betaFeedback

WorksPhysics and Astronomy

Resolving the local distortions of Ising-like moments in magnetoelectric Ho-doped langasite

Abstract

The magnetic properties of Ho-doped langasites (La:Ho)3_3Ga5_5SiO14_{14} are dominated by the Ising-like magnetic moments of the Ho3+^{3+} ions. In their saturated regime, the induced magnetic state breaks both time and space inversion symmetries, leading to a novel linear magnetoelectric effect. However, due to distortions induced by a shared Ga/Si occupancy of the 2d sites, resolving the microscopic nature of the magnetic configuration remains a difficult task. Here we combine polarized neutron diffraction and angular dependent magnetization experiments to determine the local distortions of the Ho3+^{3+} magnetic moments in doped langasites (La1−x_{1-x}Hox_x)3_3Ga5_5SiO14_{14} with x≈0.015x \approx 0.015 and x≈0.045x \approx 0.045. We propose a model for a field-induced magnetic configuration with arbitrary orientations of the local Ising axis of Ho3+^{3+} in distorted positions. The operations of broken local C2C_2 symmetry and rotations around the trigonal C3C_3 axis connect different sites, restoring the global P321 symmetry of the crystal and simplifying the description of the magnetic properties. The superposition of two distorted Ho3+^{3+} positions connected by C2C_2 symmetry determines the local magnetic susceptibility tensor, which no longer appears Ising-like at low fields.

Introduction

Compounds with the langasite structure (La3Ga5SiO14)(\mathrm{La}_3\mathrm{Ga}_5\mathrm{SiO}_{14}) [1–4] have been extensively studied for some years now and are primarily well-known for their strong piezoelectric and nonlinear optical properties [5–7]. The existence of orthogonal third and second order axes in the P321P321 space group leads to a number of unusual functional properties. For example, uniaxial pressure along the second-order axis breaks the trigonal symmetry and induces an electric polarization [6]. Significant interstitial spaces between La3+\mathrm{La}^{3+} dodecahedra and Ga3+\mathrm{Ga}^{3+} octahedra make them highly susceptible to stress [7] appreciably increasing the mechanical coupling, which is higher than that of quartz [8]. These properties, combined with structural stability over a wide range of temperatures, make langasite particularly attractive for use in piezoelectric sensors [9].

As shown in Figure 1, the unit cell of langasites consists of two alternating layers. The first layer consists of La3+\mathrm{La}^{3+} dodecahedra at the 3e3e positions with C2C_2 symmetry and of Ga3+\mathrm{Ga}^{3+} octahedra at the 1a1a positions with D3D_3 symmetry. The second layer consists of Ga3+\mathrm{Ga}^{3+} and Si3+/Ga3+\mathrm{Si}^{3+}/\mathrm{Ga}^{3+} tetrahedra occupying the 3f3f positions with C2C_2 symmetry and 2d2d positions with C3C_3 symmetry, respectively. Oxygen ions (O2−\mathrm{O}^{2-}) are located at the vertices of these polyhedra. Magnetic ions in the langasite lattice can lead to non-trivial magnetic structures along with unconventional magnetoelectric properties. For example, the well-studied iron langasites (such as Ba3NbFe3Si2O14\mathrm{Ba_3NbFe_3Si_2O_{14}}) order antiferromagnetically at TN∼27T_{\mathrm{N}} \sim27 K into a triangular spiral magnetic structure with a double chirality [10, 11]. In external magnetic fields they exhibit exotic chiral-dependent magnetoelectric coupling [12, 13, 14]. However, the presence of such magnetic structures complicates the study of the microscopic mechanisms underlying the magnetic and magnetoelectric properties of langasites, especially in external magnetic fields. From this point of view, the rare-earth langasites R3Ga5SiO14\mathrm{R_3Ga_5SiO_{14}} (R=La\mathrm{R} = \mathrm{La}, Pr\mathrm{Pr}, Nd\mathrm{Nd}, Tb\mathrm{Tb}, Dy\mathrm{Dy}, Ho\mathrm{Ho},...) are of particular interest. For instance, the well-studied compounds Nd3Ga5SiO14\mathrm{Nd_3Ga_5SiO_{14}} and Pr3Ga5SiO14\mathrm{Pr_3Ga_5SiO_{14}} [15, 16, 17, 18, 19, 20] remain paramagnetic down to very low temperatures (30 mK) within a distorted Kagome – like lattice [15, 16]. The crystal structure of R-langasites is stable only for the light rare-earth ions with R=La\mathrm{R}=\mathrm{La}, Nd, Pr. Starting from Sm, a phase separation occurs with additional garnet-like phases [21].

Image not extracted

Figure 1. Crystal structure of langasites. The crystallographic positions with ion labels and Wyckoff notation are indicated.

The compounds containing heavy rare-earth ions with large magnetic moments (Tb, Dy, Ho, ...) are therefore only stable at low concentrations of the rare-earth ions. In such compounds one can expect a strong perturbation of the ion's ground state within the crystalline field [22, 23, 24, 25, 26, 27, 28]. This determines the local magnetic anisotropy of the ions, along with macroscopic magnetic and other physical properties of the crystal. In addition, due to the non-centrosymmetric crystal structure of langasites, they may exhibit significant magnetoelectric coupling, as seen in rare-earth ferro- and alumoborates [22, 23, 24, 25, 26, 27], which possess a similar trigonal structure (R32R32 or P3121P3_12_1 space groups).

The magnetic and magnetoelectric properties of a heavy Ho3+\mathrm{Ho^{3+}} doped langasite (La1−xHox)3Ga5SiO14(\mathrm{La_{1-x}Ho_x})_3\mathrm{Ga_5SiO_{14}} (x≈0.015x\approx0.015) (Ho-LGS) were investigated for the first time in Ref. [29]. The quasi-doublet ground state of the non-Kramers Ho3+\mathrm{Ho^{3+}} ion determines the macroscopic magnetic properties of the crystal. It is largely separated from other excited levels, which makes it an Ising-like ground state. Different local orientations of the Ho3+\mathrm{Ho^{3+}} Ising axes in various positions determine its magnetic anisotropy. This picture of the Ho3+\mathrm{Ho^{3+}} ground state was proposed in Ref. [29]. However, it does not capture all relationships between the local and macroscopic magnetization, which may indeed be more complicated. This is most likely because there is a shared occupancy of the 2d positions featuring Ga and Si ions of different size (see Figure 1). This breaks the local C2C_2 symmetry of rare-earth ions at the 3e positions and leads to a deviation of local magnetization axes (i.e., the Ising axes of Ho3+\mathrm{Ho^{3+}}) from the orientations allowed by the unperturbed C2C_2 symmetry.

Polarized neutron diffraction (without polarization analysis), involves measuring a neutron flipping ratio (i.e., the intensity ratio of neutrons scattered with incoming spins parallel and antiparallel to an external magnetic field [30]). It is an exceptionally powerful tool for studying magnetization processes at local crystallographic positions. Unlike conventional magnetization measurements, which provide information about the averaged macroscopic magnetization value, polarized neutron diffraction determines the magnetization density within the unit cell. This allows an investigation of magnetization processes occurring at the microscopic level. In particular, this method has been applied to study the local susceptibility anisotropy of rare-earth ions in pyrochlore titanates R2Ti2O7\mathrm{R}_2\mathrm{Ti}_2\mathrm{O}_7 (R=Ho,Dy,Tb\mathrm{R} = \mathrm{Ho}, \mathrm{Dy}, \mathrm{Tb}) with a frustrated spin ice magnetic structure [31, 32, 33, 34, 35, 36].

In this work we will demonstrate that polarized neutron diffraction experiments are highly effective in a complementary study using both neutron scattering and magnetization in external magnetic fields, even for a very low fraction of magnetic ions within the material. To investigate the magnetization processes at the Ho3+\mathrm{Ho}^{3+} local positions and the influence of local distortions on their behavior, we performed a comprehensive experimental and theoretical study of the magnetic properties of Ho substituted paramagnetic langasite. The local susceptibility was examined by polarized neutron diffraction in a Ho-LGS (x≈0.045x \approx0.045) single crystal. Magnetization measurements were carried out under a magnetic field applied along the principal crystallographic directions. We also explored the details of the magnetization anisotropy under magnetic fields rotated within the ab∗ab^*, acac, and b∗cb^*c planes. To explain the experimental data, we propose a model that considers the influence of local distortions on the ground state of the Ising-like Ho3+\mathrm{Ho}^{3+} ion, with local positions connected by C2\mathrm{C}_2 and C3\mathrm{C}_3 symmetry operations, which preserves the global symmetry of the crystal. This model allows us to quantitatively describe the unique magnetic anisotropy of Ho-LGS. In particular, we determine the parameters of the local distortions (deviations and distributions of the Ising axis) and establish their correlation with the macroscopic magnetic properties of the crystal. These refinements are crucial to the microscopic analysis and consistent with a quantitative description of the observed Ho-LGS magnetoelectric properties (i.e. various field-induced components of electric polarization), which are currently under investigation and will be presented in further publication.

Methods

Doped langasite crystals were grown by two different methods: the Ho-LGS crystal with x≈0.015x \approx0.015 was grown using the Czochralski method by B.V. Mill [1,2]. The crystal with x≈0.045x \approx0.045 was grown using the floating-zone method by A.M. Balbashov. We determined the quality of the crystals by X-ray analysis and by scanning electron microscopy in the z-contrast mode. In both samples, only the langasite phase was detected. The magnetic properties of the langasites were studied using a MPMS-50 (Quantum Design) magnetometer in magnetic fields up to 5 T and at temperatures from 1.9 K to 60 K.

The neutron diffraction studies were performed on the x≈0.045x \approx0.045 Ho-LGS sample at the Institute Laue-Langevin (ILL), Grenoble, France. The nuclear structure factor and extinction parameters were first characterised at 2 K using the D9 diffractometer operated at λ=0.836\lambda= 0.836 Å, produced using the (220) plane of a Cu crystal monochromator, the λ/2\lambda/2 contamination was avoided by using an Er absorption filter in the transmission geometry. Integrated intensities of structural Bragg peaks were collected with standard ω\omega-scans for low-qq reflections and ω\omega-2θ2\theta scans for medium and high-qq reflections. Polarised neutron diffraction was then performed on the D3 instrument with λ=0.825\lambda= 0.825 Å. Measurements were undertaken at 5 K in a magnetic field of 0.5 T, i.e. in the linear regime. Flipping ratios were measured on a pre-aligned crystal, with directions bb, b∗b^* and cc. For this, the OrientExpress Laue instrument at ILL was used [37]. In total 933 reflections were measured, with the external magnetic field aligned parallel to the bb (305 reflections), b∗b^* (484 reflections) and cc (144 reflections) axes.

Analytically, the experimentally obtained flipping ratio, RR, is expressed as,

R=I+I−(1)R = \frac{I^{+}}{I^{-}} \tag*{(1)}

where I+I^{+} and I−I^{-} are the scattering intensities of neutrons polarized parallel and anti-parallel to the external magnetic field respectively.

The relationship between the flipping ratio and the local susceptibility parameters is laid out in detail by Gukasov and Brown in Ref. [30]. A brief summary is also provided in the supplementary material. Ultimately, the local susceptibility tensor is obtained by a least-squares refinement of the experimental flipping ratios with respect to Eq. (1).

Experiment

Magnetic properties

Angular dependence of the magnetization was measured for various temperatures and in different magnetic fields rotating in the ab∗ab^*, b∗cb^*c and acac planes. In addition, we measured the field dependence of the magnetization along different crystallographic directions in the temperature range from 1.9 K to 90 K.

The observed angular dependences provide the most relevant information. At low temperatures and in strong magnetic fields, where the magnetic moments along local axes are almost (quasi-)saturated (the Zeeman quasi-doublet splitting exceeds the thermal energy), we observe noticeable anisotropy of the magnetization (Fig. 2 a-c) associated with the peculiarities of the local Ising axis orientation of Ho3+^{3+} ions.

Image not extracted

Figure 2. Angular dependence of the magnetization in Ho-LGS (x≈0.045x\approx0.045) in (a) ab∗ab^*, (b) acac, and (c) b∗cb^*c planes at TT=1.9 K and in external field of μ0H\mu_0H = 5 T – black, 3.5 T – red, and 2 T – blue. Open symbols represent experimental data, solid lines are theoretical predictions detailed in the text. The pictograms in the right column show the orientation of the magnetic field at the points corresponding to the minima in the angular dependences, i.e. where the field is perpendicular to some of the Ising axes of Ho3+^{3+} ions. 1±^{\pm}, 2±^{\pm}, and 3±^{\pm} represent the orientations of six equivalent magnetic axes of Ho3+^{3+}, respectively.

In the quasi-saturated regime (μ0H>1\mu_0H>1 T and TT=1.9 K), a 60-degree anisotropy was observed in the ab∗ab^* (hexagonal) plane, with a maximum along the aa-axis and a minimum along the b∗b^*-axis (i. e. perpendicular to the second-order aa-axis) (see Fig. 2a). In the acac plane an asymmetric minimum occurs between the two non-equivalent maxima for H∥aH\parallel a and H∥cH\parallel c (Fig. 2b). In the b∗cb^*c plane we observe the most characteristic benchmarks of the magnetization anisotropy. When the field is oriented in this plane, the angular dependence is asymmetric with respect to the local maximum along the cc-axis (Fig. 2c). For the clockwise and anticlockwise rotation of the magnetic field relative to the cc-axis, the neighboring minima have different depths, and the following maxima have different amplitudes. The b∗b^{*}-axis is located near one of the local minima but does not coincide with it.

The depth of the magnetization minima decreases with increasing temperature and the angular dependencies become smoother, which is also observed for fields below 1 T. In the high-temperature and low-field regime, the magnetization in the acac and b∗cb^{*}c planes follows a M∼sin⁡2θHM \sim\sin2\theta_{H} dependence, while there is no anisotropy in the ab∗ab^{*} plane. The characteristic features in the angular dependence of the magnetization are the same for both compounds with x≈0.045x\approx0.045 and x≈0.015x\approx0.015 (see Supplementary Material).

In agreement with the data above, the low temperature magnetization curves demonstrate an anisotropic behavior. The magnetization saturates along local axes around a field of ∼1\sim1 T. The Van-Vleck contribution and the deviation from Ising behavior of the magnetic moments that tend to bend toward the magnetic field (see the Theory Section) leads to a small effective susceptibility for magnetic fields above 1 T and along the principal crystallographic directions aa, b∗b^{*}, and cc (Fig. 3a). In contrast, the magnetization curves with the field aligned along one of the minima in the angular dependence (i.e. when the field is orthogonal to one of the most probable Ising axes) display a stronger additional contribution to the magnetization in the saturation regime (see Fig. 3b – Hc-39°b). This additional contribution is characterized by a steeper slope and by a reduction of angular modulation with increasing magnetic field strength (see Fig. 2(c) for fields 2 T, 3.5 T, and 5 T). Finally, Figure 6, shown later in the Theory Section, demonstrates how the magnetization asymmetry in the b∗cb^{*}c plane is compatible with the crystallographic symmetry P321 and how it is associated with a deviation of the local magnetic ellipsoids from the acac plane.

Image not extracted

Figure 6. Schematic representation of susceptibility ellipsoids and the most probable directions of the Ising axes. (a) - in a three-dimensional view and (b) - projected onto the abab plane. The numbers ±1\pm1, ±2\pm2, ±3\pm3 denote the most probable directions of the Ising axes in the corresponding positions. The local coordinate system XYZXYZ, rotated by an angle β\beta around the second-order axis, is shown for position "1" only. The angle α\alpha represents the deviation of the most probable directions of the Ising axes from the aa-axis.

Image not extracted

Figure 3.

FIG. 3. (a) Magnetization curves of Ho-LGS (x≈0.045x \approx0.045) for H∥b∗H \parallel b^* and at temperatures between 1.9 K and 60 K. (b) Comparison of the magnetization curves at T=1.9T=1.9 K for H∥aH \parallel a, b∗b^*, cc, and cc-39°b∗b^* directions. The last direction corresponds to the minimum of the angular dependence in the b∗cb^*c plane (see Fig. 2(c)). (c) Comparison of the susceptibility observed from magnetization and from neutron data for H∥b∗H \parallel b^* and (d) for H∥cH \parallel c. Symbols represent experimental data, solid lines are theoretical calculations (see sections 4.1–4.3).

Neutron Scattering

Flipping ratios for a magnetic field of μ0H=0.5\mu_0 H = 0.5 T applied along the bb, b∗b^* and cc directions were refined simultaneously using the ChiLSQ subroutine of the Cambridge Crystallography Subroutine Library (CCSL). A second refinement was also performed using the Crystallographic Python Library (CrysPy), with both libraries giving equivalent outputs. Later, these refinements were further validated using the Mag2Pol software. The magnetic form factor was approximated at the dipole limit as

f(k)=⟨J0(k)⟩+(2gL−1)⟨J2(k)⟩,f(k) = \langle J_0(k) \rangle+ \left(\frac{2}{g_{\mathrm{L}}}-1\right)\langle J_2(k) \rangle,

where ⟨J0,2(k)⟩\langle J_{0,2}(k) \rangle are radial integrals of the Ho3+^{3+} electronic wavefunction and gL=5/4g_{\mathrm{L}} = 5/4 is the Landé gg-factor, accounting for the orbital contribution to the magnetic moment (see supplementary material for details [38]). The refinement quality for the three different applied magnetic field directions is shown in Figure 4.

Neutron flipping ratio refinement plots for applied fields along the crystallographic directions

Figure 4. Neutron flipping ratio refinement using CCSL and dipole approximation for applied fields of 0.5 T along the (a) b∗b^*, (b) bb, and (c) cc crystallographic directions. The measurements were performed at a sample temperature of 5 K.

The refined susceptibility tensor for a single Ho3+^{3+} ion rounded to two decimal places was found to be,

χ^=(5.403(1)0.00(5)0.002(1)χ120.2(2)0.5(4)χ13χ232.4(1))μB/T,(2)\hat{\chi} = \begin{pmatrix} 5.403(1) & 0.00(5) & 0.002(1) \\ \chi_{12} & 0.2(2) & 0.5(4) \\ \chi_{13} & \chi_{23} & 2.4(1) \end{pmatrix} \mu_{\mathrm{B}}/\mathrm{T}, \tag*{(2)}

within the orthogonal coordinate system {x}∥a\{x\}\parallel a, {y}∥b∗\{y\}\parallel b^*, and {z}∥c\{z\}\parallel c. Diagonalizing the tensor gives the following eigenvectors X=aX = a, Y=0.98b∗−0.21cY = 0.98b^* - 0.21c, and Z=0.21b∗+0.98cZ = 0.21b^* + 0.98c, such that,

χXX+χYY+χZZ=5.40X+0.12Y+2.50Z [μB/T].(3)\chi_X X + \chi_Y Y + \chi_Z Z = 5.40X + 0.12Y + 2.50Z\ [\mu_{\mathrm{B}}/\mathrm{T}]. \tag*{(3)}

A graphical representation of the corresponding magnetization ellipsoid is shown in Figure 5 (a,Figure 5c). It consists of an ellipse, mostly in the acac plane, with a slight rotation of 13∘±7∘13^\circ\pm7^\circ around the aa-axis towards the b∗b^*-axis.

Image not extracted

Figure 5. Calculated and measured magnetic susceptibility ellipsoids of Ho3+^{3+} in Ho-LGS (x≈0.045x \approx0.045) represented in the local coordinate system of a pair of distorted crystallographic positions connected by the C2C_2 symmetry. Projections of the ellipsoids on the planes of the local coordinate system are shown for XZXZ (a) and XYXY (b). The external magnetic field is μ0H=0.5\mu_0H = 0.5 T and the temperature is T=5T = 5 K. The right-most column shows the schematic orientations of the susceptibility ellipsoids within the crystallographic and local coordinates. The projection planes XZXZ (a) and XYXY (b) are indicated by the gray planes. The crystallographic coordinate system is rotated with respect to the local XYZXYZ axes by an angle β~≈4∘±2∘\tilde{\beta} \approx4^\circ\pm2^\circ around the second-order axis (C2∥aC_2 \mathbin{\|} a). Narrow blue and black ellipses show the Ising-like susceptibility of two most probable "+" and "-" directions coupled by the C2C_2 symmetry along the aa-axis. The orange and green solid lines represent the model ellipses of susceptibility accounting for small non-linear field-contributions for static fields H∥XH \mathbin{\|} X (orange) and H∥ZH \mathbin{\|} Z (green) for both distorted positions (see details in 4.4 section). Dotted lines represent the projections of the magnetic ellipsoid obtained from polarized neutron diffraction. Stars denote the susceptibility values along the local XYZXYZ axes refined from neutron measurements.

The bulk susceptibility per Ho ion is then determined by,

χij=1NHo∑k=13∑l=13RikRjlχkl,(4)\chi_{ij} = \frac{1}{N_{\mathrm{Ho}}}\sum_{k=1}^{3}\sum_{l=1}^{3} R_{ik}R_{jl}\chi_{kl}, \tag*{(4)}

where NHoN_{\mathrm{Ho}} is the number of Ho ions in the unit cell and 1<i≤31 < i \le3 and 1<j≤31 < j \le3 for all R~\widetilde{R} operators in the C3C_{3} point group. Here we apply the inverse property of symmetry rotations Rij=Rji−1R_{ij} = R_{ji}^{-1}. The resulting bulk susceptibility per Ho ion is therefore

χ^Bulk=13(8.450008.450007.17)[μB/Ho]/T.(5)\hat{\chi}_{\mathrm{Bulk}} = \frac{1}{3} \begin{pmatrix} 8.45 & 0 & 0 \\ 0 & 8.45 & 0 \\ 0 & 0 & 7.17 \end{pmatrix} [\mu_{\mathrm{B}}/\mathrm{Ho}]/T. \tag*{(5)}

The result is in fair agreement with the measured magnetic field dependence in the linear regime as shown in Fig. 3.

We note that the ellipsoid of the local magnetic susceptibility extracted from the neutron data does not correspond to the expected Ising-like behavior of Ho3+^{3+} ions proposed in Ref. [29]. Therefore, in the proceeding Theory Section we develop a more sophisticated model. It starts from the Ising-like character of Ho3+^{3+} ions and accounts for the observed susceptibility ellipsoid assuming a distribution of locally-distorted positions.

Image not extracted

Figure 3 (PDF p. 7).

Theory and Discussion

Model of disorder-induced anisotropy distribution

In Ho-LGS, the Ho3+^{3+} magnetic ions occupy three low-symmetry positions with a C2C_2 symmetry axis aligned to one of the three crystallographic directions (aa, bb, −a−b-a-b). They remain in a paramagnetic state down to low temperatures. However, the shared occupation of Ga and Si at the 2d positions leads to a breaking of the local C2C_2 symmetry in the rare-earth environment. The possible presence of a superlattice in LGS [39, 40] suggests the existence of some energetically more favorable states (i.e., arrangements of Ga and Si). It is impossible to unambiguously identify them from the present data only, but we assume that rigorous microscopic calculation may determine the set of allowed configurations.

A crystal electric field of low symmetry (C1C_1) splits the ground multiplet 5I8{}^{5}I_8 into 2J+1=172J+1 = 17 singlets. The two lowest neighboring energy levels (a quasi-doublet with a small splitting of Δcf≈3\Delta_{\mathrm{cf}} \approx3 K) primarily determine the magnetic properties of the Ho3+^{3+} non-Kramers ion in langasite (see also Ref. [29]). Such a ground state results in an Ising-like behavior. The matrix element of the magnetic moment operator between the wave functions of this quasi-doublet determines the magnetization of the Ho3+^{3+} ions, which exhibit strong anisotropy (see, for example, Ref. [41]). This is consistent with the low temperature behavior of the magnetization curves, showing a large slope in low fields, and a fast saturation at μ0H∼1\mu_0H \sim1 T (Fig. 3).

To describe the magnetic properties of Ho-LGS, we used an approach similar to Ref. [29] based on a spin Hamiltonian that includes splitting of the quasi-doublet in the crystal electric field, along with a Zeeman energy and Van-Vleck contributions:

Heff(i)=Δcfσz(i)+miHσy(i)−12Hχ^VV(i)H.(6)H_{\mathrm{eff}}^{(i)} = \Delta_{\mathrm{cf}}\sigma_z^{(i)} + \mathbf{m}_i\mathbf{H}\sigma_y^{(i)} - \frac{1}{2}\mathbf{H}\hat{\chi}_{\mathrm{VV}}^{(i)}\mathbf{H}. \tag*{(6)}

Here σz,y(i)\sigma_{z,y}^{(i)} are the Pauli matrices at the ii-th position, miH=μ0niH=−iμBgJ⟨A∣i⟩H∣B⟩\mathbf{m}_i\mathbf{H}=\mu_0 n_i\mathbf{H}=-i\mu_{\mathrm{B}}g_J\langle A|i\rangle\mathbf{H}|B\rangle (where ∣A⟩|A\rangle and ∣B⟩|B\rangle denote the wave functions of the quasi-doublet), ni\mathbf{n}_i is a unit vector of an arbitrarily oriented Ising axis, H\mathbf{H} is the external magnetic field and χ^VV(i)\hat{\chi}_{\mathrm{VV}}^{(i)} is the Van-Vleck susceptibility matrix. The energy levels of the quasi-doublet at the ii-th position, obtained by diagonalization of the spin Hamiltonian, are equal to ±εi\pm\varepsilon_i, where εi=(miH)2+Δcf2\varepsilon_i = \sqrt{(\mathbf{m}_i\mathbf{H})^2 + \Delta_{\mathrm{cf}}^2}, and determine the magnetization of Ho3+^{3+} ions (see below).

Next, we consider possible distributions of local magnetization directions (Ising axes) and their relationship to different positions within the crystal. Let's assume that the magnetization direction (Ising axis) is determined by an arbitrary unit vector n(α,β)=(cos⁡α,sin⁡αsin⁡β,sin⁡αcos⁡β)\mathbf{n}(\alpha,\beta)=(\cos\alpha,\sin\alpha\sin\beta,\sin\alpha\cos\beta). Here α\alpha is its deviation from the aa-axis, and β\beta is the angle between the projection of the Ising axis in the b∗cb^*c plane and the cc-axis (as in Figure 6). Due to the random distribution of Ga/Si ions breaking the local C2C_2 symmetry of the Ising axis, the angles α\alpha and β\beta, should also be randomly distributed around their most probable values αˉ\bar{\alpha} and βˉ\bar{\beta}, which can be determined by modeling of the magnetic properties. In this work we use a two-dimensional Gaussian distribution ρ(α,β)=Exp⁡[−(α−αˉ)22δα2]Exp⁡[−(β−βˉ)22δβ2]/2πδαδβ\rho(\alpha,\beta)=\operatorname{Exp}\left[-\frac{(\alpha-\bar{\alpha})^2}{2\delta_\alpha^2}\right]\operatorname{Exp}\left[-\frac{(\beta-\bar{\beta})^2}{2\delta_\beta^2}\right]/2\pi\delta_\alpha\delta_\beta for the Ising axes directions with dispersions δα\delta_\alpha and δβ\delta_\beta.

To preserve a global P321P321 symmetry of the Ho-LGS crystal we assume the presence of another distorted position coupled with the first one by the C‾2\overline{C}_2 symmetry operation. Taking into account that the angle αˉ\bar{\alpha} is invariant while βˉ\bar{\beta} transforms to βˉ−π\bar{\beta}-\pi under the C‾2\overline{C}_2 symmetry operation, the distribution function then becomes ρ′(α,β)=Exp⁡[−(α−αˉ)22δα2]Exp⁡[−(β−(βˉ−π))22δβ2]/2πδαδβ\rho'(\alpha,\beta)=\operatorname{Exp}\left[-\frac{(\alpha-\bar{\alpha})^2}{2\delta_\alpha^2}\right]\operatorname{Exp}\left[-\frac{(\beta-(\bar{\beta}-\pi))^2}{2\delta_\beta^2}\right]/2\pi\delta_\alpha\delta_\beta.

Thus, the superposition of ρ\rho and ρ′\rho' in distorted positions determines the distribution function ρ+=(ρ+ρ′)/2\rho_+=(\rho+\rho')/2 of the averaged local positions possessing the global C2C_2 symmetry.

The orientation of the Ising axes and the distribution functions in other positions are determined by ±120∘\pm120^\circ rotation around the trigonal (C3C_3) axis. Let us now denote the position with an arbitrarily oriented Ising axis n1(α,β)=(cos⁡α,sin⁡αsin⁡β,sin⁡αcos⁡β)\boldsymbol{n}_1(\alpha,\beta)=(\cos\alpha,\sin\alpha\sin\beta,\sin\alpha\cos\beta) and consider the distribution function ρ+\rho_+ at position “1”. Here, the form of ρ+\rho_+ reflects the presence of the two most probable directions (coupled by C2C_2 symmetry). The additional positions "2" and "3" are related to "1" by an extra C3C_3 symmetry operation. The rotation matrices 3^c±\hat{3}_{c}^{\pm} determine the pair of Ising axis directions n2,3(α,β)=3^c±n1(α,β)\boldsymbol{n}_{2,3}(\alpha,\beta)=\hat{3}_{c}^{\pm}\boldsymbol{n}_1(\alpha,\beta) in positions "2" and "3". The distribution function ρ+\rho_+ remains unchanged under the symmetry operations. Thus, we introduce a three-sublattice model with the distribution function ρ+\rho_+. It preserves the global symmetry P321P321 of the crystal despite its local deviations.

Since all three sublattices, "1", "2" and "3" include two components, they can be also treated as pairwise split into six sublattices, 1±1^{\pm}, 2±2^{\pm}, 3±3^{\pm}. The most probable Ising directions for 1±1^{\pm} are n1+(αˉ,βˉ)\boldsymbol{n}_{1+}(\bar{\alpha},\bar{\beta}) and n1−(αˉ,βˉ−π)\boldsymbol{n}_{1-}(\bar{\alpha},\bar{\beta}-\pi), with n2+,3+(αˉ,βˉ)=3^c±n1+(αˉ,βˉ)\boldsymbol{n}_{2+,3+}(\bar{\alpha},\bar{\beta})=\hat{3}_{c}^{\pm}\boldsymbol{n}_{1+}(\bar{\alpha},\bar{\beta}), n2−,3−(αˉ,βˉ)=3^c±n1−(αˉ,βˉ−π)\boldsymbol{n}_{2-,3-}(\bar{\alpha},\bar{\beta})=\hat{3}_{c}^{\pm}\boldsymbol{n}_{1-}(\bar{\alpha},\bar{\beta}-\pi) for 2±2^{\pm} and 3±3^{\pm}, respectively. They are connected by the set of threefold symmetry axis rotations on n1±\boldsymbol{n}_{1\pm} (Figure 6). The distribution function ρ+\rho_+ is the same in all positions "1", "2" and "3". For the interpretation of magnetization curves, we will employ the six-sublattice approach, while, for neutron data, it is more convenient to use the three-sublattice model with a generalized distribution function. Note that if we consider the undistorted Ho3+^{3+} positions with local C2C_2 symmetry, it is not possible to describe the angular and field dependences of the magnetization consistently (see Ref. [29]).

In the frame of this model, we can now define the total free energy of the two level (±εi\pm\varepsilon_i) system and calculate the average magnetization considering three types of magnetic positions with the distribution ρ+\rho_+ of Ising axes:

M=13nHo∑i=1,2,3(∬[mi(Hmi)+2(Δcf+Hp^iH)p^i]tanh⁡(εi/kBT)εiρ+ dαdβ+χVV(i)H).(7)\boldsymbol{M}=\frac{1}{3}n_{\mathrm{Ho}}\sum_{i=1,2,3}\left(\iint\left[\boldsymbol{m}_i(H\boldsymbol{m}_i)+2\left(\Delta_{\mathrm{cf}}+H\hat{\boldsymbol{p}}_iH\right)\hat{\boldsymbol{p}}_i\right]\frac{\tanh(\varepsilon_i/k_{\mathrm{B}}T)}{\varepsilon_i}\rho_+\,d\alpha d\beta+\chi_{\mathrm{VV}}^{(i)}H\right). \tag*{(7)}

Here, mi=μ0ni\boldsymbol{m}_i=\mu_0\boldsymbol{n}_i, p^i\hat{\boldsymbol{p}}_i, εi\varepsilon_i and ρ+\rho_+ depend on α\alpha, β\beta, nHon_{\mathrm{Ho}} is the Ho3+^{3+} ion concentration, and kBk_{\mathrm{B}} is the Boltzmann constant. We also take into account a weak renormalization of the crystal field splitting by magnetic field Δcf→Δcf+Hp^iH\Delta_{\mathrm{cf}}\to\Delta_{\mathrm{cf}}+H\hat{\boldsymbol{p}}_iH due to an influence of excited Ho3+^{3+} states (see also Ref. [42]). It is characterized by the local magnetic tensor p^′\hat{\boldsymbol{p}}', which can be represented in the crystallographic coordinate system by the rotation matrices p^i=C^i−1p^′C^i\hat{\boldsymbol{p}}_i=\hat{C}_i^{-1}\hat{\boldsymbol{p}}'\hat{C}_i [38]. The p^′\hat{\boldsymbol{p}}' tensor depends on the matrix elements of the magnetic moment between the ground and excited Ho3+^{3+} crystal field states and leads to a nonzero susceptibility when the magnetic field and Ising axis are orthogonal.

Angular dependence of magnetization

The model of Ising axis directions qualitatively explains the presence of extrema in the angular dependence of magnetization. The distinctive features of the Ising-like behavior of Ho3+^{3+} and of the disorder-induced distribution of their anisotropy axis are the most prominent in the quasi-saturated regime (μ0H>1\mu_{0}H > 1 T and T<5T < 5 K). Specifically, when the magnetic field is oriented perpendicular to one of the Ising axes, minima appear in the angular dependence of magnetization (Fig. 2a-c).

For magnetic fields rotated in the ab∗ab^{*} plane, the magnetization behaves as M∼sin⁡6φM \sim\sin6\varphi, and the equivalence of magnetic moment orientations parallel and antiparallel to the Ising axes leads to the presence of a 60-degree periodicity. We note that when the magnetic field is rotated by 90 degrees in the ab∗ab^{*} plane, three minima should be observed, for example, in the range from 0 to +90∘+90^{\circ}, when the magnetic field HH is orthogonal to the most probable axes n2+n_{2+}, n2−n_{2-}, and n1−n_{1-} (Fig. 2 right). However, the directions n2+n_{2+} and n2−n_{2-} are close to each other, thus making the corresponding minima indistinguishable (Fig. 2a).

When the field rotates by 180 degrees in the acac plane, two minima with slight asymmetry are observed (Fig. 2b). These minima are the combinations of three weaker minima. For example, when the magnetic field rotates from the cc-axis to the aa-axis, it becomes orthogonal to the axes n1−n_{1-}, n2+n_{2+} and n3+n_{3+} at the angles θH≈32∘\theta_{H} \approx32^{\circ}, 49∘49^{\circ} and 53∘53^{\circ}, respectively (right column in Fig. 2). These features are not resolved in the experiment as they are smoothed out due to the random distribution. In the angular dependences, the maximum for H∥cH \parallel c arises due to the deviation of the Ising axes from the ab∗ab^{*} plane. From the magnetization ratio for H∥cH \parallel c and H∥aH \parallel a we estimate the deviation angle as α≈32∘\alpha\approx32^{\circ} (Figure 6a).

Further characteristic features related to the orientation of the Ising axes are observed for magnetic fields in the b∗cb^{*}c plane. In this plane, the angles at which the field is orthogonal to the axes n3−n_{3-} and n2+n_{2+}, n1+n_{1+} and n1−n_{1-}, n2−n_{2-} and n3+n_{3+} coincide. As a result, only three distinct minima are observed in the 180-degree angular dependencies (Fig. 2c). The local minimum near H∥b∗H \parallel b^* occurs when the field is orthogonal to n1+\mathbf{n}_{1+} and n1−\mathbf{n}_{1-} (θH≈94∘\theta_H \approx94^\circ, right column in Fig. 2). The inequality θH≠90∘\theta_H \ne90^\circ and the asymmetry of the angular dependence in the b∗cb^*c plane, when the sign of the magnetic field projection onto the cc-axis changes (θH=90∘±δ\theta_H = 90^\circ\pm\delta, Fig. 2c), indicates a deviation of the local XZXZ plane (which includes n1+\mathbf{n}_{1+} and n1−\mathbf{n}_{1-}, see Figure 6b) from the acac plane. The estimation of this deviation is β‾≈4∘±2∘\overline{\beta} \approx4^\circ\pm2^\circ, which falls within the uncertainty bands of the angle determined by neutron refinements.

Field induced magnetic moment distribution

Using the magnetization equation in terms of the magnetic structure model, the field dependences of magnetization were simulated as well (Figure 5). From these data we can estimate the concentration of the magnetic ions in both samples (nHo=1.53%,4.45%n_{\mathrm{Ho}} = 1.53\%, 4.45\%) and the magnetic moment of the doublet (μ0=9.4μB\mu_0 = 9.4\mu_{\mathrm{B}}) for H∥aH \parallel a, b∗b^* and cc. The magnetic properties for the sample with nHo=1.53%n_{\mathrm{Ho}} = 1.53\% can be described self-consistently without additional parameters [38].

As discussed above, for H∥aH \parallel a, b∗b^* and cc, the magnetization slope in the saturation regime and for the sample with n=4.45%n = 4.45\% is rather small. In contrast, when the magnetic field is orthogonal to one of the Ising axes (Figure 5b, H∥c−39∘b∗H \parallel c - 39^\circ b^*), this slope increases significantly. We believe that this is related to the deviation of Ho3+\mathrm{Ho}^{3+} ions from pure Ising behavior. The p^′\hat{p}' components that determine this slope are equal to px=0p_x = 0, py=17⋅10−2 μB/Tp_y = 17 \cdot10^{-2}\,\mu_{\mathrm{B}}/\mathrm{T}, and pz=42.1⋅10−2 μB/Tp_z = 42.1 \cdot10^{-2}\,\mu_{\mathrm{B}}/\mathrm{T}. A finite slope in the magnetization along the aa, b∗b^* and cc-axes in the saturation regime indicates the presence of Van-Vleck contributions to the susceptibility. Summed over three positions it can be estimated as

χVVa=χVVb∗=13nHo∣∑i=1,2,3χ^VV(i)HH∣a,b∗=2.7⋅10−2 μB/T,and\chi_{\mathrm{VVa}}=\chi_{\mathrm{VVb}^*}=\frac{1}{3}n_{\mathrm{Ho}}\left|\sum_{i=1,2,3}\hat{\chi}_{\mathrm{VV}}^{(i)}\frac{\mathbf{H}}{H}\right|_{a,b^*}=2.7\cdot10^{-2}\,\mu_{\mathrm{B}}/\mathrm{T}, \qquad\text{and}
χVVc=13nHo∣∑i=1,2,3χ^VV(i)HH∣c=1.2⋅10−2 μB/T.\chi_{\mathrm{VVc}}=\frac{1}{3}n_{\mathrm{Ho}}\left|\sum_{i=1,2,3}\hat{\chi}_{\mathrm{VV}}^{(i)}\frac{\mathbf{H}}{H}\right|_{c}=1.2\cdot10^{-2}\,\mu_{\mathrm{B}}/\mathrm{T}.

The distribution of the Ising axes leads to a broadening of the minima in the angular dependencies of magnetization. In combination with deviations from the Ising behavior, a quantitative description of the depth of minima in Figs. 2a-c is obtained for the dispersion δα=6.7∘\delta_\alpha= 6.7^\circ and δβ=8∘\delta_\beta= 8^\circ.

Local magnetic susceptibility

The magnetic susceptibility calculated using the parameters obtained from the model above is also in agreement with the susceptibility obtained from the polarized neutron diffraction measurements (Figure 5). We calculated the magnetic susceptibilities of the local positions, which are superpositions of susceptibilities from two distorted Ho3+\mathrm{Ho}^{3+} positions connected by C2C_2 symmetry. We used the definition of magnetization M\mathbf{M} from Eq. (7) to determine the effective susceptibility χ^\hat{\chi} of position "1". The increment of the induced magnetization by the infinitesimal change of the magnetic field dhd\mathbf{h} is equal to dM(1)=M(1)(H+dh)−M(1)(H)d\mathbf{M}^{(1)}=\mathbf{M}^{(1)}(\mathbf{H}+d\mathbf{h})-\mathbf{M}^{(1)}(\mathbf{H}).

For the low field regime μ0H≪kBT\mu_0H \ll k_{\mathrm{B}}T, Δcf\Delta_{\mathrm{cf}} one can linearize dM(1)d\mathbf{M}^{(1)} by the increment of static field dhd\mathbf{h} and introduce it's from as:

dM(1)=nHoχ^(1)(H)dh.(8)d\mathbf{M}^{(1)}=n_{\mathrm{Ho}}\hat{\chi}^{(1)}(\mathbf{H})d\mathbf{h}. \tag*{(8)}

Here the effective local susceptibility is,

χξη(1)(H)=∬(mξmη−14(mHkBT)2)(2mξmη+mξ2)(kBT)−1ρ+dαdβ+χVVξη(9)\chi_{\xi\eta}^{(1)}(\mathbf{H})=\iint\left(m_\xi m_\eta-\frac{1}{4}\left(\frac{mH}{k_{\mathrm{B}}T}\right)^2\right)\left(2m_\xi m_\eta+m_\xi^2\right)(k_{\mathrm{B}}T)^{-1}\rho_+d\alpha d\beta+\chi_{\mathrm{VV}}^{\xi\eta} \tag*{(9)}

with ξ,η=a\xi,\eta=a, b∗b^*, cc. For simplicity, we omitted renormalization of the crystal field splitting by magnetic field due to the influence of excited Ho3+\mathrm{Ho}^{3+} states.

The linearization with respect to HH perfectly works in the low field and high temperature regime. In the crystallographic coordinate system the susceptibility tensor for T = 5 K is equal to

χ^(1)=∬(mx2000my2mymz0mymzmz2)(kBT)−1ρ dα dβ+χ^VV(1)=(8.470000.080.2300.233.32)(μB/Ho)/T(10)\begin{aligned} \hat{\chi}^{(1)} = \iint\begin{pmatrix} m_x^{2} & 0 & 0 \\ 0 & m_y^{2} & m_y m_z \\ 0 & m_y m_z & m_z^{2} \end{pmatrix}(k_{\mathrm{B}}T)^{-1}\rho\,d\alpha\,d\beta+\hat{\chi}_{\mathrm{VV}}^{(1)} \\ &= \begin{pmatrix} 8.47 & 0 & 0 \\ 0 & 0.08 & 0.23 \\ 0 & 0.23 & 3.32 \end{pmatrix}(\mu_{\mathrm{B}}/\mathrm{Ho})/T \tag*{(10)} \end{aligned}

where we took the C2_2 symmetry into account, which cancels certain components and allows us to reduce the ρ+\rho_{+} distribution function for two sites to just one site, ρ\rho. In this case, the susceptibility in position "1" is described by the components χ^(1)\hat{\chi}^{(1)} (Figure 5), and the susceptibilities of "2" and "3" are connected to "1" by ±120∘\pm120^\circ rotation (Figure 6). The rotation around the aa-axis by an angle βrot=12arctan⁡(∫sin⁡2β ρ dβ/∫cos⁡2β ρ dβ)≈4∘\beta^{\mathrm{rot}} = \frac{1}{2}\arctan(\int\sin2\beta\,\rho\,d\beta/\int\cos2\beta\,\rho\,d\beta) \approx4^\circ diagonalizes the susceptibility matrix, which is equal to βˉ\bar{\beta} and, taking uncertainties into account, is in agreement with βneutr≈13∘±7∘\beta_{\mathrm{neutr}} \approx13^\circ\pm7^\circ. Thus, in the local coordinate system, obtained by rotating the crystallographic coordinates system by an angle βrot\beta^{\mathrm{rot}} around XX parallel to the second order axis C2_2 (aa) (as shown in the right column of Figure 5), the susceptibility is well described by an ellipse, which is a cross-section of the ellipsoid x2/χxx2+y2/χyy2+z2/χzz2=1x^2/\chi_{xx}^2+y^2/\chi_{yy}^2+z^2/\chi_{zz}^2=1. The calculated eigenvector describing the magnetic ellipse of a single Ho site is given by

χxxX+χyyY+χzzZ=8.47X+0.07Y+3.34Z [μB/Ho]/T(11)\chi_{xx}X+\chi_{yy}Y+\chi_{zz}Z=8.47X+0.07Y+3.34Z\ [\mu_{\mathrm{B}}/\mathrm{Ho}]/T \tag*{(11)}

which is larger than that determined by the neutron refinements (3).

The reason for this is likely due to nonlinearities already present at 0.5 T. Specifically there is a decrease in the susceptibility under increasing applied fields (Fig 3 c,d). For magnetic fields (mH)2∼(kBT)2(mH)^2\sim(k_{\mathrm{B}}T)^2 the additional quadric term given by the static magnetic field HH must also be considered. It reduces the value of χ^(1)\hat{\chi}^{(1)} according to Eq. (9) and makes positions "1", "2" and "3" nonequivalent. Using Eq. (9) we calculated the eigenvector describing the magnetic ellipse of a single Ho site in the nonlinear regime (T=5 K and μ0H=0.5\mu_0H=0.5 T) for H∥XH\parallel X and ZZ respectively obtaining,

χxxX+χyyY+χzzZ=7.79X+0.06Y+3.02Z [μB/Ho]/T\chi_{xx}X+\chi_{yy}Y+\chi_{zz}Z=7.79X+0.06Y+3.02Z\ [\mu_{\mathrm{B}}/\mathrm{Ho}]/T
χxxX+χyyY+χzzZ=6.62X+0.05Y+2.67Z [μB/Ho]/T.(12)\chi_{xx}X+\chi_{yy}Y+\chi_{zz}Z=6.62X+0.05Y+2.67Z\ [\mu_{\mathrm{B}}/\mathrm{Ho}]/T. \tag*{(12)}

The values are now closer to those determined by the neutron refinement, but differ for H∥XH\parallel X and ZZ (Figure 5a). Here the βrot\beta^{\mathrm{rot}} dependence on HH and TT is negligible. We highlight that the neutron refinements were performed simultaneously on data collected for H∥b∗H\parallel b^*, bb and cc and considering all three Ho positions coupled by C3_3 symmetry. However, due the non-linear terms mentioned above, the response will depend on the field direction with respect to the local axes of the ellipsoid in different positions. Thus, in this particular case, the refined susceptibility tensor obtained from the neutron data (χ^neutr\hat{\chi}^{\mathrm{neutr}}) should be interpreted as an averaged and effective susceptibility.

The non-zero susceptibility for dh∥Ydh\parallel Y (Figure 5b) is the result of the distribution of the Ising axes and deviations from the Ising behavior, with the second case not being taken into account in Eqs. (8)–(12). Accounting for these deviations doubles the value χxy≈0.10\chi_{xy}\approx0.10 (μB/Ho)/T(\mu_{\mathrm{B}}/\mathrm{Ho})/T. The corresponding ellipse in the XYXY plane calculated numerically is shown on Figure 5b. There is a good agreement between the calculated and refined susceptibilities χxy\chi_{xy}. However, we note that the effective susceptibility χyyneutr\chi_{yy}^{\mathrm{neutr}} has a large relative error because of its small value. Moreover, the Gaussian distribution of Ising axes is a model approximation, which can introduce errors in the calculated susceptibilities for fields orthogonal to the Ising axes. Finally, we did not account for the specifics of the wave functions of Ho3+^{3+} to estimate the components of ρ^\hat{\rho} which describe the deviation from the Ising behavior and strongly affect the χxy\chi_{xy} magnitude. A rigorous mathematical treatment considering various configurations of Ga and Si ions and calculating the spectrum of Ho3+^{3+} ions would enable a more accurate modeling of the magnetic properties. However, in this work we used a simpler approach and were still able to obtain a good agreement between model and experiment.

In summary, the magnetization experiments provide the following approximate distribution of the magnetic anisotropy of Ho-LGS. It consists of two Ising axes n1+\mathbf{n}_{1+} and n1−\mathbf{n}_{1-}, at distorted positions connected by C2_2 symmetry and rotated from the local aa-axis by the angles α‾≈±32∘\overline{\alpha}\approx\pm32^\circ, respectively. The plane built of n1+\mathbf{n}_{1+} and n1−\mathbf{n}_{1-} roughly coincide with the acac-plane but is rotated away from it by the angle β‾≈4∘\overline{\beta}\approx4^\circ (see Figure 6). Both angles α\alpha and β\beta reveal Gaussian distributions with dispersions σα=6.7∘\sigma_{\alpha}=6.7^\circ and σβ=8∘\sigma_{\beta}=8^\circ, respectively. Four other Ho3+^{3+} positions are obtained from n1+\mathbf{n}_{1+} and n1−\mathbf{n}_{1-} applying the C3_3 symmetry operation along the cc-axis.

Taking into account the experimental errors and the model assumptions, the susceptibility agrees rather well with the refinement obtained from polarized neutron diffraction measurements. A graphical comparison of the susceptibility ellipsoids from magnetization and from neutron experiments is shown in Figure 5.

An important aspect obtained from the magnetization data is the observable influence of the local directions and the distributions of the Ising axes. To obtain further details of this magnetic configuration, more sophisticated experiments and rigorous theoretical calculations are necessary.

Conclusion

Using a comprehensive experimental and theoretical approach, we studied the effect of Ho3+^{3+} local magnetic orientations and their distortions on the macroscopic magnetic properties of doped langasites (La1−xHox)3Ga5SiO14(\mathrm{La}_{1-x}\mathrm{Ho}_{x})_{3}\mathrm{Ga}_{5}\mathrm{SiO}_{14} with x≈0.015x\approx0.015 and x≈0.045x\approx0.045.

The local magnetic susceptibility at local Ho3+^{3+} positions was determined by flipping ratio measurements using polarized neutron diffraction. Taking the C2_2 symmetry of the local 3e Ho3+^{3+} positions into account, the magnetic ellipsoids of susceptibility were determined by refinement of the neutron data. The highest susceptibility (the major axis of the ellipsoid) occurs for fields directed along the C2_2 axes XX (H∥aH\parallel a, bb, −a−b-a-b). The next-largest susceptibility is seen in a field directed along the local ZZ axis, that deviates from the acac-plane plane by an angle βneutr≈13∘±7∘\beta_{\mathrm{neutr}}\approx13^\circ\pm7^\circ. The local susceptibility along the YY-axis, which is perpendicular to the XX and ZZ axes, is an order of magnitude smaller than the other two. Here the refinement of the polarized neutron data assumed a local C2_2 symmetry and provided a good approach to capture the magnetic density at local crystallographic positions. The insight into the details of the rare-earth sites with violated C2_2 symmetry was obtained from the analysis of the angular dependent magnetization in the ab∗ab^*, b∗cb^*c, and acac planes especially in the quasi-saturated regime (μ0H>1\mu_{0}H>1 T).

We propose a simple model of the field-induced magnetic structure assuming a quasi-doublet ground state of the non-Kramers Ho3+^{3+} ions in a crystal field environment. This includes a perturbation of the Ising-like behavior with an arbitrary orientation of the local Ising axis in distorted positions due to shared occupation of the neighboring Ga/Si 2d sites in langasites. We argue that an averaged distribution of the Ising axes for the distorted Ho3+^{3+} positions are coupled by local C2_2 symmetry and by the rotations around the trigonal cc-axis, which restores the global P321P321 symmetry of the crystal. Within these assumptions it is possible to quantitatively describe the observed global magnetic properties and the ellipsoids of the local susceptibility. As a result, microscopic characteristics of the Ho3+^{3+} ground state, such as the most probable local orientations, the amplitude and distribution of the magnetic moments could be determined. These characteristics do not depend on the Ho-doping value of the studied langasite crystals, thus confirming the validity of the approach.

Finally, similar influence of local magnetic distortions on the macroscopic properties in doped langasite crystals reveals characteristic features that can be studied in other systems as well. More detailed insight into the distribution of local magnetic properties requires rigorous microscopic calculations and may also necessitate the development of new experimental approaches.

Acknowledgments

We thank Navid Quershi for validating our neutron refinement results using the Mag2Pol software. This work was supported by the Russian Science Foundation (Project No. 22-42-05004) and the Austrian Science Fund (Projects No. P32404 and No. I5539).

Supplemental Material

Rotation of the local magnetic tensor

The local aa-axis in position "1" is the second order-axis. In the spin coordinate system, the local xx-axis is parallel to the Ising axis. The connection between the local magnetic and the crystallographic coordinate systems can be represented as:

(x′,y′,z′)=B^(β)A^(α)(a,b∗,c).(13)(x',y',z')=\hat{B}(\beta)\hat{A}(\alpha)(a,b^*,c). \tag*{(13)}

Here, A^(α)=(cos⁡α0sin⁡α010−sin⁡α0cos⁡α)\hat{A}(\alpha)=\begin{pmatrix}\cos\alpha&0&\sin\alpha\\0&1&0\\-\sin\alpha&0&\cos\alpha\end{pmatrix} is the rotation around b∗b^* by the angle α\alpha and B^(β)=(1000cos⁡β−sin⁡β0sin⁡βcos⁡β)\hat{B}(\beta)=\begin{pmatrix}1&0&0\\0&\cos\beta&-\sin\beta\\0&\sin\beta&\cos\beta\end{pmatrix} is the second rotation around x′x' by the angle β\beta. Therefore, the local coordinate system of the Ho3+^{3+} ion is characterized by α\alpha and β\beta.

Taking this into account the local susceptibility tensor p^′\hat{p}' that describes the deviation from the Ising behavior, is equal to

p^i=C^i−1p^′C^i,(14)\hat{p}_i=\hat{C}_i^{-1}\hat{p}'\hat{C}_i, \tag*{(14)}

in the crystallographic coordinate system. Here ii is the position number, C^1=B^(β)A^(α)\hat{C}_1=\hat{B}(\beta)\hat{A}(\alpha), C^2/3=B^(β)A^(α)3^±\hat{C}_{2/3}=\hat{B}(\beta)\hat{A}(\alpha)\hat{3}_{\pm}, 3^±\hat{3}_{\pm} represents an additional rotation by ±120∘\pm120^\circ around the trigonal axis. Further arguments in favor of this magnetic configuration can be drawn from the magneto-electric properties of Ho-LGS, which will be analyzed in the subsequent work.

Magnetization data for Ho-langasite with x≈0.015

Angular dependencies of the magnetization of Ho-LGS

Figure S1. Angular dependencies of the magnetization of Ho-LGS (x≈0.015) in the (a) ab∗ab^*, (b) acac, and (c) b∗cb^*c planes at T=5T=5 K and for μ0H=5\mu_0H=5 T. Symbols represent the experimental data, solid lines correspond to the theory.

Magnetization curves of Ho-LGS for different field directions and temperatures

Figure 2 (PDF p. 18). Magnetization curves of Ho-LGS (x≈0.015x \approx0.015) for (a) H∥aH \parallel a, b∗b^{*}, cc, and T=1.9T = 1.9 K, and (b) H∥aH \parallel a, b∗b^{*}, cc, c−39∘b∗c-39^\circ b^{*} and T=5T = 5 K. The direction c−39∘b∗c-39^\circ b^{*} corresponds to the minimum of the angular dependence in the b∗cb^{*}c plane. Symbols represent experimental data, solid lines correspond to the theory.

Relationship between neutron flipping ratio and magnetic susceptibility

This section summaries the work of Gukasov and Brown in Ref. [30]. The scattering intensity for polarized neutrons is given by,

I=N2+2P0⋅(N′M⊥′+N′′M⊥′′)+M⊥2(15)I = N^2 + 2P_0 \cdot\left(N'M'_{\perp} + N''M''_{\perp}\right) + M_{\perp}^2 \tag*{(15)}

Here, NN is the nuclear structure factor with real (N′N') and imaginary (N′′N'') components, while M⊥′M'_{\perp} and M⊥′′M''_{\perp} are the real and imaginary components of the magnetic interaction vector M⊥M_{\perp}, defined as M⊥(k)=k×M(k)×kM_{\perp}(k) = k \times M(k) \times k. P0P_0 is the incident neutron beam polarization vector. The experimentally measured quantity is the flipping ratio,

R=I+I−(16)R = \frac{I^+}{I^-} \tag*{(16)}

where I+I^+ and I−I^- are the scattering intensity of neutrons polarized parallel and anti-parallel to an external magnetic field HH.

In order to relate the flipping ratio to the magnetic susceptibility tensor, we consider a single magnetic ion (aa) with an anisotropic susceptibility resulting from its local environment. For low magnetic fields in the linear regime, the magnetization induced by an external field HH is given by

Ma=χaH.M^a = \chi^a H.

The local magnetization of a symmetrically equivalent atom (b) is then given by,

Mb=χbH=R~χaR~−1H,(17)M^b = \chi^b H = \widetilde{R}\chi^a\widetilde{R}^{-1}H, \tag*{(17)}

where {R~:t}\{\widetilde{R}:t\} is a symmetry operator.

Taking into account the relative positions of the magnetic ions within the unit cell, and assuming a spherical distribution of the magnetic moment around each atom, the magnetization distribution in the unit cell is given by,

M(r)=∑pR~pχaR~p−1H ρ(r−R~pra−tp).(18)M(r) = \sum_p \widetilde{R}_p\chi^a\widetilde{R}_p^{-1}H\,\rho\left(r-\widetilde{R}_p r_a-t_p\right). \tag*{(18)}

Here the sum is over all NgN_g operators {R~p:tp}\{\widetilde{R}_p:t_p\} in the space group. In our analysis we consider a magnetic form factor with a dipolar approximation of the form:

f(k)=⟨j0(k)⟩+(gL−2gL)⟨j2(k)⟩,(19)f(k) = \langle j_0(k)\rangle+ \left(\frac{g_L-2}{g_L}\right)\langle j_2(k)\rangle, \tag*{(19)}

with,

⟨jk(k)⟩=∫0∞r2Rnl2(r)jk(kr) dr.(20)\langle j_k(k)\rangle= \int_0^\infty r^2R_{nl}^2(r)j_k(kr)\,dr. \tag*{(20)}

Here, Rnl=⟨r∣nl⟩R_{nl} = \langle r|nl\rangle are the radial parts of the electronic wavefunction and jk(kr)j_k(kr) are the kk-order Bessel functions. The second term in S9 accounts for the orbital contribution to the magnetic moment. Here, gLg_L is the Landé g-factor calculated for the ground multiplet quantum number at low temperature. In the case of Ho3+^{3+} (featuring 10 ff-electrons) the ground multiplet is 5I8{}^5I_8 (S = 2, L = 6, J = 8) giving a value gL=5/4g_L = 5/4. The values for ⟨Jk(k)⟩\langle J_k(k)\rangle have been computed for trivalent ions such as Ho3+^{3+} from solutions of Dirac-Fock Hamiltonians and are tabulated in the International Table of Crystallography [43].

The corresponding magnetic structure factor is then given by,

M(k)=1Naf(k)∑pR~pχaR~p−1Heik⋅(R~pra+tp)(21)M(k) = \frac{1}{N_a}f(k)\sum_p\widetilde{R}_p\chi^a\widetilde{R}_p^{-1}H e^{ik\cdot(\widetilde{R}_p r_a+t_p)} \tag*{(21)}

Here NaN_a is the number of operators qq for which Rpra+tp=raR_{p}r_a+t_p=r_a, such that NaN_a is the multiplicity of the site aa and its symmetry is determined by the point group QQ, formed by the rotational parts of the set of operators qq. With this, the flipping ratio RR can be expressed in terms of the magnetic susceptibility tensor allowing a least squares refinement of the experimental measurements.

References

  1. [12]N. Lee, Y. J. Choi, and S. W. Cheong, “Magnetic control of ferroelectric polarization in a self-formed single magnetoelectric domain of multiferroic Ba₃NbFe₃Si₂O₁₄,” Appl. Phys. Lett., vol. 104, no. 7, pp. 1–5, 2014, doi: 10.1063/1.4866187.
  2. [13]H. Narita, Y. Tokunaga, A. Kikkawa, Y. Taguchi, Y. Tokura, and Y. Takahashi, “Observation of nonreciprocal directional dichroism via electromagnon resonance in a chiral-lattice helimagnet Ba₃NbFe₃Si₂O₁₄,” Phys. Rev. B, vol. 94, no. 9, 2016, doi: 10.1103/PhysRevB.94.094433.
  3. [14]A. Yu. Tikhonovskii, V. Yu. Ivanov, A. M. Kuzmenko, A. M. Balbashov, Z. Wang, V. Skumryev, and A. A. Mukhin, “Magnetoelectric phenomena in Fe langasites,” Phys. Rev. B, vol. 105, no. 10, p. 104424, Mar. 2022, doi: 10.1103/PhysRevB.105.104424.
  4. [15]H. D. Zhou, B. W. Vogt, J. A. Janik, Y.-J. Jo, L. Balicas, Y. Qiu, J. R. D. Copley, J. S. Gardner, and C. R. Wiebe, “Partial field-induced magnetic order in the Spin-liquid kagomé Nd₃Ga₅SiO₁₄,” Phys. Rev. Lett., vol. 99, no. 23, pp. 1–4, 2007, doi: 10.1103/PhysRevLett.99.236401.
  5. [16]A. Zorko, F. Bert, P. Mendels, P. Bordet, P. Lejay, and J. Robert, “Easy-Axis Kagome Antiferromagnet: Local-Probe Study of Nd₃Ga₅SiO₁₄,” Phys. Rev. Lett., vol. 100, no. 14, p. 147201, Apr. 2008, doi: 10.1103/PhysRevLett.100.147201.
  6. [17]P. Bordet, I. Gelard, K. Marty, A. Ibanez, J. Robert, V. Simonet, B. Canals, R. Ballou, and P. Lejay, “Magnetic frustration on a Kagomé lattice in R₃Ga₅SiO₁₄ langasites with R = Nd, Pr,” J. Phys. Condens. Matter, vol. 18, no. 22, pp. 5147–5153, 2006, doi: 10.1088/0953-8984/18/22/014.
  7. [18]V. Simonet, R. Ballou, J. Robert, B. Canals, F. Hippert, P. Bordet, P. Lejay, P. Fouquet, J. Ollivier, and D. Braithwaite, “Hidden magnetic frustration by quantum relaxation in anisotropic Nd langasite,” Phys. Rev. Lett., vol. 100, no. 23, pp. 1–4, 2008, doi: 10.1103/PhysRevLett.100.0237204.
  8. [19]J. Robert, V. Simonet, B. Canals, R. Ballou, P. Bordet, P. Lejay, and A. Stunault, “Spin- Liquid Correlations in the Nd-Langasite Anisotropic Kagomé Antiferromagnet,” Phys. Rev. Lett., vol. 96, no. 19, p. 197205, May 2006, doi: 10.1103/PhysRevLett.96.197205.
  9. [20]A. Zorko, F. Bert, P. Mendels, K. Marty, and P. Bordet, “Ground State of the easy-axis rare-earth kagome langasite Pr₃Ga₅SiO₁₄,” Phys. Rev. Lett., vol. 104, no. 5, pp. 3–6, 2010, doi: 10.1103/PhysRevLett.104.057202.
  10. [21]A. Zorko, F. Bert, P. Mendels, A. Potočnik, A. Amato, C. Baines, K. Marty, P. Bordet, P. Lejay, E. Lhotel, V. Simonet, and R., “Quantum Tunneling in Half-Integer-Spin Kagome- Lattice Langasites”, arXiv preprint, arXiv:1210.8187.
  11. [22]A. K. Zvezdin, S. S. Krotov, A. M. Kadomtseva, G. P. Vorob’ev, Yu. F. Popov, A. P. Pyatakov, L. N. Bezmaternykh, and E. A. Popova, “Magnetoelectric effects in gadolinium iron borate GdFe₃(BO₃)₄,” J. Exp. Theor. Phys. Lett., vol. 81, no. 6, pp. 272– 276, Mar. 2005, doi: 10.1134/1.1931014.
  12. [23]F. Yen, B. Lorenz, Y. Y. Sun, C. W. Chu, L. N. Bezmaternykh, and A. N. Vasiliev, “Magnetic field effect and dielectric anomalies at the spin reorientation phase transition of GdFe₃(BO₃)₄,” Phys. Rev. B, vol. 73, no. 5, pp. 2–7, 2006, doi: 10.1103/PhysRevB.73.054435.
  13. [24]A.K. Zvezdin, G.P. Vorob’ev, A.M. Kadomtseva, et al., “Magnetoelectric and magnetoelastic interactions in NdFe₃(BO₃)₄ multiferroics”. JETP Lett., vol. 83, pp. 509– 514, (2006). https://doi.org/10.1134/S0021364006110099
  14. [25]A. M. Kuz’menko, A. A. Mukhin, V. Y. Ivanov, A. M. Kadomtseva, S. P. Lebedev, and L. N. Bezmaternykh, “Antiferromagnetic resonance and dielectric properties of rare-earth ferroborates in the submillimeter frequency range,” J. Exp. Theor. Phys., vol. 113, no. 1, pp. 113–120, Jul. 2011, doi: 10.1134/S106377611105013X.
  15. [26]K. C. Liang et al., “Giant magnetoelectric effect in HoAl₃(BO₃)₄,” Phys. Rev. B, vol. 83, no. 18, pp. 2–5, 2011, doi: 10.1103/PhysRevB.83.180417.
  16. [27][27*] A. M. Kadomtseva et al., “Magnetoelectric and magnetoelastic properties of rare-earth ferroborates,” Low Temp. Phys., vol. 36, no. 6, pp. 511–521, 2010, doi: 10.1063/1.3457390.
  17. [28]V. Y. Ivanov, A. M. Kuzmenko, and A. A. Mukhin, “Magnetoelectric effect in ytterbium aluminum borate YbAl₃(BO₃)₄,” JETP Lett., vol. 105, no. 7, pp. 435–441, Apr. 2017, doi: 10.1134/S0021364017070074.
  18. [29]L. Weymann et al., “Unusual magnetoelectric effect in paramagnetic rare-earth langasite,” npj Quantum Mater., vol. 5, no. 1, p. 61, Dec. 2020, doi: 10.1038/s41535-020-00263-9.
  19. [30]A. Gukasov and P. J. Brown, “Determination of atomic site susceptibility tensors from polarized neutron diffraction data,” J. Phys. Condens. Matter, vol. 14, no. 38, pp. 8831–8839, Sep. 2002, doi: 10.1088/0953-8984/14/38/307.
  20. [31]A. Gukasov and P. J. Brown, “Determination of atomic site susceptibility tensors from neutron diffraction data on polycrystalline samples,” J. Phys. Condens. Matter, vol. 22, no. 50, 2010, doi: 10.1088/0953-8984/22/50/502201.
  21. [32]I. A. Kibalin and A. Gukasov, “Local magnetic anisotropy by polarized neutron powder diffraction: Application of magnetically induced preferred crystallite orientation,” Phys. Rev. Res., vol. 1, no. 3, pp. 1–8, 2019, doi: 10.1103/PhysRevResearch.1.033100.
  22. [33]M. J. Harris, S. T. Bramwell, D. F. McMorrow, T. Zeiske, and K. W. Godfrey, “Geometrical frustration in the ferromagnetic pyrochlore Ho₂Ti₂O₇,” Phys. Rev. Lett., vol. 79, no. 13, pp. 2554–2557, 1997, doi: 10.1103/PhysRevLett.79.2554.
  23. [34]J. S. Gardner, M. J. P. Gingras, and J. E. Greedan, “Magnetic pyrochlore oxides,” Rev. Mod. Phys., vol. 82, no. 1, pp. 53–107, 2010, doi: 10.1103/RevModPhys.82.53.
  24. [35]H. Cao, A. Gukasov, I. Mirebeau, P. Bonville, C. Decorse, and G. Dhalenne, “Ising versus XY anisotropy in frustrated R₂Ti₂O₇ compounds as ‘seen’ by polarized neutrons,” Phys. Rev. Lett., vol. 103, no. 5, pp. 1–4, 2009, doi: 10.1103/PhysRevLett.103.056402.
  25. [36]H. Cao, A. Gukasov, I. Mirebeau, P. Bonville, and G. Dhalenne, “Field-induced spin-ice-like order in spin liquid Tb₂Ti₂O₇,” Phys. Rev. Lett., vol. 101, no. 19, pp. 10–13, 2008, doi: 10.1103/PhysRevLett.101.196402.
  26. [37]B. Ouladdiaf, J. Archer, G. J. McIntyre, A. W. Hewat, D. Brau, and S. York, “OrientExpress: A new system for Laue neutron diffraction, ” Physica B: Condensed Matter vol.385, pp. 1052-1054, 2006, doi: 10.1016/j.physb.2006.05.337.
  27. [38]Supplemental material.
  28. [39]A. P. Dudka, “New multicell model for describing the atomic structure of La₃Ga₅SiO₁₄ piezoelectric crystal: Unit cells of different compositions in the same single crystal,” Crystallogr. Rep., vol. 62, no. 2, pp. 195–204, 2017, doi: 10.1134/S1063774517020109.
  29. [40]Y. Yoneda, H. Takeda, T. Shiosaki, and J. Mizuki, “Difference between Local and Average Structures of La₃Ga₅SiO₁₄ Crystal,” Jpn. J. Appl. Phys., vol. 46, no. 10B, pp. 7163–7166, Oct. 2007, doi: 10.1143/JJAP.46.7163.
  30. [41]J. S. Griffith, “Spin hamiltonian for even-electron systems having even multiplicity,” Phys. Rev., vol. 132, no. 1, pp. 316–319, 1963, doi: 10.1103/PhysRev.132.316.
  31. [43]Brown, P. J., “International Tables for Crystallography” vol. C, Section 4.4.5, pp. 454-461, 2006, doi:10.1107/97809553602060000594.

Paper details

vlnc:W_92ezpadm
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents