Permanent Magnet Field and Pull Force Calculator

Calculate fields and pull forces for disc, block and ring magnets. Compare steel pull, attraction or repulsion, air gaps, grades and temperature.

#magnetic field #numerical-integration #permanent magnets #pull force #surface-charge model

This permanent magnet field and pull force calculator compares axially magnetised disc, block and ring magnets. Engineers can estimate latch forces and sensor fields, students can explore magnetic superposition, and makers can see why a small spacer changes a magnet's grip. The model keeps the finite size of every magnet: it does not replace the bodies with point dipoles.

Choose a shape and an interaction: one magnet in air, a magnet above steel, or two coaxial magnets. Enter the pole-face dimensions and thickness along the magnetisation direction. For a ring, enter the bore diameter. The block length and width describe the charged rectangular faces. Two magnets may have different dimensions and NdFeB grades; clear “Identical second magnet” to enter them separately.

The face gap is the total nonmagnetic separation. Include paint, nickel plating, adhesive, protective films and spacers when they lie between the magnetic material and its target. Drag the gap handle to compare positions. The force-against-gap chart includes a marker for the selected separation; positive means an external pull is needed, and negative means a push. Force magnitude is reported separately from the signed vertical force on the upper magnet; upward is positive. Overlapping opposite pole faces normally attract. Unequal rings with no radial face overlap can reverse their net axial force; the signed result gives the actual direction.

Choose the material grade and magnet temperature. The grade supplies remanence Br, intrinsic coercivity Hcj and representative temperature coefficients. Use measured remanence when you have a supplier certificate or a measurement-derived value. The temperature correction is reversible and linear; it does not predict recovery after overheating or permanent demagnetisation.

The section displays the computed magnetic field magnitude in air and field lines integrated through that field. N and S identify the magnetised faces. Move the probe to read the lateral and axial field components. The isolated on-axis curve answers the separate question of the magnetic field of a cylinder magnet, block or ring without steel or another magnet.

For steel, the ideal thick-plate option uses an infinite-permeability image. Finite-permeability and finite-thickness options describe an infinite, linear steel body. A thin real plate can saturate, leak flux and give substantially less pull than these models. There is no universal thickness correction without steel B–H data and the surrounding geometry.

This magnet pull force calculator estimates normal separation force, not a hanging mass or a working-load rating. Sliding also depends on friction; peeling, bending, vibration, impact and poor contact can govern a real attachment. It excludes finite plate edges, cup magnets, back iron, tilted or laterally offset magnets, and diametric or radial magnetisation. Confirm a final attachment with measurements in its actual assembly. Advanced quadrature and field-grid controls allow numerical refinement; the plotted field is an engineering view, not a prediction of microscopic edge peaks.

How it is calculated

Uniform magnetisation is prescribed along the thickness. The pole faces carry magnetic surface charge σm=M⋅n\sigma_m=\mathbf M\cdot\mathbf n, with μ0M=Br\mu_0 M=B_r. In air, B=Br4π∫Ss (r−r′)/∣r−r′∣3 dS′\mathbf B=\frac{B_r}{4\pi}\int_S s\,(\mathbf r-\mathbf r')/|\mathbf r-\mathbf r'|^3\,dS'. Closed axial cylinder, annulus and rectangle expressions follow Camacho and Sosa (2013), equations 7, 9 and 11 (PDF pages 4–5). Off axis, the charged-face integrals are evaluated as equivalent boundary integrals: Gauss–Legendre angular quadrature for circular faces and analytic corner expressions for rectangles. For a circular face, the substitution θ=csinh⁡(t)\theta=c\sinh(t), where c=(r−R)2+z2/Rrc=\sqrt{(r-R)^2+z^2}/\sqrt{Rr}, clusters nodes at the nearest rim point. On a pole plane the normal charge field uses the one-sided limit toward positive z, including the half jump inside the face. A ring is an outer cylinder minus its bore.

Block force uses the 64-corner Akoun–Yonnet (1984) analytical sum, with the contact limit evaluated explicitly. Beyond ten times the largest block dimension, an 8 × 8 Gauss integral of the analytic source field over the target pole faces evaluates the same interaction without cancellation of the corner sum; this is still a finite-body calculation, not a dipole replacement. Disc and ring force uses full numerical charge–charge integration, reduced by circular overlap: I(d)=2πd∫0R1+R2A12(s)s(s2+d2)−3/2 dsI(d)=2\pi d\int_0^{R_1+R_2}A_{12}(s)s(s^2+d^2)^{-3/2}\,ds. The change of variable u=d/s2+d2u=d/\sqrt{s^2+d^2} removes the near-contact peak; Gauss–Legendre quadrature is split at internal tangency. For ring walls below 10% of the outside radius, the annular overlap is combined before integration and the integral is split at every inner/outer circle tangency; this resolves narrow overlap features without subtracting four separately integrated large forces. At contact, I(0+)=2πA12(0)I(0^+)=2\pi A_{12}(0). Four signed face pairs give each body-to-body force. No centre-field-times-area approximation is used.

An ideal steel plane is replaced by an identical mirrored magnet: the image face gap is twice the real gap, and the force has no extra factor of one half. Finite permeability uses ρ=(μr−1)/(μr+1)\rho=(\mu_r-1)/(\mu_r+1). A finite linear slab adds images at depths 2nt2nt, with weights −(1−ρ2)ρ2n−1-(1-\rho^2)\rho^{2n-1}. These images and the original magnet produce the field in the air above the plate. An independent Maxwell-traction integral over the ideal plane checks the force. Its integration is split at every projected pole edge; at contact it uses the air-side limit below the magnet's lower face.

Material data are from the Polaris grade table at 20 °C. Br(T)=Br,20[1+αB(T−20)]B_r(T)=B_{r,20}[1+\alpha_B(T-20)] and the analogous relation for Hcj use the selected grade's coefficients. The reverse-field result samples −Hz-H_z at the upper magnet's centre, or midway through a ring wall; it is a diagnostic, not a coercivity proof.

Assumptions

Magnets are uniformly and rigidly magnetised, with recoil relative permeability taken as one; Br is treated as magnetic polarisation. Both magnets share the entered temperature, their axes coincide, and block edges are parallel. Air and nonmagnetic coatings have relative permeability one. Steel extends infinitely sideways and is isotropic and linear in the selected permeability model. The displayed section is through the symmetry axis (or the block's width midplane); magnet interiors and steel are covered by their material fills. The field grid includes the magnetisation term inside real magnets, preserving continuity of the normal flux density across pole faces for line tracing. A grid node on a lower pole face is treated as inside; one on an upper face is outside, consistently with the positive-z charge-field limit. The surface-to-boundary reduction, overlap quadrature and finite-slab image series are derived here from the stated charge model, rather than a fitted pull-force correlation.

Where it applies

The model excludes saturation, hysteresis, nonuniform magnetisation, recoil feedback and irreversible demagnetisation. Hcj is not the knee field: the sampled upper-magnet reverse field misses local edge maxima, does not screen the second magnet, and cannot establish thermal stability. Nominal grade temperature limits depend on the actual magnetic circuit. Ring “top-centre” field is measured on the bore axis and may be negative. When unequal ring faces do not overlap radially, the net axial force can reverse as the gap changes, even with unchanged pole orientation. The chart is signed: positive requires an external pull to separate the magnets; negative requires a push to hold them together.

The finite-thickness result is an approximation for real steel because permeability is held constant. It does not capture the dominant saturation loss of a very thin sheet. Plate thickness in the thick-plate modes is only a screening input; it does not alter their force or drawing. A thickness below 0.2 times the magnet diameter/length triggers a heuristic thin-sheet warning, not a calculated saturation boundary. Finite slab edges and the field below the steel are not calculated. The slab image series defaults to 18 correction images; refine the image count to check unusually thin plates. A conservative bound on the remaining force tail triggers a warning if it exceeds 1% of the result. Remote images use a coarser angular rule because their fields vary smoothly.

The default air field is solved on a roughly 73 × 42 grid, with 40 clustered Gauss–Legendre nodes on a half-circle at the default circular-face resolution of 80. Ring and finite-slab displays use fewer solved grid points to keep interaction responsive. Colours blend between solved points; field lines use bilinear interpolation and midpoint integration, so the immediate vicinity of a sharp edge or a sub-cell gap is not resolved. Increase the grid and circular-face resolution to inspect a local field. Direct probe and force results do not use that display grid. The direct probe is compared with a doubled angular rule, and the selected force with doubled force quadrature; results are withheld if their differences exceed 0.05% or the absolute floors of 0.1 µT and 0.1 µN, respectively. Move a probe away from an edge or increase resolution when it fails this test. Exactly on a pole plane, direct probes are excluded; sharp rim fields are singular in this ideal model. Ring wall / outside-radius ratios below 0.0001 are excluded to limit cancellation in annular overlap arithmetic. The minimum wall and near-edge cases have separate numerical regression and convergence tests. Force curves use 81–101 points concentrated near contact. For a finite slab, the complete finite-body kernel is solved at 141 separations concentrated near contact and interpolated with pchip for the image-series chart; its separate kernel-grid control checks this approximation.

Published pull-force comparisons illustrate model discrepancy, not a calibration: for K&J D82 and D84 at Br = 1.32 T, the ideal contact forces are approximately 34.08 N and 55.25 N versus listed measured pulls of 28.65 N and 48.40 N. Coating gaps, recoil permeability, real steel and material variation can all contribute; the source does not identify their separate shares. The corresponding regression checks allow 25% discrepancy for measured pull, while the surface-field and published cuboid-force checks have tight numerical tolerances.

References

  • Akoun, G. and Yonnet, J.-P. (1984), 3D analytical calculation of the forces exerted between two cuboidal magnets, IEEE Transactions on Magnetics 20(5), 1962–1964. https://doi.org/10.1109/TMAG.1984.1063554
  • Camacho, J. M. and Sosa, V. (2013), Alternative method to calculate the magnetic field of permanent magnets with azimuthal symmetry, Revista Mexicana de Física E 59, 8–17, especially sections 3 and 4. https://rmf.smf.mx/ojs/index.php/rmf-e/article/download/4692/5953/14417
  • Furlani, E. P., Permanent Magnet and Electromechanical Devices: Materials, Analysis, and Applications, 1st edition, Academic Press, 2001, chapter 3 (charge model). ISBN 9780122699511. https://books.google.com/books?id=irsdLnC5SrsC
  • Rosso, M., Corigliano, A. and Ardito, R. (2022), Numerical and experimental evaluation of the magnetic interaction for frequency up-conversion in piezoelectric vibration energy harvesters, Meccanica 57, 1139–1154, section 2 and Figure 3. https://doi.org/10.1007/s11012-022-01481-0
  • K&J Magnetics, Does Stacking Magnets Make Them Stronger?, D82 and D84 specification table; and Calculating Steel Thickness for Neodymium Magnets, updated 2 January 2025. https://www.kjmagnetics.com/blog/stacking-magnets ; https://www.kjmagnetics.com/blog/steel-thickness-and-magnetic-fields
  • Polaris Rare Earth Materials, NdFeB Basic Sintered Grades, magnetic properties at 20 °C and temperature coefficients, accessed 8 November 2026. https://www.polarisrem.com/products/ndfeb-basic-grades

Inputs

InputSymbolUnitDefault
Magnet shapeshapeDisc / cylinder
InteractiontargetMagnet to steel
Facing polespolarityOpposite facing poles
Diameter / block lengthD1mm20
Block widthW1mm20
Bore diameterd1mm8
Magnet thicknessL1mm6
Face gap
Include coatings, glue and any nonmagnetic spacer.
gapmm2
Identical second magnetidenticaltrue
Second diameter / lengthD2mm14
Second block widthW2mm14
Second bore diameterd2mm4
Second thicknessL2mm4
NdFeB gradegradeN42
Second NdFeB gradegrade2N42
Magnet temperatureT_mag°C20
Use measured remanencemeasuredfalse
Remanence at 20 °CBr_measT1.32
Second remanence at 20 °CBr_meas2T1.32
Steel modelsteel_modelIdeal thick plate (μ → ∞)
Plate thickness
Finite plate edges and steel saturation are excluded.
plate_tmm6
Steel relative permeability
Use a representative secant B/(μ₀ H) value near the expected operating field.
mu_steel500
Probe radial / lateral positionx_probemm8
Probe height
Height from the steel or second magnet's upper face; the single magnet starts at zero.
z_probemm13
Gap chart endgap_endmm20
Force quadrature orderN_quad48
Circular-face quadrature resolutionN_field80
Field grid columnsN_grid73
Finite-slab correction imagesN_images18
Finite-slab kernel separationsN_kernel141

Results

With the default inputs above, the calc gives:

ResultValue
Axial force magnitude32.3246 N
Force on upper magnet (+ upward)-32.3246 N
Flux density at probe0.2025 T
Probe lateral field0.1266 T
Probe axial field0.1581 T
Isolated top-centre axial field0.3383 T
Remanence at temperature1.315 T
Intrinsic coercivity at temperature955 kA/m
Sampled internal reverse field631.1 kA/m
Net interactionAttraction

Checked against sources

Each case below was calculated by this calc and agrees with the value from its source (within the tolerance where one is shown).

Published 3 mm cube force

Source: Rosso, Corigliano and Ardito (2022), Meccanica, section 2 and Figure 3, DOI 10.1007/s11012-022-01481-0: two 3.0 mm cubes, J = J′ = 1.32 T, 1 mm axial gap, peak aligned normal force 1.574 N. Axes relabelled to put magnetisation along this tool's z axis. The web transcription of its axial kernel has a U·W typo; the correct term U·V preserves x/y symmetry and gives the stated peak.

Inputs: Magnet shape = Rectangular block; Interaction = Two coaxial magnets; Diameter / block length = 3 mm; Block width = 3 mm; Magnet thickness = 3 mm; Face gap = 1 mm; Use measured remanence = yes; the others as above.

ResultSourceThis calc
Axial force magnitude1.574 ± 0.002 N1.574 N

K&J D82 field and measured pull comparison

Source: K&J Magnetics, Stacking Magnets, D82 row: 1/2 in diameter × 1/8 in thickness, N42, surface field 2952 G, measured pull 6.44 lbf. https://www.kjmagnetics.com/blog/stacking-magnets . Br = 1.32 T is the upper end of K&J's N42 range and reproduces its listed axial field. Pull comparison uses a declared 25% model-discrepancy tolerance: the ideal image result is about 34.08 N, 19% above the listed pull, so this is not a precision validation of real contact pull.

Inputs: Diameter / block length = 12.7 mm; Magnet thickness = 3.175 mm; Face gap = 0 mm; Use measured remanence = yes; the others as above.

ResultSourceThis calc
Isolated top-centre axial field0.2952 T0.2952 T
Axial force magnitude28.647 ± 7.162 N34.078 N

K&J D84 field and measured pull comparison

Source: K&J Magnetics, Stacking Magnets, D84 row: 1/2 in diameter × 1/4 in thickness, N42, 4667 G, measured pull 10.88 lbf. https://www.kjmagnetics.com/blog/stacking-magnets . Br = 1.32 T. Declared 25% model-discrepancy tolerance on measured pull, versus a tight 0.0001 T axial-field tolerance. The unmodified ideal image model gives about 55.25 N, 14% above the listed pull; real contact conditions and recoil are excluded.

Inputs: Diameter / block length = 12.7 mm; Magnet thickness = 6.35 mm; Face gap = 0 mm; Use measured remanence = yes; the others as above.

ResultSourceThis calc
Isolated top-centre axial field0.4667 T0.4667 T
Axial force magnitude48.397 ± 12.099 N55.248 N

Published unequal-cylinder repulsion

Source: Camacho and Sosa (2013), Figure 9 (PDF p. 8), measured open-circle curve read at a = 3.0 cm: approximately 0.40 N repulsion, with ±0.08 N allowance for graphical reading and the ±0.03 N sensor accuracy reported on PDF p. 7. Geometry and polarisations from Figure 7: R1 = 1.27 cm, L1 = 0.95 cm, μ0M1 = 0.89 T; R2 = 0.635 cm, L2 = 0.485 cm, μ0M2 = 0.92 T. This is a coarse experimental validation at large gap, not a digitised precision dataset. https://www.scielo.org.mx/pdf/rmfe/v59n1/v59n1a2.pdf

Inputs: Interaction = Two coaxial magnets; Facing poles = Like facing poles; Identical second magnet = no; Diameter / block length = 12.7 mm; Magnet thickness = 4.85 mm; Second diameter / length = 25.4 mm; Second thickness = 9.5 mm; Face gap = 30 mm; Gap chart end = 40 mm; Use measured remanence = yes; Remanence at 20 °C = 0.92 T; Second remanence at 20 °C = 0.89 T; the others as above.

ResultSourceThis calc
Axial force magnitude0.4 ± 0.08 N0.3639 N
Force on upper magnet (+ upward)0.4 ± 0.08 N0.3639 N

Independent near-edge disc field

Source: Independent review of board subtask #1114 supplied an adaptive surface-charge reference: 20 × 6 mm disc, Br = 1.315 T, x = 10.1 mm, z = 6.01 mm measured from the bottom face, Bz = −0.14526 T and |B| = 0.83378 T. This checks direction as well as magnitude next to the outer circular edge; the former uniform angular rule gave the wrong sign.

Inputs: Interaction = Single magnet field; Probe radial / lateral position = 10.1 mm; Probe height = 6.01 mm; the others as above.

ResultSourceThis calc
Probe axial field-0.1453 T-0.1453 T
Flux density at probe0.8338 T0.8338 T

Independent zero-gap Maxwell stress

Source: Independent review of board subtask #1114 supplied an independent Maxwell-stress reference of 96.7377 N for a 20 × 6 mm disc, Br = 1.315 T, in contact with ideal steel. The old sign(0) boundary treatment gave 40.7746 N and failed this independent momentum-balance value.

Inputs: Face gap = 0 mm; the others as above.

ResultSourceThis calc
Ideal-plane force from Maxwell stress96.738 N96.738 N
Axial force magnitude96.738 N96.738 N

Independent thin-ring interaction

Source: Independent review supplied approximately 0.119680 N from refined and adaptive charge–charge integration for two identical axially magnetised 20 / 19.98 × 2 mm rings, Br = 1.315 T, 0.01 mm gap. Independently reproduced with SciPy 1.15.3 adaptive Gauss–Kronrod integration of annular overlap using extended-precision area arithmetic (verify_review.py in the working folder): 0.119680067886 N. Tests a 0.01 mm wall, where four separately integrated disc terms previously lost 6.1%.

Inputs: Interaction = Two coaxial magnets; Magnet shape = Ring; Bore diameter = 19.98 mm; Magnet thickness = 2 mm; Face gap = 0.01 mm; the others as above.

ResultSourceThis calc
Axial force magnitude0.1197 N0.1197 N

Independent minimum-wall integration

Source: Independent adaptive charge–charge integration with SciPy 1.15.3 quad (Gauss–Kronrod), split at all annular tangencies, using NumPy longdouble overlap arithmetic; verify_review.py in the working folder. Identical 20 / 19.998 × 2 mm rings, gap 0.001 mm, Br 1.315 T: 0.0120661292258 N. Validates the accepted minimum wall / outside-radius ratio of 0.0001.

Inputs: Interaction = Two coaxial magnets; Magnet shape = Ring; Bore diameter = 19.998 mm; Magnet thickness = 2 mm; Face gap = 0.001 mm; the others as above.

ResultSourceThis calc
Axial force magnitude0.01207 N0.01207 N

Related Resources

Responses

Your email address will not be published. Required fields are marked *

L ading...