Demagnetizing Factors For Cylinders

Download as pdf or txt
Download as pdf or txt
You are on page 1of 19

IEEE TRANSACTIONS ON MAGNETICS, VOL. 21, NO.

4, JULY 1991 3601

Demagnetizing Factors for Cylinders


Du-Xing Chen, James A. Brug, Member, IEEE, and Ronald B. Goldfarb, Senior Member, IEEE

Abstract-Fluxmetric (ballistic) and magnetometric demag- magnetic poles at each end of a cylinder [ 191. This sim-
netizing factors Nf and N , for cylinders as functions of suscep- plistic model could be used only for long uniformly mag-
tibility x and the ratio y of length to diameter have been eval- netized cylinders, and the results deviated significantly
uated. Using a one-dimensional model when y 2 10, Nf was
calculated for - 1 5 x < and N,,, was calculated for x 00.
Q) +
from experimental data on ferromagnetic samples. During
Using a two-dimensional model when 0.01 5 y 5 50, an im- the 1920’s and 1930’s, there were several theoretical pa-
portant range for magnetometer measurements,N , and Nf were pers on Nf for material with constant susceptibility x. The
calculated for - 1 5 x < 00. Demagnetizing factors for x < 0 results were given as functions of x and the length-to-
are applicable to superconductors. For x = 0, suitable for diameter ratio y. These used one-dimensional models with
weakly magnetic or saturated ferromagnetic materials, Nfand
N,,, were computed exactly using inductance formulas. approximations as needed to suit the computational tech-
niques of the time. The first systematic theoretical cal-
culation of N’ for the high susceptibility case was done by
I. INTRODUCTION Wurschmidt [20], [21]. He calculated Nf of cylinders

T HE study of demagnetizing factors for ellipsoids and


their degenerate forms (spheres, infinite plates, infi-
nite cylinders) dates from the work of Poisson [ 11 and was
using a one-dimensional model in which the cylinder had
side surface poles and point end poles. He used Taylor
expansions for the magnetization and the demagnetizing
elaborated upon by Thomson [2], Evans and Smith [3], field at the midplane. The calculation was complicated,
and Maxwell [4]. Experimental investigations on finite and he completed it only for the case y = 50 and x --+ W.
right circular cylinders began in the 1870’s, when ballistic For the y and x dependence of Nf,he gave qualitative
galvanometers were first used in magnetic measurements results using the first few terms of the expansion. A sim-
of iron [5], [6]. The literature distinguishes between ilar approach with simpler expressions was used by Neu-
“magnetometric” and “fluxmetric” (or “ballistic”) de- mann and Warmuth [22], who calculated Nf for x + 00
magnetizing factors N , and Nf[7]. N , refers to an average a n d y 2 10.
of magnetization over the entire specimen and is appro- To obtain the susceptibility dependence of Nf,Stablein
priate for magnetometer measurements of small samples. and Schlechtweg [23] used a quadratic approximation and
Nf refers to an average of magnetization at the midplane two linear differential equations. The model was im-
of the sample and is appropriate for measurements made proved by substituting uniform end-surface poles for point
with short search coils. Values of the demagnetizing fac- end poles. Their results included 30 values of Nf for 10
tor were deduced from the shearing of the magnetic hys- I y 5 500 and 12.56 Ix c W . An extension of the y
teresis loop [8], a procedure originally developed by Lord region to 0 was achieved by Warmuth [24]-[26], who fit-
Rayleigh for ellipsoids [9], or by measuring the magnet- ted existing data and extrapolated graphically using the
ization and the field at the cylinder’s side [ 101. Factors demagnetizing factor N of ellipsoids as a reference. The
for cylinders of several aspect ratios were published [7], values of Nf calculated from the one-dimensional models
[8], [ l l ] , [12]. By the 1900’s, it was apparent that the were consistent with the data of ballistic measurements
values of the factors depended on the susceptibility x of on soft magnetic materials. Bozorth and Chapin [27] com-
the material [13]-[15]. piled the results, which were later plotted in Bozorth’s
Although it has been criticized from a pedagogical point book [28].
of view [16], [17], the use of fictitious magnetic poles to To obtain axial demagnetizing factors more accurately,
calculate demagnetizing fields has been universal. The especially for short cylinders, two-dimensional calcula-
first theoretical treatment of magnetic pole distributions tions are needed. The simplest case is x = 0, where N ,
in finite cylinders was by Green [18]. An early model to and Nfas functions of y can be derived analytically. The
attempt to explain experimental data considered point approximation x = 0 applies to diamagnets, paramagnets,
and saturated ferromagnets. N,,, for 25 values of y from
Manuscript received March I I , 199 1,
The authors are with the Electromagnetic Technology Division, National
0.2 to 1000 were obtained accurately to four significant
Institute of Standards and Technology, Boulder, CO 80303. figures by Brown [29] from a calculation of self-induc-
D.-X. Chen is on leave from the Electromagnetism Group, Physics De- tance [30] and listed as a table in Brown’s book [31].
partment, Universitat Autbnoma de Barcelona, 08 I93 Bellaterra, Spain.
J. A. Brug is on leave from the Thin Film Department, Hewlett-Packard
Crabtree [32] obtained the same values for the average
Laboratories, Palo Alto, CA 94303. demagnetizing factor by integration of the local field over
IEEE Log Number 9102083. the cylindrical volume. Moskowitz et al. extended

U . S . Government work not protected by U.S. Copyright

1.- i-r
3602 IEEE TRANSACTIONS ON MAGNETICS, VOL. 27, NO. 4, JULY 1991

Brown’s method to cylinders of polygonal cross section ders. Brug and Wolf [57] calculated the magnetization
[33], and KaczCr and Klem extended it to hollow cylin- distribution in disks and obtained the local demagnetizing
ders [34]. Nffor x = 0 was calculated exactly by Joseph factor for materials that undergo phase transitions.
[35]. Approximate values for N , and Nf for x = 0, ac- In Zijlstra’s book [58], Nf and N , are plotted. These
curate for large y, were calculated by Vallabh Sharma types of graphs and tables appear in other books on mag-
using uniformly magnetized volume elements [36]. Sato netism and magnetic materials, and they are widely used,
and Ishii [37] obtained a simple expression to approxi- sometimes inappropriately, in magnetic measurements of
mate N , for x = 0. Chen and Li [38], [39] obtained Nf ferromagnetic, ferrimagnetic, weakly magnetic, and su-
for x = 0 using magnetostatic potential calculations. perconducting materials. However, there remain some
The susceptibilities x = - 1 and x -+ 00 correspond to problems. For x = 0, the most accurate case, the number
perfectly diamagnetic and ideally soft ferromagnetic ma- of y values for N , and Nfis insufficient for accurate in-
terials, respectively. N , and Nffor these susceptibilities terpolation. For x # 0, almost all books give results ob-
were first treated by Taylor for perfectly conducting cyl- tained before 1950, and there are no data for x < 0. For
inders [40], [41]. He developed a method introduced by long cylinders (y > lo), there is a lack of data on the x
Smythe that expressed charge densities on the side and dependence of Nf,and there are no data on N,. For short
ends in terms of a set of orthogonal polynomials, and ex- cylinders (y < lo), there are even less data, and those
panded the electrostatic potential at the cylinder center that exist have large errors because they were obtained by
[42]-[44]. Taylor calculated electric and magnetic polar- extrapolation. In summary, there is no complete picture
izabilities for conducting cylinders for 0.25 Iy 5 4 in for the y and x dependence of Nfand N,.
both the longitudinal and transverse directions. N,(m) can In this paper, we calculate Nf and N , for a complete
be deduced from his electric polarizability results because range of y and x. Susceptibility x is traditionally assumed
of the analogy between electrostatics and magnetostatics. to be constant in the material and is therefore defined as
Because his calculation for magnetic polarizability was M / H , where M is the magnetic moment per unit volume
for a uniform quasi-static but nonpenetrating applied field, and H is the internal magnetic field. For the case x = 0,
N,( - 1) can also be deduced from his results. According in which the magnetization is uniform, we give 61 exact
to Taylor, his convergence error was less than 0.1 % for inductance calculations of N , and Nffor lop5 Iy 5 lo3.
the longitudinal direction. For x # 0, more elaborate methods are used. For y >
Using a similar approach with simpler base functions, 10, the variation of magnetization across the radius of the
Templeton et al. calculated axial Nffor x -+ 03 for 0.05 cylinder is negligible at the midplane, and we calculate
I y I250 [45], [46]. The fact that the side and end- Nfas a function of y and x (- 1 Ix < 00) based on the
pole densities have basically a 6-’13 dependence, where one-dimensional model of Stablein and Schlechtweg [23].
6 is the distance from the corner, was used to construct Unlike them, we use Taylor expansions for M ( z ) , calcu-
the set of polynomials. To estimate their error, Templeton late the demagnetizing field H d ( z ) directly at 25 points
and Arrott calculated the root-mean-square deviation of along the axis, and obtain more accurate results. The
the normalized potential from 0 and found it to be less model is also applicable to N , for x -, 00. For 0.01 Iy
than 0.3 1% [45]. Compared to an approximate formula I50, a two-dimensional finite element method is used
with 8 adjustable parameters, the deviations of their 12 that takes into account the variation of magnetic pole den-
computed Nf(oo)values were less than 0.25 %. The work sity along the side and ends of the cylinder. Values of N ,
was based on their earlier magnetostatic analysis of the and Nfare given for -1 Ix < 00.
magnetization process in soft ferromagnetic cylinders with
constant end-pole densities [47], [48]. The details of the 11. FLUXMETRIC AND MAGNETOMETRIC DEMAGNETIZING
calculation were published by Aharoni. who also calcu- FACTORS
lated the self-energy of cylinders [49], and more gener-
The demagnetizing correction is nontrivial for samples
ally, cylinders with nonuniform magnetization [50].
in open magnetic circuits. An exact correction can be ob-
For susceptibilities other than 0, - 1, and 00, different tained only for ellipsoids [4], [59], [60], where both the
techniques have been used. Archer and Guancial [511 and magnetization M and the demagnetizing field Hd are uni-
Fawzi et al. [52] calculated the distribution of magneti- form under a uniform applied field H,. If the three prin-
zation and magnetic field in long cylinders with large sus- cipal ellipsoid axes coincide with the x , y , and z axes, the
ceptibilities using volume and boundary integral equa- internal field is
tions. Using experimental resistance network analogs,
Okoshi [53] obtained Nffor x -, 00, and Yamamoto and H = H, + Hd = H, - NM, (1)
Yamada [54] obtained Nfand N,,, for large x.
Several papers have treated demagnetizing factors at where N is the demagnetizing tensor,
points. Joseph and Schlomann [55] solved for local de-
magnetizing factors in uniformly magnetized cylinders
and used a series expansion to account for nonuniform
magnetization. Kraus [56] determined the complete local
demagnetizing tensor for uniformly magnetized cylin-
I II I

CHEN et al.: DEMAGNETIZING FACTORS FOR CYLINDERS 3603

with where F(k,) and E(k,) are the complete elliptic integrals
N, + N,, + Nz = 1. (2b)
of the first and second kind of modulus k,, which is de-
fined by
If the applied field is along one of the principal axes, we k: = a 2 / ( a 2 Z2), + (7)
have
and po is the permeability of vacuum.
H = Ha + Hd = H, - NM, (3) An exact formula for the mutual inductance L, of the
where N is called the demagnetizing factor. In SI units, 0 same thin solenoid and a coaxial single-turn loop of the
5 N I 1. In cylindrical samples, which are commonly same radius at its midplane is [62]
used in magnetic measurements, the demagnetizing field Lm = (pOna/krn> [F(krn) - E(km)I, (8)
is not uniform, and two kinds of susceptibility-dependent where the modulus k, is defined by
demagnetizing factors are defined.
If the sample is located in a uniform applied field Ha ki = 4a2/(4a2 + 12). (9)
along its axis, the fluxmetric (or ballistic) demagnetizing Cohen [63] derived an exact general formula for the
factor Nf is defined as the ratio of the average demagne- mutual inductance of two concentric coaxial thin sole-
tizing field to the average magnetization at the midplane noids (denoted by subscripts 1 and 2). We have used it
perpendicular to the axis. The magnetometric demagne- successfully for these calculations, as an alternative to (6)
tizing factor N , is defined as the ratio of the average de- and (8), with a2 = a l , in the limits l2 = I , (for N,) and l2
magnetizing field to the average magnetization of the en- -+ 0 (for N f ) .
tire sample [58]:
B. Relationship between N , and L,, Nf and L,
(4) The flux density B in a magnetic material is related to
the internal field H and the magnetization M : B = p O ( H
+ M ) , where H is, in general, related to the applied and
demagnetizing fields as defined for ellipsoids in (1). Thus
+ +
B = po(Ha Hd M ) . Following Brown [29], we define
Nf and N , are functions of the ratio y of cylinder length B' as the Amperian flux density:
to diameter and the susceptibility x of the material. For
ferromagnetic or ferrimagnetic materials, this x should be B' = B - poHa = po(Hd + M). (10)
regarded as an effective x , similar to the differential sus- When x = 0, a cylinder in an axial field has a uniform
ceptibility dM/dH at the corresponding magnetic state. magnetization M . An ideal thin solenoid carrying current
In [58], the definition of N , is limited to x = 0. I through n turns over a length 21 is equivalent, with re-
spect to the B' field, to a longitudinally magnetized cyl-
111. N , AND Nf FOR x = 0 DETERMINED
BY
inder coincident with it [29]. Thus the cylinder can be
INDUCTANCE CALCULATIONS modeled as a solenoid with the same M , and its average
Hd can be obtained from M and average B' using (10).
Brown [29] showed how N , could be determined using We take the solenoid as having one turn (n = l ) , so
a self-inductance calculation in which a uniformly mag- M = I/(2l). (1 1)
netized cylinder was modeled as a solenoid. In fact, both
N , and Nf may be calculated using the mutual inductance For the entire volume, we can obtain the average B'
of two model solenoids of the same diameter. N , is ob- from the average flux 9 in the solenoid as
tained when the solenoids have the same length, and the
(B' ) = +/(m2). (1 2 4
problem reduces to the self-inductance calculation. Nf is
obtained when the length of one of the model solenoids Thus the average demagnetizing field can be obtained
approaches 0, and the problem is that of the mutual in- from (10) and (12a) as
ductance of a solenoid and a single-turn loop located at
its midplane. In this section we calculate exact values of (Hd) = +/(p~na*-
) M. (13 4
N , and Nffor x = 0 and a wide range of y. The definition of self-inductance is
L, = +/I. (144
A. Formulas for Inductance
From ( l l ) , (13a), and (14a), we obtain the final expres-
An exact formula for the self-inductance L, of a thin sion forthe magnetometric demagnetizing factor:
solenoid of length 21, radius a, and number of turns n is
N, -(Hd)/M = 1 - 2&/(pona2). (15)
[611
For N f , we obtain the average B' at the midplane from
the flux a0in the one-turn secondary loop of radius a:
* [12F(k,) + (a2 - 12)E(k,)] - a 3 } , (6) ( B ' ) = +0/(na2). ( 12b)
3604 IEEE TRANSACTIONS ON MAGNETICS, VOL. 21, NO. 4, JULY 1991

The average demagnetizing field is TABLE 1


AND MAGNETOMETRIC
EXACTFLUXMETRIC DEMAGNETIZING Nf
FACTORS
AND N,,, FOR x = 0"
(Hd) = *o/(pona2) - M. (13b)
The definition of mutual inductance is ~~
Y NAO) Nf (0) N

L, = *o/z. (14b) 0.00001 0.9999 0.9999 1 .m


0.o001 0.9994 0.9993 0.9998
The final expression for the fluxmetric demagnetizing fac- 0.001 0.9950 0.9949 0.9984
0.01 0.9650 0.9638 0.9845
tor is 0.02 0.9389 0.9364 0.9694
0.03 0.9161 0.9124 0.9546
Nf - (&)/A4 = 1 - 21L,/(po~a*). (16)
0.04 0.8954 0.8905 0.9402
0.05 0.8764 0.8703 0.9262
Equations (15) and (16) have been derived, by direct in- 0.06 0.8586 0.8513 0.9125
tegration rather than inductance formulas,' by Joseph [35]. 0.07 0.8419 0.8333 0.8991
0.08 0.8261 0.8163 0.8860
0.09 0.8110 0.8001 0.8733
C. Results 0.10 0.7967 0.7845 0.8608
Values of N,(x = 0) and Nf(x = 0) as functions of y 0.12 0.7698 0.7553 0.8367
0.14 0.7450 0.7281 0.8137
(= Z/a) computed using (6), (15), (8), and (16) are given 0.16 0.7219 0.7027 0.7917
in Table I. For N,, the data agree with those given by 0.18 0.7004 0.6789 0.7706
Brown [29], [31]. For Nf,the data agree with those ob- 0.20 0.6802 0.6565 0.7505
0.22 0.6611 0.6352 0.7312
tained by Joseph [35] and by Chen and Li [38]. In Table 0.24 0.6432 0.6151 0.7126
I we also give N for ellipsoids of revolution with longi- 0.26 0.6262 0.5960 0.6948
tudinal axes 21 and transverse axes 2a calculated from 0.28 0.6101 0.5778 0.6778
0.30 0.5947 0.5604 0.6614
well-known formulas [4], [59], [60]. 0.32 0.5801 0.5438 0.6456
0.34 0.5662 0.5279 0.6304
IV. ONE-DIMENSIONAL MODELFOR LONGCYLINDERS 0.36 0.5530 0.5127 0.6158
0.38 0.5403 0.4982 0.6017
A. Calculation of M 0.40 0.5281 0.4842 0.5b82
0.45 0.4999 0.4516 0.5563
Assume that a cylinder of length 21 and diameter 2a is 0.50 0.4745 0.4221 0.5272
located in a uniform applied field Haalong the z axis, as 0.55 0.4514 0.3952 0.5005
0.60 0.4303 0.3705 0.4758
shown in Fig. 1. The material has constant susceptibility 0.4110 0.3480 0.4531
0.65
x, which leads to 0.70 0.3933 0.3273 0.4321
0.75 0.3770 0.3082 0.4126
B = po(M + H) = poM(1 + l / x ) (17) 0.80 0.3619 0.2905 0.3944
0.90 0.3349 0.2592 0.3618
at any point inside the cylinder. Since V * B = 0, the 1 .o 0.3116 0.2322 0.3333
volume magnetic pole density, proportional to V M, - 1.1
1.2
0.2911
0.2731
0.2089
0.1886
0.3083
0.2861
equals 0 inside the cylinder; that is, all poles are on the 1.3 0.2572 0.1710 0.2664
surface. <.'' 1.4 0.2429 0.1555 0.2488
We further assume for this one-dimensional model that 1.6 0.2186 0.1298 0.2187
M,, the z component of M, is uniform in each cross sec- 1.8 0.1986 0.1096 0.1941
2.0 0.1819 0.09351 0.1736
tion of the cylinder, and can be expressed by a scalar 2.5 0.1501 0.06544 0.1351
quantity as 3.O 0.1278 0.04799 0.1087
3.5 0.1112 0.03653 0.08965
n
4 0.09835 0.02865 0.07541
M(z) = M,(z) = c M2,(Z/O2',
r=O
(18) 5
6
0.07991
0.06728
0.01889
0.01334
0.05582
0.04323
7 0.05809 0.009904 0.03461
where M2,(i = 0, 1, . , n ) are constants. ( B , Ha, and 8 0.05110 0.007635 0.02842
Hd can also be written as scalar quantities.) 9 0.04562 O.Oo6061 0.02382
For a section of cylinder of length dz at z , 10 0.04119 0.004927 0.02029
20 0.02091 0.001245 0.006749
50 0.008438 0.0001999 0.001443
$M ds = na2dM(z) + 2aaMr(z)dz = 0, (19a)
100
200
500
0.004232
0.002119
0.0008483
0.00004999
O.oooO1250
0.00000200
0.0004299
0.0001248
0.00002363
-
because $ M ds = V * M d v and V * M = 0. M r ( z ) lo00 0.0004243 0.00000050 O.OOOOO6601
is the radial component of M at the side surface. Substi- 'Factors were calculated as functions of y using inductance formulas.
tuting a ( z ) = p o M r ( z ) gives, on the side surface, the sur- For comparison, N is the demagnetizing factor for ellipsoids.
face magnetic pole density
n
On the end planes of the cylinder, we have uniform sur-
face magnetic pole densities:
U(Z) = -;poa dM(z)/dz = -(poa/l) iM21(~/l)2r-'. n

(19b)
U(* 1) = kp o ~ ( l =
) 5 PO ,E ~
r=O
2 i . (20)
I II 'I

CHEN et al. : DEMAGNETIZING FACTORS FOR CYLINDERS 3605

X uy -I----

-'-?--'-

I
-e

Fig. 1. Cylinder geometry and coordinate system.


IHa
where
F2(33 = ;(I

H d , we have
+ {)[(l + { ) 2 + t12]-1/2
+ i ( l - { ) [ ( t - {)2 + a2]-'/2 - 1.
From (3), and considering the

H~(z-
) M(z)/x
z dependence of H and

= -Ha.
Rewriting the variable {in H d l , F I I ,H d 2 , and F2 as z , and
substituting (22a) and (24a) into (21), and (18) and (21)
into (25), we obtain
(24b)

(25)

Although the side surface magnetic pole density is a n


function of z only, our assumptions of both uniform M,
and uniform a( k I) are, in fact, contradictory for constant
c [F,,(z) + F2(z) - (z/l)21/XlM21 = -Ha.
r=O
(26)
x; a uniform a +_ 1) and the side pole density would pro- This is a general equation relating the expansion coeffi-
duce a nonuniform Hdzr the z component of the demag-
cients M2rof magnetization, the applied field H,, the sus-
netizing field, which would lead to a nonuniform M,.
ceptibility x , and the position z for a cylinder of length 21
However, if y (= l / a ) is large ( 2 lo), and if we consider
and diameter 2a. In our problem, H,,I, a, and x are given,
only Nf,for which the middle part of the cylinder is more
impoltant , this one-dimensional model is a good approx-
and the n +
1 coefficients M,,are unknown. We can
choose n + 1 positions, z = zO, zI,* * * , z,, and get a
imation.
set of n +
1 linear equations. M2, (i = 0, 1, * , n ) are -
B. Calculation of Nfand N,,, for Long Cylinders then obtained by solving these equations simultaneously.
Nfand N,,, can be obtained according to (4),(5), and (25)
(7 1 10)
as
In this model, in a given Ha, the demagnetizing field
Hd is a function of z : Nf = -Hd(o)/MO = Ha/MO - l/X, (27)
H d ( z ) = Hdl(z) + Hd2(z), (21) N, = - ( H d ) / ( M ) = H a / ( M ) - l/x, (28)
where Hdl and H d 2 are the demagnetizing fields produced where the ( ) brackets denote the volume average, and

s:
by the side poles and the end poles, respectively.
n
At a point z = {on the z axis,
+I
( M ) = I-' M(z) dz = c M2,/(2i + 1).
1=O
(29)
HdI({) = -a(2/h)-' fl(z>(z - In principle, the larger the number of terms in the ex-
-I
pansion equation (26), the more accurate are the results.
[(z - {)2 + a2Ip3/*dz However, if n is too large, the computed function M ( z )
oscillates. For computation, n = 12 is a practical choice,
and
where z, = if/l2, (i = 0 , 1, * * * 9 121, (30)
P +I
H, = 1, (314
rc=! (3 1b)
Table I1 and Fig. 2(a) and 2(b) give the calculated Nf
as functions of y and 2 for -1 Ix I lo9. When x +

03, Hd is uniform in the cylinder ( H d = - H a ) , and M


close to the ends is very small. As a result, values of N,( x
-+ 0 0 ) obtained from this model are expected to be fairly

accurate due to rather small end effects. Similar data were


presented in 1641. Fig. 3 gives N , and Nfas functions of
y when x = 0.001 and lo9. The demagnetizing factors N
for ellipsoids (dashed curve) are located between N,,, and
Nffor x -+ 00.
3606 IEEE TRANSACTIONS ON MAGNETICS, VOL. 21, NO. 4, JULY 1991

TABLE I1
N, AS FUNCTION
OF y AND x CALCULATED MODEL'
USING THE ONE-DIMENSIONAL

Y 10 20 50 100 200 500 1000 2000

N/
X (IO-)) (IO-)) IO-^) (IO-') (IO-') (10-7 (IO-') (IO-')

-1 3.965 1.130 1.900 4.776 1.196 1.917 4.796 1.199


-0.8 4.239 1.162 1.931 4.852 1.215 1.945 4.865 1.216
-0.4 4.642 1.210 1.972 4.944 1.237 1.980 4.950 1.237
0 4.963 1.248 1.999 5.000 1.250 2.000 5.000 1.250
0. I 5.037 1.256 2.005 5.011 1.252 2.004 5.010 1.252
0.2 5.108 1.265 2.011 5.021 1.255 2.007 5.018 1.255
0.5 5.315 1.289 2.026 5.047 1.260 2.016 5.040 1.260
1 5.639 1.328 2.049 5.082 1.268 2.027 5.067 1.267
2 6.231 1.401 2.089 5.135 1.277 2.040 5.100 1.275
5 7.695 1.616 2.191 5.247 1.293 2.058 5.143 1.286
10 9.377 1.967 2.352 5.394 1.308 2.071 5.168 1.292
20 11.23 2.565 2.688 5.664 1.332 2.083 5.187 1.296
so 13.21 3.600 3.803 6.494 1.395 2.106 5.211 1.299
100 14.14 4.294 5.448 8.132 1.500 2.141 5.238 1.302
200 14.68 4.776 7.497 11.77 1.734 2.210 5.288 1.305
500 15.03 5.126 9.800 19.91 2.661 2.414 5.438 1.316
IO3 15.15 5.255 10.90 26.32 4.221 2.808 5.684 1.333
2 X IO3 15.21 5.322 11.53 31.12 6.275 3.889 6.186 1.369
5 X IO3 15.25 5.363 11.95 34.78 8.642 7.570 8.223 1.474
IO4 15.26 5.377 12.09 36.16 9.779 11.68 13.12 1.669
2 X IO4 15.27 5.384 12.16 36.88 10.44 15.47 22.71 2.237
5 X IO4 15.27 5.388 12.21 37.33 10.87 18.77 38.36 4.620
IO5 15.27 5.390 12.22 37.48 11.02 20.11 47.73 7.811
2 X IO5 15.28 5.390 12.23 37.55 11.10 20.83 53.77 11.10
5 X IO5 15.28 5.391 12.23 37.60 11.14 21.28 57.94 14.19
IO6 15.28 5.391 12.24 37.62 11.16 21.44 59.43 15.50
IO' 15.28 5.391 12.24 37.63 11.17 21.58 60.82 16.83
IO9 15.28 5.391 12.24 37.63 11.17 21.59 60.98 16.99
rn 15.30 12.11 37.20

"The row for x = 0 is comparable to data for Nf(0)in Table I. The last row gives A',(=) calculated by
Templeton and Arrott [45].

I" 10 100
Y Y
(a) (b)
Fig. 2. Calculated Nffrom the one-dimensional model. (a) For 10 5 y < 200, the curves from top to bottom are for x = OD,
1000, 300, 100, 30, 10, 3, 1, 0, and - 1. (b) For 10 5 y < 1000, the curves from top to bottom are for x = OD, IO5,3 X IO4,
lo4, 3 x lo3, lo3, 300, 100, 0, and - 1 .
CHEN et al.: DEMAGNETIZING FACTORS FOR CYLINDERS 3607

and use the resulting distribution to obtain the magneto-


metric and fluxmetric demagnetizing factors.
The method of dividing the surface of a magnetized
body into a set of interacting elements of uniform pole
density has been used before for the calculation of the
magnetic fields of rectangular bodies. Ruehli and Ellis
[65] assumed a constant susceptibility, and Normann and
Mende [66] used a field dependent magnetization with the
assumption that the volume pole distribution is negligible.
Both of these studies were interested primarily in the field
and magnetization distributions rather than the demagne-
tizing factors. A method that involves dividing the vol-
ume into uniformly magnetized elements was used by
Brug and Wolf [57] for the case of thin disks that undergo
magnetic phase transitions. They used a demagnetizing
matrix that was derived by Hegedus, Kadar, and Della
Torre [67], [68] for interacting volume elements in cylin-
drical geometries. Volume elements have also been used
by Soinski [69] for rectangular and ring-shaped samples.
10 100 1000 A method of obtaining demagnetizing fields in bodies of
Y arbitrary shape was presented by Vallabh Sharma [36]
Fig. 3. N,,, and N,for 10 5 y < 1000. The solid curves from top to bottom using rectangular volume elements. Templeton and Arrott
are N,(x = O ) ,y,(x --* m), N,(x -t a),and N,(x = 0). The dashed [45] used the principle that the magnetic potential is 0 at
curve is N for ellipsoids. each point on a grid inside a body with infinite suscepti-
bility. They calculated the demagnetizing factors of cyl-
Earlier results for y > 10 exist for Nf but not for N,. inders and bars and later extended this to the case of a
Wurschmidt’s result for y = 50 and x 00 is 3.4%
+
material that saturates [70].
smaller than our result [20], [21]. Neumann and War- The following method solves for the pole distribution
muth’s results deviate from our results by 1.5%, O.O%, at the surface of a cylinder for an arbitrary value of the
-3.4%, -5.3%, and -8.6% f o r y = 10, 20, 50, 100, susceptibility. The surface is divided into rings of area
and 1000 [22]. For x -+ 0 0 , the maximum deviation of 7~ ( r ; - r:) on the end planes of the cylinder and into rings
the results of Stablein and Schlechtweg from our results of area 2na(z2 - zl) on the side surface of the cylinder.
is 3 % ; but for their smallest x (12.56), the maximum dif- The cylinder diameter is 2a, rl and r2 are the inner and
ference is above 30% 1231. Templeton and Arrott’s [45] outer radii of the end-surface ring, and z , and z2 are the
results for x + 00 (Table 11) are the most accurate, and
side-surface ring limits. The z and r components of the
our data agree within 1.1 % . Compared with the exact re- demagnetizing field at a point i that results from the sur-
sults for x = 0 in Table I, the data in Table I1 have errors face poles uJ at each ring are given by
o f 0 . 7 % , 0.2%, and0.0% f o r y = 10, 20, a n d y L 50.
HLz = - c NY a , / p , , (324
v. TWO-DIMENSIONAL MODELFOR SHORT CYLINDERS J

A . Calculation of Surface Pole Density


The magnetization, in general, varies throughout the
where, for example, N ! is a scalar factor that relates the
cylinder in both the radial and axial directions. A calcu-
surface pole on the jth ring to the r component of the de-
lation of demagnetizing factors for a short cylinder (y < magnetizing field that it produces at the ith ring. A method
10) must take this variation into account, especially near
of calculating these factors is given in the Appendix.
the comers, where the magnetization sharply diverges for
The interactions between the surface poles at each ring
susceptibilities far from 0. As in the one-dimensional
is specified using the equation M = x(H, + H d ) . The
model, we assume that the cylinder consists of material
demagnetizing field Hd , given in (32), is written in terms
with constant x, so the demagnetizing fields are com-
of the magnetization at each ring using u j / p o = n * MJ
pletely specified by the surface pole density.
To obtain the distribution of poles on the surface for a
= M i , where n is the unit vector outward normal to the
surface of the ring, and MJis the magnetization at the jth
specified susceptibility, we divide the surface of the cyl-
ring. The resulting equations are for the magnetization at
inder into a set of nonoverlapping elements of uniform
each point,
pole density. With the cylindrical symmetry there is no
azimuthal dependence of the pole distribution, and we are
therefore able to use elements in the shape of rings about
the central axis. We solve the set of equations that de-
scribes the interaction between these rings of surface poles
3608 IEEE TRANSACTIONS ON MAGNETICS, VOL. 27, NO. 4, JULY 1991

TABLE 111
Nf VALUES CALCULATED AND VOLUME
BY SURFACE METHODS(NjAAND N,,,) USING 74 SIDE RINGS

X -1 0.0001 10000

Y Nfl Nf$ Nfl Nf, Nf,.

0.1 0.82568 0.82582 0.78460 0.7846 1 0.73312 0.064998


1 0.23196 0.23195 0.23258 0.23225 0.22912 0.015670
10 0.018951 0.0042004 0.0081227 0.0049267 0.016904 0.00099941

where we have constrained the applied field to be in the


axial direction. Because only the normal component of M
is needed at each point, the equations can be rewritten as
H, = x - ' M i + c NYML
J
where Hd and M, are averaged over the midplane for Nf
and over the volume for N,.
For the first method, using (37a), N f , , is calculated from
(rings on the top end-plane), (34a) ( M,).We have $ M ds = 0 for a material with constant
H, = -x-'M; + N,"M$ susceptibility. Therefore, for each transverse cross sec-
J tion of the cylinder, the average M, multiplied by the
(rings on the bottom end-plane), (34b) cross-sectional area S equals the surface integration of M,,
over the cylinder surface above the cross section. In the
o = x - l +~ C~N Y M J , case of N f , the cross section is taken to be the midplane.
J
We have, from (37a),
(rings on the side surface). (34c)
With the surface of the cylinder divided into n rings, there
will be a set of n simultaneous linear equations. These
equations are solved for M,,at each ring using matrix in- (38)
version [7 11. where SI is the surface consisting of the top half of the
The number of rings required to adequately specify the cylinder, sj is the area of thejth ring, and the sum is over
surface pole density depends on how rapidly the pole dis- the rings on the top end-plane and on the side surface
tribution varies. For both large susceptibilities and sus- above the midplane. To calculate N,, a series of cross
ceptibilities near - 1, the pole distribution diverges sections corresponding to each side-surface ring is con-
sharply near the corner of the cylinder. To reduce the total sidered, and the volume-averaged M, is calculated from a
number of rings, the density of the rings is made roughly weighted average over these cross sections.
proportional to the pole density. For the case of infinite The second method, using (37b), requires a calculation
susceptibility, the pole density near the corner at the end- of ( Hd ) . We use (32a), with i denoting an interior point
plane is approximated as [45] of the cylinder. Again, the average is taken over the mid-
uo(r, * 1 ) * [ ( 1 - r / a ) - 1 / 3- (1 + (35) plane for Nf and over the volume for N,.
Since the first method involves surface flux calcula-
We let the width of the rings A r be inversely proportional tions, while the second involves interior field calcula-
to uo near the comer of the cylinder, which gives tions, we refer to them as the surface and the volume
+ +
A r = {ao a l [ ( l - r / ~ ) - "-~ (1 r / ~ ) - ' / ~ ] } - ' . methods, respectively. Table I11 gives some examples of
the results obtained from both methods. We see from this
(36) table that the results do not agree, especially for large y
A similar equation determines the width of the rings on or large x. The differences are even larger for N,. The
the side surface of the cylinder. The adjustable parameters source of the disagreement is a systematic error that is due
a. and a I are determined by the total number of rings used almost entirely to the finite number of elements with which
in the calculation. A total of 148 rings was used, 37 on the surface pole density is calculated.
each end-plane and 74 on the side of the cylinder. For y The field produced by a magnetic pole is very sensitive
= 10, 20, and 50, the calculations were repeated with 8 to the distance r between the pole and the point at which
end-plane rings and 132 side rings. the field is considered, with an r P 2dependence. In the
two-dimensional model, the division of the surface of the
cylinder into rings will produce a discretization error be-
B. Calculation of N , and Nj cause Hd is calculated in the center of a region of Uniform
pole density. This error is mainly due to the division of
After obtaining the surface pole distribution, there are
the side surface; at the center of each ring on this surface,
two ways to calculate the demagnetizing factors, based on
the normal component of Hd is produced by poles on the
the two equations
ring itself with significant contribution from poles on ad-
Nf,m = - ( H d ) / < M , > = H , / ( M , ) - x - I , (37a) jacent rings. For the end planes, the poles on the same
I U I

CHEN et a l . : DEMAGNETIZING FACTORS FOR CYLINDERS 3609

N,' = - x ( H i ; + l), (39c)


N-' = -x(Hdl + 1). (394
We assume that the error in the surface pole density cal-
culation is not too large so that
Hdu/Hd Mzs/Mz. (39e)
Thus we have from (39)
NJl + N x ) / [ N ( l + Nvx)l = (1 + N x ) / ( 1 + N,X).
(404
Solving (40a) for N gives the interpolation/extrapolation
0 equation
N = Nr)(1 + N s x ) / ( 1 + N,,x). (40b)
-0.5 We use (40b) and the results (up to 5 significant digits)
from the two methods to obtain the final Nf and N,. They
are presented in Tables IV and V and Figs. 5, 6, and 7.
The entries in Table I11 provide data for illustrative com-
putations. Values for y I10 were obtained with the 74-
side-ring calculation. The values for y = 20 and 50 are
1 0
r/a from the 132-side-ring calculation, which gives a smaller
discretization error for these cases. As an example of the
Fig. 4 . Variation of normalized M,, on the surface of a cylinder with y =
1 for selected values of susceptibility. On the left half of the figure, cor- worst case, Nf for x = 10 000 and y = 20 is 55.46 X
responding to the cylinder side, the curves from top to bottom are for x = lop4 when computed with 74 side rings, compared to
03, - 1 , 1, -0.5, and 0. On the right half, corresponding to the cylinder 53.85 x lop4 for 132 side rings.
end, the curves are for x = 00, 1 , 0, -0.5, and - 1 .
With the corrected Nf,,we can correct M,and Hd using
(39b) and (39d) and compare them with M z s and Hdt,cal-
culated from (39a) and (39c). We find that, for all Nf and
surface do not contribute to the normal component of H d , N,, the discretization error is less than l o % , giving a de-
except for the pole at the point being considered. A region magnetizing factor error of < l %. Exceptions are when y
of uniform pole density adjacent to the point does not pro- = 20 and x 2 100 for the 74-side-ring calculation, and
duce a discretization error. when y = 50 and x 2 100 or y = 10 and x = 10 000
Fig. 4 gives the surface pole distribution of a cylinder for the 132-side-ring calculation. Thus (39e) is rather ac-
with y = 1 for x = -1, -0.5, 0.0001, 1, and 10 000, curate in most cases and the interpolation/extrapolation
calculated with the two-dimensional model. The normal- approach is valid.
ized pole density is expressed as M , / I M,, ( r = 0) I. There- A comparison can be made with less-general published
fore, all the curves have the actual sign of the pole density results obtained for specific values of x . The results agree
when H , is positive. The curves shown are for the top half with the exact self-inductance calculations in Table I to
of the cylinder with z > 0; they are odd functions of z. four significant figures except for N , for y > 1 when, in
As can be seen from the figure, when x 00, the side
--$ the worst case, the error is 0.12%. The two-dimensional
poles are broadly distributed with a larger density, and calculation for N , can be compared with N , data deduced
this should give a relatively large error. The side poles from the work of Taylor [40], [41] as shown in Section
are narrowly distributed for x < 0, giving a small error. VI-A. From Taylor's data, we calculate five values each
For the cases of small or zero side-pole density (low or of N m ( - 1) and N , ( m ) for 0.25 Iy I4 (Table VI). The
zero x ) , the error will be small or zero. The error becomes greatest difference between the two-dimensional results
larger as y increases because the size of the rings on the and Taylor's method is 0.22% for N,( - 1 ) and 0.17% for
side surface also increases. The relative effect of the end N , (00). For N f ( 0 3 ) , the results are compared with the val-
poles increases and the error decreases as y decreases. ues calculated by Templeton and Arrott [45]. The largest
This systematic error is correctable if it is not too large. deviation is 0.98% at y = 10.
For the surface and volume methods, we write the de- The results of Nf for y 1 10 can be compared with our
magnetizing factors as Ns and N,,, and write ( M , ) and one-dimensional calculation (Table 11). The deviations of
( H d ) as M,, and Hdr,, respectively. For simplicity, the the one-dimensional from the two-dimensional results are
corrected quantities are written as N , M,, and H d . Ha is -1.2%, -1.2%, 0.6%, 1.8%, 0 . 7 % , and -4.2% f o r y
arbitrarily set equal to 1. From (37), we have = 10 and x = 10 000, 100, 10, 1, 0, and - 1, respec-
tively. For y = 20 and 50 the maximum deviation reduces
N = M z- s' - x -I , (394 to 1.2% and 1.9%, respectively. Such deviations are due
mainly to the approximations made in the one-dimen-
sional model.
3610 IEEE TRANSACTIONS ON MAGNETICS, VOL. 27, NO. 4, JULY 1991

TABLE IV
N, AS OF x A N D y CALCULATED
FUNCTION USING THE TWO-DIMENSIONAL
MODEL

X -1 -0.9 -0.7 -0.5 -0.3 0.0001 0.3 1 3 10 100 10000

Y N,

0.01 9781 9728 9690 9669 9655 9639 9628 9610 9586 9562 9548 9544
0.02 9576 9513 9453 9418 9393 9365 9345 9313 9268 9225 9195 9191
0.03 9383 9316 9242 9196 9163 9125 9097 9052 8990 8929 8888 8881
0.04 9201 9132 9048 8993 8952 8906 8872 8817 8739 8663 8612 8603
0.05 9028 8957 8865 8803 8756 8704 8664 8599 8508 8419 8359 8349
0.07 8704 863 1 8528 8454 8398 8334 8285 8206 8094 7984 7909 7879
0.1 8265 8188 8073 7988 7922 7846 7788 7693 7558 7425 7335 7321
0.2 7060 6974 6840 6738 6659 6566 6494 6378 6214 6054 5945 5929
0.3 6106 6016 5879 5776 5697 5605 5535 542 1 5262 5 109 5006 4990
0.4 5309 5223 5093 4999 4926 4843 4780 4679 4539 4405 4316 4302
0.5 4626 4549 4436 4353 4292 4222 4169 4084 3968 3858 3784 3773
0.7 3513 3466 3399 335 1 3314 3273 3242 3193 3125 3060 3016 3009
1 2319 2323 2325 2325 2325 2322 2320 2315 2304 2291 2280 2278
1.5 1199 1246 131 1 1354 1385 1418 1442 1477 1520 1556 1578 1581
2 689.6 738.4 808.3 857.3 894.0 935.2 965.6 1013 1078 1138 1177 1182
2.5 449.7 487.1 543.0 584.3 616.7 654.4 683.5 731.6 801.5 871.4 920.8 927.7
3 323.7 350.0 391.1 422.8 448.6 479.9 505.0 548.4 616.1 689.2 745.0 752.6
4 199.1 212.3 233.8 251.5 266.8 286.5 303.4 334.8 390.6 461.1 522.5 531.6
5 137.6 144.9 157.0 165.4 176.6 188.9 199.9 221.8 264.8 327.6 390.1 400.1
7 77.75 80.68 85.57 89.81 93.67 99.05 104.1 114.8 139.4 185.1 244.2 255.2
10 41.38 42.47 44.28 45.84 47.27 49.27 51.17 55.41 66.43 93.25 143.1 154.5
20 11.44 11.59 11.83 12.03 12.21 12.45 12.69 13.20 14.61 19.51 42.96 53.85
50 1.937 1.946 1.961 1.974 1.984 1.999 2.012 2.039 2.110 2.337

TABLE V
OF x AND
N , AS FUNCTION CALCULATED USING THE TWO-DIMENSIONAL
MODEL

X -1 -0.9 -0.7 -0.5 -0.3 0.0001 0.3 1 3 10 100 10000

0.01 9780 9730 9695 9677 9665 965 1 9642 9628 9608 9589 9577 9574
0.02 9576 9517 9464 9434 9413 9390 9373 9347 93 10 9275 925 1 9247
0.03 9385 9323 9259 9220 9 192 9162 9139 9103 9052 9000 8967 8960
0.04 9205 9 143 907 1 9026 8993 8955 8927 8882 8819 8754 8711 8703
0.05 9035 8974 8896 8845 8808 8765 8733 8680 8605 8529 8478 8469
0.07 872 1 8660 8575 8516 847 1 8420 8380 8315 822 1 8127 8062 8050
0.1 8300 8240 8149 8082 8029 7968 7920 7841 7726 7609 7529 7515
0.2 7200 7135 7029 6947 6881 6803 6741 6637 6485 6332 6224 6207
0.3 6393 6318 6200 6108 6035 5948 5880 5768 5604 5440 5327 5306
0.4 5766 5681 5550 545 1 5373 5282 521 1 5096 4930 4768 4657 4640
0.5 5260 5167 5025 492 1 4839 4745 4674 4558 4395 4239 4134 4118
0.7 4488 4382 4226 41 14 4029 3933 3862 3749 3596 3455 3363 3349
1 3692 3576 3410 3295 3210 3116 3047 2942 2804 2682 2604 2593
1.5 2860 2744 2580 2469 2388 2301 2238 2 144 2023 1921 1858 1850
2 2339 2229 2076 1973 1898 1818 1761 1675 1567 1475 1418 141 1
2.5 1979 1878 1737 1643 1574 1501 1449 1370 1270 1184 1131 1124
3 1717 1623 1494 1407 1345 1277 1229 1156 1063 981.8 930.2 922.7
4 1358 1277 1167 1094 1040 983.1 941.8 897.0 796.8 721.7 671.0 663.5
5 1123 1053 958.0 894.5 848.4 798.6 762.7 707.7 634.1 564.1 513.5 505.8
7 834.0 779.6 705.2 655.6 619.5 580.4 552.1 508.2 447.9 386.4 336.2 328.0
10 601.5 560.7 505.1 468.0 440.9 411.5 390.1 356.6 309.6 258.5 210.2 201.3
20 316.8 293.7 262.3 241.6 226.6 210.2 198.3 179.5 152.5 120.9 82.60 72.91
50 128.4 119.1 106.4 97.78 91.53 84.70 79.70 71.77 60.18 46.05

VI. DISCUSSION magnetization is not. For x = - 1, the nonuniform mag-


netization and nonuniform demagnetizing field in the cyl-
A . Nf,,f o r x = 0,0 0 , and -1
inder combine to exactly cancel the applied field so that
In ellipsoids, a uniform applied field produces a mag- the flux density B = 0. In these three cases, demagnetiz-
netization and a demagnetizing field that are both uni- ing factors can be calculated more accurately by introduc-
form. In cylinders, the magnetization and the demagne- ing electromagnetic scalar potentials. Moreover, there are
tizing field are both nonuniform except in two cases. For some simple relations among N,, Nmy,and NmZ,the mag-
x = 0, the magnetization is uniform but the demagnetiz- netometric demagnetizing factors along the three princi-
ing field is not. For x + 03, the demagnetizing field is pal orthogonal axes.
uniform (and exactly opposite to the applied field) but the For x = 0, a cylinder has N,,, Nmy, and N,, that are
I II I

CHEN er al.: DEMAGNETIZING FACTORS FOR CYLINDERS 3611

1.o equivalent to those of a possible ellipsoid of revolution,


with
N, + Nmy + N,, = 1, (41)
according to the more general theorem of Brown and Mor-
0.8
rish [72]-[74]. As examples from Table I, cylinders with
y = 0.22, 0.80, and 1.6 are equivalent to ellipsoids with
y = 0.30, 0.90, and 1.6. A cylinder with y = 0.9065 is
equivalent to a sphere ( N = 0.3333) [36]. A cube is also
0.6 equivalent to a sphere according to the theorem [75]-[77].
c
Experimentally, for weakly magnetic and saturated fer-
z romagnetic materials, these two shape-isotropic geome-
tries can be used as alternatives to spheres. However, the
theorem of Brown and Monish on the equivalence of a
0.4 body of arbitrary geometry (including a cylinder) and a
possible ellipsoid cannot lead to an a priori estimate of
the value of N , ( x = 0) except for a body with center
symmetry, such as a cube.
0.2 For cylinders, the transverse magnetometric demagne-
tizing factors are
N , = Nmy = i ( 1 - N m J . (42)

0.0 Equations (41) and (42) are valid for N , only when x =
0.01 0.1 1 10 0. For x > 0, the sum is less than 1, while for x < 0, it
Y is greater than 1. If Nf is considered, rather than N,,,, the
Fig. 5 . Nffor 0.01 5 y 5 20 and - 1 5 x < 01 from the two-dimensional
sum is always less than 1.
model, based on Table IV. The curves are for x = - 1 , -0.5, 0, 1 , 10, For a given y, values of N f ( x 00) and N , ( x -, 03)
-+

and OD. The curve labels refer to the left side of the figure. are rather close to N for ellipsoids. We can fit the Nf(03)
and N , (03) versus y data in Table I1 and Figs. 2 and 3 by
simple equations:
1.o N~(x -+ 03, y) = N ( y ) [ l + 0.12(log - 3)1, (43a)

Nm ( X + 00, 7) = N ( y ) [1 + 0.25 (log - I)], (43b)


where the demagnetizing factor N ( y ) for ellipsoids can be
0.8 found in Table I. In the range 10 Iy I 100, (43a) and
(43b) fit the model results with maximum deviations of
1 % and 1.5 % , respectively. Including the results for y <
10 in Tables IV and V, the lower boundaries of the y
range with the same indicated errors extend to 2 and 7 for
0.6
Nfand N,, respectively.
E Taylor calculated the anisotropic electric (CY)and mag-
z netic (0)polarizabilities of conducting cylinders [40],
[4 13. In Taylor’s terminology, “conducting” means
0.4 “without field penetration,” so electrically E = 0 and

o.2i
magnetically B = 0. Therefore we can relate the longi-
tudinal CY/ and transverse CY,, electric polarizabilities to
Nm(03), and the longitudinal and transverse P,, mag-
netic polarizabilities to N , (- 1). For a conducting solid
of revolution with volume Vo,
CY/,/(EOVO/O) = Nrnz(W)-’, e a )
CYff/(%~O) = Nm(03>-’, (ab)
0.0
0.01 0.1 1 10 POP///VO = [Nrnz(-l) - 11-’9 (UC)
Y cC,P,lV, = WA-1) - 11-13 (ad)
Fig. 6. N, for 0.01 I y 5 20 and - 1 5 x < 01 from the two-dimen-
sional model, based on Table V . The curves from top to bottom are for x where eo is the permittivity of vacuum. Equations (44a)
= - 1 , -0.5, 0, 1, 10, and 00. and (44b) are in accordance with the analogy between
3612 IEEE TRANSACTIONS ON MAGNETICS, VOL. 27, NO. 4, JULY 1991

1 0.25

0.20

0.1 0.15
z z“
zSE
0.10

0.01 0.05

0.00
0.01 0.1 1 10 1 2 3 4 5 6 7 8 9 10
Y Y
(a) (b)
Fig. 7. N , (solid curves) and N,(dashed curves) from the two-dimensional model. (a) Logarithmic scale, 0.01 I y I 20. The
curve labels refer to the right side of the figure. (b) Linear scale, 1 I y I 10.

TABLE VI B. General Rules for Nf and N, as Functions of x and y


N,,(m), N,.x(m), Nmz(-1),and Nmx(-I) FORSHORTCYLINDERS~
We give some general rules for the variation of Nf and
y N,,(m) N,,(m) E(m) N,??:(-I) N,, ( - 1 ) E(-l) N , with y and x based on the tables and figures. (a) For
0 1 0 1 1 0 1
a given x , both Nf and N , decrease with increasing y. This
0.25 0.5712 0.1618 0.8948 0.6764 0.2136 1.1036 is because the demagnetizing effect is an “end effect,”
0.5 0.4111 0.2371 0.8853 0.5258 0.2928 1.1114 although the demagnetizing fields depend on both end and
1 0.2590 0.3154 0.8895 0.3692 0.3669 1.1030 side surface poles. When y increases, the effect of cylin-
2 0.1409 0.3829 0.9067 0.2341 0.4237 1.0815
4 0.06635 0.4319 0.9302 0.1361 0.4596 1.0553 der ends on the midplane and the entire volume decreases.
m 0 0.5 1 0 0.5 1 (b) When both y and x are fixed, N,,,(y, x) > Nf(y, x).
“The data are obtained from the electric and magnetic polarizabilities of
That is, the end effect is weaker at the midplane. (c) N,,,
conducting cylinders calculated by Taylor 1401, 1411. E (m) and E ( - 1) are decreases with increasing x at any y. (d) With increasing
thesumsN,,+N,,+N,:forx~mandX= -1. x , Nf increases when y > 1 and decreases when y < 1;
there is a region around y = 1 where Nf is insensitive to
x. Rules (b), (c), and (d) give the following relation for
electrostatic shielding in a conductor and magnetostatic Nf(x) and N , (x) when y > 1:
shielding in the same body but with x -+ 03.
There is a relation between and at, [41], Nrn(-l) > Nrn(0) > Nm(03) > Nf(@)
POPI[ = -CYY,mO), (454 > Nf(0)> Nf(-l). (46)
which leads to an exact relation between N,,(m) and (e) For y > 1 , the ratio N,,,/Nfincreases with increas-
Nm, ( - 1) 9
ing y and decreasing x . When y increases from 1 to 10 to
1000, N,(O)/N’(O) increases from 1.37 to 8 to 800;
N,(@) = Nmy(@)= ;[I - Nmz(-l)]. (45b)
N, (03) / N f ( m ) is smaller and increases from 1.14 to 1.32
TableVIlists Nmz(oo),Nm(@),N,,(-l), andNm(-1) to 1.46. (f) The minimum x for Nf(x) > O.99Nf(m) in-
for short cylinders based on the CY and p data given by creases with increasing y. In fluxmetric ferromagnetic
Taylor. From this table we see that the sum of the N,’s measurements, this rule tells us that can be used
deviate from 1 by about 1 1 %. For other values of x , the for dM/dH larger than a minimum value that depends on
deviation should be less, and this could help one estimate y. For y = 10, 100, and 1000, the minimum values are
N,, for different values of x and y. 200, 5 X lo4, and lo6.
I II I

CHEN er al. : DEMAGNETIZING FACTORS FOR CYLINDERS

40 1.0002
I
1

-1
3.0 1.oooo .
n
0
n
e x =0.001
U v
fl
(a) 5 2 . 0 h0.9eoE -
N -0.5 N
W v
s! SI
1.0
0.001
0.98BE -
0.0
3.0 1 0.0002 -
a o.0000
X x =0.001
h
J-o.oooz -
z
-0.0004 -

0.0006 -

Y
CI

+
(c) <
U

n
1.0

W
b
0.0

1.o -0.0002 I
1
- 1 . 0 0
1.o
0.0
:fi 0.0
z"i:
Fig. 8 . M ( z ) / M ( O ) [graphs (a) and (d)]. H d ( z ) / H a [graphs (b) and (e)], and o ( z ) / o ( + l )[graphs (c) and (f)] as functions of z / l
for a cylinder with y = 20. The curves in graphs (a), (b), and (c) are for x = - 1 , -0.5, 0.001, 30, and lo5, respectively.
Graphs (d), (e), and ( f ) are for x = 0.001 on an expanded scale.

C. Minimum x for N f ( x ) > 0 . 9 9 N f ( m ) D. Position Dependence of M , H d , and 0

For x > 0, the demagnetizing effect resembles the re- We have explained rules (a), (b), and (f). To under-
sponse of an amplifier with negative feedback. The input, stand the other rules, we have to know the details of the
output, and gain of the amplifier are Ha, M , and x , re- position dependence of M , H d , and a. Fig. 8(a), 8(b), and
spectively. After operation, M becomes H d ( H d = - N M , 8(c) gives the curves of M ( z ) / M ( O ) , Hd(Z)/Ha, and
where N is the local demagnetizing factor), with sign op- a(z)/a(+Z) as functions of z for y = 20 and x = -1,
posite to that of Ha, which feeds back to the input. The -0.5, 0.001, 30, and lo5 computed from the one-dimen-
result is that M becomes smaller and nonuniform, while sional model with n = 10. Bloomberg and Arrott [78]
H becomes smaller than Ha. There is perfect feedback derived M ( z ) / M ( O ) ,using a similar approach, for 1 Ix
when x 03, and H = 0 everywhere. Since H / H a = (1
-+ < w and y = 100. Our curves for x = 0.001 are replot-
+ N x ) - ' , where N is the local demagnetizing factor, in ted in Fig. 8(d), 8(e), and 8(f) on finer scales. When Ha
order to have N f ( x ) almost equal to N f ( m ) ,(1 + N x ) must +
is positive, M ( 0 ) and a( 1 ) are positive for x > 0 and
be sufficiently large for H I H , to be small everywhere. negative for x < 0. Therefore, for x = -0.5 and -1,
Actually Nf is the smallest local N , and the origin of rule the signs for M ( z ) and ~ ( zare ) opposite to the signs shown
(f) may be understood as follows. for M ( i ) / M ( O ) and a ( z ) / a ( + Z ) in Fig. 8(a) and 8(c).
N f ( w ) decreases with increasing y. To have N f ( x ) al- Also, M ( z ) and &(z) are even functions of z , but a ( z ) iS
most equal to Nf(03), a smaller N f ( w ) must be balanced an odd function of z .
by a larger x to fulfill the condition of a sufficiently large
(1 + N x ) . Therefore, with increasing y,the minimum x E. x Dependence of N,,,/Nf
to satisfy N f ( x ) > 0 . 9 9 N f ( w ) increases. From Table I1 For x = lo5 (that is, x w) we can see in Fig. 8(b)
-+

we can deduce this minimum x to be k / N f ( w ) with 15 < that H d ( z )is a constant equal to -Ha in the entire cylin-
k < 17. der, which makes H very small (0for x -, 00) and M ( z )

-1- 7--lr--- - --
3614 IEEE TRANSACTIONS ON MAGNETICS, VOL. 21, NO. 4 , JULY 1991

finite. M ( z ) is approximately parabolic as shown in Fig. aged N,,, increases with decreasing x even for negative x.
8(a) and as already pointed out by other authors [2 11, [27]. Finally, we have the largest value of N , / N f at x = - 1 .
The average Hd is equal to - H a . For the midplane and When y < 1, the susceptibility dependence of Nf is the
the entire volume, respectively, the average M is M ( 0 ) opposite. We can explain this as follows. For oblate cyl-
and approximately 0.67M(O). Therefore, N,,,/Nf should inders, the end poles are the main contributors to the de-
be close to 1.5. magnetizing field Hd at the midplane. For a given ( M , )
When x decreases to 30, the variation of M ( z ) is grad- at the midplane, the end-pole density is smallest when x
ual for small ? and sudden in the end regions. This in- -+ 00 since, in this case, the poles are the most uniformly

creases the volume-averaged M to more than 0.67M(O). distributed on the entire surface. Therefore Nf is the
But because Hd ( z ) becomes rather z dependent and its ab- smallest. A smaller positive x repels the poles to the end
solute value at z = 1 is much larger than at z = 0 [Fig. regions, which gives rise to a larger end-pole density and
8(b)], N , / N f is larger. a larger N f . Although all the poles are distributed on the
As x decreases to 0.001, both M ( z ) and H d ( z ) remain ends when x = 0, the pole density on the ends is not the
nearly constant in Fig. 8(a) and 8(b). The reason is that largest. This is because, when x < 0, the end-pole den-
M ( z ) is very small compared to Ha, so &( isI
even
) sity has a sign opposite to that of the side-pole density
smaller than H,. The variation of the extremely small nearby, as can be seen from Fig. 4; thus the end poles are
H d ( z )cannot be seen in Fig. 8(b), and the modification of further enhanced. As a consequence, the end pole density
the field by such a small Hd ( z ) causes an invisible change increases continuously with decreasing x regardless of its
in M in Fig. 8(a). However, if we expand the scales [Fig. sign, and Nf takes its largest value when x = - 1 .
8(d) and 8(e)], we can see that the variation of M ( z ) and
H d ( z ) continues the trend of decreasing x from lo5 to 30. G. Error Transmission in Susceptibility Measurements
This makes N , / N f even larger. From the above analysis we see that, at present, the
When x is small but negative, both M ( z ) and H d ( z ) accuracy of N , and Nf can reach 1 % in general. To know
change their signs. At this point, Nf and N,,, continue the if this accuracy is good enough for the purpose of mag-
trend without sudden change. The situation for x < 0 can netic measurements, we examine the influence of the error
be seen from the curves for x = -0.5 and - 1 in Fig. in N,,, or Nf on susceptibility measurements. We consider
8(a) and 8(b). For these two cases, H d ( z )and the absolute a cylinder consisting of material with constant x in a lon-
value of M ( z ) remain constant until z > 0.81,and then gitudinal applied field H a . An external susceptibility xe
they suddenly increase. This makes N , / N f largest when can be defined as ( M ) / H a , where the average is taken
x = - 1 . Further discussion of this in terms of pole dis- over the whole volume or the midplane, depending on
tribution is given in the next section. whether N , or Nf is considered. From (3), replacing N by
Nf,,,,and M by ( M ), we obtain
F. x Dependence of Nf
To understand the susceptibility dependence of Nf for x = xe(1 - Nf,mxc)-l* (47)
y > 1, we focus on the magnetic pole distribution shown This equation is accurate under the above assumptions and
in Fig. 8(c) and 8(f). For the largest x , a(z) varies with definitions. From (47), the relative error in x caused by
z almost linearly on the cylinder surface except for the the calculation error of Nf,,,, and the measurement error of
regions close to the ends. This means that the magnetic x e can be derived as
poles are the most uniformly distributed on the cylinder,
and Hd (0), which has a greater contribution from the poles I A X / X I = INf.rnxl lANf,rn/N~ml + ( x / x A IAxelxel.
in the central region than from the ends, is the largest.
(48)
Thus Nf is its largest for the highest x .
When x is decreasing, the variation of u(z) is progres- On the right-hand side of (48), the first term is the
sively greater in the end regions, while the magnetic pole transmission error from the erroneous Nf,,calculation to
density in the central region becomes gradually lower. the x determination. This error equals the error in Nf,,
Therefore, Nf decreases with decreasing x . When x = 0, multiplied by a factor a1 = I N f , , x ( , which can be ob-
all the poles are at the ends, a(z) is 0, and Nf should be tained from the results of this work. We examine three
its smallest. In fact, Nf continues to decrease when x be- typical cases. In the first case, I x ( and a l are small; the
comes negative. The reason is that, although a(x) remains error in Nf,,,, can be large and still result in a small trqns-
0 in a large central region for x < 0, a(z) for z close to 1 mission error. For example, when I x( = 0.1, we have a l
increases with decreasing x and its sign is opposite to that < 0.1 since Nf,,,, < 1 , and less than 10% of the N f , , error
of a ( + 1 ) [Fig. 8(c)]. The side poles produce a field at is reflected in the final x result. The second case is a mag-
the center directed opposite to the field produced by the netometric measurement for x = - 1 . c y 1 , obtained from
end poles, so that the demagnetizing effect of the end poles Table V , is 0.37, 0.17, and 0 . 0 6 for y = 1, 3, and 10;
is partially compensated by the effect of the side poles. only a small part of the error in N,,, is transmitted to the
This makes Nf a little lower than for x = 0. The side poles final x result. In the third case, high-x materials are con-
close to the ends, with signs opposite to those of the end sidered. To reduce the error due to the large a l , flux-
poles, have a different effect on N,. They greatly increase metric measurements should be made with long samples,
the value of Hd in the end regions, so the volume-aver- since Nf < N,,, and Nf decreases with increasing y. To
I II I

CHEN er a l . : DEMAGNETIZING FACTORS FOR CYLINDERS 3615

ensure that aI < 1 based on Table I1 and Fig. 2, y should our results can still be used satisfactorily for long mag-
be 12, 58, 200, and 700, for x = lo2, lo3, lo4, and lo5, netically soft materials over a wide field range. We can
respectively. regard x as the normal susceptibility x,, = M , / H , , where
The second term in (48) is the transmission error due the subscript m denotes the maximum value at the end-
to the measurement error of xe. The corresponding factor points of a symmetric magnetization loop, and use Nf(x).
is a2 x/xe. For the first case where cyI is very small, To extract an unknown x from fluxmetric measurements
a2 is very close to 1 since x = xe, so the error in x is of xe on samples with known y, one uses Nf. But a knowl-
almost the same as the xe measurement error. For our sec- edge of x is required to select the appropriate Nf value.
ond case, from (47), we have a l + a2 = 1 . This is inter- The known xe and y and the unknown x and Nfare related
esting because, when y is small so that N, + 1 and a1+ by (47) and Table I1 or Fig. 2, so the unknowns can be
1, the x error is mainly due to the error in N,, no calculated simultaneously. An iterative process may be
matter how large the xe measurement error is. For the used. Nf is estimated based on the measured xe using Ta-
high-x case, if a1 = 1 , we have a2 = 2 from (47), leading ble I1 or Fig. 2, and x is calculated from xe and Nfusing
to a double transmission error to x from the xe measuring (47). Then a better estimate of Nf is made. Since the dif-
error. As a consequence, to obtain accurate results of x ferential susceptibility is field dependent in ferromagnets,
for high-x materials, both accurate Nf and accurate xe are this treatment involves some error. The resultant x is an
strongly required if y cannot be made very large. effective susceptibility xeff.Its value is between x,, and an
In summary, to determine x accurately, the higher the averaged differential susceptibility in the sample. We have
I X I of the material, the higher is the required calculation xeff= x,, when H , --$ 0 or H , + 03, or when the sample
accuracy of Nf,,. From a magnetic measurement point of has very large y and x,, and H , is close to H , ( x,,, ,). In
view, our calculation accuracy for Nf.,at x = 0 is more other cases, including the use of N, and Tables IV and V,
than adequate, and the number of calculated points is suf- the resultant xeffmay be larger or smaller than xn, de-
ficient for accurate interpolation over a wide range of val- pending on the measurement conditions.
ues. For other x values, the requirement for calculation For measurements of semihard ferromagnets with in-
accuracy will be determined by the particular purpose of trinsic coercivity H,, Zijlstra [58] has suggested a simple
the magnetic measurement; a 1% accuracy may be suffi- method to obtain a reasonably accurate loop in which the
cient for many uses but is inadequate for others. For ex- demagnetizing corrections for M and H are performed
ample, our 0.2% accuracy for x = -1 is required for using Nf(0)and Nf(03), respectively. With our results,
superconductor calibration standards. nonzero finite effective x (corresponding to the differen-
tial susceptibilities at H = H , and H = H,) can be used
H . Application to Materials with x > 0 more properly.
Most materials have x > 0, and our results can be used
for demagnetization corrections of their magnetic mea- I. Remarks Concerning x < 0 and Nonuniform x
surements. For materials with linear or nearly linear mag- For normal diamagnetic materials with uniform x, val-
netization curves, our Nf,, values are satisfactory. These ues of N, ( x = 0) are more than adequate for experimen-
include paramagnets, spin glasses, weakly magnetic ma- tal work. However, large negative values of x arise in ac
terials, and iron-powder cores and ordinary ferromagnets magnetic measurements of normal conductors and both ac
in the initial and saturation states. However, even in these and dc measurements of superconductors, where bulk
cases, some caution is required. We give an example be- magnetic moments have their source in eddy currents and
low. shielding supercurrents, respectively. These magnetic
For magnetometric measurements of weakly magnetic moments allow us to ascribe values of M , H , and x to
materials (x < 0.01) only very small demagnetizing cor- these materials.
rections are needed. However, such materials can also be In an ideal type-I superconductor, x = -1 because B
measured by fluxmetric methods, as recommended by at and the permeability po( 1 + x) both equal 0 at every point
least one measurement standard [79]. A source of error in in the material. The same applies to a normal conductor
fluxmetric measurements is if the sample diameter is less in an ac field when the skin depth is negligible compared
than that of the measurement coil. A large demagnetizing to its dimensions. Thus there is an equivalence between
effect would occur in the measurement of M because the these cylinders and a normal perfectly diamagnetic cyl-
measured flux linkage is contributed not only by the M of inder. For these cases, our values of N,,, ( x = - 1) and
the sample but also by the Hd within the coil volume pro- Nf( x = - 1) can be used. We have verified this experi-
duced by the sample’s poles. Furthermore, fluxmetric mentally with a magnetometric low-field ac susceptibility
measurements on weakly magnetic materials require many measurement of a niobium cylinder (y = 1.033) below
coil turns, which ensures that this error will arise. The the critical temperature. The susceptometer was cali-
error in x due to this effect can be as large as 30%, even brated using the known demagnetizing factor and dipole
if the requirements of [79] are followed [38], [64]. field of a superconducting niobium sphere [80]. We ac-
The magnetization curves of ferromagnetic materials counted for a 0.4% volume decrease of both the standard
are nonlinear, and it is difficult to assign to them specific sphere and the sample cylinder upon cooling to 4 K, and
x and Nf,,(x) values except in the initial state and when deduced a value of N, equal to 0.361 0.001 from (47),
approaching saturation, as mentioned above. However, with the uncertainty based on the measured scatter in xe.
3616 IEEE TRANSACTIONS ON MAGNETICS, VOL. 27, NO. 4, JULY 1991

Our two-dimensional calculations give 0.3622. Thus we where (P, and asrepresent the potential due to a ring on
see that cylindrical superconducting standards for mag- the end-plane and side surface, respectively. The param-
netic measurements for use at low fields and temperatures eters rl and r2 are the inner and outer radii of the ring of
can be made using accurate values of N , (- 1). poles on the end surface, and zI and z2 are the limits in
The results of this work have to be used more carefully the z direction of the ring on the side surface. The limits
for materials that do not have constant susceptibility. In of integration are changed to give
these cases, an effective susceptibility should be found.
For example, the M ( H ) curve of an ideal type-I1 super-
conductor in fields below the lower critical field H,, is
+e = p F ( r , 4) dr d4 - j:' F(r, 4) dr d4, (-43)

linear, with x = - 1. In the mixed state, M ( H ) increases


when H > H,, and the effective x should be close to the
differential susceptibility at each point, which is positive.
+s = sp G(z, 4 ) dz d4 - G(z, 4) dz d4, (A4)

where F ( r , 4) and G(z, 4 ) represent the integrands in (Al)


This causes a discontinuity in the value of N , above H,, ,
and (A2). The definite integral S F ( r , 4 ) dr d4 has been
and a proper demagnetizing correction should take ac-
evaluated by Gray [83]. Using a similar method, we eval-
count of this effect. Similar caveats apply to the inter-
uate the other integrals with the results,
mediate state of superconductors.
Normal conducting cylinders in ac fields have M ( H )
loops, and a complex susceptibility with a negative real
part can be defined [81]. However, this susceptibility is
F(r, 4) dr d4 = som
( u r 2 / 2 ~ 0 ) exp (-A I zj - zi I)
due to eddy currents constrained by the skin effect, dif- h-'J1(hr2)Jo(hri) d h , ('45)
ferent from our model assumption of uniform susceptibil-
ity. Nf,,for x = - 1 may be used in the limit of small
skin depth. Otherwise, to obtain good results, y must be
large enough so that only a small correction is needed;
our Nf,,values for an effective x < 0 can be used. A
similar case arises in hard superconductors, where a por- (A61
tion of the magnetization comes from penetrated super- where Jo and J I are Bessel functions and 2a is the diam-
currents that follow the critical-state model [82]. eter of the cylinder. Similar expressions where the limits
Since most cases of magnetic measurements involve of integration are r l and z1 are obtained from the previous
nonlinear magnetization curves, the demagnetizing cor- expressions by replacing r2 with rl and z2 with z I .
rection using the factors calculated for constant x should The interaction constants NY and N:' are defined for
be made cautiously. For this, a deep understanding of the rings on the ends as
magnetization process and the ,demagnetizing effect is
most important. N : = (Lco/u) a@.,/% = fS, - f4,

APPENDIX
The demagnetizing field produced by a ring-shaped dis-
= rz,/<2lz,l>l[ r2 Iom exp (-xlz,I)Jl(hr2)JO(X~,) d h

tribution of magnetic poles is calculated below. Results


are presented both for the case where a ring has a finite
width, as on the end-plane of a cylinder, or a finite height,
- rl I exp ~ - ~ l z , l ~ J l ~ ~ d ~h ]l , ~ J ('47)
o ~ ~ ~ , ~

as on the side surface of a cylinder. The method follows N : = ( P o / 4 a+.,/ar, = f S r - f ; r


that of Gray [83], who calculated the field produced by a
disk of charge. The pole density is taken to be uniform
over the width or height of the ring.
The magnetic scalar potential at a point (rir z,) pro-
= -;[r2 som exp (-A lz, I)JlW2)Jl O r , ) dX

duced by a pole density U on the j t h ring is given by - rl iom exp ( - ~ l z , l ) J l ( ~ ~ l ) J ld( h~ ]~, l ~(A81

9, = (U/4Tpo) 1; In rj[(Zj - Zj)2 + rj + r: and, for rings on the side surface, as


N:' = (PO/U> a+,/az, = fiz- f;, ('49)
4]-'12drj d4,

+$
- 2rjrj cos

= (ua/4?rp0) S": Sz* [(zj - zi)2 + a2 + r:


(All
= - ;[a iom exp ( - h ~ ~ ~ ( ) J ~ ( X a ) Jd~h( h r , )

- 2 ~ i aCOS dz, d4, (A3


-a 1, exp ~ - ~ I
1
~d h ,~ o
(A101 ~ ~ ~ ~
I

CHEN et al.: DEMAGNETIZING FACTORS FOR CYLINDERS 3617

N: = (po/u) a+.,/&-; =fir -fir ACKNOWLEDGMENT

= [z2/(21z201 a iom
exp (-hlz2l)J,(Xa)J,(Xr;) dX
We thank A. S . Arrott, B.-Z. Li, and S . Shtrikman for
helpful comments, A. B. Kos and J. NoguCs for assis-
tance in computer programming, and T. W. Petersen for
help in preparing the figures.
- [z1/(21z101 a ’jomexp(-Xlzl()J,(Xa)Jl(Xr;) dX.
REFERENCES
(AI 1)
[I] G. Chrystal, “Magnetism,” in Encyclopaedia Brirannica, 9th ed.,
We have arbitrarily set zi = 0 for ease of notation and vol. 15, 1883, pp. 219-276.
have included the factor z,/ 1 zj I when needed to account [2] W. Thomson (Lord Kelvin), Reprinf of Papers on Elecfrosfaficsand
for the sign reversal that occurs when zj < zi. The inte- Magnerism. London: Macmillan, 1872, pp. 470-471.
[3] F. J. Evans and A. Smith, “On the magnetic character of the armour-
grals in theffactors are given in [84] and reduce to plated ships of the Royal Navy, and on the effect on the compass of
particular arrangements of iron in a ship,” Phil. Trans. Roy. Soc.
London, vol. 155, pp. 263-323, 1865.
[4] J. Clerk Maxwell, A Treafise on Elecfricifyand Magnefism, 3rd ed.,
vol. 2. Oxford: Clarendon, 1892, pp. 66-73. First published 1873.
Reprinted New York: Dover, 1954.
[5] H. A. Rowland, “On magnetic permeability, and the maximum of
magnetism of iron, steel, and nickel,” Phil. M a g . , ser. 4 , vol. 46,
pp. 140-159, Aug. 1873.
[6] H. A. Rowland, “On the magnetic permeability and maximum of
magnetism of nickel and cobalt,’’ Phil. Mag., ser. 4, vol. 48, pp.
321-340, NOV. 1874.
[7] C. R. Mann, “Demagnetization factors for cylinders,” Phys. Rev.,
vol. 3, pp. 359-369, Mar.-Apr. 1896.
[8] J . A. Ewing, “Experimental researches in magnetism,” Phil. Trans.
Roy. Soc. London, vol. 176, pp. 523-640, 1885.
[9] Lord Rayleigh (J. W. Strutt), “1. Notes on magnetism.-On the en-
ergy of magnetized iron,” Phil. M a g . , ser. 5 , vol. 22, pp. 175-183,
Aug. 1886.
[IO] D. Foster, “An experimental method for the determination of the bal-
listic demagnetization factor,” Phil. Mag., ser. 7 , vol. 8, pp. 304-
313, Sept. 1929.
[ 1 I ] A. Tanakadate, “Mean intensity of magnetization of soft iron bars of
various lengths in a uniform magnetic field,” Phil. M a g . , ser. 5 , vol.
26, pp. 450-456, Nov. 1888.
[I21 H. du Bois, The Magnetic Circuit in Theory and Practice, E. Atkin-
son, translator. London: Longmans, Green, 1896, pp. 23-43.
where a2 = r2 and b2 = ri, and [13] C. Benedicks, “Ueber die Entmagnetisirungsfactoren kreiscylin-
drischer Stabe (Demagnetizing factors of cylindrical rods),” Ann.
Physik, vol. 6 , pp. 726-740, Dec. 1901.
[I41 C. L. B. Shuddemagen, “The demagnetizing factors for cylindrical
iron rods,” Proc. Amer. Acud. Arts and Sei., vol. 43, pp. 185-256,
Sept. 1907.
1151 C. L. B. Shuddemagen, “Tables of demagnetizing factors for iron
rods,” Phys. Rev., vol. 31, pp. 165-169, Aug. 1910.
(161 F. W. Warburton, “The magnetic pole, a useless concept,” Am. Phys.
Teacher (Am. J . Phys.), vol. 2, pp. 1-6, Feb. 1934.
[I71 P. F. W. Preece. “Demagnetizing factors,” School Sei. Rev.. vol.
+ b2(2na2)Y1Mcy2,P2))z2/lz2l 52, pp. 309-315, Dec. 1970.
(a2 < b2), [I81 G. Green, An Essay on the Application of Mathematical Analysis to
(A15b) rhe Theories of Electriciry and Mugnerism. Nottingham, U.K.:
Wheelhouse, 1828, pp. 66-72. Reprinted Goteborg, Sweden:
where a2 = ri and b2 is the radius of the cylinder. The Wezata-Melins Aktiebolag, 1958.
[I91 S. P. Thompson and E. W. Moss, “On the self-demagnetizing factor
factors f;?,f&,f&, and f i r are given by the same expres- of bar magnets,” Proc. Phys. Soc. London, vol. 21. pp. 622-633,
sions except r2 is replaced by rl and z2 is replaced by zI . Dec. 1909.
For each case k = 4ab/[(a + b)2 + z2], and F ( k ) and [20] J . Wiirschmidt, “Magnetische Anfangspermeabilitat, scheinbare Re-
manenz und Verhalten bei Erschiitterungen (Magnetic initial perme-
E ( k ) are the complete elliptic integrals of the first and sec- ability, apparent hysteresis, and behavior under vibration),” Z. Phy-
ond kind. A ( a , 6 ) is related to the Heuman lambda func- sik, vol. 12, pp. 128-164, 1923.
tion and is expressed in terms of the elliptic integrals of [2 I ] J . Wiirschmidt, Theorie des Enrmagnerisierungsfakrors und der
the third kind, A(cy, P) = (1 - p)’I2(1 - k 2 / p ) ’ k I ( a , Scherung von MagnefisierunRskurven. Braunschweig: Sammlung
Vieweg, 1925.
p , 7r/2). The parameters p , cy, and 6 are specified by p [22] H. Neumann and K. Warmuth, “Uber die rechnerische Auswertung
= k 2 / [ 1 - (1 - k 2 ) sin2P], k = sin a, and sin2P = ballisticher Entmagnetisierungsfaktoren kreiszylindrischer Stabe
z2/[(a - b)2 + z 2 ] . The elliptic integrals are evaluated (Mathematical evaluation of the ballistic demagnetization factor of
circular cylindrical rods),” Wiss. VerOf. Siemens-Konzern, vol. 11,
numerically using the procedure given in [MI. pp. 25-35, Jul. 1932.
3618 IEEE TRANSACTIONS ON MAGNETICS, VOL. 21, NO. 4, JULY 1991

1231 F . Stablein and H. Schlechtweg, “Uher den Entmagnetisierungsfak- [51] W. E. Archer and E. Guancial, “Magnetization distributions in bulk
tor zylindrischer Stabe (Demagnetizing factor for cylindrical rods),” magnetic material,” IEEE Trans. Magn., vol. MAG-9, pp. 51-56,
2. Physik, vol. 95, pp. 630-646, Jul. 1935. Mar. 1973.
1241 K. Warmuth, “Die Bestimmung des ballistischen Entmagnetisi- 1521 T. H. Fawzi, K. F. Ah, and P. E. Burke, “Boundary integral equa-
erungsfaktors mit dem magnetischen Spannungsmesser an Staben von tions analysis of induction devices with rotational symmetry,” IEEE
quadratischem Querschnitt (Determination of ballistic demagnetiza- Trans. Magn., vol. MAG-19, pp. 36-44, Jan. 1983.
tion factor for specimens of square section),” Archiv Elektrotechnik, [53] T. Okoshi, “Demagnetizing factors of rods and tubes computed from
vol. 30, pp. 761-779, Dec. 1936. analog measurements,’’ J. Appl. Phys., vol. 36, pp. 2382-2387, Aug.
(251 K. Warmuth, “Zur Darstellung des ballistischen Entmagnetisierungs- 1965.
faktors zylindrischer Stabe (Ballistic demagnetization factor for cy- [54] Y. Yamamoto and H. Yamada, “New analytical expressions for flux
lindrical rods),” Archiv Elektrotechnik, vol. 31, pp. 124-130, Feb. distribution and demagnetizing factor of cylindrical core,” Electr.
1937. Eng. Jpn., vol. 102, pp. 1-8, May-Jun. 1982 [Denki Gakkai Ron-
(261 K. Warmuth, “Uber den ballistischen Entmagnetisierungsfaktor zyl- bunshi, vol. 102A. pp. 255-262, May 19821.
indrischer Stabe (Ballistic demagnetization factor of cylindrical [55] R. I, Joseph and E. Schlomann, “Demagnetizing field in nonellip-
rods),” Archiv Elektrotechnik, vol. 33, pp. 747-763, Dec. 1939. soidal bodies,” J. Appl. Phys., vol. 36, pp. 1579-1593, May 1965.
(271 R. M. Bozorth and D. M. Chapin, “Demagnetizing factors of rods,” [56] L. Kraus, “The demagnetization tensor of a cylinder,” Czech. J.
J. Appl. Phys., vol. 13, pp. 320-326, May 1942. Phys. B , vol. 23, pp. 512-519, May 1973.
(281 R. M. Bozorth, Ferromagnetism. Princeton: Van Nostrand, 1951, [57] J . A. Brug and W . P. Wolf, “Demagnetizing fields in magnetic mea-
pp. 845-849. surements. I. Thin discs,” J. Appl. Phys., vol. 57, pp. 4685-4694,
[29] W. F. Brown, Jr., “Single-domain particles: New uses of old theo- May 1985.
rems,” Am. J. Phys., vol. 28, pp. 542-551, Sept. 1960. [58] H. Zijlstra, Experimental Methods in Magnetism, vol. 2 . Amster-
[30] F. W. Grover, Inductance Calculations. New York: Van Nostrand, dam: North-Holland, 1967, pp. 69-72.
’ 1946, pp, 142-162. [59] J. A. Osborn, “Demagnetizing factors of the general ellipsoid,” Phys.
[31] W. F. Brown, Jr., Magnetostatic Principles in Ferromagnetism. Rev., vol. 67, pp. 351-357, Jun. 1945.
Amsterdam: North-Holland, 1962, pp. 187-192. [60] E. C. Stoner, “The demagnetizing factors for ellipsoids,” Phil. Mag.,
[32] G. W. Crabtree, “Demagnetizing fields in the de Haas-van Alphen ser. 7, vol. 36, pp. 803-821, Dec. 1945.
effect,” Phys. Rev. E , vol. 16, pp. 1117-1125, Aug. 1977. [61] E. B. Rosa and F. W. Grover, “Formulas and tables for the calcu-
1331 R. Moskowitz, E. Della Torre, and R. M. M. Chen, “Tabulation of lation of mutual and self-inductance,” Bull. Bureau Standards, vol.
magnetometric demagnetization factors for regular polygonal cylin- 8, pp. 1-237, Jan. 1912, equation (73).
ders,” Proc. IEEE, vol. 54, p. 1211, Sept. 1966. 1621 E. B. Rosa and F. W. Grover, “Formulas and tables for the calcu-
[34] J. KaczCr and Z. Klem, “The magnetostatic energy of coaxial cyl- lation of mutual and self-inductance,” Bull. Bureau Standards, vol.
inders and coils,” Phys. Stat. Sol. A , vol. 35, pp. 235-242, May 8, pp. 1-237, Jan. 1912, equation (54).
1976. 1631 L. Cohen, “An exact formula for the mutual inductance of coaxial
[35] R. I. Joseph, “Ballistic demagnetizing factor in uniformly magnet- solenoids,” Bull. Bureau Standards, vol. 3, pp. 295-303, May 1907.
ized cylinders,” J . Appl. Phys., vol. 37, pp. 4639-4643, Dec. 1966. [64] D.-X. Chen, Ballistic and Bridge Methods of Magnetic Measure-
(361 P. Vallabh Sharma, “Rapid computation of magnetic anomalies and menls of Materials. Beijing: China Metrology, 1990, pp. 77-85.
demagnetization effects caused by bodies of arbitrary shape,” Pure [65] A. E. Ruehli and D. M. Ellis, “Numerical calculation of magnetic
Appl. Geophys., vol. 64, no. 2, pp. 89-109, 1966. fields in the vicinity of a magnetic body,” IBMJ. Res. Develop., vol.
[37] M. Sat0 and Y. Ishii, “Simple and approximate expressions of de- 15, pp. 478-482, NOV. 1971.
magnetizing factors of uniformly magnetized rectangular rod and cyl- [66] N. Normann and H. H. Mende, “Numerical calculations of the field-
inder,” J. Appl. Phys., vol. 66, pp. 983-985, Jul. 1989. dependent magnetization of short rectangular prisms,” Appl. Phys.,
[38] D.-X. Chen and B.-Z. Li, “On the error of measurement of feebly vol. 13, pp. 15-19, May 1977.
magnetic material in regard to demagnetizing field,” Acta Metall. [67] C. J. Hegedus, G. Kadar, and E. Della Torre, “Demagnetization ma-
Sinica, vol. 19, pp. 217-224, Oct. 1983. trices for cylindrical bodies,” J . Inst. Math. Its Appl., vol. 24, pp.
[39] D.-X. Chen, Physical Basis of Magnetic Measurements. Beijing: 279-291, NOV.1979.
China Mechanical Industry, 1985, p. 242. 1681 G. Kadar, C. J. Hegedus, and E. Della Torre, “Cylindrical demag-
[40] T. T. Taylor, “Electric polarizability of a short right circular con- netization matrix,” IEEE Trans. Magn., vol. MAG-14, pp. 276-277,
ducting cylinder,” J. Res. Nut. Bur. Standards, vol. 64B, pp. 135- Jul. 1978.
143, Jul.-Sept. 1960. [69] M. Soinski, “Demagnetization effect of rectangular and ring-shaped
[41] T. T. Taylor, “Magnetic polarizability of a short right circular con- samples made of electrical sheets placed in a stationary magnetic
ducting cylinder,” J. Res. Nut. Bur. Standards, vol. 64B, pp. 199- field,” IEEE Trans. Instrum. Meas., vol. 39, pp. 704-710, Oct. 1990.
210, Oct.-Dec. 1960. 1701 T. L. Templeton, A. S . Arrott, and A. Aharoni, “Partially saturated
[42] W. R. Smythe, “Charged right circular cylinder,” J . Appl. Phys., ferromagnetic cylinders,” J. Appl. Phys., vol. 55, pp. 2189-2191,
vol. 27, pp. 917-920, Aug. 1956. Mar. 1984.
[43] W. R. Smythe, “Charged right circular cylinder,” J. Appl. Phys., [71] W. H. Press, B. P. Flannery, S. A. Teukolsky, and W. T. Vetterling,
vol. 33, pp. 2966-2967, Oct. 1962. Numerical Recipes. Cambridge, U.K. : Cambridge University Press,
[44] W. R. Smythe, Static and Dynamic Electricity, 3rd ed. New York: 1986, pp. 31-38.
McGraw-Hill, 1968, pp. 209-21 1. [72] W. F. Brown, Jr. and A. H. Morrish, “Effect of a cavity on a single-
(451 T. L. Templeton and A. S. Arrott, “Magnetostatics of rods and bars domain magnetic particle,” Phys. R e v . , vol. 105, pp. 1198-1201,
of ideally soft ferromagnetic materials,” IEEE Trans. Magn., vol. Feb. 1957.
MAG-23, pp. 2650-2652, Sept. 1987. [73] W. F. Brown, Jr., Magnetostatic Principles in Ferromagnetism.
1461 A. S. Arrott, B. Heinrich, and T. L. Templeton, “Phenomenology Amsterdam: North-Holland, 1962, pp. 49-53.
of ferromagnetism: I. Effects of magnetostatics on susceptibility,” [74] R. Moskowitz and E. Della Torre, “Theoretical aspects of demag-
IEEE Trans. Magn., vol. 25, pp. 4364-4373, Nov. 1989. netization tensors,” IEEE Trans. Magn., vol. MAG-2, pp. 739-744,
1471 A. S . Arrott, B. Heinrich, T. L. Templeton, and A. Aharoni, “Mi- Dec. 1966.
cromagnetics of curling configurations in magnetically soft cylin- [75] P. Rhodes and G. Rowlands, “Demagnetising energies of uniformly
ders,” J. Appl. Phys., vol. 50, pp. 2387-2389, Mar. 1979. magnetised rectangular blocks,” Proc. Leeds Phil. Lit. Soc., vol. 6 ,
[48] A. S. Arrott, B. Heinrich, and A. Aharoni, “Point singularities and pp. 191-210, Dec. 1954.
magnetization reversal in ideally soft ferromagnetic cylinders,” IEEE 1761 E. W. Lee and J. K. Ackers, “A note on the demagnetizing energy
Trans. Magn., vol. MAG-15, pp. 1228-1235, Sept. 1979. of a uniformly magnetized cube,” Brit. J . Appl. Phys., vol. 14, pp.
1491 A. Aharoni, “Magnetostatic energy of a saturating cylinder,” J. Appl. 46-47, Jan. 1963.
Phys., vol. 52, pp. 6840-6843, NOV. 1981. [77] T. Iwata, “A diagonal sum rule concerning demagnetization tensors
[50] A. Aharoni, “Magnetostatic energy of a ferromagnetic cylinder,” J. in composite bodies,” J. Appl. Phys., vol. 39, pp. 3094-3097, Jun.
Appl. Phys., vol. 54, pp. 488-492, Feb. 1983. 1986.
I I1 I

CHEN et al.: DEMAGNETIZING FACTORS FOR CYLINDERS 3619

[78] D. S. Bloomberg and A. S. Arrott, “Micromagnetics and magneto- surements. He was a visiting scientist at the Royal Institute of Technology,
statics of an iron single crystal whisker,” Can. J . Phys., vol. 53, pp. Stockholm, from 1985 to 1989, and at the National Institute of Standards
1454-1471, Aug. 1975. and Technology, Boulder, Colorado, during 1988 and 1990. He has pub-
[79] American Society for Testing and Materials, “Standard test methods lished two books entitled Physical Basis of Magnetic Measurements (China
for permeability of feebly magnetic materials,” A342-84, Annual Mechanical Industry, Beijing, 1985) and Ballistic and Bridge Methods of
Book of ASTM Standards, vol. 03.04. Philadelphia: ASTM, 1990, Magnetic Measurements of Materials (China Metrology, Beijing, 1990).
pp. 30-36.
[80] R. B. Goldfarb and J. V. Minervini, “Calibration of ac susceptometer
for cylindrical specimens,” Rev. Sci. Instrum., vol. 5 5 , pp. 761-764,
May 1984. James A. Brug (M’87) was born in Cody, WY, in 1954. He received the
[81] L. D. Landau, E. M. Lifshitz, and L. P. Pitaevskii, Electrodynamics B.S. degree in physics from Montana State University and the Ph.D. de-
of Continuous Media, 2nd ed. Oxford, U.K.: Pergamon, 1984, pp. gree in applied physics from Yale University.
205-207. He is currently on leave from Hewlett-Packard Laboratories in Palo Alto,
[82] C. P. Bean, “Magnetization of high-field superconductors,” Rev. CA, as an ASEE postdoctoral fellow at the National Institute of Standards
Mod. Phys., vol. 36, pp. 31-39, Jan. 1964. and Technology in Boulder, Colorado. His research has focused on noise
[83] A. Gray, “Notes on electric and magnetic field constants and their mechanisms in magnetic thin films, measurement of magnetic properties at
expression in terms of Bessel functions and elliptic integrals,” Phil. small length scales, and the magnetics of magnetoresistive recording heads.
Mag., vol. 38, PP. 201-214, Aug. 1919. Dr. Brug is a member of the American Physical Society.
[84] G. Eason, B. Noble, and I. N. Sneddon, “On certain integrals of
Lipschitz-Hankel type involving products of Bessel functions,” Phil.
Trans. Roy. Soc. London, vol. A247, pp. 529-551, Apr. 1955.
[85] W. H. Press, B. P. Flannery, S. A. Teukolsky, and W. T . Yetterling,
Numerical Recipes. Cambridge, U.K.: Cambridge University Press, Ronald B. Goldfarb (M’79-SM’86) was born in Mexico City, Mexico,
1986, pp. 183-190. II(*/2, n , k) corresponds to our ~ ( L Y p, . ~ / 2 ) in 1951. He received the B.A. degree in electrical engineering and the
with k = sin 01 and n = - p . M.S. degree in materials science from Rice University and the M.S. and
Ph.D. degrees in physics from Colorado State University.
He is a member of the Superconductor and Magnetic Measurements
Group at the National Institute of Standards and Technology (NIST) in
Du-Xing Chen was born in Shanghai, China, in 1941. He studied physics Boulder, CO. His areas of research are ac losses and coupling effects in
at Beijing University from 1958 to 1964, received the M.S. degree from low- and high-temperature superconductors, cryogenic magnetic properties
the Institute of Physics, Academia Sinica, in 1981, and received the Ph.D. of concentrated spin glasses and weakly magnetic alloys, and magnetic
degree from the Central Iron and Steel Research Institute, Beijing, in 1985. instrument development. From 1979 to 1981 he was a National Research
He is a visiting Professor in the Electromagnetism Group at the Uni- Council postdoctoral research associate at NIST.
versitat Autbnoma de Barcelona, Spain. His areas of research are magnetic Dr. Goldfarb is a member of the American Physical Society and the
measurements, magnetic materials, and superconductor theory and mea- American Society for Testing and Materials.

You might also like