The perturbation of a magnetostatic field by a toroid made of a homogeneous anisotropic magnetic material was formulated using the solutions of the Laplace equation in the toroidal coordinate system. That was straightforward in the region outside the toroid, but an affine coordinate transformation had to be employed inside the toroid. The coefficients of the series expansion of the perturbation potential in terms of appropriate toroidal basis functions were related to the coefficients of the series expansion of the source potential in terms of appropriate toroidal basis functions by a transition matrix. As a result of the solution of this novel problem, the consequences of material anisotropy on perturbing the magnetostatic field are clearly evident in the region near the toroid.
Introduction
In a source-free region, the magnetostatic problem requires solution of the Gauss equation ∇⋅B(r)=0 and the Ampere equation ∇×H(r)=0, where H(r) is the magnetostatic field and B(r) is the magnetostatic flux density. The Ampere equation allows the definition of a scalar magnetic potential Φ(r) through
H(r)=−∇Φ(r).(1)
Suppose that B(r)=μH(r) in a certain region filled with a homogeneous isotropic medium with permeability scalar μ. Then, the scalar magnetic potential satisfies the Laplace equation
∇2Φ(r)=0(2)
in that region. The Laplace equation can be solved by the method of separation of variables in thirteen three-dimensional coordinate systems: Cartesian, confocal paraboloidal, bispherical, toroidal, and ten specializations of the confocal ellipsoidal system [1, 2]. Accordingly, magnetostatic boundary-value problems can be solved, at least in principle, so long as the permeability scalar is piecewise uniform with discontinuities only across boundaries on which one of the three coordinates is invariant.
What if the permeability is not the scalar μ in a certain region filled with a homogeneous medium, but the dyadic μ? In other words, the medium is anisotropic [3, 4]. Then we get
∇⋅[μ⋅∇Φ(r)]=0(3)
in that region, so that Φ(r) is not a solution of the Laplace equation in that region. The solutions of Eq. (2) can still be applied though an affine coordinate transformation.
For the sake of illustration, in this paper we have focused on the perturbation of a magnetostatic field Hsource by a toroid made of a homogeneous anisotropic material. Perturbation by anisotropic magnetic toroids does not appear to have been treated heretofore, despite the fact that magnetic anisotropy is exhibited by both crystalline materials [4, 5, 6] and effectively homogeneous composite materials [7, 8]. Toroids made of isotropic magnetic materials are commonplace as cores in transformers and recording heads [5], and anisotropic magnetic toroids are expected to perform better by delivering optimal spatial distributions of Φ(r).
The plan of this paper is as follows. § 2 presents the formulation of the boundary-value problem and the solution procedure. § 3 provides and discusses illustrative numerical results. The paper concludes with some remarks in § 4. Throughout the paper, vectors are in boldface, unit vectors are decorated by carets, dyadics are double underlined, and column vectors as well as matrices are enclosed in square brackets.
Boundary-Value Problem
Toroidal coordinates
Figure 1 shows a toroid with mean radius Ro and cross-sectional radius ro, with the constraint Ro>ro. These two dimensions lead to the definition of the two constants
Figure 1. Toroid with mean radius Ro and cross-sectional radius ro.
ca=+Ro2−ro2=csch−1(ro/c)(4)
that are needed to define an appropriate toroidal coordinate system r≡(ξ,η,ϕ).
Cartesian coordinates {x,y,z} are given in terms of toroidal coordinates {ξ,η,ϕ} by [9]
x=ccoshξ−cosηsinhξcosϕ,(5)
y=ccoshξ−cosηsinhξsinϕ,(6)
and
z=ccoshξ−cosηsinη.(7)
Toroidal coordinates are given in terms of Cartesian coordinates as
ξ=log(x2+y2−c)2+z2(x2+y2+c)2+z21/2,(8)
η=tan−1(x2+y2+z2−c22cz),(9)
and
ϕ=tan−1(xy).(10)
The coordinate ξ=a is constant on the surface of the toroid, the interior of the toroid being denoted by
V−:{(ξ,η,ϕ)∣ξ>a,η∈[0,2π),ϕ∈[0,2π)}(11)
and the exterior by
V+:{(ξ,η,ϕ)∣ξ<a,η∈[0,2π),ϕ∈[0,2π)}.(12)
Formulation
Let the toroid be made of an anisotropic magnetic material described by the constitutive relation
B(r)=μH(r)=μ0μrA⋅A⋅H(r),ξ>a,(13)
where μ0 denotes the permeability of free space; the scalar μr>0; and the diagonal dyadic [3, 4]
A=αx−1x^x^+αy−1y^y^+z^z^(14)
contains the scalars αx>0 and αy>0. The exterior region V+ is taken to be vacuous with constitutive relation
B(r)=μ0H(r),ξ<a.(15)
Potential in region V+
Equation (2) applies in V+. Its solution in the toroidal coordinate system yields [2, 9]
contain i=−1, Γ(⋅) as the Gamma function, and 2F1(⋅,⋅,⋅;⋅) as the hypergeometric function [9, 10]. The coefficients Ajkℓn are associated with terms that are regular as ξ→0 whereas the coefficients Bjkℓn are associated with terms that are regular as ξ→∞.
is the perturbation potential. If the toroid were to be absent, Eq. (16) would hold for all r with Bjkℓn≡0∀{j,k,ℓ,n}. We proceed with the assumption that the coefficients Ajkℓn are known but the coefficients Bjkℓn are not. Furthermore, Eq. (22) is required to hold in some sufficiently large region that contains the toroidal region V− but not the region containing the source of Φsource(r).
Source potential
Sources of two types are considered for illustrative numerical results provided in § 3: (i) a uniform-magnetostatic-field source and (ii) a point-magnetic-dipole source.
For the uniform-magnetostatic-field source, we have Hsource=−Usxx^−Usyy^−Uszz^, with Usx, Usy, and Usz as constants, and hence the source potential is written as
Φsource(r)=Usxx+Usyy+Uszz.(24)
On transforming Eq. (24) to toroidal coordinates, we obtain
To find the coefficients Ajkℓn in Eq. (22) for the uniform-magnetostatic-field source, we equate Eqs. (22) and (25), multiply both sides by Ψℓn1jk(r)(coshξ−cosη1), set ξ=a, and integrate over η∈[0,2π) and ϕ∈[0,2π). This gives
For the point-magnetic-dipole source of magnetic moment m=msxx^+msyy^+mszz^, where msx, msy, and msz are constants, located at ro≡(ξo,ηo,ϕo), where ξo<a, ηo∈[0,2π), and ϕo∈[0,2π), the source coefficients can be obtained with the use of the bilinear expansion of the Green function in toroidal coordinates as [11]
where ∣m∣>0 and the unit vector m^=m/∣m∣ defines the orientation of the point magnetic dipole. See the Appendix for additional details.
Potential in region V−
Inside the magnetic toroid, the potential Φ(r) does not obey the Laplace equation; instead,
∇⋅[A⋅A⋅∇Φ(r)]=0,ξ>a.(28)
In order to solve this equation, let us make an affine coordinate transformation from r to r1≡(ξ1,η1,ϕ1) as follows [12]:
r1=A−1⋅r.(29)
Since αx>0 and αy>0, the bijective transformation (29) maps a sphere into an ellipsoid. That ellipsoid’s principal axes are the same as the Cartesian unit vectors affixed to the sphere. In such a situation, the variables (ξ1,η1,ϕ1) are given by
The unknown coefficients Cjkℓn have to be determined.
Boundary conditions
With the assumption of the interface ξ=a being current-free, the tangential component of the magnetostatic field must be continuous across the interface ξ=a[3]. Therefore, the potential must be continuous across both interfaces; hence,
where the unit vector ξ^=coshξ−cosη1−coshξcosη(x^cosϕ+y^sinϕ)−z^coshξ−cosηsinhξsinη
Transition Matrix
On multiplying both sides of Eqs. (34) and (35) by Ψℓn1j′k′(r)(coshξ−cosη1), setting ξ=a, and integrating over η∈[0,2π) and ϕ∈[0,2π), the following sets of simultaneous algebraic equations are obtained:
After truncating the indexes ℓ, ℓ′, n, and n′ so that only ℓ∈[0,N], ℓ′∈[0,N], n∈[0,N], and n′∈[0,N] are considered with N>0, Eqs. (36) and (37) can be put together in matrix form symbolically as
[I][A]+[J][B]=[K][C](44)
and
[L][A]+[M][B]=[N][C].(45)
The system of Eqs. (44) and (45) can be written as
Then, the column vector [B] containing the coefficients in the truncated series representing the perturbation potential is related to the column vector [A] containing the coefficients in the truncated series representing the source potential by
[B]=[T][A].(48)
In this equation, the transition matrix [T], capturing the perturbational characteristics of the anisotropic magnetic toroid, is given by
[T]=[U][S]−1(49)
Illustrative Numerical Results and Discussion
Preliminaries
A Mathematica{sup}TM program was written to calculate the transition matrix $[\mathcal{T}]$ of the anisotropic magnetic toroid for a chosen truncation order $N$. The column vector $[\mathcal{B}]$ containing the perturbation-potential coefficients was then calculated using {xref}Eq. (40). Also, the column vector $[\mathcal{C}]$ containing the internal-potential coefficients was calculated using {xref}Eq. (38). Thereafter, the perturbation potential and the internal potential were calculated using {xref}Eqs. (15) and (25)`, respectively. Finally, the magnetic field everywhere was calculated using
Validation of the program was performed by solving independently the problem of a toroid composed of an isotropic magnetic material and implementing this solution in a separate Mathematica{sup}TM` program. Comparison between the original program (for the anisotropic magnetic toroid) against the program for the isotropic magnetic toroid was done by setting in the original program αx=αy=1, with a being arbitrary. Excellent agreement in the results of the two programs was obtained for several different values of μr, Ro, and ro. Furthermore, on the toroidal surface, the source-potential expansions for the cases of the uniform-magnetostatic-field source and the point-magnetic-dipole source were compared against their closed-form expressions and excellent agreements were obtained.
In general, the perturbation and internal potentials depend on the
mean radius Ro and cross-sectional radius ro;
permeability μr;
anisotropy parameters αx and αy; and
source potential characteristics.
We chose
μr=αx−2+αy−2+13μave,(51)
where μave=500. Note that μr=μave when the toroid is made of an isotropic material (i.e., A=I, and hence αx=αy=1). Also, we set Ro/ro=5/3 and ro=3 cm. For later use, we define R=Ro+ro.
A convergence test was carried out with respect to N, by calculating the perturbation and internal potentials at several points in space as N was incremented by unity. The iterative process of increasing N was terminated when the perturbation and internal potentials converged within a preset tolerance of 1%. To demonstrate the convergence procedure, we defined the relative error
We show Υ(N,r) as well as the computational time in relation to N∈[0,6] in Table 1, for a toroid made of a biaxial magnetic medium with αx=1.1 and αy=1.2. These calculations were performed on the United Arab Emirates University High-Performance Computing system consisting of three nodes, each of which is equipped with Intel Xeon E5-2690 v4 @ 2.60 GHz and 256 GB RAM. The source was chosen to be a point magnetic dipole located at the origin (i.e., ξo=0 and ηo=π) and oriented along the z axis (i.e., m^=z^), while the observation point was selected to be r≡{ξ=0.9a,η=1,ϕ=1}. Per Table 1, N=6 is sufficient for satisfying the aforementioned convergence criterion. The corresponding procedure was also applied to the internal potential, and the results were found to be quantitatively similar.
N
Υ(N,r) (%)
Computational time (min)
0
-
0.133
1
100
6
2
93
48
3
4.86
150
4
8.86
514
5
3.88
1080
6
0.31
2055
Table 1. Relative error Υ(N,r) and computational time to determine Φpert(N,r) as a function of N for r≡{ξ=0.9a,η=1,ϕ=1}, when a toroid made of an anisotropic material (μave=500, αx=1.1, αy=1.2) is exposed to a point magnetic dipole m∥z^ located at the origin.
Next, we present numerical results for the variations of the magnetic field everywhere for different anisotropy parameters αx and αy of the toroid. The excitation is due to either a uniform-magnetostatic-field source or a point-magnetic-dipole source.
Uniform-magnetostatic-field source
Figures 2, 3, and 4 correspond to the cases Hsource=x^Am−1, Hsource=y^Am−1 and Hsource=z^Am−1, respectively. In each of these three figures, vector plots of the total magnetostatic field H versus x∈[−2,2]R, y∈[−2,2]R, and z∈[−2,2]R are presented, with the different panels corresponding to (a) αx=1 and αy=1, (b) αx=1.2 and αy=1.2, and (c) αx=1.1 and αy=1.2. Near the toroid, the direction of the total magnetostatic field changes significantly when Hsource is parallel to the x and y axes, but exhibits only minor variations when Hsource is parallel to the z axis. By close inspection, one can also notice that the direction of the total magnetostatic field on the toroidal surface is affected by the change in the anisotropy parameters αx and αy.
Figure 4. As in Figure 2, but for Hsource=z^Am−1.
Point-magnetic-dipole source
Now, we present the case of a point-magnetic-dipole source with the dipole located at the origin (i.e., ξo=0, ηo=π). Figure 5, Figure 6, and Figure 7 correspond to the cases m=x^, m=y^, and m=z^, respectively. In each of these three figures, vector plots of the total magnetostatic field as a function of x∈[−2,2]R~, y∈[−2,2]R~, and z∈[−2,2]R~ are presented, with the different panels corresponding to (a) αx=1 and αy=1, (b) αx=1.2 and αy=1.2, and (c) αx=1.1 and αy=1.2. In all figures, a closed-loop behavior of the total magnetostatic field lines is observed as is characteristic for a dipole excitation. In the regions near the toroid, the effect of anisotropy on the total magnetostatic field is evident in the sense that the total magnetostatic field changes substantially when anisotropy is introduced to the magnetic material of the toroid.
In general, by examining Figure 2–Figure 7, it is observed that anisotropy of the toroid has a much stronger effect on the total magnetostatic field when the source is the field of a point magnetic dipole compared to when it is a uniform magnetostatic field.
Figure 2. H(x,y,z) versus x∈[−2,2]R, y∈[−2,2]R, and z∈[−2,2]R for Hsource=x^Am−1 when (a) αx=1 and αy=1, (b) αx=1.2 and αy=1.2, and (c) αx=1.1 and αy=1.2.
Concluding Remarks
We have formulated a boundary-value problem to determine the perturbation of a magnetostatic field by a toroid made of a homogeneous anisotropic magnetic material. Whereas the solutions of the Laplace equation in the toroidal coordinate system were employed to express the total potential (and, therefore, the total magnetostatic field) as a series outside the toroid, a series of the toroidal solutions of the Laplace equation in a transformed space had to be used to express the potential inside the toroid. Standard boundary conditions were then enforced on the surface of the toroid, and their orthogonality properties on the toroidal surface were used to determine a transition matrix. This transition matrix relates the perturbation-potential coefficients to the source-potential coefficients. Illustrative numerical results show that anisotropy of the toroid material has significant effects on the magnetostatic field in the region near the toroid.
Appendix
The closed-form free-space Green function
G(r,ro)=∣r−ro∣1,(53)
where r≡(ξ,η,ϕ) and ro≡(ξo,ηo,ϕo), can be expanded in bilinear form in toroidal coordinates as [11]
The bilinear form of the Green function presented in Eq. (54) was originally derived by Morse and Feshbach [9]. However, as pointed by Bates [11], their expression does not contain the constant c and involves typographical errors. After evaluating the source potential due to a point magnetic dipole on the interface ξ=a using (i) Eq. (53), (ii) Eq. (54), and (iii) the bilinear form of the Green function reported by Morse and Feshbach [9], calculations corresponding to various dipole orientations m^ (not shown) have confirmed that Eq. (54) is the correct bilinear form of Eq. (53).
Acknowledgement. AL is grateful to the Charles Godfrey Binder Endowment at Penn State for supporting his research activities from 2006 to 2024. H. M. Alkhoori acknowledges the United Arab Emirates University for supporting this work.
References
[1]L. P. Eisenhart, Separable systems in euclidean 3-space, Physical Review 45, (1934) 427–428.DOI
[2]P. Moon and D. E. Spencer, Field Theory Handbook: Including Coordinate Systems, Differential Equations and Their Solutions, 2nd ed., (Berlin, Germany, Springer 1971).
[3]H. C. Chen, Theory of Electromagnetic Waves: A Coordinate-Free Approach (New York, USA, McGraw–Hill 1983).
[4]T. G. Mackay and A. Lakhtakia, Electromagnetic Anisotropy and Bianisotropy, 2nd ed. (Singapore, World Scientific 2020).
[5]M. McCaig, Permanent Magnets in Theory and Practice (London, United Kingdom, Pentech 1977).
[6]J. I. Gersten and F. W. Smith, The Physics and Chemistry of Materials (New York, USA, Wiley 2001).
[7]W. H. Meiklejohn and C. P. Bean, New magnetic anisotropy, Physical Review 105 (1957) 904–913.
[8]M. T. Johnson, P. J. H. Bloemen, F. J. A. den Broeder, and J. J. de Vries, Magnetic anisotropy in metallic multilayers, Reports of Progress in Physics 59 (1996) 1409–1458.DOI
[9]P. M. Morse and H. Feshbach, Methods of Theoretical Physics, Vol. II (New York, USA, McGraw–Hill 1953).
[10]M. Abramowitz and I. A. Stegun, Handbook of Mathematical Functions (New York, USA, Dover Publications 1972).
[11]J. W. Bates, On toroidal Green’s functions, Journal of Mathematical Physics 38 (1997) 3679–3691.DOI
[12]A. Lakhtakia, N. L. Tsitsas, and H. M. Alkhoori, Theory of perturbation of electrostatic field by an anisotropic sphere, Quarterly Journal of Mechanics and Applied Mathematics 74 (2021) 467–490; correction: 75 (2022) 125.
[13]A. Zangwill, Modern Electrodynamics (New York, USA, Cambridge University Press 2013).DOI
Isaac M. Craig, Madeline Van Winkle, Catherine Groschner, Kaidi Zhang, Nikita Dowlatshahi, Ziyan Zhu, Takashi Taniguchi, Kenji Watanabe, Sinéad M. Griffin, D. Kwabena Bediako ·