A Radially and Rotationally

Adjustable Magnetic Mangle for

Electron Beams

Ishaan Vohra, Isabella Vesely, Peter Morand, Aubrey Zhang, William

Lu, William Soh, Achyuta Rajaram, Daniel Jeon

Myriad Magnets

Phillips Exeter Academy, Exeter, NH, USA

 A Radially and Rotationally Adjustable
Magnetic Mangle for Electron Beams

Contents

1 Introduction 1

2 Why we want to go 1

3 Aim of the experiment 1

4 Magnet design 1
4.1 Adjustable magnetic mangle . . . . . . . . . . . . . . . . . . . . . 1
4.2 Brace structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

5 Experiment design 8
5.1 Detector setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
5.2 Schedule . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
5.3 Data analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

6 What we hope to take away from the experience 11

7 Outreach activity 11

8 Acknowledgements 12

9 References 13

10 Appendix 13
10.1 Defining the corresponding ideal field . . . . . . . . . . . . . . . . 13
10.2 Rotating the magnets . . . . . . . . . . . . . . . . . . . . . . . . 14
1 Introduction
Excessive energy usage is a major contributor to climate change, a pressing world
issue [1]. Electromagnets used at accelerator facilities can consume large quanti-
ties of energy – it is therefore worthwhile to investigate alternative technologies
that provide the same capabilities.
We propose a multifunctional variant of the Halbach cylinder known as a
magnetic mangle [2, 3], designed for 0.5 – 6 GeV electron beams1 . Our mangle
can be radially adjusted to achieve a continuous range of interior magnetic
flux densities, and rotationally adjusted to switch between generating dipole
and quadrupole fields. Similar to conventional Halbach cylinders, our magnetic
mangle produces a strong magnetic field inside the array while generating very
weak magnetic fields externally, making it safer to use near other electronics
(such as cellphones, detectors, or pacemakers) than typical magnet setups.
In our experiment, we aim to demonstrate the viability of our mangle design.
We hope that our design and experiment will act as a proof of concept for future
electromagnet alternatives in accelerators.

2 Why we want to go
As passionate physics students, going to CERN/DESY will bring us amazing
insights into particle physics first-hand, allowing us to work with cutting edge
equipment alongside experts. We are very eager to turn our proposal into a
reality as we strongly believe that our design will help make future particle
physics experiments safer and more sustainable.

3 Aim of the experiment

We aim to validate our magnetic mangle’s ability to bend and focus electron
beams in the 0.5 – 6 GeV range. We will compare the observed electron deflection
distributions with our simulated deflection distributions in GEANT4 to analyze
how well our predictions align with the real performance of the magnet [4].
We also aim to further investigate any discrepancies that arise between the
experiment and simulation.

4 Magnet design
4.1 Adjustable magnetic mangle
A magnetic mangle consists of N permanently diametrically-magnetized cylinders
arranged in a circle such that the desired flux density and polarity is achieved in
1We have chosen to design our mangle for electrons over other charged particles because

both CERN and DESY can provide high purity electron beams (see Section 5).

1
the central cavity, while only a very weak flux density is produced externally [5].
One rotational arrangement of the cylinders creates a dipole field in the cavity,
while another creates a quadrupole field (see Figure 1). Moreover, by moving
each cylinder radially outward or inward from the center of the array, the flux
density within the cavity can be decreased or increased, respectively.

(a) Dipole arrangement (b) Quadrupole arrangement

Figure 1: Simulated flux diagrams of the dipole and quadrupole rotational

arrangements for N = 8 cylinders. Black arrows denote the direction of magne-
tization of each cylinder.

Since magnetic mangles consist of a finite number of cylinders with fixed

directions of magnetization, the fields they generate deviate from ideal dipole and
quadrupole fields, as can be seen near the edges of the interior field in Figure 1.
In general, as N increases, the deviation from the ideal field decreases, improving
the accuracy with which electrons are deflected. However, if too many cylinders
are used, it becomes impractical to rotate each one between configurations.2
To optimally balance these two factors, we used the ANSYS Maxwell simu-
lation software [6] to model 2D cross-sections of mangles with various N . The
deviation of each mangle field from the corresponding ideal field (see Appendix
Section 10.1) was quantified using the relative absolute error (RAE) given by
v
uX
u n
t ~ mangle
|B i
~ ideal |2
B i
i=1
RAE = v , (1)
uX
u n
t ~ ideal |2
|B i
i=1

2 Our mangle design provides the flexibility to individually rotate each cylinder to allow for

magnetic field arrangements other than dipole and quadrupole.

2
 where B~ mangle and B
~ ideal are the mangle field and corresponding ideal field
i i
vectors at a given sample point i out of n total sample points. The results of
the analysis are shown below in Figure 2.

Figure 2: Plot of the RAE for 2D cross-sections of magnetic mangles with

di↵erent numbers of cylinders.

As shown, N = 8 is the fewest number of cylinders needed to have both

dipole and quadrupole RAE under 5%. Thus, we opted for 8 cylinders, allowing
the mangle to generate fields that do not deviate much from the ideal while
maintaining rotation efficiency.
We then conducted a cost analysis based on the experiment budget and
settled on the design shown in Figure 3.

3
Figure 3: Simulated design of our magnetic mangle to be constructed using
N35-grade NdFeB. We aim to use sustainably-sourced rare-earth metals in order
to minimize the environmental footprint of the experiment.

The radial adjustment range of d = 5.5–10.0 cm considers not only the 5x5 cm
minimum cavity size, but also the coercivity limit of typical N35 NdFeB magnets
– according to our simulations, this range provides a 20% margin between the
maximum flux density experienced by any part of the mangle and the magnets’
µ0 Hc ⇡ 1.2T. Above this limit, the cylinders’ directions of magnetization may
change undesirably.
For more detail on the expected cost and sourcing of the mangle’s components,
refer to our cost analysis.3
Further simulations of the final mangle design were also performed (see
Figures 4 and 5) to model B, the dipole flux density, and g, the quadrupole
flux density gradient, for di↵erent values of d, as well as to demonstrate that
the external field grows weaker with increasing distance. Prior to installing
the mangle in the beamline, we intend to measure its field pattern in di↵erent
configurations using a magnetometer to compare against the simulated results.
3 See our cost analysis spreadsheet at: https://docs.google.com/spreadsheets/d/
1oWpocO46t6OGfqjlob-5VgnrV-Hy8mJZ3GOcQm68XXM/edit?usp=sharing

4
(a) Dipole: magnetic flux density magnitude (sampled from center of cavity) for di↵erent
radial distances of the cylinders.

(b) Quadrupole: magnetic flux density gradient for di↵erent radial distances of the
cylinders.

Figure 4: Flux density and flux density gradient plots for di↵erent dipole and
quadrupole radial configurations in the final mangle design.

5
Figure 5: Mean flux density at a given radial sample distance from the cavity
center for dipole and quadrupole configurations with d = 5.5 cm. As shown by
the dotted black line in the top right diagram, the mean flux density at a given
radial distance is found by uniformly sampling points around a circular ring and
averaging the flux density. As seen in the blue region, the exterior flux density
decreases to about 0.2 mT within 20 cm from the cavity center – this is a level
considered safe for most electronic devices.

4.2 Brace structure

Our brace structure (Figures 6 and 7) enables continuous radial adjustments and
rotational freedom around the magnets’ central axes. A radial linkage translates
the magnets and may be manually adjusted by shortening and extending four
aluminum turnbuckle-style rods connecting the moving pin joints. We determined
through our simulations that the magnets exert up to 350 N of radially-directed
force and 4 Nm of torque on each other; the titanium brace structure has thus
been designed to withstand this. The linkage structure itself is attached to a 76.2
mm-thick standing block of ultra high molecular weight polyethylene, providing
cost-efficient stability. Since polyethylene is non-magnetic and aluminum and
titanium are only weakly paramagnetic, we expect the brace structure to have
a negligible e↵ect on the mangle’s magnetic field. For a description of how the
magnets can be rotated, see Appendix Section 10.2.

6
Figure 6: CAD design of the brace structure with magnets in place: magnets
(black) are attached to the inner linkage joints through aluminum casings (blue).
Set screws are installed optimally along the axial line and at a 60 displacement
angle, applying friction and thus restricting rotational freedom to any chosen
angle and allowing for both dipole and quadrupole arrangements in any direction.

(a) 9.2 cm radius arrangement (b) 5.6 cm radius arrangement

Figure 7: Two configurations for varied radii. The orange, turn-buckle style
rods can be screwed and unscrewed to adjust their length, thus adjusting the
magnets’ distance from center.

7
5 Experiment design
To minimize the number of stray particles in our beam, we plan to use either the
0.5–4 GeV high-purity electron beam at CERN or the 1–6 GeV electron beam
at DESY [7].

5.1 Detector setup

Figure 8: Detector setup to measure the electron deflection through the magnetic
mangle.

• Before entering the mangle, the beam telescope measures the incoming
angle of electrons with high spatial resolution. At this stage, the beam
spot is only about 2 cm in diameter, so the beam will pass within the 2x2
cm sensor area.
• Two DWC’s track the positions of electrons deflected by the magnetic
mangle. The first DWC is placed near the mangle and the second is placed
further away to study the beam divergence and the deflection due to the
magnet.
• Data from the Cherenkov detector, halo counter, and calorimeter can be
used to filter events during data analysis (see Section 5.3).

5.2 Schedule
We plan to run our experiment in about 9 days with the following schedule:

• Days 1-4: Set up experiment, align detectors, integrate the DAQ.

• Day 5: Control run with detectors but without magnetic mangle (to study
intrinsic beam divergence).

• Day 6: Dipole configuration at 4 energies.

8
 • Day 7: Dipole configuration (with a radial adjustment) at the same 4
energies.
• Day 8: Quadrupole configuration at 4 energies.
• Day 9: Quadrupole configuration (with a radial adjustment) at the same 4
energies.

During the following days, we could analyze data and conduct additional runs
if necessary. Extra runs could involve the same above setups, or if time permits,
we could investigate the electron deflections using other angular configurations
besides dipole and quadrupole, such as a field that both bends and focuses
electrons (a gradient magnet).

5.3 Data analysis

Each event is analyzed particle-by-particle. Halo counter data is first used
to filter out any uncollimated stray particles. Although the electron beam is
known to be of high purity, the Cherenkov detector is used to tag and filter out
remaining pions or muons.
Calorimeter data is used to reject electrons which have lost & 100MeV.4
These electrons may have collided with the air and changed direction or lost
too much energy emitting synchrotron radiation in the mangle’s magnetic field
(among other e↵ects). Such backgrounds and mechanisms of energy loss would
cause erroneous signals in the DWCs, and would therefore need to be filtered
out.
From the remaining electron deflection events, plots are created to characterize
the observed deflections over di↵erent energies and mangle configurations. The
plots can then be compared to the simulated plots in GEANT4 (Figures 9 and
10) to assess the true performance of our mangle.
4100 MeV is a temporary estimate for the energy threshold of the DWCs – this value may

be adjusted when better understood.

9
 (a) No magnetic mangle present in beam- (b) Dipole configuration with radial ar-
line rangement of d = 6.0 cm (B = 0.29 T)

(c) Quadrupole configuration with radial

arrangement of d = 7.0 cm (g = 6.1
T/m)

Figure 9: GEANT4 simulation: position distributions of 1500 electrons on a

tracking detector 4.9 m from the mangle. The plots were simulated using a
circular particle source 2 cm in diameter and a simplified detector volume for
the tracking detector.

10
Figure 10: GEANT4 simulation: Normalized deflection angle distributions at
1.0 GeV and 2.0 GeV passing through the mangle dipole configuration.

6 What we hope to take away from the experi-

ence
Working on this project has proven invaluable to our development as budding
physicists. As we complete our research, we are gaining first-hand exposure to the
practical considerations, physics knowledge, and dedication involved in improving
energy efficiency at the bleeding edge of high-energy science. Additionally, by
collaborating with the researchers at CERN and DESY, we hope to learn
invaluable skills that we can use throughout our scientific careers.

7 Outreach activity
We hope to extend our e↵orts to visit local elementary and middle schools to
expose younger students to STEM. We plan to collaborate with our school’s
service organization to organize these events, making use of their close connections
with the local libraries and primary schools. We plan to foster curiosity and
excitement in physics by holding approachable physics conversations, focusing

11
on concepts rather than math. By leading engaging and easily understandable
talks, we hope to inspire the next generation of scientists.
We also plan on holding several talks through our physics club about our
proposal. We will prioritize reaching out to members of our school community
who are not yet involved in science clubs through social media, posters, and
word of mouth.

8 Acknowledgements
We would like to sincerely thank our physics teachers, Mr. James DiCarlo and Mr.
Scott Saltman for granting us permission to compete in the Beamline for Schools
Competition and for supporting our project. We are also extremely grateful to
Dr. Mark Adams, Dr. Don Lincoln, Dr. Elvin Harms Jr, Dr. Josh Barrow, and
Mr. Daniel MacLean (all of Fermilab) for guiding us along the way with their
valuable advice. We would also like to thank Dr. Efstathios Stefanidis (former
ATLAS physicist), Dr. Rasmus Bjork (of the Denmark Technical University),
and Dr. Marco Roda (of the University of Liverpool) for their guidance and
support. This proposal also would not have been possible without the assistance
of Dr. Markus Joos and Dr. Margherita Boselli (of BL4S) who swiftly answered
all of our questions, as well as Mr. Kenneth Cecire, the United States national
contact for the Beamline for Schools Competition. Lastly, we would like to
thank our school, Phillips Exeter Academy, for providing us with an exceptional
learning environment to pursue our passion for physics.

12
9 References
[1] K. Gillingham, R. G. Newell, and K. Palmer, “Energy efficiency economics
and policy,” Annu. Rev. Resour. Econ., vol. 1, no. 1, pp. 597–620, 2009.
[2] R. Bjørk, C. R. H. Bahl, A. Smith, and N. Pryds, “Optimization and
improvement of halbach cylinder design,” Journal of Applied Physics, vol. 104,
no. 1, p. 013910, 2008.
[3] C. Fernandes, J. Ventura, and D. Silva, “Rod mangle rotation patterns for
adjustable magnetic field generation,” Journal of Magnetism and Magnetic
Materials, vol. 565, p. 170227, 2023.
[4] S. Agostinelli, J. Allison, K. a. Amako, J. Apostolakis, H. Araujo, P. Arce,
M. Asai, D. Axen, S. Banerjee, G. Barrand, et al., “Geant4—a simulation
toolkit,” Nuclear instruments and methods in physics research section A:
Accelerators, Spectrometers, Detectors and Associated Equipment, vol. 506,
no. 3, pp. 250–303, 2003.
[5] H. A. Leupold, “Mangle magnetic structure,” Apr. 23 2002. US Patent
6,376,959.
[6] A. S. Martyanov and N. I. Neustroyev, “Ansys maxwell software for elec-
tromagnetic field calculations,” Eastern European Scientific Journal, no. 5,
2014.
[7] “Beams and detectors.”

10 Appendix
10.1 Defining the corresponding ideal field
For each set of cross-sectional mangle fields with a given N , we define the
corresponding ideal fields (centered at the origin) to be

~ dip (x, y) = [0, B]

B (2)
in the dipole case and

~ quad (x, y) = g[ x, y]
B (3)
in the quadrupole case. The magnitude of the ideal dipole’s flux density, B, is
obtained from the flux density at the center of the array. The ideal quadrupole’s
magnetic flux gradient, g, is obtained through a linear regression of the flux
density magnitude plotted against sample distance from the center of the cavity.
For each mangle field, thousands of coordinates were sampled uniformly
from a circular area of the cavity with radius 3.54 cm (corresponding to the
circumcircle of the 5x5 cm minimum cavity size through which the electron beam

13
passes). From these coordinates, the RAE was then calculated to quantify the
deviation of the mangle field from the corresponding ideal field (see Section 4.1).

10.2 Rotating the magnets

To rotate the magnets, the mangle is first set to its largest radial configuration
(d = 10.0 cm) to reduce the forces and torques between the magnets. Custom
cranks are attached to each magnet and tightened with a set screw; the magnet
casing is then loosened, and the crank rotates them to the desired rotational
configuration. Markings on the casing align with the magnets’ markings at
common angle intervals (every 5 ).

(a) Casing with all four set screws visible (b) Set screws inside the casing

14