Numerische Mathematik Numerische

https://doi.org/10.1007/s00211-022-01298-9 Mathematik

Large-stepsize integrators for charged-particle dynamics

over multiple time scales

Ernst Hairer1 · Christian Lubich2 · Yanyan Shi2,3,4

Received: 25 January 2021 / Revised: 9 April 2022 / Accepted: 10 May 2022

© The Author(s) 2022

Abstract
The Boris algorithm, a closely related variational integrator and a newly proposed
filtered variational integrator are studied when they are used to numerically integrate
the equations of motion of a charged particle in a mildly non-uniform strong magnetic
field, taking step sizes that are much larger than the period of the Larmor rotations.
For the Boris algorithm and the standard (unfiltered) variational integrator, satisfactory
behaviour is only obtained when the component of the initial velocity orthogonal to
the magnetic field is filtered out. The particle motion shows varying behaviour over
multiple time scales: fast gyrorotation, guiding centre motion, slow perpendicular
drift, near-conservation of the magnetic moment over very long times and conserva-
tion of energy for all times. Using modulated Fourier expansions of the exact and
numerical solutions, it is analysed to which extent this behaviour is reproduced by
the three numerical integrators used with large step sizes that do not resolve the fast
gyrorotations.

Keywords Charged particle · Strong magnetic field · Boris algorithm · Variational

integrator · Filtered variational integrator · Modulated Fourier expansion · Long-term
behaviour

B Christian Lubich
[email protected]
Ernst Hairer
[email protected]
Yanyan Shi
[email protected] ; [email protected]

1 Dept. de Mathématiques, Univ. de Genève, CH-1211, Genève 24, Switzerland

2 Mathematisches Institut, Univ. Tübingen, D-72076 Tübingen, Germany
3 LSEC, Academy of Mathematics and Systems Science, Chinese Academy of Sciences, Beijing
100190, China
4 University of Chinese Academy of Sciences, Beijing 100049, China

123
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
 E. Hairer et al.

Mathematics Subject Classification (2010) 65L05 · 65P10 · 78A35 · 78M25

1 Introduction

The time integration of the equations of motion of charged particles is a basic algo-
rithmic task for particle methods in plasma physics [2]. In this paper we consider
the case of a non-uniform strong magnetic field in the asymptotic scaling known as
maximal ordering [4, 21], with a small parameter ε  1 whose inverse corresponds to
the strength of the magnetic field. The particle motion then shows different behaviour
over multiple time scales:
– fast Larmor rotation over the time scale ε,
– guiding centre motion over the time scale ε0 ,
– slow drift perpendicular to the magnetic field over the time scale ε−1 ,
– near-conservation of the magnetic moment over time scales ε−N with arbitrary
N > 1,
– and energy conservation for all times.
In this paper we are interested in using numerical integrators with step sizes h that
are much larger than the quasi-period 2π ε of the Larmor rotation. We thus have the
two small parameters h and ε, which we will assume to be related by

0 < ε ≤ h 2  1. (1.1)

We study the behaviour of the numerical integrators over the time scales ε0 , ε−1 , and
ε−N for N > 1.
The papers [23, 26] are similarly motivated by the objective to numerically integrate
charged-particle dynamics accurately while stepping over the fast time scale of Larmor
rotation. We are, however, not aware of any rigorous analysis of the error behaviour
of numerical integrators in a large-stepsize regime in the existing literature. With an
emphasis on different aspects, recent papers on numerical methods for charged-particle
dynamics in a strong magnetic field include [5–7, 10–12, 15, 16, 24].
In Section 2 we formulate the equations of motion in the scaling considered here
and illustrate the solution behaviour over various time scales.
In Section 3 we describe the three numerical integrators studied in this paper:
the Boris algorithm [3, 9, 14, 22], a closely related variational integrator [15, 25],
and a newly proposed filtered variational integrator, which only requires a minor
algorithmic modification of the standard variational integrator and can be interpreted
as the standard variational integrator for a Lagrangian with an anisotropically modified
kinetic energy term.
In Section 4 we give modulated Fourier expansions of the exact solution and of the
numerical solutions of the three numerical methods used with step sizes (1.1). The dif-
ferential equations for the dominant modulation functions are the key to understanding
the method behaviour over the times scales ε0 and ε−1 for all three methods. For the
Boris algorithm and the standard (unfiltered) variational integrator, the initial velocity
needs to be modified such that its component perpendicular to the magnetic field is

123
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
 Charged-particle dynamics over multiple time scales

O(ε)-small. The complete modulated Fourier expansion will be used for studying
the long-time near-conservation of the magnetic moment and energy for the filtered
variational integrator.
In Section 5 we obtain O(h 2 ) error bounds uniformly in ε for all three (formally
second-order) numerical methods over the time scale ε0 . This is not an obvious result
for large step sizes (1.1) but here it follows directly from a comparison of the modulated
Fourier expansions of the exact and numerical solutions.
In Section 6 we show that all three methods reproduce the perpendicular drift with an
O(h 2 ) or O(h) error over the time scale ε−1 . This is again obtained via the modulated
Fourier expansions, which also yield an O(ε) approximation to the perpendicular drift
by the solution of a slow differential equation over times O(ε−1 ).
In Section 7 we consider the long-term energy behaviour. For the standard varia-
tional integrator with the modified starting velocity we prove near-conservation of the
total energy up to time O(ε−1 ). For the filtered variational integrator we prove near-
conservation of magnetic moment and energy over times ε−N with arbitrary N > 1
for non-resonant step sizes, using the Lagrangian structure of the modulation system.
Moreover, we show results of numerical experiments for the energy behaviour of the
three methods over long times.
The conclusion of our investigation is that the new filtered variational integrator
with non-resonant large step sizes (1.1) reproduces the characteristic features well
over all time scales, and this is fully explained by our theory. The Boris algorithm and
the standard (unfiltered) variational integrator also work remarkably well for large
stepsizes (1.1) on the time scales ε0 and ε−1 in accordance with our theory, provided
that the initial velocity is modified such that the component perpendicular to the
magnetic field is reduced to size O(ε). With this filtering of the starting velocity, the
long-time energy behaviour of the Boris method and the standard variational integrator
appears to be better in our numerical experiments than we can explain by theory.

2 Multiple time scales in the continuous problem

We study the time integration of the equations of motion of a charged particle in a

strong magnetic field, with position x(t) ∈ R3 and velocity v(t) = ẋ(t) at time t,

ẍ(t) = ẋ(t) × B(x(t)) + E(x(t))

1 (2.1)
with B(x) = B0 + B1 (x) for 0 < ε  1,
ε

where B0 is a fixed vector in R3 of unit norm, |B0 | = 1. The non-constant magnetic

field B1 (x) is assumed to have a known vector potential A1 (x), i.e. B1 (x) = ∇x ×
A1 (x). This gives B(x) = ∇x × A(x) with the vector potential A(x) = − 21 x × B0 /ε +
A1 (x). We always assume that B1 : R3 → R3 and E : R3 → R3 are smooth with
derivatives bounded independently of ε on bounded subsets of R3 . The above scaling
corresponds to what is known as maximal ordering in the literature; see [4, 21].

123
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
 E. Hairer et al.

For the initial position and velocity we always assume boundedness independently
of ε:

|x(0)| ≤ C0 , |ẋ(0)| ≤ C1 . (2.2)

For studying the perpendicular drift, we need further assumptions on B1 and E that
are specified in Section 6. When it comes to studying the long-time energy behaviour,
we further assume that the force field has a scalar potential, E(x) = −∇φ(x). The
total energy is then

H (x, v) = 21 |v|2 + φ(x), (2.3)

which is conserved along every trajectory and is bounded independently of ε under

condition (2.2). We further consider the magnetic moment (rescaled with ε),

1 |v × B(x)|2
I (x, v) = . (2.4)
2ε |B(x)|3

We note that, with v⊥ denoting the velocity component orthogonal to B(x),

1 |v × B0 + O(ε)|2
I (x, v) = = 21 |v⊥ |2 (1 + O(ε)) + O(ε2 ), (2.5)
2 1 + O(ε)

for (x, v) in any region that is bounded independently of ε. The magnetic moment is
an adiabatic invariant: it is conserved up to O(ε) over very long times t ≤ ε−N with
arbitrary N > 1; see e.g. [1, 15, 19, 20].
In Figure 1 we illustrate the solution behaviour on various time scales. We show the
fast Larmor rotation of angular frequency ε−1 and amplitude O(ε) on the time scale ε
and the guiding centre motion on the time scale ε0 in the first picture, and in addition
the slow drift perpendicular to the magnetic field on the time scale ε−1 in the second
picture (here: horizontal drift for the magnetic field in vertical direction). Finally,
the third picture shows the long-time near-conservation of the magnetic moment and
the conservation of energy. Our objective is to understand how the behaviour on the
various time scales can be replicated by numerical methods with large time steps that
do not resolve the fast Larmor rotations.
In Figure 1 we take the electromagnetic fields and the vector and scalar potentials
as
⎛ ⎞ ⎛ ⎞ ⎛ ⎞ ⎛ ⎞
0 x (x − x2 ) x2 1
1⎝ ⎠ ⎝ 1 3 1
B(x) = 0 + x2 (x1 − x3 )⎠ with A(x) = − ⎝−x1 ⎠ + x1 x2 x3 ⎝1⎠ ,
ε 1 x (x − x ) 2 0 1
3 2 1

E(x) = −x with φ(x) = 2 |x| ,

1 2

and the initial values x(0) = (0.3, 0.2, −1.4) and ẋ(0) = (−0.7, 0.08, 0.2) .

123
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
 Charged-particle dynamics over multiple time scales

1
0

-0.5 0
z

z
-1
-1

0.22 0.32 0.2

0.2 0.3 0 0.2
0
0.18 0.28 -0.2 -0.2
y 0.16 0.26 x y -0.4 -0.4 x

2
energy
1.5

1
magnetic moment
0.5

0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
104

Fig. 1 Trajectories of the particle for t ≤ π/2 (top left) and t ≤ 5/ε (top right), both with ε = 10−2 .
Energy and magnetic moment for t ≤ ε −4 (bottom) for ε = 10−1 . The analogous picture for ε = 10−2
instead of ε = 10−1 would show the magnetic moment as a horizontal straight line

Remark 2.1 The results of this paper pertain to the above situation (2.1) – (2.2) of a
mildly non-uniform strong magnetic field. For a strongly non-uniform magnetic field
B(x) =  B(x)/ε with  B(x) independent of ε with non-vanishing gradient, different
numerical phenomena arise, which require a different analysis; cf. [15] for a long-time
analysis of a variational integrator in a small-stepsize regime and [26] for numerical
results for an ingeniously modified Boris method used with large step sizes. The
modification of the force field proposed in [26] is not needed for the mildly non-
uniform situation (2.1) studied in this paper, but it is essential in the case of strongly
non-uniform strong magnetic fields.

3 Three numerical integrators

We now describe the three numerical integrators for (2.1) that are studied in this paper
when applied with large step sizes h ε.

3.1 Boris algorithm

The Boris method, introduced in [3], is the standard integrator for particle-in-cell codes
for plasma simulation; see e.g. [2, 8]. Given the position and velocity approximation

123
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
 E. Hairer et al.

(x n , v n−1/2 ), the algorithm computes (x n+1 , v n+1/2 ) as follows, with B n = B(x n )

and E n = E(x n ):

1 1
n− 2 h n
v+ = v n− 2 + E
2
1 1
n+ 2 n− 2 h n+ 21 n−
1
v− − v+ = (v− + v+ 2 ) × B n
2
1 1
n+ h
v n+ 2 = v− 2 + E n
2
1
x n+1 = x n + hv n+ 2 , (3.1)

where the starting value is chosen as v 1/2 = v 0 + h2 v 0 × B 0 + h2 E 0 .

The method has the equivalent two-step formulation

x n+1 − 2x n + x n−1 x n+1 − x n−1

2
= × Bn + E n , (3.2)
h 2h

with the velocity approximation

x n+1 − x n−1
vn = . (3.3)
2h

It is known from [9] that the Boris algorithm is not symplectic unless B is a constant
magnetic field. The energy behaviour over long times, which is not fully satisfactory,
has been studied in [14] for step sizes with h|B|  1, which in our case (2.1) would
read h  ε in contrast to (1.1).
In the large-stepsize regime (1.1) the starting velocity needs to be modified. Instead
of setting v 0 equal to the initial data ẋ(0) we choose v 0 such that its component v⊥ 0

orthogonal to the magnetic field is O(ε)-small. We propose to take v = v + v⊥ with

0 0 0

 
v 0 = P0 ẋ(0), v⊥
0
= ε v 0 × B1 (x 0 ) + E(x 0 ) × B0 , (3.4)

where P0 = B0 B0 is the orthogonal projection in the direction of B0 . (This choice of

v⊥0 will be explained in Section 4 right after Theorem 4.2.) Without such a modification

of the starting velocity, the Boris algorithm shows highly oscillatory behaviour with a
large amplitude proportional to (h 2 /ε)|v⊥ 0 |; cf. [23].

Since the Boris method with large step size (1.1) and the proposed filtering of
the initial velocity will give an approximation to the guiding centre rather than to
the oscillatory trajectory, it is reasonable to take the guiding centre approximation
x(0) + ε ẋ(0) × B0 instead of x(0) as the starting position x 0 .
We note that while the one-step map (x n , v n−1/2 ) → (x n+1 , v n+1/2 ) is volume-
preserving [22], the starting-value map (x(0), ẋ(0)) → (x 0 , v 0 ) and also the map
(x 0 , v 0 ) → (x 1 , v 1/2 ) are far from volume-preserving for step sizes (1.1).

123
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
 Charged-particle dynamics over multiple time scales

3.2 Standard variational integrator

The variational integrator to be studied here is constructed in the same way as is

done in the interpretation of the Störmer–Verlet method as a variational integrator; see
e.g. [17, Chap. VI, Example 6.2] and [25]. The integral of the Lagrangian L(x, v) =

2 |v| + A(x) v − φ(x) over a time step is approximated in two steps: the path x(t)
1 2

of positions is approximated by the linear interpolant of the endpoint positions, and

the integral is approximated by the trapezoidal rule. This approximation to the action
integral is then extremized. With the derivative matrix A (x) = (∂ j Ai (x))i,3
j=1 and

its transpose A (x) , this variational integrator becomes the following:

x n+1 − 2x n + x n−1
=
h2
x n+1 − x n−1 A(x n+1 ) − A(x n−1 )
A (x n ) − + En, (3.5)
2h 2h

or equivalently, written as a perturbation to the Boris algorithm and using that v ×

B(x) = A (x) v − A (x)v,

x n+1 − 2x n + x n−1 x n+1 − x n−1

= × Bn + E n
h2 2h
x n+1 − x n−1 A(x n+1 ) − A(x n−1 )
+ A (x n ) − . (3.6)
2h 2h

We note that the correction to the Boris method as given in the second line vanishes
for linear A(x). In the situation of the magnetic field of (2.1), we can therefore replace
A by A1 in (3.6). The variational integrator coincides with the Boris algorithm in the
case of a constant magnetic field (B1 ≡ 0).
This method is again complemented with the velocity approximation (3.3). It can
be given a one-step formulation similar to the Boris algorithm, with the correction
term of (3.6) added in the second line of (3.1). It is, however, an implicit method,
because the vector potential A is evaluated at the new position x n+1 .
For the case of a strong magnetic field and for step sizes with h|B| ≤ Const., the
variational integrator has been shown to have excellent near-preservation of energy
and magnetic moment over very long times [15].
For large step sizes (1.1), the variational integrator requires the same modification
of the starting velocity as the Boris method in order to suppress high oscillations of
large amplitude in the numerical solution.

123
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
 E. Hairer et al.

3.3 Filtered variational integrator

As a new method to be studied here, we propose the following modification of the

variational integrator: with the filter functions

tanh(ζ /2) 1 ζ
ψ(ζ ) = tanch(ζ /2) = , ϕ(ζ ) = = ,
ζ /2 sinch(ζ ) sinh(ζ )

which are even functions and take the value 1 at ζ = 0, and with the skew-symmetric
matrix B0 defined by − B0 v = v × B0 for all v ∈ R3 , we define the filter matrices

h h
=ψ − B0 = I + 1 − tanc B02 ,
ε 2ε
h h −1 2
= ϕ − B0 = I + 1 − sinc B0 ,
ε ε

where the rightmost expressions are obtained from a Rodriguez formula; see [16,
Appendix]. Here, tanc(ξ ) = tan(ξ )/ξ and sinc(ξ ) = sin(ξ )/ξ . The filter matrices
and are symmetric and act as the identity on vectors in the direction of B0 .
We put the filter matrix in front of the right-hand side of (3.5):

x n+1 − 2x n + x n−1
=
h2
(x n+1 − x n−1 ) A(x n+1 ) − A(x n−1 )
A (x n ) − + En . (3.7)
2h 2h
This is combined with the velocity approximation

x n+1 − x n−1 h −1
vn = + ε 1 − sinc E n × B0 . (3.8)
2h ε

This filtered variational integrator coincides with the filtered Boris algorithm of [16]
for the special case of a constant magnetic field B(x) = B0 /ε. If additionally also E
is constant, then this method yields the exact position and velocity, as was shown for
the filtered Boris algorithm.
For stepsizes h with tan(h/(2ε)) ≥ c > 0, the filter matrix is positive definite. The
above integrator can then be interpreted as a variational integrator corresponding to a
discrete Lagrangian where the kinetic energy term has the modified mass matrix −1 .
Its eigenvalues corresponding to the eigenvectors orthogonal to B0 are 1/ tanc(h/(2ε))
and are thus proportional to h/ε, which is greater than h −1 under condition (1.1). The
discrete Lagrangian reads

h n+1/2  −1 n+1/2
L h (x n , x n+1 ) = (v ) v
2
A(x n ) + A(x n+1 ) n+1/2 φ(x n ) + φ(x n+1 )
+h v −h ,
2 2

123
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
 Charged-particle dynamics over multiple time scales

where v n+1/2 = (x n+1 − x n )/h. The standard (unfiltered) variational integrator has
the same discrete Lagrangian except for the identity matrix in place of the matrix −1 .
The filtered variational integrator for (2.1) can be written and implemented as the
following implicit one-step method:

1
n− 2 1 h
v+ = v n− 2 + En
2
1 1
n+ 2 n− 21 1 n+ 21 n−
v− − v+ =h (v− + v+ 2 ) × B n
2
1 n+ 1 n−
1
A1 (x n+1 ) − A1 (x n−1 )
+ A1 (x n ) (v− 2 + v+ 2 ) −
2 2h
1 1
n+ 2 h
v n+ 2 = v− + En
2
1
x n+1 = x n + hv n+ 2 .

This can be solved by a fixed-point iteration for x n+1 , where a good starting iterate
is obtained from a Boris step. The first velocity is chosen as follows: we set v 1/2 =
v̄ + 21 δv with h v̄ = 21 (x 1 − x −1 ) and h δv = x 1 − 2x 0 + x −1 , where in view of (3.8)
for n = 0,

h −1
−1
v̄ = v 0 − ε 1 − sinc E n × B0
ε

and δv is implicitly determined (and computed via fixed-point iteration) from (3.7)
with n = 0, i.e. from the equation

A1 (x 1 ) − A1 (x −1 )
δv = h v̄ × B(x 0 ) + A1 (x 0 )v̄ − + E(x 0 ) ,
2h

where x ±1 = x 0 ± h v̄ + 21 h δv.
In contrast to the Boris algorithm and the unfiltered variational integrator, we here
take the original initial data v 0 = ẋ(0) and x 0 = x(0).

4 Modulated Fourier expansions

We give modulated Fourier expansions of the exact solution of (2.1) and the numerical
solutions of the three integrators for large step sizes h 2 ≥ c ε (in the following we set
the irrelevant positive constant c equal to 1 for simplicity). Analogous expansions for
step sizes h ≤ Cε were previously given in [13, 15, 16]; see also [17, Ch. XIII]. In
particular, we explicitly state the differential equations for the dominant modulation
functions up to O(ε2 ) for the exact solution, and up to O(h 2 ) for the numerical
solutions.

123
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
 E. Hairer et al.

4.1 Modulated Fourier expansion of the exact motion

We write the solution of (2.1) as


x(t) ≈ z k (t) eikt/ε (4.1)
k

with coefficient functions z k (t) for which all time derivatives are bounded indepen-
dently of ε.
We diagonalize the linear map v → v × B0 , which has eigenvalues λ1 = i, λ0 = 0
and λ−1 = −i (recall the normalization |B0 | = 1). The normalized eigenvectors are
denoted v1 , v0 = B0 , v−1 = v 1 . We let P j = v j v ∗j be the orthogonal projections onto
the eigenspaces. We write the coefficient functions of (4.1) in the basis (v j ),

z k = z 1k + z 0k + z −1
k
, z kj (t) = P j z k (t).

The following theorem is a variant of Theorems 4.1 in [15, 16], proved by the same
arguments but in a technically simplified way, since here we have the constant fre-
quency 1/ε and constant projections P j , as opposed to the state-dependent frequency
and projections in [15, 16].

Theorem 4.1 Let x(t) be a solution of (2.1) with an initial velocity bounded indepen-
dently of ε (|ẋ(0)| ≤ C1 ), which stays in a compact set K for 0 ≤ t ≤ T (with K
and T independent of ε). For an arbitrary truncation index N ≥ 1 we then have an
expansion


x(t) = z k (t) eikt/ε + R N (t)
|k|≤N

with the following properties:

(a) The modulation functions z k together with their derivatives (up to order N ) are
−1
bounded as z 0j = O(1) for j ∈ {−1, 0, 1}, z 11 = O(ε), z −1 = O(ε), and for the
remaining (k, j) with |k| ≤ N ,

z kj = O(ε|k|+1 ).

−k
They are unique up to O(ε N ) and are chosen to satisfy z − j = z j . Moreover, ż ±1
k 0
0 = O(ε).
together with its derivatives is bounded as ż ±1
(b) The remainder term and its derivative are bounded by

R N (t) = O(t 2 ε N ), Ṙ N (t) = O(tε N ) for 0 ≤ t ≤ T .

123
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
 Charged-particle dynamics over multiple time scales

0 , z 1 , z −1 satisfy the differential equations

(c) The functions z 00 , z ±1 1 −1

  i
z̈ 00 = P0 ż 0 × B1 (z 0 ) + E(z 0 ) + 2P0 Re z 1 × B1 (z 0 )z −1 + O(ε2 ),
ε
 
0
ż ±1 = ±iε P±1 ż 0 × B1 (z 0 ) + E(z 0 ) + O(ε2 ),
±1  ±1 
ż ±1 = P±1 z ±1 × B1 (z 0 ) + O(ε2 ).

All other modulation functions z kj are given by algebraic expressions depending

−1
on z 0 , ż 00 , z 11 , z −1 .
(d) Initial values for the differential equations of item (c) are given by

z 0 (0) = x(0) + ε ẋ(0) × B0 + O(ε2 ),

 
ż 00 (0) = P0 ẋ(0) − ε P0 (ẋ(0) × B0 ) × B1 (x(0)) + O(ε2 ),
±1
z ±1 (0) = ∓iε P±1 ẋ(0) + O(ε2 ).

The constants symbolized by the O-notation are independent of ε and t with 0 ≤ t ≤ T ,

but depend on N , on the velocity bound M, on bounds of derivatives of B1 and E on
the compact set K , and on T .

4.2 Resonant modulated Fourier expansion of the Boris algorithm and the
standard variational integrator for h2 ≥ "

When the Boris method is applied to the linear differential equation ẍ = ẋ × B0 /ε

with |B0 | = 1 (that is, B1 and E are not present in (2.1)), then diagonalization of B0
shows that x n is a linear combination (with coefficients independent of n) of terms 1,
nh and e±inhω , where

h
hω = 2 arctan .
2ε

If h/ε is large, then hω is close to π . In particular, if h 2 ≥ ε, then hω = π − γ h with

γ > 0 bounded independently of h and ε with h 2 ≥ ε, and so e±inhω = (−1)n e∓inhγ ,
where we note that e∓itγ is a smooth function of t all of whose derivatives are bounded
independently of ε and h. In the general case of (2.1), we have the following result.

Theorem 4.2 Let x n be the numerical solution obtained by applying either the Boris
algorithm or the variational integrator to (2.1) with a stepsize h satisfying

h 2 ≥ ε. (4.2)

We assume that the starting velocity v 0 is bounded independently of ε and h and that
its component orthogonal to B0 , i.e. v⊥
0 = (I − P )v 0 , is small:
0

|v⊥
0
| ≤ c1 ε. (4.3)

123
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
 E. Hairer et al.

We further assume that the numerical solution x n stays in a compact set K for 0 ≤
nh ≤ T (with K and T independent of ε and h). For an arbitrary truncation index
N ≥ 2, we then have a decomposition

x n = y(t) + (−1)n z(t) + R N (t), t = nh, (4.4)

with the following properties:

(a) The functions y(t) and z(t) together with their derivatives (up to order N ) are
bounded as y = O(1), z = O(h 2 ). They are unique up to O(h N ). Moreover, we
have ẏ × B0 = O(ε) and z · B0 = O(h 4 ).
(b) The remainder term is bounded by

R N (t) = O(t 2 h N ) for 0 ≤ t ≤ T .

(c) The functions y j = P j y ( j = 0, ±1) and z ±1 = P±1 z satisfy the differential

equations
 
ÿ0 = P0 ẏ × B1 (y) + E(y) + O(h 2 ),
 
ẏ±1 = ±iε P±1 ẏ × B1 (y) + E(y) + O(εh 2 ),
ε
ż ±1 = ∓4i 2 z ±1 + O(εh 2 ).
h
The function z 0 = P0 z is given by an algebraic expression depending on y, ẏ0 and
z ±1 .
(d) Initial values for the differential equations of item (c) are given by

y(0) = x 0 + O(h 2 ),
ẏ0 (0) = P0 (x 0 )v 0 + O(h 2 ),
ih 2  
z ±1 (0) = ∓ P±1 v 0 ∓ iε P0 v 0 × B1 (x 0 ) + E(x 0 ) + O(h 4 ).
4ε
The constants symbolized by the O-notation are independent of ε, h and n with 0 ≤
nh ≤ T , but depend on the velocity bound, on bounds of derivatives of B1 and E on
the compact set K , and on T .

We note that the differential equations for y agree with those for z 0 of the exact
±1
solution up to O(h 2 ). The differential equations for z ±1 and for z ±1 of the exact
solution differ, but we still have

d ∗
 
|z ±1 |2 = 2 Re z ±1 ż ±1 = O ε|z ±1 |2 + O(εh N ) = O(εh 4 ),
dt
which is to be compared with

d ±1 2 ±1 ∗ ±1
|z | = 2 Re (z ±1 ) ż ±1 = O(ε3 ).
dt ±1

123
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
 Charged-particle dynamics over multiple time scales

To obtain an O(h 2 ) approximation to the guiding centre z 0 (t) over bounded time
intervals, we run the Boris algorithm with the modified initial velocity v 0 = P0 ẋ(0)
instead of ẋ(0), or even better, determine P±1 v 0 such that z ±1 (0) = O(h 4 ), which
holds true with the proposed choice (3.4).

Proof The bounds of parts (a) and (b) are proved as in previous proofs of modulated
Fourier expansions; see e.g. [15] and [17, Ch. XIII]. Here we just show (c) and (d),
assuming that the bounds of (a) and (b) are already available.
To derive the differential equations of (c), we insert (4.4) into the two-step formu-
lation of the numerical method, expand y(t ± h) and z(t ± h) into Taylor series at t,
expand the nonlinear functions B1 and E at y(t) and separate the terms without and
with the factor (−1)n . This gives us the equations

B0
ÿ + O(h 2 ) = ẏ × + ẏ × B1 (y) + E(y) + O(h 2 )
ε
4 B0
− 2 z − z̈ + O(h 2 ) = −ż × + ż × B1 (y) + ẏ × B1 (y)z + E (y)z + O(h 2 ).
h ε
In the equation for z we note that also z̈ and the last three terms on the right-hand side
are O(h 2 ) as z and its derivatives are O(h 2 ), and the indicated O(h 2 ) terms are then
actually O(h 4 ).
Taking the projection P0 on both sides of the differential equation for y yields
the stated second-order differential equation for y0 on noting that P0 ( ẏ × B0 ) = 0.
Moreover, since P±1 ( ẏ × B0 ) = ±i ẏ±1 , we obtain

i  
∓ ẏ±1 = − ÿ±1 + P±1 ẏ × B1 (y) + E(y) + O(h 2 ).
ε

Differentiating this equation and multiplying with iε yields ÿ±1 = O(ε), which is
O(h 2 ) under condition (4.2). So we obtain the stated first-order differential equation
for y±1 .
Taking the projection P0 in the above equation for z yields − h42 z 0 = O(h 2 ), and
hence z 0 = O(h 4 ). Taking the projections P±1 yields

4 i
− z ±1 = ∓ ż ±1 + O(h 2 ),
h 2 ε
which can be rearranged into the stated differential equation for z ±1 .
In view of (4.4) for n = 0 and z(0) = O(h 2 ), we have y(0) = x 0 + O(h 2 ). Since
we obtain by inserting (4.4) for n = −1, 1

x 1 − x −1
v0 = = ẏ(0) − ż(0) + O(h 2 ),
2h

we obtain the stated expression for ẏ0 (0) on taking the projection P0 . Taking the
projections P±1 and using the differential equations for y±1 and z ±1 , we arrive at the
stated expression for z ±1 (0). 


123
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
 E. Hairer et al.

4.3 Non-resonant modulated Fourier expansion of the filtered variational

integrator for h2 ≥ "

As the filtered integrator is exact for the linear equation ẍ = ẋ × B0 /ε, it has the same
high frequency 1/ε. When this integrator is applied to (2.1), it has a modulated Fourier
equation that is very similar to that of the exact solution given in Theorem 4.1.

Theorem 4.3 Let x n be a solution of the filtered variational integrator applied to (2.1)
with a stepsize h satisfying

h2 ≥ ε (4.5)

and, for some N ≥ 1, the non-resonance conditions

 kh   kh 
 
sin  ≥ c > 0, cos ≥c>0 (k = 1, . . . , N ),
2ε 2ε
  (4.6)
 kh h 
tan − tan ≥c>0 (k = 2, . . . , N ),
2ε 2ε

where c is a positive constant. We assume that the initial velocity v 0 = ẋ(0) is bounded
independently of ε and h, as in (2.2). We further assume that the numerical solution
x n stays in a compact set K for 0 ≤ nh ≤ T (with K and T independent of ε and h).
We then have an expansion, at t = nh,

xn = z k (t)eikt/ε + R N (t) (4.7)
|k|≤N

with the following properties:

(a) The bounds of parts (a) of Theorem 4.1 for the modulation functions are valid also
in this case, except z 0k = O(hε|k| ) for |k| ≥ 1.
(b) The remainder at t = nh is bounded, for arbitrary M > 1, by

P0 R N (t) = O(t 2 h M ) + O(t 2 ε N ), P±1 R N (t) = O(t 2 εh M−1 ) + O(t 2 ε N ).

0 , z 1 , z −1 satisfy the differential equations

(c) The functions z 00 , z ±1 1 −1

 
z̈ 00 = P0 ż 0 × B1 (z 0 ) + E(z 0 ) + O(h 2 ),
 
0
ż ±1 = ±iε P±1 ż 0 × B1 (z 0 ) + E(z 0 ) + O(εh),
±1 ε h  ±1 
ż ±1 = sin P±1 z ±1 × B1 (z 0 ) + O(ε2 ).
h ε

All other modulation functions z kj are given by algebraic expressions depending

−1
on z 0 , ż 00 , z 11 , z −1 .

123
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
 Charged-particle dynamics over multiple time scales

(d) Initial values for the differential equations of item (c) are given by

z 0 (0) = x 0 + O(h 2 ),
ż 00 (0) = P0 v 0 + O(h 2 ),
±1
z ±1 (0) = ∓iε P±1 v 0 + O(εh).

The constants symbolized by the O-notation are independent of ε and t with 0 ≤ t ≤ T ,

but depend on M and N , on the velocity bound (2.2), on bounds of derivatives of B1
and E on the compact set K , and on T .

Proof Parts (a) and (b) are again proved as in previous proofs of modulated Fourier
expansions; see e.g. [15] and [17, Ch. XIII]. Here we only show (c) and (d), assuming
that the bounds of (a) and (b) are already available.
To derive the differential equations of (c), we insert (4.7) into the two-step for-
mulation (3.7) (or equivalently (3.6) with an extra factor on the right-hand side),
expand z k (t ± h) into a Taylor series at t, use Lemma 5.1 of [16] to expand the first
and second-order difference quotients for z k (t)eikt/ε for 0 < |k| ≤ N , and expand B1
and E at z 0 (t). We then separate the terms multiplying eikt/ε for |k| ≤ N . Moreover,
we consider the components z kj = P j z k for j = 0, ±1.
For k = 0, j = 0 we obtain

z̈ 00 + O(h 2 ) = P0 (ż 0 + O(h 2 )) × B1 (z 0 ) + E(z 0 ) + O(ε2 /h) ,

where the O(h 2 ) terms result from the Taylor expansions of the second and first order
difference quotients of z 0 , and the (smaller) O(ε2 /h) term results from the Taylor
expansion of B1 and E at z 0 and the bound z k = O(ε|k| ). This yields the first equation
of (c).
For k = 0, j = 1 we obtain

2ε h i 0
z̈ 10 + O(εh 2 ) = tan (ż + O(h 2 ))
h 2ε ε 1

+P1 (ż 0 + O(h 2 )) × B1 (z 0 ) + E(z 0 ) + O(ε2 /h) .

We solve this equation for ż 10 , which appears in the dominant term with a factor h −1 ,
and recall that | tan(h/(2ε))| ≥ c > 0 by the non-resonance condition (4.6). Using
that z̈ 10 and its higher derivatives are O(ε) by part (a), this yields

ż 10 = iε P1 ż 0 × B1 (z 0 ) + E(z 0 ) + O(εh),

which is the differential equation for z 10 stated in (c). The case j = −1 is obtained by
taking complex conjugates.

123
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
 E. Hairer et al.

For k = 1, j = 1 we find for y11 (t) = z 11 (t)eit/ε , using Lemma 5.1 of [16] and the
O(ε) bound for z 11 and its derivatives of part (a),

y11 (t + h) − 2y11 (t) + y11 (t − h)

h2
4 h 1 2i h 1
= eit/ε − 2 sin2 z (t) + sin ż (t) + O(ε)
h 2ε 1 h ε 1

and

y11 (t + h) − y11 (t − h) i h 1 h 1
= eit/ε sin z (t) + cos ż (t) + O(εh) .
2h h ε 1 ε 1

The latter formula yields

h y11 (t + h) − y11 (t − h) B0
P1 tanch B0 ×
2ε 2h ε
2ε h i it/ε i h 1 h 1
= tan e sin z (t) + cos ż (t) + O(εh)
h 2ε ε h ε 1 ε 1
4 h 1 2i h h 1
= eit/ε − 2 sin2 z (t) + tan cos ż (t) + O(ε)
h 2ε 1 h 2ε ε 1

and similarly

h y11 (t + h) − y11 (t − h)
P1 tanch B0 × B1 (z 0 (t))
2ε 2h
2ε h i h 1
= eit/ε tan P1 sin z (t) × B1 (z 0 (t)) + O(ε) .
h 2ε h ε 1
We insert the modulated Fourier expansion (4.7) into the two-step formulation of the
filtered variational integrator, which we wirte as (3.6) with the extra filter factor
on the right-hand side, and we collect the terms with the factor eit/ε . The dominant
terms after projecting with P1 are given by the above formulas. The remaining term
on the right-hand side (as in the second line of (3.6) but multiplied with ) is of size
(2ε/h) tan(h/2ε) · O(ε + h 2 ) = O(hε) for h 2 ≥ ε under condition (4.6). We thus
obtain
4 h 1 2i h 1
− sin2 z (t) + sin ż (t) + O(ε)
h2 2ε 1 h ε 1
4 h 1 2i h h 1
= − 2 sin2 z (t) + tan cos ż (t) + O(ε)
h 2ε 1 h 2ε ε 1
2ε h i h 1
+ tan P1 sin z (t) × B1 (z 0 (t)) + O(ε) .
h 2ε h ε 1
Here the dominant terms are the first terms on the left-hand and the right-hand sides,
which are the same and thus cancel. The dominant terms then become the terms

123
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
 Charged-particle dynamics over multiple time scales

containing the factor (2 i/h)ż 11 (t). Since a calculation shows that we have, with ξ =
h/(2ε) for short,

sin(2ξ ) − tan(ξ ) cos(2ξ ) = (tan(2ξ ) − tan(ξ )) cos(2ξ ) = tan(ξ ),

the above equation yields the differential equation for z 11 as stated in part (c) of the
−1
theorem. The result for z −1 is obtained by taking complex conjugates.
The formulae for the initial values are obtained by the same arguments as in the
proof of Theorem 4.2, using here that (x 1 − x −1 )/(2h) is related to v 0 by (3.8) for
n = 0. 


5 Time scale "0 : error bounds for position and parallel velocity

Comparing the modulated Fourier expansions of the numerical solution with that of
the exact solution, we obtain the following error bounds from Theorems 4.1–4.3.

Theorem 5.1 Consider applying the Boris method, the variational integrator and the
filtered variational integrator to (2.1) over a time interval 0 ≤ t ≤ T (with T inde-
pendent of ε) using a stepsize h with

h 2 ≥ ε.

Suppose that the conditions of Theorem 4.2 are satisfied in the case of the Boris method
and the variational integrator (in particular, small perpendicular starting velocity:
v⊥0 = O(ε)), and that the conditions of Theorem 4.3 are satisfied in the case of the

filtered variational integrator (in particular, the non-resonance conditions (4.6) and
bounded initial velocity (2.2)). For each of the three methods, the errors in position x
and parallel velocity v = P0 v at time tn = nh ≤ T are then bounded by

|x n − x(tn )| ≤ Ch 2 , |v n − v (tn )| ≤ Ch 2 (tn ≤ T ),

where C is independent of ε, h and n with h 2 ≥ ε and nh ≤ T (but depends on T ).

Proof The result is obtained by representing the exact and numerical solutions by
their modulated Fourier expansions and using the bounds and differential equations
of the modulation functions as given in Theorems 4.1–4.3. Note that the differential
equations of the dominating modulation functions for the three methods and for the
exact solution coincide up to defects of size O(h 2 ), which lead to an O(h 2 ) error in the
positions. Inserting the modulated Fourier expansion of the numerical solution into
the formula for the approximate velocity v n for each method and comparing with the
time-differentiated modulated Fourier expansion of the exact solution then yields the
O(h 2 ) error bound for the parallel velocity. 


Remark 5.2 For h 2 ∼ ε, the above error bounds are thus O(ε). For all three methods,
the error bounds remain in general O(ε) also for smaller stepsizes h ∼ ε. This can be

123
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
 E. Hairer et al.

104
3
104 10
102

101
10 2 100

10-1
10-2
100
10-3 10-4

10-2 10-5
10-5 10-4 10-3 10-2 10-5 10-4 10-3 10-2 10-5 10-4 10-3 10-2

10-1

10-1
10-2 100

10-2

10-3
10-3

10-4
10-4

10-5 10-5 10-1

10-5 10-4 10-3 10-2 10-5 10-4 10-3 10-2 10-5 10-4 10-3 10-2

10-1

10 -1 100

10-2
10-2

10-3
10-3 10-2

10-4
10-4

10-5 10-4
10-5 10-4 10-3 10-2 10-5 10-4 10-3 10-2 10-5 10-4 10-3 10-2

Fig. 2 Global error vs. ε (ε = 1/2 j , j = 6, · · · 17) with different h for the Boris algorithm with starting
values x(0), v(0) (top row), with modified starting values (3.4) (centre row), and for the filtered variational
integrator with starting values x(0), v(0) (bottom row)

shown by comparing the modulated Fourier expansions for such stepsizes, as given
in [15] for the standard variational integrator. The filtered Boris method of [16], used
with h ∼ ε, has an O(ε2 ) error in the position and the parallel velocity, and an O(ε)
error in the perpendicular velocity.
Numerical experiment. For the example of Section 2, Figure 2 shows the relative
errors in x, v and v⊥ at time t = π/2 versus ε for various step sizes h for three
numerical approaches:

123
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
 Charged-particle dynamics over multiple time scales

(i) in the top row for the Boris algorithm with the original initial data as starting
values,
(ii) in the centre row for the Boris algorithm with modified starting values (3.4),
(iii) in the bottom row for the filtered variational integrator with the original initial data
as starting values.
For any step size h, the errors in x and v increase roughly proportionally to h 2 /ε
when ε → 0 in case (i), whereas in cases (ii) and (iii) the errors tend to a constant
error level proportional to h 2 .

6 Time scale "−1 : perpendicular drift

6.1 Perpendicular drift of the exact motion

We let P = P0 = B0 B0 be the orthogonal projection onto the span of B0 , and

P⊥ = P1 + P−1 = I − P the orthogonal projection onto the plane orthogonal to B0 .
We decompose x ∈ R3 as

x = x + x⊥ with x = P x, x⊥ = P⊥ x.

We assume that (with slight abuse of notation for B1 )

B1 (x) = B1 (x⊥ ) + ε B2 (x), E(x) = E ⊥ (x⊥ ) + E (x) + εE 2 (x), (6.1)

with E ⊥ · B0 = 0 and E × B0 = 0, and where the functions B1 , B2 and E ⊥ , E , E 2

on the right-hand side and all their derivatives are bounded independently of ε. We
thus only allow a weak dependence of the magnetic field and the perpendicular electric
field on x . We then have the following result.

Theorem 6.1 Let x(t) be a solution of (2.1) with (6.1), with an initial velocity bounded
independently of ε (|ẋ(0)| ≤ M), which stays in a compact set K for 0 ≤ t ≤ c ε−1
(with K and c independent of ε). Then, the solution y⊥ (t) of the initial-value problem
for the slow differential equation

ẏ⊥ (t) = εE ⊥ (y⊥ (t)) × B0 , y⊥ (0) = x⊥ (0), (6.2)

remains O(ε)-close to the perpendicular component of x(t) over times O(ε−1 ):

|x⊥ (t) − y⊥ (t)| ≤ Cε, 0 ≤ t ≤ c/ε. (6.3)

The constant C is independent of ε and t with 0 ≤ t ≤ c/ε, but depends on the initial
velocity bound M, on bounds of derivatives of B1 and E on the compact set K , and
on c.

Remark 6.2 It is well known in the physical literature (going back to [20, Eq. (13)])
that the perpendicular velocity is largely determined by the E × B term, as is justified

123
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
 E. Hairer et al.

by averaging techniques; see also, e.g., [10, Eq. (6)] in the numerical literature. An
O(ε) bound over times O(ε−1 ) as in (6.3) was recently proved in [12] in the more
restricted setting of a constant magnetic field (B1 ≡ 0) and an electric field with
E ≡ 0.

Proof The proof uses the modulated Fourier expansion of Theorem 4.1, in particular
0 and z ±1 in part (c), and the familiar argument of
the differential equations for z ±1 ±1
Lady Windermere’s fan [18]. We structure the proof into four parts (a)–(d).
(a) Over the (short) time interval 0 ≤ t ≤ 1, Theorem 4.1 yields that
−1
x⊥ (t) = z ⊥
0
(t) + z 11 (t)eit/ε + z −1 (t)e−it/ε + O(ε2 ),

0 (t) = z 0 (t) + z 0 (t) and z ±1 (t) satisfy the differential equations

where z ⊥ 1 −1 ±1

0
ż ⊥ = ε (ż 0 + ż ⊥
0
) × B1 (z ⊥
0
) + E ⊥ (z ⊥
0
) × B0 + O(ε2 ),
 
ż 11 = P1 z 11 × B1 (z ⊥0
) + O(ε2 ),

−1  
and z −1 = z 11 . We note that ż 0 = ż 00 = ẋ + O(ε), because we have dtd z 01 eit/ε =
(iz 01 /ε + ż 01 )eit/ε = O(ε). Moreover, the implicit differential equation for z ⊥
0 can be
0
solved for ż ⊥ to yield

0
ż ⊥ = ε ż 0 × B1 (z ⊥
0
) + E ⊥ (z ⊥
0
) × B0 + O(ε2 ).

(b) On every time interval n ≤ t ≤ n + 1 (with n ≤ c/ε) we can do the same and,
[n] 0 on this interval and by z [n] the function z 1 , we have
denoting by y⊥ the function z ⊥ 1 1

[n]  
x⊥ (t) = y⊥ (t) + 2 Re z 1[n] eit/ε + O(ε2 ), n ≤ t ≤ n + 1,

[n]
where y⊥ and z 1[n] solve the initial value problems

[n] [n] [n]

ẏ⊥ = ε ẋ × B1 (y⊥ ) + E ⊥ (y⊥ ) × B0 ,
[n]  
y⊥ (n) = x⊥ (n) − 2 Re z 1[n] (n)ein/ε ,

and
 [n] 
ż 1[n] = P1 z 1[n] × B1 (y⊥ ) ,
z 1[n] (n) = −iε P1 ẋ(n).

We consider these initial value problems on the time interval n ≤ t ≤ c/ε. By

the error bound of the modulated Fourier expansion on the interval [n, n + 1] (in
particular at t = n + 1) as stated by Theorem 4.1 (b), by the essential uniqueness
of the coefficient functions of the modulated Fourier expansion and their bounds as

123
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
 Charged-particle dynamics over multiple time scales

stated by Theorem 4.1 (a), and by the approximation of the modulation functions z ⊥0
[n] [n]
and z 11 on the interval [n, n + 1] by the functions y⊥ and z 1 defined above, which
has an O(ε2 ) error because of Theorem 4.1 (c) and (d), we obtain
[n] [n+1]
y⊥ (n + 1) = y⊥ (n + 1) + O(ε2 ),
z 1[n] (n + 1) = z 1[n+1] (n + 1) + O(ε2 ).

In view of the factor ε in front of the right-hand side of the differential equations for
[n+1] [n]
y⊥ and y⊥ , this estimate implies that

[n+1] [n]
y⊥ (t) − y⊥ (t) = O(ε2 ), n + 1 ≤ t ≤ c/ε.

Moreover, taking the inner product of the differential equation for z 1[n] with z 1[n] shows
that

d [n] 2 
|z 1 | = 2 Re z 1[n] ż 1[n] = 0,
dt

and hence

|z 1[n] (t)| = |z 1[n] (n)|, n ≤ t ≤ c/ε.

[0]
(c) Next we study the difference between y⊥ (t) and y⊥ (t) of (6.2). We have
 t
[0] [0] [0]
y⊥ (t) − y⊥ (t) = y⊥ (0) − y⊥ (0) + ε E ⊥ (y⊥ (s)) − E ⊥ (y⊥ (s)) × B0 ds
0
 t
 [0] 
+ ε ẋ (s) × B1 (y⊥ (s)) × B0 ds.
0

The difference of the initial values is O(ε), and the last integral term is bounded using
partial integration:
 t
[0] 
ε ẋ (s) × B1 (y⊥ (s)) × B0 ds
0
[0] [0]
= ε x (t) × B1 (y⊥ (t)) − x (0) × B1 (y⊥ (0)) × B0
 t
∂ B1 [0] [0]
− ε x (s) × (y⊥ (s)) ẏ⊥ (s) × B0 ds.
0 ∂ x ⊥

[0]
This is O(ε) for 0 ≤ t ≤ c/ε, because x is bounded by assumption and ẏ⊥ (s) =
O(ε). With a Lipschitz bound of E and the Gronwall lemma, this yields that the
[0]
difference between y⊥ (t) and y⊥ (t) of (6.2) is bounded by

[0]
y⊥ (t) − y⊥ (t) = O(ε), 0 ≤ t ≤ c/ε.

123
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
 E. Hairer et al.

(d) With the above estimates we obtain, for n ≤ t ≤ n + 1 ≤ c/ε,

[n]    
x⊥ (t) − y⊥ (t) = x⊥ (t) − y⊥ (t) − 2 Re z 1[n] (t)eit/ε + 2 Re z 1[n] (t)eit/ε

n−1
[ j+1] [ j] [0]
+ y⊥ (t) − y⊥ (t) + y⊥ (t) − y⊥ (t)
j=0

= O(ε) + O(ε) + O(nε2 ) + O(ε) = O(ε),

which is the stated result. 



6.2 Perpendicular drift of numerical approximations

For the Boris algorithm with large step size (1.1) and a small perpendicular component
of the starting velocity we obtain the following result from Theorem 4.2.

Theorem 6.3 Under the assumptions of Theorem 4.2 (in particular (4.2)–(4.3)), and
provided that the numerical solution x n of the Boris method stays in a compact set K
for 0 ≤ t ≤ c ε−1 (with K and c independent of ε and h), the solution y⊥ (t) of the
initial-value problem for the slow differential equation (6.2) remains O(h 2 )-close to
the perpendicular component of x n over times O(ε−1 ):

|x⊥
n
− y⊥ (tn )| ≤ Ch 2 , 0 ≤ tn = nh ≤ c/ε. (6.4)

The constant C is independent of ε and h and n with 0 ≤ nh ≤ c/ε, but depends on

the initial velocity bound, on bounds of derivatives of B1 and E on the compact set
K , and on c.

Proof The proof uses Theorem 4.2 and Lady Windermere’s fan in the same way as
in the proof of Theorem 6.1, without any additional difficulty. We therefore omit the
details. 


Analogous results hold true also for the standard and filtered variational integra-
tors, for the latter with non-resonant stepsizes (4.6), using the corresponding modulated
Fourier expansions as given in Theorems 4.2 and 4.3 . We note that for the filtered vari-
ational integrator we do not need the smallness assumption (4.3) for the perpendicular
component of the velocity required for the Boris and standard variational integrators,
but the mere boundedness of the initial velocity suffices for the filtered variational
integrator. However, in view of the O(εh) remainder term (instead of O(εh 2 )) in the
0 in part (c) of Theorem 4.3, the error bound of x n for the
differential equation for z ±1 ⊥
filtered variational integrator is only O(h) instead of O(h 2 ).
Numerical experiment. For the example of Section 2 and for the methods (i)–(iii) of
the numerical experiments of Section 5, Figure 3 shows the projection of the computed
particle trajectory onto the plane perpendicular to B0 = e3 up to time T = 5/ε,
for the fixed step size h = 10−2 and three values of ε. The exact solution has a
gyroradius of O(ε), which is too small to be visible in the figure. It is observed

123
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
 Charged-particle dynamics over multiple time scales

0.5 0.5 0.5

0 0 0

-0.5 -0.5 -0.5

-0.5 0 0.5 -0.5 0 0.5 -0.5 0 0.5

Fig. 3 Particle trajectory for times t ≤ 5/ε projected onto the perpendicular plane as computed by the
Boris algorithm with starting values x(0), v(0) (top row), with modified initial values (3.4) (centre row),
and by the filtered variational integrator with starting values x(0), v(0) (bottom row). The step size used is
h = 10−2 in all cases

that the Boris algorithm with the original initial velocity as starting velocity shows a
substantially enlarged gyroradius for h ε, while after modifying the starting velocity
to (3.4), the Boris algorithm shows correct results. The same behaviour is observed
also for the standard variational integrator (not shown here, since the pictures are
indistinguishable). In contrast, the filtered variational integrator shows correct results
both for the original initial values (as shown) and for the modified starting velocity
(not shown here).

7 Long-term near-conservation of magnetic moment and energy

7.1 Time scale "−1 : Standard variational integrator

For the standard (unfiltered) variational integrator with step sizes (1.1) and the modified
starting velocity (3.4) we can show energy conservation up to O(h 2 ) over time ε−1 ,

123
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
 E. Hairer et al.

provided that h 6 ≤ ε. We do not have, and do not expect, such a result for the Boris
algorithm in a non-uniform magnetic field (2.1).

Theorem 7.1 Under the assumptions of Theorem 4.2 , and provided that the numerical
solution x n of the variational integrator with step size (1.1) and starting velocity
(3.4) stays in a compact set K for 0 ≤ t ≤ c ε−1 (with K and c independent of ε
and h), the total energy (2.3) remains O(h 2 )-close to the initial energy over times
c min(ε−1 , h −6 ):

|H (x n , v n ) − H (x 0 , v 0 )| ≤ Ch 2 , 0 ≤ nh ≤ c min(ε−1 , h −6 ). (7.1)

Moreover, with the modified initial velocity, the magnetic moment (2.4) remains O(ε2 )
small over times c ε−1 :

|I (x n , v n )| ≤ Cε2 , 0 ≤ nh ≤ c ε−1 . (7.2)

The constants C are independent of ε and h and n with 0 ≤ nh ≤ c/ε, but depend on
bounds of derivatives of B1 and E on the compact set K , and on c.

Proof The proof uses Theorem 4.2 and arguments from the proof of Proposition 6.2
in [13]. We first consider the energy behaviour over a short time interval of length
1, over which we can apply Theorem 4.2. With D = d/dt and the shift operator
eh D , with δ(ζ ) = ζ − ζ −1 and ρ(ζ ) = ζ − 2 + ζ −1 , and with the expansions
δ(eh )/(2h) = (1 + α2 h 2 + α4 h 4 + . . . ) and ρ(eh )/h 2 = (1 + β2 h 2 + β4 h 4 + . . . ),
we insert the decomposition (4.4) into the two-step formulation (3.2) of the numerical
method and obtain the equation for the function y(t) in (4.4) as

B0
ÿ + β2 h 2 y (4) + β4 h 4 y (6) + . . . = ( ẏ + α2 h 2 y (3) + α4 h 4 y (5) + . . .) ×
ε
δ(eh D ) δ(eh D )
+ A1 (y) y− A1 (y) − ∇φ(y) + O(|z|2 ) + O(h N ), (7.3)
2h 2h

where the left-hand side contains only even-order derivatives of y, and the right-hand
side contains only odd-order derivatives of y. We multiply both sides of (7.3) with
ẏ  . The multiplied left-hand side is the time derivative of an expression in which the
appearing second and higher derivatives of y can be substituted as functions of (y, ẏ)
via the differential equation for y in part (c) of Theorem 4.2; cf. [14]. On the right-hand
side we have

− ẏ  ε−1 B0 (α2 h 2 y (3) + α4 h 4 y (5) + . . .)

1 d
+ ẏ  A1 (y) δ(eh D )y − δ(eh D )A1 (y) − φ(y) + O(|z|2 ) + O(h N ). (7.4)
2h dt

The first term is O(h 2 ) because ẏ⊥ = ẏ1 + ẏ−1 and its derivatives are O(ε) by
Theorem 4.2. Since B0 is a skew-symmetric matrix, the first term is again the time
derivative of an expression in which the appearing second and higher derivatives of y

123
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
 Charged-particle dynamics over multiple time scales

can be substituted as functions of (y, ẏ); cf. [14]. The same holds true for the second
term, as is shown in the proof of Proposition 6.2 of [13].
We have thus found a function Hh (x, v) with the properties that uniformly for all
x in a bounded domain and all bounded v with v⊥ = O(ε) we have

Hh (x, v) = H (x, v) + O(h 2 ), (7.5)

d
Hh (y(t), ẏ(t)) = O(|z(t)|2 ) + O(h N ). (7.6)
dt

We now consider the equation for z. With the starting velocity (3.4) we have |z(0)| ≤
c0 h 4 for some constant c0 ; see part (d) of Theorem 4.2. The differential equation for
z ⊥ = z 1 + z −1 can be written as

4ε
ż ⊥ = z ⊥ × B0 + O(ε|z ⊥ |) + O(εh N ).
h2

Multiplying this equation with 2(z ⊥ ) and noting that 2(z ⊥ ) ż ⊥ = (d/dt)|z ⊥ |2 , we
obtain

d
|z ⊥ |2 ≤ Cε|z ⊥ |2 + O(εh N ),
dt

which shows that |z ⊥ (t)| ≤ ecεt |z ⊥ (0)| + O(tεh N ). Moreover, from the proof of
Theorem 4.2 we have |z 0 (t)| ≤ Ch 2 |z ⊥ (t)|. Patching many short time intervals of
length 1 together as in part (b) of the proof of Theorem 6.1, we find that on each
of these intervals up to time c ε−1 (but not on longer time intervals ε−α with α > 1
because of the ecεt exponential growth of our bound of z ⊥ ), we can apply Theorem 4.2
and the oscillatory component z on the interval remains of size O(h 4 ). By (7.6) we
thus have

Hh (y(t), ẏ(t)) = Hh (y(0), ẏ(0)) + O(th 8 ).

(Different to Section 6, we now do not put a superscript on y and z to designate the

interval of length 1 in which t lies). Together with

H (x n , v n ) = Hh (x n , v n ) + O(h 2 ) = Hh (y(tn ), ẏ(tn )) + O(h 2 ),

this yields the stated result for the energy.

The long-term smallness of the magnetic moment follows from (2.5) and the relation
v⊥n = ẏ (t ) − (−1)n ż (t ) + O(εh 2 ). This yields v n = O(ε) by the differential
⊥ n ⊥ n ⊥
equations in part (c) of Theorem 4.2 for y±1 and z ±1 , which contain a factor ε on the
right-hand side. These functions are again patched together over many short intervals
as is done in the proof of Theorem 6.1. 


123
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
 E. Hairer et al.

7.2 Time scale "−N for N > 1: Filtered variational integrator

We have the following result on the long-term near-conservation of magnetic moment

and energy by the filtered variational integrator with non-resonant large step sizes (4.5)
with (4.6).
Theorem 7.2 Let M > N be arbitrary positive integers. Under the assumptions of
Theorem 4.3 (in particular (4.5)–(4.6) and an initial velocity bounded independently
of ε), and provided that the numerical positions x n of the filtered variational integrator
stay in a compact set K for 0 ≤ t ≤ c ε−N (with K and c independent of ε and h), the
magnetic moment and the total energy along the numerical solution (x n , v n ) remain
almost conserved over such long times:

| I (x n , v n ) − I (x 0 , v 0 ) | ≤ Ch
for 0 ≤ t ≤ c min(h −M , ε−N ).
|H (x n , v n ) − H (x 0 , v 0 )| ≤ Ch

The constant C is independent of ε and h and n with 0 ≤ nh ≤ c min(h −M , ε−N ),

but depends on the initial velocity bound, on bounds of derivatives of B1 and E on the
compact set K , on c, and on the choice of M and N .
Proof The proof uses arguments that are very similar to the proofs of Theorems 2.2 and
2.3 of [15] on the long-term near-conservation properties of the standard variational
integrator for step sizes h ≤ cε. We therefore only indicate the main steps in the proof,
which are marked as items (i)-(iv) below.
To simplify the expressions for the remainder terms, we assume in the following
the mild stepsize restriction h m ≤ ε for some fixed m > 2 and we choose M ≥ m N .
This is only done for ease of presentation and allows us to cover the time scale ε−N .
Without this assumption we arrive at the stated time scale min(h −M , ε−N ).
(i) (Lagrangian structure of the modulation equations; cf. [15, (5.23)]) Over a
time interval of length 1 we consider the modulation functions z k (t) of Theorem 4.3
multiplied with the corresponding highly oscillatory exponentials:

y k (t) = z k (t)eikt/ε for |k| ≤ N and y k (t) = 0 for |k| > N .

We write y = (y k )k∈Z and define the extended potentials

 1 (m) 0 α
U(y) = φ (y )y
m!
0≤m≤N
s(α)=0
⎛ ⎞
⎜  1 (m) 0 α ⎟
A(y) = (Ak (y))k∈Z = ⎜
⎝ A (y )y ⎟
⎠ ,
m!
0≤m≤N
s(α)=k k∈Z

where the sums are taken over all multi-indices α = (α 1 , . . . , α m ) with α j ∈ Z \

{0} with prescribed sum s(α) = α 1 + . . . + α m , and where we use the notation

123
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
 Charged-particle dynamics over multiple time scales

φ (m) (y 0 )yα = φ (m) (y 0 )(y α , . . . , y α ) for the m-linear mth derivative of φ evaluated
1 m

at (y α , . . . , y α ), and analogously for A(m) (y 0 )yα . The terms for m = 0 are to be

1 m

interpreted as φ(y 0 ) and A(y 0 ).

The system of modulation equations of the filtered variational integrator can then
be written, up to O(ε N ), as the discrete Euler-Lagrange equations corresponding to
the discrete Lagrangian

h n+1/2 ∗ −1
Lh (yn , yn+1 ) = (v ) ( ⊗ I )vn+1/2
2
h ∗ h 
+ A(yn ) + A(yn+1 ) vn+1/2 − U(yn ) + U(yn+1 )
2 2

with vn+1/2 = (yn+1 − yn )/h, which differs from that of the standard variational
integrator only by the modified kinetic energy term with −1 . We thus have

 ∂A j ∗
∂U ∗
−1 2 k
δh y = (y) δ2h y j − δ2h Ak (y) − (y) + O(ε N ), (7.7)
∂ yk ∂ yk
j∈Z

where δ2h f (t) = ( f (t + h) − f (t − h))/(2h) and δh2 f (t) = ( f (t + h) − 2 f (t) +

f (t − h))/h 2 denote the first-order and second-order symmetric difference quotients,
respectively.
(ii) (Almost-invariant close to the magnetic moment; cf. [15, Theorem 5.2]) With
the group action S(λ)y = (eikλ y k )k∈Z (for λ ∈ R), we have

U(S(λ)y) = U(y), A(S(λ)y) = S(λ)A(y) for all λ.

Differentiation with respect to λ (at λ = 0) yields

 ∂U
ik (y)y k = 0
∂ yk
k∈Z
 ∂Ak
ij (y)y j = ikAk (y) for k ∈ Z.
∂yj
j∈Z

Multiplying (7.7) with −ik(y k )∗ , summing over k and using these relations yields that
the function
i  k ∗ −1 k
Ih [y](t) = − ky (t) y (t + h)
εh
k
i 
+ k Ak (y(t))∗ y k (t + h) − y k (t)∗ Ak (y(t + h))
2ε
k

satisfies

Ih [y](t) − Ih [y](t − h) = O(hε N +1 ) (7.8)

123
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
 E. Hairer et al.

and is thus an almost-invariant of the modulation system. Using the bounds of the
modulation functions, we find that

i
Ih [y](t) = − y 1 (t)∗ −1 y 1 (t + h) − y −1 (t)∗ −1 y −1 (t + h)
εh
i
+ 2 y 1 (t)∗ (y 1 (t + h) × B0 ) − y −1 (t)∗ (y −1 (t + h) × B0 ) + O(ε).
2ε

Here, a calculation shows that the first term equals (1 + cos(h/ε))|z 11 |2 /ε2 + O(h),
and the second term equals − cos(h/ε)|z 11 |2 /ε2 + O(h). So we obtain

1 1 2
Ih [y](t) = |z (t)| + O(h).
ε2 1

On the other hand, since

x n+1 − x n−1
vn = + O(ε)
2h
x n+1 − x n−1
= I + (1 − sinc(h/ε)−1 ) B̂02 + O(ε)
2h
i 1 −1
= ż 0 (t) + z (t)eit/ε − z −1 (t)e−it/ε + O(h) at t = nh
ε 1

and ż 0 (t) × B0 = O(ε), we find that

1 1 2
I (x n , v n ) = 21 |v n × B0 |2 + O(ε) = |z (t)| + O(h).
ε2 1

So we obtain that the magnetic moment along the numerical solution is O(h)-close to
the almost-invariant:

I (x n , v n ) = Ih [y](nh) + O(h). (7.9)

(iii) (Almost-invariant close to the total energy; cf. [15, Theorem 5.3]) Multiplying
(7.7) with ( ẏ k )∗ and summing over k gives

  d d
( ẏ k )∗ −1 2 k
δh y − Ak (y)∗ δ2h y k − ( ẏ k )∗ δ2h Ak (y) + U(y)
dt dt
k k
= O(ε N ). (7.10)

The arguments in the proof of Theorem 5.3 in [15] show that each of the three terms
on the left-hand side is a total differential up to O(ε N ). So there exists a function

Hh [y](t) = Kh [y](t) + Mh [y](t) + U[y](t),

123
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
 Charged-particle dynamics over multiple time scales

where the time derivatives of the three terms on the right-hand side equal the three
corresponding terms on the left-hand side of (7.10), and we have

d
Hh [y](t) = O(ε N ).
dt

We now determine the dominant part of Hh [y]. We find

2(z 1 )∗ −1 z 1 h
Kh [y] = 21 (ż 0 )∗ −1 0
ż + sin(h/ε) − 2 sin2 (h/2ε) + O(h)
h2 ε
2|z 11 |2 h
= 21 |ż 0 |2 + sin(h/ε) − 2 sin2 (h/2ε) + O(h)
tanc(h/2ε)h 2 ε
|z 11 |2
= 21 |ż 0 |2 + (1 + cos(h/ε)) + O(h)
ε2
|z 1 |2
Mh [y] = − cos(h/ε) 12 + O(h)
ε
U[y] = φ(z 0 ) + O(h).

Thus we have

|z 11 (t)|2
Hh [y](t) = 21 |ż 0 (t)|2 + + φ(z 0 (t)) + O(h). (7.11)
ε2

On the other hand, from the formula for v n in (ii) we have, at t = nh,

|z 11 (t)|2
2 |v | = 21 |ż 0 (t)|2 + + O(h).
1 n 2
ε2

The energy along the numerical solution is therefore

|z 11 (t)|2
H (x n , v n ) = 21 |v n |2 + φ(x n ) = 21 |ż 0 (t)|2 + + φ(z 0 (t)) + O(h).
ε2

and hence we have

H (x n , v n ) = Hh [y](t) + O(h).

(iv) (From short to long time intervals; cf. [15, Section 4.5], [17, Section XIII.7]).
The stated long-time near-conservation results are now obtained by patching together
the short-time near-conservation results of (ii) and (iii) over many intervals of length
1, via an often-used argument that involves the uniqueness up to O(ε N +1 ) of the
modulation functions. 


123
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
 E. Hairer et al.

10

5
energy error

-5

-10
0 1 2 3 4 5 6 7 8 9 10
106
10

5
energy error

-5

-10
0 1 2 3 4 5 6 7 8 9 10
106

0.08
energy error

0.06
0.04
0.02
0
-0.02
0 1 2 3 4 5 6 7 8 9 10
106

Fig. 4 Energy error H (xn , vn ) − H (x0 , v0 ) along the numerical solutions of the Boris algorithm (top), of
the standard variational integrator (centre) and of the filtered variational integrator (bottom), obtained with
ε = 10−4 and h = 10−2

Numerical experiment. We illustrate the energy behaviour of the numerical methods

for the magnetic field
⎛ ⎞ ⎛ ⎞
1 x − x3
1⎝ ⎠ ⎝ 2
B(x) = 0 + x1 + x3 ⎠
ε 0.5 x −x 2 1

and the scalar potential φ(x) = x13 − x23 + 15 x14 + x24 + x34 . We take the initial values
x(0) = (0, 1, 0.1) , v(0) = (0.09, 0.05, 0.2) .
We apply the three numerical integrators of Section 3 with ε = 10−4 , step size h =
−2
10 , and final time T = 107 . Figure 4 shows the energy error H (xn , vn ) − H (x0 , v0 )
along the numerical solutions of the Boris algorithm, the standard variational integrator
and the filtered variational integrator, taking the initial values x(0), v(0) as starting
values for all three methods.

123
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
 Charged-particle dynamics over multiple time scales

10-3
2
energy error

-2

-4

-6

0 1 2 3 4 5 6
107
10-3

2
energy error

-2

-4
0 1 2 3 4 5 6
107

Fig. 5 Energy error H (xn , vn ) − H (x0 , v0 ) along the numerical solutions of the Boris algorithm (top)
and of the standard variational integrator (bottom) obtained with modified initial values (3.4) and step size
h = 10−2 , for ε = 10−4

The energy errors of the Boris algorithm and the variational integrator (top and
centre picture) appear to behave randomly. Running several trajectories corresponding
to random perturbations of the initial data of magnitude 10−14 showed energy errors
that look like random walks with a deviation of magnitude 10 for t ≤ 106 . For
larger times, some of the trajectories showed blow-up behaviour. A similar random
walk behaviour was observed also for the deviation of the magnetic moment, but the
deviation was less than 10−1 for t ≤ 107 .
In contrast, the energy error of the filtered variational integrator oscillates with
a small amplitude without drift (bottom picture of Figure 4). The error I (xn , vn ) −
I (x0 , v0 ) of the magnetic moment along the numerical solution of the filtered varia-
tional integrator has a very similar behaviour (not shown here).
If we apply the Boris algorithm and the standard variational integrator with modified
initial values (3.4), then the magnetic moment remains small over very long time,
oscillating between 0 and approximately 2 · 10−6 over the whole time interval. In this
case of modified initial velocity, we observe very good near-conservation of energy for
the variational integrator while there is a linear drift for the Boris algorithm. However,
this drift becomes dominant over the small oscillations in the energy only for times
∼ 107 ; see Figure 5.

Acknowledgements This work was partially supported by the Swiss National Science Foundation, grant
No. 200020_192129, and by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation)
– Project-ID258734477 – SFB 1173. The work by Yanyan Shi was done at the University of Tübingen

123
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
 E. Hairer et al.

during her one-year research stay, which was funded by a scholarship provided by the University of the
Chinese Academy of Sciences (UCAS).

Funding Open Access funding enabled and organized by Projekt DEAL.

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which
permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give
appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence,
and indicate if changes were made. The images or other third party material in this article are included
in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If
material is not included in the article’s Creative Commons licence and your intended use is not permitted
by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the
copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.

References
1. Benettin, G., Sempio, P.: Adiabatic invariants and trapping of a point charge in a strong nonuniform
magnetic field. Nonlinearity 7(1), 281 (1994)
2. Birdsall, C.K., Langdon, A.B.: Plasma Physics via Computer Simulation. Taylor and Francis Group,
New York (2005)
3. Boris, J. P.: Relativistic plasma simulation-optimization of a hybrid code. Proceeding of Fourth Con-
ference on Numerical Simulations of Plasmas (November 1970), 3–67
4. Brizard, A.J., Hahm, T.S.: Foundations of nonlinear gyrokinetic theory. Rev. Modern Phys. 79(2),
421–468 (2007)
5. Chartier, P., Crouseilles, N., Lemou, M., Méhats, F., Zhao, X.: Uniformly accurate methods for Vlasov
equations with non-homogeneous strong magnetic field. Math. Comp. 88(320), 2697–2736 (2019)
6. Chartier, P., Crouseilles, N., Lemou, M., Méhats, F., Zhao, X.: Uniformly accurate methods for three
dimensional Vlasov equations under strong magnetic field with varying direction. SIAM J. Sci. Comput.
42(2), B520–B547 (2020)
7. Crouseilles, N., Lemou, M., Méhats, F., Zhao, X.: Uniformly accurate particle-in-cell method for the
long time solution of the two-dimensional Vlasov-Poisson equation with uniform strong magnetic
field. J. Comput. Phys. 346, 172–190 (2017)
8. Derouillat, J., Beck, A., Pérez, F., Vinci, T., Chiaramello, M., Grassi, A., Flé, M., Bouchard, G.,
Plotnikov, I., Aunai, N., et al.: Smilei: A collaborative, open-source, multi-purpose particle-in-cell
code for plasma simulation. Computer Physics Commun. 222, 351–373 (2018)
9. Ellison, C.L., Burby, J.W., Qin, H.: Comment on ”Symplectic integration of magnetic systems”: A
proof that the Boris algorithm is not variational. J. Comput. Phys. 301, 489–493 (2015)
10. Filbet, F., Rodrigues, L.M.: Asymptotically stable particle-in-cell methods for the Vlasov-Poisson
system with a strong external magnetic field. SIAM J. Numer. Anal. 54(2), 1120–1146 (2016)
11. Filbet, F., Rodrigues, L.M.: Asymptotically preserving particle-in-cell methods for inhomogeneous
strongly magnetized plasmas. SIAM J. Numer. Anal. 55(5), 2416–2443 (2017)
12. Filbet, F., Rodrigues, L.M., Zakerzadeh, H.: Convergence analysis of asymptotic preserving schemes
for strongly magnetized plasmas. Numer. Math. 149(3), 549–593 (2021)
13. Hairer, E., Lubich, C.: Symmetric multistep methods for charged particle dynamics. SMAI J. Comput.
Math. 3, 205–218 (2017)
14. Hairer, E., Lubich, C.: Energy behaviour of the Boris method for charged-particle dynamics. BIT 58,
969–979 (2018)
15. Hairer, E., Lubich, C.: Long-term analysis of a variational integrator for charged-particle dynamics in
a strong magnetic field. Numer. Math. 144(3), 699–728 (2020)
16. Hairer, E., Lubich, C., Wang, B.: A filtered Boris algorithm for charged-particle dynamics in a strong
magnetic field. Numer. Math. 144(4), 787–809 (2020)
17. Hairer, E., Lubich, C., Wanner, G.: Geometric Numerical Integration. Structure-Preserving Algo-
rithms for Ordinary Differential Equations, 2nd ed. Springer Series in Computational Mathematics
31. Springer-Verlag, Berlin, 2006

123
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
 Charged-particle dynamics over multiple time scales

18. Hairer, E., Nørsett, S. P., Wanner, G.: Solving Ordinary Differential Equations I. Nonstiff Problems,
2nd ed. Springer Series in Computational Mathematics 8. Springer, Berlin, 1993
19. Kruskal, M.: The gyration of a charged particle. Rept. PM-S-33 (NYO-7903), Princeton University,
Project Matterhorn (1958)
20. Northrop, T. G.: The adiabatic motion of charged particles. Interscience Tracts on Physics and Astron-
omy, Vol. 21. Interscience Publishers John Wiley & Sons New York-London-Sydney, 1963
21. Possanner, S.: Gyrokinetics from variational averaging: existence and error bounds. J. Math. Phys. 59,
8 (2018), 082702, 34
22. Qin, H., Zhang, S., Xiao, J., Liu, J., Sun, Y., Tang, W. M.: Why is Boris algorithm so good? Physics
of Plasmas 20, 8 (2013), 084503.1–4
23. Ricketson, L. F., Chacón, L.: An energy-conserving and asymptotic-preserving charged-particle orbit
implicit time integrator for arbitrary electromagnetic fields. J. Comput. Phys. (2020), 109639
24. Wang, B., Zhao, X.: Error estimates of some splitting schemes for charged-particle dynamics under
strong magnetic field. SIAM J. Numer. Anal. 59(4), 2075–2105 (2021)
25. Webb, S.D.: Symplectic integration of magnetic systems. J. Comput. Phys. 270, 570–576 (2014)
26. Xiao, J., Qin, H.: Slow manifolds of classical Pauli particle enable structure-preserving geometric
algorithms for guiding center dynamics. Computer Physics Comm. 265, 107981 (2021)

Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps
and institutional affiliations.

123
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
Terms and Conditions
Springer Nature journal content, brought to you courtesy of Springer Nature Customer Service Center
GmbH (“Springer Nature”).
Springer Nature supports a reasonable amount of sharing of research papers by authors, subscribers
and authorised users (“Users”), for small-scale personal, non-commercial use provided that all
copyright, trade and service marks and other proprietary notices are maintained. By accessing,
sharing, receiving or otherwise using the Springer Nature journal content you agree to these terms of
use (“Terms”). For these purposes, Springer Nature considers academic use (by researchers and
students) to be non-commercial.
These Terms are supplementary and will apply in addition to any applicable website terms and
conditions, a relevant site licence or a personal subscription. These Terms will prevail over any
conflict or ambiguity with regards to the relevant terms, a site licence or a personal subscription (to
the extent of the conflict or ambiguity only). For Creative Commons-licensed articles, the terms of
the Creative Commons license used will apply.
We collect and use personal data to provide access to the Springer Nature journal content. We may
also use these personal data internally within ResearchGate and Springer Nature and as agreed share
it, in an anonymised way, for purposes of tracking, analysis and reporting. We will not otherwise
disclose your personal data outside the ResearchGate or the Springer Nature group of companies
unless we have your permission as detailed in the Privacy Policy.
While Users may use the Springer Nature journal content for small scale, personal non-commercial
use, it is important to note that Users may not:

1. use such content for the purpose of providing other users with access on a regular or large scale
basis or as a means to circumvent access control;
2. use such content where to do so would be considered a criminal or statutory offence in any
jurisdiction, or gives rise to civil liability, or is otherwise unlawful;
3. falsely or misleadingly imply or suggest endorsement, approval , sponsorship, or association
unless explicitly agreed to by Springer Nature in writing;
4. use bots or other automated methods to access the content or redirect messages
5. override any security feature or exclusionary protocol; or
6. share the content in order to create substitute for Springer Nature products or services or a
systematic database of Springer Nature journal content.
In line with the restriction against commercial use, Springer Nature does not permit the creation of a
product or service that creates revenue, royalties, rent or income from our content or its inclusion as
part of a paid for service or for other commercial gain. Springer Nature journal content cannot be
used for inter-library loans and librarians may not upload Springer Nature journal content on a large
scale into their, or any other, institutional repository.
These terms of use are reviewed regularly and may be amended at any time. Springer Nature is not
obligated to publish any information or content on this website and may remove it or features or
functionality at our sole discretion, at any time with or without notice. Springer Nature may revoke
this licence to you at any time and remove access to any copies of the Springer Nature journal content
which have been saved.
To the fullest extent permitted by law, Springer Nature makes no warranties, representations or
guarantees to Users, either express or implied with respect to the Springer nature journal content and
all parties disclaim and waive any implied warranties or warranties imposed by law, including
merchantability or fitness for any particular purpose.
Please note that these rights do not automatically extend to content, data or other material published
by Springer Nature that may be licensed from third parties.
If you would like to use or distribute our Springer Nature journal content to a wider audience or on a
regular basis or in any other manner not expressly permitted by these Terms, please contact Springer
Nature at

[email protected]