Nuclear Instruments and Methods in Physics Research A 427 (1999) 408}411

Particle trajectory tracing in ANSYS

Miklos Gyimesi, Vladimir Zhulin*, Dale Ostergaard
ANSYS, Inc. Southpointe, 275 Technology Dr., Canonsburg, PA 15317, USA

Abstract

The paper describes the elecromagnetic particle trajectory tracing capability of the ANSYS commercial "nite element
program.  1999 Elsevier Science B.V. All rights reserved.

1. Introduction MIT Plasma Fusion Lab, 3M [1}4]. Its submodel-

ing capability, allowing detailed analysis in one
Accurate particle trajectory tracing is of central part of the domain, has been exploited in ion im-
importance in the design of particle accelerators, plant analysis [5].
synchrotrons, electron microscopes, lithography ANSYS Revision 5.5 [6], to be released in 1998,
machines, cathode ray tubes, etc. ANSYS is a gen- will allow users to e$ciently compute and conve-
eral purpose commercial "nite element program niently visualize/animate particle trajectories de-
with electromagnetic, thermal, mechanical, #uid rived from analysis results obtained by state-of-
dynamic and acoustic capabilities. It has the fol- the-art "eld computation technology. The tracing
lowing technical features: algorithm provides machine precision accuracy
within an element. The paper summarizes ANSYS's
E element library (physics/geometry),
magnetic formulations an particle trajectory trac-
E coupled "eld problems,
ing capability.
E solid modeling, meshing,
E 3D graphics with animation,
E postprocessing,
2. Electromagnetic 5nite elements
E parameter language,
E design optimization.
Electromagnetic "eld problems are governed by
ANSYS has been used for collider, tokamak and Maxwell's equations:
cyclotron design at CERN, Fermi National Lab,
* *
curl H"J# D, curl E"! B,
*t *t
*
div B"0, div [J# D]"0, div D"o,
* Corresponding author: Tel.: #1-412-514-3103; fax: #1- *t
412-514-3118.
E-mail address: [email protected] (V. Zhulin) B"k H, D"e E, J"p E,

0168-9002/99/$ } see front matter  1999 Elsevier Science B.V. All rights reserved.
PII: S 0 1 6 8 - 9 0 0 2 ( 9 8 ) 0 1 5 2 8 - 9
 M. Gyimesi et al. / Nuclear Instruments and Methods in Physics Research A 427 (1999) 408}411 409

where E and H are the electric and magnetic "elds, functions over brick, wedge, tetrahedron and pyr-
B is the #ux density, J and D are the electric and amid elements; but this formulation is restricted to
displacement currents, o is the space charge, k is the iron free domains [7].
permeability, e is the permittivity and p is the con- To remove this restriction ANSYS recently de-
ductivity. Maxwell's equations can be reformulated veloped new edge-#ux formulation using 1st order
by potentials: non-conforming edge elements over brick [7], tet-
rahedron, pyramid and wedge [6,14] shaped ele-
Electrostatics with scalar potential, U, ments.
3D magnetostatic analysis can be simulated e$-
E "!grad ;,!div [e grad ;]"o; ciently by scalar potential formulations. ANSYS
Magnetostatics with vector potential, A, has 1st order brick, wedge, pyramid and tetrahed-
ron as well as 2nd order tetrahedron elements.
1 These are based on
B"curl A, curl[ curl A]"J
k
E E$cient reduced potential formulation [8] for
or with magnetic scalar potential, <, general problems,
E Di!erential potential formulation [9] to avoid
div k [H !grad <]"0, curl H "J. numerical cancellation when the iron region is
E E
Time dependent problems with A and time integ- singly connected,
rated scalar potential, =: E General potential [10}12] for multiply connec-
ted iron domains.
*
B"curl A, E"! [A#grad =] The Biot}Savart "eld of coils and current carry-
*t ing bars is computed by accurate and e$cient in-
tegration scheme [13].
   
1 * *
curl curl A "! p#e [A#grad =] ANSYS o!ers 1st and 2nd order high frequency
k *t *t
elements over brick, wedge and tetrarahedron

 
* * geometries to study electromagnetic wave pro-
div p#e [A#grad =]"0. pagation.
*t *t
Other ANSYS elements can solve multiphysics
The "nite element method satis"es the potential problems with electromagnetics coupled to electric
equations by minimizing an energy functional over circuit, thermal, structural, #uid dynamics or
small volumes (elements) by Ritz}Galerkin tech- acoustics analyses.
nique [6].

4. Particle tracing
3. ANSYS magnetic elements
Once the electromagnetic "eld is computed, par-
Electrostatic problems can be solved by scalar ticle trajectories can be evaluated by solving the
potential formulation using 2nd order shape func- motion equations:
tions over quadrangle and triangle shaped 2D ele-
m a"F"q (E#v;B)
ments or over brick, wedge, tetrahedron and
pyramid shaped 3D elements. where F is the Lorentz force, m is the mass, q is the
2D magnetostatic and eddy current problems charge, a is the acceleration, and * is the velocity of
can be solved by vector potential formulation using a particle.
2nd order shape functions over quadrangle and The tracing follows from element to element: the
triangle shaped elements. 3D magnetostatic and exit point of an old element becomes the entry
eddy current problems can be solved by nodal point of a new element. Given the entry location
vector potential formulation using 1st order shape and velocity for an element, the exit location and
410 M. Gyimesi et al. / Nuclear Instruments and Methods in Physics Research A 427 (1999) 408}411

velocity can be obtained by integrating the motion 5. Illustrative example

equations.
ANSYS current particle tracing algorithm ex- A TOKAMAK like example has been chosen to
ploits the following assumptions: demonstrate ANSYS magnetic particle trajectory
tracing capabilities. Emphasis has been put on ob-
E no relativistic e!ects,
taining nice graphical output; dimensions and exci-
E electrostatic or magnetostatic analysis,
tations are "ctitious. The investigated region is
E either electric tracing B"0 or magnetic tracing
a torus with 10 m major and 5 m minor radii. The
E"0,
domain is meshed by tetrahedra as shown in Fig. 1.
E Constant E or B within an element,
Fig. 1 also shows the applied Cartesian coordinate
E quadrangle, triangle, hexahedron, tetrahedron,
system. The #ux density is azimuthal; its average
wedge or pyramid element shapes bounded by
value is 1 Tesla.
planar surfaces.
Figs. 2 and 3 show two computed trajectories.
These simpli"cations signi"cantly reduce the The q/m ratio has been chosen as unity in both
computation time of the tracing algorithm be- cases. The entry point is at Cartesian coordinates
cause the trajectory can be given in an analytic (11, 0, ! 2.5) for both cases. In the "rst case shown
form: in Fig. 2, the entry velocity is (0, 0.7, 0.7). In the
second case shown in Fig. 3, the entry velocity is
E parabola in the case of electric tracing,
(0, 0.3, 0.3).
E helix in the case of magnetic tracing.
In both cases, marked helical motion can be
The exit point from an element is the point where observed as it loops around the torus. In the "rst
the particle trajectory meets the bounding planar case the radius of the helix is larger due to the larger
surface of the element. It can be easily computed initial transverse velocity component. The particle
when the trajectory is a parabola. However, when tracing is terminated when the trajectory exits the
the trajectory is a helix, a transcendental equation studied toroidal region. Since the second case is
must be solved. A Newton}Raphson algorithm is
implemented in ANSYS with a carefully selected
starting point to ensure convergence to the correct
solution. This is far from obvious: about 70 sub-
cases are di!erentiated by the algorithm [6]. This
procedure allows accurate particle tracing within
an element up to machine precision. This does not
mean that the tracing is exact since the element "eld
solution may be inexact. However, with mesh re-
"nement, this error can be controlled.
Once a trajectory is computed, any available
physical items can be printed or plotted along the
path. For example, elapsed time, traveled distance,
particle velocity, components, temperature, "eld
components, potential values, #uid velocity, acous-
tic pressure, mechanical strain, etc. Animation is
also available.
The plotted particle traces consist of two
branches: the "rst is a trajectory for a given start
point at a given velocity (forward ballistics); the
second is a trajectory for a particle to hit a given
target location at a given velocity (backward
ballistics). Fig. 1. Torus domain meshed with tetrahedra.
 M. Gyimesi et al. / Nuclear Instruments and Methods in Physics Research A 427 (1999) 408}411 411

Physically this is plausible since the instability in

the circular motion around the torus is caused by
the transverse velocity. Plots for this case are omit-
ted since graphically it is not illustrative.

References

[1] R.H. Wands, H.J. Krebs, Magnetic "eld and force for
superconducting super collider, Fermi Nat. Lab.,
Cryogenics Eng. Conf, July 1989.
[2] M. Modena, P. Sievers, Computation of a pulsed focus
Quadrupole, CERN, Proc. IEEE, 12th Int. Conf. on Mag-
net Technology, Leningrad, USSR, June 1991.
[3] R.L. Myatt, P.H. Titus, 3D coupled electromagnetic, ther-
mal, current di!usion in the "nger joints of the Alcator
C}Mod Toroidal Field Coils, MIT Plasma Fusion Center,
Proc. 17th IEEE/NPSS Symp. on Fusion Engineering, San
Fig. 2. Particle trajectory in the 1st case. Diego, 1997.
[4] D. Lindeman, Charged particle tracking algorithm for use
with ANSYS, Minnesota Mining and Manufacturing,
ANSYS Conf., August 1998.
[5] W.R. Edwards, Submodeling Electromagnetics used in ion
implanters, ion Implants, Varian Assoc, ANSYS Conf.,
August 1998.
[6] ANSYS Theory Reference Manual, Revision 5.5, Section
5.7, to be released.
[7] M. Gyimesi, D. Ostergaard, Non-conforming hexahedral
edge elements for magnetic analysis, COMPUMAG RIO,
1997, IEEE Trans. Magn., 1998.
[8] J. Simkin, C.W. Trowbridge, Int. J. Numer. Meth. Eng. 14
(1975) 423.
[9] I.D. Mayergoyz, M.V.K. Chari, J. D'Angelo, IEEE Trans.
Magn. 23 (6) (1987) 3889.
[10] M. Gyimesi, D. Lavers, IEEE Trans. Magn. 28 (4) (1992)
1924.
[11] M. Gyimesi, D. Lavers, Application of general potential
formulation to "nite elements, in: T. Honma, I. Sebestyen
(Eds.), Applied Electromagnetics in Materials and Com-
putational Technology, Hokkaido University Press, Hok-
kaido, 1992.
Fig. 3. Particle trajectory in the 2nd case. [12] M. Gyimesi, D. Lavers, D. Ostergaard, T. Pawlak, IEEE
Trans. Magn. 29 (1993) 1345.
more stable, the particle travels more helix loops [13] M. Gyimesi, D. Lavers, D. Ostergaard, T. Pawlak,
Biot}Savart for bars/arcs, IEEE Trans. Magn. 29 (6) (1993)
before exiting the studied region. 2389.
By further decreasing the entry velocity compo- [14] M. Gyimesi, D. Ostergaard, Mixed shape non-conforming
nents more and more helical loops can be observed. edge elements, CEFC 98, Tucson, Arizona, June, 1998.