Tutorial: Theory and applications

of the Maxwell stress tensor

Stanley Humphries, Copyright 2012

Field Precision
PO Box 13595, Albuquerque, NM 87192 U.S.A.
Telephone: +1-505-220-3975
Fax: +1-617-752-9077
E mail: [email protected]
Internet: http://www.fieldp.com

1
 The Maxwell stress tensor may be used to calculate electric and magnetic
forces on objects. The method is seldom introduced in introductory texts on
electromagnetism. Advanced texts often present the Maxwell stress tensor
as a mathematical abstraction without explaining why is is useful. In reality,
the method is critical for practical force calculations in numerical codes. In
this tutorial, I will review the theory and emphasize the specific reasons why
the method is used. To illustrate, I will concentrate on magnetostatic forces.
The Lorentz force density is familiar from basic electromagnetism courses:

f = j B. (1)
The vector quantity f is the force per volume in a region with magnetic flux
density B and current density j. The total magnetic force on an object is
given by a volume integral of the force density over its volume:
Z Z Z
F= dV j B. (2)
It is important to recognize that the quantity j is the total current density
within the object, the sum of applied and atomic contributions. The applied
current density arises from currents in coils. The magnitude, location and
direction of the applied current density follows from the known coil geometry
and drive currents. Therefore, is relatively easy to apply Eq. 2 in numerical
codes to find coil forces. In contrast, volume integrals are not practical to
find force contributions from magnetically-active materials (field excluders,
iron and permanent magnets). In this case, the current may be concentrated
in thin surface layers.
To find an alternate force expression, we start from Amperes law:

B. = 0 j. (3)
Note that the relative magnetic permeability (r ) does not appear in Eq. 3
because j represents the sum of all currents (applied and material). Substi-
tution in Equation 2 gives an expression for the body force entirely in terms
of the magnetic flux density:
1 Z Z Z
F= dV ( B) B. (4)
0
Expanding the curl and the cross product, the x component of Equation 4 is

!
1 Z Z Z Bx Bz By Bx
Fx = dV Bz Bz By + By . (5)
o z x x y

2
 Consider taking the divergence of the following vector:

Bx2 By2 Bz2

" ! #
1
Sx = x + y Bx By + z Bx Bz (6)
0 2
The result is

Sx =
!
Bx By Bz By By Bx Bx
Bx + + By Bz + By + Bz . (7)
dx dy dz x x y z

The condition B = 0 implies that the terms in parenthesis sum to zero.

Note that the remaining terms are identical to those in Equation 5. We can
therefore write the x component of body force as
1 Z Z Z
Fx = dV Sx . (8)
o
Finally, we can apply the divergence theorem to convert the volume integral
to a surface integral:
1 Z Z
Fx = dA Sx n. (9)
o
The surface in Equation 9 may be any closed surface surrounding the object.
The quantity n is a unit vector pointing out of the surface. Applying similar
operations to the other force components leads to the general force law
1 Z Z
F= dA S n. (10)
o
The quantity S is the Maxwell stress tensor for magnetostatic fields:

Bx2 B 2 /2

Bx By Bx Bz
1
S= By Bx By2 B 2 /2 By Bz ,

(11)
0

2 2
Bz Bx Bz By Bz B /2
where B 2 = Bx2 + By2 + Bz2 .
There are three reasons why Equation 11 is better suited to a numerical
calculation than Equation 2:

The accuracy of the integral is not affected by the distribution of cur-

rent density within the onject. In particular, the thin surface layers of
magnetically-active materials present no problem.

3
Figure 1: Surface integral of the Maxwell stress tensor. a)Valid surfaces for
integrals around magnetically-active materials. A: region surfaces bounded
by air elements. B :surface of an air region that encloses the magnetic ma-
terials. b)Invalid surface for an integral. The enclosed material current is
undefined on the common boundary.

The surface need not coincide with the physical surface of the object.
If the object has sharp corners and regions of field enhancement, we
can improve accuracy be using a diagnostic surface removed from the
physical surface.

The integral depends only on the field distribution outside the object. It
is not necessary to know the exact current density distributions within
complex anisotropic or nonlinear materials.

Some care must be exercised in applying the method. Because the integral
determines the force on all applied and material currents inside the surface,
we must avoid ambiguities in the enclosed current. The surface must be in
an air region surrounding the object. Here, the term air implies that the
region contains no material currents ( = 0 ). Figure 1 illustrates the logic
of the calculation. If we have an assembly of ferromagnetic and permanent-
magnet objects surrounded by an air volume, then we can take the integral
of Eq. 10 over any enclosing surface. One choice is the outer boundary of
the assembly components (designated surface A in Fig. 1). In this case, we
evaluate magnetic field values in the air elements near the surface to ensure
that the integral encloses all material currents of the object. We could also
define an arbitrary surface by enclosing the assembly inside a diagnostic air
region. The integral over the surface marked B in Fig. 1 gives the same result
(to within the numerical accuracy of field interpolations). Figure 1b shows a
case where the stress tensor integral may not give the correct result. In this
case the integral extends over the surface of ferromagnetic region A that is in

4
contract with another iron or permanent-magnet region (B ). The field values
in region B along the common boundary include the effects of the surface
currents of both regions. The calculation gives the force on region A plus an
indeterminate portion of the force on region B.