WorksPhysics and Astronomy
Openable Force-Balanced Halbach Magnets: From Fibonacci Sphere Simulations to Icosahedral Realizations
Abstract
A long-standing goal in magnet design is to completely surround a volume of highly homogeneous magnetic field with permanent magnets while maintaining practical access to that volume. In this work, we present a theoretical and experimental investigation of mechanically accessible spherical magnets in Halbach configuration that can be opened with minimal or vanishing force. Focusing on dipolar Halbach spheres composed of discrete magnetic subunits, we derive conditions for force-free opening along specific cutting planes. These conditions define a continuous set of geometries for which tensile magnetic forces cancel, leaving only shear components, enabling mechanically effortless opening. The theoretical predictions are validated experimentally using icosahedral approximations of the Halbach sphere, for which both opening forces and magnetic field properties are measured. The results demonstrate that excellent field homogeneity can be preserved while reducing opening forces by orders of magnitude. Although discussed in detail for the dipolar case, the theoretical framework is general and applicable to higher-order multipole Halbach systems. Finally, the concepts are extended to spherocylindrical Halbach configurations, highlighting their potential for large-volume, highly homogeneous, and mechanically accessible permanent-magnet systems for magnetic resonance
Graphical Abstract
Figure 1.
Highlights
Numerically computed cutting planes enable force-free opening of discrete spherical Halbach magnets.
The theoretical framework is general and directly applicable to higher-order multipole Halbach configurations.
Icosahedral arrangement of magnets demonstrate reduced opening forces while preserving magnetic field homogeneity.
Spherocylindrical geometries enable scalable systems with larger, expandable accessible volumes of highly homogeneous fields.
Very efficient design in terms of magnetic material usage relative to the generated homogeneous field volume.
Overcomes the limited accessibility of spherical magnets, enabling general applications.
Enables compact, energy-efficient permanent-magnet systems for magnetic resonance and related applications.
Introduction
Magnetic resonance techniques such as nuclear magnetic resonance (NMR) and electron spin resonance (ESR) rely on the application of strong, spatially homogeneous static magnetic fields. In NMR in particular, magnetic field homogeneity directly limits spectral resolution, while field strength determines both sensitivity and achievable resolution. Although superconducting magnets dominate high-field laboratory systems, permanent magnets have become increasingly attractive for specialized applications in recent years [1].
Permanent magnets offer two decisive advantages. First, they deliver exceptionally high magnetic field strength per unit volume: modern rare-earth materials (e.g., NdFeB) can produce flux densities comparable to those generated by coils requiring thousands of ampere-turns and, in the same confined volume, incurring substantial resistive losses—often on the order of hundreds of watts. In contrast, permanent magnets operate entirely passively, requiring neither electrical power nor active cooling to maintain the field. Together, these characteristics enable compact, energy-efficient and portable magnetic resonance systems suitable for mobile and field-deployable applications [1].
Achieving strong and homogeneous magnetic fields with permanent magnets requires carefully designed arrangements that optimally superimpose the fields of individual magnet elements. The Halbach array [2] provides an elegant and well-established solution to this problem. There is a great deal of literature on the design and construction of Halbach cylinders, as summarized in a recent review [3].
In an ideal realization, a Halbach magnet would be fabricated from a magnetic material with continuously varying magnetization directions—a spatial waveform whose frequency determines the polarity and distribution of the resulting field. Since such materials are not readily manufacturable, practical implementations rely on discretization into multiple magnetic subunits, each with a uniform magnetization direction, arranged to optimize field strength and/or homogeneity [4]. This modular construction not only approximates the ideal field distribution, but also allows, in principle, relative motion between the discrete subunits.
This relative motion has been exploited in several ways. One approach uses the rotation of two or more concentric Halbach assemblies to add or subtract magnetic fields, thus tuning the overall field strength [3, 5, 6, 7]. Another possibility is to mechanically open the magnet structure, for example, by means of a hinge, which ideally requires no force [8]. In that publication, and subsequently in [3], force-free angles were analyzed for cylindrical Halbach arrangements composed of magnetic dipoles.
Shortly thereafter, similar concepts were proposed for spherical Halbach configurations [9, 10]. However, only recently has it become possible to realize spherical Halbach magnets approximated by regular (Platonic) and semi-regular (Archimedean) solids [11]. That work demonstrated that the highest magnetic field homogeneity is associated with icosahedral symmetry.
Despite their excellent field properties, the practical use of spherical Halbach magnets has long been considered severely constrained, because of the limited access to their interior or the large magnetic forces encountered when opening the structure. Depending on geometry and magnet strength, opening forces can reach several kilonewtons, posing substantial challenges in handling, safety, and mechanical design. Previous attempts to mitigate this problem have included introducing apertures or slots into the magnet structure or superimposing magnetized regions with spatially varying remanence [12, 13]. Although these approaches can provide limited access, they generally compromise field homogeneity, increase stray fields and weight, or require complex and impractical fabrication processes.
In the present work, the mechanical opening of discretized spherical Halbach magnets is investigated both theoretically and from a practical perspective. Although the discussion focuses on the dipolar Halbach sphere, the underlying theory is general and applies equally to higher-order multipole magnet systems. Guided by the theoretical predictions, icosahedral magnet arrangements were constructed, and both the forces required to open them and the resulting magnetic field properties are experimentally characterized.
Finally, the concepts developed here are extended to a combination of Halbach cylinders and hemispheres that form a spherocylindrical magnet (capsule-shaped) [3, 14]. In principle, such magnets can be fabricated with arbitrary lengths and are therefore capable of enclosing extended objects (e. g., humans) while providing highly homogeneous magnetic fields over a large fraction of the internal volume. Access to the interior of such a magnetic “sarcophagus” can be easily achieved by opening the structure at an appropriate angle, ideally requiring only minimal force –or even exploiting a controlled repelling force to compensate for the weight of the moving section.
This represents something of a magnet designer’s ideal: a volume of highly homogeneous magnetic field, fully enclosed by magnets yet entirely accessible, and maximally efficient in terms of magnet mass relative to the volume of the homogeneous field.
Theory
Separation force of discretized Halbach spheres
As real Halbach spheres will always feature a degree of discretization with smaller or larger permanent magnets, the basic concept of a force-free cut can be illustrated by considering an arrangement of a sufficiently large number of point dipoles located on a sphere with radius .
As an example, an arrangement of dipoles distributed on the surface of a sphere is shown in fig:1a. The orientation of the field in the center of the sphere is indicated by and is oriented along the -axis. The normal of the cutting plane,
forms the polar angle with , while the azimuthal angle is in this drawing. The cutting plane is allowed to be shifted along , where is the distance of the cutting plane from the center of the sphere.
To calculate forces between the magnets on different sides of the cutting plane we recall that the force between two magnetic dipoles with magnetic moments and , separated by the vector , is given by
where and denotes the permeability of free space [15]. Applying the gradient operator then gives (for example, with as the unit vector of and as its magnitude) [16]
The force between the two clusters separated by the cutting plane can then be obtained as the sum of such dipole-dipole interactions, and the result is illustrated in Fig. 1. Here we consider dipoles located on a Fibonacci lattice [17], which provides an approximately uniform equal-area distribution of points on the sphere. Interactive explorations of its magnetic properties are provided in [18, 19]. The locations of the dipoles are shown as dots, cyan below the cutting plane, and magenta above. They are oriented according to the Halbach condition, namely the polar orientation angle is twice the polar angle of its location. Every second dipole orientation is indicated as an arrow.
In addition, an example of a cutting plane is shown. The distance of this plane from the center of the sphere is measured in units of its radius , and in this drawing.
An overview of the force component between the two spherical caps separated by the cutting plane is given in fig:1d. Attractive forces are shown in blue and repelling forces are shown in red. For a qualitative understanding of the sign change of the force, it helps to consider the line along and to interpret each half sphere as a single dipole. For we are dealing with 2 parallel dipoles located one behind the other - this yields an attractive force. For the two parallel dipoles are located side by side - they repel each other. The points of interest, where the force is zero, must lie somewhere between these extremes. They are indicated by the solid green line. The kinks in that line occur whenever the cutting plane crosses a (point) dipole. The dashed line denotes zero force for 20,000 magnets, which can be considered as an approximation of a continuous distribution of magnetization, an argument which is elucidated in further detail below.
Figure 1 (PDF p. 5). The force to open a spherical Fibonacci arrangement of 300 dipoles. (a) Only every second dipole is indicated by an arrow pointing along . Their arrangement forms a homogeneous field indicated by (magenta), where red arrows are almost parallel to , blue antiparallel, green perpendicular. The vector (cyan) is normal to the cutting plane (cyan circle). (b) is numerically determined (red dots). The parameters of the fitted solid gray line are given in the legend. (c) for three different angles : attracting (blue), almost forceless (green), repelling (red). The lines are smoothing splines to guide the eye. (d) : the green solid line denotes . The zero-force line for 20,000 dipoles is dashed green.
The line along is shown in quantitative detail in Fig. 1b. The force is given in units of
Thus, is the sum of all the magnitudes of dipole moments located on the surface of a hemisphere. By this scaling, one obtains numbers for the force that are fairly independent of the number of dipoles considered. The numerically calculated forces are indicated by red dots. For comparison, a curve
is also shown in the plot, where is a fit parameter. This is a purely empirical ansatz, inspired by the force between two parallel dipoles, where would be 3. The fact that we are getting 2.6 instead indicates that the simplified model of two point dipoles interacting at a fixed distance—which does explain the qualitative features—is not sufficient for a quantitative description.
Figure 1c shows three line cuts through the contour shown in Fig. 1d. At a cutting angle of , we have an attractive force, which becomes weaker with increasing distance of the cutting plane from the center. For the force is still below zero for , but slightly above for larger values of . At , the force is clearly repulsive, becoming weaker with increasing . The solid smoothing splines in this plot are just intended as guides for the eye. Note that the data points are not perfectly symmetric with respect to the line because the Fibonacci distribution of the dipoles does not have exact point symmetry with respect to the center of the dipole cluster.
As shown in Fig. 1d, the force is much more sensitive to changes in the tilt angle than to variations in the height of the cutting-plane , but the tendency that the force is reduced with increasing – a reduced opening aperture – is visible.
As a side remark on the Fibonacci distribution of permanent magnets, it is noticeable that it is also a practical approach to achieve homogeneous magnetic fields inside a sphere. Although it does not offer a clear symmetry of the magnetic field around the center and is thus inferior to the fourth-order saddle points provided by the configurations with icosahedral symmetry, it can outperform those configurations once a finite deviation from the center field is accepted. To give an example: The volume with a deviation from the center is larger for a Fibonacci arrangement of 120 magnets than for a distribution of those magnets at the vertices of a truncated icosidodecahedron. However, the latter arrangement wins if only deviations on a ppm-level are acceptable [18].
A smaller number of magnets is clearly more practical than the 300 dipoles shown above. 12 magnets, located at the vertices of an icosahedron, and oriented in the spherical Halbach arrangement, form a very homogeneous magnetic field in their center, precisely speaking a saddle point of fourth order [11]. Thus, they constitute a very effective and useful spherical arrangement that merits investigation of its opening force. The geometry is shown in Figure 2a. Note that the position of the vertices is chosen to preserve symmetry with respect to the -, - and -plane.
Figure 2. The force to open an icosahedron. (a) Same as Fig. 1a, for 12 magnets in an icosahedral arrangement. (b) Same as Fig. 1d. (c) Same as Fig. 1b. .
The opening force for this configuration is shown in Figure 2b. The fact that this is a function with jump discontinuities is much more prominent here compared to the Fibonacci arrangement with 300 magnets given above. The most practicable opening plane is certainly the one with , a great circle that cuts the sphere exactly into two hemispheres. It should be noted that the angle of a force-free opening is smaller here compared to the configuration discussed above, namely . The force near this angle crosses the zero line with a finite slope of about as shown in Figure 2c. The unsteady jump near is caused by the fact that two dipoles exchange their side with respect to the cutting plane here.
Having noticed that the force-free opening angle increases to for 300 dipoles, and to for 20000 dipoles, one can speculate that it reaches the magic angle in the limit of a homogeneous Halbach magnetization of the sphere, although there is no mathematical proof at this point, instead we can only present numerical evidence in Figure 3.
Figure 3. The opening angle for Fibonacci spheres increases with the number of dipoles. (a) The dots (magenta) are numerically obtained. The fitting function (blue) with two parameters is given in the legend. The fitting range is indicated by a solid line, and the extrapolated part is dashed. The asymptotic value obtained from that fit is indicated as the red dashed line. (b) This inset shows on a log-log plot. (c) Same data shown as function of . In this case, is revealed as the crossing of the asymptotic straight line with the line .
Figure 3 shows that the cutting angle increases with increasing number of dipoles. Although one might think that a distribution of 20000 vertices is already a very good approximation of a homogeneous Halbach distribution of magnetization, the plot clearly shows that we are still dealing with an error on a percentage level here. Specifically speaking, the deviation of the cutting angle from the asymptotic one according to the fit is .
Figure 3b (the inset of Figure 3a) confirms that the asymptotic range is reached for , and shows the power -0.5 by the asymptotic slope in this log-log plot. Once this power is known, the asymptotic cutting angle can also be determined graphically, by plotting as a function of , as done in Figure 3b. The asymptotic value is read from the crossing of with the zero line.
We take the study shown in Figure 3 as convincing evidence that the asymptotic value of the cutting angle is the magic angle:
This result is consistent with interpreting both hemispheres as point dipoles of equal strength and parallel magnetization. According to Eq. (2), their force would depend on the cutting angle like . This term crosses zero at the magic angle. However, it must be noted that this argument might be useful for memorizing the result, but it is not exact because it ignores the forces that arise from higher-order moments of the hemisphere.
A technical note on the numerical summation of forces is useful here. For the example shown in Fig. 1a, there are 150 × 150 pairs to add. If only one dipole is left on one side of the cutting plane, by choosing a sufficiently large , the number of interactions reduces to 299 in this particular case. A direct summation of forces is certainly possible for a small number of dipoles, but it might cause numerical problems for larger numbers. The largest number of dipoles we considered here is 20000, as shown in Figure 3. In that case pairs have to be added, where the largest distance is 2 and the smallest distance about 0.02, thus their distances differ by two orders of magnitude. However, due to the strong dependence on distance, namely , this means adding up numbers, where the force terms differ by 8 orders of magnitude. This problem can be reduced slightly by summing up potentials, the term within the bracket in eq. 1, because the potential drops with only the third power of the distance. The gradient is then obtained numerically by the difference of two potentials calculated at a small distance.
Force-Balanced Halbach spherocylinders
Spherocylinders are Halbach cylinders closed by spherical caps to minimize the distortions of the homogeneous field at their ends, as described in [3, 14], and illustrated in
To find the appropriate radius for the end caps, it helps to recall that an ideal Halbach spherical shell with an inner radius and an outer radius generates a magnetic field of [13]
that is 4/3 stronger than that of an infinitely long Halbach cylinder with the same inner and outer radii ( and ). Thus, a reduction of the outer sphere radius to
makes both fields equal. If we now cut a long cylindrical shell into two semi-infinite ones, the field at the end is reduced by a factor 1/2. The same is true for a sphere cut into 2 hemispheres. Using a hemisphere as an end cap for the cylinder, the superposition of both fields retains the original strength of the infinite cylinder here. This argument is strict only at the cutting plane of the cylinder, it does not apply for every distance from that plane. Moreover, it only represents an upper bound as a finite-length Halbach cylinder produces weaker fields. Depending on the length of the cylindrical section, the outer diameter of the hemisphere can be adapted to minimize the end effects. The corresponding correction factors are given in [3], Appendix C. The design of discrete approximations of a spherical geometry needs further modifications as discussed below.
The opening angle for Halbach cylinders has been discussed in Ref. [8]. Their approach is based on the fact that the total force is dominated by nearest-neighbor interactions in the limit of an infinitely thin cylinder. This allows the force-free opening angle to be determined in two limiting cases: the infinitely long and the very short cylinder. For infinitely long cylinders the force between two parallel line dipoles is proportional to , where is the angle between their magnetic moments and their separation vector. The force vanishes at , which is the analog of the magic angle for the line dipoles. Following the idea behind Eq. (5) of Ref. [8], one gets a force-free opening angle of in this case of an infinitely long cylinder. For an infinitely short cylinder (a ring), the force between parallel point dipoles is proportional to . Following the arguments in [8], the opening angle is , which according to Figure 5 is the same as that obtained for homogeneous spherical shells. Combining both arguments, one can conclude that the finite size cylinder will have a force-free opening angle in the range . That range could be matched by a Fibonacci hemisphere, provided that more than dipoles are used. The truncated icosahedron discussed in Figure 5 with its optimal opening angle of would also fit this range nicely with only 30 magnets.
Figure 5. The optimal opening angle for four magnet configurations: (a) icosahedron, (b) dodecahedron, (c) truncated icosahedron, (d) truncated icosidodecahedron. The cutting angle is shown as a colored line. The color provided in the color bar indicates the distance from the cutting plane to the nearest magnet. The blue circle indicates the optimal choice, with the corresponding numbers in the legend.
This theoretical framework of continuously magnetized magnetic surfaces outlines a design space for permanent-magnet assemblies that balances mechanical accessibility with the magnetic field quality required for high-resolution magnetic resonance applications. We consider it as a useful guide for designs using discrete magnets.
Practical implications
In principle, there is a wide range of possible cutting heights for a force-free opening. Choosing the largest radius of the opening disk by setting seems to be the most practical, both in terms of interior accessibility and mechanical stability. It must be stressed that for the so called force-free opening angle there are shear forces, which must be accommodated by appropriately robust hinges. Furthermore, and more importantly, as noted in the Introduction, magnetic materials with such geometries and continuously varying magnetization directions are not available. Therefore, cylindrical and spherical Halbach magnets must be approximated by using discrete segments with uniform magnetization, each positioned and oriented appropriately. In practice, condition will not be met exactly. However, even partial force compensation can reduce opening forces by one or two orders of magnitude, significantly enhancing mechanical safety and usability while preserving magnetic field strength and homogeneity.
When the opening plane overlaps with a magnet, one can make use of the possibility to turn the cutting plane around the -direction by an azimuthal angle , as discussed in Appendix G in [3] for rotational symmetric arrangements. However, there is no strict rotational symmetry for spherical arrangements of discrete magnets, so each configuration needs individual analysis. We present such a study for four different geometries in Figure 5.
For this illustration, we have chosen the icosahedron formed by 12 magnets at its vertices, the dodecahedron (20 magnets), the truncated icosahedron (60 magnets) and the truncated icosidodecahedron (120 magnets). These configurations have icosahedral symmetry, and consequently their central magnetic field forms a saddle of fourth order [11]. To find the optimal opening angle, we restricted the investigation to , because this cut offers the largest opening aperture. For the normal of the cutting plane, we are now left with two parameters due to the lack of circular symmetry, the polar angle and the azimuthal angle . In Figure 5 the force-free opening angle is determined as a function of . This function is plotted in different colors, where the color indicates the minimal distance from the cutting plane to the nearest magnet . It is normalized here by the common edge length of the Platonic or Archimedean solid, i. e., the distance between neighboring magnets. To achieve maximum mechanical stability of the plastic frame while ensuring secure fixation of the magnets bonded to it, should be maximized. That maximum is indicated by the red color and highlighted by a blue circle.
For the icosahedron characterized in Figure 5a one obtains – the value already introduced in the discussion of Figure 2 – for an azimuthal angle of . In that case, the magnet closest to the cutting plane has a distance of .
The dodecahedron elucidated in Figure 5b brings a negative surprise in the pursuit to find an optimal cutting position: For any , the cutting plane for a force-free opening touches a magnet, as indicated by the scale for , which is of order here. This small number is surely just a numerical artifact without any physical meaning. The fact that the cutting plane hits a magnet does not make it impossible to build such an apparatus, but it would certainly be a nuisance for the construction and stability of the magnet holding frame.
The geometry of the truncated icosahedron (the soccer ball shape), a magnet configuration introduced in [11] and elucidated in Figure 5c, seems to offer a realistic change to realize a force-free opening, with its maximal distance , which is obtained at an opening angle and the azimuthal angle .
The same is true for the truncated icosidodecahedron with its 120 magnets, see Figure 5d. For this configuration, it is noticeable that it contains large holes of decagon shape, which make it comparatively easy to reach the interior anyway. These large openings might reduce the need to open the whole construction when compared to the truncated icosahedron.
It should be noted that the tendency of the cutting angle to increase with the number of magnets, , is also evident in the study shown in Figure 5, consistent with the behavior observed for Fibonacci structures as increases. Furthermore, Figure 5 demonstrates that, due to the discretized nature of the magnetic structures, the resulting forces and feasible opening planes must be determined through numerical simulations of the magnet design.
The same applies to composite configurations. For example, the magnetic field inside a spherocylinder can be understood as the superposition of a central Halbach cylinder and two hemispherical end caps, as discussed in Figure 4. However, the number, positions, and dimensions of the individual magnets must be optimized to balance field contributions and manufacturability using available magnet sizes (see Figure 6). The optimization is carried out here with the Magpylib simulation tool [20], and its result is illustrated in Figure 6b.
Figure 4. Halbach hemispheres as end caps of a Halbach cylinder form a Halbach spherocylinder (capsule shape). A front quarter is cut away to reveal the interior. The red arrows indicate the magnetization directions, and the large yellow arrow denotes the resulting magnetic field. The blue transparent plane marks the surface for an opening at an angle of between the normal (cyan) and the magnetic field (yellow).
Figure 6. Magpylib simulation of a spherocylindrical magnet assembled from cuboidal magnets: (a) The magnet consists of a central Halbach cylinder (magnets with north poles in orange and south poles in green), discretized into 16 cuboidal magnets (10 mm side length, 100 mm total length) arranged with their centers on a radius of 40 mm, and two icosahedral hemispherical end caps (magnets north in red, south in blue). Each end cap is constructed from four full cubic magnets (20 mm side length) and four half-cubes, with magnet centers located on hemispheres of radius 40 mm. The resulting magnetic field in the inside is oriented along the -direction. (b) Axial magnetic field component evaluated along the central -axis. Contributions from the icosahedral end caps (red/magenta), the Halbach cylinder (blue), and the combined structure (black) are shown. Colored background rectangles indicate the axial extent or open inner volume of each substructure, using the same color coding.
Experimental Design and Results
It should be noted that the two prototypes presented in the following were developed primarily to validate the concept of actuating magnetic opening and closing mechanisms with minimal force. To enable rapid fabrication, the systems were constructed using readily available, low-cost permanent magnets with short delivery times; however, these components exhibited limited reproducibility. In addition, the quantity procured was only sufficient for system assembly, precluding any selection or optimization from a larger pool of magnets. Consequently, the achievable magnetic field strength and homogeneity were not fully optimized.
Icosahedron
The first prototype, shown in Figure 7, is a Halbach sphere approximated by an icosahedral arrangement, a variation of the magnet assembly discussed in detail in Ref. [11]. This assembly can be opened by means of the hinge located between the brass nuts. The hinge position was designed to open a cutting plane defined by , , and . Its pivot point is located at a distance mm from the center of the icosahedron.
Figure 7. Hinged icosahedral magnet composed of twelve NdFeB cubes (side length 20 mm with their centers at a radius of 40.5 mm), cutting plane defined by , , and . (a) and (b) Closed configuration shown from two different perspectives. The hinge is realized by an M8 screw from stainless steel. Two bollards allow the magnet to be secured. The measured opening torque is N m. The coordinate system is overlayed (cf. Figure 2). (c) Open configuration. The surface of the inner (sample) volume ( mm) is colored white for improved visibility. Two of the six openings ( mm) are visible here.
The assembly consists of 12 permanent cubic magnets (purchased from Magnethandel item #3982, Dortmund, Germany; dimensions ; , grade N45, specified remanence – T). From the magnitude of the measured magnetic field mT (Figure 8), an effective remanence of T was obtained by fitting.
Figure 8. Magnetic field of the icosahedron shown in Figure 7, measured with a scanning Hall probe: (a) all three field components along the -axis. Note that the inner opening of the shell has a diameter of only 56 mm as indicated by the gray background. (b) component in a zoomed region spanning 20 mm. (c) Field distribution for a central cylindrical volume (8 mm diameter, 20 mm length). The green scale on top refers to -NMR frequencies.
The 12 magnets were bonded into two 3D-printed supports with epoxy glue such that their centers lay on a sphere of radius 40.5 mm, resulting in an inner free volume with radius mm and a mass of 1040 g (see Figure 7a-Figure 7c). Additional access for field measurement in the closed state is provided by three 15 mm diameter through-holes, yielding six openings aligned along the Cartesian axes.
The torque needed to close the magnet was measured using a spring scale yielding N m. The negative sign means that the force is attractive, i.e., the shell has a weak tendency to close by itself.
For a quantitative comparison, we need the force in the units given in Eq. (3), which for the parameters , , and is calculated to be
which means that we are more than four orders of magnitude below the maximum here, where according to Figure 2 the maximum is 0.6 in the same units.
To calculate the torque with respect to the hinge with its distance from the center of the icosahedron, we use the general formula
where is the torque with respect to the origin, here the center of the icosahedron. This torque turns out to be zero for this configuration (and in general to all Halbach spheres with icosahedral symmetry). The closest point of the hinge pin, with its orientation along the direction , is located at the position
a vector perpendicular to . The scalar opening force introduced above is the force component along
Thus, it is the force component perpendicular to , which simplifies the calculation of , the torque component of the vector along the direction of the hinge pin. In SI units, according to the experimental parameters given above, this component then reads:
For the interpretation of this result, an error estimate is necessary. The angle can only be determined within a -range. This corresponds to a torque range of N m to N m.
The measured torque is found to be outside this range, which could be due to the fact that the calculation is based on point dipoles (or spheres), but not on cubes. Cubes with their tips bring parts of their volume closer to their neighbor, and due to the fact that forces scale with the inverse fourth power of the distance, these parts tend to raise the average force.
The value of the measured torque of N m can be compared to the maximum torque expected for this arrangement, which is 3.3 N m. Thus, the reduction obtained here is more than one order of magnitude.
The magnetic field inside both Halbach magnets investigated here was measured using a three-axis Hall probe (MV2, Metrolab Technology SA, Plan-les-Ouates, Switzerland) mounted on a custom-built three-dimensional positioning stage. The stage is driven along the , and axes by three stepper motors, which are controlled by an Arduino Mega R3 microcontroller using custom-developed software. Figure 8a shows the three magnetic field components along the -axis for the icosahedral assembly depicted in Figure 7 in its closed configuration. The field homogeneity is moderate, likely approaching the best achievable given the relatively low grade of the magnets and the limited precision of the 3D printing process. In particular, the acrylonitrile butadiene styrene (ABS) material used for printing exhibited slight deformation (approximately 1 mm) under magnetic forces in this configuration.
Despite this, the central region (Figure 8b) displays reasonable homogeneity along the same axis. The homogeneity over a larger volume can be inferred from the magnetic field distribution (histogram) shown in Figure 8c. This distribution is based on nearly 2000 Hall probe measurements acquired within a sampled central cylindrical volume (diameter 8 mm, length 20 mm). For convenience, the field values are additionally expressed as NMR frequencies on a secondary (green) axis. The normalized peak exhibits a full width at half maximum (FWHM) of approximately 2.4 mT (102 kHz), corresponding to about 12000 ppm.
Spherocylinder
In addition, a spherocylinder was constructed following the idea described in Figure 4. The apparatus is shown in Figure 9. The cylindrical section was designed to consist of 16 cuboid magnets of dimensions , magnetized parallel to one of the 10 mm sides. Since magnets of this size were not commercially available, each cuboid was assembled from two magnets and one magnet (Magnethandel items #3860 and #3977, Dortmund, Germany). The magnets were arranged with the shorter magnet positioned centrally and flanked on either side by the longer magnets. The magnets were glued into 3D-printed supports so that the center of each cuboid lies on a circle of radius mm, resulting in a cylindrical cavity with an inner radius of mm. The outer diameter of the support cylinder is 120 mm.
Figure 9. Hinged spherocylinder composed of a 100 mm long Halbach cylinder and two hemispherical halves of an icosahedron, openable to an angle defined by and . (a) Closed configuration, with the hinge located on the right (realized using an M6 threaded brass rod). The measured opening torque is . (b) Open configuration. The surface of the inner sample volume ( mm) is shown in white for improved visibility. The openings at the top and bottom are bisected by the opening plane.
The icosahedral hemispheres attached at each end of the cylinder contain two types of magnet: four cubes and four cuboids. The cubic magnets were assembled from two magnets (magnetized parallel to the 10 mm side; Magnethandel item #3445). The magnets, magnetized along the 20 mm side, were not available as stock items and were therefore constructed from four cubes each (obtained from Maqna, Germany). All magnets in the assembly had the same listed magnetic properties (NdFeB, grade N45, specified remanence: – T).
These magnets were glued into 3D-printed supports, again with their centers positioned on a circle of radius mm, yielding an inner radius of mm. An opening with a diameter of 20 mm was left along the cylinder axis (). The cylindrical section and the hemispheres were connected using 12 threaded brass rods of 5 mm diameter, together with a 6 mm rod serving as a hinge. The completed assembly has a total mass of 3.18 kg.
The opening angle was set at , placing it in between the next pair of magnets in the cylinder. The torque required to open the spherocylinder was measured accordingly to be N m. The fact that the force is attractive is understandable, because the deliberately chosen angle is well below the range of force-free opening angles for cylinders (–). The influence of the caps is expected to give a weak repelling force at this angle, which is clearly too small to compensate for the attractive force between the cylinder halves.
The same measurements and data processing as described for the icosahedron were applied to the magnetic field distribution within the closed spherocylinder shown in Figure 9. The results are presented in Figure 10. The field profile is considerably more symmetric than that of the icosahedral configuration shown in Figure 8, the homogeneity inferred from the field distribution in Figure 10c – although evaluated over a slightly larger cylindrical volume (diameter 10 mm, length 20 mm) – is approximately three times better (FWHM mT or 35 kHz, corresponding to approximately 4100 ppm). A possible explanation for this is that there were more opportunities to choose magnetic blocks when optimizing the spherocylinder.
Figure 10. Magnetic field of the spherocylinder shown in Figure 9 (same representation as in Figure 8): (a) all three field components along the -axis. The inner accessible volume extends to 78 mm as indicated by the gray area. (b) component zoomed into the central region covering a distance 2 times larger than in Figure 8b. (c) Field distribution calculated for a central cylindrical volume with a diameter of 10 mm and a length of 20 mm.
Conclusion and Outlook
In this work, we have developed and experimentally validated a general framework for the design of mechanically accessible spherical Halbach magnets that can be opened with minimal or vanishing force while preserving high magnetic field homogeneity. Starting from a dipolar Halbach sphere composed of discrete magnetic elements, the magic angle was identified as the asymptotic angle of the cutting plane. Selecting this plane with the help of a hinge eliminates the tensile forces across the interface, leaving only shear components that can be easily managed by the appropriate mechanical design.
The theoretical predictions were confirmed using prototype icosahedral assemblies, which demonstrated that the required opening forces can be reduced by orders of magnitude without significantly compromising the quality of the magnetic field. Residual asymmetries observed in the field measurements were attributed primarily to practical limitations such as minor geometric misalignments and fabrication tolerances, rather than fundamental constraints of the concept, hence more carefully built systems should also result in better homogeneities.
The extension of this approach to spherocylindrical geometries further demonstrates its versatility while enhancing mechanical accessibility and scalability to larger volumes. In particular, combining Halbach cylinders with hemispherical end caps enables the generation of highly homogeneous magnetic fields over expandable volumes with exceptionally efficient use of magnetic material in terms of both mass and volume. At the same time, it preserves the key advantage of allowing controlled opening with manageable forces.
Overall, the presented concept addresses a long-standing challenge in permanent magnet design: combining excellent field homogeneity with full mechanical accessibility. The results open new possibilities for compact, energy-efficient magnetic resonance systems and other applications requiring well-defined magnetic fields in enclosed yet accessible volumes.
Future work may focus on optimized magnet selection, improved assembly precision, and extension to higher-order multipole configurations to further enhance performance.
References
- [1]M. L. Johns, E. O. Fridjonsson, S. J. Vogt, A. Haber (Eds.), Mobile NMR and MRI: Developments and Applications, The Royal Society of Chemistry, 2015. doi:10.1039/9781782628095. URL https://doi.org/10.1039/9781782628095
- [2]K. Halbach, Design of permanent multipole magnets with oriented rare earth cobalt material, Nuclear Instruments and Methods 169 (1) (1980) 1–10. doi: 10.1016/0029-554X(80)90094-4.DOI
- [3]H. Soltner, P. Blümler, Practical Concepts for Design, Construction and Application of Halbach Magnets in Magnetic Resonance, Applied Magnetic Resonance 54 (2023) 1701–1739. doi:10.1007/s00723-023-01602-2.DOI
- [4]I. Rehberg, P. Blümler, Analytic approach to creating homogeneous fields with finite-size magnets, Phys. Rev. Appl. 23 (2025) 064029. doi:10.1103/9nnk-jytn.DOI
- [5]H. Leupold, A. Tilak, E. Potenziani, Adjustable mutli-tesla permanent magnet field sources, IEEE Transactions on Magnetics 29 (6, 1) (1993) 2902–2904, 1993 IEEE International Magnetics Conference (INTERMAG 93), Stockholm, Sweden, Apr 13–16, 1993. doi:10.1109/20.281092.DOI
- [6]C. Bauer, H. Raich, G. Jeschke, P. Blümler, Design of a permanent magnet with a mechanical sweep suitable for variable-temperature continuous-wave and pulsed epr spectroscopy, Journal of Magnetic Resonance 198 (2) (2009) 222–227. doi:10.1016/j.jmr.2009.02.010.DOI
- [7]O. Tretiak, P. Blümler, L. Bougas, Variable single-axis magnetic-field generator using permanent magnets, AIP Advances 9 (11) (2019). doi:10.1063/1.5130896.DOI
- [8]C. W. Windt, H. Soltner, D. van Dusschoten, P. Blümler, A portable Halbach magnet that can be opened and closed without force: The NMR-CUFF, Journal of Magnetic Resonance 208 (1) (2011) 27–33. doi:10.1016/j.jmr.2010.09.020.DOI
- [9]P. Blümler, H. Soltner, Halbach magnet for magnetic resonance that can be opened and closed without effort, European Patent EP2365353 (2011).
- [10]P. Blümler, F. Casanova, Hardware developments: Halbach magnet arrays, in: M. Johns, E. O. Fridjonsson, S. Vogt, A. Haber (Eds.), Mobile NMR and MRI: Developments and Applications, Royal Chemical Society, Cambridge, 2015, book section 4, pp. 133–157. doi:10.1039/9781782628095-00133.
- [11]I. Rehberg, P. Blümler, Discretized Halbach spheres: Icosahedral symmetry for optimal field homogeneity, Phys. Rev. Appl. 25 (2026) 054009. doi:10.1103/hyv2-s2tf. URL https://link.aps.org/doi/10.1103/hyv2-s2tfDOI
- [12]H. Leupold, E. Potenziani, A. Tilak, Lightweight, distortion-free access to interiors of strong magnetic-field sources, Journal of Applied Physics 76 (10, 2) (1994) 6856–6858, 6th Joint Magnetism and Magnetic Materials-Intermag Conference, Albuquerque, NM, Jun. 20–23, 1994. doi:10.1063/1.358530.DOI
- [13]H. Leupold, A. Tilak, E. Potenziani, Permanent magnet spheres: Design, construction, and application (invited), Journal of Applied Physics 87 (9, 2) (2000) 4730–4734, 44th Annual Conference on Magnetism and Magnetic Materials, San Jose, CA, Nov. 15–18, 1999. doi:10.1063/1.373141.DOI
- [14]J. Chen, C. Xu, Design and analysis of the novel test tube magnet as a device for portable nuclear magnetic resonance, IEEE Transactions on Magnetics 43 (9, 1) (2007) 3555–3557. doi:10.1109/TMAG.2007.901888.DOI
- [15]E. P. Furlani, Permanent Magnet and Electromechanical Devices: Materials, Analysis, and Applications, Academic Press, San Diego, 2001. doi:10.1016/B978-012269951-1/50004-8.
- [16]K. W. Yung, P. B. Landecker, D. D. Villani, An analytic solution for the force between two magnetic dipoles, Magnetic and Electrical Separation 9 (1) (1998) 39–52. doi:10.1155/1998/79537.DOI
- [17]R. Swinbank, R. James Purser, Fibonacci grids: A novel approach to global modelling, Quarterly Journal of the Royal Meteorological Society 132 (619) (2006) 1769–1793. onlinelibrary.wiley.com/doi/pdf/10.1256/qj.05.227, doi:https://doi.org/10.1256/qj.05.227. URL https://rmets.onlinelibrary.wiley.com/doi/abs/10.1256/qj.05.227
- [18]I. Rehberg, P. Blümler, Halbach_two_point_oh: Optimize Uniform Fields with Permanent Magnets (Jun. 2026). URL https://doi.org/10.5281/zenodo.20669385
- [19]I. Rehberg, Dipole Cluster Inspector - A Python GUI for Exploring Thousands of Magnetic Configurations (Jul. 2026). doi:10.5281/zenodo.10084573. URL https://doi.org/10.5281/zenodo.10084573
- [20]M. Ortner, L. Bandeira, Magpylib: a free python package for magnetic field computation, SoftwareX 11 (2020) 100466. doi:10.1016/j.softx.2020.100466.DOI