Development and Manufacturing of a Nb3Sn
Quadrupole Magnet Model at CEAlSaclay
for the TESLA Interaction Region
M. Durante, A Devred, M. Fratini, D. Leboeuf, M.
Segreti, P. V 6drine
18th International Conference on Magnet Technology,
Morioka (Japon) October 20-24. 2003
Departement d'Astrophysique, de Physique des Particules, de Physique Nucleaire et de I'lnstrumentation Associee
DSM/DAPNIA, CEA/Saclay F - 91191 Gif-sur-Yvette Cedex
Tel: (1)69082402 Fax: (1) 69089989
Development and Manufacturing of a Nb3Sn
Quadrupole Magnet Model at CEAlSaciay
for the TESLA Interaction Region
M. Durante, A. Devred, M. Fratini, D. Lebreuf, M. Segreti and P. Vedrine
superconducting properties far exceed those of NbTi, is the
fabrication of short and powerful quadrupole magnets for the
Interaction Regions (IR) of large particle accelerators. In some
cases, as in the future linear collider TESLA, the quadrupole
magnets are inside the detector solenoid and must operate in its
background field. This situation gives singular Lorentz force
distribution in the ends of the magnet. To learn about Nb3Sn
technology, evaluate fabrication techniques and test the
interaction with a solenoidal' field,
CEAJSaclay has started the manufacturing of a I-m-long, 56-mm
single-aperture quadrupole magnet model. The model relies on
the same coil geometry as the LHC arc quadrupole magnets, but
has no iron yoke. It wiD produce a nominal field gradient of
211 Tim at 11,870 A. The coils are wound from Rutherford-type
cables insulated with glass fiber tape, before being heat-treated
and vacuum-impregnated with epoxy resin. Laminated collars,
locked around the coll assembly by means of keys, restrain the
Lorentz forces. After a recall of the conceptnal design, the paper
reviews the progress in the manufacturing and test of the main
components as well as in the design and delivery of the main
tooling. The first coil is expected to be wound and heat treated
during the last quarter of 2003.
possible application of Nb3Sn, whose
Index Terms---superconducting Quadrupole Magnet, Nb3Sn,
Wind & React.
superconducting dipole and quadrupole magnets, which :rre
located close to the interaction points. As the InteractIOn
Regions (IR's) are very crowded, see for instance the design of
the Large Hadron Collider (LHC) Interaction Regions [1),
there are definite advantages in increasing the field integral or
the focalization power of these magnets to reduce their length
and save some space.
HE fmal focalization of high intensity beams in large
particle colliders usually requires sets of strong
Manuscript received October 21, 2003.
M. Durante is with CENDSMlDAPNWSACM, 91191 G i f - s u r - Y v e t ~ e
Cedex, France (phone: 33169087127;
A Devred is with CEAlDSM/DAPNWSACM and CERN/ATIMAS, CR
1211, Geveve 23, SWITZERLAND (email: a r n a u d . d e v r e d @ ~ e m . c h ) ..
M. Fratini was with CENDSMIDAPNWSACM. S h e . l ~ now :V1th TEA
Sistemi, piazza Mazzini 1, 56100 Pisa, Italy (e-mail: m.frattnt@Cpr.lt).
M. Segreti and P. Vedrine are with CENDSMIDAPNWSACM.
D. Lebamf is with CENDSMIDAPNWSIS.
fax: 33169089283; e-matl:
Furthermore, in some accelerator designs, such as in the
TeV Superconducting Linear Accelerator (TESLA), under
study at DESY, the final focusing quadrupole magnets end up
inside the detector magnet, and must sustain a 4 T solenoidal
field [2). The requirement for high field and high field
gradient, or the ability to operate in a sizable background field,
precludes the use of NbTi. It has long been thought that
interaction region magnets could serve as a test bed for Nb3Sn
applications to large particle accelerators. For instance,
Twente University in the Netherlands is working since 1998
on a Nb3Sn model of beam-separation dipole magnet that
could advantageously replace the low-field magnets presently
considered for the LHC at CERN (3). Similarly, a
collaboration made up of Fermilab, Brookhaven National
Laboratory (BNL) and Lawrence
Laboratory (LBNL) is promoting a US-LHC Accelerator
Research Program (LARP) aimed at upgrading the final
focusing quadrupole magnets of LHC [4). More recently, a
collaboration involving 6 European institutes is pushing
forward a proposal to build the Next European Dipole (NED),
which targets also LHC IR upgrade [5).
CEAlSaclay has undertaken, since 1996, an R&D program
aimed at designing and building a short Nb3Sn quadrupole
magnet model. To save time and money, the model design is
based on the design of the quadrupole magnets for the arcs of
the Large Hadron Collider (LHC). The 3-m-Iong, 56-mm-twin
aperture LHC arc quadrupole magnets, relying on NbTi cables
and polyimide insulation, have been developed at Saclay for
CERN [6) and are now under series production at ACCEL
Instrument GmbH (7).
The Nb3Sn magnet model described here is l-m-Iong and
has a single-aperture of 56-mm and no iron yoke. It is
developed in collaboration with AlstomIMSA which was
responsible for producing the conductor.
The four Nb3Sn coils will be realized following the so called
"wind, react and impregnate" technique. The coils will. be
manufactured separately. They will be wound in the desrred
saddle-shape, starting from a cable made up of Nb3Sn
precursors and copper stabilizer, and then heat treated at
6600C in an inert gas flow for about 240 hours. After heat
treatment, the coils will be vacuum-impregnated with epoxy
resin to confer them a rigid shape and facilitate subsequent
II. FINAL DESIGN OVERVIEW
A. Electromagnetic design
The quadrupole magnet model relies on the same coil
geometry as the LHC arc quadrupole m a g n e t s ~ but it has no
iron yoke. At a nominal current of 11,870 A, the field gradient
is 211 Tim. The components of the Lorentz force over a coil
octant at 11,870 A are: 400 kN/m (outwardly) along the pole
mating planes and 711 kN/m (downwardly) along a
perpendicular direction. The high order multipole components
are all below 10-4 units.
B. Mechanical Design
A cross-sectional view of the quadrupole magnet cold mass
is shown in Fig. 1. The two layers of each coil will be wound
"in one go" without internal splice. After coil heat treatment
and impregnation with epoxy resin, mechanical massaging will
be carried out on the whole length of each pole using the coil
size measurement machine. This massaging is necessary to
reduce the effect of the singularity observed during the initial
loading of conductor stacks representative of coil straight
section (see section IV). T h e n ~ the four coils will be assembled
together and covered with quench protection heaters and
polyimide ground plane insulation. The coil assembly will be
restrained by laminated, 2-mm thick, austenitic steel collars
locked by eight, full-length, tapered keys. The 8 keys will be
driven into keyways on the collar outer surface by means of a
press. As for the LHC arc quadrupole m a g n e t s ~ the collaring
will be performed vertically. Finally, the collared-coil
assembly will be centered within a precisely-machined steel
inertia tube delimiting the region of liquid helium circulation.
The aim of the collaring operation is to apply a large
azimuthal pre-compression to the coil by the use of the collar
poles. The pre-compression is needed both to compensate the
thermal shrinkage differentials during cool-down and to
compensate stress redistribution during excitation.
Fig. 1. Cross-sectional view ofNb)Sn quadrupole magnet modeL
The fmite element (FE) model of the structure has been
developed using the COF AST3D approach , which is a
module developed specially for the CASTEM software
package . COFAST3D is based on a decomposition of the
structure into sub-structures for an easier iterative resolution
scheme. It shows a great reduction of computing time when
complex and multiple contact zones are taken into account,
especially in a 3 D approach like here.
The FE model is restricted to 1I4th ofthe quadrupole magnet
cross section. All successive steps of loading history, from
collaring to cool-down and to excitation are described. The
thermo-mechanical properties of conductor blocks have been
measured on ten-stack samples fabricated following processes
similar to those foreseen for coil production .
Fig. 2 shows the computed azimuthal stress distribution in
the coil at 4.2 K and 11,870 A. All the parts of coil assembly
remain in compression and the peak stress on the conductors is
always below the critical level of 150 MPa. At the pole area,
the minimum stress remains superior to 13 MPa thereby
avoiding a separation at the coiVcollar pole interface during
energization and preventing field quality distortions.
Fig. 2. Azimuthal coil stress distribution at 11,870 A.
III. CABLE DEVELOPMENT
A. Model Cable Manufacturing
Cable design has been developed in collaboration with
AIstomlMSA following closely
specifications. The chosen conductor is a Rutherford-type
cable, 15J-rom wide, 1.48-mm thick (mid-thickness), with a
keystone angle of 0.9°. 36 Nb3Sn strands, with a nominal
diameter of 0.825 rom and a copper-to-non-copper ratio of
about 1.4 to 1, are arranged in two layers, separated by a
13-mm wide, 25-/lm-thick, stainless steel core. Strand design
is based on internal-tin process. Each strand is made up of 19
multi filamentary bundles, spaced by 6 CuSn elements and
surrounded by a double Nb/Ta diffusion barrier and a copper
~ _ 2 0
LHC outer cable
The critical current density in the non-copper of the final
strand has been measured to be of the order of 1850 Nmm2 at
4.2 K and 7 T (yielding 765 Nmm2 at 4.2 K and 12 T). The
effective diameter of the filaments after the heat treatment was
measured to be -19 j.Un .
AlstomlMSA has started final strand and cable production.
The delivery at Saclay of the fmal cable is planned for the end
of 2003. Critical current measurements will be carried out on
virgin and extracted strands.
B. New Cable Development
A new wire and cable development has been initiated in
2002 with AlstomIMSA to boost the critical current density of
the wire up to 2000 Almm2 at 4.2 K and 12 T. The other
parameters are the same as for the model cable. The boost in
critical current density is required to meet the performance
specifications for the TESLA fmal focusing quadrupole
IV. INSULATION DEVELOPMENT AND TESTS
A. Insulation Development
Cable insulation for the quadrupole magnet model relies on
a mineral fiber tape, double-wrapped around the cable prior to
winding. After heat treatment, the tape wrapping is completed
by a vacuum impregnation with epoxy resin, enhancing
dielectric strength and providing a rigid bonding. The tape
ensures a proper spacing between coil turns and prevent resin
crack. According to electromagnetic design, the turn-to-turn
insulation thickness must be of the order of 220 Ilm under
80 MPa. This tight dimensional constraint is imposed by the
fact that we are re-using the electromagnetic design of the
LHC arc quadrupole magnets, that had been optimized for a
After an R&D program, based on tensile and dielectric
strength measurements, to compare various insulation
systems , the chosen tape was a 15-mm-wide, 60-llm-thick
quartz fiber tape. Quartze1™ fibers with diameters as small as
17 gIkm are commercially available but they need to be woven
on old wooden looms to limit tension and risk ofbreakage.
successfully. Nevertheless, we had some difficulties in
realizing cable wrapping. This is due to the fact that the tape
sizing must be removed prior to wrapping to prevent
subsequent pollution during coil heat treatment. Tape sizing
removal is carried out by carbonization in air at 350
Without sizing, the tape becomes fragile and cannot be
wrapped around the cable using an automatic wrapping
machine and only hand-made wrapping turned out to be
Another problem appeared to be the control of tape width.
Indeed the reliance on an archaic 100m does not allow to
perform all the on-line controls that can be performed on more
modem machines. The woven tapes exhibited some width
irregularities, eventually leading to non-reproducible wrapping
or over thicknesses. Further tests varying the tape weaving
underwent all tests
parameters (number of warp and fill ends) did not lead to
better results. This problem could be resolved by modifying or
rebuilding the loom - a work well beyond the scope and
financial reach ofthe project.
Keeping in mind our primary goal of developing industrial
fabrication processes, we preferred to check other solutions,
using thicker glass fiber tapes. 80-llm-thick E-glass fiber tapes
and 100-llm-thick S2-glass fiber tapes have been taken into
account. As described in the next section, tests on stacks made
up of 10 insulated conductors have been carried out to study
the influence of tape overlapping (from 20% to 50%) and tape
thickness on composite behavior.
B. Ten-stack test results
Two series of stacks of ten insulated cables were
manufactured and tested in compression. Cable wrapping was
performed in one single layer with an overlap of respectively
20,33 and 47 % for S-glass fiber tape and 33,40 and 50% for
E-glass fiber tape. The cable stacking was alternated to
compensate for the keystone edge. Stack height was controlled
during heat treatment and impregnation by means of carefully
Each sample was mounted in a V-shape, stainless-steel
holder and was loaded in compression along its transverse
direction through a stainless steel upper bar. The stress-strain
curve of each sample was monitored for three successive
cycles at room temperature. Each stack was tested twice.
Composite behavior, as illustrated in Fig. 3, is similar to that
observed on quartz-tiber-based samples . The first loading
curve for a virgin stack exhibits a pronounced non-linear
behavior, whose average slope is smaller than the slopes ofthe
subsequent loadings. Then, the curve exhibits an hysteresis
between loading and unloading, which evolves from cycle to
cycle and seems to be more or less stabilized after three
loadings. For the second test, the first loading still remains
different than the following ones, but in a less dramatic way
than for a virgin stack.
1 00 -,---~--~"-----~"~--------~,-----~.---"-~-,--~---~--.
height ( mm)
Fig. 3. Compressive test curves for a S2-glass-fiber-based sample.
The strong non-linear behavior of virgin stacks could be
explained as a rearrangement of the strands in the cables.
Additional tests on ten stacks made up of non insulated cables
seem to confirm this hypothesis.
In Fig. 4, ten-stack dimensions under 80 MPa for the first
cycle ofthe second test (corresponding to the collaring process
in the "massaged" coil history) are related to initial dimensions
(measured with a micrometer after vacuum impregnation,
equal to the impregnation mould cavity height). The final
dimension of the stack seems to be driven by the initial height
of the impregnation mold, more or less independently from
overlapping percentage. The fitting curves are a little bit
different for the two insulation systems, but show a linear
behavior with a slope equals to 1. Based on these tests, the S2
glass system was chosen for cable insulation.
S 18.6 - , - - - - - - ~ - - T ~ - - - - - - - - ~ - ~ - _ , ~ - - ~ ~ - - - - - - ~ - _ r _ - - ~ - ~ - - - - _ ,
E 18.4 +---------}-------------!----------tz..--------f
;; 18.2 - 4 - - - - - - ~ - - - - - - - - - - + - - - - - ~ + _ - - - - _ _ _ 1
~ ? 18 +-----------+-------+_"87""-----"'....---------1
~ 17.6 + - - - - - - ~ - - - + - - - ~ 7 " ' 7 , £ - - l - - - - - - - - - - - - . _ - - - - - - - - _ _ _ ;
§ 17.4 t - - - - - - - - - - + : , . I " ~ - - - - - - ~ - - - - - - - - + - - - - - - - - ~
1: 17.2 + - - - - - ~ I f I # Z - - - - - + - - - - - - + - - - - - - - _ _ j
17 t - - - 7 " ' ~ ~ - - + - - - - - - - - - - - - - + - - - - - - - - - - - + - - - - - - - - - - -
% 16.8 +-------}---------+----------------+--------/
Initial height (mm)
Fig. 4. Final height versus initial height for S2-glass and E-glass based
samples. Measured values and fitting curves.
V. TOOLING AND COMPONENTS
The detailed design of components and tooling is almost
finished. Unlike initially foreseen , pole pieces are no
more integrated in the impregnated coil. Temporary pole
pieces will be used during heat treatment and impregnation,
and will subsequently be replaced by collar pole pieces during
collaring (as in LHC arc quadrupole magnet assembly
process). 0.5-nun-thick, stainless-steel protection sheets will
be placed between the coil and the pole pieces to prevent
damage of coil insulation.
Coil end spacers (Fig. 5) will be realized in AI-80 wt% Cu
alloy, machined using a 5-axis EDM tool and will be insulated
from the coil by means of O.l-mm-thick mica sheets. Tum-to
turn insulation pieces, cut out from O.l-mm-thick mica sheets,
will be added in the coil end regions.
The insulation spacer between coil layers will be made up of
a ceramic fiber sheet, 0.5-mm thick, that will be laser-cut to
avoid unweaving. Interlayer insulation will be completed by
coil vacuum impregnation with epoxy resin.
Fig. 5. End spacer machined from copper-aluminum alloy
VI. MANUFACTORY AND TESTS
Coil winding is expected to start before the end of the year
2003. Two dummy poles (pole 00 and pole 0) will be
manufactured (using dummy Nb3Sn cables), to validate
fabrication tools and procedures. Then, five poles will be
fabricated from the final cable. After fabrication and
mechanical test, the dummy coils will be cut and assembled to
validate collaring procedure. Both dummy and model coils
will be instrumented with capacitive-stress-gauges. Stress
evolution in the coils will be monitored during all steps of
magnet fabrication and test. The final coils will also be
instrumented with voltage taps.
To evaluate the influence of an external field on the coil
ends (as will be the case for the TESLA fmal-focusing
quadrupole magnets), the magnet model will be tested in the
aperture of a 2-T, 530-mm inner-bore MRI magnet available at
CEAlSaclay. The cold mass is expected to take place in June
The authors would thank the technical staff of SACM and
SIS involved in this project and especially A. Acker and
 ? R Ostojic, "The LHC Insertion Magnets,"-1§§!i,_T.,:!!.ns ... Appl.
Supercond., V o L l ~ N o . l J P p . · ) ? 6 : ~ O l , _ ~ ~ ~ q ~ ..,.. ~ . ~ " ,
 - ~ · R ' : " " S f i n K m a n r i : N. Walker, et aI., "A new delivery beam system (BDS)
for the TESLA linear collider," Proceeding ofEPA C 2000, p. 513.
 ? A. den Duden H.H.1. ten Kate et al., "A 10 T Nb3Sn model separator
dipole magnet for the CERN LHC." In L. Liangzhen, S. Guolia and
Y.-Luguang (eds.), Proceedings of 15th International Conference on
Magnet Technology (MI15),
pp. 137-140, 1998.
 ? S.A. Gourlay, "High field magnet R&D in the USA." in these
 ? A. Devred, D.E. Baynham, et al., "High Field Accelerator Magnet R&D
in Europe," in these Proceedings.
 ? M. Peyrot, 1.M. Rifflet, et al., "Construction of the new prototype of
main quadrupole cold masses for the arc short straight section ofLHC,"
IEEETrans. Appl. Supercond., V o L I 9 . N o . - 1 . ~ ) 10 .. 173 2000
 ? f):--lGische4P' Schmidt, et al., "Industrialization of the LHC Main
Quadrupole Cold Masses up to Series Fabrication," in these
 ?C. Blanzi:, L Champaney, P. Vedrine, "Contact problems in the design
of a superconducting quadrupole
Computations, Vol. 17 No.2, pp. 136-153,2000.
 ? CASTEM is a trademark of CEAlSaclay, France (web page:
 M. ? Reytier, A. Devred, et al., "Characterization of the ~ e r m o
mechanical behavior of insulated cable stacks representatIve of
accelerator magnet c o i l s : : J E E E . I m . t . l ~ . . : : ! I ! E l . . . . s J & P J ! f 9 E ! ! ! ! : : _ . . v 01. 11 No.1,
[Ilj M. 'Durante:P. Bredy, etal., "Development ofa Nb3Sn multifilamentary
wire for accelerator magnet applications", P ~ i c a S
A:i:Jewea; J>. Bredy, et al., "Insulation systems for Nb3Sn accelerator ?
magnet coils fabricated by the wind & react technique," 1 ~ ~ : _ C r y o . Eng. ?
(Materials), Vol. 46 .. p'p.. 143:-J5Q, ~ ( ) . 9 9 . _ ?
[13fT. ?Gourdm", A. Devred, et al., "Mechanical design of a NbJSn ?
quadrupole magnet," ProceedingofEPAC 2000, p. 2136. ?
Beijing, China: Science Press,