IEEE Xplore At-A-Glance
  • Abstract

Parameters for Optimizing the Forces Between Linear Multipole Magnet Arrays

Multipole magnet arrays have the potential to achieve greater forces than homogeneous magnets for linear spring applications. This letter investigates the effects of varying key parameters of linear multipole magnet arrays in relation to their force-bearing potential. Equal sized arrays in repulsive configurations are vertically displaced to each other; only vertical forces are compared. The force versus displacement characteristic is dependent on the aspect ratio of the arrays, the wavelength of magnetization, and the total number of magnets used in the array. Some general design guidelines for optimizing the repulsive forces are established based on the results.



Linear springs constructed with permanent magnets can behave in interesting ways. In levitation contexts, the natural gravity-opposing characteristic with zero power input makes permanent magnets an appealing choice for force generation (e.g., see [Íñiguez 2010]). In the context of supporting a variable-mass load for vibration isolation, the nonlinear force versus displacement characteristic decreases the amount of variability in the resonance frequency of the structure, since as the mass increases and closes the gap between the magnets, the stiffness also increases [Bonisoli 2007]. When used in attraction, the negative stiffness can be used to decrease the resonance frequencies of a supported mass, applicable for ‘‘high-static-low-dynamic’’ or ‘‘quasi-zero stiffness’’ springs [Carrella 2008, Robertson 2009].

With well-known closed form solutions for calculating the forces between cuboid permanent magnets of parallel magnetization [Akoun 1984] and solutions for the forces with orthogonal magnetizations recently published [Janssen 2009, Allag 2009b], it is now possible to analyze a wide variety of magnet configurations that previously required semianalytical or finite-element analysis techniques. This letter investigates the force characteristics between linear multipole magnet arrays as a function of array size and magnet arrangement using cuboid-shaped magnets. The forces are calculated using the force equations derived by the aforementioned researchers.

The results presented in this letter are reproducible [Buckheit 1995] with code located at This is a MATLAB software package written by the authors for calculating the forces between magnets and multipole arrays of magnets, and is freely available to be used by the public. The directory examples/magspring/ contains the code that has been used to directly generate the figures in this letter. Each figure caption refers to the file from which it is generated.



Consider two cuboid-shaped permanent magnets of equal size, facing in repulsion, with vertical magnetization and with displacement between them only in the vertical direction. The shape of the cuboid has a dramatic effect on the force versus displacement characteristic of the system. Assume square facing sides and a height-to-length ratio of α.For fixed magnet volume V, the height of the cuboid is Formula, and the face size length is Formula. The force versus displacement characteristic between two such magnets each with constant volume V = 1000 mm3 over a range of aspect ratios αis shown in Fig. 1. Displacement is measured between the near faces of the magnets. (Note that all forces in this letter are calculated assuming a magnetization of 1 T for all magnets.This value has been chosen to in-effect normalize the output forces by the magnetization strength.) These results show that for a constant volume of magnetic material, a greater facing area yields greater forces.

Figure 1
Fig. 1. Force versus displacement for pairs of repulsive magnets of constant volume each of 1000 mm3, comparing various magnet height-to-length ratios α. Calculated in the file ‘‘mag_ratio.m.’’


Large, flat, thin magnets can be difficult to obtain and hard to work with. While multiple smaller magnets can be stacked together to approximate a single large magnet, greater forces can be achieved by varying the magnetization pattern between adjacent magnets in the stack. In the late 1970s, it was discovered that magnets with sinusoidal magnetization produced single-sided magnetic fields [Halbach 1980, Shute 2000], although it was known much earlier for magnetic bearings that stacks of shorter ring magnets with alternating magnetizations produced greater forces than a longer single ring magnet [Backers 1961].



In this letter, we will consider a linear stack of magnets with N magnets aligned along a horizontal axis (often referred to as a linear Halbach array); planar and volumetric stacks that have multiple magnets in the other directions will not be considered here. As the number of magnets per wavelength of magnetization R increases, the magnetization pattern of the array more closely approximates true sinusoidal magnetization, as shown in Fig. 2.

Figure 2
Fig. 2. Three Halbach arrays of equal length, facing up, each with a single wavelength of magnetization and composed of R ∊ {2, 4,8} magnets per wavelength, respectively.
Figure 3
Fig. 3. Geometry of a linear Halbach array with four magnets of length m per wavelength of magnetization λ.This array contains two wavelengths of magnetization with an end magnet for symmetry, i.e., W = 2 and R = 4.

The magnetic flux pattern of the array is also dependent on the wavelength of magnetization λ. As the wavelength of magnetization decreases, the total number of magnets used in the array increases, for a fixed array length I. The relationship between wavelength, array length, and number of magnets is shown in Fig. 3.Note the one extra magnet included in Figs. 2 and 3 such that the total number of wavelengths W = (lm)/λ and the total number of magnets N = WR + 1. This magnet is necessary to balance the forces in the horizontal direction such that only vertical forces are generated between the arrays.

There are two independent variables to consider when choosing the parameters for a linear Halbach array of a certain size: number of magnets per wavelength R, and total number of wavelengths W in the array.

A. Varying Magnetization Discretization and Wavelength

Consider two linear Halbach arrays of equal size with height h = 10 mm, square cross section, and length l = 100 mm. Their strong sides are aligned towards each other and their magnetization pattern is such that there is a repulsive force between them. The vertical displacement between their centers is δ, which can be normalized by the height of the arrays; δ /h = 1 corresponds to the position at which the faces of the two arrays are touching. By calculating the forces between the arrays using superposition of the forces between each permutation of magnet pairs in the two arrays [Allag 2009a], the force versus normalized vertical displacement was calculated for the number of magnets per wavelength R ∊ {2, 4, 8} and the number of wavelengths W ∊ {1, 2, 4} and compared with the forces generated between a pair of equivalently sized magnets of homogeneous magnetization. These results are shown in Fig. 4.

Figure 4
Fig. 4. Force versus displacement normalized by the array height h between two facing linear Halbach arrays with a varying number of magnets per wavelength R and a varying number of wavelengths of magnetization W. The dashed line is the force between two single magnets of equal size to the arrays. Calculated in the file ‘‘multipole_compare.m.’’

It can be seen in Fig. 4(a) that for a small number of wavelengths, the discretization of the magnetization makes little difference to the force characteristic. But as the number of wavelengths increases the number of magnets per wavelength has an increasing effect. As seen most prominently in Fig. 4(c), increasing the number of magnets per wavelength R increases the forces over all values of displacement considered.

Therefore, as a general design guideline, it is only necessary to use a large number of magnets per wavelength if there are at least several wavelengths of magnetization in total in the array. In the results shown in Fig. 4, the ratio in forces between R  = 2 and R = 4 is greater than the ratio in forces between R = 4 and R = 8; most of the benefit of increasing the number of magnets is realized using four magnets per wavelength of magnetization (i.e., 90°rotations between successive magnets such as shown in Fig. 3).In the cases where there are many more wavelengths of magnetization again(as shown in Fig. 5 later), there is a greater advantage to using R = 8 over R = 4. Therefore, the greater the number of wavelengths of magnetization, the greater the force improvement in increasing the number of magnets per wavelength.

Figure 5
Fig. 5. Force characteristic with arrays each composed of 50 magnets of length m = 2 mm, over a variety of number of magnets per wavelength R. The dashed line is the force between two homogeneous magnets of length 100 mm. Calculated in ‘‘multipole_const_Nmag.m.’’

The use of multipole arrays can have a significant effect on the useful range of the force/displacement characteristic. As the number of wavelengths increases, the magnetic field of each array becomes stronger but the magnetic field lines exhibit smaller excursions outside the magnet array before returning.Thus, the forces become stronger but over a smaller displacement, and therefore the stiffness of the magnetic spring is increased as well. For some purposes and in some cases, this can be detrimental in that it can increase the resonance frequency of the system, resulting in poorer vibration isolation properties.

B. Constant Number of Magnets

The results shown previously have in general indicated that improvements to the force characteristic are seen with a greater number of magnets. However, given a minimum magnet thickness that can be fabricated, and hence, for a given array length a maximum number of magnets in total, the question arises:is it better to maximize the number of wavelengths W or the number of magnets per wavelength R? Consider an array of the same outer dimensions as the previous example composed of magnets each of length m = 2 mm and of cross-sectional area 10 mm × 10 mm, such that there are 50 magnets in the array. The force characteristic for this system, again with R ∊{2, 4, 8}, is shown in Fig. 5. In this extreme example with large W, the strong region of the field is close to the surfaces of the arrays and there is a considerable difference in the curves for each value of R; maximizing R produces stronger results providing there are sufficiently many wavelengths of magnetization along the length of the array. When W is small (e.g., W < 5) for a fixed magnet size, these general results do not hold and the design possibilities must be evaluated on a case-by-case basis.

C. Unequal Magnet Sizes

While the force characteristic of an eight-magnet wavelength (R = 8) array can outperform the four-magnet (R = 4) array, the latter can be improved in some cases by adjusting the relative sizes of the magnets in the array. Consider the four-magnet array shown in Fig. 6 in which the horizontally polarized magnets of length b are smaller than the vertically polarized magnets of length a. Magnet size ratio γ = b/a is the measure used here to compare different array configurations, for which γ = 0 corresponds to an array composed only of vertically oriented magnets, and γ = 1 corresponds to equally sized magnets of both horizontal and vertical magnetizations(as considered previously in this letter).

Figure 6
Fig. 6. Schematic of a four-magnet Halbach array with variable magnet sizes. The number of magnets per wavelength R = 4 for all arrays of this type unless b = 0, in which case R =2.
Figure 7
Fig. 7. Force characteristic between two modified Halbach arrays of W =2 with magnet length ratio γ between the sizes of vertically to horizontally polarized magnets. Calculated in “linear_quasi_example.m.”

Fig. 7 compares the force characteristic with a variety of magnet size ratios for arrays composed of nine magnets (i.e., two wavelengths of magnetization with a symmetry magnet), of length l  =100 mm, and of cross-sectional area 10 mm × 10 mm. As expected from the previous results, γ =0 results in smaller forces than for γ =1; however, γ = 0.5 results in slightly greater forces again: an increase of 5 % at a displacement of approximately δ = 1.3 h, tapering off as the displacement increases. (This value of γ is close to optimum for this system; see Fig. 8.)This result can be justified intuitively with the recognition that there is a stronger vertical force between opposing vertically polarized magnets than between horizontally polarized magnets; dedicating a greater proportion of the magnet volume to the vertical magnets yields an increase in the total force.

Figure 8
Fig. 8. Integral of the force-displacement characteristic versus magnet length ratio γ of two modified Halbach arrays of size 100 mm × 10 mm × 10 mm over a displacement range of 10–20 mm, shown with varying number of wavelengths of magnetization W. Calculated in ‘‘linear_quasi_ratios.m.’’

However, as the number of wavelengths of magnetization W increases, there is a decrease in the improvement offered by reducing the size of the horizontal magnets. This can be quantified by comparing the integral of force over the displacement range of interest for a variety of magnet length ratios γ.In Fig. 8, such results are shown(for the same arrays discussed previously) comparing the relative difference of the force-displacement integral as a function of the magnet length ratio, normalized by the integral results for γ =1. Since the force improvement with adjusting γ is only significant for low numbers of wavelength of magnetization, this technique is only suitable for increasing the forces when a small total number of magnets are to be used, perhaps for ease of construction of the magnet array. Otherwise, it is more efficient simply to increase the number of magnets than to change the magnet size ratio.



In optimizing the forces between linear Halbach magnet arrays, it has been shown that there is a relationship on the force versus displacement characteristic from both the wavelength of magnetization and the number of magnets in the array. As the wavelength of magnetization decreases while keeping the array length constant, the effect of increasing the number of magnets per wavelength increases. In order to achieve significantly larger forces over homogeneous magnetization, a large number of magnets should be used; a minimum of around 16 + 1 magnets with 90° magnetization rotations or 32 + 1 magnets with 45° magnetization rotations. When only a small number of magnets are used, small improvements to the forces can be achieved by using magnets of smaller volume which are magnetized parallel to the array length and magnets of larger volume which are polarized in the facing direction.


Corresponding author: W. Robertson (


1. 3D analytical calculation of the forces exerted between two cuboidal magnets

G Akoun, J-P Yonnet

IEEE Trans. Magn., Vol. MAG-20, issue (5), pp. 1962 –1964, 1984, 10.1109/TMAG.1984.1063554

2. 3D analytical calculation of forces between linear Halbach-type permanent-magnet arrays

H Allag, J-P Yonnet, M Latreche

Proc. 8th Int. Symp. Adv. Electromech. Motion Syst. Electric Drives Joint Symp. (ELECTROMOTION), 2009a, pp. 1–6, 10.1109/ELECTROMOTION.2009.5259084

3. 3D analytical calculation of interactions between perpendicularly magnetized magnets—Application to any magnetization direction

H Allag, J-P Yonnet, M Fassenet, M Latreche

Sens. Lett., Vol. 7, issue (3), pp. 486–491, 2009b, 10.1166/sl.2009.1094

4. A magnetic journal bearing

F T Backers

Philips Tech. Rev., vol. 22, issue (7), p. 232–238, 1961

5. Passive elasto-magnetic suspensions:Nonlinear models and experimental outcomes

E Bonisoli, A Vigliani

Mech.Res. Commun., Vol. 34, issue (4), pp. 385 –394, 2007, 10.1016/j.mechrescom.2007.02.005

6. Wavelab and reproducible research

J B Buckheit, D L Donoho

Lecture Notes Stat., vol. 103, p. 55–81. Available at:, 1995

7. On the design of a high-static-low-dynamic stiffness isolator using linear mechanical springs and magnets

A Carrella, M J Brennan, T P Waters, K Shin

J. Sound Vibration, Vol. 315, issue (3 ), pp. 712–720, 2008, 10.1016/j.jsv.2008.01.046

8. Design of permanent multipole magnets with oriented rare earth cobalt material

K Halbach

Nucl.Instrum. Methods, Vol. 169, issue (1 ), pp. 1–10, 1980, 10.1016/0029-554X(80)90094-4

9. Numerical simulation of a simple low-speed model for an electrodynamic levitation system based on a Halbach magnet array

J Íñiguez, V Raposo

J. Magn. Magn. Mater., Vol. 322, issue (9–12), pp. 1673–1676, 2010, 10.1016/j.jmmm.2009.04.035

10. Analytical calculation of interaction force between orthogonally magnetized permanent magnets

J L G Janssen, J J H Paulides, E Lomonova, F Bölöni, A Tounzi, F Piriou

Sens. Lett., Vol. 7, issue (3), pp. 442–445, 2009, 10.1166/sl.2009.1049

11. Theoretical design parameters for a quasi-zero stiffness magnetic spring for vibration isolation

W S Robertson, M R F Kidner, B S Cazzolato, A C Zander

J. Sound Vibration, Vol. 326, issue ( 1–2), pp. 88–103, 2009, 10.1016/j.jsv.2009.04.015

12. One-sided fluxes in planar, cylindrical, and spherical magnetized structures

H A Shute, J C Mallinson, D T Wilton, D J Mapps

IEEE Trans. Magn., Vol. 36, issue (2), pp. 440–451, 2000, 10.1109/20.825805


No Photo Available

Will Robertson

No Bio Available
No Photo Available

Ben Cazzolato

No Bio Available
No Photo Available

Anthony Zander

No Bio Available

Cited By

No Citations Available


IEEE Keywords

No Keywords Available

INSPEC: Controlled Indexing

electromagnetism, magnetisation

More Keywords

No Keywords Available


No Corrections


No Content Available

Indexed by Inspec

© Copyright 2011 IEEE – All Rights Reserved