J. Plasma Phys. (2025), vol. 91, E34 © The Author(s), 2025.

1
Published by Cambridge University Press
This is an Open Access article, distributed under the terms of the Creative Commons Attribution licence (http://
creativecommons.org/licenses/by/4.0/), which permits unrestricted re-use, distribution and reproduction,
provided the original article is properly cited.
doi:10.1017/S0022377824001648

Application of Lagrangian techniques for

calculating the on-axis rotational transform

S. Guinchard 1, †, W. Sengupta 2 and S.R. Hudson 1

1 Princeton Plasma Physics Laboratory, Princeton, NJ 08540, USA

2 Department of Astrophysical Sciences, Princeton University, Princeton, NJ 08543, USA

(Received 26 April 2024; revised 2 December 2024; accepted 3 December 2024)

The Floquet exponents of periodic field lines are studied through the variations of the
magnetic action on the magnetic axis, which is assumed to be elliptical. The near-axis
formalism developed by Mercier, Solov’ev and Shafranov is combined with a Lagrangian
approach. The on-axis Floquet exponent is shown to coincide with the on-axis rotational
transform. A discrete solution suitable for numerical implementation is introduced, which
gives the Floquet exponents as solutions to an eigenvalue problem. This discrete formalism
expresses the exponents as the eigenvalues of a 6 × 6 matrix.
Key words: magnetic action, rotational transform, near-axis expansion, Lagrangian integration,
Mercier, Solov’ev-Shafranov, variation

1. The need for and the origin of the rotational transform

Magnetostatic equilibria are characterized by the following equations:

∇p = J × B, ∇ × B = J, ∇ · B = 0, (1.1a–c)

where B is the magnetic field, J is the current density and p is the scalar pressure. With
the assumption that p is not constant in a small region and the surface lies in a bounded
volume of space, the boundary must be topologially toroidal (Kruskal & Kulsrud 1958).
The phase portrait of B, where the magnetic field lines are treated like integral curves of
a Hamiltonian dynamical system, is characterized by the topology of the level sets of p.
A magnetic field line action can be defined (Cary & Littlejohn 1983), which serves as a
starting point for the Lagrangian integration carried out in this paper.
For magnetic confinement of plasmas in toroidal geometries, that the magnetic field
lines rotate poloidally (the short way) as they rotate toroidally (the long way) around the
torus is essential for cancelling charged particle drifts, which would otherwise lead to loss
(Spitzer 1958). The number of poloidal rotations that a field line achieves per toroidal
period is called the rotational transform, ι (Spitzer 1958).
Kruskal & Kulsrud consider the case where ‘p is reasonably smooth and not constant
in any region’, and this entails that the magnetic field lines lie on nested flux surfaces. In
† Email address for correspondence: [email protected]

Downloaded from https://www.cambridge.org/core. IP address: 213.255.248.59, on 12 Feb 2025 at 06:14:27, subject to the Cambridge Core terms of use, available at
https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0022377824001648
 2 S. Guinchard, W. Sengupta and S.R. Hudson
the following, we do not place such a constraint on the pressure and we do not require that
the magnetic field has nested flux surfaces. The existence of nested magnetic surfaces,
their link with continuous symmetries and the preservation of magnetic surfaces under
small symmetry-destroying perturbation are widely discussed in Kolmogorov (1954),
Arnol’d (1963) and Moser (1962) and also in texts and reviews: Moser (1973), Arnold
(1978), Lichtenberg & Lieberman (1992) and Meiss (1992). Herein, we only require that
the magnetic field is ‘toroidal’, which we describe as follows. We consider a magnetic
configuration in the domain Ω, which is assumed to be a solid torus, with B · ∇φ > 0
everywhere where φ is the toroidal angle, and with B · n = 0 where n is normal to the
boundary. For any poloidal section Σ of Ω, the Poincaré return map is defined by the
intersection of field lines and Σ after one toroidal period. Brouwer’s fixed-point theorem
(Brouwer 1910) ensures that there will be at least one fixed point of the Poincaré first return
map.
Taking such a point as the origin of polar-like coordinates, e.g. (r, θ ), we may consider
the effect of iterating the mapping on nearby points, where the point (rn , θn ) gets mapped
to (rn+1 , θn+1 ). If, upon iterating the return map, the points rotate around the fixed point,
then the fixed point is called elliptic. (A more rigorous definition of elliptic fixed points is
given in Meiss (1992).) There may be more than one fixed point, and not all fixed points
are elliptic. Even though the methods of Lagrangian integration as described below can
be applied more generally, for clarity of exposition we restrict our attention to the case
of an isolated elliptic fixed point. Stellarators are typically designed to have one easily
identifiable elliptic fixed point about which most field lines rotate, and this is called the
magnetic axis.
This sequence of angular displacements, θ = θn+1 − θn , enables one to define the
rotational transform of a field line:

1 
N
ι := lim θn . (1.2)
N↑∞ 2πN
n

Even though the angle coordinate becomes degenerate at the origin in polar-like
coordinates, the common approach is to define the on-axis rotational transform by taking
the limit as the starting point gets closer to the axis.
Note that the this definition of the rotational transform is purely geometric, in that the
rotational transform effectively measures how many times a given magnetic field line links
the magnetic axis. A discussion of the rotational transform as an asymptotic linking of
neighbouring field lines can be found in Arnold (2014).
Mercier (see Mercier 1964; Helander 2014) expressed the on-axis rotational transform
as an integral along the axis:
  
1 d J0 
ι=N+ +δ −τ , (1.3)
2π cosh η 2B0
with  denoting the arc length. The on-axis current density and magnetic field are denoted
by J0 and B0 , respectively, and the torsion of the axis by τ . The eccentricity of flux surfaces
is described by η and δ is a parameter describing their rotation around the axis, with δ  :=
dδ/d. Here N is an integer coming from the phase of the rotation term δ (see Pfefferlé
et al. 2018). Note that Spitzer (1958) had identified independently that a way to generate
some rotational transform is to give torsion to an axis. This expression provides invaluable
insight, showing how rotational transform can be produced by plasma currents, as is used
by tokamaks, or by geometrical shaping, as is used by stellarators.

Downloaded from https://www.cambridge.org/core. IP address: 213.255.248.59, on 12 Feb 2025 at 06:14:27, subject to the Cambridge Core terms of use, available at
https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0022377824001648
 Application of Lagrangian techniques 3
One may also encounter different definitions of the rotational transform. Assuming that
the magnetic configuration possesses magnetic surfaces, ensuring that the toroidal and
poloidal magnetic fluxes ψ and χ respectively, can be properly defined, then as shown by
Mercier et al. (1974), ι can also be expressed as the ratio of differential fluxes:

dχ
ι= . (1.4)
dψ

Mercier’s expression (1.3) for instance was derived starting from the flux definition
(1.4). The fluxes were expanded in a power series of the distance to the axis by
constructing custom polar coordinates, so-called Mercier coordinates, which are described
in Appendix C.
In this paper, we present a derivation of Mercier’s formula using Lagrangian integration.
This approach expresses the rotational transform as a Floquet exponent of the field lines.
The relationship between the rotational transform and the Floquet exponent was described
by Greene (1979), and more recently by Duignan & Meiss (2021). In § 2, the magnetic
field-line action is defined. Stationary curves of the action are shown to be field lines,
enabling one to identify a magnetic axis. Assumption is made in this paper that the axis
is elliptical, but the formalism can be applied to the hyperbolic case. In § 3, Mercier’s
formula (1.3) is derived from the second variation of the action. The result is obtained
through a near-axis expansion of the null eigenspace of the second variation operator. The
periodicity enables the use of a Floquet description of the solutions. This same result is
derived through the theory of the Hill infinite determinant, analogous to Schrödinger’s
equation in a periodic potential. In both cases, the rotational transform is shown to be a
Floquet exponent of the null eigenspace of the second variation operator. In § 4, a discrete
formalism is introduced so that the Floquet exponents can be solved for numerically.

2. Description of the magnetic field-line action

Let us start by defining the magnetic field line action, as introduced by Cary & Littlejohn
(1983). This is defined as a line integral that, for a given magnetic field, depends only the
integration contour. Herein, the integration contours are assumed to be curves that close
after one toroidal period. Let us consider a closed, differentiable curve C ⊂ R3 with total
length L, closing after one toroidal transit. The latter assumption that the curve closes after
one toroidal period is a necessary condition for a magnetic axis. Curve C is parametrized
by the C1 , L-periodic vector-valued function

x : [0, L] → R3
(2.1)
 → x() ∈ C ,

 being the arc length and we note x := dx/d. The action is defined as the circulation
along C of the magnetic vector potential A, with B = ∇ × A in Ω:

S := d A · x . (2.2)
C

Properties of the magnetic field are accessible through calculus of variations, performed
on the action. The variations are performed with respect to changes in the geometry of the

Downloaded from https://www.cambridge.org/core. IP address: 213.255.248.59, on 12 Feb 2025 at 06:14:27, subject to the Cambridge Core terms of use, available at
https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0022377824001648
 4 S. Guinchard, W. Sengupta and S.R. Hudson
curve. The first variation with respect to a variation δx is

δ S [δx] = d x × B · δx, (2.3)
C

which shows that stationary curves are tangential to the magnetic field and hence field
lines. We used that δA[δx] = δx · ∇A. From (2.3), a magnetic axis can be identified.
We focus in particular on elliptical axes, but one can apply the following formalism to
hyperbolic axes as well. The difference would be seen in the flux-surface functions around
the axis, and in the nature of the Floquet exponents. Let us denote an elliptical axis by
Ca . Additional properties of the field appear at higher orders of variations of S . For the
rotational transform in particular, the second-order variation δ 2 S is of interest.
3. Verification of Mercier’s formula for the Floquet exponent
3.1. The second variation as an operator
In order to express the on-axis rotational transform, the second-order variation of the
action applied to the axis needs to be derived. Assume that an elliptical axis Ca has
been found as a stationary curve of the action. The formalism is identical for hyperbolic
axes, but in this paper, we focus on the elliptic case. The second variation of the action
performed from the axis Ca is
 

δ S [δx] =
2
d δ(x × B · δx) = d δx · (δx × B + x × δB), (3.1)
Ca Ca

where f  := d f /d for any f . Using δB = δx · ∇B and the Einstein summation

convention, we write the second variation as an operator:

δ2S
δ2S = d δxi i j δx j , (3.2)
Ca δx δx
where
δ2S d
= ijk Bk + imk xm ∂j Bk , (3.3)
δx δx
i j d
and i, j, k, m ∈ {1, 2, 3}, which in matrix form reads
δ2S d
= − (I × B)
M := + x × (∇B)T . (3.4)
δxδx d
The identity tensor is denoted by I . Note that the covariant tensor M can be easily
symmetrized (see Hudson & Dewar 2009). However, the direction in which the derivative
d/d is taken has to be chosen carefully. We continue with the non-symmetric form (3.4).
We show that the null eigenspace of M is of particular interest. Indeed, by periodicity
of the field lines considered to compute the action, the eigenspaces of the operator M are
related to geometric properties of the neighbouring field lines, including the rotational
transform. Let v be a null eigenfunction of M such that
M v = 0. (3.5)
For v to be non-trivial, (3.5) rewrites
det (M ) = 0. (3.6)
We will demonstrate that the condition (3.5) is satisfied by solutions v, such that their
Floquet exponent is the on-axis rotational transform, and (3.6) will serve in the discrete

Downloaded from https://www.cambridge.org/core. IP address: 213.255.248.59, on 12 Feb 2025 at 06:14:27, subject to the Cambridge Core terms of use, available at
https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0022377824001648
 Application of Lagrangian techniques 5
formalism introduced in § 4. To solve (3.5), the near-axis formalism developed by Mercier,
Solov’ev and Shafranov (Solov’ev & Shafranov 1970; Mercier et al. 1974) is used.
Near-axis formalism in the inverse coordinate approach, where the flux surface ψ is
used as a coordinate, has proved to be very successful in understanding quasi-symmetry
(see Garren & Boozer 1991; Landreman & Sengupta 2018, 2019; Rodríguez, Sengupta
& Bhattacharjee 2023). However, for this work, the near-axis formalism in direct
or Mercier–Solov’ev–Shafranov coordinates (Jorge, Sengupta & Landreman 2020b,a;
Sengupta et al. 2024) is more relevant.
3.2. Derivation of the Floquet exponent from the second variation
The operator is expanded in the Solov’ev–Shafranov coordinates, introduced in Appendix
C.2. The latter set of coordinates is closely related to the Mercier coordinates, presented
in Appendix C.1.
In what follows, it is assumed that the magnetic axis is a closed curve Ca ⊂ R3
parametrized by the vector-valued function r 0 , with the arc length  as parameter, and
a total length L. The basis of expansion is taken to be {e1 , e2 , e3 }, whose expressions in
terms of N , B and t are given in Appendix C.2, but the results will always be expressed in
terms of the Solov’ev–Shafranov vectors {N , B, t}. Additionally, the following notation
is adopted: f  = df /d for any f .
3.2.1. Magnetic field expansion near an elliptic magnetic axis
From (3.4), the magnetic field B needs to be described in the vicinity of the axis r 0 for
M to be expanded in the near-axis formalism. The expansion of B in Solov’ev–Shafranov
coordinates is carried in the limit x, y 1, or in other words, is limited to linear terms
only. The magnetic field can be written in contravariant form as
   
|g|B = |g|B1 e1 + |g|B2 e2 + |g|B3 e3 , (3.7)
with
  
|g|B1 = a1 x + a2 y, |g|B2 = b1 x + b2 y, |g|B3 = B0 + c1 x + c2 y, (3.8a–c)

√ a1 , a2 , b1 , b2 , c1 , c2 are periodic functions of , B0 the zeroth-order magnitude of B

where
and |g| = h as defined in Appendix C.2. Therefore, in the limit x, y 1, h = 1 and the
magnetic field reads

B = (B0 + c1 x + c2 y) t + a1 x + (a2 + u B0 )y N

+ (b1 − u B0 )x + b2 y B + O (x2 + y2 ). (3.9)
A direct calculation shows that
⎫
∇B = κB0 tn + a1 N N + b2 BB + B0 tt + (a2 + u B0 )BN ⎪
⎬
+ (b1 − u B0 )N B + (c1 N + c2 B)t,
  ⎪
⎭
x × (∇B)T = κB0 bt + B a1 N + (a2 + u B0 )B − N (b1 − u B0 )N + b2 B ,
(3.10)
which yields, using dyadic algebra
⎫
∇ · B = I : ∇B = B0 + a1 + b2 , ⎪
⎬
× 
∇ × B = −I .∇B = (b1 − a2 − 2u B0 )t + (c2 − κB0 sin δ)N (3.11)
⎪
⎭
−(c1 − κB0 cos δ)B.

Downloaded from https://www.cambridge.org/core. IP address: 213.255.248.59, on 12 Feb 2025 at 06:14:27, subject to the Cambridge Core terms of use, available at
https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0022377824001648
 6 S. Guinchard, W. Sengupta and S.R. Hudson
Here κ denotes the local curvature of the axis r 0 . Additionally, (C10a,b) has been used to
rewrite b in terms of N and B. Requiring that ∇ · B = 0 and ∇ × B = J0 t, one finds
b1 − a2 J0
a1 + b2 = −B0 , = u + , c1 = κB0 cos δ, c2 = κB0 sin δ. (3.12a–d)
2B0 2B0
So far, only two constraints have been derived for the four expansion functions
a1 , a2 , b1 , b2 . Turning to the Mercier representation (D2a,b) and (D7a–c) of B brings in
additional conditions. Together with the definitions of N and B,
B   
= t 1 + κ x2 + y2 cos (u − δ)
B0
⎡ ⎛ ⎞ ⎤
J0
  −τ +δ 
⎢ 1 B0 ⎜ −J0 2B0 ⎟ ⎥
+N ⎢ −
⎣ 2 B0 + η 
x + ⎜
⎝ 2B0 + sinh η⎟ ⎥
⎠ y⎦
cosh η

⎡⎛ ⎞ ⎤
J0
− τ + δ    
⎢⎜ J0 ⎟ ⎥
+B⎢⎜ +
2B0
sinh η ⎟ x − 1 B0 − η y⎥
⎣⎝ 2B0 cosh η ⎠ 2 B0 ⎦

+ O (x2 + y2 ), (3.13)
where δ, as explained in Appendix C, describes the rotation of elliptical flux surfaces
around the expansion axis r 0 . As for η, it describes the eccentricity of the flux surfaces
around r 0 . Defining
J0
− τ + δ
2B0
Ω0 () = , (3.14)
cosh η
the comparison between the Mercier representation (3.13) and the Solov’ev–Shafranov
representation (3.9) yields for the expansion coefficients
    ⎫
a1 1 B0 a2 J0 /2 ⎪
=− +η ,
=− u + 
+ Ω0 sinh η,⎪
⎪
⎬
B0 2 B0 B0 B0
     (3.15)
b2 1 B0 b1 J0 /2 ⎪
⎪
=− −η ,
= u + 
+ Ω0 sinh η, ⎭⎪
B0 2 B0 B0 B0
which satisfy the relations (3.12a–d). From the definition of Ω0 , we can further simplify
a2 and b1 as
a2 b1
= −Ω0 e−η , = Ω0 eη . (3.16a,b)
B0 B0
As a check of correctness for the above expressions, one can show that B · ∇ψ = 0, where
√ √ √ √
gB · ∇ = gB1 ∂x + gB2 ∂y + gB3 ∂
= B0 ∂ + (a1 x + a2 y)∂x + (b1 x + b2 y)∂y . (3.17)
The flux surface function ψ for the elliptic case is given in the Mercier representation by
(D10) or equivalently in the Solov’ev–Shafranov representation by
ψ(x, y) = B0 (eη x2 + e−η y2 ). (3.18)

Downloaded from https://www.cambridge.org/core. IP address: 213.255.248.59, on 12 Feb 2025 at 06:14:27, subject to the Cambridge Core terms of use, available at
https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0022377824001648
 Application of Lagrangian techniques 7
Note that in the case of a hyperbolic axis, the plus sign from the flux function (3.18) would
be replaced by a minus sign.

3.2.2. The second variation tensor and its null eigenspace

Now that the magnetic field has been properly expressed and expanded in
Solov’ev–Shafranov coordinates, the obtained B and ∇B can be substituted in (3.4) to
expand the second variation tensor. From (3.10), we find that it is given by
⎫
d ⎪
M := M 1 + M 2, ⎪
⎪
d ⎪
⎪
⎪
⎬
M 1 = (N B − BN )B0 ,
 (3.19)
⎪
M 2 = κB0 [B cos δ − N sin δ] t + B a1 N + (a2 + u B0 )B ⎪
⎪
⎪
 ⎪
⎪
 ⎭
− N (b1 − u B0 )N + b2 B .

Observing the dyadic form of M , it can be shown that tangential components v t of

the eigenvectors v = (v t , v N , v B )T will not contribute to M v = 0. Alternatively, the
underlying argument is that the arc length parametrization of C constrains the admissible
variations δx = v, removing the dependence in the tangential component. Therefore, v t
can be absorbed by redefining v N and v B . Thus, the null eigenvectors are chosen to be of
the form v = v N N + v B B. We note that

d     
v = v N  + u v B N + v B − u v N B + κ v N cos (u − δ) + v B sin (u − δ) t.
d
(3.20)
The tangential terms ∝ t are not relevant since they do not contribute to M v = 0. The
equations for v N , v B then read
      N
0 +1 d v N −b1 −b2 v
B0 + = 0. (3.21)
−1 0 d v B a1 a2 vB

From (3.21), by periodicity of ai /B0 and bi /B0 , i = 1, 2, the system can immediately be
rewritten in the form of a periodic system:
 
dv vN
= A()v, v= . (3.22a,b)
d vB

Here, A() is periodic in . The Floquet theorem, (A6), can be applied to conclude that the
solution must be of the form
v = U ()eC /L , (3.23)

where U is a symplectic periodic matrix (with period L) and C is a constant Hamiltonian

matrix (Duignan & Meiss 2021). The eigenvalues of C are the Floquet exponents, which
must be purely imaginary near an elliptic axis and real in the hyperbolic case.
Although the system is already in a form that allows one to solve for the exponents ν by
means of (3.23), identifying the matrices C and U can be somewhat troublesome. For that
reason, (3.21) can be alternatively rewritten in the following ordinary differential equation

Downloaded from https://www.cambridge.org/core. IP address: 213.255.248.59, on 12 Feb 2025 at 06:14:27, subject to the Cambridge Core terms of use, available at
https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0022377824001648
 8 S. Guinchard, W. Sengupta and S.R. Hudson
(ODE) form:
dv N dv B d
= = , (3.24)
a1 v + a 2 v
N B b1 v + b 2 v
N B B0
which shows that the eigenvector v satisfies the general characteristic equation along the
magnetic field:
dX dY d
√ =√ =√ . (3.25)
gB1 (X , Y , ) gB2 (X , Y , ) gB3 (X , Y , )
The eigenvector v components v N , v B can be identified with X , Y , the displacements of
the magnetic field line from the closed field line r 0 along the rotated normal and binormal
directions. The expansion coordinates (x, y) ought not to be confused with (X , Y ), which
are the solutions of the characteristic ODEs (3.25). The equations for (X , Y ) are
  ⎫
1 B0 ⎪
X +
+ η X + Ω0 e Y = 0,⎪
 −η
⎪
⎬
2 B0
   (3.26)
1 B0 ⎪
⎪

Y +  +η
− η Y − Ω0 e X = 0. ⎭ ⎪
2 B0
Introducing the variables X, Y defined through
1 1
X = √ e−η/2 X(), Y = √ e+η/2 Y(), (3.27a,b)
B0 B0
which is possible since the periodicity is conserved, the system reduces to the simple
harmonic oscillator:
J0 /2
− τ + δ
  B0
X + Ω0 Y = 0, Y − Ω0 X = 0, Ω0 = , (3.28a–c)
cosh η
with ‘time-dependent’ frequency Ω0 (),  being the time-like parameter. Using the
complex variable Z = X + i Y, the system reshapes as a single complex ODE:
 
Z  − i Ω0 Z = 0 ⇒ Z() = Z0 exp i Ω0 (s) ds. (3.29)
0

Separating the periodic and non-periodic parts of the exponential, we obtain

  ⎫
⎪
Z() = Zp ()ei ν/L
, Zp () = Z0 exp i Ω̃0 (s) ds,⎪
⎪
⎬
0
 ⎪
(3.30)
1 L
⎪
⎪
Ω̄ := Ω0 (s) ds, Ω̃0 = Ω0 − Ω̄. ⎭
L 0

Comparing (3.30) with the Floquet form (3.23), we find that ν is given by
J0 (s)/2
 L 
− τ (s) + δ  (s)
B0 (s)
ν= Ω0 (s) ds (mod 2π) = ds (mod 2π) = 2πι.
0 cosh η(s)
(3.31)
This matches the expression for the rotational transform (1.3) up to a factor 2π, as
ν represents an angle and ι the number of turns that this angle constitutes. It follows

Downloaded from https://www.cambridge.org/core. IP address: 213.255.248.59, on 12 Feb 2025 at 06:14:27, subject to the Cambridge Core terms of use, available at
https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0022377824001648
 Application of Lagrangian techniques 9
that solving for the null eigenspace of the second variation tensor expanded in the
Solov’ev–Shafranov near-axis formalism yields the correct on-axis rotational transform.
The exact same approach can be followed in the hyperbolic case.
3.2.3. Derivation of the Floquet exponent from the Hill’s determinant equation
Here we pursue an alternative approach to obtain the Floquet exponent, using the theory
of the Hill’s infinite determinant. This way of solving for the Floquet exponents of a
periodic system has been known for a long time (Magnus 1953); however, to the authors’
knowledge, it is the first time that such an approach is used from the magnetic action. We
start with the system (3.26) written as
1 1
X = (a1 X + a2 Y ), Y = (b1 X + b2 Y ). (3.32a,b)
B0 B0
Eliminating Y from (3.32a,b), we obtain the following second-order ODE for X :
X  + 2C1 X  + C2 X = 0, (3.33)
where
 
a B a2 a1 1
2C1 = − 2 + 2 0 , C2 = D − , D= (a1 b2 − a2 b1 ). (3.34a–c)
a2 B0 B0 a2 B20
We can eliminate the first derivative term from (3.33) by the change of variables
   √
a2
X = exp − C1 d Ψ = X, (3.35)
B0
which leads to
Ψ  + ω2 Ψ = 0, ω2 ≡ C2 − C1 − C12 . (3.36a,b)
Equation (3.36a,b) is in the form of Hill’s equation or a Schrödinger equation with a
both X
periodic potential ω2 . We note that the linear transformation (3.35) implies that √
and Ψ have the same Floquet exponent. This is because the multiplication factor a2 /B0
is periodic in nature and therefore does not change the Floquet exponent.
Leveraging the periodicity of ω2 , ω2 ( + L) = ω2 (), we can Fourier expand

ω2 () = Ωk ei (2π/L)k , (3.37)
k∈Z

where {Ωk }k∈Z denote the Fourier coefficients of ω2 . Let the fundamental solutions of
(3.36a,b) be given by Ψ± () such that
Ψ+ (0) = 1, Ψ− (0) = 0, Ψ+ (0) = 0, Ψ− (0) = 1, (3.38a–d)
where these conditions have been chosen so that the basis functions Ψ± are linearly
independent and with unit Wronskian. The Floquet solutions are given by

Ψ+ () = e+i ν σ+ (), σ+ ( + L) = σ+ (), σ+ (0) = 1,
(3.39)
Ψ− () = e−i ν σ− (), σ− ( + L) = σ− (), σ− (0) = 0,
where σ+ and σ− satisfy the periodicity condition and the linear independence, and ν
denotes the Floquet exponents for which we seek. Nothing more needs to be known about

Downloaded from https://www.cambridge.org/core. IP address: 213.255.248.59, on 12 Feb 2025 at 06:14:27, subject to the Cambridge Core terms of use, available at
https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0022377824001648
 10 S. Guinchard, W. Sengupta and S.R. Hudson
the functions σ± since their coefficients do not appear in the final expression for the
Floquet exponents. Let us now Fourier expand the Floquet solution as
    
i ν i (2πn/L) 2πn
Ψ+ =e bn e = bn exp i  +ν , (3.40)
n∈Z n∈Z
L

where {bn }n∈Z is the set of Fourier coefficients of σ+ . The Fourier expansion of (3.36a,b)
reads

  
2πn 2
 
2nπ

Ψ+ () + ω ()Ψ+ () = −
2
bn ν + exp i  ν +
n∈Z
L L
    
  2πn
i (2π/L)k
+ Ωk e bn exp i  +ν
k∈Z n∈Z
L
  2   
  2πn 2πn
= Ωk bn−k − + ν bn exp i  +ν
n∈Z k
L L
= 0. (3.41)

In order for (3.41) to be verified, the following condition has to hold:

  2
2πn
Ωk bn−k − + ν bn = 0, n ∈ Z, (3.42)
k∈Z
L

which can be rewritten in the following matrix form, after dividing (3.42) by Ω0 −
(2πn/L + ν)2 :

Bnm bm = 0, n ∈ Z, (3.43)
m∈Z

where the matrix B is defined as

Ωn−m
Bnn = 1, Bnm =  2 . (3.44)
2πn
Ω0 − +ν
L
This yields the following determinant equation:

det Bnm = 0. (3.45)

The above determinant can be considered a function of the Floquet exponents ν:

⎛ ⎞
⎜ Ωn−m ⎟
⎜ ⎟
det Bnm (ν) := (ν) ≡ det ⎜δnm +  2 ⎟, n, m ∈ Z. (3.46)
⎝ 2πn ⎠
Ω0 − +ν
L
For (3.45) to be verified, we seek for the Floquet exponents ν such that (ν) = 0.
Following Wang, Guo & Xia (1989), we now show that the exponents can be deduced from
the very simple expression (3.50). One notes that (ν) is a 2π-periodic function, with

Downloaded from https://www.cambridge.org/core. IP address: 213.255.248.59, on 12 Feb 2025 at 06:14:27, subject to the Cambridge Core terms of use, available at
https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0022377824001648
 Application of Lagrangian techniques 11
√
poles in ν = ± Ω0 − 2πn/L, n ∈ Z. It can be shown that the determinant is absolutely
convergent in the whole ν-plane, except in these poles. Therefore,  is meromorphic.
Moreover, as the imaginary part Im(ν) → ±∞, (ν) → 1. Let us now define the complex
function f as
L  L 
f (ν) := cot (ν − Ω0 ) − cot (ν + Ω0 ). (3.47)
2 2
It is useful to introduce f since it has the same poles and periodicity as . Moreover, it is
bounded as Im(ν) → ±∞. This way, there has to exist a constant K ∈ C such that

D(ν) ≡ (ν) + Kf (ν) (3.48)

has no singularity in the whole ν plane. Together with the fact that it is bounded as |ν| →
∞, according to Liouville’s theorem, D is a constant function. In the limit |ν| → ∞, we
see that D = 1. To determine K, take ν = 0:
1 − (0) 1 − (0)
ν = 0 =⇒ K = = √ . (3.49)
f (0) Ω0 L
2 cot
2
Using the value of K from (3.49), the Floquet exponent equation reduces to
√
2 L Ω0 L
sin ν = (0) sin 2
. (3.50)
2 2
Although (3.50) is very simple, one obstacle remains to compute the exponents: one
has to evaluate the infinite determinant, (0). We refer to Wang et al. (1989) for some
approximations of (0). For instance, when Ωn are sufficiently small, (0) can be
approximated by the order-3 determinant with B00 as central element (we recall that the
determinant involves summation over all Z), providing

2Ω12 2Ω12 Ω2 Ω22

(0)  1 + + − . (3.51)
Ω0 (4 − Ω0 )2 Ω0 (4 − Ω0 )2 (4 − Ω0 )2
However, this approximation breaks down when the coefficients become too large as |n|
increases. As of the exponents computed, they might not all be suitable for ι. One has to
discard the results that are not relevant. Moreover, we emphasize that the exponents may
be shifted by 2kπ/L, with k an integer, without changing the mathematics of the system,
by periodicity. However, one has to carefully choose the adapted value for ι, by setting the
appropriate phase shift – see (1.3).
In a nutshell, the second variation of the magnetic action, at an elliptical axis, has
been expanded in the Solov’ev–Shafranov near-axis formalism. It has been shown that
the null eigenspace of this operator yields the correct on-axis rotational transform, upon
applying Floquet theory to solve for the latter. On the other hand, solving for the null
eigenspace has led to a system that could be rewritten in the form of a Hill equation. The
Floquet exponents appear as solutions of an infinite determinant equation, that can only be
approximated analytically. The derivation of the on-axis rotational transform as a Floquet
exponent of the null eigenvectors of δ 2 S has also been carried on in Mercier coordinates,
to support the previous conclusions – see Appendix D.
Magnetic confinement device design needs fast and accurate computation of ι.
Currently, the most widely used method to compute ι is field-line tracing (Todoroki 2003).

Downloaded from https://www.cambridge.org/core. IP address: 213.255.248.59, on 12 Feb 2025 at 06:14:27, subject to the Cambridge Core terms of use, available at
https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0022377824001648
 12 S. Guinchard, W. Sengupta and S.R. Hudson
Field-line tracing methods compute ι as an infinite time limit of following a field line. In
practice, it can be done by following a field line sufficiently long to achieve convergence.
As an alternative, one can also compute Greene’s residue of the axis, which is related to ι
(see Greene 1979; Hanson & Cary 1984; Hudson 2004). Even though it still involves ODE
integration, we can limit ourselves to just one circuit around the torus.
In the following section, we introduce a new method to compute ι as a solution to a
discrete problem involving the magnetic action. The action is discretized in a way similar
to that in Mackay & Meiss (1983) and Hudson & Suzuki (2014).
4. Discrete formalism: piecewise action
We have seen that the problem described in the previous section, involving the
second-order variation of the magnetic action to determine the on-axis Floquet exponent,
leads to the operator equation (3.5). In another approach, the magnetic axis can be
discretized, and the rotational transform can be determined from the multipliers of the
latter curve. The multipliers have been shown to be linked to the residue of the curve (see
Greene 1979; Mackay & Meiss 1983). This discrete approach is explored in what follows.
We consider
n−1 

S= A · d
i=1 Ci


n−1
= S(xi , xi+1 ), (4.1)
i=1

meaning that the curve has been discretized with n ∈ N points. So far, the way that Ci
are defined is not important. Only the endpoints matter. The particular case where Ci are
segments is given in Appendix E. The following notation is adopted for the derivatives:

S [i,i+1]
1 := ∇ xi S[i,i+1] = ∇ xi S(xi , xi+1 ),
(4.2)
S [i,i+1]
2 := ∇ xi+1
S [i,i+1]
= ∇ x i+1
S(xi , x i+1 ),

as well as for the second-order derivatives:

[i,i+1]
S 21 := ∇ xi+1 S [i,i+1]
1 , (4.3)

and similarly for S 12 , S 11 and S 22 . Let us also define generalized periodic orbits of type
(q), as orbits with
xi+q = xi , (4.4)
for some q ∈ N. Therefore, the magnetic axis as discretized above is a general periodic
orbit of type (n). The terminology ‘orbit’ for field lines is justified by the Hamiltonian
behaviour of the magnetic field. Moreover, the discretized action (4.1) satisfies the
periodicity condition


n−1 
n−1
S(xi+n , xi+n+1 ) = S(xi , xi+1 ) + C, (4.5)
i=1 i=1

where the constant C = 0. In the one-dimensional case, as dealt with in Mackay & Meiss
(1983), the existence of extrema of the action is ensured by an additional convexity

Downloaded from https://www.cambridge.org/core. IP address: 213.255.248.59, on 12 Feb 2025 at 06:14:27, subject to the Cambridge Core terms of use, available at
https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0022377824001648
 Application of Lagrangian techniques 13
condition on the Lagrangian, −L12 > 0. In our three-dimensional case, this can be
generalized in that the second-order derivative tensors S [i,i+1]
12 ought to be negative definite
for all i:

xT S 12 [i,i+1] x < 0, x ∈ R∗3 , 1 ≤ i ≤ n. (4.6)

The fact that C = 0 together with the condition that S [i,i+1]

12 guarantees that the action of
periodic orbits of type (n) is bounded below and the Poincaré–Birkhoff theorem (Birkhoff
1913) ensures the existence of at least two stationary trajectories among the space of all
generalized periodic paths of type (n), one that minimizes the action and is elliptic, and one
that is a saddle called minimax and is usually hyperbolic but can be alternating hyperbolic,
where hyperbolic and elliptic describe the behaviour of nearby trajectories.
For an (n)-periodic curve that extremizes the action, the latter has to be stationary with
respect to an arbitrary variation in its geometry δxi , and the stationarity condition can be
expressed in terms of the previously defined derivatives:
 
δS[δxi ] = ∇ xi S[i−1,i] + ∇ xi S[i,i+1] · δxi = 0,
(4.7)
⇔ S [i−1,i]
2 + S [i,i+1]
1 = 0.

Note that (4.7) holds for any 1 ≤ i ≤ n, and the stationarity condition is expressed for each
point in terms of the two nearest neighbours. This way, a magnetic axis can be identified.
Similarly to the continuous case dealt with in § 3, the neighbouring field lines have to
satisfy

 
∇ xi−1 S [i−1,i]
2 δxi−1 + ∇ xi+1 S [i,i+1]
1 δxi+1 + ∇ xi S [i−1,i]
2 + S [i,i+1]
1 δxi = 0, (4.8)

which we rewrite in terms of the second-order derivatives of the action as

 
[i−1,i]
S 12 δxi−1 + S [i,i+1]
21 δxi+1 + S [i−1,i]
22 + S [i,i+1]
11 δxi = 0, 1 ≤ i ≤ n. (4.9)

From Mackay & Meiss (1983), we know that the multipliers λ of an (n)-periodic orbit are
defined by the existence of a tangent orbit satisfying

δxi+n = λδxi , (4.10)

so the following holds:

δxn+1 = λδx1 , δx0 = λ−1 δxn . (4.11a,b)

They can be written in their Floquet form (Greene 1979):

λ = ei ν , (4.12)

where the Floquet exponent ν describes the average rotation angle per period of the
(n)-orbit, therefore, the rotational transform ι. Since (4.9) is valid for any 1 ≤ i ≤ n,

Downloaded from https://www.cambridge.org/core. IP address: 213.255.248.59, on 12 Feb 2025 at 06:14:27, subject to the Cambridge Core terms of use, available at
https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0022377824001648
 14 S. Guinchard, W. Sengupta and S.R. Hudson
together with (4.11a,b), it can be rewritten in tensor form:
⎛  [01]  ⎞⎛ ⎞
S 22 + S 11
[12] [12]
S 12 λ−1 S [01]
21
δx1
⎜ .. ⎟ ⎜ .. ⎟
⎜ [12]
S 21
[23]
S 12 . ⎟⎜ . ⎟
⎜ ⎟⎜ ⎟
⎜ ⎟⎜ ⎟ = 0, (4.13)
⎜ .. .. ⎟⎜ . ⎟
⎝ . . [n−1,n]
S 12 ⎠ ⎝ .. ⎠
 
λS [n,n+1]
12
[n−1,n]
S 21
[n−1,n]
S 22 + S [n,n+1]
11 δxn
where the block-tridiagonal form with corners arises naturally. Denoting by M the matrix
of second derivatives, we get an equation for the multipliers:
M (λ)δx = 0. (4.14)
The blank spaces in M are blocks of 0. Note that each block in M is of size 3 × 3
since S lm contain the second derivatives of a line integral embedded in three-dimensional
space. Thus, for an (n)-periodic curve, M ∈ M3n×3n , and δx ∈ R3n . For (4.14) to hold, the
determinant of M ought to be zero to avoid the trivial solution δx = 0.
For a block-tridiagonal matrix defined as in (4.13) with λ ∈ C, the determinant can be
expressed analytically (Molinari 2008):
 n ⎫
(−1)3n ! ⎪
⎪
det M (λ) = det ( T − λ I ) det S [i, i + 1] , ⎪
⎪
(−λ) 3 S 6 12 ⎪
⎬
i=1
n   ⎪ (4.15)
! [i,i+1] −1 [i,i+1] [i−1,i] ⎪
− S −1 ( + ) − ⎪
[i−1,i] [i,i+1]
S S S S
TS = 12 22 11 12 12 ,⎪
⎪
⎭
I3 0
i=1

which requires one to define the so-called transfer matrix T S . Our generalized convexity
condition (4.6) that ensures the existence of minimizing orbits guarantees that the
determinant of the transfer matrix can be computed as S [i,i+1] 12 is invertible ∀i. The
multipliers λ are then given by the solutions of
λ−3 det (T S − λI ) = 0, (4.16)
so they are the non-zero eigenvalues of T S . Once the multipliers have been computed, the
Floquet exponent can be easily determined from (3.31).
Alternatively, note that (4.13) gives a recursive relation for δxi :
 [01]  ⎫
S 22 + S 11 δx1 + S 12 δx2 + λ−1 S 21 δxn = 0,
[12] [12] [01]
⎪
⎪
 [n−1,n]  ⎬
λS 12
[n,n+1]
δx1 + S 21 [n−1,n]
δxn−1 + S 22 + S 11[n,n+1]
δxn = 0, (4.17)
  ⎪
⎪
⎭
[k−1,k]
S 21 δxk−1 + S 22 [k−1,k]
+ S 11 [k,k+1]
δxk + S 12[k,k+1]
xk+1 = 0; 2 ≤ k ≤ n − 1.
Equation (4.17) can be rewritten as
   
δxn+1 δx1
=J , (4.18)
δxn δx0
with
n 
! [k,k+1] [k,k+1]

J :=
−S −1
12 (S 22
[k−1,k]
+ S 11
[k,k+1]
) −S −1
12
[k−1,k]
S 12
. (4.19)
I 0
k=1

Downloaded from https://www.cambridge.org/core. IP address: 213.255.248.59, on 12 Feb 2025 at 06:14:27, subject to the Cambridge Core terms of use, available at
https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0022377824001648
 Application of Lagrangian techniques 15
The multipliers of a periodic orbit are the eigenvalues of the derivatives of the return map
around the orbit, J (Mackay & Meiss 1983), so upon comparison with (4.15), J = T S
confirming the result from (4.16).
The strength of this method resides in the fact that the problem is reduced to finding
the eigenvalues of a 6 × 6 matrix. In fact, the size of the operator matrix M increases
linearly with the number of discretization points, but of interest are solely the eigenvalues
of the matrix T S , whose size is 6 × 6, no matter how many discretization points have
been used. Although the number of matrices in the product J also increases linearly with
the number of discretization points, the most commonly used algorithms to determine
the eigenvalues of a matrix have a cost that is generally cubic (Francis 1961, 1962;
Kublanovskaya 1962). Since an m × m matrix has at most m distinct eigenvalues, reducing
its size is crucial for computational efficiency. The derivation of the aforementioned
results in the case where the curve of interest is discretized by piecewise linears is given
in Appendix E.
5. Conclusion
In this paper, after having introduced the rotational transform, the Hamiltonian
behaviour of toroidal magnetic fields was used to motivate the definition of a magnetic
action. The latter action served as starting point to express the on-axis rotational
transform from a novel approach. The focus has been on elliptical magnetic axes, but
this method applies to hyperbolic axes as well, and more generally to any periodic field
line.
The action and resulting properties were studied through the lens of the calculus of
variations, where variations of the curves’ geometry were performed. The first variation
led to the result that extremizing curves are magnetic field lines, enabling one to identify a
magnetic axis. The second variation sheds light on the geometry of the neighbouring field
lines. Studying the null eigenspace of the second variation enabled the derivation of the
on-axis rotational transform, with a focus on elliptical axes. The second variation, seen
as an operator, was expanded in the near-axis formalism developed by Mercier, Solov’ev
and Shafranov, to yield a system of periodic differential equations. The periodicity of the
system allowed the use of Floquet theory to solve for the eigenspace. The key result is that
the Floquet exponents of the axis were shown to match Mercier’s expression (1.3) of the
rotational transform. Additionally, solving for the null eigenspace was shown to lead to a
Hill equation, from which the Floquet exponents were expressed as solutions of an infinite
determinant equation.
Following this continuous derivation of the on-axis rotational transform through the
Floquet exponents of the null eigenspace of the second variation, a discrete approach was
introduced. It consists of discretizing the field line of interest and by linearity, to define
the action as a sum of piecewise actions. This approach was described by Mackay &
Meiss (1983) for one-dimensional Lagrangian systems. We provide a generalization of
this method as our action is based on field lines which are in essence three-dimensional.
Solving for the Floquet exponents was shown to be closely related to solving for the
multipliers of the curve, described by Mackay & Meiss (1983) and Greene (1979), and
the parallel between the two approaches was made as a consistency check. The on-axis
Floquet multiplier is computed by finding the eigenvalues of a 6 × 6 matrix, made of
second derivatives of the action. Its efficiency in comparison with field-line tracing will
be studied in future work.
Finally, this paper applies the near-axis formalism to recovering the rotational transform
from a Lagrangian approach. Although the method has been introduced for elliptical axes,
also called O-points, it can be applied to hyperbolic axes, or X-points, with the only

Downloaded from https://www.cambridge.org/core. IP address: 213.255.248.59, on 12 Feb 2025 at 06:14:27, subject to the Cambridge Core terms of use, available at
https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0022377824001648
 16 S. Guinchard, W. Sengupta and S.R. Hudson
difference being that the Floquet exponents are real, not purely imaginary; and, more
generally, to all periodic orbits. The discrete approach stands as an alternative to compute
the rotational transform instead of following field lines. We aim at using this method in
stellarator optimization.

Acknowledgements
We dedicate this paper to Bob Dewar. The authors would like to thank A. Bhattacharjee
and E.J. Paul for stimulating discussions and helpful suggestions.

Editor P. Helander thanks the referees for their advice in evaluating this article.

Funding
W.S. was supported by a grant from the Simons Foundation/SFARI (560651, AB),
and the Department of Energy Award No. DE-SC0024548. This paper is based upon
work supported by the US Department of Energy, Office of Science, Office of Fusion
Energy Sciences, and has been authored by Princeton University under Contract Number
DE-AC02-09CH11466 with the US Department of Energy.

Declaration of interests
The authors report no conflict of interest.

Data availability
The data that support the findings of this study are available from the corresponding
author upon reasonable request.

Appendix A. Floquet theory

Floquet theory was formulated by Gaston Floquet towards the end of the 19th century, in
his attempt to solve linear differential equations with periodic coefficients (Floquet 1883).
Suppose one needs to solve the linear system

ẋ = A(t)x, x(t0 ) = x0 , (A1a,b)

where A ∈ Mn×n is periodic in t with period T. Considering n linearly independent

solutions {x1 , . . . xn } of (A1a,b), it is useful to introduce the so-called fundamental matrix
X by grouping xi together:

X (t, t0 ) := (x1 ; x2 ; . . . ; xn ) , (A2)

where xi are column vectors, and the second argument has been added to specify that the
initial condition occurs at time t0 . Thus, (A1a,b) can be rewritten as

d
X (t, t0 ) = A(t)X (t, t0 ). (A3)
dt

If X (t0 , t0 ) = I , X is called the principal fundamental matrix. We now state

some well-known theorems of Floquet theory, the proofs of which can be found
in Meiss (2007).

Downloaded from https://www.cambridge.org/core. IP address: 213.255.248.59, on 12 Feb 2025 at 06:14:27, subject to the Cambridge Core terms of use, available at
https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0022377824001648
 Application of Lagrangian techniques 17
T HEOREM A.1 (Abel). The determinant of the fundamental matrix X is
 t
det X(t, t0 ) = exp trA(s) ds. (A4)
t0

Moreover, when (t0 , t) = (0, T), it can be rewritten as the product of the so-called Floquet
multipliers:
! n
det X(T, 0) = eνi T , (A5)
i=1

where νi are the Floquet exponents.

This second property simply states that the Floquet multipliers are the eigenvalues of
the monodromy matrix m := X (T, 0) of the linear system (A3).
T HEOREM A.2 (Floquet–Lyapunov). The fundamental matrix X solution of the system
(A3) is of the form
X(t, t0 ) = P(t)e(t−t0 )B , (A6)
where the matrix P is symplectic and T-periodic, with P(t0 ) = I, and B is a constant
Hamiltonian matrix, that is,
 
T T 0 −I
JB = B J , J = . (A7a,b)
I 0

Appendix B. Frenet–Serret frame

The Frenet–Serret frame is a local basis that spans the three-dimensional space R3 .
The terminology ‘local’ arises from the fact that this frame is defined locally along a
curve C ⊂ R3 . Let us define the basis vectors and some essential properties of the frame.
We refer to Duignan & Meiss (2021) as they provide the reader with a brief yet efficient
description of this frame. The original mathematical developments can be found in Frenet
(1852).
The frame is composed of the three well-known tangent, normal and binormal vectors,
defined and related to each other as follows. Provided that the curve C is described by the
vector-valued function r 0 ∈ R3 and parametrized by the arc length , the tangent vector t
is defined as
dr 0
t := = r 0 . (B1)
d
The normal vector accounts for the normalized rate of change of the tangent along the
curve: " "−1
n := t "t " . (B2)
Finally, the binormal vector is the cross product of the tangent and the normal vector:
b := t × n. (B3)
The curvature κ and the torsion τ of the curve C can be expressed from the derivatives of
r0:
  
 
r 0 × r 0 · r 
κ := |r 0 | = |t |, τ := 0
, (B4a,b)
κ2
provided that the curvature is non-vanishing. The rate of change of the Frenet–Serret
frame along the curve C is written in terms of the curvature and the torsion. The resulting

Downloaded from https://www.cambridge.org/core. IP address: 213.255.248.59, on 12 Feb 2025 at 06:14:27, subject to the Cambridge Core terms of use, available at
https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0022377824001648
 18 S. Guinchard, W. Sengupta and S.R. Hudson
expressions are the so-called Frenet–Serret formulae:
⎛ ⎞ ⎛ ⎞⎛ ⎞
d ⎝ n 0 τ () −κ() n
b ⎠ = ⎝ −τ () 0 0 ⎠⎝ b ⎠. (B5)
d t κ() 0 0 t

With the previous relations, the reader is equipped with what is necessary to delve into
near-axis expansion coordinates, based on the Frenet–Serret frame.
Appendix C. Near-axis expansion coordinates
This appendix is dedicated to the description of two sets of coordinates that have
been proven to be powerful in the expansion of operators or physical quantities in the
vicinity of field lines. The terminology near-axis expansion is used here, as in this
paper expansions are carried out around magnetic axes, but those coordinates remain
suitable for any near-field-line expansion. Appendix C.1 is dedicated to the description
of Mercier coordinates, introduced by Mercier (1964), and Appendix C.2 deals with the
Solov’ev–Shafranov coordinates, as described in Solov’ev & Shafranov (1970). They are
both closely related, and a correspondence can easily be established between the two. In
addition, they both are based on the Frenet–Serret frame, as described in Appendix B.
In what follows, the coordinate systems are introduced in the context of magnetic fields,
and each curve that is dealt with is assumed to be a field line. In addition, we emphasize
that the following two subsections do not describe new material, introduced in the above
references. The reader might want to refer to Jorge, Sengupta & Landreman (2020c) and
Sengupta et al. (2024) for more insight on how near-axis expansion has been used in the
context of plasma physics.
C.1. Mercier coordinates
Mercier’s frame {ρ, ω, t} is based on the Frenet–Serret frame {n, b, t}, and related to the
latter by a rotation of the normal and binormal vectors by a polar angle θ , which is a
purely geometric quantity, as shown in figure 1. The field line (that can be considered to
be a magnetic axis here) C ⊂ R3 and described by r 0 is assumed closed and parametrized
by the arc length  with a total length L. Following Mercier et al. (1974), the near-axis
expansion is based on the construction of a tube of radius ρ around the axis, such that any
neighbouring point can be described by ρ()ρ():

r() = r 0 () + ρ()ρ(), r 0 () = t(), (C1a,b)

where the dependence on  has been made explicit, but is omitted in what follows for the
sake of readability. However, it is made clear whenever the dependence on quantities is
not obvious. Therefore {ρ, ω, t} is a right-handed triad related to {n, b, t} by

ρ = n cos θ + b sin θ,
ω = b cos θ − n sin θ. (C2a,b)
#
It can be shown by direct computation that (ρ, ω = θ + τ d , ), with τ denoting
the torsion of the axis r 0 (see Appendix B), forms an orthogonal coordinate system with
metric
ds2 = dρ 2 + ρ 2 dω2 + h2 d2 , h = 1 − κρ cos θ, (C3a,b)
where κ denotes the curvature of C . The product κρ can be seen as a measure of the ratio
between the radius of curvature of r 0 and the radius of expansion around the latter. The

Downloaded from https://www.cambridge.org/core. IP address: 213.255.248.59, on 12 Feb 2025 at 06:14:27, subject to the Cambridge Core terms of use, available at
https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0022377824001648
 Application of Lagrangian techniques 19

F IGURE 1. Mercier’s triad {ρ, ω, t} related to the usual Frenet–Serret {n, b, t}. The solid black
line is a field line.

F IGURE 2. Solov’ev–Shafranov triad {N , B, t} related to the usual Frenet–Serret {n, b, t}.

The solid black line is a field line.

metric tensor reads ⎛ ⎞

1 0 0
gij = ⎝0 ρ2 0 ⎠. (C4)
0 0 h2
Therefore, from the previous characterization of the Mercier coordinates, a few differential
identities can be derived. Since the basis considered for Mercier coordinates is the triad
{ρ, ω, t}, the identity tensor is trivially written as
I = ρρ + ωω + tt. (C5)
By means of the metric expression (C1a,b), the gradient can be derived:
ω t
∇ = ρ∂ρ + ∂ω + ∂ , (C6)
ρ h
where the subscripts denote with respect to which coordinate the partial derivative is taken.
Moreover, some pseudo-Poisson formulae can be derived for the derivatives of ρ and ω,
to read
ρ ,ω = ω, ρ , = −t κ cos θ, ω,ω = −ρ, ω, = t κ sin θ, (C7a–d)
where the subscript comma represents partial differentiation. Finally, using (C6), the
gradient of each basis vector can be easily expressed:
κ 1 κ cos θ κ sin θ 1
∇t = tn, ∇ρ = ωω − tt, ∇ω = tt − ωρ. (C8a–c)
h ρ h h ρ
In the formalism developed by Mercier, when expanding quantities in powers of ρ, it can
be useful to have an additional ‘phase’ term δ that may simplify expressions. The latter

Downloaded from https://www.cambridge.org/core. IP address: 213.255.248.59, on 12 Feb 2025 at 06:14:27, subject to the Cambridge Core terms of use, available at
https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0022377824001648
 20 S. Guinchard, W. Sengupta and S.R. Hudson
phase is a function of the arc length  and enables one to define the so-called Mercier
angle u:

u := θ + δ = ω − τ d + δ. (C9)

In fact, the δ term describes the rotation of trajectories around the axis r 0 . One notes that,
by periodicity, δ has to satisfy δ( + L) = δ() + 2πn, with n ∈ Z. The rotation function
δ is important for the Solov’ev–Shafranov near-axis expansion as the latter is based on the
construction of ellipses around r 0 , whose rotation is naturally important.

C.2. Solov’ev–Shafranov coordinates

The Solov’ev–Shafranov coordinate system is closely related to the Mercier triad, but
differs in that the expansion is not carried out by constructing a tube of radius ρ around
the axis, but an ellipse of semi-axes varying along r 0 as depicted in figure 2. Let {N , B, t}
be the orthogonal triad related to the Frenet–Serret frame through a rotation by function δ
introduced in the Mercier formalism such that

N = n cos δ − b sin δ = ρ cos u − ω sin u, B = n sin δ + b cos δ = ρ sin u + ω cos u,

(C10a,b)
where {ρ, ω, t} form the Mercier basis, with Mercier angle u as defined in Appendix C.1.
Equation (C10a,b) shows how intrinsically related are the two frames. Again, the field line
considered, r 0 , forms a closed curve C parametrized by the arc length  and a total length
L. It is obvious that N and B depend on , enabling one to write

N  = −κ cos δt − u B, B = −κ sin δt + u N , (C11a,b)

where κ denotes the curvature of r 0 , which is a local property. Following Solov’ev &
Shafranov (1970), the position vector r of any point in the vicinity of the axis can be
expressed as
r = r 0 + xN + yB, (C12)
where x and y are expansion parameters depending on the position of the frame {N , B, t}
through . In Mercier coordinates, x = ρ cos u and y = ρ sin u. Differentiating r with
respect to the length parameter :

dr
= (x + u y)N + ( y − u x)B + ht, (C13)
d
where the prime denotes total differentiation with respect to , and r 0 = t. Additionally,
h = 1 − κ(x cos δ − y sin δ) = 1 − κρ cos θ and u = δ  − τ . It induces the following
metric:
 
ds2 = dx2 + dy2 + 2u ( y dx − x dy) d + h2 + u (x2 + y2 ) d2 .
2
(C14)

The covariant and contravariant forms gij and gij of the metric tensor for ds2 read
⎛ ⎞
1 0 u y
gij = ⎝ 0 1 −u x ⎠ (C15)
u y −u x 2 2
h + u (x + y )
2 2

Downloaded from https://www.cambridge.org/core. IP address: 213.255.248.59, on 12 Feb 2025 at 06:14:27, subject to the Cambridge Core terms of use, available at
https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0022377824001648
 Application of Lagrangian techniques 21
and ⎛ 2 ⎞
1 h + u 2 y2 −u 2 xy −u y
g = 2
ij ⎝ −u 2 xy 1 u x ⎠ (C16)
h
−u y u x 1
√
with |g| = h. A direct computation shows that the following basis vectors lead to the
same metric tensors:
⎫
u y u x 1 ⎪
e1 = ∇x = N − t, e2 = ∇y = B + t, e3 = ∇ = t, ⎬
h h h
e1 = h (∇y × ∇) = N , e2 = h (∇ × ∇x) = B, e3 = h (∇x × ∇y) ⎪
⎭
= ht + u (−xB + yN ),
(C17)
which will be useful and, in fact, more convenient in the near-axis expansion of the
quantities of interest in this paper. The gradient in the basis {ej }j is expressed as
∂
∇ ≡ ej gij (C18)
∂α i
with α 1 = x, α 2 = y, α 3 =  and the Einstein summation convention.
Appendix D. Derivation of the rotational transform from the action in Mercier
coordinates
D.1. Near-axis expansion of the second variation tensor
As we did in the Solov’ev–Shafranov geometry in the body of this paper, in order to expand
the operator
δ2S d
M := = − (I × B) + x × (∇B)T (D1)
δxδx d
in Mercier coordinates, B is expanded in the parameter ρ such that κρ 1. To evaluate M
to lowest order in ρ, the magnetic field needs to be expanded up to first order. Therefore,
we set
 
B = B0 ()t + ρB1 , B1 = Bρ1 ρ + Bω1 ω + Bt1 t . (D2a,b)
The next step in the expansion of M is to compute ∇B. It is straightforward from the
differential identity (C6):
1 
∇B = tt B0 () + tn κB0 + ρ B1 + ρ∇B, (D3)
h
such that, to leading order
  
x × ∇BT = κB0 bt + Bρ1 ω − Bω1 ρ ρ + ωω (∂ω Bρ1 − Bω1 ) − ωρ (∂ω Bω1 + Bρ1 ),
(D4)
−(I × B) = (ρω − ωρ)B0 .
Finally, M as given by (3.4) simplifies to
⎫
d ⎪
M := M 1 + M 2, ⎪
⎪
d ⎬
M 1 = (ρω − ωρ)B0 , (D5)
⎪
⎪
 ρ  ρ ρ ⎭
⎪
ω ω ω
M 2 = κB0 bt + B1 ω − B1 ρ ρ + ωω (∂ω B1 − B1 ) − ωρ (∂ω B1 + B1 ).
So far, the power expansion of the gradient of the magnetic field relied on a generic
expression of B in terms of Bρ1 , Bω1 and Bt1 . The latter functions can be determined enforcing

Downloaded from https://www.cambridge.org/core. IP address: 213.255.248.59, on 12 Feb 2025 at 06:14:27, subject to the Cambridge Core terms of use, available at
https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0022377824001648
 22 S. Guinchard, W. Sengupta and S.R. Hudson
additional constraints such as ∇ · B = 0 and ∇ × B = J 0 = J0 ()t. The divergence and
curl of B can be evaluated directly from the ∇B tensor, using dyadic algebra. From

∇ · B = I : ∇B = B0 + 2Bρ1 + ∂ω Bω1 ,
(D6)
∇ × B = −I ×.∇B = t (2Bω1 − ∂ω Bρ1 ) + (ω − ρ∂ω ) (B0 κ cos θ − Bt1 ),

where ∂ω := ∂/∂ω, it can be easily shown that

1 1
Bt1 = κB0 cos θ = B1t , Bρ1 = − (B0 + ∂ω b1 ) = B1ρ , Bω1 = J0 + b1 = B1ω ,
2 2
(D7a–c)
where b1 satisfies the harmonic equation
 
∂ω2 + 4 b1 = 0. (D8)

Following Mercier et al. (1974), the solution of (D8) can be expressed as

 
b1  J0 η
= bc2 cos (2u) + bs2 sin (2u) = tanh η() δ − τ + cos (2u) + sin (2u),
B0 2B0 2
(D9)
where u is the Mercier angle as introduced in Appendix C.1. The functions η() and
δ() represent, respectively, the eccentricity and the rotation of the elliptic flux surfaces
winding around the axis, given by the flux function

ψ = ρ 2 B0 [cosh(η) + sinh(η) sin (2u)] . (D10)

D.2. Floquet exponents and rotational transform

Now that the magnetohydrodynamic constraints have been enforced, let us obtain the
null eigenvector v = v t t + v ρ ρ + v ω ω of M such that M v = 0, in order to determine the
Floquet exponents. We take advantage of the orthogonality of Mercier coordinates to work
with the covariant representation of v, so the notation is less heavy. The metric tensor
for this set of coordinates is given by (C4). We state that the components vt , vρ , vω are
functions of ω and . The M 1 terms can be shown to be

d 
M1 v = B0 ρ (vω − κ sin θ vt ) − ω(vρ + κ cos θ vt ) . (D11)
d

Similarly,

M 2v = ρ B0 vω + vω (B0 + B1ρ ) − B1ω vρ

+ ω −B0 vρ + vω (B1ω − J0 ) + B1ρ vρ . (D12)

Thus, we observe that M v = 0 only has ρ, ω components with  derivatives of vρ , vω . The

tangential component vt is thus an arbitrary constant, which can be absorbed by redefining

Downloaded from https://www.cambridge.org/core. IP address: 213.255.248.59, on 12 Feb 2025 at 06:14:27, subject to the Cambridge Core terms of use, available at
https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0022377824001648
 Application of Lagrangian techniques 23
vρ , vω . We therefore set vt = 0. The condition M v = 0 leads to the system
      
0 +1 d vρ B1ω [B0 + B1ρ ] vρ
B0 + = 0. (D13)
−1 0 d vω B1ρ [B1ω − J0 ] vω

Substituting the constraints (D7a–c) into (D13), we obtain the equivalent system of
coupled linear partial differential equations:
     ⎫
B0 ∂ω b1 J0 /2 b1 ⎪
vω + − vω − + vρ = 0,⎪
⎪
⎬
2B0 2B0 B0 B0
     (D14)
B0 ∂ω b1 J0 /2 b1 ⎪
⎪

vρ + + vω + − vω = 0.⎪
⎭
2B0 2B0 B0 B0

Therefore, the the system (D14) can be represented in the form

dṽ
= A()ṽ, ṽ = (vρ , vω )T . (D15)
d
Here, A is a matrix with components periodic in  with period L. Note that this
representation is possible since all the quantities between brackets are depending on 
along the field line from which the expansion is carried out, ω included (Solov’ev &
Shafranov 1970; Duignan & Meiss 2021). We can directly use the Floquet theorem on
(D15) to conclude that the solution must be of the form

ṽ = U ()eC /L , (D16)

where U is a symplectic periodic matrix and C is a constant Hamiltonian matrix. The

eigenvalues of C are the Floquet exponents, which must be of the form ±i ν (ν ∈ R) near
an elliptic axis.
An equivalent approach to solve the system from (D14) and hence identify the exponents
ν is to use the fact that b1 only has second harmonics in u. This allows us to seek a solution
where ṽ only has first harmonics in u. From now on, we follow this approach. We show
that there exist solutions to M v = 0 with
     
vρ () vρc () vρs ()
= cos u + sin u. (D17)
vω () vωc () vωs ()

Substituting the above into (D14), together with the expressions for B1ρ , B1ω from (D7a–c),
yields first- and third-order harmonics in u. Equating the terms with first harmonics in u,
the vω equation yields
  ⎫
J0 /2 B0 1 ⎪
 
vωc − (τ − δ )vωs − vρc + vωc + (ac2c + as2s ) = 0,⎪⎪
⎬
B0 2B0 2
  (D18)
J0 /2 B0 1 ⎪
⎪
 
vωs + (τ − δ )vωc − vρs + ⎪
vωs + (as2c − ac2s ) = 0. ⎭
B0 2B0 2

Equating the third harmonic terms leads to the following constraints:

ac2c = as2s , as2c = −ac2s , (D19a,b)

Downloaded from https://www.cambridge.org/core. IP address: 213.255.248.59, on 12 Feb 2025 at 06:14:27, subject to the Cambridge Core terms of use, available at
https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0022377824001648
 24 S. Guinchard, W. Sengupta and S.R. Hudson
where the a terms have the following definitions:
⎫
bc2 bs2 bs2 bc2
ac2c = − vρc − vωc , as2s = − vρs + vωs ,⎪
⎪
⎬
B0 B0 B0 B0
(D20)
⎪
vωs .⎪
bs2 bc2 bc2 bs2
as2c = − vρc + vωc , ac2s = − vρs − ⎭
B0 B0 B0 B0
The constraint (D19a,b) and the definitions (D20) imply, together with b2s2 + b2c2 = 0, that

vρs = vωc , vρc = −vωs , (D21a,b)

which allows us to rewrite (D18) solely in terms of vωc , vωs . Simplification leads to
     ⎫
B0 bs2 J0 /2 bc2 ⎪

vωc + − vωc + −τ +δ + 
vωs = 0,⎪⎪
⎬
2B0 B0 B0 B0
     (D22)
B0 bs2 J0 /2 bc2 ⎪
⎪

vωs + + vωs − − τ + δ − vωc = 0. ⎪
⎭
2B0 B0 B0 B0
Finally, substituting (D9), we get
     ⎫
B0 η e+η J0 /2 ⎪

vωc + − vωc + − τ + δ vωs = 0,⎪

⎪
⎬
2B0 2 cosh (η) B0
     (D23)
B0 η e−η J0 /2 ⎪
⎪

vωs + + vωs − − τ + δ  vωc = 0. ⎪
⎭
2B0 2 cosh (η) B0
It is possible to further simplify the system by introducing new variables X, Y:
1 1
vωc =   e+η/2 X(), vωs =   e−η/2 Y() (D24a,b)
B0 B0

such that (D23) reduces to

J0 /2
− τ + δ
B0
X  + Ω0 ()Y = 0, Y  − Ω0 ()X = 0, Ω0 () = . (D25a–c)
cosh η
One notes that (D24a,b) has the exact same form as (3.27a,b), which describe a harmonic
oscillator system with a ‘time’-dependent frequency Ω0 . Once again, using the complex
variable Z = X + iY, the system can be rewritten as a single complex ODE:
 
Z  − i Ω0 Z = 0 ⇒ Z() = Z0 exp i Ω0 (s) ds. (D26)
0

Separating the periodic and non-periodic parts of the exponential we get

  ⎫
⎪
Z() = Zp ()ei ν/L
, Zp () = Z0 exp i Ω̃0 (s) ds,⎪
⎪
⎬
0
 ⎪
(D27)
1 L
⎪
⎪
Ω̄ ≡ Ω0 (s) ds, Ω̃0 = Ω0 () − Ω̄. ⎭
L 0

Downloaded from https://www.cambridge.org/core. IP address: 213.255.248.59, on 12 Feb 2025 at 06:14:27, subject to the Cambridge Core terms of use, available at
https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0022377824001648
 Application of Lagrangian techniques 25
Comparing (D27) with (D16), we find that ν, given by
J0 (s)/2
 L  − τ (s) + δ  (s)
B0 (s)
ν= Ω0 (s) ds (mod 2π) = ds (mod 2π), (D28)
0 cosh η(s)
is the Floquet exponent for the system. It confirms the result obtained in the
Solov’ev–Shafranov coordinates, and it is a second confirmation that the second variation
of the magnetic field action, when expanded by means of a near-axis formalism and
combined with Floquet theory, yields the correct on-axis rotational transform. This was
expected since the result is independent of coordinates.
Appendix E. Discrete formalism: the piecewise linear discretization
A discrete method to compute the on-axis Floquet exponent has been introduced in § 4.
Here, we give the computations in the particular case of the curve being broken down to a
concatenation of segments. Recall that the discrete action was introduced as
n−1 
S= A · d. (E1)
i=1 Ci

Thus, taking Ci to be segments, the action sums up to a sum of integrals along piecewise
linears Ci := {R3  x = ζ (xi+1 − xi ) + xi | ζ ∈ [0, 1]}:
⎫

n−1
⎪
⎪
S= S(xi , xi+1 ), ⎪
⎪
⎪
⎪
⎪
⎪
i=1
 1 ⎪
⎬
S(xi , xi+1 ) = dζ A (ζ (xi+1 − xi ) + xi ) · (xi+1 − xi )⎪ (E2)
⎪
⎪
0 ⎪
⎪
 1 ⎪
⎪
⎪
⎪
= dζ A (v(x(ζ ))) · u, ⎭
0

with x(0) = xi and x(1) = xi+1 , and the vector functions u and v defined as follows:

u(xi , xi+1 ) = xi+1 − xi ,
(E3)
v(ζ, xi , xi+1 ) = ζ (xi+1 − xi ) + xi ,
such that ∇ xi u(xi , xi+1 ) = −I , ∇ xi+1 u(xi , xi+1 ) = I , ∇ xi v(ζ, xi , xi+1 ) = (1 − ζ )I , ∇ xi+1 v(ζ,
xi , xi+1 ) = ζ I . Since
∇ xi [A(v(xi )) · u(xi )] = J TA◦v (xi )u(xi ) + J Tu (xi )A(v(xi ))

= J Tv (xi )J TA (v(xi )) u(xi ) + J Tu (xi )A(v(xi )), (E4)
where J Ta stands for the transpose of the Jacobian matrix of the vector field a, one gets the
following first-order derivatives of the action:
 1 ⎫
⎪
S1
[i,i+1]
= (1 − ζ )J A (ζ (xi+1 − xi ) + xi ) (xi+1 − xi ) − A (ζ (xi+1 − xi ) + xi ) dζ,⎪
T
⎪
⎬
0
 ⎪
1
⎪
S [i−1,i]
2 = ζ J TA (ζ (xi − xi−1 ) + xi−1 ) (xi − xi−1 ) + A (ζ (xi − xi−1 ) + xi−1 ) dζ. ⎪
⎭
0
(E5)

Downloaded from https://www.cambridge.org/core. IP address: 213.255.248.59, on 12 Feb 2025 at 06:14:27, subject to the Cambridge Core terms of use, available at
https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0022377824001648
 26 S. Guinchard, W. Sengupta and S.R. Hudson
A direct computation of the second derivatives reads
 ⎫
1  ⎪
[i−1,i]
S 12 = dζ ζ (1 − ζ )∇ v J A (v)(xi − xi−1 ) − J A (v) + (1 − ζ )J A (v), ⎪
T T
⎪
⎪
⎪
⎪
0
 ⎪
⎪
 ⎪
⎪
1
⎪
⎪
[i,i+1]
S 21 = dζ (1 − ζ ) ζ ∇ v J A (v)(xi+1 − xi ) + J A (v) + ζ J A (v),
T T
⎪
⎬
0
 ⎪
(E6)
1    ⎪
⎪
[i−1,i]
S 22 = T
dζ ζ ζ ∇ v J A (v)(xi − xi−1 ) + J A (v) + J A (v) , T
⎪
⎪
⎪
⎪
0
 ⎪
⎪
   ⎪
⎪
1
⎪
[i,i+1]
S 11 = dζ (1 − ζ ) (1 − ζ )∇ v J A (v)(xi+1 − xi ) − J A (v) + J A (v) .⎪
T T
⎭
0

The second-order derivatives above can be easily computed numerically. We recall that
the multipliers λ of the curve are given by solutions of
 n ⎫
! ⎪
⎪
det M (λ) = (−1)n λ−3 det (T S − λI 6 ) det S 12 [i, i + 1] , ⎪ ⎪
⎪
⎬
i=1
n   ⎪ (E7)
! [i,i+1] −1 [i,i+1] [i−1,i] ⎪
− S −1 ( + ) − ⎪
[i−1,i] [i,i+1]
S S S S
TS = 12 22 11 12 12 ,⎪
⎪
⎭
I3 0
i=1

and that they are linked to the Floquet exponent by λ = ei ν . From the set of expressions
(E6), the determinant of M can be computed, leading to the multipliers. The above can be
implemented numerically as a novel tool to compute the on-axis Floquet exponent.

REFERENCES

A RNOL ’ D , V.I. 1963 Proof of a theorem of A. N. Kolmogorov on the invariance of quasi-periodic motions
under small perturbations of the Hamiltonian. Russ. Math. Surv. 18 (5), 9.
A RNOLD , V.I. 1978 Mathematical methods of Classical Mechanics. Springer.
A RNOLD , V.I. 2014 The Asymptotic Hopf Invariant and Its Applications, pp. 357–375. Springer.
B IRKHOFF , G.D. 1913 Proof of Poincaré’s geometric theorem. Trans. Am. Math. Soc. 14 (1), 14–22.
B ROUWER , L.E.J. 1910 Über eineindeutige, stetige transformationen von flächen in sich. Math. Ann. 69
(2), 176–180.
C ARY, J.R. & L ITTLEJOHN , R.G. 1983 Noncanonical Hamiltonian mechanics and its application of
magnetic field line flow. Ann. Phys. 151(1), 1–34.
D UIGNAN , N. & M EISS , J.D. 2021 Normal forms and near-axis expansions for Beltrami magnetic fields.
Phys. Plasmas 28 (12), 122501.
F LOQUET, G. 1883 Sur les équations différentielles linéaires à coefficients périodiques. Ann. Sci. École
Norm. Sup. 12, 47–88.
F RANCIS , J.G.F. 1961 The QR transformation a unitary analogue to the LR transformation. Part 1. Comput.
J. 4 (3), 265–271.
F RANCIS , J.G.F. 1962 The QR transformation. Part 2. Comput. J. 4 (4), 332–345.
F RENET, F. 1852 Sur les courbes à double courbure. J. Math. Pures Appl. 17, 437–447.
GARREN , D.A. & BOOZER , A.H. 1991 Existence of quasihelically symmetric stellarators. Phys. Fluids B
3 (10), 2822–2834.
G REENE , J.M. 1979 A method for determining a stochastic transition. J. Math. Phys. 20 (6), 1183–1201.
H ANSON , J.D. & CARY, J.R. 1984 Elimination of stochasticity in stellarators. Phys. Fluids 27 (4), 767.
H ELANDER , P. 2014 Theory of plasma confinement in non-axisymmetric magnetic fields. Rep. Prog. Phys.
77 (8), 087001.

Downloaded from https://www.cambridge.org/core. IP address: 213.255.248.59, on 12 Feb 2025 at 06:14:27, subject to the Cambridge Core terms of use, available at
https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0022377824001648
 Application of Lagrangian techniques 27
H UDSON , S.R. 2004 Destruction of invariant surfaces and magnetic coordinates for perturbed magnetic
fields. Phys. Plasmas 11 (2), 677–685.
H UDSON , S.R. & D EWAR , R.L. 2009 Are ghost surfaces quadratic-flux-minimizing? Phys. Lett. A 373
(48), 4409–4415.
H UDSON , S.R. & S UZUKI , Y. 2014 Chaotic coordinates for the Large Helical Device. Phys. Plasmas 21
(10), 102505.
J ORGE , R., S ENGUPTA , W. & L ANDREMAN , M. 2020a Construction of quasisymmetric stellarators using
a direct coordinate approach. Nucl. Fusion 60 (7), 076021.
J ORGE , R., S ENGUPTA , W. & L ANDREMAN , M. 2020b Near-axis expansion of stellarator equilibrium at
arbitrary order in the distance to the axis. J. Plasma Phys. 86 (1), 905860106.
J ORGE , R., S ENGUPTA , W. & L ANDREMAN , M. 2020c Near-axis expansion of stellarator equilibrium at
arbitrary order in the distance to the axis. J. Plasma Phys. 86 (1), 905860106.
KOLMOGOROV, A.N. 1954 On conservation of conditionally periodic motions for a small change in
Hamilton’s function. Dokl. Akad. Nauk SSSR 98, 527–530.
K RUSKAL , M.D. & K ULSRUD , R.M. 1958 Equilibrium of a magnetically confined plasma in a toroid.
Phys. Fluids 1 (4), 265–274.
K UBLANOVSKAYA , V.N. 1962 On some algorithms for the solution of the complete eigenvalue problem.
USSR Comput. Maths Math. Phys. 1 (3), 637–657.
L ANDREMAN , M. & S ENGUPTA , W. 2018 Direct construction of optimized stellarator shapes. Part 1.
Theory in cylindrical coordinates. J. Plasma Phys. 84 (6), 905840616.
L ANDREMAN , M. & S ENGUPTA , W. 2019 Constructing stellarators with quasisymmetry to high order.
J. Plasma Phys. 85 (6), 815850601.
L ICHTENBERG , A.J. & L IEBERMAN , M.A. 1992 Regular and Chaotic Dynamics, 2nd edn. Springer.
M ACKAY, R.S. & M EISS , J.D. 1983 Linear stability of periodic orbits in lagrangian systems. Phys. Lett.
A 98 (3), 92–94.
M AGNUS , W. 1953 Infinite Determinants in the Theory of Mathieu’s and Hill’s Equations, vol. 1.
Mathematics Research Group, Washington Square College of Arts and Science.
M EISS , J.D. 1992 Symplectic maps, variational principles, and transport. Rev. Mod. Phys. 64, 795–848.
M EISS , J.D. 2007 Differential Dynamical Systems. SIAM.
M ERCIER , C. 1964 Equilibrium and stability of a toroidal magnetohydrodynamic system in the
neighbourhood of a magnetic axis. Nucl. Fusion 4 (3), 213.
M ERCIER , C., E URATOM & C OMMISSION OF THE E UROPEAN C OMMUNITIES 1974 Lectures in
Plasma Physics: The Magnetohydrodynamic Approach to Plasma Confinement in Closed Magnetic
Configurations. EUR 5127 EN/1987. Commission of the European Communities.
M OLINARI , L.G. 2008 Determinants of block tridiagonal matrices. Linear Algebra Appl. 429 (8–9),
2221–2226.
M OSER , J. 1962 On invariant curves of area-preserving mappings of an annulus. Nachr. Akad. Wiss.
Göttingen, Math. Phys. Kl. II 1,1 Kl (1), 1.
M OSER , J. 1973 Stable and Random Motions. Princeton Univ. Press.
P FEFFERLÉ , D., G UNDERSON , L., H UDSON , S.R. & N OAKES , L. 2018 Nonplanar elasticae as optimal
curves for the magnetic axis of stellarators. Phys. Plasmas 25 (9), 092508.
RODRÍGUEZ , E., S ENGUPTA , W. & B HATTACHARJEE , A. 2023 Constructing the space of quasisymmetric
stellarators through near-axis expansion. Plasma Phys. and Control. Fusion 65 (9), 095004.
S ENGUPTA , W., RODRIGUEZ , E., J ORGE , R., B HATTACHARJEE , M. & L ANDREMAN , A. 2024
Stellarator equilibrium axis-expansion to all orders in distance from the axis for arbitrary plasma
beta.. J. Plasma Phys. 90 (4). doi:10.1017/S002237782400076X
S OLOV ’ EV, L.S. & S HAFRANOV, V.D. 1970 Reviews of Plasma Physics, vol. 5. ed. M. A. Leontovich.
Consultants Bureau.
S PITZER , L. J R . 1958 The stellarator concept. Phys. Fluids 1 (4), 253–264.
T ODOROKI , J. 2003 Calculating rotational transform following field lines. J. Plasma Fusion Res. 79 (4),
321–322.
WANG , Z.X., G UO , D.R. & X IA , X.J. 1989 Special Functions. World Scientific Pub Co Inc.

Downloaded from https://www.cambridge.org/core. IP address: 213.255.248.59, on 12 Feb 2025 at 06:14:27, subject to the Cambridge Core terms of use, available at
https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0022377824001648