitative determination
of small spin-orbit effects such as magnetic anisotropies or
magnetostriction coefficients still poses severe problems (see
orbital moment ( 0.01 μB / Mn )

3 Refs. 50, 58 and the references therein). In the present MnSi

films, the lattice stretches are sizable so that the induced
M || (111): Mn 1a
Mn 3a uniaxial anisotropy can be estimated from electronic-structure
M || (0 1 -1 ): Mn 1a
Mn 3b calculations, although the intrinsic cubic anisotropies in this
Mn 3b material that derive from higher-order spin-orbit coupling
2.5 Mn 3b
effects are too small to be calculated reliably by DFT methods.
To address the limit of relatively thick films, we neglect the
proper surface-induced anisotropies.
2 The isotropic in-plane strain in the (111)-oriented films
(d)
transforms the cubic lattice from the structure described
86 87 88 89 90 91 92 93 by space group P 21 3 (#198) into a rhombohedrally dis-
rhombohedral angle Ξ torted structure with symmetry R3 (#146). We performed
calculations using a full-potential local orbital method, as
FIG. 11. (Color online) Results from density functional theory
implemented in the FPLO code59 with the generalized gradient
calculations for strained MnSi: (a) the dependence of the energy
approximation (GGA) parametrized by Perdew, Burke, and
increase
E on the rhombohedral angle  with fixed lattice constant
Ernzerhof.60 The variation of total energy with respect to the
a = 0.4556 nm and the calculated anisotropy for the magnetization
along the rhombohedral axis, parallel to [111], and perpendicular
rhombohedral angle  is evaluated from a full relativistic cal-
to the axis along [110]. (b) Dependence of strain components on culation with the quantization axis (magnetization direction) in
 for the deformation with constant lattice parameter. Note the various directions.61 In the calculations, the lattice constant has
factor 10 for the volume strain ⊥ + 2  . (c) Corresponding evolution been kept at a fixed value of a = 0.4556 nm, as was determined
of spin moments for different sites in the rhombohedral cell. (d) previously from calculations by optimizing the structure of the
Comparison of orbital magnetic moments on the different Mn sites cubic cell, including the cell parameters. The calculated results
for the magnetization parallel and perpendicular to the rhombohedral for the cubic MnSi structure are in good agreement with earlier
axis. In the latter case, the equivalence of the 3b sites is lost and the DFT studies.62 In the rhombohedrally distorted cell, the site
orbital magnetic moments split. positions 4a for the Mn and Si in the cubic structure are split

094429-8
CHIRAL MODULATIONS AND REORIENTATION EFFECTS . . . PHYSICAL REVIEW B 85, 094429 (2012)

into two crystallographic inequivalent positions 1a and 3b.

For each rhombohedral angle  = π/2 − 2( ⊥ −  )/3, the
lattice cell of the distorted bulk MnSi was first optimized
by a scalar relativistic calculation. In these calculations,
the site positions of all atoms have been relaxed so that
forces on the atom were less than 10−3 eV/Å. In these
optimized structures, full relativistic calculations are used
to estimate the magnetocrystalline anisotropy. The energy
was converged to values better than 4 × 10−7 eV/unit cell FIG. 12. (Color online) Diagram of a skewed conical phase in
and the density to better than 10−6 . It was checked that the an in-plane magnetic field, where n̂ is the film normal and x̂ is the
reciprocal space integration with a k-mesh subdivision of direction of the applied magnetic field H . The wavefronts of this
22 × 22 × 22 is sufficient to obtain satisfactory convergency phase remain parallel to the surface, while the axis of the cone ĉ cants
of the fully relativistic total energies. As expected, the method in the direction of the field by an angle φ.
is unable to resolve the anisotropy for different magnetization
directions in the plane, which was checked by performing
calculations with
√ the quantization axis along the [112], [110], S DFT = 1.00μB /Mn. As a result of the enhanced spin moment
and [112] + 6[110] directions. Thus, the sixfold basal plane in the DFT calculations that neglect quantum-dynamical spin
anisotropy in the isotropically strained MnSi(111) should be fluctuations, the MAE is also overestimated. In order to
below a few μeV/f.u. Therefore, we only compare results for provide a more realistic estimate on the uniaxial anisotropy
the quantization along the crystallographic [111] axis with energy Ea from the calculations, we assume that the MAE
those with the axis along the in-plane direction [110] in the scales with the magnetization. A renormalization considering
following discussion. spin fluctuations should behave as Ea =MAE (Sexp /S DFT )n .
Figure 11(a) shows the dependence of the total energy
E The effects of thermal and quantum dynamics at the zero
on  with respect to the minimum energy for the cubic struc- point both cause a reduction of the magnetization, which
ture  = 90◦ and the magnetocrystalline anisotropy energy essentially influences the macroscopic magnetic properties in
(MAE) evaluated as the difference of the total energies for the the same way. Therefore, we can rely on recent results on the
two different quantization axes, MAE= Etot 111
− Etot
01−1
. The temperature dependence of the magnetic uniaxial anisotropies
figure shows that the isotropic in-plane expansion induces an in band ferromagnets, which find a scaling of MAE with the
easy-plane magnetic anisotropy, while compressively strained magnetization and an exponent n = 2.48–50 Thus, a simple
films would acquire an out-of-plane easy-axis anisotropy. rescaling of the MAE by a factor (0.42/1.00)2 = 0.18 is
Figure 11(b) shows the dependence of the different strain appropriate to estimate a theoretical magnetoelastic coupling
components on rhombohedral angle , which are almost coefficient of bulk MnSi of B2 = −0.37 MJ/m3 and a
linear over the relevant range of angles, while the volume nonlinear coefficient of D1 = −3.2 MJ/m3 . In view of the
strain ⊥ + 2  is almost zero for the deformation mode used large uncertainties regarding the theoretical treatment of the
in the calculations. The magnetoelastic coupling coefficient spin-fluctuation effect, there is still a reasonable agreement of
B2DFT = −2.1 ± 0.2 MJ/m3 is calculated from the slope of this value with the experimental data shown in Fig. 7.
MAE( = 90◦ ). Similarly, the leading nonlinear magnetoe-
lastic coupling coefficient D1DFT = −17.9 ± 1.5 MJ/m3 can
be estimated from a nonlinear fit. The total magnetic moment B. Field-driven modulated structures
remains constant under strain. However, there is a marked
The profiles from the PNR in applied magnetic fields
redistribution of spin moment between the different Mn and
(Fig. 9) are unusual. If one assumes a one-dimensional chiral
Si sites, as shown in Fig. 11(b). A sizable induced spin
modulated state to explain these profiles, then one is led to
moment on Si is found, which undergoes minor changes in
consider a skewed conical phase with anharmonic distortions.
the strained lattice. The redistribution of spin moments on Mn
The lth harmonic of the magnetic moment for such a structure
is accompanied by a complex dependency of the corresponding
as a function of depth z,
orbital magnetic moments on the different Mn sites that also
depend on the quantization direction [Fig. 11(c)]. These orbital
moments are associated with the magnetic anisotropy energy. μl (z) = μl cos(θ )ĉ + μl sin(θ )[cos(lQ · n̂z)â
The observed dependence indicates an intricate mechanism + sin(lQ · n̂z)b̂], (10)
due to the strain that makes all the MnSi 4a sites inequivalent
for a magnetization perpendicular to the rhombohedral axis.
The calculated orbital moments on Si are very small. is expressed in terms of an orthonormal coordinate system
MnSi is described as a weak band ferromagnet subject
to strong spin fluctuations, which strongly reduce the static â = cos(φ)x̂ − sin(φ)n̂, b̂ = ŷ, ĉ = sin(φ)x̂ + cos(φ)n̂.
spin polarization in the low-temperature magnetically or-
(11)
dered state.63 Thus, the experimentally determined magnetic
moment that is 0.39 μB /Mn in bulk MnSi is enhanced
to about Sexp = 0.42μB /Mn in the epitaxial films with a The coordinate axes are written in terms of φ, the angle of
slight volume expansion. As in previous DFT calculations the cone axis with respect to n̂, and in terms of the unit
on MnSi with B20-structure,62 we find a net spin moment of vectors x̂[110], ŷ[112], and n̂[111], as shown in Fig. 12.

094429-9
E. A. KARHU et al. PHYSICAL REVIEW B 85, 094429 (2012)

The skewed cone cants in the direction of the field, while the one-dimensional lattice at the nucleation transition Hh . This
wavefronts of the spin density wave remain perpendicular to means that the helicoid and the competing conical state,
n̂, in contrast to a conical phase where the wavefronts are propagating transverse to it, have wide ranges of coexistence
perpendicular to the ĉ direction. For the case of μ0 H = 0.5 in applied fields, and during the transformation process, it
T, a first harmonic μ1 = 0.42μB /Mn and second harmonic is impossible to destroy domains with kink-like structures
μ2 = 0.12μB /Mn with θ = 56◦ and φ = 40◦ are able to even in high applied fields. The characteristic field Hα in the
accurately reproduce the magnetic SLD profile. Field-induced experiments may be explainable by such a process, which may
second harmonics have been observed in bulk MnSi,43,64 and involve skewed helices instead of flat helicoids. Most likely,
were also predicted by Plumer and Walker.65 However, such a transitions from a helicoid into free kinks and a transformation
skewed conical phase has not been seen previously. into a conical phase at higher fields takes place in many of
The one-dimensional skewed conical phase rationalizes the the films. The first process is exactly the type of nucleation
experimental observations and is one of the possible structures transition of helicoidal kinks, predicted by Dzyaloshinskii in
to explain the unusual magnetization processes. It is clear from his seminal work.1 The second process seems to proceed by a
an unrestricted search for one-dimensional static magnetic domain process at higher fields.
states that skewed conical helices do not exist as solutions The existence of a skyrmion lattice is a possible explana-
in the basic magnetic model with a homogeneous easy-plane tion for the unusual magnetization process and the unusual
uniaxial anisotropy [Eq. (1)]. However, it is equally clear magnetic depth profiles from the PNR data. However, the
that epitaxial thin films may possess a number of additional magnetic anisotropy for the film with thickness d = 26.7 nm is
interactions stemming from the surfaces, which could even- found to be too low in the magnetic phase diagram (data point
tually stabilize such distorted modulated structures. There is #2 in Fig. 2) in comparison to the theoretical threshold for
ample choice of possible competing and inhomogeneously stable skyrmion phases in the basic model, so that additional
distributed anisotropies, e.g., the intrinsic cubic anisotropies, magnetic couplings would be required to stabilize such states.
that may cause anharmonicities, and competing easy-axis On the other hand, the demagnetizing field makes it difficult
surface anisotropies. Moreover, in the strained films, the to transform a helicoidal or a conical structure into a phase
isotropy of the chiral DM interaction may be lost, so that with skyrmion axes running along the direction of an applied
the in-plane and out-of-plane helical modulations become in-plane magnetic field.
energetically different. Therefore, it is possible that in all experiments done so
The phase diagram, Fig. 2, enumerates the basic confor- far, the skyrmion lattices are kinetically suppressed, although
mations of the possible magnetic modulations in an almost the estimated anisotropies of the MnSi films cover a range
isotropic chiral magnet with an easy-plane anisotropy: the where the basic model displays thermodynamically stable
flat helicoidal structure in the field plane, the conical helices, skyrmions in applied fields, as shown in Fig. 2. There are
and skyrmion lattices. It also shows that there is a close clearly additional effects that may confound the formation of
competition between these states around a triple point, where such a state and favor either the helicoidal order or other types
these three structures can coexist. All transformations between of spiral order, even in equilibrium. A surface anisotropy may
these states are expected to be first-order processes. Moreover, also be present with an easy axis that may distort any of these
this indicates the possibility of severe history dependence modulated phases. Moreover, it is possible that the isotropy
and hysteresis in the magnetization processes. Weaker in- of the Lifshitz invariants m · ∇ × m is broken by the strain in
teractions, e.g., intrinsic cubic anisotropies2 and additional the films since the DM interaction and the uniaxial anisotropy
surface effects, are not considered in the phase diagram. These may have the same origin of a pairwise anisotropic exchange
couplings could lead to modifications or distortions of the basic in a band ferromagnet.49
modulated structures. Quantitative agreement with experiment Based on the experimental observations, it is not possible to
is expected for the saturation of the conical phase at field HC2 conclusively decide the question of which magnetic structures
because this transition is continuous and starts out from the are responsible for the anomalous field-driven evolution in the
unique saturated state. This field is used to determine the experiments. It is clear that different MnSi films may even
induced anisotropy Ku of the films in Fig. 2. For all other show different sequences of states. The fact that some films
transition lines and transformation processes between these display two transitions at lower fields is suggestive of different
different states, a quantitative agreement is not to be expected magnetization processes. Be that as it may, a comparison of
as their energies are very close and relative thermodynamical the magnetic profiles from the PNR data with depth profiles
stability may be shifted by minor additional effects. of the layer-averaged component of the magnetization in field
The thicker films that show a lower easy-plane anisotropy direction Mx  for different structures allows some conclusions
display a hysteretic transformation process at the characteristic on the possible magnetic modulations. Figure 13 shows
field Hα , which is just below the field Hh , for transformations profiles from numerical solutions to Eq. (1) that correspond
from flat helicoids into the saturated state. It is important to the 26.7-nm-thick film in a 0.5-T field reported in Fig. 9.
to understand the peculiar nature of this transformation, The skewed anharmonic conical helix yields a good fit to
which does not destroy the helical kinks (360◦ rotations) the magnetic depth profile. In particular, the modulation falls
of the magnetization in an infinite system. Rather, the short of the saturation magnetization in the negative direction.
transformation retains the localized core of these kinks In contrast, Mx (z) of the flat helicoids covers the full
while their tails are stretched to infinity, so that the period of range of the magnetization, from +Ms to Ms , albeit with a
the helicoid diverges. The reverse transformation requires the strong anharmonicity. The distorted conical helix has constant
nucleation of single kinks, which then assemble into a periodic value of magnetization in field direction. Finally, the profiles

094429-10
CHIRAL MODULATIONS AND REORIENTATION EFFECTS . . . PHYSICAL REVIEW B 85, 094429 (2012)

1 VI. CONCLUSION
average magnetization <Mx / Ms> (z)

The results of this work establish epitaxial MnSi films as a

chiral cubic helimagnetic system where complex magnetiza-
0.5
tion processes can be studied under the influence of induced
magnetic anisotropies. The major part of this work shows
that the in-plane tensile strain in MnSi(111) films produces
0
an easy-plane uniaxial anisotropy, which can be explained
by a homogeneous magnetoelastic effect. Both the theoretical
results within a simplified model and the experiments reveal a
-0.5 exp. profile (PNR)
skew conical helix
number of modulated states that are metastable or do not exist
conical helix in bulk MnSi, which is a nearly isotropic cubic helimagnet.
periodic helicoid
periodic Skyrmion lattice Depending on the value of the uniaxial anisotropy, different
-1
0 0.5 1 1.5 2
types of magnetization processes are found. Our experimental
reduced depth z / LD data give clear indications for such processes in the MnSi films.
One of the most interesting transformation processes that
FIG. 13. (Color online) Experimental magnetization profile as may be observed in this system is the creation of solitonic
normalized layer-averaged magnetization in field direction Mx vs states,13 such as the nucleation of free kinks in the helicoidal
position in the layer (normalized by wavelength of helix modulation) phase in a transverse magnetic field, or alternatively the
for the 26.7-nm-thick film in a 0.5-T field compared to numerical so- formation of skyrmion lattices. These structures have their
lutions for different modulated states. Data and solutions correspond own mesoscale lengths, fixed by the competition between
to reduced anisotropy and field (k = 0.04, h = 0.50). Dzyaloshinskii-Moriya and direct exchange. This poses new
problems for their stability and transformational behavior in
confining geometries of magnetic nanostructures. Magnetic
microscopic imaging now will play a major role in order to
of the metastable and periodic skyrmion lattice solutions discriminate between the different chirally modulated states
have a similar character as the experimental data. For the and to unravel the magnetic phase diagram relying on the
observations, we have then at least two possible explanations: intricate competition between these states. We also show that
(i) If one-dimensional modulations occur as intermediate states quantitative models for the magnetic behavior of these films
in the MnSi films, they should have a complex skewed conical will require very detailed determination of their magnetic
structure. Additional magnetic couplings such as competing properties, including higher-order anisotropies and surface
anisotropies would be required to explain such unusual helices effects.
in MnSi. (ii) A transverse distorted skyrmion phase does
exist as a field-driven phase in easy-plane chiral magnets, ACKNOWLEDGMENT
as shown by the theoretical magnetic phase diagram. This
explanation also requires additional magnetic couplings to This work was supported by the NSERC, the Canada
achieve a quantitative agreement between the model and the Research Chairs Program, and DFG through Project No. RO
experimental observations. 2238/9-1.

* 10
theodore.monchesky@dal.ca S. Heinze, K. von Bergmann, M. Menzel, J. Brede, A. Kubetzka,
1
I. E. Dzyaloshinskii, Zh. Eksp. Teor. Fiz. 47, 992 (1964) [Sov. Phys. R. Wiesendanger, G. Bihlmayer, and S. Blügel, Nat. Phys. 7, 713
JETP 20, 665 (1965)]. (2011).
2 11
P. Bak and M. H. Jensen, J. Phys. C: Solid State 13, L881 (1980). M. Bode, M. Heide, K. von Bergmann, P. Ferriani, S. Heinze,
3
B. Lebech, J. Bernhard, and T. Freltoft, J. Phys.: Condens. Matter G. Bihlmayer, A. Kubetzka, O. Pietzsch, S. Blügel, and
1, 6105 (1989). R. Wiesendanger, Nature (London) 447, 190 (2007).
4 12
A. N. Bogdanov and D. A. Yablonskii, Zh. Eksp. Teor. Fiz. 95, 178 A. Tonomura, X. Yu, K. Yanagisawa, T. Matsuda, Y. Onose,
(1989) [Sov. Phys.–JETP 68, 101 (1989)]. N. Kanazawa, H. S. Park, and Y. Tokura, Nano Lett. 12, 1673
5
A. N. Bogdanov and U. K. Rößler, Phys. Rev. Lett. 87, 037203 (2012).
13
(2001). U. K. Rößler, A. A. Leonov, and A. N. Bogdanov, J. Phys.: Conf.
6
U. K. Rößler, A. N. Bogdanov, and C. Pfleiderer, Nature (London) Ser. 303, 012105 (2011).
14
442, 797 (2006). J. Fukuda and S. Zumer, Nat. Commun. 2, 246 (2011).
7 15
A. N. Bogdanov, U. K. Rößler, and C. Pfleiderer, Phys. B N. S. Kiselev, A. N. Bogdanov, R. Schäfer, and U. K. Rößler,
(Amsterdam) 359, 1162 (2005). J. Phys. D: Appl. Phys. 44, 392001 (2011).
8 16
X. Z. Yu, Y. Onose, N. Kanazawa, J. H. Park, J. H. Han, Y. Matsui, Y. Ishikawa, K. Tajima, D. Bloch, and M. Roth, Solid State
N. Nagaosa, and Y. Tokura, Nature (London) 465, 901 (2010). Commun. 19, 525 (1976).
9 17
X. Z. Yu, N. Kanazawa, Y. Onose, K. Kimoto, W. Z. Zhang, S. M. L. Plumer and M. B. Walker, J. Phys. C: Solid State 14, 4689
Ishiwata, Y. Matsui, and Y. Tokura, Nat. Mater. 10, 106 (2011). (1981).

094429-11
E. A. KARHU et al. PHYSICAL REVIEW B 85, 094429 (2012)

18 39
A. I. Okorokov, S. V. Grigoriev, Y. O. Chetverikov, S. V. Maleyev, A. N. Bogdanov, U. K. Rößler, and K. H. Müller, J. Magn. Magn.
R. Georgii, P. Böni, D. Lamago, H. Eckerlebe, and K. Pranzas, Mater. 238, 155 (2002).
40
Phys. B 356, 259 (2005). A. Thiaville and A. Fert, J. Magn. Magn. Mater. 113, 161 (1992).
19 41
S. Kusaka, K. Yamamoto, T. Komatsubara, and Y. Ishikawa, Solid D. Bloch, J. Voiron, V. Jaccarino, and J. H. Wernick, Phys. Lett. A
State Commun. 20, 925 (1976). 51, 259 (1975).
20 42
K. Kadowaki, K. Okuda, and M. Date, J. Phys. Soc. Jpn. 51, 2433 Y. Ishikawa, G. Shirane, J. A. Tarvin, and M. Kohgi, Phys. Rev. B
(1982). 16, 4956 (1977).
21 43
B. Lebech, P. Harris, J. S. Pedersen, K. Mortensen, C. I. Gregory, S. V. Grigoriev, S. V. Maleyev, A. I. Okorokov, Y. O. Chetverikov,
N. R. Bernhoeft, M. Jermy, and S. A. Brown, J. Magn. Magn. Mater. P. Böni, R. Georgii, D. Lamago, H. Eckerlebe, and K. Pranzas,
140, 119 (1995). Phys. Rev. B 74, 214414 (2006).
22 44
S. V. Grigoriev, S. V. Maleyev, A. I. Okorokov, Y. O. Chetverikov, S. V. Grigoriev, V. A. Dyadkin, E. V. Moskvin, D. Lamago, T. Wolf,
and H. Eckerlebe, Phys. Rev. B 73, 224440 (2006). H. Eckerlebe, and S. V. Maleyev, Phys. Rev. B 79, 144417 (2009).
23 45
S. M. Stishov, A. E. Petrova, S. Khasanov, G. K. Panova, A. A. S. V. Grigoriev, S. V. Maleyev, V. A. Dyadkin, D. Menzel,
Shikov, J. C. Lashley, D. Wu, and T. A. Lograsso, J. Exp. Theor. J. Schoenes, and H. Eckerlebe, Phys. Rev. B 76, 092407 (2007).
46
Phys. 106, 888 (2008). C. Kittel, Rev. Mod. Phys. 21, 541 (1949).
24 47
S. Mühlbauer, B. Binz, F. Jonietz, C. Pfleiderer, A. Rosch, U. Gradmann, J. Magn. Magn. Mater. 54-57, 733 (1986).
48
A. Neubauer, R. Georgii, and P. Böni, Science 323, 915 J. B. Staunton, S. Ostanin, S. S. A. Razee, B. L. Gyorffy,
(2009). L. Szunyogh, B. Ginatempo, and E. Bruno, Phys. Rev. Lett. 93,
25
C. Pappas, E. Lelièvre-Berna, P. Falus, P. M. Bentley, E. Moskvin, S. 257204 (2004).
49
Grigoriev, P. Fouquet, and B. Farago, Phys. Rev. Lett. 102, 197202 O. N. Mryasov, U. Nowak, K. Y. Guslienko, and R. W. Chantrell,
(2009). Europhys. Lett. 69, 805 (2005).
26 50
H. Wilhelm, M. Baenitz, M. Schmidt, U. K. Rößler, A. A. Leonov, J. B. Staunton, L. Szunyogh, A. Buruzs, B. L. Gyorffy, S. Ostanin,
and A. N. Bogdanov, Phys. Rev. Lett. 107, 127203 (2011). and L. Udvardi, Phys. Rev. B 74, 144411 (2006).
27 51
H. Wilhelm, M. Baenitz, M. Schmidt, C. Naylor, R. Lortz, U. K. G. Wedler, J. Walz, A. Greuer, and R. Koch, Surf. Sci. 454–456,
Rößler, A. A. Leonov, and A. N. Bogdanov, J. Phys.: Condens. 896 (2000).
52
Matter (to be published). M. Komelj and M. Fähnle, Phys. Rev. B 65, 092403 (2002).
28 53
A. A. Leonov, A. N. Bogdanov, and U. K. Rößler, e-print E. du Trémolet de Lacheisserie, Magnetostriction: Theory and
arXiv:1001.1292. Applications of Magnetoelasticity (CRC Press, Boca Raton, 1993).
29 54
A. B. Butenko, A. A. Leonov, U. K. Rößler, and A. N. Bogdanov, S. V. Maleyev, J. Phys.: Condens. Matter 21, 146001 (2009).
55
Phys. Rev. B 82, 052403 (2010). E. Fawcett, J. P. Maita, and J. H. Wernick, Int. J. Magn.1, 29 (1970).
30 56
E. Karhu, S. Kahwaji, T. L. Monchesky, C. Parsons, M. D. E. Franus-Muir, M. L. Plumer, and E. Fawcett, J. Phys. C: Solid
Robertson, and C. Maunders, Phys. Rev. B 82, 184417 (2010). State 17, 1107 (1984).
31 57
E. A. Karhu, S. Kahwaji, M. D. Robertson, H. Fritzsche, B. J. J. B. Staunton, Rep. Prog. Phys. 57, 1289 (1994).
58
Kirby, C. F. Majkrzak, and T. L. Monchesky, Phys. Rev. B 84, T. Burkert, O. Eriksson, P. James, S. I. Simak, B. Johansson, and
060404 (2011). L. Nordström, Phys. Rev. B 69, 104426 (2004).
32 59
E. Magnano, F. Bondino, C. Cepek, F. Parmigiani, and M. C. K. Koepernik and H. Eschrig, Phys. Rev. B 59, 1743 (1999).
60
Mozzati, Appl. Phys. Lett. 96, 152503 (2010). J. P. Perdew, K. Burke, and M. Ernzerhof, Phys. Rev. Lett. 77, 3865
33
K. Schwinge, C. Müller, A. Mogilatenko, J. J. Paggel, and (1996).
61
P. Fumagalli, J. Appl. Phys. 97 (2005). The full relativistic option of the FPLO code, http://www.fplo.de,
34
A. Hubert and R. Schäfer, Magnetic Domains: The Analysis of relies on an all-electron solution of the Dirac equation and
Magnetic Microstructures (Springer, Berlin, 1998). incorporates spin-orbit coupling to all orders.
35 62
P. Bruno and J.-P. Renard, Appl. Phys. A 49, 499 (1989). T. Jeong and W. E. Pickett, Phys. Rev. B 70, 075114 (2004).
36 63
H. Fritzsche, J. Kohlhepp, and U. Gradmann, Phys. Rev. B 51, T. Moriya, Spin Fluctuations in Itinerant Electron Magnetism
15933 (1995). (Springer-Verlag, Berlin, 1985).
37 64
B. Heinrich and J. F. Cochran, Adv. Phys. 42, 523 (1993). Y. Ishikawa, T. Komatsubara, and D. Bloch, Phys. B & C
38
W. J. M. de Jonge, P. J. H. Bloemen, and F. J. A. den Broeder, (Amsterdam) 86, 401 (1977).
65
in Ultrathin Magnetic Structures, edited by J. A. C. Bland and M. L. Plumer and M. B. Walker, J. Phys. C: Solid State 15, 7181
B. Heinrich (Springer, Berlin, 1994). (1982).

094429-12