Proc. CERN Accelerator School on Measurement and Alignment of Accelerator and
Detector Magnets, April 11-17, 1997, Anacapri, Italy, CERN-98-05, pp.1-26
BASIC THEORY OF MAGNETS
Animesh K. Jain
RHIC Project, Brookhaven National Laboratory, Upton, New York 11973-5000, USA
Abstract
The description of two dimensional magnetic fields of magnets used in accelerators is discussed in terms of a harmonic expansion. The expansion is derived for cylindrical components and extended to
Cartesian components. The Cartesian components are also described in terms of a complex field. The rules for transformation of the expansion coefficients under various types of coordinate transformation are given. The relationship between a given current distribution and the resulting field harmonics is explored in terms of the vector and complex potentials. Explicit results are presented for some simple geometries. Finally, the harmonics allowed under various symmetries in the magnet current are discussed.
1. MULTIPOLE EXPANSION OF A TWO-DIMENSIONAL FIELD
For most practical purposes related to magnetic measurements in an accelerator magnet, one is interested in the magnetic field in the aperture of the magnet, which is in vacuum and carries no current. Also, most accelerator magnets tend to be long compared to their aperture.
Thus, a two dimensional description is valid for most of the magnet, except at the ends. We shall at first confine ourselves to a description of a purely two dimensional field in a current free region. The relationship between the field and the currents will be treated later in Sec. 5 and onwards.
In free space, with no true currents, the curl of the magnetic field, H, is zero. Also, the magnetic induction, B, is given by m0H, where m0 = 4p ´10-7 Henry/m is the permeability of free space. Consequently, the magnetic induction, B, can be expressed as the gradient of a magnetic scalar potential, Fm:
Ñ ´ H = 0Þ Ñ ´ B = 0Þ B = -ÑFm (1)
The magnetic induction, B = m0H also has zero divergence. Combining this fact with Eq. (1), we get Laplace’s equation for the scalar potential:
Ñ2Fm = 0 (2)
We choose a cylindrical coordinate system with the Z-axis along the length of the magnet and the origin located at the center of the magnet aperture. For a two dimensional field having no axial component, the scalar potential has no z-dependence. Laplace’s equation can be solved by separation of variables and imposing the boundary conditions that Fm be periodic in the angular coordinate, q, and be finite at r = 0. The resulting general solution is a series expansion of the scalar potential. The radial and the azimuthal components of B are then
2
obtained by taking the gradient of the scalar potential. The components of B in cylindrical coordinates can be written in the form :
B r r C n r R r n m n ref n ( ,q) ( ) sin[ ( n )]
¶
¶
= - q a æ è ç ö ø ÷
=
æ è çç ö ø ÷÷
-
=
¥ - å F
1
1
(3)
B r r C n r R m n n ref n q q n
¶
¶ q q a ( , ) ( ) cos[ ( )] = -æè ç öø ÷ æ è ç ö ø ÷
=
æ è çç ö ø ÷÷
-
=
¥ - å 1
1
1
F
(4) where C(n) and an are constants and Rref is an arbitrary reference radius, typically chosen to be 50-70% of the magnet aperture. For a given r, the n-th term in Br(r,q ) has n maxima and n minima as a function of the azimuthal angle q. These angular positions may be regarded as the locations of magnetic poles. Thus, the n-th terms in Eqs. (3) and (4) correspond to a
2n-pole field. Accordingly, C(n) is said to be the amplitude of the 2n-pole component of the total field. The locations of the south and the north poles in the 2n-pole field are: q p a p a p
= + + +a
2
5
2
9 n n n n 2n n ; ; ; .... SOUTH POLES (5) q p a p a p
= + + + a
3
2
7
2
11
n n n n 2n n ; ; ; ....NORTH POLES (6)
The parameter an defines the orientation of the 2n-pole component of the field with respect to the chosen X-axis and is called the phase angle. Eqs. (3) and (4) represent the multipole expansion of the components of a two dimensional field in a current free region.
1.1 The Cartesian Components
Accelerator physicists often prefer to work with the Cartesian components of the field.
The relationship between the Cartesian and the cylindrical components is shown in Fig. 1.
B
Br
B
By
Bx
X-axis
Y-axis
q r q q q
Fig. 1 Cylindrical and Cartesian components of the magnetic induction vector, B
3
Using the expansions for Br and Bq from Eqs. (3) and (4), we get the Cartesian components: B r B B C n r R x r n n n ref n ( ,q ) = cosq - q sinq = ( ) sin[( )q a n ] æ è çç ö ø ÷÷
- -
=
¥ - å 1
1
1 (7)
B r B B C n r R y r n n n ref n ( ,q) = sinq + q cosq = ( ) cos[( )q a n ] æ è çç ö ø ÷÷
- -
=
¥ - å 1
1
1 (8)
1.2 The Complex Field
The components of a two dimensional field can be very conveniently described in terms of a complex field, B(z), defined as a function of the complex variable z = x + iy = r • exp(iq )
[a bold and italic font is used for complex quantities in this paper]. Looking at the expansions in Eqs. (7) and (8), one could define the complex field as:
B z [ ] z ( ) = ( , ) + ( , ) = ( )exp(- ) æ è çç ö ø ÷÷
=
¥ -
B x y iB x y å C n in y x n R n ref n a
1
1
(9)
The divergence and the curl (in a source free region) of the vector B are zero. In two dimensions, we get:
Ñ = Þ æ è ç ö ø ÷
+
æ è ç ö ø ÷
=
æ è ç ö ø ÷
= - æ è ç ö ø ÷
.B 0 0,
¶
¶
¶
¶
¶
¶
¶
¶
B
x
B
y
B
y
B
x x y or y x (10)
(Ñ ´ ) = Þ , æ è ç ö ø ÷
-
æ è ç ö ø ÷= æ è ç ö ø ÷
=
æ è ç ö ø ÷
B z y x y x B x B y B x B y 0 0
¶
¶
¶
¶
¶
¶
¶
¶ or (11)
Eqs. (10) and (11) are nothing but the well known Cauchy-Riemann conditions for the complex field B(z) to be an analytic function of the complex variable z. The complex field is sometimes defined as Bx(x, y) – iBy(x, y), which is also an analytic function of z. It should be noted that the quantity Bx + iBy is not an analytic function of z. The analytic property of the complex field has been exploited in solving many two dimensional problems [1-6].
2. TWO-DIMENSIONAL BEHAVIOUR OF THE INTEGRAL FIELD
In magnets of a finite length, the two dimensional representation of the field is valid only in the body of the magnet, sufficiently away from the ends. In regions near the ends of the magnet, the field is three dimensional and the usual multipole expansion is no longer valid. Some examples of a three dimensional treatment may be found in Refs. [7]-[9].
However, for most practical purposes, one is interested in the integral of the field (or of its derivatives) over the length of the magnet. This is because most magnets are short compared to the wavelength of the betatron oscillations in the machine and the details of axial variation are of little consequence. Also, a typical rotating coil of a finite length only measures the
4
integral of the field over its length. It can be shown that the integral field behaves as a two dimensional field provided the integration is carried out over an appropriate region [10].
In general, for three dimensional fields in a current free region, the scalar potential satisfies the Laplace’s equation:
Ñ = + + é ë êê ù û úú
2 =
2
2
2
2
2
2 F 0
F F F m x y z m m m x y z
( , , )
¶
¶
¶
¶
¶
¶
(12)
Integrating along the Z-axis from Z1 to Z2, we get: ò ¶
¶
¶
¶
¶
¶
¶
¶
¶
¶
¶
¶
2
2
2
2
2
2
2
2
2
2
2
2
1
2
1
2
1
2
0
F F F
F
m m m F
Z
Z m Z
Z
m
Z
Z x y z dz x y x y z dz z + + dz é ë êê ù û úú ó õ ôô
= + é ë êê ù û úú
+
ó õ ô
( , , ) = (13)
We define the z-integrated scalar potential as Fm òFm
Z
Z
(x, y) = (x, y, z)dz
1
2
. From Eq. (13):
¶
¶
¶
¶
¶
¶
2
2
2
2 2 1
1
F F F 2 m m m
Z
Z z z x y z
+ B x y Z B x y Z é ë êê ù û úú
= - é ë ê ù û ú
= ( , , ) - ( , , ) (14)
If the region of integration is so chosen that the Z-component of the field is zero at the boundaries of this region, then the right hand side vanishes and the z-integrated scalar potential satisfies the two dimensional Laplace’s equation. For example, the points Z1 and Z2 could both be chosen well outside the magnet on opposite ends to include the integral of the field over the entire magnet. This situation is commonly encountered in the measurement of the integral field of relatively short magnets with a long integral coil. Alternatively, one could choose Z1 well outside the magnet and Z2 well inside the magnet, where the field is again two dimensional. Such a situation would apply to the measurement of the end region in a long magnet with a short measuring coil. The measuring coil for this purpose must have sufficient length so that the condition in Eq. (14) can be satisfied.
NORMAL AND SKEW COMPONENTS
The multipole expansion of the complex field is given by Eq. (9). The normal and skew components are defined as the real and the imaginary parts of the expansion coefficients:
C(n)exp(–inan) = (2n-pole NORMAL Term) + i(2n-pole SKEW Term) (15)
Unfortunately, the index n in the expansion coefficient is not the same as the corresponding power (n–1) of z in Eq. (9). This has led to two different conventions being followed in denoting the normal and the skew terms. The “US Convention” denotes the 2n-pole normal and skew terms with an index of (n–1), to match the corresponding power of z in Eq. (9):
2 - pole Normal Term =
2 - pole Skew Term = n C n n B n C n n A n n n n
( )cos( )
( )sin( ) a a
º
- º
-
-
1
1
“U.S. CONVENTION” (16)
5
On the other hand, the “European Convention” denotes the 2n-pole normal and skew components with an index of n to retain the simple relationship between the index and the number of poles:
2 - pole Normal Term =
2 - pole Skew Term = n C n n B n C n n A n n n n
( )cos( )
( )sin( ) a a
º
- º
“EUROPEAN CONVENTION” (17)
Even though the two conventions are commonly referred to as the “US” and the
“European” notations, their use is not necessarily restricted to the respective geographic regions. This calls for exercising caution in interpreting the notations. One possible indicator of the notation is the presence or the absence of B0 and A0 terms.
In terms of the normal and the skew components, the expansion of B(z) is:
B z [ ] z ( ) = + = + æ è çç ö ø ÷÷
=
¥å
B iB B iA y x n n R n ref n 0
“U.S. CONVENTION” (18)
B z [ ] z ( ) = + = + æ è çç ö ø ÷÷
=
¥ -
B iB å B iA y x n n R n ref n 1
1
“EUROPEAN CONVENTION” (19)
The corresponding equations for the radial and the azimuthal components of the field are: B r [ ] r R r B n A n n ref n ( ,q) = n sin{( )q} n cos{( )q} æ è çç ö ø ÷÷
+ + +
=
¥å
0
1 1
“U.S. Convention” (20)
B r [ ] r R
B n A n n ref n q ( ,q) = n cos{( )q} n sin{( )q} æ è çç ö ø ÷÷
+ - +
=
¥å
0
1 1
“U.S. Convention” (21)
B r [ ] r R r B n A n n ref n ( ,q ) = n sin( q ) n cos( q ) æ è çç ö ø ÷÷
+
=
¥ - å 1
1
“European Convention” (22)
B r [ ] r R
B n A n n ref n q ( ,q) = n cos( q) n sin( q) æ è çç ö ø ÷÷
-
=
¥ - å 1
1
“European Convention” (23)
It should be noted that sometimes the skew component is defined with a sign opposite to that used here [11]. In view of the different notations in existence, it is important to recognize the convention being used in any particular work.
It is possible to assign a physical significance to the normal and the skew components in terms of the derivatives of the field. Using Eq. (18) or Eq. (19), it is easy to show