Download as pdf or txt
Download as pdf or txt
You are on page 1of 6

The 8

th
INTERNATIONAL SYMPOSIUM ON ADVANCED TOPICS IN ELECTRICAL ENGINEERING
May 23-25, 2013
Bucharest, Romania
Experimental characterization of electromagnets for
particle accelerators
Daniel DAN
1,2
, Daniel IOAN
1
1
University POLITEHNICA of Bucharest, Splaiul Independentei nr. 313, 060042, Bucharest, Romania
2
INCDIE ICPE - CA (ICPE - Advanced Researches), Splaiul Unirii, nr. 313, 030138, Bucharest, Romania
daniel.dan@icpe-ca.ro, daniel@lmn.pub.ro
AbstractThe paper analyses two measurement procedures
of magnetic eld in electromagnets for particle accelerators. In
these devices, it is of interest the magnetic eld distribution, in
the neighborhood of their axis. This distribution is not measured
directly, but by using the measurements of magnetic ux density,
in a nite number of points selected by users. Based on these
measurements, an analytical method is used to determine eld
variation in the circular aperture of electromagnet, and to
extract the coefcients of their multipole (harmonic) expansion.
Experimental characterizations, by using measurements along
radius and along aperture periphery are analyzed and compared.
It is proven that the circular measurements are more robust,
because the radial ones leads to a weak conditioned problem,
which is not well-posed, in the sense of Hadamard.
I. INTRODUCTION
A particle accelerator is a device that uses electromagnetic
elds to propel charged particles to high speeds and to
contain them in well-dened beams. Transverse acceleration is
described by the magnetic component of the Lorentz force, to
be precise, the vector product between velocity and magnetic
induction.
Electromagnetic multipoles are components of particle ac-
celerators. These multipoles can be dipole, quadrupole, sex-
tupole etc. [1].
In particle accelerators, ordinary dipoles are used to deect
particles; the role of quadrupoles is to focus the electron beam;
and sextupoles role is to correct the focus of electrons with
different energies.
Because the Lorentz force is orthogonal on velocity, the
electromagnets change the movement direction, but they did
not make change of the particle energy. In order to predict
their effect on particle beam, the study of the magnetic eld
in the aperture of these devices is very important.
The magnetic eld in the aperture of electromagnet, are
currently determined by four methods [2]: Single Stretched
Wire Method- SSW, Vibrating Wire Method - VW, Hall Probe
Method - HP and Rotating / Harmonic Coil Method - HC.
SSW method consists in moving a wire having high con-
ductivity inside aperture of magnet, which is subject to char-
acterization [2]. The voltage induced in the wire is measured.
VW method is based on the principle of mechanical oscil-
lations created by an alternating current running through the
wire. The method proposed by A. Temnykh has a sensitivity
to detect submicrometer magnetic axis of a magnet [3].
HP method is a classical method, used to determine the ux
density point by point in the electromagnets aperture, so that
it is possible to create a complete mapping of the magnetic
eld distribution [3].
HC method consists in the measurement of induced voltage
in a rectangular coil arranged radially or tangentially, placed
inside of a rotating cylinder in the aperture. Rapid angular
encoders and integrated voltage measurements systems spe-
cialized for data acquisition are used to determine eld in
hundreds of points by rotating measurement cylinder up to 10
revolutions per second.
For electromagnets of particle accelerators, the harmonic
coil technique (HC) is currently the most convenient, accurate
and widespread measurement technique for magnetic eld [4].
A cross-sectional schematic representation of such coils (the
radial version) is given in [5].
The current paper analyses two procedures of magnetic eld
measurement in electromagnets for particle accelerators. In
these devices, we are interested in the distribution of magnetic
eld, in the neighborhood of their axis. This distribution is not
measured directly, but by using measurements of magnetic ux
density, in a nite number of points selected by users. Based
on these measurements, analytical method is used to determine
eld variation in the circular aperture of electromagnet, and
to extract the coefcients of their multipole (harmonic) expan-
sion. Experimental characterizations, by using measurements
along radius and along aperture periphery are analyzed and
compared.
II. SEXTUPOL MAGNET AND SOLUTION OF LAPLACE
EQUATION
The subject of our study is a sextupole electromagnet. It is
considered in a normal position, if it has two poles along the
axis Oy and the Ox axis is located between poles (Fig. 1).
If two of the poles are along the right axis Ox and the axis
Oy lies between the other two poles (equidistant), then the
electromagnet is called skew sextupole.
Let consider the electromagnet aperture, represented by a
2D circular domain, and its border:
G =
{
(x, y) R
2

x
2
+y
2
< r
2
0
}
;
G = C =
{
(x, y) R
2

x
2
+y
2
= r
2
0
}
.
(1)
Laplaces equation satised by the magnetic scalar potential
inside aperture and its Dirichlet boundary condition in polar
978-1-4673-5980-1/13/$31.00 2013 IEEE
Fig. 1. Sextupole in a normal position [6]
coordinates are:

r
2

2
V
m
r
2
+r
V
m
r
+

2
V
m

2
= 0,
V
m
(r
0
, ) = f() =

n=0
(
n
cos (n) +
n
sin (n)) .
(2)
The Fourier coefcients of boundary condition are:

0
=
2

2
0
f () d;

n
=
1

2
0
f () cos (n) d;

n
=
1

2
0
f () sin (n) d.
(3)
By separation of variables [7], is obtained:
V
m
(r, ) =
0
+

n=0
r
n
(
n
cos (n) +
n
sin (n)) , (4)
and based on B = V
m
, are computed the radial and
azimuthal components of the magnetic eld:

B
r
(r, ) =

n=1
(
r
r
0
)
n1
(B
n
sin (n) +A
n
cos (n)) ,
B

(r, ) =

n=1
(
r
r
0
)
n1
(B
n
cos (n) A
n
sin (n)) ,
(5)
where
A
n
= n
n
, B
n
= n
n
and B = B =

B
2
r
+B
2

.
Let consider now an ideal sextupol in skew position having
as boundary condition: B
r
(r
0
, ) = B
0
sin(3) with B
0
=
0.11T.
In Fig. 2, 3, 4 are depicted spatial distribution of: B
n
(r, ),
B
t
(r, ), |B| (r, ) for this ideal device inside its aperture.
III. ANALYSIS OF EXPERIMENTAL DATA
A. Experimental data
The electromagnet depicted in Fig. 1 rotated in skew posi-
tion was experimentally characterized by HP method. Table I
presents the values of radial magnetic ux density measured,
Fig. 2. Radial component of the magnetic ux density
Fig. 3. Azimuthal component of the magnetic ux density
Fig. 4. The module of the magnetic ux density
along horisontal Ox axes for different excitation currents
between: I = 50A and I = 580A. The measured data are
plotted in Fig. 5, 6, 7.
Let consider the spatial variation of radial magnetic eld
for a reference current I
0
= 300A:
B
r
= [0.00, 0.0034, 0.0181, 0.0426, 0.0770, 0.1216];
x

= [0, 0.2000, 0.4000, 0.6000, 0.8000, 1.0000].


(6)
Where x

= r/r
0
with r
0
= 0.05m, and B
r
are the average
values for symmetric positive and negative coordinates. Since
TABLE I
EXPERIMENTAL DATA OF B
r
[T] [8]
x[m] I
1
= 50A I
2
= 100A I
3
= 200A I
4
= 300A I
5
= 400A I
6
= 500A I
7
= 580A
-0.05 0.020242 0.039452 0.078291 0.117118 0.155458 0.191329 0.213700
-0.04 0.012976 0.024993 0.049257 0.073475 0.097344 0.119578 0.133432
-0.03 0.007351 0.013833 0.026891 0.039899 0.052654 0.064424 0.071698
-0.02 0.003409 0.006005 0.011156 0.016300 0.021297 0.025759 0.028406
-0.01 0.001107 0.001415 0.002014 0.002574 0.003061 0.003299 0.003318
0.00 0.000457 0.000109 0.000648 0.001424 0.002211 0.003188 0.003892
0.01 0.001410 0.002012 0.003157 0.004308 0.005432 0.006240 0.006703
0.02 0.004000 0.007174 0.013530 0.019865 0.026131 0.031800 0.035338
0.03 0.008261 0.015618 0.030443 0.045250 0.059941 0.073532 0.082093
0.04 0.014156 0.027311 0.053920 0.080519 0.106886 0.131478 0.147069
0.05 0.021755 0.042434 0.084248 0.126067 0.167577 0.206448 0.230987
Fig. 5. Radial magnetic ux density vs. current
Fig. 6. Radial magnetic ux density vs. radius
the rst value is perturbed by an obvious error, it was replaced
with B
r
= 0, the correct value, due to the symmetry.
B. Polynomial approximation
The spatial variation of magnetic eld B
rexp
for the
reference current (6) was tted with a 10th degree, even
Fig. 7. Radial magnetic ux density vs. radius and current
polynomial function:
B
rmod
(x

) = c
2
x
2
+c
4
x
4
+c
6
x
6
+c
8
x
8
+c
10
x
10
. (7)
Using CFTOOL module of MATLAB, the coefcients pre-
sented in Table IV were obtained, for several polynomial
degrees.
A graphical approximation, of 10th degree polynomial ap-
proximation, compared with experimental data is given in Fig.
8. Relative standard deviation, estimated with:
=
max
k=1,11
(B
rexp,k
B
rmod,k
)
max
k=1,11
(B
rexp,k
)
, (8)
is 2.21%.
C. Linear model
Figure 5 suggests that the magnetic core is not strongly
saturated. If the magnetic eld has a linear dependence versus
excitation current, we can consider:
B
r
= B
ref
I
I
0
, (9)
where B
ref
is the eld corresponding to the reference current
I
0
. It will be computed, starting from the matrix in Table I,
with n = 11 rows (corresponding to n measuring positions)
TABLE II
THE REFERENCE FIELD COEFFICIENT AND ITS STANDARD DEVIATION FOR SEVERAL POSITION AND CURRENTS
B
ref
(i, j)[T]
x[m] I II III IV V VI VII

B
ref
std %
-0.05 0.1215 0.1184 0.1174 0.1171 0.1166 0.1148 0.1105 0.1166 2.8882
-0.04 0.0779 0.0750 0.0739 0.0735 0.0730 0.0717 0.0690 0.0734 2.3549
-0.03 0.0441 0.0415 0.0403 0.0399 0.0395 0.0387 0.0371 0.0402 1.9108
-0.02 0.0205 0.0180 0.0167 0.0163 0.0160 0.0155 0.0147 0.0168 1.6246
-0.01 0.0066 0.0042 0.0030 0.0026 0.0023 0.0020 0.0017 0.0032 1.4414
0.00 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
0.01 0.0085 0.0060 0.0047 0.0043 0.0041 0.0037 0.0035 0.0050 1.4566
0.02 0.0240 0.0215 0.0203 0.0199 0.0196 0.0191 0.0183 0.0204 1.5962
0.03 0.0496 0.0469 0.0457 0.0452 0.0450 0.0441 0.0425 0.0456 1.9052
0.04 0.0849 0.0819 0.0809 0.0805 0.0802 0.0789 0.0761 0.0805 2.3282
0.05 0.1305 0.1273 0.1264 0.1261 0.1257 0.1239 0.1195 0.1256 2.8915
Fig. 8. Polynomial approximation versus measured values
and m = 7 columns (corresponding to m excitation currents),
which contain the n m measured data with:
B
ref;i,j
= B
r;i,j
I
0
I
j
. (10)
The average value of these reference coefcients, for each
measuring position is presented in the Table II (where is
included also its standard deviation).
If the current is considered smaller or equal to I
0
, the
results are presented the Table III. For currents lower than
the reference current, the error of linear model is acceptable,
being under 1.1%. So, we can conclude that the reference
current do not saturates the magnetic core.
TABLE III
THE REFERENCE FIELD COEFFICIENT AND ITS STANDARD DEVIATION FOR
CURRENTS LOWER THAN I
0
B
ref
(i, j)[T]
I II III IV

B
ref
std %
-0.05 0.1215 0.1184 0.1174 0.1171 0.1186 1.0153
-0.04 0.0779 0.0750 0.0739 0.0735 0.0750 1.0141
-0.03 0.0441 0.0415 0.0403 0.0399 0.0415 0.9693
-0.02 0.0205 0.0180 0.0167 0.0163 0.0179 0.9579
-0.01 0.0066 0.0042 0.0030 0.0026 0.0041 0.9357
0.00 0.0000 0.0000 0.0000 0.0000 0.0000 00.0000
0.01 0.0085 0.0060 0.0047 0.0043 0.0059 0.9583
0.02 0.0240 0.0215 0.0203 0.0199 0.0214 0.9532
0.03 0.0496 0.0469 0.0457 0.0452 0.0468 0.9978
0.04 0.0849 0.0819 0.0809 0.0805 0.0821 1.0287
0.05 0.1305 0.1273 0.1264 0.1261 0.1276 1.0487
The errors under 3% in the Table II validate the linear model
(10), also for higher currents. Therefore, the radial ux density
of the characterized electromagnet, along horizontal radius,
within its aperture has following variation:
B
r
(r, I) = f
(
r
r
0
)

I
I
0
=
[
c
2
(
r
r
0
)
2
+c
4
(
r
r
0
)
4
+
+c
6
(
r
r
0
)
6
+c
8
(
r
r
0
)
8
+c
10
(
r
r
0
)
10
]
I
I
0
.
(11)
TABLE IV
FITTING WITH EVEN POLYNOMIALS
Degree II IV VI VIII X
c
2
0.1208 0.1158 0.1093 0.0976 0.06939
c
4
0.005962 0.02744 0.1062 0.4759
c
6
-0.01515 -0.1611 -1.586
c
8
0.0789 2.13
c
10
-0.9678
suma 0.1208 0.1218 0.1216 0.1216 0.1215
sse 5.048e 6 2.037e 6 1.141e 6 5.333e 6 1.642e 6
RMSE 0.001005 0.0007136 0.0006167 0.0005164 4.052 10
9
Fig. 9. Even polynomials
D. Least-squares method
The experimental data sets, preprocessed by averaging val-
ues, measured at symmetric positions were tted with even
polynomials, by using the least squares method. The most
reliable model is that having 6th degree. At higher grades,
the coefcients do not decrease monotonically with the order,
probably due to measurement error and numerical instability,
results can be seen in the Table IV, and in Fig. 9,. Similar
results can be seen in the Table V, and in Fig. 10, where the
results are obtained by using measurements in all 11 positive
and negative positions, without previous averaging.
Fig. 10. Fitting of the 11 experimental values
TABLE V
FITTING THE 11 EXPERIMENTAL VALUES WITH EVEN POLYNOMIALS
Degree II IV VI VIII X
c
2
0.1208 0.1158 0.1093 0.0976 0.06939
c
4
0.005962 0.02744 0.1062 0.4759
c
6
-0.01515 -0.1611 -1.586
c
8
0.0789 2.13
c
10
-0.9678
suma 0.1208 0.1218 0.1216 0.1216 0.1215
sse 9.712e 8 9.11e 5 8.931e 5 8.809e 5 8.703e 5
RMSE 0.003116 0.003182 0.003341 0.003547 0.003808
IV. THE FIELD DISTRIBUTION IN THE APERTURE, BASED
ON PERIPHERAL MEASUREMENTS.
Fig. 11. The 30 degree slice extracted from circular aperture
A "slice" of 30 degree is extracted from circular aperture
disk, (Fig. 11). The boundaries of this sub-domain D are:
C
1
on symmetry axe of the pole;
C
2
on the polar inter-axes;
C
3
on the aperture border, with radius r
0
= 50mm.
Magnetic eld satises the following boundary conditions:
(c1): B
n
= 0 on C
1
, because the magnetic eld lines are
tangential along this radial direction.
(c2): H
t
= 0 on C
2
, because the magnetic eld lines are
perpendicular to this radial direction.
(c3): B
n
= B
r
known - in this procedure, the radial ux
density is measured on C
3
- the aperture periphery.
If we will consider that the angle originates from C
1
, then
from (5) and (c1) follows that all coefcients B
n
are zero.
From (c2) and (5) we have:
A
n
cos
(
n
6
)
= 0
(2k + 1)

2
=
n
6
n = 3(2k + 1).
(12)
For any value of the index n, which is not a multiple of
three times odd numbers A
n
= 0. Therefore, the magnetic
scalar potential, in aperture has only cos harmonics with orders
3, 9, 15, . . . and the radial component has polynomial variation
vs. radius of orders 2, 8, 14, . . .:
B
r
(x

) =

(
A
n
x
n1
cos (n)
)
, n = 3, 9, 15, . . . (13)
For = 0, it results following variation on C
1
:
B
r
(x

) =

(A
n
x
n1
) =
= A
3
x
2
+A
9
x
8
+ = c
2
x
2
+c
8
x
8
+
(14)
The presence of non-zero terms c
4
x
4
, c
6
x
6
and c
10
x
10
in (7)
has three possible explanations:
structural asymmetries;
measurement errors;
computational errors.
V. THE FIELD DISTRIBUTION IN THE APERTURE, BASED ON
RADIAL MEASUREMENTS
If these errors would be zero, then the coefcients c
2
, c
8
, c
14
extracted from the measurement, determining the Fourier co-
efcients of harmonic solution: A
3
= c
2
, A
9
= c
8
, A
15
= c
14
.
With them, using (5) can be determining the variation of the
ux density vector eld, throughout the aperture. Magnetic
eld distribution in the electromagnets aperture can be deter-
mined with (5) where coefcients A
n
, B
n
can be computed
by using the model (11), extracted from experimental data.
It should be determined the coefcients c
2
, c
8
and c
14
from
the experimental data, looking for the the best approximation
of a polynomial with these coefcients (ascending order:
2, 8, 14).
Unfortunately the conditioning number of high polynomial
interpolation is extremely high. According Gautschs theorem,
in our case, this number has the approximate value = (1 +

2)
n+1
= 16238 for n = 10 [9].
This means that a disturbance of 1% in experimental data
is amplied ten thousand times in the computed polynomial
approximation. This behavior makes totally irrelevant the
numerical result.
In this approach, it was actually solved the Laplace equation
with Dirichlet conditions on C
1
and C
2
. Moreover there are
not boundary conditions on C
3
, but instead an additional Neu-
mann condition was imposed on C
1
. Although this problem
has an unique solution, it is not well formulated problem,
because it is numerically unstable [10]. Moreover, this problem
is very similar with Hadamards famous counterexample [11].
Unlike this problem, if Dirichlet or Neumann the boundary
conditions are imposed in each point of the domains border
D, including on C
3
, as in section V, then the problem is well-
formulated, in Hadamard sense. Consequently, its conditioning
number is not greater than 1 [10].
VI. CONCLUSIONS
The eld problem based on peripheral measurements is well
conditioned, because small changes in boundary data are not
amplied, and the corresponding variations of the eld in any
point within the aperture D has a relative error not greater
than the relative measurement error.
It should be noted that, from practical point of view, the
effort to make measurements is the same. With the same
number of measurements, the error in determining the eld
inside aperture is much higher, when measurement are used
along radius, than in the case of using circular measurements.
According to (14) the best polynomial approximation with
three terms is:
B
r
(x

) = c
2
x
2
+c
8
x
8
+c
14
x
14
. (15)
Their coefcients are represented in Table VI and its graphic
is depicted in Figure 12. However, as it was proven, this is not
a reliable result, because its conditioning number is 551614.
TABLE VI
FITTING WITH THE POLYNOMIAL B
r
(x

) = c
2
x
2
+ c
8
x
8
+ c
14
x
14
c
2
0.1161
c
8
0.02028
c
14
-0.01479
sum 0.12159
sse 1.910 10
6
RMSE 0.0007978
Fig. 12. Fitting of experimental values with the polynomial B
r
(x

) =
c
2
x
2
+ c
8
x
8
+ c
14
x
14
REFERENCES
[1] S. Russenschuck, Field Computation for Accelerator Magnets: Analyti-
cal and Numerical Methods for Electromagnetic Design and Optimiza-
tion. John Wiley and Sons, May 11, 2010.
[2] L. Walckiers, Magnetic measurement with coils and wires, CERN-
2010-004, pp. 357385, March 2011. http://arxiv.org/abs/1104.3784v1.
[3] A. Temnykh, Y. Levashovb, and Z. Wolf, A study of undulator magnets
characterization using the vibrating wire technique, Stanfort - LCLS-
TN-10-2.
[4] A. K. Jain, Measurements of Field Quality Using Harmonic
Coils. http://www.bnl.gov/magnets/staff/gupta/scmag-course/uspas01/
AJ01/HarmonicCoil_Slides.pdf.
[5] D. Dan and D. Ioan, Eroarea de metod a n determinarea experimen-
tal analitic a a cmpului unui electromanget pentru acceleratoarele de
particule, SNET12, Bucharest, 2012. http://snet.elth.pub.ro/snet2012/.
[6] ICPE-CA, Electromagneti si surse de alimentare pentru proiectul fair
- poveste de succes pentru icpe - ca, 2012.
[7] I. G. Shabac, Matematici speciale 2. Bucure sti, Romnia: Editura
Didactic a si Pedagogic a, 1965.
[8] ICPE-CA, Institutul National de Cercetare Proiectare pentru Ingineria
electric a. http://www.icpe-ca.ro/ro/http://www.icpe-ca.ro/ro/.
[9] K. A. Gallivan, Polynomial interpolation, 2013.
[10] S. Zaglmayr, High Order Finite Element Methods for Electromagnetic
Field Computation. Thesis - Linz Univ, 2006.
[11] R. Corless, Elementary Partial Differential Equations - Hadamards
example. http://www.apmaths.uwo.ca, 1998.

You might also like