Beyond-all-order asymptotics for homoclinic snaking of localised

patterns in reaction-diffusion systems.

Edgardo Villar-Sepúlveda
School of Engineering Mathematics and Technology
University of Bristol
arXiv:2501.02698v1 [nlin.PS] 6 Jan 2025

Abstract
Spatially localised stationary patterns of arbitrary wide spatial extent emerge from subcritical Turing
bifurcations in one-dimensional reaction-diffusion systems. They lie on characteristic bifurcation curves
that oscillate around a Maxwell point in a homoclinic snaking phenomenon. Here, a generalisation
of the exponential asymptotics method by Chapman & Kozyreff is developed to provide leading-order
expressions for the width of the snaking region close to a Turing bifurcation’s super/sub-critical transition
in arbitrary n-component reaction-diffusion systems. First, general expressions are provided for the
regular asymptotic approximation of the Maxwell point, which depends algebraically on the parametric
distance from the codimension-two super/sub-critical Turing bifurcation. Then, expressions are derived
for the width of the snaking, which is exponentially small in the same distance. The general theory is
developed using a vectorised form of regular and exponential asymptotic expansions for localised patterns,
which are matched at a Stokes line. Detailed calculations for a particular example are algebraically
cumbersome, depend sensitively on the form of the reaction kinetics, and rely on the summation of a
weakly convergent series obtained via optimal truncation. Nevertheless, the process can be automated,
and code is provided that carries out the calculations automatically, only requiring the user to input the
model, and the values of parameters at a codimension-two bifurcation. The general theory also applies
to higher-order equations, including those that can be recast as a reaction-diffusion system. The theory
is illustrated by comparing numerical computations of localised patterns in two versions of the Swift-
Hohenberg equation with different nonlinearities, and versions of activator-inhibitor reaction-diffusion
systems.

1 Introduction
The spontaneous formation of spatially periodic patterns through a finite wavenumber instability of systems
of reaction-diffusion equations arises in several areas of applied mathematics, beginning with Alan Turing’s
seminal work on the chemical basis of morphogenesis [1]. See also [2, 3], for recent reviews and updates. If
the Turing bifurcation is subcritical, giving rise to unstable periodic patterns that subsequently restabilise,
then there should be a family of transverse heteroclinic connections between the background steady state
and stable periodic patterns [4, 5, 6]. Unfolding this family leads to localised structures of arbitrarily wide
spatial extent that are formed by homoclinic connections to the background with arbitrary oscillations
near the periodic pattern [7]. These connections organise themselves in what has been dubbed a snaking
bifurcation diagram as depicted in Figure 1. See, for example, [8, 4, 9, 10, 11] and references therein.
Beck et al [12] provided proof of the existence and properties of the snaking bifurcation diagram in a
large class of reversible systems under generic assumptions on the existence of the heteroclinic connection.
Computational methods can find much of the fine structure of homoclinic snaking in examples, but in general
analytical descriptions of the localised patterns remain problematic, even in one spatial dimension. See [13]
for a review of the state of the art in one or more spatial dimensions.
One possibility to get an analytical handle on localised pattern formation due to homoclinic snaking is in
the asymptotic limit of a codimension-two super/sub-critical Turing bifurcation. From such a bifurcation,
there emerges a Maxwell point at which the background-to-periodic heteroclinic connection exists. Unfortu-
nately, it is known that the normal form of such a bifurcation, up to any algebraic order, is integrable and
so any heteroclinic connection is non-transverse and there can be no homoclinic snake within the normal

1
Figure 1: Qualitative description of the snake of localised patterns snaking around a Maxwell point in a one-parameter
diagram. See text for details. The inset shows a two-parameter diagram where the magenta curves represent loci of
the outermost folds of the snake, surrounding the Maxwell point (black curve) arising from a codimension-two Turing
bifurcation (the point of transition between red and blue lines, which represent supercritical and subcritical Turing
bifurcation lines, respectively). The asymptotic scaling of the width between the two magenta curves as we approach
the codimension-two point is what the analysis in the paper seeks to explain. While intended to be qualitative, the
numerical curves are taken from the Swift-Hohenberg 2-3 studied in Section 8.

form (see e.g. [14, 7]). Therefore, the snaking region must emerge beyond all algebraic orders in parameters
that unfold the normal form. The computation of this beyond-all-orders width about the Maxwell point, for
arbitrary reaction-diffusion systems posed on the real line forms the subject of the present paper.
At present, this beyond-all-orders computation has only been attempted in a few special cases [5, 6, 15].
A canonical example of pattern formation is the fourth-order Swift-Hohenberg partial differential equation
(PDE). As we shall see, such higher-order equations can be brought into the framework of the class of reaction-
diffusion equations studied in this paper, provided we allow for a singular temporal evolution operator, which
is fine because we are primarily concerned with the existence of stationary patterns here, not their dynamics.
This paper is motivated by the groundbreaking work by Chapman & Kozyreff [5, 16] who used exponential
asymptotics to study homoclinic snaking in the Swift-Hohenberg equation with quadratic-cubic nonlinearities
posed on the line. They used matching of inner and outer expansions of exponentially growing and decaying
solutions, upon crossing a Stokes line in the complex plane. The problem is reduced to the computation
of a single coefficient of an exponentially small amplitude, which can be achieved through an iteration
scheme. They found the iteration to be slowly converging and instead simply fit this coefficient to numerical
results. Later, Dean et al. [6] applied the same technique to a version of the Swift-Hohenberg with cubic-
quintic nonlinearities. The additional odd symmetry made the calculation slightly more straightforward and
they were able to make their iteration scheme converge to an amplitude that matched extremely well with
numerical results. More recently, the same method was implemented by De Witt [15] for a modified version
of the Schnakenberg system, which is a reaction-diffusion equation with two components.

1.1 Problem description

We pose a general reaction-diffusion system on the line as:

M ∂t u = f (u; a, b) + D̂ ∂xx u, x ∈ R, (1)

2
where u = u(x, t) ∈ Rn , n ≥ 2, is a vector of real functions that are assumed to be well-defined for all
(x, t) ∈ R × R+ 0 , D̂ is a diffusion matrix, which can be any square matrix as long as (1) is well-posed and
the assumptions below hold, a, b ∈ R are parameters of the system, and M is the mass matrix of the system,
which can be any square matrix, as long as (1) is well-posed. Note that writing the system in this way allows
the analysis to be applied to systems of n ≥ 1 equations containing higher, even order spatial derivatives. For
example, a fourth-order equation ∂t u1 = ∂xxxx u1 can be written as (1) simply by adding the extra equation
∂xx u1 − u2 = 0.
In this article, we aim to generalize the method of exponential asymptotics to study homoclinic snaking
near codimension-two Turing bifurcation points under the assumption that such a phenomenon occurs.
Specifically, we will find conditions for the existence of Maxwell points in which the homogeneous steady
state has the same energy as a large-amplitude periodic orbit and construct approximations of the localized
structures that appear close to such points.
We assume that (1) has a homogeneous steady state P = P(a, b) such that f (P; a, b) = 0 for all a, b ∈ R,
with Ju f (P; a, b), the Jacobian matrix of f (u; a, b) at u = P, being an invertible matrix.
Furthermore, we assume that P goes through a codimension-two Turing bifurcation point when (a, b) =
(a0 , b0 ). That is, there exists k ∗ > 0, the wavenumber, such that
 
2
det Ju f (P; a0 , b0 ) − (k ∗ ) D̂ = 0,
∂   
det Ju f (P; a0 , b0 ) − k 2 D̂ = 0, and
∂k k=k∗
  
2
dim ker Ju f (P; a0 , b0 ) − (k ∗ ) D̂ = 1.

Furthermore, we assume that

 
det Ju f (P; a0 , b0 ) − k 2 D̂ ̸= 0, (2)

for all k ∈ R \ {k ∗ }; and the existence of localized structures close to this codimension-two bifurcation point.
Without loss of generality, we assume that P = 0 for all a, b. If not, we can make the change u → u − P,
which translates P to the origin. Furthermore, for the same reason, without loss of generality, we assume
that (a0 , b0 ) = (0, 0). Moreover, for simplicity of notation, we will omit the dependence on a and b and drop
the star sign of the critical wavenumber k ∗ when there is no place for confusion.
The process we are about to follow comprises the following steps:
1. Obtain an expression for the amplitude of a localized solution close to a codimension-two Turing
bifurcation point (Section 2).
2. Expand the system about the singularities of such amplitudes, which involves the computation of inner
and outer solutions (Sections 3 and 4).

3. Determine the order of n at which we truncate our asymptotic expansion optimally and study the
equation for the remainder (Sections 5 and 6).
4. Find conditions that let us ensure that the remainder tends to zero as its corresponding independent
variable tends to ±∞, which ensures that our expansion will be well-defined, providing us with an
estimation for the width of the homoclinic snaking (end of Section 6).

1.2 Outline
The outline of this article is as follows. Section 2 starts by introducing the standard notation used throughout
the article, and computes a regular asymptotic expansion up to fifth order of the kind of solutions we are
interested in. For later use, it is required to go up to seventh order; those results are presented in Appendix
B. Sections 3 and 4 consider the behaviour of the asymptotic expansion as the order tends to ∞, with the
two sections dealing respectively with late-term expansion in an inner and an outer region. Section 5 then
develops the equation for the remainder up to fifth order after truncating the expansion at a high order,

3
without the forcing due to truncation. As for the amplitude equation, it will be required to develop the
equation for the remainder up to seventh order; those results are presented in Appendix C. The effect of
this forcing is considered in Section 6, which gives a final condition in order to determine the width of the
snaking. Next, Section 7 shows how to match solutions at the boundaries of the domains where different
expansions are valid in order to join fronts. To illustrate the theory, Appendix A shows what happens in a
case in which a generic assumption does not hold, and Section 8 presents results from several examples in
which this assumption holds or does not hold. In each case, we evaluate the specific regular and beyond-all-
orders coefficients and compare the results with numerical computations. Note that this evaluation step is
computationally cumbersome, but we provide the reader with code in an associated GitHub repository [17],
which can be used to replicate the results, as well as compute the snaking width for other examples. Finally,
Section 9 presents a brief discussion and suggests avenues for future work.

2 Regular asymptotics
We start the analysis of our system by performing a regular asymptotic expansion. This process has been
carried out in different ways in the past in order to study the criticality of Turing bifurcations (see, e.g. [18]).
However, near a codimension-two Turing bifurcation point, localized solutions are expected to appear apart
from the periodic states. To study such solutions, we need to obtain a general and accurate expression for
the amplitude of different solutions that arise near a codimension-two Turing bifurcation. As we explain
below, such an expression will have singularities that need to be studied to control the convergence of the
solution we aim to build through this process.
We start by introducing some notation to be used throughout this article.

2.1 Including parameters and kinetics in a general way

First, we scale x → x/k and expand (1) as:
M M
∂u X a X b
∂2u
M = ai bj fi,j (u) + k 2 D̂ , (3)
∂t i=0 j=0
∂x2

where Ma , Mb > 0 are the degrees of the right-hand side of (3) in terms of a and b, respectively (they could
be infinite) Here, we assume that for every i, j, fi,j is an analytic function on u, which implies that we can
expand it using Taylor series up to order infinity. In particular, if we Taylor-expand each function fi,j around
u = 0, the system becomes
Ma X Mb n
!m
∂u X i j
X 1 X ∂ ∂2u
M = a b ur fi,j (0) + k 2 D̂ .
∂t i=0 j=0
m! r=1 ∂ur ∂x2
m≥1

∂u
We are interested in the study of stationary solutions, so we assume that = 0. Let 0 < ε ≪ 1 be a small
∂t
2
positive number. Define a long spatial variable by X = ε x. Thus, the system becomes
Ma XMb n
!m  2 2 2
X 1 
X
i j
X ∂ ∂ u 2 ∂ u 4 ∂ u
0= a b ur fi,j (0) + k 2 D̂ + 2 ε + ε . (4)
i=0 j=0
m! r=1 ∂ur ∂x2 ∂x∂X ∂X 2
m≥1

Furthermore, consider the following expansions:

N
X
u→ εr u[r] + RN ,
r=1
X
a→ ap εp ,
p≥1
X
b→ bq εq + δb,
q≥1

4
where RN stands for the remainder of our approximation of the solution after truncating the asymptotic
expansion at order N and, for every pair of integers p, q ≥ 1, ap , bq are real numbers. Our first goal is to
find an approximation of the Maxwell point that exists close to codimension-two Turing bifurcation points,
together with an approximation of a solution that joins the homogeneous steady state 0 to a large-amplitude
periodic orbit with the same minimal energy as 0. We do this by finding suitable parameters in the asymptotic
expansion to find a Maxwell point, whilst making sure that said asymptotic expansion is valid.
Now, for each integer 0 ≤ i ≤ Ma , 0 ≤ j ≤ Mb , we introduce the following symmetric, multilinear vector
functions
 
1  X ∂ 2 fi,j (u) 
F2,i,j (v1 , v2 ) = v1,p1 v2,p2 ,
2! ∂up1 ∂up2
1≤p1 ,p2 ≤n
u=0
 
3
1  X ∂ fi,j (u) 
F3,i,j (v1 , v2 , v3 ) = v1,p1 v2,p2 v3,p3 ,
3! ∂up1 ∂up2 ∂up3
1≤p1 ,p2 ,p3 ≤n
u=0
 
4
1  X ∂ fi,j
F4,i,j (v1 , v2 , v3 , v4 ) = v1,p1 v2,p2 v3,p3 v4,p4  ,
4! ∂up1 ∂up2 ∂up3 ∂up4
1≤p1 ,p2 ,p3 ,p4 ≤n
u=0
 
1  X ∂ 5 fi,j
F5,i,j (v1 , v2 , v3 , v4 , v5 ) = v1,p1 v2,p2 v3,p3 v4,p4 v5,p5  ,
5! ∂up1 ∂up2 ∂up3 ∂up4 ∂up5
1≤p1 ,p2 ,p3 ,p4 ,p5 ≤n
u=0
(5)
⊺
where vℓ = (vℓ,1 , . . . , vℓ,n ) , for ℓ = 1, . . . , 5, and we note that these functions can be naturally generalized
to higher orders.
Also, for simplicity of notation, we define
Mℓ = Ju f0,0 (0) − ℓ2 k 2 D̂,
for each non-negative integer ℓ, and we note that Mℓ is invertible for every ℓ ̸= 1.

2.2 Construction of a regular amplitude equation

This section aims to show how to develop a regular asymptotic expansion for the amplitude of Turing
[1]
patterns, A1 = A1 (X) ∈ C, under the ansatz u ∼ A1 eix ϕ1 where k is the wavenumber of the Turing
[1]
bifurcation and ϕ1 is an eigenvector of M1 . In particular, we want to show that A1 satisfies an amplitude
equation of the form
2 2 4
α1 A′′1 + i α2 A′1 + i α3 |A1 | A′1 + i α4 A21 Ā′1 + α5 A1 + α6 |A1 | A1 + α7 |A1 | A1 = 0, (6)
where αi ∈ R for each i = 1, . . . , 7. For this, we perform an asymptotic expansion of system (4) in terms of
ε. We do this order by order as follows.

Order O(ε). At this order, (4) becomes

∂ 2 u[1]
0 = Ju f0,0 (0) u[1] + k 2 D̂ ,
∂x2
which implies
[1]
u[1] = A1 eix̂ ϕ1 + c.c., (7)
[1]
where c.c. stands for the complex-conjugate of the term to the left of it, ϕ1 ̸= 0 fulfills
[1]
M1 ϕ1 = 0, (8)
and x̂ = x − χ̂, where χ̂ is a real variable that stands for a spatial translation in x, which plays a key degree
of freedom one needs for matching solutions in Section 7. However, for simplicity of notation, when there is
no confusion, we will drop write x instead of x̂.

5


Order O ε2 . At this order, (4) becomes
  ∂ 2 u[2]
0 = Ju f0,0 (0) u[2] + a1 Ju f1,0 (0) u[1] + b1 Ju f0,1 (0) u[1] + F2,0,0 u[1] , u[1] + k 2 D̂ ,
∂x2
which implies
∂2
 
[1] [1]
Ju f0,0 (0) + k D̂ 2 u[2] = −a1 A1 eix Ju f1,0 (0) ϕ1 − b1 A1 eix Ju f0,1 (0) ϕ1
2
∂x
  
[1] [1] 2
− F2,0,0 ϕ1 , ϕ1 |A1 | + A21 e2ix + c.c.

Therefore, as this equation is linear, we have that

2 [2] [2] [2] [1]
u[2] = |A1 | W0 + A1 eix W1 + A21 e2ix W2 + A2 eix ϕ1 + c.c.,
where A2 = A2 (X),
 
[2] [1] [1]
M0 W0 = −F2,0,0 ϕ1 , ϕ1 ,
[2] [1] [1]
M1 W1 = −a1 Ju f1,0 (0) ϕ1 − b1 Ju f0,1 (0) ϕ1 , (9)
and
 
[2] [1] [1]
M2 W2 = −F2,0,0 ϕ1 , ϕ1

Here, we recall that det (M1 ) = 0. Therefore, (9) might not have a solution for every value of a1 , b1 ∈ R.
For this expansion to be valid, we need all the equations to have a solution, so we make use of the Fredholm
alternative. It states that for a generic Fredholm operator, L, we have that
Im(L)⊥ = ker (L∗ ) , (10)
n ∗
In particular, using the usual inner product of vectors in C , we have that for a real matrix M , M = M ⊺ .
Considering this, we define ψ± = e±ix ψ, where ψ ̸= 0 fulfills
M⊺1 ψ = 0. (11)
With this, and using (10), we choose a1 , b1 ∈ R such that
D E
[1] [1]
a1 Ju f1,0 (0) ϕ1 + b1 Ju f0,1 (0) ϕ1 , ψ = 0, (12)

which ensures that (9) has a solution.

Order O ε3 . At this order, (4) becomes

0 = Ju f0,0 (0) u[3] + a1 Ju f1,0 (0) u[2] + b1 Ju f0,1 (0) u[2] + a2 Ju f1,0 (0) u[1] + b2 Ju f0,1 (0) u[1]
 
+a21 Ju f2,0 (0) u[1] + a1 b1 Ju f1,1 (0) u[1] + b21 Ju f0,2 (0) u[1] + 2 F2,0,0 u[1] , u[2]
     
+a1 F2,1,0 u[1] , u[1] + b1 F2,0,1 u[1] , u[1] + F3,0,0 u[1] , u[1] , u[1]
 2 [3]
∂ 2 u[1]

2 ∂ u
+k D̂ +2 ,
∂x2 ∂x∂X
which implies
∂2
 
Ju f0,0 (0) + k D̂ 2 u[3] = −a1 Ju f1,0 (0) u[2] − b1 Ju f0,1 (0) u[2] − a2 Ju f1,0 (0) u[1]
2
∂x
− b2 Ju f0,1 (0) u[1] − a21 Ju f2,0 (0) u[1] − a1 b1 Ju f1,1 (0) u[1]
   
− b21 Ju f0,2 (0) u[1] − 2 F2,0,0 u[1] , u[2] − a1 F2,1,0 u[1] , u[1]
    ∂ 2 u[1]
− b1 F2,0,1 u[1] , u[1] − F3,0,0 u[1] , u[1] , u[1] − 2 k 2 D̂ ,
∂x∂X

6
Therefore, the solution to this equation is given by:
2 [3] [3] 2 [3] [3] [3]
u[3] = |A1 | W0 + A1 eix W1 + |A1 | A1 eix W1,2 + i A′1 eix W1,3 + A21 e2ix W2
[3] [2] [2] [1] [2]
+ A31 e3ix W3 + 2 A1 Ā2 W0 + A2 eix W1 + A3 eix ϕ1 + 2 A1 A2 e2ix W2 + c.c.,
where A3 = A3 (X), the sign ′ denotes differentiation with respect to X,
 
[3] [2] [2] [1] [2]
M0 W0 = −a1 Ju f1,0 (0) W0 − b1 Ju f0,1 (0) W0 − 2 F2,0,0 ϕ1 , W1
   
[1] [1] [1] [1]
− a1 F2,1,0 ϕ1 , ϕ1 − b1 F2,0,1 ϕ1 , ϕ1 ,

[3] [2] [2] [1] [1]

M1 W1 = −a1 Ju f1,0 (0) W1 − b1 Ju f0,1 (0) W1 − a2 Ju f1,0 (0) ϕ1 − b2 Ju f0,1 (0) ϕ1
[1] [1] [1]
(13)
− a21 Ju f2,0 (0) ϕ1 − a1 b1 Ju f1,1 (0) ϕ1 − b21 Ju f0,2 (0) ϕ1 ,
   
[3] [1] [2] [2] [1] [1] [1]
M1 W1,2 = −2 F2,0,0 ϕ1 , 2 W0 + W2 − 3 F3,0,0 ϕ1 , ϕ1 , ϕ1 , (14)

[3] [1]
M1 W1,3 = −2 k 2 D̂ ϕ1 , (15)

 
[3] [2] [2] [1] [2]
M2 W2 = −a1 Ju f1,0 (0) W2 − b1 Ju f0,1 (0) W2 − 2 F2,0,0 ϕ1 , W1
   
[1] [1] [1] [1]
− a1 F2,1,0 ϕ1 , ϕ1 − b1 F2,0,1 ϕ1 , ϕ1 ,

and
   
[3] [1] [2] [1] [1] [1]
M3 W3 = −2 F2,0,0 ϕ1 , W2 − F3,0,0 ϕ1 , ϕ1 , ϕ1 .

Note that, as stated after (2), when (a, b) = (0, 0), we are at a codimension-two Turing bifurcation point.
Therefore, from [18], we have that
D     E
[1] [2] [2] [1] [1] [1]
2 F2,0,0 ϕ1 , 2 W0 + W2 + 3 F3,0,0 ϕ1 , ϕ1 , ϕ1 , ψ = 0,

which implies that (14) has a solution.

On the other hand, to ensure that (13) has a solution, we need that a1 , b1 , a2 , b2 ∈ R fulfill
D
[2] [2] [1] [1]
a1 Ju f1,0 (0) W1 + b1 Ju f0,1 (0) W1 + a2 Ju f1,0 (0) ϕ1 + b2 Ju f0,1 (0) ϕ1
E
[1] [1] [1]
+a21 Ju f2,0 (0) ϕ1 + a1 b1 Ju f1,1 (0) ϕ1 + b21 Ju f0,2 (0) ϕ1 , ψ = 0.

Next, to see that (15) has a solution, we consider the following standard result:
[1]
Lemma 1. For ϕ1 and ψ defined in (8) and (11), respectively, we have that
D E
[1]
D̂ ϕ1 , ψ = 0.

[1] [1]
Proof. For each k ∈ R, we can think of ϕ1 = ϕ1 (k) as a function of k is defined by the equation
 
[1] [1]
Ju f0,0 (0) − k 2 D̂ ϕ1 = λ(k) ϕ1 ,

where λ(k) represents the eigenvalue of Ju f0,0 (0) − k 2 D̂ with the largest real part such that λ (k ∗ ) = 0 for
the value k = k ∗ where the Turing bifurcation occurs. With this, after differentiating this expression with
respect to k, we have
  dϕ[1] dλ
[1]
dϕ1
[1] [1]
−2 k D̂ ϕ1 + Ju f0,0 (0) − k 2 D̂ 1
= (k) ϕ1 + λ(k) .
dk dk dk

7
Therefore, when evaluating k = k ∗ —and dropping the star sign—, we obtain
[1]
dϕ1 [1]
M1 = 2 k D̂ ϕ1 .
dk
Thus, when applying the inner product with respect to ψ, we see that
* + * +
[1] [1]
D
[1]
E dϕ1 dϕ1 ⊺
2 k D̂ ϕ1 , ψ = M1 ,ψ = , M1 ψ = 0,
dk dk

which concludes the proof.

Order O ε4 . At this order, equation (4) becomes

0 = Ju f0,0 (0) u[4] + a1 Ju f1,0 (0) u[3] + b1 Ju f0,1 (0) u[3] + a2 Ju f1,0 (0) u[2] + b2 Ju f0,1 (0) u[2]
+ a3 Ju f1,0 (0) u[1] + b3 Ju f0,1 (0) u[1] + a21 Ju f2,0 (0) u[2] + a1 b1 Ju f1,1 (0) u[2]
+ b21 Ju f0,2 (0) u[2] + 2 a1 a2 Ju f2,0 (0) u[1] + (a1 b2 + a2 b1 ) Ju f1,1 (0) u[1]
+ 2 b1 b2 Ju f0,2 (0) u[1] + a31 Ju f3,0 (0) u[1] + a21 b1 Ju f2,1 (0) u[1] + a1 b21 Ju f1,2 (0) u[1]
     
+ b31 Ju f0,3 (0) u[1] + 2 F2,0,0 u[1] , u[3] + F2,0,0 u[2] , u[2] + 2 a1 F2,1,0 u[1] , u[2]
     
+ 2 b1 F2,0,1 u[1] , u[2] + a2 F2,1,0 u[1] , u[1] + b2 F2,0,1 u[1] , u[1]
     
+ a21 F2,2,0 u[1] , u[1] + a1 b1 F2,1,1 u[1] , u[1] + b21 F2,0,2 u[1] , u[1]
     
+ 3 F3,0,0 u[1] , u[1] , u[2] + a1 F3,1,0 u[1] , u[1] , u[1] + b1 F3,0,1 u[1] , u[1] , u[1]
 2 [4]
∂ 2 u[2]

  ∂ u
+ F4,0,0 u[1] , u[1] , u[1] , u[1] + k 2 D̂ + 2 .
∂x2 ∂x∂X
Therefore,
2 [4] 4 [4] [4] [4] 2 [4] [4]
u[4] = |A1 | W0 + |A1 | W0,2 + i Ā1 A′1 W0,3 + A1 eix W1 + |A1 | A1 eix W1,2 + i A′1 eix W1,3
[4] 2 [4] [4] [4] [4]
+ A21 e2ix W2 + |A1 | A21 e2ix W2,2 + i A1 A′1 e2ix W2,3 + A31 e3ix W3 + A41 e4ix W4
[3] [2] 2 [2] [3] [3]
+ 2 A1 Ā2 W0 + 2 A1 Ā3 W0 + |A2 | W0 + A2 eix W1 + A21 Ā2 eix W1,2
2 [3] [3] [2] [1] [3]
+ 2 |A1 | A2 eix W1,2 + i A′2 eix W1,3 + A3 eix W1 + A4 eix ϕ1 + 2 A1 A2 e2ix W2
[2] [2] [3]
+ 2 A1 A3 e2ix W2 + A22 e2ix W2 + 3 A21 A2 e3ix W3 + c.c.

where A4 = A4 (X),
[4] [3] [3] [2] [2]
M0 W0 = −a1 Ju f1,0 (0) W0 − b1 Ju f0,1 (0) W0 − a2 Ju f1,0 (0) W0 − b2 Ju f0,1 (0) W0
 
[2] [2] [2] [1] [3]
− a21 Ju f2,0 (0) W0 − a1 b1 Ju f1,1 (0) W0 − b21 Ju f0,2 (0) W0 − 2 F2,0,0 ϕ1 , W1
     
[2] [2] [1] [2] [1] [2]
− F2,0,0 W1 , W1 − 2 a1 F2,1,0 ϕ1 , W1 − 2 b1 F2,0,1 ϕ1 , W1
     
[1] [1] [1] [1] [1] [1]
− a2 F2,1,0 ϕ1 , ϕ1 − b2 F2,0,1 ϕ1 , ϕ1 − a21 F2,2,0 ϕ1 , ϕ1
   
[1] [1] [1] [1]
− a1 b1 F2,1,1 ϕ1 , ϕ1 − b21 F2,0,2 ϕ1 , ϕ1 ,

     
[4] [1] [3] [2] [2] [2] [2]
M0 W0,2 = −2 F2,0,0 ϕ1 , W1,2 − 2 F2,0,0 W0 , W0 − F2,0,0 W2 , W2
   
[1] [1] [2] [2] [1] [1] [1] [1]
− 3 F3,0,0 ϕ1 , ϕ1 , 2 W0 + W2 − 3 F4,0,0 ϕ1 , ϕ1 , ϕ1 , ϕ1 ,

8
  
[4] [1] [3]
M0 W0,3 = −2 F2,0,0 ϕ1 , W1,3 ,

[4] [3] [3] [2]

M1 W1 = −a1 Ju f1,0 (0) W1 − b1 Ju f0,1 (0) W1 − a2 Ju f1,0 (0) W1
[2] [1] [1] [2]
− b2 Ju f0,1 (0) W1 − a3 Ju f1,0 (0) ϕ1 − b3 Ju f0,1 (0) ϕ1 − a21 Ju f2,0 (0) W1
[2] [2] [1]
− a1 b1 Ju f1,1 (0) W1 − b21 Ju f0,2 (0) W1 − 2 a1 a2 Ju f2,0 (0) ϕ1 (16)
[1] [1] [1]
− (a1 b2 + a2 b1 ) Ju f1,1 (0) ϕ1 − 2 b1 b2 Ju f0,2 (0) ϕ1 − a31 Ju f3,0 (0) ϕ1
[1] [1] [1]
− a21 b1 Ju f2,1 (0) ϕ1 − a1 b21 Ju f1,2 (0) ϕ1 − b31 Ju f0,3 (0) ϕ1 ,
 
[4] [3] [3] [1] [3] [3]
M1 W1,2 = −a1 Ju f1,0 (0) W1,2 − b1 Ju f0,1 (0) W1,2 − 2 F2,0,0 ϕ1 , 2 W0 + W2
   
[2] [2] [2] [1] [2] [2]
− 2 F2,0,0 W1 , 2 W0 + W2 − 2 a1 F2,1,0 ϕ1 , 2 W0 + W2
    (17)
[1] [2] [2] [1] [1] [2]
− 2 b1 F2,0,1 ϕ1 , 2 W0 + W2 − 9 F3,0,0 ϕ1 , ϕ1 , W1
   
[1] [1] [1] [1] [1] [1]
− 3 a1 F3,1,0 ϕ1 , ϕ1 , ϕ1 − 3 b1 F3,0,1 ϕ1 , ϕ1 , ϕ1 ,

[4] [3] [3] [2]

M1 W1,3 = −a1 Ju f1,0 (0) W1,3 − b1 Ju f0,1 (0) W1,3 − 2 k 2 D̂ W1 , (18)

[4] [3] [3] [2] [2]

M2 W2 = −a1 Ju f1,0 (0) W2 − b1 Ju f0,1 (0) W2 − a2 Ju f1,0 (0) W2 − b2 Ju f0,1 (0) W2
 
[2] [2] [2] [1] [3]
− a21 Ju f2,0 (0) W2 − a1 b1 Ju f1,1 (0) W2 − b21 Ju f0,2 (0) W2 − 2 F2,0,0 ϕ1 , W1
       
[2] [2] [1] [2] [1] [2] [1] [1]
− F2,0,0 W1 , W1 − 2 a1 F2,1,0 ϕ1 , W1 − 2 b1 F2,0,1 ϕ1 , W1 − a2 F2,1,0 ϕ1 , ϕ1
       
[1] [1] [1] [1] [1] [1] [1] [1]
− b2 F2,0,1 ϕ1 , ϕ1 − a21 F2,2,0 ϕ1 , ϕ1 − a1 b1 F2,1,1 ϕ1 , ϕ1 − b21 F2,0,2 ϕ1 , ϕ1 ,
   
[4] [1] [3] [3] [2] [2]
M2 W2,2 = −2 F2,0,0 ϕ1 , W1,2 + W3 − 4 F2,0,0 W0 , W2
   
[1] [1] [2] [2] [1] [1] [1] [1]
− 6 F3,0,0 ϕ1 , ϕ1 , W0 + W2 − 4 F4,0,0 ϕ1 , ϕ1 , ϕ1 , ϕ1 ,
 
[4] [1] [3] [2]
M2 W2,3 = −2 F2,0,0 ϕ1 , W1,3 − 8 k 2 D̂ W2 ,
   
[4] [3] [3] [1] [3] [2] [2]
M3 W3 = −a1 Ju f1,0 (0) W3 − b1 Ju f0,1 (0) W3 − 2 F2,0,0 ϕ1 , W2 − 2 F2,0,0 W1 , W2
     
[1] [2] [1] [2] [1] [1] [2]
− 2 a1 F2,1,0 ϕ1 , W2 − 2 b1 F2,0,1 ϕ1 , W2 − 3 F3,0,0 ϕ1 , ϕ1 , W1
   
[1] [1] [1] [1] [1] [1]
− a1 F3,1,0 ϕ1 , ϕ1 , ϕ1 − b1 F3,0,1 ϕ1 , ϕ1 , ϕ1 ,
and
   
[4] [1] [3] [2] [2]
M4 W4 = −2 F2,0,0 ϕ1 , W3 − F2,0,0 W2 , W2
   
[1] [1] [2] [1] [1] [1] [1]
− 3 F3,0,0 ϕ1 , ϕ1 , W2 − F4,0,0 ϕ1 , ϕ1 , ϕ1 , ϕ1 .

Here, we need to ensure that equations (16), (17) and (18) have a solution so we require
D
[3] [3] [2] [2]
a1 Ju f1,0 (0) W1 + b1 Ju f0,1 (0) W1 + a2 Ju f1,0 (0) W1 + b2 Ju f0,1 (0) W1
[1] [1] [2] [2]
+a3 Ju f1,0 (0) ϕ1 + b3 Ju f0,1 (0) ϕ1 + a21 Ju f2,0 (0) W1 + a1 b1 Ju f1,1 (0) W1
[2] [1] [1]
+b21 Ju f0,2 (0) W1 + 2 a1 a2 Ju f2,0 (0) ϕ1 + (a1 b2 + a2 b1 ) Ju f1,1 (0) ϕ1
[1] [1] [1] [1]
+2 b1 b2 Ju f0,2 (0) ϕ1 + a31 Ju f3,0 (0) ϕ1 + a21 b1 Ju f2,1 (0) ϕ1 + a1 b21 Ju f1,2 (0) ϕ1
E
[1]
+b31 Ju f0,3 (0) ϕ1 , ψ = 0,

9
 D  
[3] [3] [1] [3] [3]
a1 Ju f1,0 (0) W1,2 + b1 Ju f0,1 (0) W1,2 + 2 F2,0,0 ϕ1 , 2 W0 + W2
   
[2] [2] [2] [1] [2] [2]
+2 F2,0,0 W1 , 2 W0 + W2 + 2 a1 F2,1,0 ϕ1 , 2 W0 + W2 (19)
   
[1] [2] [2] [1] [1] [2]
+2 b1 F2,0,1 ϕ1 , 2 W0 + W2 + 9 F3,0,0 ϕ1 , ϕ1 , W1
    E
[1] [1] [1] [1] [1] [1]
+3 a1 F3,1,0 ϕ1 , ϕ1 , ϕ1 + 3 b1 F3,0,1 ϕ1 , ϕ1 , ϕ1 , ψ = 0,

and
D E
[3] [3] [2]
a1 Ju f1,0 (0) W1,3 + b1 Ju f0,1 (0) W1,3 + 2 k 2 D̂ W1 , ψ = 0. (20)

An important thing to note here is that equations (12), (19) and (20) can be written as a linear system of
equations

υ1 a1 + υ2 b1 = 0,
υ3 a1 + υ4 b1 = 0, (21)
υ5 a1 + υ6 b1 = 0,

where υi ∈ R for all i = 1, . . . , 6. This implies that, if the determinant of any 2 × 2 submatrix of the matrix
of coefficients of this system is different from zero, then a1 = b1 = 0. That is, generically, these variables
need to be equal to zero. Nevertheless, we carry on with a general asymptotic analysis as there may be
systems in which one can carry out the analysis with these variables being different from zero.


Order O ε5 . At this order, (4) becomes

0 = Ju f0,0 (0) u[5] + a1 Ju f1,0 (0) u[4] + b1 Ju f0,1 (0) u[4] + a2 Ju f1,0 (0) u[3] + b2 Ju f0,1 (0) u[3]
+ a3 Ju f1,0 (0) u[2] + b3 Ju f0,1 (0) u[2] + a4 Ju f1,0 (0) u[1] + b4 Ju f0,1 (0) u[1] + a21 Ju f2,0 (0) u[3]
+ a1 b1 Ju f1,1 (0) u[3] + b21 Ju f0,2 (0) u[3] + 2 a1 a2 Ju f2,0 (0) u[2] + (a1 b2 + a2 b1 ) Ju f1,1 (0) u[2]
+ 2 b1 b2 Ju f0,2 (0) u[2] + 2 a1 a3 Ju f2,0 (0) u[1] + (a1 b3 + a3 b1 ) Ju f1,1 (0) u[1] + 2 b1 b3 Ju f0,2 (0) u[1]
+ a22 Ju f2,0 (0) u[1] + a2 b2 Ju f1,1 (0) u[1] + b22 Ju f0,2 (0) u[1] + a31 Ju f3,0 (0) u[2] + a21 b1 Ju f2,1 (0) u[2]
+ a1 b21 Ju f1,2 (0) u[2] + b31 Ju f0,3 (0) u[2] + 3 a21 a2 Ju f3,0 (0) u[1] + 2 a1 a2 b1 Ju f2,1 (0) u[1]
+ a2 b21 Ju f1,2 (0) u[1] + a21 b2 Ju f2,1 (0) u[1] + 2 a1 b1 b2 Ju f1,2 (0) u[1] + 3 b21 b2 Ju f0,3 (0) u[1]
+ a41 Ju f4,0 (0) u[1] + a31 b1 Ju f3,1 (0) u[1] + a21 b21 Ju f2,2 (0) u[1] + a1 b31 Ju f1,3 (0) u[1] + b41 Ju f0,4 (0) u[1]
       
+ 2 F2,0,0 u[1] , u[4] + 2 F2,0,0 u[2] , u[3] + 2 a1 F2,1,0 u[1] , u[3] + 2 b1 F2,0,1 u[1] , u[3]
       
+ a1 F2,1,0 u[2] , u[2] + b1 F2,0,1 u[2] , u[2] + 2 a2 F2,1,0 u[1] , u[2] + 2 b2 F2,0,1 u[1] , u[2]
       
+ a3 F2,1,0 u[1] , u[1] + b3 F2,0,1 u[1] , u[1] + 2 a21 F2,2,0 u[1] , u[2] + 2 a1 b1 F2,1,1 u[1] , u[2]
     
+ 2 b21 F2,0,2 u[1] , u[2] + 2 a1 a2 F2,2,0 u[1] , u[1] + (a1 b2 + a2 b1 ) F2,1,1 u[1] , u[1]
       
+ 2 b1 b2 F2,0,2 u[1] , u[1] + a31 F2,3,0 u[1] , u[1] + a21 b1 F2,2,1 u[1] , u[1] + a1 b21 F2,1,2 u[1] , u[1]
     
+ b31 F2,0,3 u[1] , u[1] + 3 F3,0,0 u[1] , u[1] , u[3] + 3 F3,0,0 u[1] , u[2] , u[2]
     
+ 3 a1 F3,1,0 u[1] , u[1] , u[2] + 3 b1 F3,0,1 u[1] , u[1] , u[2] + a2 F3,1,0 u[1] , u[1] , u[1]
     
+ b2 F3,0,1 u[1] , u[1] , u[1] + a21 F3,2,0 u[1] , u[1] , u[1] + a1 b1 F3,1,1 u[1] , u[1] , u[1]
     
+ b21 F3,0,2 u[1] , u[1] , u[1] + 4 F4,0,0 u[1] , u[1] , u[1] , u[2] + a1 F4,1,0 u[1] , u[1] , u[1] , u[1]
   
+ b1 F4,0,1 u[1] , u[1] , u[1] , u[1] + F5,0,0 u[1] , u[1] , u[1] , u[1] , u[1]

10
 ∂ 2 u[5] ∂ 2 u[3] ∂ 2 u[1]
 
+ k 2 D̂ 2
+2 + . (22)
∂x ∂x∂X ∂X 2
At this point, we are not interested in the full solution to this equation (find the complete development up
to order seven in Appendix B). For now, we only need to ensure it exists. To do this, we have to obtain a
solvability condition from the terms that are multiples of eix on the right-hand side of (22), which are given
by
[5] [5] [5] 2 [5] [5] [5] 2 [5] 4
−Q1 A′′1 − i Q2 A′1 − i Q3 |A1 | A′1 − i Q4 A21 Ā′1 − Q5 A1 − Q6 |A1 | A1 − Q7 |A1 | A1 + . . . ,
where “. . .” are terms that depend on A2 , A3 , A4 and do not influence the solvability condition, and
[5] [3] [1]
Q1 = −2 k 2 D̂ W1,3 + k 2 D̂ ϕ1 ,
[5] [4] [4] [3] [3]
Q2 = a1 Ju f1,0 (0) W1,3 + b1 Ju f0,1 (0) W1,3 + a2 Ju f1,0 (0) W1,3 + b2 Ju f0,1 (0) W1,3
[3] [3] [3] [3]
+ a21 Ju f2,0 (0) W1,3 + a1 b1 Ju f1,1 (0) W1,3 + b21 Ju f0,2 (0) W1,3 + 2 k 2 D̂ W1 ,
     
[5] [1] [4] [4] [2] [3] [1] [1] [3] [3]
Q3 = 2 F2,0,0 ϕ1 , W0,3 + W2,3 + 4 F2,0,0 W0 , W1,3 + 6 F3,0,0 ϕ1 , ϕ1 , W1,3 + 4 k 2 D̂ W1,2 ,
     
[5] [1] [4] [2] [3] [1] [1] [3] [3]
Q4 = −2 F2,0,0 ϕ1 , W0,3 − 2 F2,0,0 W2 , W1,3 − 3 F3,0,0 ϕ1 , ϕ1 , W1,3 + 2 k 2 D̂ W1,2 ,
[5] [4] [4] [3] [3]
Q5 = a1 Ju f1,0 (0) W1 + b1 Ju f0,1 (0) W1 + a2 Ju f1,0 (0) W1 + b2 Ju f0,1 (0) W1
[2] [2] [1] [1]
+ a3 Ju f1,0 (0) W1 + b3 Ju f0,1 (0) W1 + a4 Ju f1,0 (0) ϕ1 + b4 Ju f0,1 (0) ϕ1
[3] [3] [3] [2]
+ a21 Ju f2,0 (0) W1 + a1 b1 Ju f1,1 (0) W1 + b21 Ju f0,2 (0) W1 + 2 a1 a2 Ju f2,0 (0) W1
[2] [2] [1]
+ (a1 b2 + a2 b1 ) Ju f1,1 (0) W1 + 2 b1 b2 Ju f0,2 (0) W1 + 2 a1 a3 Ju f2,0 (0) ϕ1
[1] [1] [1]
+ (a1 b3 + a3 b1 ) Ju f1,1 (0) ϕ1 + 2 b1 b3 Ju f0,2 (0) ϕ1 + a22 Ju f2,0 (0) ϕ1
[1] [1] [2] [2]
+ a2 b2 Ju f1,1 (0) ϕ1 + b22 Ju f0,2 (0) ϕ1 + a31 Ju f3,0 (0) W1 + a21 b1 Ju f2,1 (0) W1
[2] [2] [1] [1]
+ a1 b21 Ju f1,2 (0) W1 + b31 Ju f0,3 (0) W1 + 3 a21 a2 Ju f3,0 (0) ϕ1 + 2 a1 a2 b1 Ju f2,1 (0) ϕ1
[1] [1] [1] [1]
+ a2 b21 Ju f1,2 (0) ϕ1 + a21 b2 Ju f2,1 (0) ϕ1 + 2 a1 b1 b2 Ju f1,2 (0) ϕ1 + 3 b21 b2 Ju f0,3 (0) ϕ1
[1] [1] [1] [1]
+ a41 Ju f4,0 (0) ϕ1 + a31 b1 Ju f3,1 (0) ϕ1 + a21 b21 Ju f2,2 (0) ϕ1 + a1 b31 Ju f1,3 (0) ϕ1
[1]
+ b41 Ju f0,4 (0) ϕ1 ,
[5] [4] [4] [3] [3]
Q6 = a1 Ju f1,0 (0) W1,2 + b1 Ju f0,1 (0) W1,2 + a2 Ju f1,0 (0) W1,2 + b2 Ju f0,1 (0) W1,2
[3] [3] [3]
+ a21 Ju f2,0 (0) W1,2 + a1 b1 Ju f1,1 (0) W1,2 + b21 Ju f0,2 (0) W1,2
   
[1] [4] [4] [2] [3] [3]
+ 2 F2,0,0 ϕ1 , 2 W0 + W2 + 2 F2,0,0 W1 , 2 W0 + W2
   
[3] [2] [2] [1] [3] [3]
+ 2 F2,0,0 W1 , 2 W0 + W2 + 2 a1 F2,1,0 ϕ1 , 2 W0 + W2
   
[1] [3] [3] [2] [2] [2]
+ 2 b1 F2,0,1 ϕ1 , 2 W0 + W2 + 2 a1 F2,1,0 W1 , 2 W0 + W2
   
[2] [2] [2] [1] [2] [2]
+ 2 b1 F2,0,1 W1 , 2 W0 + W2 + 2 a2 F2,1,0 ϕ1 , 2 W0 + W2
   
[1] [2] [2] [1] [2] [2]
+ 2 b2 F2,0,1 ϕ1 , 2 W0 + W2 + 2 a21 F2,2,0 ϕ1 , 2 W0 + W2
   
[1] [2] [2] [1] [2] [2]
+ 2 a1 b1 F2,1,1 ϕ1 , 2 W0 + W2 + 2 b21 F2,0,2 ϕ1 , 2 W0 + W2
   
[1] [1] [3] [1] [2] [2]
+ 9 F3,0,0 ϕ1 , ϕ1 , W1 + 9 F3,0,0 ϕ1 , W1 , W1
   
[1] [1] [2] [1] [1] [2]
+ 9 a1 F3,1,0 ϕ1 , ϕ1 , W1 + 9 b1 F3,0,1 ϕ1 , ϕ1 , W1
   
[1] [1] [1] [1] [1] [1]
+ 3 a2 F3,1,0 ϕ1 , ϕ1 , ϕ1 + 3 b2 F3,0,1 ϕ1 , ϕ1 , ϕ1
     
[1] [1] [1] [1] [1] [1] [1] [1] [1]
+ 3 a21 F3,2,0 ϕ1 , ϕ1 , ϕ1 + 3 a1 b1 F3,1,1 ϕ1 , ϕ1 , ϕ1 + 3 b21 F3,0,2 ϕ1 , ϕ1 , ϕ1 ,

11
      
[5] [1] [4] [4] [2] [3] [3] [2] [2]
Q7 = 2 F2,0,0 ϕ1 , 2 W0,2 + W2,2 + 2 F2,0,0 W2 , W3 + 2 F2,0,0 W1,2 , 2 W0 + W2
   
[1] [1] [3] [3] [1] [2] [2]
+ 3 F3,0,0 ϕ1 , ϕ1 , 3 W1,2 + W3 + 12 F3,0,0 ϕ1 , W0 , W0
   
[1] [2] [2] [2] [1] [1] [1] [2] [2]
+ 6 F3,0,0 ϕ1 , W2 , 2 W0 + W2 + 8 F4,0,0 ϕ1 , ϕ1 , ϕ1 , 3 W0 + 2 W2
 
[1] [1] [1] [1] [1]
+ 10 F5,0,0 ϕ1 , ϕ1 , ϕ1 , ϕ1 , ϕ1 .

Again, we need to ensure that (22) has a solution and, as before,

D we canE apply the inner product with
[5]
ψ. Finally, we arrive at the amplitude equation (6), where αi = Qi , ψ for i = 1, . . . , 7, and we require
α1 ̸= 0.

2.3 Solving the amplitude equation

In this section, we look for heteroclinic orbits in (6), between the trivial steady state A1 = 0 and a finite-
amplitude periodic pattern, whose amplitude is to be determined. We are interested in such fronts because
they help us define the Maxwell point and, generically, give rise to homoclinic snaking, which we aim to
study through beyond-all-orders terms.
To solve (6), we write A1 in polar coordinates as A1 = R1 ei φ1 , where R1 = R1 (X) ∈ R and φ1 =
φ1 (X) ∈ R. With this, (6) becomes
2
α1 R1′′ − α1 R1 (φ′1 ) − α2 R1 φ′1 − α3 R13 φ′1 + α4 R13 φ′1 + α5 R1 + α6 R13 + α7 R15
+i α1 R1 φ′′1 + 2 α1 R1′ φ′1 + α2 R1′ + (α3 + α4 ) R12 R1′ = 0,

which can be split into two equations given by

2
α1 R1′′ − α1 R1 (φ′1 ) − α2 R1 φ′1 − α3 R13 φ′1 + α4 R13 φ′1 + α5 R1 + α6 R13 + α7 R15 = 0, (23)

and

α1 R1 φ′′1 + 2 α1 R1′ φ′1 + α2 R1′ + (α3 + α4 ) R12 R1′ = 0. (24)

Now, note that (24) is equivalent to

α3 + α4 3 ′ α2
R12 φ′′1 + 2 R1 R1′ φ′1 = − R1 R1 − R1 R1′ ,
α1 α1
which can be integrated directly to obtain
α3 + α4 2 α2
φ′1 = − R1 − . (25)
4 α1 2 α1
On the other hand, when replacing (25) into (23), we get
dV
2 R1′′ = , (26)
dR1
where
dV
= 2 β1 R1 + 4 β3 R13 + 6 β5 R15 ,
dR1
α2 + 4 α1 α5
β1 = − 2 ,
4 α12
α2 (α3 − α4 ) + 2 α1 α6
β3 = − , (27)
4 α12
(α3 + α4 ) (3 α3 − 5 α4 ) + 16 α1 α7
β5 = − .
48 α12

12
Note that (26) is invariant under the change X → υ − X for every υ ∈ R. Furthermore, we have that V
represents an energy potential, which, for simplicity, we subject to the condition V (0) = 0. This implies that

V (R1 ) = β1 R12 + β3 R14 + β5 R16 ,

which has critical points at R1 = 0 and

s p
−β3 ± β32 − 3 β1 β5
R1,± = .
3 β5

We need these critical points to be real so we require β1 β5 > 0, β3 β5 < 0 and β32 ≥ 3 β1 β5 .
Furthermore, note that

d2 V
(0) = 2 β1 , and
dR12
p
d2 V 8 β32 − 3 β1 β5 − β3 β32 − 3 β1 β5
2 (R1,+ ) = . (28)
dR1 3 β5

To find a Maxwell point, we need to find the condition under which R1 = 0 and R1 = R1,+ have the
same minimal energy, which implies that we need β1 > 0, β3 < 0 and β5 > 0. Furthermore, with these
assumptions, we have that (28) is positive, which implies that R1,+ is also a minimum.
In addition, we require R1 = 0 to have the same minimal energy as R1,+ . Therefore, we need

V (0) = V (R1,+ ) ⇐⇒ β32 = 4 β1 β5 , (29)

which defines a first approximation to the Maxwell point and is an assumption we make from now on.
With this, if we multiply (26) by R1′ and integrate it once, we obtain
2
(R1′ ) = R12 β1 + β3 R12 + β5 R14 ,


(30)

where we note that the right-hand side is a positive multiple of a parabola in R12 with a positive leading-
order coefficient and a discriminant equal to zero due to (29). This implies that the right-hand side of (30)
is non-negative, making (30) well-defined.
Let us now consider the following change R12 = 1/P, which implies:
2
P′ P′

β1 β3 β5
R1′ =− and = + 2 + 3.
2 P 3/2 2 P 3/2 P P P

Therefore,
2
(P ′ ) = 4 β1 P 2 + β3 P + β5 .

We proceed to solve this equation using the method of separation of variables. If we consider the negative
branch of this equation when taking the square root, we have that
Z
1 dP
p = −X + C,
2 β1 P 2 + β3 P + β5
p
where C ∈ R is a constant of integration. Now, let Q = β1 P 2 + β3 P + β5 . Then,

β1 P 2 + β3 P + β5 − Q2 = 0,

which implies
p √
−β3 ± β32 − 4 β1 (β5 − Q2 ) −β3 ± 2 β1 Q dQ
P= = and dP = ± √ .
2 β1 2 β1 β1

13
With this, we have
Z Z
1 dP 1 dQ 1 1 p 
p = √ = √ log(Q) = √ log β1 P 2 + β3 P + β5 ),
2 β 1 P 2 + β3 P + β5 2 Q β1 2 β1 2 β1

where we choose the positive sign for P, as it needs to be non-negative, by definition.

With this, we note that
1 p 
√ log β1 P 2 + β3 P + β5 = −X + C
2 β1
 p 
⇐⇒ β1 P 2 + β3 P + β5 = C 2 exp −4 β1 X ,

which yields
q √


1 −β3 + β32 − 4 β1 β5 − C 2 exp −4 β1 X
P= =
R12 2 β1
√ √

−β3 + 2 β1 C exp −2 β1 X
=
2 β1
Therefore, we conclude that
4 β1
R12 = √
, (31)
−2 β3 + exp −2 β1 X

1
where we have taken C = √ , for simplicity. Note that (31) corresponds to an up-front since
4 β1
2 β1
lim R12 = 0, and lim R12 = − > 0.
X→−∞ X→∞ β3

On the other hand, as equation (26) is invariant under the change X → −X, then the down-front

4 β1
R̃12 (X) = √
, (32)
−2 β3 + exp 2 β1 X

is also a solution to (26). Figure 2 depicts these two fronts. Specifically, the red (respectively, blue) dashed
line represents the up-front (respectively, down-front) given by (31) (respectively, (32)). On the other hand,
the green continuous lines show an oscillatory pattern enclosed by these two fronts. These two solutions’
existence is key for studying late terms in the asymptotic expansion, as their existence and matching provides
conditions on some parameters of the expansion (see Section 7).
To keep consistency with notation, we now focus on the up-front, (31). The analysis for the down-front
is analogous. This yields

β1 (α3 + α4 ) α
φ′1 = √

− 2 ,
α1 2 β3 − exp −2 β1 X 2 α1

so φ1 takes the following form:

  p  p  p 
φ1 = η log 1 − 2 β3 exp 2 β1 X − 2 ξ β1 X − ξ log (−2 β3 ) − η log 2 β1 , (33)

where
√
α2 (α3 + α4 ) β1
ξ= √ , and η= .
4 α1 β1 4 α1 β3

14
 u

Figure 2: Form of solutions captured by the amplitude equation, (6), at the Maxwell point. The red and blue dashed
curves represent the envelope of oscillatory localized solutions
r to (3), represented by R1 . Specifically, the blue dashed
2 β1
line represents an up-front joining R1 = 0 to R1 = − (represented by the top and bottom black horizontal
β3
lines) from left to right, whilst the red dashed curve represents a translated reflection of the blue curve, which turns
out to be a down-front joining the same steady states in the opposite direction. Furthermore, the green continuous
curve represents the oscillatory pattern enclosed by that envelope.

Now, to simplify notation, we consider the following translation:

log (−2 β3 )
X→− √ + X,
2 β1

which transforms (31) and (33) into

√

2 β1 exp 2 β1 X
R12 =− √
,
β3 1 + exp 2 β1 X

(2 j + 1) π i
which has singularities at Xj = √ for j ∈ Z, and
2 β1
  p  p  p 
φ1 = η log 1 + exp 2 β1 X − 2 ξ β1 X − η log 2 β1 .

respectively. With this, we note that

A1 = R 1 e i φ 1
s
 p 1−η i 1   p η i  
− 12
p
= 2 β1 (−2 β3 ) √
1 + exp 2 β1 X exp −2 i β1 ξ X .
1 + exp −2 β1 X

Now, we note that the leading-order expansion of different functions associated with A1 (X) around X0 =
πi
√ , which turns out to be one of the singularities of R1 that is closest to the real axis in X, together
2 β1
with X−1 , is given by
r
[1] −1 ηi [1]
A1 ϕ1 ∼ −i CA (X0 − X) ϕ1 ,
X0 − X
r
[1] −1 −ηi [1]
Ā1 ϕ1 ∼ −i CĀ (X0 − X) ϕ1 ,
X0 − X

15
 2 [2] [2] −1 [2]
|A1 | W0 = R12 W0 ∼ CA CĀ (X0 − X) W0 ,
[2] −1+2 η i [2]
A21 W2 ∼ CA
2
(X0 − X) W2 ,
[2] −1−2 η i [2]
Ā21 W2 ∼ CĀ
2
(X0 − X) W2 ,
 3/2
2 [3] [3] −1 ηi [3]
|A1 | A1 W1,2 = R12 A1 W1,2 ∼ i CA2
CĀ (X0 − X) W1,2 ,
X0 − X
 3/2
2 [3] 2 [3] 2 −1 −ηi [3]
|A1 | Ā1 W1,2 = R1 Ā1 W1,2 ∼ i CA CĀ (X0 − X) W1,2 ,
X0 − X
r
[3] i −1 −1+ηi [3]
i A′1 W1,3 ∼ − (2 η + i) CA (X0 − X) W1,3 ,
2 X0 − X
r
[3] i −1 −1−ηi [3]
−i Ā′1 W1,3 ∼ − (2 η − i) CA (X0 − X) W1,3 ,
2 X0 − X
 3/2
3 [3] 3 −1 3ηi [3]
A1 W3 ∼ i CA (X0 − X) W3 ,
X0 − X
 3/2
[3] −1 −3 η i [3]
Ā31 W3 ∼ i CĀ
3
(X0 − X) W3 ,
X0 − X
4 [4] [4] −2 [4]
|A1 | W0,2 = R14 W0,2 ∼ CA 2 2
CĀ (X0 − X) W0,2 ,
[4] i −2 [4]
i Ā1 A′1 W0,3 ∼ (1 − 2 η i) CA CĀ (X0 − X) W0,3 ,
2
[4] i −2 [4]
−i A1 Ā′1 W0,3 ∼ − (1 + 2 η i) CA CĀ (X0 − X) W0,3 ,
2
2 [4] [4] −2+2 η i [4]
|A1 | A21 W2,2 = R12 A21 W2,2 ∼ CA 3
CĀ (X0 − X) W2,2 ,
2 [4] [4] −2−2 η i [4]
|A1 | Ā21 W2,2 = R12 Ā21 W2,2 ∼ CA CĀ
3
(X0 − X) W2,2 ,
[4] i −2+2 η i [4]
i A1 A′1 W2,3 ∼ (1 − 2 η i) CA 2
(X0 − X) W2,3 ,
2
[4] i −2−2 η i [4]
−i Ā1 Ā′1 W2,3 ∼ − (1 + 2 η i) CĀ
2
(X0 − X) W2,3 .
2
where
 p 1/2
−1/2
CA = i eπ ξ 2 β1 (−2 β3 ) ,
 p 1/2
−1/2
CĀ = i e−π ξ 2 β1 (−2 β3 ) .

Remark 1. Note that

CĀ = −e−2 π ξ C̄A . (34)
Furthermore, observe that the expression of the leading-order term of A1 at X0 can be simplified even further
if we assume
r r
−1 1
=i .
X0 − X X0 − X
Nonetheless, as we are now working with a complex variable to carry out the analysis, we have to bear in
mind that, in the complex plane, some functions are not uniquely defined. We will then state everything in
this way to keep the development general.

3 Late-term expansion — inner solution

In this section, we aim to study the asymptotic behaviour of our solution, under the assumption that we
keep going with the asymptotic expansion up to a high order N ≫ 1. In particular, as explained in [5],

16
to understand the leading-order asymptotics in this case, which allows us to obtain an approximation of
the width of the homoclinic snaking, we must find two kinds of solutions: an inner and an outer solution.
The inner solution refers to an approximation that is valid only sufficiently close to a singularity, and the
outer solution lets us understand the behaviour of solutions away from it. Of course, as these solutions are
intended to represent different approximations of the same function, they need to be matched as X tends to
the singularity.
In particular, using the natural extensions of the symmetric multilinear vector functions we defined in
(5), let M > 0 be an integer and define p = (p1 , p2 , . . .), q = (q1 , q2 , . . .), n = (n1 , . . . , nM ), r = (r1 , . . . , rM ),
eM = (1, . . . , 1) (the vector with M ones) and e = (1, 2, . . . , M ). With this, assuming that n is sufficiently
large, and Taylor expanding Equation (4), we see that the n-th order equation is given by
  
X n−1−p
n−1 X Y Y
0= Cp,q  aps s   bqss  Ju fp,q (0) u[n−p−q]
p=0 q=0 s≥1 s≥1
p·e=p q·e=q
  
X n−1−p
n−1 X n−p−q
X Y Y  
+ Cp,q  aps s   bqss  FM,p,q u[n1 ] , . . . , u[nM ]
p=0 q=0 M =2 s≥1 s≥1
p·e=p q·e=q n·eM =n−p−q

∂ 2 u[n] ∂ 2 u[n−2] ∂ 2 u[n−4]

 
2
+ k D̂ + 2 + , (35)
∂x2 ∂x∂X ∂X 2

where Cp,q are the coefficients associated with the powers of a and b in the Taylor expansion of fp,q , for
each pair of vectors, p, q defined above. From the results we have obtained so far, we let
n
X
u[n] = Wn,r eirx , (36)
r=−n

where n ≥ 1 is an integer, and Wn,r = Wn,r (X) for every integers n ≥ 1, −n ≤ r ≤ n. In particular, taking
into account only the terms that are dominant at each mode as X → X0 , we still have Wn,−r = Wn,r for
each integers n ≥ 1, −n ≤ r ≤ n, r ̸= 0, and we note that
[1]
W1,1 eix + c.c. = A1 eix ϕ1 + c.c.,
2 [2]
W2,0 = 2 |A1 | W0 ,
[2]
W2,2 e2ix + c.c. = A21 e2ix W2 + c.c.,
2 [3] [3]
W3,1 eix + c.c. ∼ |A1 | A1 eix W1,2 + i A′1 eix W1,3 + c.c.,
[3]
W3,3 e3ix + c.c. = A31 e3ix W3 + c.c.,
4 [4] [4]
W4,0 ∼ |A1 | W0,2 + i Ā1 A′1 W0,3 + c.c.,
2 [4] [4]
W4,2 e2ix + c.c. ∼ |A1 | A21 e2ix W2,2 + i A1 A′1 e2ix W2,3 + c.c.

Therefore, when replacing (36) into (35) we obtain, for n ≥ 1 large and each r ∈ [−n, n],
  
X n−1−p
n−1 X Y Y
0= Cp,q  aps s   bqss  Ju fp,q (0) Wn−p−q,r
p=0 q=0 s≥1 s≥1
p·e=p q·e=q
  
X n−1−p
n−1 X n−p−q
X X Y Y
+ Cp,q  aps s   bqss  FM,p,q (Wn1 ,r1 , . . . , WnM ,rM )
p=0 q=0 M =2 r·eM =r s≥1 s≥1
p·e=p q·e=q n·eM =n−p−q |rℓ |≤nℓ
′ ′′
+ k 2 D̂ −r2 Wn,r + 2 i r Wn−2,r


+ Wn−4,r , (37)

17
3.1 Asymptotic expansion
Again, what we need to do now is to solve a linear equation order by order. We highlight that The following
analysis is valid for any singularity, but we focus on X0 , as it is one of the singularities closest to the real
axis for X. In particular, based on [5], for our inner solution we use the ansatz
Vn,r
Wn,r (X) = n
−r η i
, (38)
(X0 − X) 2
which implies
′ n − 2 − 2rηi Vn−2,r
Wn−2,r = n
−r η i
,
2 (X0 − X) 2
′′ (n − 2 − 2 r η i)(n − 4 − 2 r η i) Vn−4,r
Wn−4,r = n
−r η i
.
4 (X0 − X) 2
Therefore, when replacing (38) into (37), we obtain
  
X n−1−p
n−1 X Y Y Vn−p−q,r
p q
0= Cp,q  as s   bss  Ju fp,q (0) n−p−q
−r η i
p=0 q=0 s≥1 s≥1 (X0 − X) 2
p·e=p q·e=q
  
X n−1−p
n−1 X n−p−q
X X Y Y
+ Cp,q  aps s   bqss 
p=0 q=0 M =2 r·eM =r s≥1 s≥1
p·e=p q·e=q n·eM =n−p−q |rℓ |≤nℓ
!
Vn1 ,r1 VnM ,rM
× FM,p,q n1
−r1 η i
,..., nM
−rM η i
(X0 − X) 2
(X0 − X) 2
!
2 2 Vn,r n − 2 − 2rηi Vn−2,r
+ k D̂ −r n
−r η i
+ 2ir n
−r η i
(X0 − X) 2 2 (X0 − X) 2
!
(n − 2 − 2 r η i) (n − 4 − 2 r η i) Vn−4,r
+ n
−r η i
,
4 (X0 − X) 2
which is equivalent to
  
X n−1−p
n−1 X Y Y Vn−p−q,r
0= Cp,q  aps s   bqss  Ju fp,q (0) −p−q
p=0 q=0 s≥1 s≥1 (X0 − X) 2
p·e=p q·e=q
  
X n−1−p
n−1 X n−p−q
X X Cp,q Y Y
+ −p−q
 aps s   bqss  FM,p,q (Vn1 ,r1 , . . . , VnM ,rM )
p=0 q=0 M =2 r·eM =r (X0 − X) 2 s≥1 s≥1
p·e=p q·e=q n·eM =n−p−q |rℓ |≤nℓ
  
n − 2 − 2rηi
+ k 2 D̂ −r2 Vn,r + 2 i r Vn−2,r
2

(n − 2 − 2 r η i)(n − 4 − 2 r η i)
+ Vn−4,r ,
4
and we observe that the only terms that play a role in the inner solution are those with no parameters.
Therefore, we can simplify our equation for the inner solution as
n
X X
0 = Ju f0,0 (0) Vn,r + FM,0,0 (Vn1 ,r1 , . . . , VnM ,rM )
M =2 r·eM =r
n·eM =n |rℓ |≤nℓ
   
n − 2 − 2rηi (n − 2 − 2 r η i) (n − 4 − 2 r η i)
+ k 2 D̂ −r2 Vn,r + 2 i r Vn−2,r + Vn−4,r ,
2 4

18
which is equivalent to
  n
X X
0 = Ju f0,0 (0) − r2 k 2 D̂ Vn,r + FM,0,0 (Vn1 ,r1 , . . . , VnM ,rM )
M =2 r·eM =r
n·eM =n |rℓ |≤nℓ (39)
   
n − 2 − 2rηi (n − 2 − 2 r η i) (n − 4 − 2 r η i)
+ k 2 D̂ 2ir Vn−2,r + Vn−4,r ,
2 4

Here, note that if we change r into −r in the previous expression, we have

  n
X X
0 = Ju f0,0 (0) − r2 k 2 D̂ Vn,−r + FM,0,0 (Vn1 ,−r1 , . . . , VnM ,−rM )
M =2 r·eM =r
n·eM =n |rℓ |≤nℓ
   
2 n − 2 + 2rηi (n − 2 + 2 r η i)(n − 4 + 2 r η i)
+ k D̂ −2 i r Vn−2,−r + Vn−4,−r .
2 4

which gives us the conjugate of (39) if, based on (34), we consider

Vn,−r = (−1)n e−2 r π ξ Vn,r ,

for every integers n ≥ 1 and −n ≤ r ≤ n.

Now, in order to solve (39), we expand each vector Vn,r as
  X [s]
n−1 vr
Vn,r = κn−1 Γ + γr , (40)
2 ns
s≥0

where κ and γr are complex variables yet to be determined.

Now, noting that
          
n−1 n−3 n−3 n−3 n−5 n−5
Γ + γr = + γr Γ + γr = + γr + γr Γ + γr ,
2 2 2 2 2 2

and replacing (40) into (39), we obtain

  X [s]

2 2

n−1 n−1 vr
0 = Ju f0,0 (0) − r k D̂ κ Γ + γr
2 ns
s≥0
n
X X
+ FM,0,0 (Vn1 ,r1 , . . . , VnM ,rM )
M =2 r·eM =r
n·eM =n |rℓ |≤nℓ

   X [s]
n − 2 − 2rηi n−3 vr
+ k 2 D̂ 2 i r κn−3 Γ + γr
2 2 (n − 2)s
s≥0

 X [s]
(n − 2 − 2 r η i)(n − 4 − 2 r η i) n−5 vr
+κn−5 Γ + γr ,
4 2 (n − 4)s
s≥0

19
which is equivalent to

  X vr[s]
0 = Ju f0,0 (0) − r2 k 2 D̂ κ4
ns
s≥0
 −1 n
5−n n−1 X X
+κ Γ + γr FM,0,0 (Vn1 ,r1 , . . . , VnM ,rM )
2
M =2 r·eM =r
n·eM =n |rℓ |≤nℓ

 X [s]
n − 2 − 2rηi vr
+ k 2 D̂ 2 i r κ2
n − 3 + 2 γr (n − 2)s
s≥0

[s]
(n − 2 − 2 r η i)(n − 4 − 2 r η i) X vr
+ ,
(n − 3 + 2 γr ) (n − 5 + 2 γr ) (n − 4)s
s≥0

Furthermore, by expanding the previous expression in powers of 1/n, assuming that n is large, we have

  X vr[s]
0 = Ju f0,0 (0) − r2 k 2 D̂ κ4
ns
s≥0
 −1 n
5−n n−1 X X
+κ Γ + γr FM,0,0 (Vn1 ,r1 , . . . , VnM ,rM )
2
M =2 r·eM =r
n·eM =n |rℓ |≤nℓ
!
    [2]
2 2 −2 γr − 2 r η i + 1 (2 γr − 3) (2 γr + 2 r η i − 1) [0] 1 2 [1] vr
+ k D̂ 2 i r κ 1 + + vr + + vr + 2
n n2 n n2 n
!!
    [2]
−4 γr − 4 r η i + 2 (−2 i γr + 2 η r + i) (−6 i γr + 2 r η + 9 i) [0] 1 4 [1] vr
+ 1+ − vr + + vr + 2
n n2 n n2 n
 
1
+O . (41)
n3

Now, as n is large, then powers of 1/n can be considered small numbers, which lets us follow the usual
approach to study this equation following an asymptotic approach.

3.2 Solution at O(1)

At this order, (41) becomes

M (κ, r) vr[0] = −NL(1), (42)

where
   
M (κ, r) = κ4 Ju f0,0 (0) − r2 k 2 D̂ + κ2 2 i r k 2 D̂ + k 2 D̂,

and NL(1), are terms of order O(1) associated with the nonlinearities of system (3), which depend on γr .
We highlight that we will deal with this problem using dominance of modes as follows. We define γ as the
value of γr with the highest real part as r takes integer values between −∞ and ∞. With this, we can note
that NL(1) = 0 for the values of r such that γr = γ, assuming they are finite. To see why this is not an
assumption, consider the following result:
Lemma 2. The implicit function κ = |κ(r)| determined by det(M (κ, r)) = 0 is even and bounded away from
r = ±1. Furthermore, it is finite for r = ±1 and lim κ(r) = 0.
|r|→∞

20
Proof. Note that

M (κ, r) = M (κ̄, −r),

which proves that κ = |κ(r)| is an even function.

Now, note that det(M (κ, r)) is a polynomial of degree 4 n in κ for all |r| ̸= 1. On the other hand, when
|r| = 1, the degree of det(M (κ, r)) decreases to 4 n − 4. However, due to the fundamental theorem of algebra,
these polynomials will have well-defined finite complex roots for all values of r.
On the other hand, when |r| gets large,


det(M (κ, r)) ∼ det r k 2 κ2 −r κ2 + 2 i D̂ .

which implies that κ must tend to zero as |r| → ∞ in order to fulfill the equation det(M (κ, r)) = 0.
[0]
Now, to solve (42), we want vr to be non-zero for the values of r such that γr = γ. To determine those
values of r, We need to find the values of κ with the largest amplitude such that det(M (κ, r)) = 0, together
with the values of r that give rise to those values of κ, as these will generate the leading-order solution.
Observe that
   
2ir 1
M (κ, r) = κ4 Ju f0,0 (0) + −r2 + 2 + 4 k 2 D̂ .
κ κ
 
Next, note that there might be complex roots in ℓ to the equation det Ju f0,0 (0) − ℓ2 D̂ = 0. Nevertheless,
as we are focusing on the patterns that arise at a codimension-two Turing bifurcation point, then we only
focus on the case ℓ ∈ R, which implies that ℓ = k is the only solution in that case. Thus, det(M (κ, r)) = 0 if

2ir 1 i
−r2 + + 4 = −1 ⇐⇒ κ2 = ,
κ2 κ (r ± 1)

for r ̸= ±1, whilst κ2 = ±i/2 when r = ±1. This implies that the dominant modes for κ are attained
when r ∈ {0, ±2}. There is one case that needs extra care, though. That is the case in which system (3) is
invariant under the change u → −u. In such a case, the modes r = 0, ±2 lead to trivial solutions, making
the following analysis unfeasible [6]. We will consider the general case for now and will develop the changes
needed in the symmetric case in Appendix A.
Now, we define κ2± = ±i, and take the following values of γ in order of dominance, taking into account
that an increase of 1 in the value of n increases the power of X0 − X in the denominator of (38) by 1/2:
1 p−2
γ0,±2 = γ, γ±1 = γ − , and γ±p = γ − , for p ≥ 3. (43)
2 2

3.3 The case κ2 = κ2+

In this case, we have
 √   
M ± i, r = − Ju f0,0 (0) − (r − 1)2 k 2 D̂ .

In particular,
 √   √   
M ± i, 0 = M ± i, 2 = − Ju f0,0 (0) − k 2 D̂ .

21
 [0] [0] [1] [0] [0] [1] [0] [0]
This implies that v0 = c0 ϕ1 , and v2 = c2 ϕ1 , where c0 , c2 ∈ R, and
..
.
[0]
v−4 = 0,
[0]
v−4 = 0,
[0]
v−2 = 0,
[0] [0] [2]
v−1 = −2 κ3 CĀ c0 W2 ,
 
[0] [0] [0] [2]
v1 = −2 κ3 CĀ c2 + CA c0 W0 ,
[0] [0] [2]
v3 = −2 κ3 CA c2 W2 ,
[0] 2 [0] [3]
v4 = −3 i CA c2 W 3 ,
..
.

With this, we continue to the following order.

Order O(1/n). Following the same procedure at this order, we obtain

..
.
[1] [3] [0] 2
v−2 = −6 i W3 c0 CĀ ,
  
[1] 2 [0] [4] [0] [4] [4]
v−1 = −2 κ CĀ CĀ c2 W2,2 + c0 3 CA CĀ W2,2 + (η + i (γ − 1)) W2,3 ,
   
[1] [0] [3] [3] 2 [0] [3]
v0 = − c0 (1 − 2 γ) W1,3 + 4 i CA CĀ W1,2 + 2 i CĀ c2 W1,2 ,
    
[1] [0] [4] [4] [0] [4] [4]
v1 = −2 κ CA c0 4 CA CĀ W0,2 + (η + γ i) W0,3 + CĀ c2 4 CA CĀ W0,2 + (3 η − γ i) W0,3 ,
   
[1] [0] [0] [3] [0] [3]
v2 = −i 2 CA 2 c2 CĀ + CA c0 W1,2 + c2 (−2 γ i + 4 η + i) W1,3 ,
  
[1] 2 [0] [4] [0] [4] [4]
v3 = −2 κ CA CA c0 W2,2 + c2 3 CA CĀ W2,2 + (−γ i + 3 η + i) W2,3 ,
..
.

With this, we are ready to determine a solvability condition at the following order.

Order O(1/n2 ) Using the information we have already obtained at previous orders, at this order we obtain
two solvability conditions that are given by

2 2

[0]
Sc,0 = − (4 γ (γ − 2) + 3) α1 + 4 CA CĀ ((η + (γ − 1) i) α3 − (2 η + i) α4 ) + 12 CA CĀ α7 c0
2 [0]
+ 2 CĀ ((−2 η + i) α3 + (4 η + i (1 − 2 γ)) α4 − 4 CA CĀ α7 ) c2 = 0,

and
2 [0]
S0,2 = −2 CA ((2 η + i) α3 + i (1 − 2 γ) α4 + 4 CA CĀ α7 ) c0
− ((2 γ − 3 + 4 η i)(2 γ − 1 + 4 η i) α1
2 2

[0]
+4 CA CĀ ((3 η + i (1 − γ)) α3 + (−2 η + i) α4 ) + 12 CA CĀ α7 c2 = 0.
[0] [0]
Note that this is a linear system of equations for c0 and c2 , which has nontrivial solutions when β32 = 4 β1 β5
if and only if

γ ∈ {−1 − η i, −η i, 2 − η i, 3 − η i}.

22
Therefore, as γ was defined to be the value of γr with the highest real part, then we take γ = 3 − η i, which
yields
[0] h1 [0]
c2 = c , (44)
k1 0
where
2 2
h1 = (4 γ (γ − 2) + 3) α1 + 4 CA CĀ ((η + (γ − 1) i) α3 − (2 η + i) α4 ) + 12 CA CĀ α7 ,
2
k1 = 2 CĀ ((−2 η + i) α3 + (4 η + i (1 − 2 γ)) α4 − 4 CA CĀ α7 ) .

3.4 The case κ2 = κ2−

In this case, we have
 √   
M ± −i, r = − Ju f0,0 (0) − (r + 1)2 k 2 D̂ .
[0] [0] [1] [0] [0] [1]
Therefore, v−2 = c−2 ϕ1 , and v0 = c0 ϕ1 , and
..
.
[0] 2 [0] [3]
v−4 = 3 i CĀ c−2 W3 ,
[0] [0] [2]
v−3 = −2 κ3 CĀ c−2 W2 ,
 
[0] [2] [0] [0]
v−1 = −2 κ3 W0 CĀ c0 + CA c−2 ,
[0] [0] [2]
v1 = −2 κ3 CA c0 W2 ,
[0]
v2 = 0,
[0]
v3 = 0,
[0]
v4 = 0,
..
.

Order O(1/n) At this order, we obtain

..
.
  
[1] 2 [0] [4] [0] [4] [4]
v−3 = −2 κ CĀ CĀ c0 W2,2 + c−2 3 CA CĀ W2,2 + (3 η + i (γ − 1)) W2,3 ,
   
[1] [0] [3] [3] 2 [0] [3]
v−2 = − c−2 (2 γ − 1 − 4 η i) W1,3 − 4 i CA CĀ W1,2 − 2 i CĀ c0 W1,2 ,
    
[1] [0] [4] [4] [0] [4] [4]
v−1 = −2 κ CA c−2 4 CA CĀ W0,2 + (3 η + γ i) W0,3 + c0 CĀ 4 CA CĀ W0,2 + (η − γ i) W0,3 ,
   
[1] [0] [3] [0] [0] [3]
v0 = − (1 − 2 γ) c0 W1,3 − 2 i CA 2 CĀ c0 + CA c−2 W1,2 ,
  
[1] 2 [0] [4] [0] [4] [4]
v1 = −2 CA κ CA c−2 W2,2 + c0 3 CA W2,2 CĀ + (η + (1 − γ) i) W2,3 ,
[1] 2 [0] [3]
v2 = 6 i CA c0 W 3 ,
..
.


which implies that the solvability conditions at order O 1/n2 are given by

Sc,−2 = (−(2 γ − 3 − 4 i η)(2 γ − 1 − 4 i η) α1

2 2

[0]
+4 CA CĀ (−i (γ − 3 η i − 1) α3 + (2 η + i) α4 ) − 12 CA CĀ α7 c−2
2 [0]
− 2 CĀ ((2 η − i) α3 + i (2 γ − 1) α4 + 4 CA CĀ α7 ) c0 = 0,

23
 2 [0]
Sc,0 = −2 CA ((2 η + i) α3 + (−4 η + i (1 − 2 γ)) α4 + 4 CA CĀ α7 ) c−2
2 2

[0]
− (4 γ (γ − 2) + 3) α1 + 4 CA CĀ ((η − i (γ − 1)) α3 − (2 η − i) α4 ) + 12 CA CĀ α7 c0 = 0.
Once again, as before, it can be proven that this system of equations has nontrivial solutions when β32 =
4 β1 β5 if and only if
γ ∈ {−1 + η i, η i, 2 + η i, 3 + η i}.
Furthermore, when γ takes one of these values, we have
[0] h̄1 [0]
c−2 = c .
k̄1 0
[0] [0]
Moreover, from (40), we conclude that c0 and c±2 are key terms of our inner solution, which can be obtained
via the following limit:
D E
−(n−1) [1]
1 κ V n,r , ϕ 1
c[0]
r = D [1] E lim n−1

, (45)
[1] n→∞ Γ 2 +γ
ϕ ,ϕ 1 1

for r = −2, 0, 2 and the corresponding value of κ, provided that these limits exist.

3.5 Combining eigenvectors

Now, we note that the solution comprising the dominant values of κ is given by
n n
[n]
X
irx
X Vn,r
u = Wn,r (X) e = n eirx
2 −r ηi
r=−n r=−n (X0 − X)
n
+ γr X vr[s] irx
n−1
n−1


X κ Γ 2
= n e
2 −r
ηi ns
r=−n (X0 − X) s≥0
!
X √ n−1 Γ n−1 + γ  √ n−1 Γ n−1 + γ



2 [0] irx 2 [0] irx
∼
n→∞
i n v e + − i
−r η i r n
−r η i r
v e
X→X0 r=0,2
(X0 − X) 2 (X0 − X) 2
!
X √ n−1 Γ n−1 + γ
 √ n−1 Γ n−1 + γ

2 [0] irx 2 [0] irx
+ −i n v e + − −i
−r η i r n v e
−r η i r
r=−2,0 (X0 − X) 2 (X0 − X) 2
!
Γ n−1


2 +γ i(n−1)π/4 n h1 e2ix [1]
= n e (λ1 + (−1) λ2 ) 1 + ϕ1
(X0 − X) 2 k1 (X0 − X)−2 η i
!
Γ n−1


+ γ̄ −2ix
−i(n−1)π/4 h̄ 1 e [1]
+ 2
n e (λ3 + (−1)n λ4 ) 1 + 2ηi
ϕ1 , (46)
(X0 − X) 2 k̄ 1 (X0 − X)

[0]
where λi correspond to the corresponding coefficients cr we found before for the corresponding dominant
value of κ.
Last but not least, for the inner solution we have
W2 n,2 r−1 = 0, and W2 n−1,2 r = 0, ∀ n ∈ N, r ∈ Z.
Therefore, as u[n] , has even harmonics, we need λ2 = λ1 , and λ4 = λ3 , which implies
!
Γ n−1


[n] n−1
2 +γ n h1 e2ix [1]
u ∼ i 2 λ1 n (1 + (−1) ) 1 + −2 η i
ϕ1
(X0 − X) 2 k1 (X0 − X)
! (47)
n−1


n−1 Γ 2 + γ̄ n h̄1 e−2ix [1]
+ (−i) 2 λ3 n (1 + (−1) ) 1 + ϕ1 .
(X0 − X) 2 k̄1 (X0 − X)2 η i
Thus, from now on, we assume that n is even.

24
Remark 2. Note that, for each dominant value of κ we consider, there are two associated dominant values
of r. Therefore, as there are four dominant values of κ, the summations in (46) and (47) comprise eight
terms in total. Nevertheless, something similar will happen to the other four terms so they are omitted.

4 Late term expansion — outer solution

The previous procedure allowed us to approximate the solution when close to singularities. Nevertheless,
we note that the ansatz (38) is not so accurate away from them as it lacks the information provided by
the parameters of the system. For that reason, we need to consider a more general ansatz away from the
singularities. Let

κn−1 Γ n−1 X gr[s/2] (X)

2 + γr
Wn,r = n−1 , (48)
(X0 − X) 2
+γr
s≥0
ns/2

where we have used the summation over natural halved orders to take into account all the terms we obtained
in our expansion for the amplitude up to order five (see Section 2). With this, note that

Γ n−3

 ′ 
1 2 + γr 1
X 
′
Wn−2,r = κn−3 n−1 (n − 3 + 2 γ r ) g [s/2]
r + 2 (X 0 − X) g [s/2]
r ,
2 (X0 − X) 2
+γr
s≥0
(n − 2)s/2

1 n−5 Γ n−5


2 + γr 1
X 
′′
Wn−4,r = κ n−1 s/2
(n − 5 + 2 γr ) (n − 3 + 2 γr ) gr[s/2]
4 (X0 − X) 2 +γr
s≥0
(n − 4)
 ′  ′′ 
2
+4 (n − 5 + 2 γr ) (X0 − X) gr[s/2] + 4 (X0 − X) gr[s/2] .

Thus, when replacing (48) into (37), we obtain

  
X n−1−p
n−1 n−p−q−1 [s/2]


X Y Y Γ + γr X gr
0= Cp,q  aps s   bqss  n−1−p−q
Ju fp,q (0) κ 2
n−p−q−1
p=0 q=0 s≥1 s≥1 (X0 − X) 2 +γ
r
s≥0
(n − p − q)s/2
p·e=p q·e=q
  
X n−1−p
n−1 X n−p−q
X X Y Y
+ Cp,q  aps s   bqss 
p=0 q=0 M =2 r·eM =r s≥1 s≥1
p·e=p q·e=q n·eM =n−p−q −ni ≤ri ≤ni

× FM,p,q (An1 ,r1 Wn1 ,r1 , . . . , AnM ,rM WnM ,rM )


Γ n−1 X gr[s/2]


2 + γr
+ k 2 D̂ −r2 κn−1 n−1
(X0 − X) 2
+γr
s≥0
ns/2
 
n−3

 
1 Γ 2 + γr X 1   ′
+ 2 i r  κn−3 n−1 s/2
(n − 3 + 2 γr ) gr[s/2] + 2 (X0 − X) gr[s/2] 
2 (X0 − X) 2 +γr
s≥0
(n − 2)

1 n−5 Γ n−5


2 + γr 1
X 
+ κ n−1 s/2
(n − 5 + 2 γr ) (n − 3 + 2 γr ) gr[s/2]
4 (X0 − X) 2 +γr
s≥0
(n − 4)
 ′  ′′ 
2
+4 (n − 5 + 2 γr ) (X0 − X) gr[s/2] + 4 (X0 − X) gr[s/2] ,

25
which is equivalent to
  
 −1 n−1 n−1−p n−p−q


n−1 X X Y Y Γ + γr
0=Γ + γr Cp,q  aps s   bqss  4−p−q
Ju fp,q (0) κ 2
−p−q
2 p=0 q=0 (X0 − X) 2
s≥1 s≥1
p·e=p q·e=q
[s/2]
X gr
×
s≥0
(n − p − q)s/2
  

n−1
−1
n−1 X n−1−p
n−1 X n−p−q
X X Y Y
2 +γr
+ κ5−n Γ + γr (X0 − X) Cp,q  aps s   bqss 
2 p=0 q=0 r·eM =r
M =2 s≥1 s≥1
p·e=p q·e=q n·eM =n−p−q −ni ≤ri ≤ni

× FM,p,q (An1 ,r1 Wn1 ,r1 , . . . , AnM ,rM WnM ,rM )


X gr[s/2]
+ k 2 D̂ −r2 κ4
s≥0
ns/2
 ′ 
X 1 2 
+ 2 i r κ2 gr
[s/2]
+ (X 0 − X) g [s/2]
r
s≥0
(n − 2)s/2 n − 3 + 2 γr
 ′
X 1 [s/2] 4 
[s/2]
+ v r + (X 0 − X) g r
s≥0
(n − 4)s/2 n − 3 + 2 γr
4  ′′ 
2 [s/2]
+ (X0 − X) gr , (49)
(n − 5 + 2 γr ) (n − 3 + 2 γr )

Now, as usual, we go forward by following an order-by-order expansion.

4.1 The case κ2 = κ2+

Order O(1). At this order, as for the inner solution, (49) becomes

M (κ, r) gr[0] = NL(1).

[0] [0] [1] [0] [0] [1]
Now, as κ2 = κ2+ , then g0 = h0 ϕ1 , and g2 = h2 ϕ1 , which implies that

..
.
[0]
g−4 = 0,
[0]
g−3 = 0,
[0]
g−2 = 0,
[0] [0] [2]
g−1 = −2 κ3 Ā1 h0 W2 ,
 
[0] [0] [0] [2]
g1 = −2 κ3 Ā1 h2 + A1 h0 W0 ,
[0] [0] [2]
g3 = −2 κ3 A1 h2 W2 ,
[0] [0] [3]
g4 = −3 i A21 h2 W3 ,
..
.

26


Order O 1/n1/2 At this order,

..
.
[1/2]
g−3 = 0,
[1/2]
g−2 = 0,
[1/2]
√ p [0] [3]
g−1 = −2 2 i X0 − X Ā1 h0 W2 ,
[1/2]
√ p [0] [2]
g0 = − 2 κ3 X0 − X h0 W1 ,
[1/2]
√ p 
[0] [0]

[3]
g1 = −2 2 i X0 − X Ā1 h2 + A1 h0 W0 ,
[1/2]
√ p [0] [2]
g2 = − 2 κ3 X0 − X h2 W1 ,
[1/2]
√ p [0] [3]
g3 = −2 2 i X0 − X A1 h2 W2 ,
..
.

Order O(1/n). At this order,

..
.
[1] [0] [3]
g−2 = 6 i (X − X0 ) Ā21 h0 W3 ,

[1] [0] [4] 2 [0] [4]
g−1 = 2 κ (X − X0 ) Ā31 h2 W2,2 + 3 |A1 | Ā1 h0 W2,2
  ′  
[0] [4] [0] [4] ′ [0] [4]
+Ā1 2 h0 W2 − i h0 W2,3 − i Ā1 h0 W2,3 ,
  ′ 
[1] 2 [0] [3] 2 [0] [3] [0] [3] [0] [3]
g0 = −2 (X − X0 ) −2 i |A1 | h0 W1,2 − i Ā1 h2 W1,2 − i h0 W1 − h0 W1,3 ,
     ′ 
[1] [0] 2 [4] [4] 2 [0] [4] [0] [4]
g1 = 2 κ (X − X0 ) A1 2 h0 2 |A1 | W0,2 + W0 + 4 Ā1 h2 W0,2 − i h0 W0,3
  ′    
[0] [4] [0] [4] [0] [0] [4]
+Ā1 2 h2 W0 + i h2 W0,3 − i Ā′1 h2 − A′1 h0 W0,3 ,
    ′ 
[1] [0] 2 [3] [3] 2 [0] [3] [0] [3]
g2 = 2 i (X − X0 ) h2 2 |A1 | W1,2 + W1 + A1 h0 W1,2 + i h2 W1,3 ,
     ′  
[1] [0] 2 [4] [4] [0] [4] 3 [0] [4] ′ [0] [4]
g3 = 2 κ (X − X0 ) A1 h2 3 |A1 | W2,2 + 2 W2 + i h2 W2,3 + A1 h0 W2,2 + i A1 h2 W2,3 ,

..
.


Order O 1/n3/2 At this order,

..
.
√
[3/2] κ X0 − X  [0]  2 [4] [2] [4]

g0 = √ h0 16 (X − X0 ) |A1 | W1,2 + i (4 γ − 5) W1 + 8 (X − X0 ) W1
2 2
  ′ 
[0] [4] [0] [4]
+8 (X − X0 ) Ā21 h2 W1,2 − i h0 W1,3
√
[3/2] κ X0 − X  [0]  2 [4] [2] [4]

g2 = √ h2 16 (X − X0 ) |A1 | W1,2 + i (4 γ − 5) W1 + 8 (X − X0 ) W1
2 2
 ′ 
[0] [4] [0] [4]
+8 (X − X0 ) A21 h0 W1,2 + 8 i (X − X0 ) h2 W1,3

27
 ..
.

With this, we are ready to get the solvability conditions at the next order.


Order O 1/n2 At this order, we obtain two solvability conditions, which given by
 ′′    ′
[0] 2 [0]
Sc,0 = α1 h0 − i α2 + α3 |A1 | h0

 [0]
2 4
+ α5 + 2 α6 |A1 | + 3 α7 |A1 | − i α3 A1 Ā′1 + 2 α4 Ā1 A′1 h0
 ′  
[0] 2 [0]
− i α4 Ā21 h2 + α6 Ā21 + 2 α7 Ā21 |A1 | − i α3 Ā1 Ā′1 h2 = 0, (50)

and
 ′′    ′
[0] 2 [0]
Sc,2 = α1 h2 + i α2 + α3 |A1 | h2

 [0]
2 4
+ α5 + 2 α6 |A1 | + 3 α7 |A1 | + i α3 Ā1 A′1 + 2 α4 A1 Ā′1 h2
 ′  
[0] 2 [0]
+ i α4 A21 h0 + α6 A21 + 2 α7 A21 |A1 | + i α3 A1 A′1 h0 = 0. (51)

4.2 The case κ2 = κ2−

[0] [0] [1] [0] [0] [1]
In this case, we have g−2 = h−2 ϕ1 and g0 = h0 ϕ1 . Furthermore, as before, following an asymptotic
expansion using an order-by-order approach, we have that.

Order O(1). At this order,

..
.
[0] [0] [3]
g−4 = 3 i Ā21 h−2 W3 ,
[0] [0] [2]
g−3 = −2 κ3 Ā1 h−2 W2 ,
 
[0] [0] [0] [2]
g−1 = −2 κ3 Ā1 h0 + A1 h−2 W0 ,
[0] [0] [2]
g1 = −2 κ3 A1 h0 W2 ,
[0]
g2 = 0,
[0]
g3 = 0,
[0]
g4 = 0,
..
.

28


Order O 1/n1/2 . At this order,

..
.
[1/2]
√ p [0] [3]
g−3 = 2 i 2 X0 − X Ā1 h−2 W2 ,
[1/2]
√ p [0] [2]
g−2 = − 2 κ3 X0 − X h−2 W1 ,
[1/2]
√ [3]
p 
[0] [0]

g−1 = 2 2 i W0 X0 − X Ā1 h0 + A1 h−2 ,
[1/2]
√ p [0] [2]
g0 = − 2 κ3 X0 − X h0 W1 ,
[1/2]
√ p [0] [3]
g1 = 2 2 i X0 − X A1 h0 W2 ,
[1/2]
g2 = 0,
[1/2]
g3 = 0,
..
.

Order O(1/n). At this order,

..
.

[1] [0] [4] 2 [0] [4]
g−3 = 2 κ (X − X0 ) Ā31 h0 W2,2 + 3 |A1 | Ā1 h−2 W2,2
  ′  
[0] [4] [0] [4] ′ [0] [4]
+Ā1 2 h−2 W2 − i h−2 W2,3 − i Ā1 h−2 W2,3 ,
    ′ 
[1] 2 [0] [3] 2 [0] [3] [0] [3] [0] [3]
g−2 = −2 (X − X0 ) 2 i |A1 | h−2 W1,2 + i Ā1 h0 W1,2 + h−2 W1 + h−2 W1,3 ,
     ′ 
[1] [0] 2 [4] [4] 2 [0] [4] [0] [4]
g−1 = 2 κ (X − X0 ) A1 2 h−2 2 |A1 | W0,2 + W0 + 4 Ā1 h0 W0,2 − i h−2 W0,3
  ′    
[0] [4] [0] [4] [0] ′ [0] ′ [4]
+Ā1 2 h0 W0 + i h0 W0,3 − i h0 Ā1 − h−2 A1 W0,3 ,
    ′ 
[1] [0] 2 [3] [3] 2 [0] [3] [0] [3]
g0 = −2 i (X − X0 ) h0 2 |A1 | W1,2 + W1 + A1 h−2 W1,2 + i h0 W1,3 ,
     ′ 
[1] [0] 2 [4] [4] [0] [4]
g1 = 2 κ (X − X0 ) A1 h0 3 |A1 | W2,2 + 2 W2 + i h0 W2,3

[0] [4] [0] [4]
+A31 h−2 W2,2 + i A′1 h0 W2,3 ,
[1] [0] [3]
g2 = −6 i (X − X0 ) A21 h0 W3 ,
..
.

29
Order O(1/n3/2 ). At this order,
..
.
√ 
[3/2] κ X0 − X [0]

2 [4] [2] [4]

g−2 = √ h−2 16 (X − X0 ) |A1 | W1,2 − i (4 γ − 5) W1 + 8 (X − X0 ) W1
2 2
  ′ 
2 [0] [4] [0] [4]
+8 (X − X0 ) Ā1 h0 W1,2 − i h−2 W1,3 ,
√
[3/2] κ X0 − X  [0]  2 [4] [2] [4]

g0 = √ h0 16 (X − X0 ) |A1 | W1,2 − i (4 γ − 5) W1 + 8 (X − X0 ) W1
2 2
 ′ 
2 [0] [4] [0] [4]
+8 (X − X0 ) A1 h−2 W1,2 + 8 i (X − X0 ) h0 W1,3 .

..
.


Order O 1/n2 . Finally, at this order, we obtain two solvability conditions given by
 ′′   ′
[0] 2 [0]
Sc,−2 = α1 h−2 − i α2 + α3 |A1 | h−2

 [0]
2 4
+ α5 + 2 α6 |A1 | + 3 α7 |A1 | − i α3 A1 Ā′1 + 2 α4 Ā1 A′1 h−2
 ′  
[0] 2 [0]
− i α4 Ā21 h0 + α6 Ā21 + 2 α7 |A1 | Ā21 − i α3 Ā1 Ā′1 h0 = 0, (52)

 ′′    ′
[0] 2 [0]
Sc,0 = α1 h0 + i α2 + α3 |A1 | h0

 [0]
2 4
+ α5 + 2 α6 |A1 | + 3 α7 |A1 | + i α3 Ā1 A′1 + 2 α4 A1 Ā′1 h0
 ′  
[0] 2 [0]
+ i α4 A21 h−2 + α6 A21 + 2 α7 |A1 | A21 + i α3 A1 A′1 h−2 = 0. (53)

4.3 Solving a coupled system of amplitude equations

Note that the pairs of equations (50)–(51) and (52)–(53) are symmetric. Thus, it is enough to solve one pair
of them in order to know the behaviour of the solutions for both systems. In particular, we focus on solving
 ′′    ′ 
 [0]
[0] 2 [0] 2 4
α1 h0 − i α2 + α3 |A1 | h0 + α5 + 2 α6 |A1 | + 3 α7 |A1 | − i α3 A1 Ā′1 + 2 α4 Ā1 A′1 h0
 ′  
[0] 2 [0]
−i α4 Ā21 h2 + α6 Ā21 + 2 α7 Ā21 |A1 | − i α3 Ā1 Ā′1 h2 = 0, (54)
 ′′    ′  
[0] 2 [0] 2 4 [0]
+ α5 + 2 α6 |A1 | + 3 α7 |A1 | + i α3 Ā1 A′1 + 2 α4 A1 Ā′1 h2


α1 h2 + i α2 + α3 |A1 | h2
 ′  
[0] 2 [0]
+i α4 A21 h0 + α6 A21 + 2 α7 A21 |A1 | + i α3 A1 A′1 h0 = 0. (55)

To solve these equations, we note that (55) is the complex conjugate version of (54). Furthermore, based on
[5, 6], and by looking at the equations R1 and φ1 solve, (23) and (24), we set
[0] [0]
h0 = (R2 − i R1 φ2 ) e−i φ1 and h2 = (R2 + i R1 φ2 ) ei φ1 ,
where R2 and φ2 are real functions.
Therefore, when we replace this into (54), we obtain

− 48 α12 β5 R15 φ2 − 32 α12 β3 R13 φ2 + 16 α12 φ2 R1′′ + 8 α1 α3 R12 R2′ + 8 α1 α4 R12 R2′
+ 32 α12 R1′ φ′2 + 24 α1 α3 R2 R1 R1′ + 24 α1 α4 R2 R1 R1′ − 16 α12 β1 R1 φ2 + 16 α12 R1 φ′′2
+ i −240 α12 β5 R2 R14 − 4 α32 R2 R14 + 12 α42 R2 R14 + 8 α3 α4 R2 R14 − 8 α1 α3 R13 φ′2
+24 α1 α4 R13 ϕ′2 − 96 α12 β3 R2 R12 − 16 α12 β1 R2 + 16 α12 R2′′ = 0.

30
Here, as all the coefficients and functions are real, we have that the real and imaginary parts of the previous
equation must equal zero to fulfill it.
Then, after splitting the equation and simplifying it using (23), (24), and β32 = 4 β1 β5 , we obtain

4 α1 R1′ φ′2 + 3 (α3 + α4 ) R1 R2 R1′ + R1 (2 α1 φ′′2 + (α3 + α4 ) R1 R2′ ) = 0, (56)

and

4 α12 (R2′′ − β1 R2 )
− R12 R2 24 α12 β3 + R12 60 α12 β5 + (α3 + α4 ) (α3 − 3 α4 ) + 2 α1 (α3 − 3 α4 ) R1 φ′2 = 0. (57)

Here, note that (56) can be directly solved for φ′2 . Specifically, we have

(α3 + α4 ) R1 R2 4 α1 K3
φ′2 = − + ,
2 α1 R12

where K3 ∈ R is a constant of integration. Now, when replacing this expression into (57), we obtain

R2′′ − β1 + 6 β3 R12 + 15 β5 R14 R2 = 2 (α3 − 3 α4 ) K3 R1 ,

(58)

which is a linear non-homogeneous equation in R2 . To solve it, note that the choice K3 = 0 makes the
equation homogeneous, with a solution given by
Z X !
′ ds
R2,h = R1 K1 + K2 ′ 2
.
X0 (R1 )

Now, observe that if K1 = K2 = K3 = 0, then φ2 = K4 ∈ R is a constant, which leads to

[0]
h0 = −i K4 R1 e−i φ1 = −i K4 Ā1 ,
[0]
h2 = i K4 R1 ei φ1 = i K4 A1 .

On the other hand, when K2 = K3 = 0, we have that R2,h = K1 R1′ , which implies

′
′ (α3 + α4 ) R12
φ2 = −K1 = −K1 φ′′1 .
4 α1
This yields φ2 = −K1 φ′1 (setting the constant of integration to zero), which implies
[0]
h0 = (K1 R1′ − i K1 R1 φ′1 ) e−i φ1 = K1 Ā′1 ,
[0]
h2 = (K1 R1′ + i K1 R1 φ′1 ) ei φ1 = K1 A′1 .

Now, if K3 ̸= 0, then we can use the method of variation of parameters to find the solution to the non-
homogeneous equation, (58). In particular, we have
Z X
R12
R2,p = (α3 − 3 α4 ) K3 R1′ ′ 2
ds.
X0 (R1 )

Therefore, we conclude that, when K1 = 0,

X
K2 + (α3 − 3 α4 ) K3 R12
Z
R2 = R1′ 2 ds,
X0 (R1′ )
which implies
!
X s X
K2 + (α3 − 3 α4 ) K3 R12
Z Z Z
α3 + α4 ds
φ2 = − R1 R1′ 2 dσ ds + 4 α1 K3 .
2 α1 X0 X0 (R1′ ) X0 R12

31
Thus, when gathering all the information we have obtained, we conclude
Z X
[0] ′ ′ K2 + (α3 − 3 α4 ) K3 R12
h2 = K1 A1 + i K4 A1 + R1 2 ds
X0 (R1′ )
Z X Z s ! Z X !!
α3 + α4 ′ K2 + (α3 − 3 α4 ) K3 R12 ds
+i R1 − R1 R1 ′ )2
dσ ds + 4 α1 K3 2 ei φ 1 ,
2 α1 X0 X0 (R 1 X0 R1

and
X
K2 + (α3 − 3 α4 ) K3 R12
Z
[0]
h0 = K1 Ā′1 − i K4 Ā1 + R1′ 2 ds
X0 (R1′ )
! !!
X s X
K2 + (α3 − 3 α4 ) K3 R12
Z Z Z
α3 + α4 ds
−i R1 − R1 R1′ 2 dσ ds + 4 α1 K3 e−i φ1 .
2 α1 X0 X0 (R′ ) X0 R12

Remark 3. We highlight that we solved the amplitude equations in this section by assuming that R2 and φ2
were real functions. Nevertheless, this is not the case when making an integration involving X0 in the limits
of integration. In reality, we are only interested in the cases where X is real. However, as we have been
focusing on the study of singularities, we need to keep focusing on them in every step in order to complete
our analysis.

4.4 Matching inner and outer solutions

We have obtained an inner and an outer solution for our late-term expansion. To match these, we need
  [0]   [0]
n−1 n−1 gr (X) n−1 n−1 vr
κ Γ + γr n−1 ∼κ Γ + γr n
−r η i
,
2 (X0 − X) 2
+γr 2 (X0 − X) 2
which implies
[0]
vr
gr[0] (X) ∼ 1
−r η i−γr
.
(X0 − X) 2
In particular, we note that
[0] [0] [0]
[0] v−2 [0] v0 [0] v2
g−2 (X) ∼ 1
+2 η i−γ
, g0 (X) ∼ 1
−γ
, g2 (X) ∼ 1
−2 η i−γ
.
(X0 − X) 2 (X0 − X) 2 (X0 − X) 2
Furthermore, when γ = 3 − η i, we have
[0] [0] [0]
[0] v−2 [0] v0 [0] v2
g−2 (X) ∼ − 52 +3 η i
, g0 (X) ∼ − 25 +η i
, g2 (X) ∼ − 25 −η i
.
(X0 − X) (X0 − X) (X0 − X)
Now, we observe that
 3/2 !
[0] i πξ
 p 1/2
−1/2 ηi −1
h2 ∼ K1 (2 η + i) e 2 β1 (−2 β3 ) (X0 − X)
2 X0 − X
√
−1/2
 
πξ 1/2 2+η i
1 (3 + 2 η i) e 2 β1 (−2 β3 ) (X0 − X)
+ K2 − q 
6 −1
X0 −X
√
1/2
 

−1/2 1+η i
i α3 (3 + 4 η (η − i)) − 3 α4 (2 η − i)2 eπ ξ 2 β1 (−2 β3 ) (X0 − X)
+ K3  q 
−1
6 η X0 −X
 r 
πξ
 p 1/2
−1/2 −1 ηi
+ K4 i e 2 β1 (−2 β3 ) (X0 − X) ,
X0 − X

32
which implies that the analysis done previously for the inner solution is matched only by the dominant term
associated with K2 , which lets us conclude that
Z X Z X Z s ! !
[0] ′ ds α3 + α4 ′ dσ
h2 ∼ K2 R1 ′ 2
− i R1 R1 R1 ′ 2
ds ei φ1 .
X0 (R1 ) 2 α1 X0 X0 (R1 )

Furthermore, by matching this solution with (47) as X → X0 , we obtain

√
1/2 −1/2
6 i e−π ξ 2 β1 (−2 β3 )
K2 = − λ1 . (59)
(3 + 2 η i)
Furthermore, we conclude that the contribution to the solution provided by X0 is given by
Γ n−1


2 +γ
n−1
 
[0] [0] [1]
u[n] ∼ i 2 n−1 (1 + (−1)n ) h0 + h2 e2ix ϕ1
2 +γ
(X0 − X)
(60)
Γ n−1


2 + γ̄
n−1
 
[0] [0] [1]
+ (−i) 2
n−1 (1 + (−1)n ) h0 + h−2 e−2ix ϕ1 ,
+γ̄
(X0 − X) 2

as n → ∞. Similarly, due to symmetry, we have that the contribution to the solution provided by X̄0 = X−1
is given by
Γ n−1


[n] n−1
2 + γ̄ n

[0] [0] −2ix

[1]
u ∼ (−i) 2
n−1 (1 + (−1) ) h0 + h 2 e ϕ1
2 +γ̄
X̄0 − X
Γ n−1


2 + γ̄
n−1
 
[0] [0] [1]
+i 2

n−1 (1 + (−1)n ) h0 + h−2 e2ix ϕ1 ,
+γ̄
X̄0 − X 2
as n → ∞.
Remark 4. We highlight that the last few expressions are just an abuse of notation. Let us recall that R2
[2] [0]
and φ2 should be real functions (see Remark 3), which would make h0 = h2 . Nevertheless, as we are
working at complex singularities, this is no longer the case. One technically needs to be extra careful with
the term that is being complex-conjugated and consider the fact that there are more constants of integration
fulfilling different rules that can be deduced in the same way as when we matched solutions in this section.
In any case, the development shown here is enough for our purposes and it is consistent with [5, 6].

5 Estimation of the residual

In this section, we start the study of the residual that arises after truncating the asymptotic expansion we
have been carrying out. Specifically, if we assume that we have performed the expansion up to a sufficiently
large order, then the remainder is expected to be small. Therefore, we are only interested in studying the
linearized equation for the remainder, which turns out to approximate the residual well and will let us obtain
more conditions to ensure that such a term tends to zero as X → ±∞.
We begin by introducing new notation for the linearized equation of the residual. In particular, we define
the following symmetric multilinear vector functions in terms of the Jacobian of (3):
n
!
X ∂Ju fi,j (u)
Ju F1,i,j (v1 ) = v1,p1 ,
p1 =1
∂up1
u=0
n
!
1 X ∂ 2 Ju fi,j (u)
Ju F2,i,j (v1 , v2 ) = v1,p1 v2,p2 ,
2! p ,p =1 ∂up1 ∂up2
1 2 u=0
n
!
1 X ∂ 3 Ju fi,j (u)
Ju F3,i,j (v1 , v2 , v3 ) = v1,p1 v2,p2 v3,p3 ,
3! p ,p ,p =1 ∂up1 ∂up2 ∂up3
1 2 3 u=0

33
 n
!
1 X ∂ 4 Ju fi,j (u)
Ju F4,i,j (v1 , v2 , v3 , v4 ) = v1,p1 v2,p2 v3,p3 v4,p4 .
4! p1 ,p2 ,p3 ,p4 =1
∂up1 ∂up2 ∂up3 ∂up4
u=0

Lemma 3. For every pair of non-negative integers i, j, and every vectors v1 , v2 , v3 , v4 , v5 , we have

Ju F1,i,j (v1 ) v2 = 2 F2,i,j (v1 , v2 ) ,

Ju F2,i,j (v1 , v2 ) v3 = 3 F3,i,j (v1 , v2 , v3 ) ,
Ju F3,i,j (v1 , v2 , v3 ) v4 = 4 F4,i,j (v1 , v2 , v3 , v4 ) ,
Ju F4,i,j (v1 , v2 , v3 , v4 ) v5 = 5 F5,i,j (v1 , v2 , v3 , v4 , v5 ) .

Proof. The proof is similar for each equality. Therefore, for simplicity, we give explicit proof of the first
equality only. Observe that
n
! n X n
!
X ∂Ju fi,j (u) X ∂ 2 fi,j (u)
Ju F1,i,j (v1 ) v2 = v1,p v2 = v1,p v2,q = 2 F2,i,j (v1 , v2 ) ,
p=1
∂up p=1 q=1
∂uq ∂up
u=0 u=0

which proves the first equality and, therefore, as explained above, the result.

The next step is to estimate the remainder of the approximation we are carrying out. To do that, we
consider that our solution can be written as

u = S + RN ,

N
X
where S = εn u[n] is the approximation of the solution we have developed so far, and RN is the remainder,
n=1
which is assumed to be small. We can therefore linearize (3) about S, considering that S solves the equation
up to order N , obtaining
Mb
Ma X
∂ 2 RN ∂ 2 RN ∂ 2 RN
X  
p q 2
0= a b Ju fp,q (S) RN + k D̂ 2
+ 2 ε2 + ε4 + E(δb)
p=0 q=0
∂x ∂x∂X ∂X 2 (61)
+ forcing due to truncation,


where ‘forcing due to truncation’ corresponds to terms of order O εN +1 that come out of the evaluation
of S in the linear operator due to the truncation of the expansion, and E(δb) corresponds to the term that
appears when b gets away from the Maxwell point. Now, as usual, we need to solve (61) going order by order
by expanding RN as
∞
[p]
X
RN = ε p RN .
p=1

Fortunately, the corresponding solutions at each order can be written in terms of the solutions we already
obtained in Section 2, as stated below.

Order O(ε): At this order, (61) becomes

2
 
2 ∂ [1]
Ju f0,0 (0) + k RN = 0,
∂x2

which has a solution

[1] [1]
RN = B1 eix ϕ1 + c.c.,

where B1 = B1 (X).

34


Order O ε2 : At this order, (61) becomes
[2]
[2] [1] [1]
 
[1] ∂ 2 RN
0 = Ju f0,0 (0) RN + a1 Ju f1,0 (0) RN + b1 Ju f0,1 (0) RN + Ju F1,0,0 u[1] RN + k 2 D̂ .
∂x2
Therefore,
[2] [2] [2] [2] [1]
RN = 2 A1 B̄1 W0 + B1 eix W1 + 2 A1 B1 e2ix W2 + B2 eix ϕ1 + c.c.,

where B2 = B2 (X).


Order O ε3 : At this order, (61) becomes
[3] [2] [2] [1] [1]
0 = Ju f0,0 (0) RN + a1 Ju f1,0 (0) RN + b1 Ju f0,1 (0) RN + a2 Ju f1,0 (0) RN + b2 Ju f0,1 (0) RN
 
[1] [1] [1] [2]
+ a21 Ju f2,0 (0) RN + a1 b1 Ju f1,1 (0) RN + b21 Ju f0,2 (0) RN + Ju F1,0,0 u[1] RN
       
[1] [1] [1] [1]
+ Ju F1,0,0 u[2] RN + a1 Ju F1,1,0 u[1] RN + b1 Ju F1,0,1 u[1] RN + Ju F2,0,0 u[1] , u[1] RN
!
[3] [1]
2 ∂ 2 RN ∂ 2 RN
+ k D̂ +2 ,
∂x2 ∂x∂X

which implies that

[3] [3] [3] [3] 2 [3] [3]
RN = 2 A1 B̄1 W0 + B1 eix W1 + A21 B̄1 eix W1,2 + 2 |A1 | B1 eix W1,2 + i B1′ eix W1,3
[3] [2] [3] [2] [2]
+ 2 A1 B1 e2ix W2 + 2 A2 B1 e2ix W2 + 3 A21 B1 e3ix W3 + 2 A1 B̄2 W0 + 2 A2 B̄1 W0
[2] [1] [2]
+ B2 eix W1 + B3 eix ϕ1 + 2 A1 B2 e2ix W2 + c.c.,

where B3 = B3 (X).


Order O ε4 : At this order, (61) becomes
[4] [3] [3] [2] [2]
0 = Ju f0,0 (0) RN + a1 Ju f1,0 (0) RN + b1 Ju f0,1 (0) RN + a2 Ju f1,0 (0) RN + b2 Ju f0,1 (0) RN
[1] [1] [2] [2] [2]
+ a3 Ju f1,0 (0) RN + b3 Ju f0,1 (0) RN + a21 Ju f2,0 (0) RN + a1 b1 Ju f1,1 (0) RN + b21 Ju f0,2 (0) RN
[1] [1] [1] [1]
+ 2 a1 a2 Ju f2,0 (0) RN + a1 b2 Ju f1,1 (0) RN + a2 b1 Ju f1,1 (0) RN + 2 b1 b2 Ju f0,2 (0) RN
[1] [1] [1] [1]
+ a31 Ju f3,0 (0) RN + a21 b1 Ju f2,1 (0) RN + a1 b21 Ju f1,2 (0) RN + b31 Ju f0,3 (0) RN
       
[3] [1] [2] [2]
+ Ju F1,0,0 u[1] RN + Ju F1,0,0 u[3] RN + Ju F1,0,0 u[2] RN + a1 Ju F1,1,0 u[1] RN
       
[2] [1] [1] [1]
+ b1 Ju F1,0,1 u[1] RN + a1 Ju F1,1,0 u[2] RN + b1 Ju F1,0,1 u[2] RN + a2 Ju F1,1,0 u[1] RN
       
[1] [1] [1] [1]
+ b2 Ju F1,0,1 u[1] RN + a21 Ju F1,2,0 u[1] RN + a1 b1 Ju F1,1,1 u[1] RN + b21 Ju F1,0,2 u[1] RN
     
[2] [1] [1]
+ Ju F2,0,0 u[1] , u[1] RN + 2 Ju F2,0,0 u[1] , u[2] RN + a1 Ju F2,1,0 u[1] , u[1] RN
!
2 [4] 2 [2]
 
[1]
 
[1] ∂ R ∂ R
+ b1 Ju F2,0,1 u[1] , u[1] RN + Ju F3,0,0 u[1] , u[1] , u[1] RN + k 2 D̂ N
+2 N
,
∂x2 ∂x∂X

35
which has a solution given by
[4] [4] 2 [4] [4] [4] [4]
RN = 2 A1 B̄1 W0 + 4 |A1 | A1 B̄1 W0,2 + i Ā1 B1′ W0,3 + i A′1 B̄1 W0,3 + B1 eix W1
[4] 2 [4] [4] [4]
+ A21 B̄1 eix W1,2 + 2 |A1 | B1 eix W1,2 + i B1′ eix W1,3 + 2 A1 B1 e2ix W2
[4] 2 [4] [4] [4]
+ A31 B̄1 e2ix W2,2 + 3 |A1 | A1 B1 e2ix W2,2 + i A1 B1′ e2ix W2,3 + i A′1 B1 e2ix W2,3
[4] [4] [3] [3] [2]
+ 3 A21 B1 e3ix W3 + 4 A31 B1 e4ix W4 + 2 A1 B̄2 W0 + 2 A2 B̄1 W0 + 2 A2 B̄2 W0
[2] [2] [3] [3] 2 [3]
+ 2 A1 B̄3 W0 + 2 A3 B̄1 W0 + B2 eix W1 + A21 B̄2 eix W1,2 + 2 |A1 | B2 eix W1,2
[3] [3] [3] [3]
+ 2 A1 A2 B̄1 eix W1,2 + 2 A1 Ā2 B1 eix W1,2 + 2 Ā1 A2 B1 eix W1,2 + i B2′ eix W1,3
[2] [1] [3] [3] [2]
+ B3 eix W1 + B4 eix ϕ1 + 2 A1 B2 e2ix W2 + 2 A2 B1 e2ix W2 + 2 A2 B2 e2ix W2
[2] [2] [3] [3]
+ 2 A1 B3 e2ix W2 + 2 A3 B1 e2ix W2 + 3 A21 B2 e3ix W3 + 6 A1 A2 B1 e3ix W3 + c.c., (62)

where B4 = B4 (X).


Order O ε5 : At this order, we have
[5] [4] [4] [3] [3]
0 = Ju f0,0 (0) RN + a1 Ju f1,0 (0) RN + b1 Ju f0,1 (0) RN + a2 Ju f1,0 (0) RN + b2 Ju f0,1 (0) RN
[2] [2] [1] [1] [3]
+ a3 Ju f1,0 (0) RN + b3 Ju f0,1 (0) RN + a4 Ju f1,0 (0) RN + b4 Ju f0,1 (0) RN + a21 Ju f2,0 (0) RN
[3] [3] [2] [2]
+ a1 b1 Ju f1,1 (0) RN + b21 Ju f0,2 (0) RN + 2 a1 a2 Ju f2,0 (0) RN + a1 b2 Ju f1,1 (0) RN
[2] [2] [1] [1]
+ a2 b1 Ju f1,1 (0) RN + 2 b1 b2 Ju f0,2 (0) RN + 2 a1 a3 Ju f2,0 (0) RN + a1 b3 Ju f1,1 (0) RN
[1] [1] [1] [1]
+ a3 b1 Ju f1,1 (0) RN + 2 b1 b3 Ju f0,2 (0) RN + a22 Ju f2,0 (0) RN + a2 b2 Ju f1,1 (0) RN
[1] [2] [2] [2]
+ b22 Ju f0,2 (0) RN + a31 Ju f3,0 (0) RN + a21 b1 Ju f2,1 (0) RN + a1 b21 Ju f1,2 (0) RN
[2] [1] [1] [1]
+ b31 Ju f0,3 (0) RN + 3 a21 a2 Ju f3,0 (0) RN + 2 a1 a2 b1 Ju f2,1 (0) RN + a2 b21 F1,2 (0) RN
[1] [1] [1] [1]
+ a21 b2 Ju f2,1 (0) RN + 2 a1 b1 b2 Ju f1,2 (0) RN + 3 b21 b2 Ju f0,3 (0) RN + a41 Ju f4,0 (0) RN
[1] [1] [1] [1]
+ a31 b1 Ju f3,1 (0) RN + a21 b21 Ju f2,2 (0) RN + a1 b31 Ju f1,3 (0) RN + b41 Ju f0,4 (0) RN
       
[4] [1] [3] [2]
+ Ju F1,0,0 u[1] RN + Ju F1,0,0 u[4] RN + Ju F1,0,0 u[2] RN + Ju F1,0,0 u[3] RN
       
[3] [3] [1] [1]
+ a1 Ju F1,1,0 u[1] RN + b1 Ju F1,0,1 u[1] RN + a1 Ju F1,1,0 u[3] RN + b1 Ju F1,0,1 u[3] RN
       
[2] [2] [2] [2]
+ a1 Ju F1,1,0 u[2] RN + b1 Ju F1,0,1 u[2] RN + a2 Ju F1,1,0 u[1] RN + b2 Ju F1,0,1 u[1] RN
       
[1] [1] [1] [1]
+ a2 Ju F1,1,0 u[2] RN + b2 Ju F1,0,1 u[2] RN + a3 Ju F1,1,0 u[1] RN + b3 Ju F1,0,1 u[1] RN
       
[2] [2] [2] [1]
+ a21 Ju F1,2,0 u[1] RN + a1 b1 Ju F1,1,1 u[1] RN + b21 Ju F1,0,2 u[1] RN + a21 Ju F1,2,0 u[2] RN
     
[1] [1] [1]
+ a1 b1 Ju F1,1,1 u[2] RN + b21 Ju F1,0,2 u[2] RN + 2 a1 a2 Ju F1,2,0 u[1] RN
     
[1] [1] [1]
+ a1 b2 Ju F1,1,1 u[1] RN + a2 b1 Ju F1,1,1 u[1] RN + 2 b1 b2 Ju F1,0,2 u[1] RN
     
[1] [1] [1]
+ a31 Ju F1,3,0 u[1] RN + a21 b1 Ju F1,2,1 u[1] RN + a1 b21 Ju F1,1,2 u[1] RN
     
[1] [3] [1]
+ b31 Ju F1,0,3 u[1] RN + Ju F2,0,0 u[1] , u[1] RN + 2 Ju F2,0,0 u[1] , u[3] RN
     
[2] [1] [2]
+ 2 Ju F2,0,0 u[1] , u[2] RN + Ju F2,0,0 u[2] , u[2] RN + a1 Ju F2,1,0 u[1] , u[1] RN
     
[2] [1] [1]
+ b1 Ju F2,0,1 u[1] , u[1] RN + 2 a1 Ju F2,1,0 u[1] , u[2] RN + 2 b1 Ju F2,0,1 u[1] , u[2] RN
     
[1] [1] [1]
+ a2 Ju F2,1,0 u[1] , u[1] RN + b2 Ju F2,0,1 u[1] , u[1] RN + a21 Ju F2,2,0 u[1] , u[1] RN

36
      
[1] [1] [2]
+ a1 b1 Ju F2,1,1 u[1] , u[1] RN + b21 Ju F2,0,2 u[1] , u[1] RN + Ju F3,0,0 u[1] , u[1] , u[1] RN
     
[1] [1] [1]
+ 3 Ju F3,0,0 u[1] , u[1] , u[2] RN + a1 Ju F3,1,0 u[1] , u[1] , u[1] RN + b1 Ju F3,0,1 u[1] , u[1] , u[1] RN
!
[5] [3] [1]

[1] [1] [1] [1]

[1] 2 ∂ 2 RN ∂ 2 RN ∂ 2 RN
+ Ju F4,0,0 u , u , u , u RN + k D̂ +2 + ,
∂x2 ∂x∂X ∂X 2

Again as in Section 2, at this order, we need to determine a solvability condition associated with the resonant
terms that are multiples of eix . Said part of this equation is given by
 
[5] [5] [5] [5] 2 [5] [5] [5] 2
Ju f0,0 (0) − k 2 D̂ RN = −Q1 B1′′ − i Q2 B1′ − i Q3 |A1 | B1′ − i Q4 A21 B̄1′ − Q5 B1 − 2 Q6 |A1 | B1
[5] 4 [5] [5] [5]
− 3 Q7 |A1 | B1 − i Q3 Ā1 A′1 B1 − 2 i Q4 A1 Ā′1 B1 − Q6 A21 B̄1 (63)
[5] 2 [5]
− 2 Q7 |A1 | A21 B̄1 − i Q3 A1 A′1 B̄1 − non-homogeneous part − . . . ,

where ‘non-homogeneous part’ represents the terms associated with E(δb) that appear at this order and ‘. . .’
stands for terms that are multiples of A2 , A3 , A4 , B2 , B3 .B4 and are not required to establish this solvability
condition as they are non-resonant.
With this, when applying the inner product with ψ on both sides of (63), we obtain
2
α1 B1′′ + i α2 B1′ + i α3 |A1 | B1′ + i α3 A1 A′1 B̄1 + i α3 Ā1 A′1 B1 + i α4 A21 B̄1′ + 2 i α4 A1 Ā′1 B1
2 2 4 dδ (64)
+α5 B1 + α6 A21 B̄1 + 2 α6 |A1 | B1 + 2 α7 |A1 | A21 B̄1 + 3 α7 |A1 | B1 + (A1 ) = 0,
dA1
dδ
where is the function one obtains when applying the inner product with the non-homogeneous part of
dA1
dδ
dδ dδ
(63), and satisfies R1 eix = (R1 ) eix (note that the term (A1 ) is an abuse of notation used
dA1 dR1 dA1
i φ1
to simplify notation). Now, if we set B1 = (RB + i R1 φB ) e , where RB and φB are real functions, then
we can split the resulting equation into
 
dδ 2 ′
8 α1 2 + R1 (R1 (10 α7 R1 RB − (α3 − 3 α4 ) ϕB ) + 6 α6 RB ) + 2 α5 RB
dR1 (65)
′′
+16 α12 RB + RB 4 α22 + 11 α32 − 2 α4 α3 − 13 α42 R14 + 24 α2 (α3 − α4 ) R12 = 0,

and

4 R1 2 α1 (3 α3 RB R1′ + 3 α4 RB R1′ + 2 α1 φ′′B + 2 α5 φB ) + α22 φB

+16 α12 (φB R1′′ + 2 R1′ φ′B ) + 3 α32 − 2 α4 α3 − 5 α42 + 16 α1 α7 R15 φB

(66)
′
+8 (α2 (α3 − α4 ) + 2 α1 α6 ) R13 φB + 8 α1 (α3 + α4 ) R12 RB = 0.

Now, in order to simplify expressions, note that to make all these expansions possible, we need the coefficient
α1 to be different from zero, which lets us find the coefficients α5 , α6 and α7 from (27):

α22
α5 − α1 β1 − ,
4 α1
α2 (α4 − α3 )
α6 = − 2 α1 β3 ,
2 α1
(α3 + α4 ) (3 α3 − 5 α4 )
α7 = −3 α1 β5 −
16 α1
With this, (66) becomes

(α3 + α4 )
′
R12 φ′′B + 2 R1 R1′ φ′B = − R13 RB ,
2 α1

37
which implies
α3 + α4 L3
φ′B = − R1 RB + 4 α1 2 . (67)
2 α1 R1

Therefore, when replacing (67) into (65), we obtain

′′ 1 dδ
− β1 + 6 β3 R12 + 15 β5 R14 RB = 2 (α3 − 3 α4 ) L3 R1 −


RB (R1 ) = 0, (68)
α1 dR1

which is a non-homogeneous second-order equation just as (58), but with an extra non-homogeneous term.
The solutions will then have the same form plus an extra term. In particular, the homogeneous solution to
(68) is given by
Z X !
ds
RB,h (X) = R1′ L1 + L2 ′ 2
,
X0 (R1 )

whilst the particular one is given by

!
X X
R12
Z Z
1 δ (R1 )
RB,p (X) = R1′ (α3 − 3 α4 ) L3 ′ 2 ds − α 2 ds .
X0 (R1 ) 1 X0 (R1′ )

Therefore, by excluding L1 , we have

!
X X
L2 + (α3 − 3 α4 ) L3 R12
Z Z
1 δ (R1 )
RB (X) = R1′ 2 ds − 2 ds ,
X0 (R1′ ) α1 X0 (R1′ )

which implies
!
X ω ω
L2 + (α3 − 3 α4 ) L3 R12
Z Z Z
α3 + α4 1 δ (R1 )
φB (X) = − R1 R1′ 2 ds − 2 ds dω
2 α1 X0 X0 (R1′ ) α1 X0 (R1′ )
Z X
1
+ 4 α1 L3 ds,
X0 R12

and

B1 = (RB + i R1 φB ) ei φ1
!
X X
L2 + (α3 − 3 α4 ) L3 R12
Z Z
1 δ (R1 )
= L1 A′1 − i L4 Ā1 + R1′ 2 ds − 2 ds (69)
X0 (R1′ ) α1 X0 (R1′ )
Z X
1
+ i R1 4 α1 L3 ds
X0 R12
X ω
L2 + (α3 − 3 α4 ) L3 R12
Z Z
α3 + α4
− R1 R1′ 2 ds
2 α1 X0 X0 (R1′ )
Z ω ! !!
1 δ (R1 )
− ds dω ei φ1 (X) ,
α1 X0 (R1′ )2

Remark 5. Once again, RB and φB must, technically, be real functions (see Remark 3). Nevertheless, we
make the integration on the complex plane to continue our analysis at singularities.

38
6 Remainder with forcing due to truncation
In Section 5, we analyzed the equation for the remainder whilst ignoring the forcing due to truncation. That
helped us obtain a general expression for the amplitude of the residual, but the solution we obtained for it
misses an extra term. In particular, we note that, due to the linearity of equation (61), some constants of
integration need to be determined in order to know the actual expression of our remainder. To find such
constants, we focus our attention on a Stokes line, at which a key dominant term gets switched on, and will
let us focus on the key part of the remainder. With the forcing due to truncation, equation (61) becomes:
Mb
Ma X
∂ 2 RN ∂ 2 RN ∂ 2 RN
X  
0= ap bq Ju fp,q (S) RN + k 2 D̂ 2
+ 2 ε2 + ε4
p=0 q=1
∂x ∂x∂X ∂X 2
 2 [N −1] (70)
∂ 2 u[N −3] ∂ 2 u[N −1]

N +1 2 ∂ u N +3 2
+ε k D̂ 2 + + ε k D̂ .
∂x∂X ∂X 2 ∂X 2

Now, let us recall that our leading-order solution provided by X0 is given by (60). Therefore, the dominant
terms triggered by X0 , present in the forcing of (70), are given by

Γ N2 + γ


X
N +1 N ix̂
ε κ e N ei(r−1)x̂ Cr ,
+γ
(X0 − X) 2
r=0,2

where
1 + 2 i κ2 r
   
N 1 [1]
Cr = h[0]
r − ε2 +γ k 2 D̂ ϕ1 .
κ4 2 κ2 (X − X0 )

Furthermore, two consecutive terms have a ratio given, approximately, by

Γ N2+2 + γ


N +3 N +2
ε κ N +2 eirx !
+γ N
(X0 − X) 2 2 2 2 +γ
∼ε κ .
Γ N2 + γ


X0 − X
εN +1 κN e irx
N
+γ
(X0 − X) 2

We want consecutive terms to become the same order as N → ∞, so we need

N
2 +γ N |X0 − X|
ε2 |κ| 2
∼1 then +γ ∼ 2 + ν,
|X0 − X| 2 ε2 |κ|

where ν is a small finite number used to ensure that N is an even natural number.
We seek to find the Stokes line, which is a singularity that triggers the dominant part of the remainder.
With that in mind, we make the change X − X0 = ρ eiθ , where ρ > 0 and θ ∈ R, which implies that
N ρ iθ
+γ ∼ 2 + ν, and X = X0 + ρ e . Therefore, when using the phase shifts introduced in Equation
2 2
ε |κ|
(7), we have
2 2 2
−iχ̂
eix̂ = ei(X−X0 +X0 )/ε = ei(X−X0 )/ε eiX0 /ε e−iχ̂ .

Furthermore, by using Stirling’s approximation, we note that

Γ N2 + γ


i(X−X0 )/ε2 N +2 γ N +2 γ
e ε κ N
+γ
(X0 − X) 2
s ! N2 +γ
N
i(X−X0 )/ε2 N +2 γ N +2 γ 1 2π 2 +γ
∼e ε κ N N
(X0 − X) 2 +γ
2 +γ
e

39
  N2−1 +γ
√

i(X−X0 )/ε2 N +2 γ N +2 γ 1 N 1
= 2πe ε κ N +γ N
(X0 − X) 2 +γ 2 e 2 +γ

 N2−1 +γ iρ eiθ /ε2

√

1 N e
∼ 2 π εN +2 γ κN +2 γ N +γ N
(−ρ ei θ ) 2
+γ 2 e 2 +γ
ρ
! +ν− 21 iθ
√ ε2 |κ|2 /ε2
eiρ e
   
2 ρ
+ν 2 ρ
+ν 1 ρ
∼ 2πε ε2 |κ|2 κ ε2 |κ|2
ρ
+ν 2 +ν ρ
+ν
ε2 |κ|


ρ ei (θ+π) ε2 |κ|2 e ε2 |κ|2
2

ρ
+ν
 ! 2 ρ 2 +ν !− 12 iθ
√ 2 |κ| /ε2
κ ε2 |κ|2 1 |κ| ν ε
ρ eiρ e
= 2π   ρ
+ν
1 + ε2 2 +ν ρ
+ν
2 ρ
+ν ei(θ+π)


ε2 |κ|2 ρ ε2 |κ| e ε2 |κ|2
|κ| ε2 |κ|2
 
ρ
2 +ν iθ
/ε2
√ κ ε2 |κ|2 1 ν ε |κ| eiρ e
∼ 2π   ρ
+ν
e 1 ρ
ρ
ρ 2 e ε2 |κ|2 +ν
2 +ν


|κ| ε2 |κ|2 ei(θ+π) ε2 |κ|2

 
ρ ρ
2 +ν − ε2 |κ|
√ κ ε2 |κ|2 e 2
iθ
/ε2
= 2 π ε |κ| 
ρ

ρ
+ν
eiρ e .
1 2 +ν ei(θ+π) ε2 |κ|2
ρ |κ|
2 ε2 |κ|2

Here, observe that this quantity is exponentially small except in a neighborhood of the Stokes line, θ = − π2 .
Therefore, we let θ = − π2 + ε θ̂, which implies
 
ρ ρ
2 +ν − ε2 |κ|
N √


i(X−X0 )/ε 2
N +2 γ N +2 γ Γ +γ κ ε2 |κ|2 e 2
iε θ̂
/ε2
e ε κ 2
N ∼ 2 π ε |κ|   ρ
+ν
eρ e
+γ 2 ρ
+ν
 
(X0 − X) 2 1
ρ 2 |κ| ε2 |κ|2
ei( 2 +ε θ̂)
π ε2 |κ|2

 
ρ ρ
2 +ν − ε2 |κ|
√ κ ε2 |κ|2 e 2
 2 2

ρ 1+i ε θ̂− ε 2θ̂ /ε2
∼ 2 π ε |κ| 
ρ
   ρ
+ν
e .
1 2 +ν ε2 |κ|2
ρ 2 |κ| ε2 |κ|2
i ei ε θ̂

Now, if we take κ2 = κ2+ , then

ρ
− ε2 |κ|
N √


Γ +γ 1 e 2 2 2
 
i(X−X0 )/ε2 N +2 γ N +2 γ 2 ρ 1+i ε θ̂− ε 2θ̂ /ε2
e ε κ N ∼ 2 π ε |κ| 1 ρ
+ν
e
+γ ρ2
 
(X0 − X) 2 ei ε θ̂
ε2 |κ|2

√
2 π ε |κ| θ̂ 2
∼ 1 e−ρ 2

ρ 2

Furthermore, with the scales we have defined, we obtain

1 + 2 i κ2 r
   
N 1 [1]
Cr = h[0]
r − ε 2
+ γ k 2 D̂ ϕ1
κ4 2 κ2 (X − X0 )
! !
[0] 1 + 2 i κ2 r 2 ρ 1 [1]
∼ hr −ε 2 +ν k 2 D̂ ϕ1
κ4 ε2 |κ| κ2 ρ ei (− 2 +ε θ̂)
π

!
[0] 1 + 2 i κ2 r i [1]
∼ hr − 2 k 2 D̂ ϕ1
κ4 |κ| κ e2 i ε θ̂
!
[0] 2
hr 1 + 2iκ r i  
[1]
∼ 2 − 2 1 − i ε θ̂ k 2 D̂ ϕ1
κ κ2 |κ|
 
[1]
∼ h[0]
r 2 r − 2 + i ε θ̂ k 2 D̂ ϕ1 ,

which implies
   
[0] [1] [0] [1]
C0 ∼ h0 −2 + i ε θ̂ k 2 D̂ ϕ1 , and C2 ∼ h2 2 + i ε θ̂ k 2 D̂ ϕ1 .

40
Therefore, the second line of (70) is, asymptotically, given by

Γ N2 + γ


X
N +1 N ix
ε κ e N ei(r−1)x̂ Cr
2 +γ
(X0 − X) r=0,2
N

!
1−2 γ −2 γ iX0 /ε2 −iχ̂ i(X−X0 )/ε2 N +2 γ N +2 γ Γ +γ X
=ε κ e e e ε κ 2
N ei(r−1)x Cr
+γ
(X0 − X) 2
r=0,2
√ !
2 2 π ε |κ| θ̂ 2
X
∼ ε1−2 γ κ−2 γ eiX0 /ε e−iχ̂ 1 e−ρ 2 ei(r−1)x̂ Cr
ρ2 r=0,2
√
2 2 π |κ| k 2 θ̂ 2
 
[0]
 
[0]

[1]
∼ ε2−2 γ κ−2 γ eiX0 /ε e−iχ̂ 1 e−ρ 2 −2 + i θ̂ ε h0 e−ix̂ + 2 + i θ̂ ε h2 eix̂ D̂ ϕ1 .
ρ 2

Moreover,
∂ ∂θ ∂X ∂ ∂ ∂
= = i ρ ei θ ε = ρ ei ε θ̂ ε ,
∂ θ̂ ∂ θ̂ ∂θ ∂X ∂X ∂X
which yields

X0 − X = −ρ eiθ = i ρ ei ε θ̂ ,
i ε θ̂
/ε2 2
eix̂ = eρ e eiX0 /ε e−iχ̂
∂ e−iε θ̂ ∂
= .
∂X ρ ε ∂ θ̂
Thus, (70) becomes
Mb
Ma X
∂ 2 RN,2 ε ∂ 2 RN,2 −i ε θ̂ ε2 ∂ 2 RN,2 −2 i ε θ̂
X  
p q 2
0= a b Ju fp,q (S) RN,2 + k D̂ +2 e + 2 e
p=0 q=0
∂x2 ρ ∂x∂ θ̂ ρ ∂ θ̂2
√
2−2 γ −2 γ iX0 /ε2 −iχ̂ 2 π |κ| k 2 −ρ θ̂2  
[0] −ix
 
[0] ix

[1]
+ε κ e e 1 e 2 −2 + i θ̂ ε h 0 e + 2 + i θ̂ ε h2 e D̂ ϕ1 ,
ρ2
which can be expanded as
Mb
Ma X
∂ 2 RN,2  ∂2R ε2   ∂2R 

X ε N,2 N,2
0= ap bq Ju fp,q (S) RN,2 + k 2 D̂ + 2 1 − i ε θ̂ + 1 − 2 i ε θ̂
p=0 q=0
∂x2 ρ ∂x∂ θ̂ ρ2 ∂ θ̂2
√
2 2 π |κ| k 2 −ρ θ̂2  
[0]
 
[0]

[1]
+ε2−2 γ κ−2 γ eiX0 /ε e−iχ̂ 1 e 2 −2 + i θ̂ ε h0 e−ix + 2 + i θ̂ ε h2 eix D̂ ϕ1 , (71)
ρ2
Now, we proceed to solve (71). To do so, we use the following expansion:
2 X [j]
RN,2 = ε−2 γ eiX0 /ε εj RN,2 .
j≥1

Therefore, when solving this equation order by order, we have

Order O(ε): At this order, (71) becomes

∂2
 
[1]
Ju f0,0 (0) + k 2 D̂ 2 RN,2 = 0,
∂x
which has the following solution:
     
[1] [1]
RN,2 = C1 θ̂ eix + C−1 θ̂ e−ix ϕ1 ,

and we will omit the dependence on θ̂ from now on when there is no confusion.

41


Order O ε2 : At this order, (71) becomes

∂2
   
[2] [1] [1] [1]
Ju f0,0 (0) + k 2 D̂ RN,2 = −Ju F1,0,0 u[1] RN,2 − a1 Ju f1,0 (0) RN,2 − b1 Ju f0,1 (0) RN,2
∂x2
2 [1] √
2 2 ∂ RN,2 2 π |κ| −ρ θ̂2  [0] −ix  2 [1]

− k D̂ + κ−2 γ e−iχ̂ 1 e 2 h0 e 2 k D̂ ϕ1
ρ ∂x∂ θ̂ ρ2
√
2 π |κ| −ρ θ̂2 
[0]

[1]

− κ−2 γ e−iχ̂ 1 e 2 h2 eix 2 k 2 D̂ ϕ1 .
ρ2
Therefore,

[2] [2] [2] 1
[3]

RN,2 = 2 A1 C−1 + Ā1 C1 W0 + C1 eix + C−1 e−ix W1 − i C1′ eix − i C−1
′
e−ix W1,3



ρ
√ √
2 π |κ| −ρ θ̂ [0] −ix [3]
2 2 π |κ| −ρ θ̂2 [0] ix [3]
−κ−2 γ e−iχ̂ 1 e 2 h
0 e W 1,3 + κ−2 γ −iχ̂
e 1 e 2 h2 e W1,3
ρ2 ρ2
[2]
+2 A1 C1 e2ix + Ā1 C−1 e−2ix W2 .

Order O ε3 : At this order, (71) becomes

∂2
     
2 [3] [1] [1]
Ju f0,0 (0) + k D̂ 2 RN,2 = −Ju F1,0,0 u[2] RN,2 − a1 Ju F1,1,0 u[1] RN,2
∂x
   
[1] [1]
− b1 Ju F1,0,1 u[1] RN,2 − Ju F2,0,0 u[1] , u[1] RN,2
 
[2] [2] [2]
− Ju F1,0,0 u[1] RN,2 − a1 Ju f1,0 (0) RN,2 − b1 Ju f0,1 (0) RN,2
[1] [1] [1]
− a2 Ju f1,0 (0) RN,2 − b2 Ju f0,1 (0) RN,2 − a21 Ju f2,0 (0) RN,2
2 [2]
[1] [1] 2 2 ∂ RN,2
− a1 b1 Ju f1,1 (0) RN,2 − b21 Ju f0,2 (0) RN,2 − k D̂
ρ ∂x∂ θ̂
2 [1] [1]
θ̂ 2 ∂ RN,2 1 ∂ 2 RN,2
+ 2i k D̂ − 2 k 2 D̂
ρ ∂x∂ θ̂ ρ ∂ θ̂2
√ 2
2 π |κ| k −ρ θ̂ 2 
[0] [0]

[1]
− i κ−2 γ e−iχ̂ 1 e 2 θ̂ h0 e−ix + h2 eix D̂ ϕ1 .
ρ2
Now, as usual, we need to determine solvability conditions to ensure that this equation has a solution. In
particular, we obtain two solvability conditions given by

′′
√ θ̂ 2 [0]

D E 2 k 2 C−1 − i κ−2 γ e−iχ̂ 2 π |κ| ρ3/2 e−ρ 2 θ̂ h0
[3]
D̂ W1,3 , ψ = 0, (72)
ρ2
 √ θ̂ 2

D E 2 k 2 C1′′ − i κ−2 γ e−iχ̂ 2 π |κ| ρ3/2 e−ρ 2 θ̂ h[0]
2
[3]
D̂ W1,3 , ψ = 0, (73)
ρ2

where ′ denotes the derivative with respect to θ, and we have assumed that a1 + b1 = 0 to ensure the
convergence of the expressions we develop below noting that, generically, a1 = b1 = 0 (see, Equation (21)).
This yields

′′
√ θ̂ 2 [0]
C−1 = i κ−2 γ e−iχ̂ 2 π |κ| ρ3/2 e−ρ 2 θ̂ h0 ,
√ θ̂ 2 [0]
C1′′ = i κ−2 γ e−iχ̂ 2 π |κ| ρ3/2 e−ρ 2 θ̂ h2 .

42
Now, when integrating these expressions, we note that
√ 1 [0]
Z θ̂
ν2 ρ
C−1 = −i κ−2 γ e−iχ̂ 2 π |κ| ρ 2 h0 e− 2 dν,
−∞
√ 1 [0]
Z θ̂
ν2 ρ
C1 = −i κ−2 γ e−iχ̂ 2 π |κ| ρ 2 h2 e− 2 dν,
−∞

which implies
Z √ ρ2 θ̂
−2 γ −iχ̂
√ [0] 2
C−1 = −2 i κ e π |κ| h0 e−ν dν,
−∞
Z √ ρ2 θ̂
√ [0] 2
C1 = −2 i κ−2 γ e −iχ̂
π |κ| h2 e−ν dν.
−∞

where we have set the condition C−1 , C1 → 0 as θ̂ → −∞. Thus, as we cross the Stokes line as θ̂ goes from
−∞ to ∞,
2
 
[0] [0] [1]
−2 i π ε−2 γ κ−2 γ eiX0 /ε e−iχ̂ |κ| h2 eix + h0 e−ix ϕ1 (74)

gets triggered by κ2+ . Furthermore, by symmetry, we have that

2
 
[0] [0] [1]
2 i π ε−2 γ̄ κ−2 γ̄ e−iX̄0 /ε eiχ̂ |κ| h2 e−ix + h0 eix ϕ1

gets triggered by κ2− .

Now, let us recall that the outer solution we had obtained for the remainder in Section 5 is given by
[1] [1]
RN = B1 eix ϕ1 + c.c., where B1 is given by (69). Now, as in [5], we note that L1 (respectively, L4 )
corresponds to translating the solution u[1] in X (respectively, x). Therefore, these are not relevant to our
analysis. The constant we need to focus on is L2 , which controls the solution triggered at the Stokes line.
Therefore, for simplicity, we take L1 = L3 = L4 = 0, which yields
Z X Z X !
′ L2 1 δ (R1 )
B1 = R1 ′ 2
ds − ds
X0 (R1 ) α1 X0 (R1′ )2
Z X Z ω Z ω ! !
α3 + α4 L 2 1 δ (R1 )
−i R1 R1 R1′ ′ 2
ds − ds dω ei φ1 (X) .
2 α1 X0 X0 (R1 ) α1 X0 (R1′ )2

Furthermore, if we focus on the contribution of the term related to L2 in order to match solutions, we have
that Z X Z X Z ω ! !
′ ds α3 + α4 ′ ds
B1 = L2 R1 ′ )2
−i R1 R1 R1 ′ )2
dω ei φ1 (X) ,
X0 (R 2 α1 X0 X0 (R
1 1 (75)
L2 [0]
= h ,
K2 2
which has an asymptotic behavior given by,
√ s
2 β3 β1 √
B1 ∼ −L2 2 − (1 + 2 η i) e2 β1 X ei φ1 (X) , (76)
8 β1 β3
as X → ∞.
Moreover, when matching (75) and (74) with κ2 = κ2+ , we obtain
2
−2 γ −2 γ iX0 /ε
L2 = L+
2 = −2 i π ε κ e e−iχ̂ |κ| K2 . (77)
On the other hand, when considering κ2 = κ2− , we have
2
L2 = L−
2 = 2iπε
−2 γ̄ −2 γ̄ −iX̄0 /ε
κ e eiχ̂ |κ| K̄2 .

43
Remark 6. We highlight a key difference between our formulation in equation (70) and [5, Equation 120]
and [6, Equation 7.4]. Although the approach followed here follows the same ideas as those developed in
those articles, we have to bear in mind that the Swift-Hohenberg equation is a fourth-order partial differential
equation, which implies that the definition of X = ε2 x produces several terms out of ∂xxxx u. Therefore, the
authors of those papers simplified the equation of the remainder by using the explicit form of their model
whilst making an abuse of notation by using terms that are not part of the expansion to simplify expressions.
In our context, as a stationary reaction-diffusion equation is only a second-order partial differential equation,
then that is not necessary, and the forcing due to truncation gets reduced to the last three terms in (70).

6.1 Final prediction of the width of homoclinic snaking

Merging all the information we have developed so far, we summarize the expressions that need to be used
to determine the width of the homoclinic snaking close to codimension-two Turing bifurcation points.
In particular, from (76), we have that the homogeneous part of the amplitude of the remainder, consid-
ering the contribution from κ2+ and κ2− , is given by:
√ √ !s
2 β3 2 β β1 √
− 3
B1,h ∼ −L+
2 − L2 − (1 + 2 η i) e 2 β1 X i φ1 (X)
e ,
8 β12 8 β12 β3

as X → ∞. On the other hand, the term associated with the separation from the Maxwell point introduced
in equation (64) will produce a particular solution with the form
s
β1 √
B1,p ∼ R(ε) δb − (1 + 2 η i) e2 β1 X ei φ1 (X) ,
β3

as X → ∞, where R(ε) is a rational function in ε that needs to be determined for each system one wants to
study (see the examples in Section 8). Thus, by considering these two contributions, we have that B1 has
an asymptotic behaviour given by
√ !s
2 β3 β1 √
+ 2 β1 X i φ1 (X)


B1 ∼ − Re L2 + p(δb, ε) − (1 + 2 η i) e e
4 β12 β3
   
√ −π
2
√ 1 2 +η
2 β3 π |K2 | e k β1 ε

= − cos (K2o − χ̂ + 2 η log(ε)) + R(ε) δb (78)

 
2 β12 ε6
s
β1 √
× − (1 + 2 η i) e2 β1 X ei φ1 (X) ,
β3
o
as X → ∞, where K2 = |K2 | ei K2 . Therefore, in order to have a valid√ expansion, we need to ensure that
this expression tends to zero as X → ∞, so we need the coefficient of e2 β1 X to equal 0. Thus, as the cosine
function is bounded between - 1 and 1, we have that there exists a value of χ̂ so that said coefficient equals
zero if and only if
 
√ −π
2
√1
β1 ε2
+η
2 β3 π |K2 | e k

|δb| ≤ , (79)
2 β12 R(ε) ε6

where the coefficients β1 and β3 are defined in (27) and K2 , defined in (59), requires the determination of
[0]
λ1 , which is a generic expression for the terms cr defined by the limit (45). We highlight that (79) provides
an exponentially decaying bound for the width of the snaking as ε → 0+ .

44
7 Joining fronts
Let us recall that we have been analyzing two solutions at the same time (see sub-Section 2.3). They
correspond to fronts that join two steady states in both directions, up and down (see (31) and (32)). Now,
we need to join those solutions to construct a homoclinic orbit, which turns out to be a localized solution of
system (3) (see Figure 2), and obtain an approximation of the homoclinic snaking bifurcation curve.
Note that the solution we obtained for
the remainder in Sections 5 and 6 is exponentially growing on X
and becomes order 1 when X ∼ O 1/ε2 . Therefore, if we let 2 L be the width of the support of this solution
on x in (3), then the localized solution can be constructed by joining an up-front in the range 0 ≤ X ≤ L/ε2
with a down-front in the range L/ε2 ≤ X ≤ 2 L/ε2 . With this, we now proceed to perform the matching.
The up-front solution is given by
N
X
u= εr u[r] + RN , (80)
r=1

where
[1]
u[1] = A1 ei(x−χ̂) ϕ1 + c.c.,
[1]
u[3] = A3 ei(x−χ̂) ϕ1 + c.c.,

and a first-order approximation of the remainder has also been determined. Still, it depends on the function
δ (R1 ), which is related to the separation of one parameter from the Maxwell point (see equation (64)), which
will be studied in the examples developed in Section 8.
Now, for 1 ≪ X ≪ 1/ε2 , we have
 √ 
A1 ∼ ∆1,1 + ∆1,2 e−2 β1 X eΦ i ,

where
s s
2 β1 2 β1 −1 + 2 η i
∆1,1 = − , ∆1,2 = − ,
β3 β3 2
p  p 
Φ = 2 (η − ξ) β1 X + ζ − η log 2 β1 ,

and ζ = ζ(Ξ) is a linear function on Ξ = ε2 X obtained in Appendix B.

Furthermore, from Appendix B, we have that A3 has an asymptotic behaviour, as X → ∞, given by
 √ 
A3 ∼ ∆3,1 + ∆3,2 X e−2 β1 X eΦ i ,

where
2−5/2
∆3,1 = − √ 7/2
α2 β33 (α1,4 + 2 α1 ζΞ ) + α22 β33
α1 2 β1 (−β3 )
+2 α1 β33 α8,4 − 8 β13 β7,3 + 4 β12 β3 β5,3 − 2 β1 β32 β3,3 + 4 i α1 β1 β33 ω4



,
−3/2
2
∆3,2 = 7/2
(1 − 2 η i) α2 β33 (α2 + α1,4 + 2 α1 ζΞ )
α12 (−β3 )
+ α1 −β12 β32 (α6,4 + α7,4 ) + 2 β33 α8,4 − 4 β13 β7,3 .

With this, we note that the up-front solution is, asymptotically, given by

u ∼ ε u[1] + ε3 u[3]
 √   √ 
[1]
=ε ∆1,1 + ∆1,2 e−2 β1 X + ε2 ∆3,1 + ∆3,2 X e−2 β1 X eΦ i ei(x−χ̂) ϕ1 + c.c., (81)

45
and, by the symmetric properties of the amplitude equation, (6) (see sub-Section 2.3), we have that the down-
front is given by (81) when we change X → L/ε2 − X, and x → L/ε4 − X. However, the corresponding two
solutions may have different phase shifts. Therefore, for 1 ≪ L/ε2 −X ≪ 1/ε2 , an asymptotic approximation
of the down-front is given by
 √ 2
  √ 2

∆1,1 + ∆1,2 e−2 β1 (L/ε −X ) + ε2 ∆3,1 + ∆3,2 L/ε2 − X e−2 β1 (L/ε −X )


u∼ε
 p −η i √ 2 4
e2 i (η−ξ) β1 (L/ε −X ) ei ζ ei(L/ε −x−χ̌) ϕ1 + c.c.,
[1]
× 2 β1 (82)

where χ̌ ∈ R is the phase shift for the down-front.

Now, we highlight that,
for our expansion to be valid, we need to ensure that these match for X = O ε2
and L/ε2 − X = O 1/ε2 . Nevertheless, the expressions of these fronts will not be uniformly valid in these
√
limits if ∆3,2 due to the term X e−2 β1 X present in A3 . Therefore, in the general case, we must construct
an extra outer expansion valid inside the extended periodic domain (see the periodic pattern when the fronts
merge in Figure 2) to match it with these two fronts.

7.1 Outer expansion inside the extended periodic domain

In order to obtain a suitable outer approxiation, we take the usual expansion, (80), only focusing on the
limit as X → ∞ (see the periodic pattern when the fronts merge in Figure 2). In particular, we note that

A1 = ∆1,1 eΦ i

solves the amplitude equation at fifth order, (6). With this, we can solve the equation for the remainder
at order five, (64), by setting B1 (X, Ξ) = (Λ1 (X, Ξ) + i Ω1 (X, Ξ)) ei Φ(X,Ξ) , which transforms said equation
into

α1 Λ1XX − γ1,2,− Ω1X + γ1,1 Λ1 = 0, (83)

α1 Ω1XX + γ1,2,+ Λ1X = 0, (84)

where
4 (α4 (α3 + α4 ) + 4 α1 α7 ) β12
γ1,1 = ,
3 α1 β32
γ1,2,± = 2 α1 ΦX + (α3 ± α4 ) ∆21 + α2 .

Now, if we integrate (84) once, we obtain

γ1,2,+
Ω1X = − Λ1 ,
α1
which transforms (83) into

α1 Λ1XX − 4 α1 β1 Λ1 = 0,

Therefore,
√ √
Λ1 = Ψ(Ξ) e2 β1 X
+ Π(Ξ) e−2 β1 X
,

which implies
 √ √ 
Ω1 = 2 η Ψ(Ξ) e2 β1 X − Π(Ξ) e−2 β1 X .

46
With this, we can obtain the leading-order solution of the amplitude equation at order seven, (106) (see
Appendix B):
2 2 4
α1 A3XX + i α2 A3X + i α3 |A1 | A3X + i α4 A21 Ā3X + α5 A3 + 2 α6 |A1 | A3 + 3 α7 |A1 | A3
2
+ i α3 Ā1 A1X A3 + 2 i α4 A1 Ā1X A3 + α6 A21 Ā3 + 2 α7 |A1 | A21 Ā3 + i α3 A1 A1X Ā3
2 4 2
+ i α1,4 A1X + i α2,4 |A1 | A1X + i α3,4 |A1 | A1X + i α4,4 A21 Ā1X + i α5,4 |A1 | A21 Ā1X
2 2 2 4 6
+ α6,4 Ā1 (A1X ) + α7,4 A1 |A1X | + α8,4 A1 + α9,4 |A1 | A1 + α10,4 |A1 | A1 + α11,4 |A1 | A1
2
+ 2 α1 A1XΞ + i α2 A1Ξ + i α3 |A1 | A1Ξ + i α4 A21 Ā1Ξ = 0,

In particular, note that

A3 = ∆3,1 eΦ i

solves the seventh-order equation for A3 .

Now, we note that the full homogeneous equation for the remainder at order seven is given by Equation
(112) (see Appendix C). However, for simplicity, as stated in Appendix B, we can assume that A2 = B2 = 0
as that is, generically, the case (the right-hand side of the differential equations to determine A2 and B2 can
be forced to equal zero). This transforms the equation for the remainder at order seven into:
2
α1 B3XX + i α2 B3X + i α3 |A1 | B3X + i α3 Ā1 A1X B3 + i α3 A1 A1X B̄3 + i α4 A21 B̄3X
2 2 4
+ 2 i α4 A1 Ā1X B3 + α5 B3 + α6 A21 B̄3 + 2 α6 |A1 | B3 + 2 α7 |A1 | A21 B̄3 + 3 α7 |A1 | B3
2
+ i α1,3 B1XXX + α2,3 B1XX + α3,3 |A1 | B1XX + α3,3 Ā1 A1XX B1 + α3,3 A1 A1XX B̄1
2
+ α4,3 A21 B̄1XX + 2 α4,3 A1 Ā1XX B1 + i α5,3 B1X + i α6,3 |A1 | B1X + i α6,3 Ā1 A1X B1
4 2 2
+ i α6,3 A1 A1X B̄1 + i α7,3 |A1 | B1X + 2 i α7,3 |A1 | Ā1 A1X B1 + 2 i α7,3 |A1 | A1 A1X B̄1
2
+ i α8,3 A21 B̄1X + 2 i α8,3 A1 Ā1X B1 + i α9,3 |A1 | A21 B̄1X + i α9,3 A31 Ā1X B̄1
2
+ 3 i α9,3 |A1 | A1 Ā1X B1 + α10,3 A21X B̄1 + 2 α10,3 Ā1 A1X B1X + α11,3 A1 Ā1X B1X
2 2
+ α11,3 A1 A1X B̄1X + α11,3 |A1X | B1 + α12,3 B1 + α13,3 A21 B̄1 + 2 α13,3 |A1 | B1
2 4 4 6
+ 2 α14,3 |A1 | A21 B̄1 + 3 α14,3 |A1 | B1 + 3 α15,3 |A1 | A21 B̄1 + 4 α15,3 |A1 | B1
2
+ 2 α1 B1XΞ + i α2 B1Ξ + i α3 |A1 | B1Ξ + i α3 Ā1 A1Ξ B1 + i α3 A1 A1Ξ B̄1 + i α4 A21 B̄1Ξ
+ 2 i α4 A1 Ā1Ξ B1 + i α3 Ā1 A3 B1X + i α3 A1 Ā3 B1X + i α3 Ā1 A3X B1 + i α3 A1 A3X B̄1
+ i α3 A1X Ā3 B1 + i α3 A1X A3 B̄1 + 2 i α4 A1 A3 B̄1X + 2 i α4 A1 Ā3X B1
+ 2 i α4 Ā1X A3 B1 + 2 α6 Ā1 A3 B1 + 2 α6 A1 Ā3 B1 + 2 α6 A1 A3 B̄1 + 2 α7 A31 Ā3 B̄1
2 2 2
+ 6 α7 |A1 | A1 Ā3 B1 + 6 α7 |A1 | Ā1 A3 B1 + 6 α7 |A1 | A1 A3 B̄1 = 0.

Now, if we set B3 = (Λ3 (X, Ξ) + i Ω3 (X, Ξ)) ei Φ , we obtain two equations given by
√ √
α1 Λ3XX + γ1,3 Ω3X − γ1,4 Λ3 = (γ1,5 Ψ + γ1,6 ΨΞ ) e2 β1 X
+ (γ1,7 Π − γ1,6 ΠΞ ) e−2 β1 X
, (85)
√ √
2 β1 X −2 β1 X
α1 Ω3XX + γ1,8 Λ3X = (γ1,9 Ψ + γ1,8 ΨΞ ) e + (γ1,10 Π + γ1,8 ΠΞ ) e , (86)

where
(α3 − 3 α4 ) β1
γ1,3 = ,
β3
γ1,3 γ1,8
γ1,4 = 4 α1 β1 − ,
α1
1
γ1,5 = 2 α1 −2 α3 β1 α1 β32 28 β12 (α3,4 + α5,4 ) − 15 β3 β1 (α2,4 + α4,4 ) + 9 β32 α1,4
24 α13 β35
 p 
+6 α1 β32 3 ζΞ − 2 β1 ω4 + α4 4 β12 β3 (7 β3 α6,4 + 15 β5,3 ) + 15 β33 α8,4 − 120 β13 β7,3

47
 
3/2
−30 β1 β32 β3,3 + 15 α42 β1 β32 ω4 − 3 2 α1 α4 β1 β32 −28 β12 (α3,4 + α5,4 ) + 15 β1 β3 (α2,4 + α4,4 )


 p 
−9 β32 α1,4 + 6 α1 β32 2 β1 ω4 − 3 ζΞ + α42 β1 4 β12 β3 (7 β3 α6,4 + 15 β5,3 ) + 15 β33 α8,4
−120 β13 β7,3 − 30 β1 β32 β3,3 + 16 α12 β32 −2 β33 α8,4 + 4 β13 β7,3 − 4 β12 β3 β5,3 + 3 β1 β32 β3,3



     
5/2 3/2
+6 α43 β1 β32 ω4 + α32 β1 β3 β3 28 β12 α6,4 + 15 β3 α8,4 − 30 β1 β3,3 − 6 α4 β1 ω4 + 60 β12 β5,3

5/2
−120 β13 β7,3 + 6 α33 β1 β32 ω4 + 3 α2 β33 (10 (α3 − 3 α4 ) α1 β1 ((α3 + α4 ) ζΞ − 2 β1 α6,4 )

+8 α12 β3 (4 β3 α1,4 + 3 (α4 − α3 ) β1 ) + 5 (α3 + α4 ) (α3 − 3 α4 ) β1 α1,4 + 72 α13 β32 ζΞ

+3 α22 β33 36 α12 β32 + 5 (α3 + α4 ) (α3 − 3 α4 ) β1 ,

√


β1 α32 − 2 α3 α4 − 3 α42 β1 + 8 α12 β32
γ1,6 =− ,
2 α1 β32
 
3/2 2
(α3 − 3 α4 ) β1 (α3 + α4 ) β1 + 4 α12 β32
γ1,7 = γ1,5 − ω4 ,
α12 β33
(α3 + α4 ) β1
γ1,8 =− ,
β3
1
2 α1 12 α12 (α3 + α4 ) β32 4 β1 β3 β3 2 β12 α6,4 + β3 α8,4 − 2 β1 β3,3 + 4 β12 β5,3



γ1,9 = 4
√ 6
48 α1 β1 β3
 p   p 
−8 β13 β7,3 + α3 β12 β32 2 β1 ω4 − ζΞ + α4 β12 β32 2 β1 ω4 + 3 ζΞ

− 48 α13 β1 β34 4 β12 (α3,4 + α5,4 ) − 2 β1 β3 (α2,4 + α4,4 ) + β32 α1,4

− 2 α1 (α3 + α4 ) (α3 − 3 α4 ) β12 β32 4 β12 (α3,4 + α5,4 ) − 3 β1 β3 (α2,4 + α4,4 ) + 3 β32 α1,4

2
+ (α3 + α4 ) (α3 − 3 α4 ) β12 4 β12 β3 (β3 α6,4 + 3 β5,3 ) + 3 β33 α8,4 − 24 β13 β7,3 − 6 β1 β32 β3,3


p 
+96 α14 β1 β36 β1 ω4 − ζΞ + 3 α2 β1 β33 8 α13 β32 (5 (α3 + α4 ) ζΞ − 8 β1 α6,4 )
+ 2 (α3 + α4 ) (α3 − 3 α4 ) α1 β1 ((α3 + α4 ) ζΞ − 2 β1 α6,4 ) − 8 (α3 + α4 ) α12 β3 ((α3 − α4 ) β1

2
−2 β3 α1,4 ) + (α3 + α4 ) (α3 − 3 α4 ) β1 α1,4 + 3 α22 (α3 + α4 ) β1 β33 20 α12 β32
+ (α3 + α4 ) (α3 − 3 α4 ) β1 )) ,
 
2
2 β1 (α3 + α4 ) β1 + 4 α12 β32
γ1,10 = −γ1,9 + ω4 .
α1 β32

Next, note that we can integrate (86) with respect to X, to obtain

√ √
 
1 1 2 β1 X 1 −2 β1 X
Ω3X = √ (γ1,9 Ψ + γ1,8 ΨΞ ) e − √ (γ1,10 Π + γ1,8 ΠΞ ) e − γ1,8 Λ3 ,
α1 2 β1 2 β1
which can be replaced into (85), yielding
p √ p √
α1 Λ3XX − 4 α1 β1 Λ3 = −4 α1 β1 (γ1,11 Ψ + ΨΞ ) e2 β1 X − 4 α1 β1 (γ1,11 Π − ΠΞ ) e−2 β1 X , (87)

where
1
γ1,11 = √ 3 α22 −144 α14 β34 − 28 α12 (α3 − 3 α4 ) (α3 + α4 ) β1 β32
384 α16 β1 β37

2 2
+ (α3 + α4 ) (α3 − 3 α4 ) β12 β33 − 3 α2 288 α15 β34 ζΞ + 16 α14 β33 ((α3 − 15 α4 ) β1 + 8 β3 α1,4 )
+ 8 α13 (α3 − 3 α4 ) β1 β32 (12 β1 α6,4 + 7 (α3 + α4 ) ζΞ )
+ 8 α12 (α3 + α4 ) (α3 − 3 α4 ) β1 β3 ((α3 − α4 ) β1 + 4 β3 α1,4 )
2
− 2 α1 (α3 − 3 α4 ) (α3 + α4 ) β12 ((α3 + α4 ) ζΞ − 2 β1 α6,4 )

48
 
2 2
− (α3 + α4 ) (α3 − 3 α4 ) β12 α1,4 β33 + 2 α1 −48 α13 (α3 − 3 α4 ) β1 4 β12 (α3,4 + α5,4 )
−3 β1 β3 (α2,4 + α4,4 ) + 3 β32 α1,4 β34 + 96 α14 8 β13 β7,3 − 8 β12 β3 β5,3

+β32 (6 β3,3 + (9 α4 − 3 α3 ) ζΞ ) β1 − 4 β33 α8,4 β34

2
− 2 α1 (α3 + α4 ) (α3 − 3 α4 ) β12 β32 4 (α3,4 + α5,4 ) β12 − 3 β3 (α2,4 + α4,4 ) β1 + 3 β32 α1,4

− 4 α12 (α3 − 3 α4 ) (α3 + α4 ) β1 β32 60 β13 β7,3 − β12 β3 (β3 (31 α6,4 + 7 α7,4 ) + 16 β5,3 )
+β32 (3 (α3 − 3 α4 ) ζΞ − 6 β3,3 ) β1 + 24 β33 α8,4

2 2

− (α3 − 3 α4 ) (α3 + α4 ) β12 24 β13 β7,3 − 4 β3 (β3 α6,4 + 3 β5,3 ) β12 + 6 β1 β32 β3,3 − 3 β33 α8,4 .

Now, as the term on the right-hand side of (87) is secular, we need to set

Ψ = Ψ0 e−γ1,11 Ξ , and Π = Π0 eγ1,11 Ξ ,

in order to ensure the solution is bounded.

In summary, we can conclude that the far-field solution inside the localized pattern is given by
[1]
u∗ ∼ ε (∆1,1 + Λ1 + i Ω1 ) + ε3 ∆3,1 eΦ i ei(x+χ̃) ϕ1 + c.c.


 √ √ 
[1]
= ε ∆1,1 + (1 + 2 η i) Ψ0 e2 β1 X e−γ1,11 Ξ + (1 − 2 η i) Π0 e−2 β1 X eγ1,11 Ξ + ε2 ∆3,1 eΦ i ei(x−χ̃) ϕ1
+ c.c., (88)

where χ̃ is the phase shift for the fast oscillations.

The determination of the constants Ψ0 and Π0 can also be carried out by matching the fronts and the
far-field solutions inside the localized patterns. Specifically, they can be found by using [5]

lim u x, X, 0− ∼ lim+ u∗ (x, X, Ξ),

(89)
X→∞ Ξ→0

for matching (81) to (88), whilst

u x, X, L+ ∼ lim− u∗ (x, X, Ξ).

lim (90)
X−L/ε2 →−∞ Ξ→L

for matching (82) to (88).

However, these limits depend on the function δ = δ (R1 ), which is related to the separation of one
parameter from the Maxwell point (see equation (64)), and depends on the specific form of the vector field
posed in (1).

7.2 Summary of Section 7

The calculations we carried out up until Section 6 let us obtain an approximation of the width of the
homoclinic snaking close to codimension-two bifurcation points. Nevertheless, from the beginning, we stated
that we have been studying two solutions at the same time, an up-front and a down-front (see sub-Section
2.3). Section 7 gives the final conditions that need to be met to match these fronts. We have to bear in mind
that these conditions depend on δb, the separation of parameter b from the Maxwell point and, therefore
after determining bounds for that parameter by using (79), one can use conditions (89) and (90) to obtain
approximations of the homoclinic snaking as a function of ε.
We highlight that, in the end, although the limits (89) and (90) seem complicated to evaluate, what
one really needs to do to match the fronts is to equate the corresponding coefficients of the constant and
exponential terms in X in (81), (82), and (88). It is worth noting that, in the limit (90), one needs to match
the corresponding exponentials by making use of the complex conjugates of (81), (82), and (88), for the
equations to be well-posed (see e.g. [5]).

49
 Variable C4 E0 E2 E4 α1 α3 α5 α6 α7
√ √ √ √
114 20919 63711 114 16 734 8820
Value 1 4 4 -1 8 −
38 1083 10069012 19 19 361
Table 1: Values of some of the main parameters in the asymptotic expansion of system (91).

8 Examples
We now proceed to illustrate the theory by computing the necessary components of the beyond-all-orders
theory for several common examples. Readers are invited to run the code for each example themselves which
is provided at [17], using Python and Mathematica. In each case, we compare our analytical width of the
snake given by (79) with a numerical evaluation of fold points of homoclinic trajectories of the associated
spatial-dynamics problem on the real line, using AUTO [19].

8.1 Swift-Hohenberg 2-3

We start by applying our general theory to the Swift-Hohenberg equation, which was first deduced from the
equations for thermal convection. Said equation, with quadratic and cubic nonlinearities, was studied in [5],
and served as the main inspiration for this paper. We highlight that said equation has been studied from
different points of view and different nonlinearities as it is one the of the simplest pattern-forming equations
(see, e.g. [6, 20, 21, 22], and references therein). The Swift-Hohenberg 2-3 equation can be written as
2
∂t u = −C u − 3 E u2 − u3 − (1 + ∂xx ) u,

where u = u(x, t) is a real variable. Nevertheless, as explained in Section 1, this type of equation can be
written as (1) simply by defining an auxiliary variable v = ∂xx u as:

∂t u = −(1 + C) u − 3 E u2 − u3 − 2 v − ∂xx v,
(91)
0 = v − ∂xx u,

and we use a = C, and b = E.

To evidence the use of the code shared to study this type of problems, we state the parameter expansionsas
they were considered in [5]:
C = C4 ε4 ,
E = E0 + E2 ε2 + E4 ε4 + δE.
Now, we note that we can obtain the values of the variables shown in Table 1 simply by running the Python
code at [17]. In particular, we highlight that the values obtained are consistent with the results obtained in
[5].
To find the width of the snaking, we need to focus on the fifth-order equation for the remainder, (64). To
study said equation, we note that the only terms that will affect the remainder are given by the linearization
of (91) at (u, v) = (0, 0), and we add −3 δE u2 to the equation for u in order to take into account the
separation of E from the Maxwell point. In particular, we need such a
term to appear in the solvability
condition at order five. With this in mind, we note that F2,0,1 u[1] , u[1] will not have resonant terms at


said order. The first of those resonant terms will turn up due to 2 F2,0,1 u[1] , u[2] at order five, which lets
[4]
us know that the first term coming from the quadratic expression needs to be part of RN for the resonant
[4]
term to appear at order five. In particular, when using the same notation as before, we have that RN is
given by (62) plus

δE 2 [4] δE [4]
|A1 | W∗ + 2 A21 e2ix W∗,2 + c.c.,
ε2 ε

50
where
 
[4] [1] [1]
M0 W∗ = −F2,0,1 ϕ1 , ϕ1 ,
 
[4] [1] [1]
M2 W∗,2 = −F2,0,1 ϕ1 , ϕ1 ,

[4]

and the terms ε2 in the denominators are put to match the orders of RN and F2,0,1 u[1] , u[1] .
With this, we note that the equation for the amplitude of the remainder, (64), becomes
2 2
α1 B1′′ + i α3 |A1 | B1′ + i α3 A1 A′1 B̄1 + i α3 Ā1 A′1 B1 + α5 B1 + α6 A21 B̄1 + 2 α6 |A1 | B1
2 4
√ δE 2
+2 α7 |A1 | A21 B̄1 + 3 α7 |A1 | B1 + 2 114 2 |A1 | A1 = 0,
ε
dδ √ 2
which implies that (A1 ) = 2 114 δE |A1 | A1 . Therefore, (see (78))
dA1
s √ √ √ !
β1 2 β3 L+ 2 β3 L− 57 δE
B1 ∼ − − 2
− 2
+ 2 (1 + 2 η i) eX
β3 8 β12 8 β12 2 ε α1 β3
   
s √ −π2 η+ √ 1 2 √
1 β1  2 β3 π |K2 | e β1 ε
57 δE 
= − − 2 cos(K2o − χ̂ + 2 η log(ε)) + 2 X
 (1 + 2 η i) e ,
2 β3 β1 ε6 ε α1 β3

o
as X → ∞, where β1 , β3 are defined in (27) and K2 = |K2 | ei K2 was defined in (59).
Thus, in order to ensure that our solution remains bounded, we need to force the coefficient of this
expression to be zero, which implies that we need
 
√ −π
2 η+ √ 1 2
2 α1 β32 π |K2 | e β1 ε

|δE| ≤ √ , (92)
57 β12 ε4

for there to exist a value of χ̂ so that B1 decays to zero as X tends to infinity. Note that (92) is an
exponentially small term as ε → 0+ .
Now, note that all the variables in (92) are known except for K2 . To determine said variable, we need to
run the recurrence given by (39) with the initial conditions given by
√ √
19 4 98861726 i
A1 = ,
√ √ 734 √ √ √ √

19 4 734 −345333190 367 + 954973269 2 π + 2602107 367 π i + 236726744 2 i
A3 = ,
360647576448
which correspond to the leading order coefficients of A1 and A3 , respectively, at X0 (see ansatz (38)). This
[0]
is done to obtain an approximation of c2 , given by (45). In particular, when running n up to 181, we obtain
the graphs shown in Figure 3, where panel (a) (respectively, (b)) shows the convergence of the magnitude
[0]
(respectively, angle) of c2 . In particular, the dotted lines correspond to the approximation of each of these
quantities for different values of n/2 and the continuous line is a graph of the best curve of the form
c1 c2
f (n) = c0 + + 2 (93)
n n
that fits each set of points, ignoring the first few points. These curves were obtained based on the procedure
[0]
in [6], and said lines seem to approximate the curves well. In particular, this lets us know that c2 →
−2.57629 i
0.0303085 e as n → ∞, which implies that (see (59)):
−1/2
6 i (−2 β3 ) [0]
K2 = − c0 ≈ 0.0300695 + 0.0195978 i,
(3 − 2 η i)

51
 Abs[c2 [0] ]
0.035 Arg[c2 [0] ] n
-0.5
20 40 60 80 2

0.030 -1.0

-1.5
0.025

-2.0

0.020
-2.5

n
0.015 -3.0
0 20 40 60 80 2
(a) (b)
[0]
Figure 3: Result of the iteration of (39) to approximate c2 for system (91). The points correspond to the iteration
of the argument in the limit (45) for different values of n and continuous lines are the best fit for these points with
a function given by (93). (a) Behaviour of the absolute value of the argument of the limit in (45). (b) Behaviour of
the angle of the argument of the limit in (45) for r = 2.

and

|K2 | = 0.0358922, K2o = 0.577605.

With this, we can see that the width of the snaking in terms of ε is approximately given by
π
2.82934 e− ε2
|δE| ≤ , (94)
ε4
which is close to the function obtained by numerical fitting in [5, Equation (163)]. A numerical graph showing
the fit of this function compared to numerics is shown in Figure 4, which shows that our approximation of
the width of the homoclinic snaking is remarkably good.

8.2 Swift-Hohenberg 3-5

As a second example, we study the same type of equation we studied previously but with different nonlin-
earities. In particular, we now apply our theory to the Swift-Hohenberg 3-5 equation, which can be written
as
∂t u = r u − u − 2 v + s u3 − u5 − vxx ,
(95)
0 = v − ∂xx u,
and we use a = s and b = r. This example is motivated by the fact that it has the symmetry (u, v) → −(u, v),
which changes the dominant value of κ and the dominant modes to be taken into account to study the width
of the snaking [6]. In particular, we show that the theory we have developed can also be applied in these
cases (see Appendix A).
First, we note that a degenerate Turing bifurcation in (95) occurs at (r, s) = (0, 0) with a wavenumber
k = 1. Furthermore, the parameters obtained when making the expansions as developed throughout this
article are shown in Table 2.
As in the previous example, to obtain an approximation of the width of the snaking, it is important to
[0] [0]
know c0 and c2 . Nevertheless, due to the special symmetry (u, v) → −(u, v) (95) has, these terms equal

52
 0
0.2
log(2 | E|)

-10

||u||L2
0.15
-20

-30 0.1
0.2 0.4 0.6 0.8 1 0.4028 0.40284 0.40288 0.40292
E
(a) (b)

1
0.5
0
u
-0.5
-1
-1.5
-200 -100 0 100 200
x
(c)

Figure 4: (a) Numerical comparison between the analytical formula for the width of the homoclinic snaking, (94) (red
continuous line), compared to numerical results obtained with Auto (blue dashed line) for system (91). (b) Graph of
one branch of the homoclinic snaking curve (blue curve), together with three vertical dashed lines, representing the
folds and the Maxwell point at ε ≈ 0.471138 (point marked in black in panel (a)). (c) Graph of the homoclinic orbit
to the homogeneous steady state P = 0 at the red point marked in panel (b), corresponding to E ≈ 0.40284133406.

Variable r4 s2 α1 α5 α6 α7 β1 β3 β5

27 27 27 3 5
Value − 1 4 − 3 −10 −
160 160 640 8 6
Table 2: Values of some of the main parameters in the asymptotic expansion of system (95).

53
 [0] [0]
zero. The dominant modes in this case are r = ±1, ±3, so we need to compute c1 and c3 (see sub-Section
[0]
3.2). The convergence of c1 when using our code with initial conditions
3 √ 3 √
10 4 4 3 i 10 4 4 3
A1 = , and A3 = −
10 16
[0] [0]
is shown in Figure 5. The convergence for c3 looks very similar to the one for c1 and is, therefore, omitted
(the result can be checked by using the code that is contained in the GitHub repository provided in [17]).
[0]
We highlight that the results obtained are consistent with [6]. In fact, we also conclude that c1 ≈ 0.101 i,
which is the same value as was found in [6], but with a different sign. The change in sign here is consistent
with the choice of κ2 = κ2+ /2 to make our analysis (see Appendix A for an explanation of the division by 2).
Next, to complete this example, we note that the calculation in this case can be carried out using the
same ansatzes at each stage, but with a few minor changes (see Appendix A). Now, note that the parameter
r only influences the linear term in (95). Therefore, when considering the equation for the remainder, we
only need to add a linear term to Equation (64):
2
α1 B1′′ + i α2 B1′ + i α3 |A1 | B1′ + i α3 A1 A′1 B̄1 + i α3 Ā1 A′1 B1 + i α4 A21 B̄1′ + 2 i α4 A1 Ā′1 B1
2 2 4 δr
+α5 B1 + α6 A21 B̄1 + 2 α6 |A1 | B1 + 2 α7 |A1 | A21 B̄1 + 3 α7 |A1 | B1 + 4 A1 = 0,
ε
dδ δr
which implies that (A1 ) = 4 A1 . Therefore, when taking into account the little changes one needs to
dA1 ε
make in order to study the width of the snaking for (95) (see Appendix A), we have
s   √
1 2 β1 + β3 − β3 δr
B1 ∼ − −L2 − L2 − e2 β1 X ei φ1
8 β1 β3 β1 β1 α1
−√ π
s !
1 2 β1 β3 16 π |K2 | e β1 ε2 o δr √
= − − 6
cos (K2 − 2 χ̂) − 4 e 2 β 1 X e i φ1 ,
8 β1 β3 β1 ε ε α1

as X → ∞. With this, similar to the previous example, we need the condition

π
−√
β 1 ε2
α1 β3 16 π |K2 | e
|δr| ≤ − ,
β1 ε2
to ensure that there is a condition to ensure that the remainder tends to 0 as X → ∞, where
3 √
4
|K2 | ≈ 0.0202 10 4 3.

Therefore, we conclude that

π
267.183114 e−4.868645 ε2
|δr| ≤ .
ε2
Finally, when making the numerical check with actual snaking curves, we obtain the graph shown in
Figure 6, where we can see that the matching is, once again, pretty good.

8.3 Modified Schnakenberg system

To finish the study of models that have already been studied in the past to conclude that the theory we have
developed is, indeed, a generalization of what was already known. In particular, we deal with a modified
version of the Schnakenberg system, which turns out to be a simple model for glycolysis that has been widely

54
Abs[c1 [0] ] Arg[c1 [0] ]
0.20 2.0

1.8
0.15

1.6
0.10
1.4

0.05
1.2

n+1 n+1
0.00 1.0
0 20 40 60 80 2 0 20 40 60 80 2
(a) (b)
[0]
Figure 5: Similar to Figure 3 but to approximate c1 for model (95).

0.1
log(2 | r|)

-10
||u||L2

0.09
-20

0.08
-30
0.6 0.7 0.8 0.9 1 -0.07107134 -0.0710713 -0.07107126 -0.07107122
r
(a) (b)

0.5

u 0

-0.5

-1
-150 -100 -50 0 50 100 150
x
(c)

Figure 6: Similar to Figure 4, but for model (95). The snaking in panel (b) (respectively, the solution in panel (c))
corresponds to ε ≈ 0.805579 (respectively, r ≈ −0.071071309973).

55
used to study pattern formation (see, e.g. [11, 10, 15]). The modified version of the Schnakenberg system
we are to study here is given by:
 2
1
∂t u = −u + u2 v + σ u− + ∂xx u,
v
 2 (96)
2 1
∂t v = λ − u v − σ u − + d ∂xx v,
v

and is motivated by the study carried out in [15]. Said system has a homogeneous steady state given by
 
1
P = (u, v) = λ, .
λ
√
Furthermore, if we fix d = 3 + 8, then this steady state goes through a codimension-two Turing bifurcation
at
√ !
21 √ √ √
q q
15 2 15 31
(σ, λ) = 2 9407 − 6651 2 − + 9407 − 6651 2 − , 1 .
22 22 11 11

Now, let us take a = λ and b = σ, and consider the translation (u, v) → (u, v) + P. With this, system (96)
becomes
!
2

u σ λ u v + u + λ2 v
∂t u = λ u v + u + λ v + + 1 + ∂xx u,
λ (λ v + 1)2

2
u2 (λ v + 1) + 2 λ u (λ v + 1) + λ3 v σ λ u v + u + λ2 v
∂t v = − − + d ∂xx v.
λ (λ v + 1)2
Now, we note that, when getting away from the Maxwell point with σ, said parameter influences infinitely
many degrees of nonlinearities. Nevertheless, we only need to consider the dominant nonlinearities which,
in this case, are the quadratic and cubic ones. In particular, following the same idea as in Example 8.1, we
have that the equation for the amplitude of the remainder, (64), becomes
2 2
α1 B1′′ + i α3 |A1 | B1′ + i α3 A1 A′1 B̄1 + i α3 Ā1 A′1 B1 + α5 B1 + α6 A21 B̄1 + 2 α6 |A1 | B1
√ δσ
q
2 4 2 2
+2 α7 |A1 | A21 B̄1 + 3 α7 |A1 | B1 − 9407 − 6651 2 2 |A1 | A1 = 0.
3 ε
Furthermore, in this case, after running the recurrence, (39), with initial conditions

A1 = 0.833888471879154 i,
A3 = −0.534219592319241 − 0.432892956121214 i,
[0]
we obtain that c2 ≈ 0.121435 e−0.292384 i , which implies that |K2 | ≈ 0.201897174319627 (the graph showing
[0]
the convergence of c2 is shown in Figure 7).
With this, we have that
√ √ p √ ! s
2 β3 L+ 2 β3 L− 9407 − 6651 2 (1 + 2 η i) β1 √
B1 ∼ − 2
2
− 2
2
− √ δσ − e2 β1 X
4 β1 4 β1 3 2 α1 β3 ε2 2 β3
  
√ −π2
√ 1 2 +η
2 π β3 |K2 | e β1 ε

∼ − cos (K2o − χ̂ + 2 η log(ε))


2
β1 ε6
p √ ! s
9407 − 6651 2 (1 + 2 η i) β1 √
− √ δσ − e2 β1 X ,
3 2 α1 β3 ε 2 2 β3

56
Abs[c2 [0] ]
Arg[c2 [0] ] n
0.0
0.14 5 10 15 2

0.12 -0.2

0.10
-0.4
0.08
-0.6
0.06

-0.8
0.04

0.02 -1.0
n
0.00
0 5 10 15 2 -1.2
(a) (b)
[0]
Figure 7: Similar to Figure 3 but to approximate c2 for model (96).

Variable Value

λ4 1
√ p √ √ √
q p
−2595800404 − 27511240 2 9407 − 6651 2 − 1525844 9407 − 6651 2 + 1869381209 2
σ2 − p √
242 1617 − 1139 2

σ4 3.24595161831014

√
α1 4−2 2
√ p √ p √
76294 2 1140 9407 − 6651 2 444 18814 − 13302 2 113852
α3 − − − +
363 121 121 363
p √ p √ √
626 2 18814 − 13302 2 10 9407 − 6651 2 421 2
α4 − + + +
33 33 33 33
√
α5 2−2 2

α6 2.83351727667069
p √ p √ √
213917678 622015 18814 − 13302 2 6617678 9407 − 6651 2 450818570 2
α7 − − + +
14641 14641 43923 43923

Table 3: Values of some of the main parameters in the asymptotic expansion of system (96).

57
 0
1.419
|)

-10 1.418

||u||L2
log(2 |

1.417
-20
1.416
-30
0.2 0.25 0.3 0.35 0.4 0.45 0.5 -1.56 -1.555 -1.55 -1.545

(a) (b)
2 2

1.5
1.5
u 1 v
0.5 1

0 0.5
-200 -100 0 100 200 -200 -100 0 100 200
x x
(c) (d)

Figure 8: Similar to Figure 4, but for model (96). The snaking in panel (b) (respectively, the solution in panels (c)
and (d)) corresponds to εapprox0.394849 (respectively, σ ≈ −1.5488164814).

as X → ∞. Therefore, as in the previous examples, the condition we need to ensure the remainder to
converge to 0 as X → ∞ is given by
 
−π
2
√ 1
+η
β1 ε2
6 π α1 β32 |K2 | e
|δσ| ≤ √ ,
ε4
p
β12 9407 − 6651 2

which is, approximately, given by

π
15.217545 e−0.594604 ε2
|∆σ| ≤ .
ε4
The graph showing the match, in this case, is shown in Figure 8, which shows, once again, that the theory
we have developed matches numerics pretty well. We highlight that the result shown in [15] does not appear
to be successful because of the choice of the parameters to do the matching. In fact, if one takes b = λ, then
all the powers of (δλ)p will appear in the final expression for B1 , for every integer p ≥ 1, and finding a√bound
in that case is much more intricate. In that case, it would have been better to scale t → λ t and x → λ x in
order to carry out this analysis. We also remark that the recurrence (39) takes longer to run for this model
in comparison to the previous two examples due to the presence of the variable v in the denominator of (96).
[0]
Fortunately, c2 has a tidy behaviour from the beginning (see Figure 7), so not many terms were necessary
to conclude.

8.4 Brusselator
In [11], the authors developed pattern formation and pattern localization for Schnakenberg-type models used
to model chemical reactions or root-hair development. In particular, they proved several results for each
model but highlighted that the Brusselator is simple enough so that codimension-two points can be found
explicitly, but complex enough to show localized structures (in fact, they proved that there is a way to scale

58
the model so that the coordinates of its codimension-two Turing bifurcation points do not change under a
change in the diffusion coefficient). Therefore, we focus on the Brusselator system, which can be written as:
∂t u = a − c u + u2 v + δ 2 ∂xx u, (97)
2
∂t v = (c − 1) u − u v + ∂xx v,
and we use b = c. This system has only one homogeneous steady state given by
 
c−1
P = a, .
a
which goes through a Turing bifurcation whenever c = a2 δ 2 + 2 a δ + 2.
1 √

The authors of [11] proved this model has codimension-two points at a± = 21 ± 313 . In partic-
√ 16 δ
21 + 313
ular, if we fix δ = , then the variables of the expansion for the codimension-2 point at a = a+ can
48
be obtained and are shown in Table 4. Now, note that when making the change
 
c−1
(u, v) → u + a, v + ,
a
we obtain the system given by
u2 u2
 
2 2
∂t u = a v − + 2auv + u v − 2u + c u + + δ 2 ∂xx u,
a a
u2 u2
 
2 2
∂t v = −a v + − 2auv − cu − u v + u − c u + + ∂xx v.
a a

Here, we note that a change in c away from the Maxwell point, which will be called ∆c to avoid confusion
with δ, will affect two terms, a linear and a quadratic term: u + u2 /a. Nevertheless, we only need to care
about the dominant term, which is the linear one, which implies that the equation for the amplitude of the
remainder becomes
2 2
α1 B1′′ + i α3 |A1 | B1′ + i α3 A1 A′1 B̄1 + i α3 Ā1 A′1 B1 + α5 B1 + α6 A21 B̄1 + 2 α6 |A1 | B1
2 4 1  √  ∆γ
+2 α7 |A1 | A21 B̄1 + 3 α7 |A1 | B1 + 37 − 313 A1 = 0.
66 ε4
Now, in order to determine the width of the snaking, let us recall that we need to run the recursion given
by (39) with the initial conditions given by
A1 ≈ 0.109085 i,
A3 ≈ 0.145930 − 0.009474 i.
Figure 9 shows the results of the convergence of the argument of the limit in (45) for r = 2, for different
values of n/2. In particular, when fitting a curve as in the Swift-Hohenberg equation, we note that
[0]
c2 ≈ 0.000302336 e−1.83337 i .

Furthermore, we have
s √ √ √ !
1 β1 2 β3 L+ 2 β3 L− 313 − 37 √
B1 ∼ − − 2
− 2
+ √ ∆c (1 + 2 η i) e2 β1 X
8 β1 β3 β1 β1 33 2 α1 ε 4
  
s √ −π
2
√ 1 2 +η
1 β1  4 2 β3 π |K2 | e β1 ε

= − − 6
cos (K2o − χ̂ + 2 η log(ε))
8 β1 β3 β1 ε
√ !
313 − 37 √
+ √ ∆c (1 + 2 η i) e2 β1 X
,
33 2 α1 ε4

59
 Variable Value

a2 1
√
545 + 29 313
c2
192
√
27181639265 1139313695 313
c4 +
160777328256 53592442752
√
519128111361361961282269542371 149360653887596807674222710397 313
c6 −
70461515885712580123818315840 352307579428562900619091579200
√
58 + 2 313
α1
33
√
58 + 2 313
α2 −
99
√
8549 + 49 313
α3
7128
√
316 + 20 313
α4 −
891
√
−25180885801 + 399883939 313
α5
165801619764
√
55 + 3 313
α6
162
√
265583 + 5555 313
α7 −
497664

Table 4: Values of some of the main parameters in the asymptotic expansion of system (97).

Abs[c2 [0] ] Arg[c2 [0] ] n

0.0004 -1.5
20 40 60 80 2

0.0003
-2.0

0.0002

-2.5
0.0001

n -3.0
0.0000
0 20 40 60 80 2
(a) (b)

Figure 9: Similar to Figure 3 but for model (97).

60
as X → ∞.
Thus, to ensure that the remainder tends to zero as X → ∞, we need
 
−π
2
√ 1
+η
β 1 ε2
264 α1 β3 π |K2 | e
|∆c| ≤ − √ ,
ε2


β1 313 − 37

where |K2 | ≈ 0.0101224029550248. Therefore,

π
7.473090 e−4.808385 ε2
|∆c| ≤ . (98)
ε2
The result of the matching of this function with respect to numerical computation of the snaking for model
(97) is shown by the black line in Figure 10a, which shows that although the prediction we have obtained
seems to follow the same behaviour as the actual width of the snaking, the coefficient in front of (98) needs
to be higher. In fact, in this case, we have that

h1
≈ 0.006504 ≪ 1,
k1

which implies that the mode r = 2 is much less dominant than r = 0 (see Equation (44)). Now, observe the
[0]
convergence of c0 , shown in Figure (11). In this case, we note that we have two limits to which the curve
seems to converge. This happens because the mode r = 0 is present for every dominant value of κ2 = κ2± .
That is, there is an interaction between these modes.
To summarize, if we take the average between the two limits of convergence, which turn out to be 0.038285
and 0.089638, and divide the result by 4, considering that there are four different values of κ (that is, four
different solutions, see Section 3), we obtain
[0]
c0 ≈ 0.0159904.

[0]
Now, if we use this value for computing K2 , instead of c2 , taking into account that the matching does not
[0] [0]
change when considering that h0 is defined as the complex-conjugate of h2 when the integrals that defined
them are set to be real (see Equation (59)), we obtain
π
395.248011 e−4.808385 ε2
|∆c| ≤ , (99)
ε2
which leads to the red line in Figure 10a, leading to a remarkable match, even for high values of ε.

8.5 4-component Brusselator

Finally, we will show how our code can be used to carry out the calculations for higher-component systems.
In particular, once again, motivated by the Brusselator, we now work with two Brusselator models coupled
linearly, which forms a 4-component reaction-diffusion equation studied in [23]. Said equation is given by

∂t u = a − (b + 1) u + u2 v + α (w − u) + δ 2 ∂xx u,
∂t v = b u − u2 v + β (z − v) + ∂xx v,
(100)
∂t w = a − (b + 1) w + w2 z + α (u − w) + δ 2 ∂xx w,
∂t z = b w − w2 z + β (v − z) + ∂xx z,

and it is a model that has a homogeneous steady state given by

 
b b
P = a, , a, .
a a

61
 0
5.784
-5
c|)

5.782
-10

||u||L2
5.78
log(2 |

-15 5.778
-20 5.776

-25 5.774
0.7 0.75 0.8 0.85 0.9 0.95 1 17.51892 17.51893 17.51894 17.51895 17.51896
c
(a) (b)
5 6
5.5
4 5
u v 4.5
3
4
3.5
2
-100 -50 0 50 100 -100 -50 0 50 100
x x
(c) (d)

Figure 10: Similar to Figure 4, but for model (97). Here, panel (a) shows the comparison between the width of the
snaking with respect to Equations (98) (black line) and (99) (red line). The snaking in panel (b) (respectively, the
solution shown in panels (c) and (d)) corresponds to ε ≈ 0.901788 (respectively, c ≈ 17.518934615).

Abs[c0 [0] ]
0.20 Arg[c0 [0] ] n
0.0
20 40 60 80 2

-0.5
0.15

-1.0

0.10
-1.5

-2.0
0.05

-2.5
n
0.00 -3.0
0 20 40 60 80 2
(a) (b)

Figure 11: Similar to Figure 3 but for model (97).

62
 Variable Value

a2 1
√
29 313 545
b2 + 192
192
√
27181639265 1139313695 313
b4 +
160777328256 53592442752
√
519128111361361961282269542371 149360653887596807674222710397 313
b6 −
70461515885712580123818315840 352307579428562900619091579200
√
4 313 116
α1 +
33 33
√
116 4 313
α2 − −
99 99
√
49 313 8549
α3 +
3564 3564
√
40 313 632
α4 − −
891 891
√
25180885801 399883939 313
α5 − +
82900809882 82900809882
√
313 55
α6 +
27 81
√
265583 5555 313
α7 − −
248832 248832

Table 5: Values of some of the main parameters in the asymptotic expansion of system (100).

Now, if we fix
√ !
1 21 + 313
(α, β, δ) = 1, , ,
2 48

then P goes through a codimension-two Turing bifurcation at

√
 
1 
(a, b) = 3, 841 + 37 313 .
128
The variables of the expansion at this codimension-two point are shown in Table 5. Now, if we make the
translation (u, v, w, z) → (u, v, w, z) + P, we obtain the system
u2
 
∂t u = b u + + v (2 a + u)2 − 2 u + w + δ 2 ∂xx u,
2a
   
1 u (2 a + u)
− v 2 (2 a + u)2 + 1 + z + ∂xx v,


∂t v = −b
2 a
w2
 
∂t w = b w + + z (2 a + w)2 − 2 w + u + δ 2 ∂xx w,
2a
   
1 w (2 a + w)
− z 2 (2 a + w)2 + 1 + v + ∂xx z.


∂t z = −b
2 a

63
Abs[c0 [0] ]
Arg[c0 [0] ] n
0.14 0.0
10 20 30 40 50 2
0.12
-0.5
0.10
-1.0
0.08
-1.5
0.06

0.04 -2.0

0.02 -2.5
n
0.00
0 10 20 30 40 50 2 -3.0
(a) (b)
[0]
Figure 12: Similar to Figure 3 but only for the magnitude of c0 for model (100).

Now, as in the previous examples, let us recall that we need to run the iteration (39) in order to be able to
estimate the width of the snaking in this case, with initial conditions given by

A1 ≈ 0.109084591223343 i,
A3 ≈ 0.145930074159533 − 0.00947358642130998 i.

The result of said iteration is shown in Figure 12. We highlight here two facts. First, as in the previous
[0]
example, the most dominant mode turns out to be r = 0, reason why we show the convergence of c0 instead
[0]
of c2 . On the other hand, we have to bear in mind that, as (100) is a 4-component system, numerical error
grows larger quickly due to catastrophic cancellation [24]. In fact, the branch at the top of Figure 12a does
not seem to be converging just yet, but that is expected to be caused by numerical error. In fact, as n grows,
the little numerical errors generated at lower orders get stronger, leading the iteration away from the good
behaviour they show at lower orders. In this case, we use only the value the iteration seems to converge to,
which is, approximately, 0.0380283. Therefore, following the same idea as before, if we simply divide this
value by 4, we obtain that
[0]
c0 ≈ 0.00950707,

which implies |K2 | ≈ 0.31830279378449.

To summarize the process, by following the same ideas as in the previous example, we have to solve the
following equation for the amplitude of the remainder at order five, (64) given by:
2 2
α1 B1′′ + i α3 |A1 | B1′ + i α3 A1 A′1 B̄1 + i α3 Ā1 A′1 B1 + α5 B1 + α6 A21 B̄1 + 2 α6 |A1 | B1
2 4 1  √  δb
+2 α7 |A1 | A21 B̄1 + 3 α7 |A1 | B1 + 37 − 313 A1 = 0.
33 ε4

64
With this, we have
s √ √ √ !
1 β1 2 β3 L+ 2 β3 L− 313 − 37 √
B1 ∼ − − 2
− 2
+ √ δb (1 + 2 η i) e2 β1 X
4 β1 β3 2 β1 2 β1 33 2 α1 ε 4
  
s √ −π
2
√ 1 2 +η
1 β1  2 2 β3 π |K2 | e β1 ε

= − − 6
cos (K2o − χ̂ + 2 η log(ε))
4 β1 β3 β1 ε
√ !
313 − 37 √
+ √ δb (1 + 2 η i) e2 β1 X ,
33 2 α1 ε 4

as X → ∞. Therefore, as in the previous examples, we conclude that

 
−π
2
√ 1
+η
β 1 ε2
132 α1 β3 π |K2 | e
|δb| ≤ √ , (101)
ε2


β1 313 − 37

which can be written, approximately, as

π
234.994153440791 e−4.808385 ε2
|δb| ≤ ,
ε2
The black line in Figure (13a) shows the matching in this case. Although it is not perfect due to the
numerical inaccuracies we have highlighted, we can see that the shape of the analytical and numerical curves
match well. In reality, the coefficient in front of the right-hand side of (101) should be a little higher (in
fact, the branch at the top in Figure 12 is not expected to go down in that way, so that the average becomes
a little higher, by following the same procedure as in the previous example). In fact, if we multiply that
coefficient by 1.7, we obtain the red line shown in Figure 13a which, once again, shows that the theory we
have developed produces a great match when the constants are computed fully accurately.

Remark 7. We close this section by highlighting that it is hard to work with small values of ε in the numerical
computations. We omit the details, but using suitable tolerances in AUTO, we had to take extremely long
domains in order to find reliable estimates for the width of the homoclinic snaking when it becomes smaller
than O(10−7 ). Given the exponentially small asymptotic estimates of the theory, in practice, this means that
the smallest values of ε for which we can demonstrate the agreement of the numerics with the theory is only
about 0.4. Nevertheless, we note that the homoclinic snaking occurs, generically, at a distance of order ε2
or ε4 from the codimension-two Turing bifurcation line, which allowed us to demonstrate good agreement
between our analysis and numerics by computing how the width of the snake scales with ε.

9 Discussion
In this paper, we have achieved what we set out to do. We have successfully generalized the theory of
exponential asymptotics used to study homoclinic snaking close to a codimension-two Turing bifurcation in
the Swift-Hohenberg equation [5, 6] to arbitrary systems of reaction-diffusion equations undergoing the same
instability. Accompanying codes to do all the calculations automatically in Python and Mathematica for
any reaction-diffusion systems have also been provided [17], and the codes were used to do the calculations
for each of the examples shown in this paper. Further study will include the use of these codes to study
homoclinic snaking in models where it has been found but not much has been said about the region in the
parameter space where such phenomenon can be found. Although this calculation is extensive, the purpose
of this article is to provide tools for everyone to be able to understand it easily and use it as needed.
We highlight that some details in this paper have not been treated, as it is already quite lengthy. For
example, we have not explained how to use information on the phase of the exponentially small estimate
in order to study the fine details of the homoclinic snaking. Furthermore, with additional symmetry or
conserved quantities, there can be additional branches. Moreover, it is well known that examples such as

65
 0
8.14
log(2 | b|)

-10

||u||L2
8.135
-20
8.13
-30
0.5 0.6 0.7 0.8 0.9 1 16.358985 16.35899 16.358995 16.359
b
(a) (b)
5 6
5.5
4 5
u v 4.5
3 4
3.5
2
-60 -40 -20 0 20 40 60 -60 -40 -20 0 20 40 60
x x
(c) (d)
5 6
5.5
4 5
w z 4.5
3 4
3.5
2
-60 -40 -20 0 20 40 60 -60 -40 -20 0 20 40 60
x x
(e) (f)

Figure 13: Similar to Figure 10 but for model (100). The snaking in panel (b) (respectively, the solution in panels
(c), (d), (e), and (f)) corresponds to ε ≈ 0.88773 (respectively, b ≈ 16.35899033).

66
the Swift-Hohenberg equation, which have variational structures give rise to ‘ladder’ asymmetric stationary
localised patterns connecting the two interleaving snaking branches.
Also, we have not discussed the temporal stability of patterns. Under certain conditions on the matrix M
and forms of f , much can be said at the general level, without needing to resort to model-specific calculations.
Details are left for future work.
A more challenging open question is to consider analogous structures that arise near Turing bifurcations
corresponding to the dispersion curve having a non-zero imaginary part. Similar codimension-two bifurcation
points occur for such finite wavenumber Hopf bifurcations (also known as wave bifurcations), but are more
complex due to the presence of both standing and travelling waves. Nevertheless, in earlier work [25], the
present author has computed the regular asymptotic expansion to produce the amplitude equations for wave
bifurcations up to order five. The same approach as that presented here is expected to be generalisable, but
a general calculation is expected to be yet more cumbersome.
Another generalisation under consideration is to study so-called slanted snaking, in the presence of a zero
wavenumber mode in addition to the Turing instability [26]. Particular classes of systems in higher spatial
dimensions would also be interesting to study, although complete generalisations there seem a long way off.
Furthermore, as the theory developed here is related to the WKB theory [27], which has been used to
study Turing bifurcations and localized patterns in heterogeneous reaction-diffusion equations (see e.g. [28]),
a generalization of this approach to study localized solutions in heterogeneous systems is also under consid-
eration.
One last thing to highlight is that the process carried out in this article is a formal derivation of the region
in the parameter space where homoclinic snaking can be found close to codimension-two Turing bifurcation
points. It would be nice to develop generalities about a more rigorous approach to deal with this kind of
problem as often such approaches follow from more formal asymptotic procedures, like the ones used here.

Acknowledgements
I want to thank Alan R. Champneys for his incredible support and insistence on this project. I also want to
acknowledge Andrew Dean, Jon Chapman, Gregory Kozyreff, Philippe Trinh and Hannes de Witt for their
insightful input in different stages of this project. Moreover, I would like to thank Bastián Jiménez-Sepúlveda
for his invaluable support in optimizing the codes developed to make the intensive computations efficient.
This project was funded by ANID, Beca Chile Doctorado en el extranjero, number 72210071.

References
[1] A.M. Turing. The chemical basis of morphogenesis. Philosophical Transactions of the Royal Society of
London B: Biological Sciences, 237:37–72, 1952.
[2] J.H.P. Dawes. After 1952: The later development of Alan Turing’s ideas on the mathematics of pattern
formation. Historia Mathematica, 43(1):49–64, 2016.

[3] A.L. Krause, E.A. Gaffney, P.K. Maini, and V. Klika. Modern perspectives on near-equilibrium analysis
of turing systems. Philosophical Transactions of the Royal Society A, 379(2213):20200268, 2021.
[4] J. Burke and E. Knobloch. Homoclinic snaking: structure and stability. Chaos: An Interdisciplinary
Journal of Nonlinear Science, 17(3), 2007.

[5] S.J. Chapman and G. Kozyreff. Exponential asymptotics of localised patterns and snaking bifurcation
diagrams. Physica D: Nonlinear Phenomena, 238(3):319–354, 2009.
[6] A. Dean, P.C. Matthews, S.M. Cox, and J.R. King. Exponential asymptotics of homoclinic snaking.
Nonlinearity, 24(12):3323, 2011.
[7] P.D. Woods and A.R. Champneys. Heteroclinic tangles and homoclinic snaking in the unfolding of a
degenerate reversible Hamiltonian–Hopf bifurcation. Physica D: Nonlinear Phenomena, 129:147–170,
1999.

67
 [8] A.R. Champneys. Editorial to Homoclinic snaking at 21: In memory of Patrick Woods. IMA J Appl
Math, 86:845–855, 2021.
[9] E. Knobloch. Spatial localization in dissipative systems. Annual Review of Condensed Matter Physics,
6:325–359, 2015.

[10] F. Al Saadi, A. Champneys, and N. Verschueren. Localized patterns and semi-strong interaction, a
unifying framework for reaction–diffusion systems. IMA Journal of Applied Mathematics, 86(5):1031–
1065, 2021.
[11] E. Villar-Sepúlveda and A. Champneys. Degenerate Turing bifurcation and the birth of localized
patterns in activator-inhibitor systems. SIAM Journal on Applied Dynamical Systems, 22(3):1673–1709,
2023.
[12] M. Beck, J. Knobloch, D.J.B. Lloyd, B. Sandstede, and T. Wagenknecht. Snakes, ladders, and isolas of
localized patterns. SIAM Journal on Mathematical Analysis, 41:936–972, 2009.
[13] J.J Bramburger, D.J. Hill, and D.J.B. Lloyd. Localized multi-dimensional patterns. arXiv preprint
arXiv:2404.14987, 2024.

[14] M. Haragus and G. Iooss. Local Bifurcations, Center Manifolds, and Normal Forms in Infinite-
Dimensional Dynamical Systems. Springer, 2011.
[15] H. de Witt. Beyond all order asymptotics for homoclinic snaking in a schnakenberg system. Nonlinearity,
32(7):2667, 2019.

[16] G. Kozyreff and S.J. Chapman. Asymptotics of large bound states of localized structures. Physical
Review Letters, 97:044502, 2006.
[17] E. Villar-Sepúlveda. Exponential asymptotics for homoclinic snaking. https://github.com/
edgardeitor/Exponential-asymptotics.

[18] Edgardo Villar-Sepúlveda and Alan Champneys. Computation of turing bifurcation normal form for n-
component reaction-diffusion systems. ACM Transactions on Mathematical Software, 49(4):1–24, 2023.
[19] E.J. Doedel and B. Oldeman et al. Auto-07p: Continuation and bifurcation software for ordinary
differential equations, 2020. Latest version at https://github.com/auto-07p.
[20] Lambertus A Peletier and Vivi Rottschäfer. Pattern selection of solutions of the swift–hohenberg
equation. Physica D: Nonlinear Phenomena, 194(1-2):95–126, 2004.
[21] J. Burke and E. Knobloch. Normal form for spatial dynamics in the Swift-Hohenberg equation. In
Conference Publications, volume 2007, page 170. American Institute of Mathematical Sciences, 2007.
[22] John Burke and Edgar Knobloch. Snakes and ladders: localized states in the Swift–Hohenberg equation.
Physics Letters A, 360(6):681–688, 2007.
[23] Jennifer K Castelino, Daniel J Ratliff, Alastair M Rucklidge, Priya Subramanian, and Chad M Topaz.
Spatiotemporal chaos and quasipatterns in coupled reaction–diffusion systems. Physica D: Nonlinear
Phenomena, 409:132475, 2020.
[24] Jean-Michel Muller, Nicolas Brisebarre, Florent De Dinechin, Claude-Pierre Jeannerod, Vincent Lefevre,
Guillaume Melquiond, Nathalie Revol, Damien Stehlé, Serge Torres, et al. Handbook of floating-point
arithmetic, volume 1. Springer, 2018.
[25] Edgardo Villar-Sepúlveda and Alan Champneys. Amplitude equations for wave bifurcations in reaction–
diffusion systems. Nonlinearity, 37(8):085012, 2024.

[26] E. Knobloch. Localized structures and front propagation in systems with a conservation law. IMA J
Appl Math, 81:457–487, 2016.

68
[27] HL Berk, William McCay Nevins, and KV Roberts. New stokes’ line in wkb theory. Journal of
Mathematical Physics, 23(6):988–1002, 1982.
[28] Andrew L Krause, Václav Klika, Edgardo Villar-Sepúveda, Alan R Champneys, and Eamonn A
Gaffney. Pattern localisation in swift-hohenberg via slowly varying spatial heterogeneity. arXiv preprint
arXiv:2409.13043, 2024.

69
 [0] [0]
A The case in which c0 = c2 = 0
[0] [0]
Generically, we have that the constants c0 or c2 let us estimate the width of the homoclinic snaking
successfully, close to codimension-two Turing bifurcation points (see Section 6). Nevertheless, there are
cases in which both of these constants equal zero, which does not let us give a proper estimation for this
width (see e.g. [6]). This happens, generically, when the system (1) has the symmetry u → −u, a case in
which the even terms of the expansion vanish. To overcome this issue, the same general ansatzes we used in
Section 3 are still valid, but some key changes need to be made. In particular,
1. The dominant values of κ2 become ±i/2, and the dominant modes for which these values are attained
are r = ±1, ±3 (see sub-Section 3.2).
2. The ansatz for the values of γr , (43) becomes
1 r−3
γ0 = γ±2 = γ − , γ±1 = γ, γ±r = γ − , for r ≥ 3.
2 2
3. After making the same expansion as in Equation (40) to obtain the inner solution with these new
values of γr (see Section 3 for the ideas that need to be followed to study this case), we conclude that
κ ∈ {−1 − 2 η i, −2 η i, 2 − 2 η i, 3 − 2 η i},
from where we take, once again, the one with the highest real part, γ = 3 − 2 η i.
4. As the dominant multiples of the critical wavenumber, k, are now odd, we need to assume that n is
odd. Furthermore, we need to consider the changes of the modes in the forcing due to truncation of
the equation for the remainder (see Section 6). In particular, the forcing due to truncation provided
by κ2 = κ2+ /2 is now given by
Γ N2 + γ


X
N +1 N 2ix
ε κ e N ei(r−2)x̂ Cr ,
+γ
(X0 − X) 2 r=1,3

where
 
[1]
Cr = 2 h[0]
r −2 (2 − r) + 2 i ε θ̂ k 2 D̂ ϕ1 ,

which implies
   
[0] [1] [0] [1]
C1 = h1 −4 + 4 i ε θ̂ k 2 D̂ ϕ1 , C3 = h3 4 + 4 i ε θ̂ k 2 D̂ ϕ1 ,
[0] [0] [0] [0]
and h1 , h3 are the analogs of h0 , h2 , respectively, for this case. We highlight that the equations
[0] [0] [0] [0]
that determine h1 , h3 coincide with the ones obtained for h0 , h2 in Section 4.
5. Following the same steps as in Section 6, we note that
2 2 2
−2iχ̂
e2ix̂ = e2i(X−X0 +X0 )/ε = e2i(X−X0 )/ε e2iX0 /ε e−2iχ̂ ,
which yields
N √


2i(X−X0 )/ε2 Γ +γ 2
e ε N +2 γ
κ N +2 γ 2
N ∼ 2 π ε |κ| e−ρ θ̂ .
+γ
(X0 − X) 2

Therefore,
[2] [2] [2] 1
[3]
RN,2 = 2 A1 C−1 + Ā1 C1 W0 + C1 eix + C−1 e−ix W1 − i C1′ eix − i C−1
′
e−ix W1,3



ρ
√ √
2 π |κ| −ρ θ̂2 [0] −ix [3] 2 π |κ| −ρ θ̂2 [0] ix [3]
−2 κ−2 γ e−iχ̂ 1 e h0 e W1,3 + 2 κ−2 γ −iχ̂
e 1 e h2 e W1,3
ρ 2 ρ2
[2]
+2 A1 C1 e2ix + Ā1 C−1 e−2ix W2 ,

70


which implies that the analog of the solvability conditions at order O ε3 , (72) and (73), yield

′′
√ [0] 2
C−1 = 2 2 π i θ̂ h0 ρ3/2 |κ| κ−2 γ e−θ̂ ρ−iχ̂
√ [0] 2
C1′′ = 2 2 π i θ̂ h2 ρ3/2 |κ| κ−2 γ e−θ̂ ρ−i χ̂ ,

which yield
[0]
C−1 = −i π κ−2 γ h0 e−iχ̂ ,
[0]
C1 = −i π κ−2 γ h0 e−iχ̂ .

Therefore, in this case, as we cross the Stokes line, the following function gets triggered by κ2 = κ2+ /2:
2
 
[0] [0] [1]
−i π ε−2 γ κ−2 γ e2iX0 /ε e−2iχ̂ h2 eix + h0 e−ix ϕ1 ,

which implies
2
−2 γ −2 γ 2iX0 /ε
L2 = L+
2 = −i π ε κ e e−2iχ̂ K2 ,

with the corresponding development also for κ2− /2.

In summary, the only change one needs to make in this case to apply the same theory in order to estimate
1
the width of the snaking is to scale (77) by a factor of , take the factor |κ| out of it, and multiply the power
2
of the exponential terms by 2.

B Expansion up to order 7
To iterate (39), we need to obtain the dominant mode of A3 and use it as part of the initial condition.
Therefore, we need to study the amplitude equation at order 7. To do this, we define an extra variable small
for the expansion: Ξ = ε2 X = ε4 x.
Therefore,

∂2u ∂2u 2
 2
∂2u ∂2u ∂2u

2 ∂ u 4 ∂ u
2
= 2
+ 2 ε + ε 2
+ 2 + 2 ε6 + ε8 .
∂x ∂x ∂x∂X ∂X ∂x∂Ξ ∂X∂Ξ ∂Ξ2

Thus, the equation at order five gets an extra term given by

∂ 2 u[1]
2 k 2 D̂ ,
∂x∂Ξ
which does not affect the Ginzburg-Landau equation, (6), but it does imply that A1 = A1 (X, Ξ), which only
implies that φ1 = φ1 (X, Ξ) = φ1 (X) + ζ(Ξ), where ζ is a real function still to determine.
dA
Here, we use the notation AX = ∂X A or AX = to denote partial or ordinary derivatives, disregarding
dX

71
the change of notation depending on whether functions depend on one or more variables. With this, we have
2 [5] 4 [5] [5]
u[5] = |A1 | W0 + |A1 | W0,2 + i Ā1 A1X W0,3
[5] [5] 2 [5] [5]
+ A1XX eix W1 + i A1X eix W1,2 + i |A1 | A1X eix W1,3 + i A21 Ā1X eix W1,4
[5] 2 [5] 4 [5] [3]
+ A1 eix W1,5 + |A1 | A1 eix W1,6 + |A1 | A1 eix W1,7 + i A1Ξ W1,3 eix
[5] 2 [5] [5]
+ A21 e2ix W2 + |A1 | A21 e2ix W2,2 + i A1 A1X e2ix W2,3
[5] 2 [5] [5]
+ A31 e3ix W3 + |A1 | A31 e3ix W3,2 + i A21 A1X e3ix W3,3
[4] 2 [4] [4] [4]
+ 2 A1 Ā2 W0 + 4 |A1 | A1 Ā2 W0,2 + i A1X Ā2 W0,3 + i Ā1 A2X W0,3
[3] 2 [3] [2] [2]
+ 2 A1 Ā3 W0 + |A2 | W0 + 2 A2 Ā3 W0 + 2 A1 Ā4 W0
[4] [4] 2 [4] [4]
+ A2 eix W1 + A21 Ā2 eix W1,2 + 2 |A1 | A2 eix W1,2 + i A2X eix W1,3
[3] 2 [3] [3] [3] 2 [3]
+ A3 eix W1 + 2 A1 |A2 | eix W1,2 + Ā1 A22 eix W1,2 + A21 Ā3 eix W1,2 + 2 |A1 | A3 eix W1,2
[3] [2] [1]
+ i A3X eix W1,3 + A4 eix W1 + A5 eix ϕ1
[4] 2 [4] [4] [4]
+ 2 A1 A2 e2ix W2 + 3 |A1 | A1 A2 e2ix W2,2 + A31 Ā2 e2ix W2,2 + i A1 A2X e2ix W2,3
[4] [3] [3] [2] [2]
+ i A1X A2 e2ix W2,3 + 2 A1 A3 e2ix W2 + A22 e2ix W2 + 2 A1 A4 e2ix W2 + 2 A2 A3 e2ix W2
[4] [3] [3]
+ 3 A21 A2 e3ix W3 + 3 A21 A3 e3ix W3 + 3 A1 A22 e3ix W3 + . . . + c.c.,

where ‘. . .’ represents terms that are multiples of e4ix or e5ix , and

[5] [4] [4] [3] [3]
M0 W0 = −a1 Ju f1,0 (0) W0 − b1 Ju f0,1 (0) W0 − a2 Ju f1,0 (0) W0 − b2 Ju f0,1 (0) W0
[2] [2] [3] [3]
− a3 Ju f1,0 (0) W0 − b3 Ju f0,1 (0) W0 − a21 Ju f2,0 (0) W0 − a1 b1 Ju f1,1 (0) W0
[3] [2] [2]
− b21 Ju f0,2 (0) W0 − 2 a1 a2 Ju f2,0 (0) W0 − (a1 b2 + a2 b1 ) Ju f1,1 (0) W0
[2] [2] [2] [2]
− 2 b1 b2 Ju f0,2 (0) W0 − a31 Ju f3,0 (0) W0 − a21 b1 Ju f2,1 (0) W0 − a1 b21 Ju f1,2 (0) W0
   
[2] [1] [4] [2] [3]
− b31 Ju f0,3 (0) W0 − 2 F2,0,0 ϕ1 , W1 − 2 F2,0,0 W1 , W1
     
[1] [3] [1] [3] [2] [2]
− 2 a1 F2,1,0 ϕ1 , W1 − 2 b1 F2,0,1 ϕ1 , W1 − a1 F2,1,0 W1 , W1
     
[2] [2] [1] [2] [1] [2]
− b1 F2,0,1 W1 , W1 − 2 a2 F2,1,0 ϕ1 , W1 − 2 b2 F2,0,1 ϕ1 , W1
     
[1] [1] [1] [1] [1] [2]
− a3 F2,1,0 ϕ1 , ϕ1 − b3 F2,0,1 ϕ1 , ϕ1 − 2 a21 F2,2,0 ϕ1 , W1
     
[1] [2] [1] [2] [1] [1]
− 2 a1 b1 F2,1,1 ϕ1 , W1 − 2 b21 F2,0,2 ϕ1 , W1 − 2 a1 a2 F2,2,0 ϕ1 , ϕ1
     
[1] [1] [1] [1] [1] [1]
− (a1 b2 + a2 b1 ) F2,1,1 ϕ1 , ϕ1 − 2 b1 b2 F2,0,2 ϕ1 , ϕ1 − a31 F2,3,0 ϕ1 , ϕ1
     
[1] [1] [1] [1] [1] [1]
− a21 b1 F2,2,1 ϕ1 , ϕ1 − a1 b21 F2,1,2 ϕ1 , ϕ1 − b31 F2,0,3 ϕ1 , ϕ1 ,

72
  
[5] [4] [4] [1] [4]
M0 W0,2 = −a1 Ju f1,0 (0) W0,2 − b1 Ju f0,1 (0) W0,2 − 2 F2,0,0 ϕ1 , W1,2
     
[2] [3] [2] [3] [2] [3]
− 4 F2,0,0 W0 , W0 − 2 F2,0,0 W1 , W1,2 − 2 F2,0,0 W2 , W2
     
[1] [3] [1] [3] [2] [2]
− 2 a1 F2,1,0 ϕ1 , W1,2 − 2 b1 F2,0,1 ϕ1 , W1,2 − 2 a1 F2,1,0 W0 , W0
     
[2] [2] [2] [2] [2] [2]
− 2 b1 F2,0,1 W0 , W0 − a1 F2,1,0 W2 , W2 − b1 F2,0,1 W2 , W2
   
[1] [1] [3] [3] [1] [2] [2] [2]
− 3 F3,0,0 ϕ1 , ϕ1 , 2 W0 + W2 − 6 F3,0,0 ϕ1 , W1 , 2 W0 + W2
   
[1] [1] [2] [2] [1] [1] [2] [2]
− 3 a1 F3,1,0 ϕ1 , ϕ1 , 2 W0 + W2 − 3 b1 F3,0,1 ϕ1 , ϕ1 , 2 W0 + W2
   
[1] [1] [1] [2] [1] [1] [1] [1]
− 12 F4,0,0 ϕ1 , ϕ1 , ϕ1 , W1 − 3 a1 F4,1,0 ϕ1 , ϕ1 , ϕ1 , ϕ1
 
[1] [1] [1] [1]
− 3 b1 F4,0,1 ϕ1 , ϕ1 , ϕ1 , ϕ1 ,

 
[5] [4] [4] [1] [4]
M0 W0,3 = −a1 Ju f1,0 (0) W0,3 − b1 Ju f0,1 (0) W0,3 − 2 F2,0,0 ϕ1 , W1,3
     
[2] [3] [1] [3] [1] [3]
− 2 F2,0,0 W1 , W1,3 − 2 a1 F2,1,0 ϕ1 , W1,3 − 2 b1 F2,0,1 ϕ1 , W1,3 .

Moreover, to find the resonant solution to (22), we note that the corresponding equation will be given by

[5] [5] [5] [5] 2 [5] [5] [5] 2 [5] 4

M1 u1 = −Q1 A′′1 − i Q2 A′1 − i Q3 |A1 | A′1 − i Q4 A21 Ā′1 − Q5 A1 − Q6 |A1 | A1 − Q7 |A1 | A1 .
(102)

Unfortunately, we cannot simply solve the equation for each term separately, as we cannot ensure that
[5]
Qp ∈ im (M1 ) for every p = 1, . . . , 7. Nevertheless, we know that
2 2 4
α1 A′′1 + i α2 A′1 + i α3 |A1 | A′1 + i α4 A21 Ā′1 + α5 A1 + α6 |A1 | A1 + α7 |A1 | A1 = 0,

which implies

2 2 4
 ψ
α1 A′′1 + i α2 A′1 + i α3 |A1 | A′1 + i α4 A21 Ā′1 + α5 A1 + α6 |A1 | A1 + α7 |A1 | A1 = 0.
⟨ψ, ψ⟩

Therefore, we see that (102) is equivalent to

     
[5] [5] ψ [5] ψ [5] ψ 2
M1 u1 = −Q1 + α1 A′′1 + i −Q2 + α2 A′1 + i −Q3 + α3 |A1 | A′1
⟨ψ, ψ⟩ ⟨ψ, ψ⟩ ⟨ψ, ψ⟩
   
[5] ψ [5] ψ
+ i −Q4 + α4 A21 Ā′1 + −Q5 + α5 A1
⟨ψ, ψ⟩ ⟨ψ, ψ⟩
   
[5] ψ 2 [5] ψ 4
+ −Q6 + α6 |A1 | A1 + −Q7 + α7 |A1 | A1 ,
⟨ψ, ψ⟩ ⟨ψ, ψ⟩

which satisfies that the coefficient of each term related to the amplitude, A1 , is a vector that belongs to
im (M1 ). Therefore,
[5] [5] α1
M1 W1 = −Q1 + ψ,
⟨ψ, ψ⟩

[5] αp
M1 W1,p = −Q[5]
p + ψ for p = 2, . . . , 7.
⟨ψ, ψ⟩

73
Furthermore,
[5] [4] [4] [3] [3]
M2 W2 = −a1 Ju f1,0 (0) W2 − b1 Ju f0,1 (0) W2 − a2 Ju f1,0 (0) W2 − b2 Ju f0,1 (0) W2
[2] [2] [3] [3]
− a3 Ju f1,0 (0) W2 − b3 Ju f0,1 (0) W2 − a21 Ju f2,0 (0) W2 − a1 b1 Ju f1,1 (0) W2
[3] [2] [2]
− b21 Ju f0,2 (0) W2 − 2 a1 a2 Ju f2,0 (0) W2 − (a1 b2 + a2 b1 ) Ju f1,1 (0) W2
[2] [2] [2] [2]
− 2 b1 b2 Ju f0,2 (0) W2 − a31 Ju f3,0 (0) W2 − a21 b1 Ju f2,1 (0) W2 − a1 b21 Ju f1,2 (0) W2
     
[2] [1] [4] [2] [3] [1] [3]
− b31 Ju f0,3 (0) W2 − 2 F2,0,0 ϕ1 , W1 − 2 F2,0,0 W1 , W1 − 2 a1 F2,1,0 ϕ1 , W1
     
[1] [3] [2] [2] [2] [2]
− 2 b1 F2,0,1 ϕ1 , W1 − a1 F2,1,0 W1 , W1 − b1 F2,0,1 W1 , W1
     
[1] [2] [1] [2] [1] [1]
− 2 a2 F2,1,0 ϕ1 , W1 − 2 b2 F2,0,1 ϕ1 , W1 − a3 F2,1,0 ϕ1 , ϕ1
     
[1] [1] [1] [2] [1] [2]
− b3 F2,0,1 ϕ1 , ϕ1 − 2 a21 F2,2,0 ϕ1 , W1 − 2 a1 b1 F2,1,1 ϕ1 , W1
     
[1] [2] [1] [1] [1] [1]
− 2 b21 F2,0,2 ϕ1 , W1 − 2 a1 a2 F2,2,0 ϕ1 , ϕ1 − a1 b2 F2,1,1 ϕ1 , ϕ1
     
[1] [1] [1] [1] [1] [1]
− a2 b1 F2,1,1 ϕ1 , ϕ1 − 2 b1 b2 F2,0,2 ϕ1 , ϕ1 − a31 F2,3,0 ϕ1 , ϕ1
     
[1] [1] [1] [1] [1] [1]
− a21 b1 F2,2,1 ϕ1 , ϕ1 − a1 b21 F2,1,2 ϕ1 , ϕ1 − b31 F2,0,3 ϕ1 , ϕ1 ,

 
[5] [4] [4] [1] [4] [4]
M2 W2,2 = −a1 Ju f1,0 (0) W2,2 − b1 Ju f0,1 (0) W2,2 − 2 F2,0,0 ϕ1 , W1,2 + W3
     
[2] [3] [2] [3] [3] [2] [3]
− 4 F2,0,0 W0 , W2 − 2 F2,0,0 W1 , W1,2 + W3 − 4 F2,0,0 W2 , W0
   
[1] [3] [3] [1] [3] [3]
− 2 a1 F2,1,0 ϕ1 , W1,2 + W3 − 2 b1 F2,0,1 ϕ1 , W1,2 + W3
     
[2] [2] [2] [2] [1] [1] [3] [3]
− 4 a1 F2,1,0 W0 , W2 − 4 b1 F2,0,1 W0 , W2 − 6 F3,0,0 ϕ1 , ϕ1 , W0 + W2
   
[1] [2] [2] [2] [1] [1] [2] [2]
− 12 F3,0,0 ϕ1 , W1 , W0 + W2 − 6 a1 F3,1,0 ϕ1 , ϕ1 , W0 + W2
   
[1] [1] [2] [2] [1] [1] [1] [2]
− 6 b1 F3,0,1 ϕ1 , ϕ1 , W0 + W2 − 16 F4,0,0 ϕ1 , ϕ1 , ϕ1 , W1
   
[1] [1] [1] [1] [1] [1] [1] [1]
− 4 a1 F4,1,0 ϕ1 , ϕ1 , ϕ1 , ϕ1 − 4 b1 F4,0,1 ϕ1 , ϕ1 , ϕ1 , ϕ1 ,

   
[5] [4] [4] [1] [4] [2] [3]
M2 W2,3 = −a1 Ju f1,0 (0) W2,3 − b1 Ju f0,1 (0) W2,3 − 2 F2,0,0 ϕ1 , W1,3 − 2 F2,0,0 W1 , W1,3
   
[1] [3] [1] [3] [3]
− 2 a1 F2,1,0 ϕ1 , W1,3 − 2 b1 F2,0,1 ϕ1 , W1,3 − 8 k 2 D̂ W2 ,

74
 [5] [4] [4] [3] [3]
M3 W3 = −a1 Ju f1,0 (0) W3 − b1 Ju f0,1 (0) W3 − a2 Ju f1,0 (0) W3 − b2 Ju f0,1 (0) W3
 
[3] [3] [3] [1] [4]
− a21 Ju f2,0 (0) W3 − a1 b1 Ju f1,1 (0) W3 − b21 Ju f0,2 (0) W3 − 2 F2,0,0 ϕ1 , W2
     
[2] [3] [2] [3] [1] [3]
− 2 F2,0,0 W1 , W2 − 2 F2,0,0 W2 , W1 − 2 a1 F2,1,0 ϕ1 , W2
     
[1] [3] [2] [2] [2] [2]
− 2 b1 F2,0,1 ϕ1 , W2 − 2 a1 F2,1,0 W1 , W2 − 2 b1 F2,0,1 W1 , W2
     
[1] [2] [1] [2] [1] [2]
− 2 a2 F2,1,0 ϕ1 , W2 − 2 b2 F2,0,1 ϕ1 , W2 − 2 a21 F2,2,0 ϕ1 , W2
     
[1] [2] [1] [2] [1] [1] [3]
− 2 a1 b1 F2,1,1 ϕ1 , W2 − 2 b21 F2,0,2 ϕ1 , W2 − 3 F3,0,0 ϕ1 , ϕ1 , W1
     
[1] [2] [2] [1] [1] [2] [1] [1] [2]
− 3 F3,0,0 ϕ1 , W1 , W1 − 3 a1 F3,1,0 ϕ1 , ϕ1 , W1 − 3 b1 F3,0,1 ϕ1 , ϕ1 , W1
     
[1] [1] [1] [1] [1] [1] [1] [1] [1]
− a2 F3,1,0 ϕ1 , ϕ1 , ϕ1 − b2 F3,0,1 ϕ1 , ϕ1 , ϕ1 − a21 F3,2,0 ϕ1 , ϕ1 , ϕ1
   
[1] [1] [1] [1] [1] [1]
− a1 b1 F3,1,1 ϕ1 , ϕ1 , ϕ1 − b21 F3,0,2 ϕ1 , ϕ1 , ϕ1 ,

     
[5] [1] [4] [4] [2] [3] [2] [3]
M3 W3,2 = −2 F2,0,0 ϕ1 , W2,2 + W4 − 4 F2,0,0 W0 , W3 − 2 F2,0,0 W2 , W1,2
   
[1] [1] [3] [3] [1] [2] [2] [2]
− 3 F3,0,0 ϕ1 , ϕ1 , W1,2 + 2 W3 − 3 F3,0,0 ϕ1 , W2 , 4 W0 + W2
   
[1] [1] [1] [2] [2] [1] [1] [1] [1] [1]
− 4 F4,0,0 ϕ1 , ϕ1 , ϕ1 , 2 W0 + 3 W2 − 5 F5,0,0 ϕ1 , ϕ1 , ϕ1 , ϕ1 , ϕ1 ,

and
     
[5] [1] [4] [2] [3] [1] [1] [3] [3]
M3 W3,3 = −2 F2,0,0 ϕ1 , W2,3 − 2 F2,0,0 W2 , W1,3 − 3 F3,0,0 ϕ1 , ϕ1 , W1,3 − 18 k 2 D̂ W3 .

Order O ε6 . At this order, (4) becomes

0 = Ju f0,0 (0) u[6] + a1 Ju f1,0 (0) u[5] + b1 Ju f0,1 (0) u[5] + a2 Ju f1,0 (0) u[4] + b2 Ju f0,1 (0) u[4]
+ a3 Ju f1,0 (0) u[3] + b3 Ju f0,1 (0) u[3] + a4 Ju f1,0 (0) u[2] + b4 Ju f0,1 (0) u[2] + a5 Ju f1,0 (0) u[1]
+ b5 Ju f0,1 (0) u[1] + a21 Ju f2,0 (0) u[4] + a1 b1 Ju f1,1 (0) u[4] + b21 Ju f0,2 (0) u[4]
+ 2 a1 a2 Ju f2,0 (0) u[3] + (a1 b2 + a2 b1 ) Ju f1,1 (0) u[3] + 2 b1 b2 Ju f0,2 (0) u[3]
+ 2 a1 a3 Ju f2,0 (0) u[2] + (a1 b3 + a3 b1 ) Ju f1,1 (0) u[2] + 2 b1 b3 Ju f0,2 (0) u[2]
+ 2 a1 a4 Ju f2,0 (0) u[1] + (a1 b4 + a4 b1 ) Ju f1,1 (0) u[1] + 2 b1 b4 Ju f0,2 (0) u[1]
+ a22 Ju f2,0 (0) u[2] + a2 b2 Ju f1,1 (0) u[2] + b22 Ju f0,2 (0) u[2] + 2 a2 a3 Ju f2,0 (0) u[1]
+ (a2 b3 + a3 b2 ) Ju f1,1 (0) u[1] + 2 b2 b3 Ju f0,2 (0) u[1] + a31 Ju f3,0 (0) u[3]
+ a21 b1 Ju f2,1 (0) u[3] + a1 b21 Ju f1,2 (0) u[3] + b31 Ju f0,3 (0) u[3] + 3 a21 a2 Ju f3,0 (0) u[2]
+ 2 a1 a2 b1 Ju f2,1 (0) u[2] + a2 b21 Ju f1,2 (0) u[2] + a21 b2 Ju f2,1 (0) u[2] + 2 a1 b1 b2 Ju f1,2 (0) u[2]
+ 3 b21 b2 Ju f0,3 (0) u[2] + 3 a21 a3 Ju f3,0 (0) u[1] + 2 a1 a3 b1 Ju f2,1 (0) u[1] + a3 b21 Ju f1,2 (0) u[1]
+ a21 b3 Ju f (0) u[1] + 2 a1 b1 b3 Ju f1,2 (0) u[1] + 3 b21 b3 Ju f0,3 (0) u[1] + 3 a1 a22 Ju f (0) u[1]
+ 2 a1 a2 b2 Ju f2,1 (0) u[1] + a1 b22 Ju f1,2 (0) u[1] + a22 b1 Ju f2,1 (0) u[1] + 2 a2 b1 b2 Ju f1,2 (0) u[1]
+ 3 b1 b22 Ju f (0) u[1] + a41 Ju f4,0 (0) u[2] + a31 b1 Ju f3,1 (0) u[2] + a21 b21 Ju f2,2 (0) u[2]
+ a1 b31 Ju f1,3 (0) u[2] + b41 Ju f0,4 (0) u[2] + 4 a31 a2 Ju f4,0 (0) u[1] + 3 a21 a2 b1 Ju f3,1 (0) u[1]
+ 2 a1 a2 b21 Ju f2,2 (0) u[1] + a2 b31 Ju f1,3 (0) u[1] + a31 b2 Ju f3,1 (0) u[1] + 2 a21 b1 b2 Ju f2,2 (0) u[1]
+ 3 a1 b21 b2 Ju f1,3 (0) u[1] + 4 b31 b2 Ju f0,4 (0) u[1] + a51 Ju f5,0 (0) u[1] + a41 b1 Ju f4,1 (0) u[1]
+ a31 b21 Ju f3,2 (0) u[1] + a21 b31 Ju f2,3 (0) u[1] + a1 b41 Ju f1,4 u[1] + b51 Ju f0,5 (0) u[1]

75
        
+ 2 F2,0,0 u[1] , u[5] + 2 F2,0,0 u[2] , u[4] + F2,0,0 u[3] , u[3] + 2 a1 F2,1,0 u[1] , u[4]
       
+ 2 b1 F2,0,1 u[1] , u[4] + 2 a1 F2,1,0 u[2] , u[3] + 2 b1 F2,0,1 u[2] , u[3] + 2 a2 F2,1,0 u[1] , u[3]
       
+ 2 b2 F2,0,1 u[1] , u[3] + a2 F2,1,0 u[2] , u[2] + b2 F2,0,1 u[2] , u[2] + 2 a3 F2,1,0 u[1] , u[2]
       
+ 2 b3 F2,0,1 u[1] , u[2] + a4 F2,1,0 u[1] , u[1] + b4 F2,0,1 u[1] , u[1] + 2 a21 F2,2,0 u[1] , u[3]
       
+ 2 a1 b1 F2,1,1 u[1] , u[3] + 2 b21 F2,0,2 u[1] , u[3] + a21 F2,2,0 u[2] , u[2] + a1 b1 F2,1,1 u[2] , u[2]
     
+ b21 F2,0,2 u[2] , u[2] + 4 a1 a2 F2,2,0 u[1] , u[2] + 2 (a1 b2 + a2 b1 ) F2,1,1 u[1] , u[2]
     
+ 4 b1 b2 F2,0,2 u[1] , u[2] + 2 a1 a3 F2,2,0 u[1] , u[1] + (a1 b3 + a3 b1 ) F2,1,1 u[1] , u[1]
       
+ 2 b1 b3 F2,0,2 u[1] , u[1] + a22 F2,2,0 u[1] , u[1] + a2 b2 F2,1,1 u[1] , u[1] + b22 F2,0,2 u[1] , u[1]
       
+ 2 a31 F2,3,0 u[1] , u[2] + 2 a21 b1 F2,0,1 u[1] , u[2] + 2 a1 b21 F2,1,2 u[1] , u[2] + 2 b31 F2,0,3 u[1] , u[2]
     
+ 3 a21 a2 F2,3,0 u[1] , u[1] + 2 a1 a2 b1 F2,2,1 u[1] , u[1] + a2 b21 F2,1,2 u[1] , u[1]
       
+ a21 b2 F2,2,1 u[1] , u[1] + 2 a1 b1 b2 F2,1,2 u[1] , u[1] + 3 b21 b2 F2,0,3 u[1] , u[1] + a41 F2,4,0 u[1] , u[1]
       
+ a31 b1 F2,3,1 u[1] , u[1] + a21 b21 F2,2,2 u[1] , u[1] + a1 b31 F2,1,3 u[1] , u[1] + b41 F2,0,4 u[1] , u[1]
     
+ 3 F3,0,0 u[1] , u[1] , u[4] + 6 F3,0,0 u[1] , u[2] , u[3] + F3,0,0 u[2] , u[2] , u[2]
     
+ 3 a1 F3,1,0 u[1] , u[1] , u[3] + 3 b1 F3,0,1 u[1] , u[1] , u[3] + 3 a1 F3,1,0 u[1] , u[2] , u[2]
     
+ 3 b1 F3,0,1 u[1] , u[2] , u[2] + 3 a2 F3,1,0 u[1] , u[1] , u[2] + 3 b2 F3,0,1 u[1] , u[1] , u[2]
     
+ a3 F3,1,0 u[1] , u[1] , u[1] + b3 F3,0,1 u[1] , u[1] , u[1] + 3 a21 F3,2,0 u[1] , u[1] , u[2]
     
+ 3 a1 b1 F3,1,1 u[1] , u[1] , u[2] + 3 b21 F3,0,2 u[1] , u[1] , u[2] + 2 a1 a2 F3,2,0 u[1] , u[1] , u[1]
     
+ (a1 b2 + a2 b1 ) F3,1,1 u[1] , u[1] , u[1] + 2 b1 b2 F3,0,2 u[1] , u[1] , u[1] + a31 F3,3,0 u[1] , u[1] , u[1]
     
+ a21 b1 F3,2,1 u[1] , u[1] , u[1] + a1 b21 F3,1,2 u[1] , u[1] , u[1] + b31 F3,0,3 u[1] , u[1] , u[1]
     
+ 4 F4,0,0 u[1] , u[1] , u[1] , u[3] + 6 F4,0,0 u[1] , u[1] , u[2] , u[2] + 4 a1 F4,1,0 u[1] , u[1] , u[1] , u[2]
     
+ 4 b1 F4,0,1 u[1] , u[1] , u[1] , u[2] + a2 F4,1,0 u[1] , u[1] , u[1] , u[1] + b2 F4,0,2 u[1] , u[1] , u[1] , u[1]
     
+ a21 F4,2,0 u[1] , u[1] , u[1] , u[1] + a1 b1 F4,1,1 u[1] , u[1] , u[1] , u[1] + b21 F4,0,2 u[1] , u[1] , u[1] , u[1]
   
+ 5 F5,0,0 u[1] , u[1] , u[1] , u[1] , u[2] + a1 F5,1,0 u[1] , u[1] , u[1] , u[1] , u[1]
   
+ b1 F5,0,1 u[1] , u[1] , u[1] , u[1] , u[1] + F6,0,0 u[1] , u[1] , u[1] , u[1] , u[1] , u[1]
 2 [6]
∂ 2 u[4] ∂ 2 u[2] ∂ 2 u[2]

2 ∂ u
+ k D̂ +2 +2 + .
∂x2 ∂x∂X ∂x∂Ξ ∂X 2

76
From this, we note that
2 [6] 4 [6] 6 [6] 2 [6] [6] 2 [6]
u[6] = |A1 | W0 + |A1 | W0,2 + |A1 | W0,3 + |A1X | W0,4 + i Ā1 A1X W0,5 + i |A1 | Ā1 A1X W0,6
[6] [4] [6] [6] 2 [6]
+ Ā1 A1XX W0,7 + i Ā1 A1Ξ W0,3 + A1XX eix W1 + i A1X eix W1,2 + i |A1 | A1X eix W1,3
[6] [6] 2 [6] 4 [6] [4]
+ i A21 Ā1X eix W1,4 + A1 eix W1,5 + |A1 | A1 eix W1,6 + |A1 | A1 eix W1,7 + i A1Ξ eix W1,3
[6] 2 [6] 4 [6] 2 [6] [6]
+ A21 e2ix W2 + |A1 | A21 e2ix W2,2 + |A1 | A21 e2ix W2,3 + (A1X ) e2ix W2,4 + i A1 A1X e2ix W2,5
2 [6] [6] [6] [4]
+ i |A1 | A1 A1X e2ix W2,6 + i A31 Ā1X e2ix W2,7 + A1 A1XX e2ix W2,8 + i A1 A1Ξ e2ix W2,3
[5] 2 [5] [5] [5] 2 [4] [4]
+ 2 A1 Ā2 W0 + 4 |A1 | A1 Ā2 W0,2 + i A1X Ā2 W0,3 + i Ā1 A2X W0,3 + |A2 | W0 + 2 A1 Ā3 W0
2 [4] 2 2 [4] [4] [4] [4]
+ 4 |A1 | A1 Ā3 W0,2 + 4 |A1 | |A2 | W0,2 + 2 A21 Ā22 W0,2 + i A1X Ā3 W0,3 + i Ā1 A3X W0,3
[4] [3] [3] [2] [2] 2 [2]
+ i Ā2 A2X W0,3 + 2 A1 Ā4 W0 + 2 A2 Ā3 W0 + 2 A1 Ā5 W0 + 2 A2 Ā4 W0 + |A3 | W0
[5] [5] 2 [5] [5]
+ A2XX eix W1 + i A2X eix W1,2 + i |A1 | A2X eix W1,3 + i A1 A1X Ā2 eix W1,3
[5] [5] [5] [5]
+ i Ā1 A1X A2 eix W1,3 + i A21 Ā2X eix W1,4 + 2 i A1 Ā1X A2 eix W1,4 + A2 eix W1,5
2 [5] [5] 4 [5] 2 [5]
+ 2 |A1 | A2 eix W1,6 + A21 Ā2 eix W1,6 + 3 |A1 | A2 eix W1,7 + 2 |A1 | A21 Ā2 eix W1,7
[3] [4] 2 [4] [4] 2 [4]
+ i A2Ξ eix W1,3 + A3 eix W1 + 2 |A1 | A3 eix W1,2 + Ā1 A22 eix W1,2 + 2 |A2 | A1 eix W1,2
[4] [4] [3] 2 [3] [3]
+ A21 Ā3 eix W1,2 + i A3X eix W1,3 + A4 eix W1 + 2 |A1 | A4 eix W1,2 + A21 Ā4 eix W1,2
[3] [3] [3] 2 [3]
+ 2 A1 A2 Ā3 eix W1,2 + 2 A1 Ā2 A3 eix W1,2 + 2 Ā1 A2 A3 eix W1,2 + |A2 | A2 eix W1,2
[3] [2] [1] [5] 2 [5]
+ i A4X eix W1,3 + A5 eix W1 + A6 eix ϕ1 + 2 A1 A2 e2ix W2 + 3 |A1 | A1 A2 e2ix W2,2
[5] [5] [5] [4] [4]
+ A31 Ā2 e2ix W2,2 + i A1 A2X e2ix W2,3 + i A1X A2 e2ix W2,3 + A22 e2ix W2 + 2 A1 A3 e2ix W2
2 [4] [4] 2 [4] 2 [4]
+ 3 |A1 | A1 A3 e2ix W2,2 + A31 Ā3 e2ix W2,2 + 3 A21 |A2 | e2ix W2,2 + 3 |A1 | A22 e2ix W2,2
[4] [4] [4] [3]
+ i A1X A3 e2ix W2,3 + i A1 A3X e2ix W2,3 + i A2 A2X e2ix W2,3 + 2 A1 A4 e2ix W2
[3] [2] [2] [2]
+ 2 A2 A3 e2ix W2 + 2 A1 A5 e2ix W2 + 2 A2 A4 e2ix W2 + A23 e2ix W2 + . . . + c.c.
where ‘. . . ’ represents terms that are multiples of e3ix , e4ix , e5ix or e6ix , and
[6] [5] [5] [4] [4]
M0 W0 = −a1 Ju f1,0 (0) W0 − b1 Ju f0,1 (0) W0 − a2 Ju f1,0 (0) W0 − b2 Ju f0,1 (0) W0
[3] [3] [2] [2]
− a3 Ju f1,0 (0) W0 − b3 Ju f0,1 (0) W0 − a4 Ju f1,0 (0) W0 − b4 Ju f0,1 (0) W0
[4] [4] [4] [3]
− a21 Ju f2,0 (0) W0 − a1 b1 Ju f1,1 (0) W0 − b21 Ju f0,2 (0) W0 − 2 a1 a2 Ju f2,0 (0) W0
[3] [3] [2]
− (a1 b2 + a2 b1 ) Ju f1,1 (0) W0 − 2 b1 b2 Ju f0,2 (0) W0 − 2 a1 a3 Ju f2,0 (0) W0
[2] [2] [2]
− (a1 b3 + a3 b1 ) Ju f1,1 (0) W0 − 2 b1 b3 Ju f0,2 (0) W0 − a22 Ju f2,0 (0) W0
[2] [2] [3] [3]
− a2 b2 Ju f1,1 (0) W0 − b22 Ju f0,2 (0) W0 − a31 Ju f3,0 (0) W0 − a21 b1 Ju f2,1 (0) W0
[3] [3] [2] [2]
− a1 b21 Ju f1,2 (0) W0 − b31 Ju f0,3 (0) W0 − 3 a21 a2 Ju f3,0 (0) W0 − 2 a1 a2 b1 Ju f2,1 (0) W0
[2] [2] [2] [2]
− a2 b21 Ju f1,2 (0) W0 − a21 b2 Ju f2,1 (0) W0 − 2 a1 b1 b2 Ju f1,2 (0) W0 − 3 b21 b2 Ju f0,3 (0) W0
[2] [2] [2] [2]
− a41 Ju f4,0 (0) W0 − a31 b1 Ju f3,1 (0) W0 − a21 b21 Ju f2,2 (0) W0 − a1 b31 Ju f1,3 (0) W0
     
[2] [1] [5] [2] [4] [3] [3]
− b41 Ju f0,4 (0) W0 − 2 F2,0,0 ϕ1 , W1,5 − 2 F2,0,0 W1 , W1 − F2,0,0 W1 , W1
     
[1] [4] [1] [4] [2] [3]
− 2 a1 F2,1,0 ϕ1 , W1 − 2 b1 F2,0,1 ϕ1 , W1 − 2 a1 F2,1,0 W1 , W1
     
[2] [3] [1] [3] [1] [3]
− 2 b1 F2,0,1 W1 , W1 − 2 a2 F2,1,0 ϕ1 , W1 − 2 b2 F2,0,1 ϕ1 , W1
     
[2] [2] [2] [2] [1] [2]
− a2 F2,1,0 W1 , W1 − b2 F2,0,1 W1 , W1 − 2 a3 F2,1,0 ϕ1 , W1
     
[1] [2] [1] [1] [1] [1]
− 2 b3 F2,0,1 ϕ1 , W1 − a4 F2,1,0 ϕ1 , ϕ1 − b4 F2,0,1 ϕ1 , ϕ1

77
      
[1] [3] [1] [3] [1] [3]
− 2 a21 F2,2,0 ϕ1 , W1 − 2 a1 b1 F2,1,1 ϕ1 , W1 − 2 b21 F2,0,2 ϕ1 , W1
     
[2] [2] [2] [2] [2] [2]
− a21 F2,2,0 W1 , W1 − a1 b1 F2,1,1 W1 , W1 − b21 F2,0,2 W1 , W1
     
[1] [2] [1] [2] [1] [2]
− 4 a1 a2 F2,2,0 ϕ1 , W1 − 2 (a1 b2 + a2 b1 ) F2,1,1 ϕ1 , W1 − 4 b1 b2 F2,0,2 ϕ1 , W1
     
[1] [1] [1] [1] [1] [1]
− 2 a1 a3 F2,2,0 ϕ1 , ϕ1 − (a1 b3 + a3 b1 ) F2,1,1 ϕ1 , ϕ1 − 2 b1 b3 F2,0,2 ϕ1 , ϕ1
     
[1] [1] [1] [1] [1] [1]
− a22 F2,2,0 ϕ1 , ϕ1 − a2 b2 F2,1,1 ϕ1 , ϕ1 − b22 F2,0,2 ϕ1 , ϕ1
     
[1] [2] [1] [2] [1] [2]
− 2 a31 F2,3,0 ϕ1 , W1 − 2 a21 b1 F2,2,1 ϕ1 , W1 − 2 a1 b21 F2,1,2 ϕ1 , W1
     
[1] [2] [1] [1] [1] [1]
− 2 b31 F2,0,3 ϕ1 , W1 − 3 a21 a2 F2,3,0 ϕ1 , ϕ1 − 2 a1 a2 b1 F2,2,1 ϕ1 , ϕ1
     
[1] [1] [1] [1] [1] [1]
− a2 b21 F2,1,2 ϕ1 , ϕ1 − a21 b2 F2,2,1 ϕ1 , ϕ1 − 2 a1 b1 b2 F2,1,2 ϕ1 , ϕ1
     
[1] [1] [1] [1] [1] [1]
− 3 b21 b2 F2,0,3 ϕ1 , ϕ1 − a41 F2,4,0 ϕ1 , ϕ1 − a31 b1 F2,3,1 ϕ1 , ϕ1
     
[1] [1] [1] [1] [1] [1]
− a21 b21 F2,2,2 ϕ1 , ϕ1 − a1 b31 F2,1,3 ϕ1 , ϕ1 − b41 F2,0,4 ϕ1 , ϕ1 ,

[6] [5] [5] [4] [4]

M0 W0,2 = −a1 Ju f1,0 (0) W0,2 − b1 Ju f0,1 (0) W0,2 − a2 Ju f1,0 (0) W0,2 − b2 Ju f0,1 (0) W0,2
 
[4] [4] [4] [1] [5]
− a21 Ju f2,0 (0) W0,2 − a1 b1 Ju f1,1 (0) W0,2 − b21 Ju f0,2 (0) W0,2 − 2 F2,0,0 ϕ1 , W1,6
     
[2] [4] [2] [4] [2] [4]
− 4 F2,0,0 W0 , W0 − 2 F2,0,0 W1 , W1,2 − 2 F2,0,0 W2 , W2
     
[3] [3] [3] [3] [3] [3]
− 2 F2,0,0 W0 , W0 − 2 F2,0,0 W1 , W1,2 − F2,0,0 W2 , W2
     
[1] [4] [1] [4] [2] [3]
− 2 a1 F2,1,0 ϕ1 , W1,2 − 2 b1 F2,0,1 ϕ1 , W1,2 − 4 a1 F2,1,0 W0 , W0
     
[2] [3] [2] [3] [2] [3]
− 4 b1 F2,0,1 W0 , W0 − 2 a1 F2,1,0 W1 , W1,2 − 2 b1 F2,0,1 W1 , W1,2
     
[2] [3] [2] [3] [1] [3]
− 2 a1 F2,1,0 W2 , W2 − 2 b1 F2,0,1 W2 , W2 − 2 a2 F2,1,0 ϕ1 , W1,2
     
[1] [3] [2] [2] [2] [2]
− 2 b2 F2,0,1 ϕ1 , W1,2 − 2 a2 F2,1,0 W0 , W0 − 2 b2 F2,0,1 W0 , W0
     
[2] [2] [2] [2] [1] [3]
− a2 F2,1,0 W2 , W2 − b2 F2,0,1 W2 , W2 − 2 a21 F2,2,0 ϕ1 , W1,2
     
[1] [3] [1] [3] [2] [2]
− 2 a1 b1 F2,1,1 ϕ1 , W1,2 − 2 b21 F2,0,2 ϕ1 , W1,2 − 2 a21 F2,2,0 W0 , W0
     
[2] [2] [2] [2] [2] [2]
− 2 a1 b1 F2,1,1 W0 , W0 − 2 b21 F2,0,2 W0 , W0 − a21 F2,2,0 W2 , W2
     
[2] [2] [2] [2] [1] [1] [4] [4]
− a1 b1 F2,1,1 W2 , W2 − b21 F2,0,2 W2 , W2 − 3 F3,0,0 ϕ1 , ϕ1 , 2 W0 + W2
   
[1] [2] [3] [3] [1] [3] [2] [2]
− 6 F3,0,0 ϕ1 , W1 , 2 W0 + W2 − 6 F3,0,0 ϕ1 , W1 , 2 W0 + W2
   
[2] [2] [2] [2] [1] [1] [3] [3]
− 3 F3,0,0 W1 , W1 , 2 W0 + W2 − 3 a1 F3,1,0 ϕ1 , ϕ1 , 2 W0 + W2
   
[1] [1] [3] [3] [1] [2] [2] [2]
− 3 b1 F3,0,1 ϕ1 , ϕ1 , 2 W0 + W2 − 6 a1 F3,1,0 ϕ1 , W1 , 2 W0 + W2
   
[1] [2] [2] [2] [1] [1] [2] [2]
− 6 b1 F3,0,1 ϕ1 , W1 , 2 W0 + W2 − 3 a2 F3,1,0 ϕ1 , ϕ1 , 2 W0 + W2
   
[1] [1] [2] [2] [1] [1] [2] [2]
− 3 b2 F3,0,1 ϕ1 , ϕ1 , 2 W0 + W2 − 3 a21 F3,2,0 ϕ1 , ϕ1 , 2 W0 + W2
   
[1] [1] [2] [2] [1] [1] [2] [2]
− 3 a1 b1 F3,1,1 ϕ1 , ϕ1 , 2 W0 + W2 − 3 b21 F3,0,2 ϕ1 , ϕ1 , 2 W0 + W2
   
[1] [1] [1] [3] [1] [1] [2] [2]
− 12 F4,0,0 ϕ1 , ϕ1 , ϕ1 , W1 − 18 F4,0,0 ϕ1 , ϕ1 , W1 , W1

78
    
[1] [1] [1] [2] [1] [1] [1] [2]
− 12 a1 F4,1,0 ϕ1 , ϕ1 , ϕ1 , W1 − 12 b1 F4,0,1 ϕ1 , ϕ1 , ϕ1 , W1
   
[1] [1] [1] [1] [1] [1] [1] [1]
− 3 a2 F4,1,0 ϕ1 , ϕ1 , ϕ1 , ϕ1 − 3 b2 F4,0,1 ϕ1 , ϕ1 , ϕ1 , ϕ1
   
[1] [1] [1] [1] [1] [1] [1] [1]
− 3 a21 F4,2,0 ϕ1 , ϕ1 , ϕ1 , ϕ1 − 3 a1 b1 F4,1,1 ϕ1 , ϕ1 , ϕ1 , ϕ1
 
[1] [1] [1] [1]
− 3 b21 F4,0,2 ϕ1 , ϕ1 , ϕ1 , ϕ1 ,

     
[6] [1] [5] [2] [4] [2] [4]
M0 W0,3 = −2 F2,0,0 ϕ1 , W1,7 − 4 F2,0,0 W0 , W0,2 − 2 F2,0,0 W2 , W2,2
     
[3] [3] [3] [3] [1] [1] [4] [4]
− F2,0,0 W1,2 , W1,2 − F2,0,0 W3 , W3 − 3 F3,0,0 ϕ1 , ϕ1 , 2 W0,2 + W2,2
   
[1] [2] [3] [1] [3] [2] [2]
− 6 F3,0,0 ϕ1 , W2 , W3 − 6 F3,0,0 ϕ1 , W1,2 , 2 W0 + W2
   
[2] [2] [2] [2] [2] [2]
− 4 F3,0,0 W0 , W0 , W0 − 6 F3,0,0 W0 , W2 , W2
   
[1] [1] [1] [3] [3] [1] [1] [2] [2]
− 4 F4,0,0 ϕ1 , ϕ1 , ϕ1 , 3 W1,2 + W3 − 24 F4,0,0 ϕ1 , ϕ1 , W0 , W0
   
[1] [1] [2] [2] [2] [1] [1] [1] [1] [2] [2]
− 12 F4,0,0 ϕ1 , ϕ1 , W2 , 2 W0 + W2 − 10 F5,0,0 ϕ1 , ϕ1 , ϕ1 , ϕ1 , 3 W0 + 2 W2
 
[1] [1] [1] [1] [1] [1]
− 10 F6,0,0 ϕ1 , ϕ1 , ϕ1 , ϕ1 , ϕ1 , ϕ1 ,

 
[6] [3] [3] [2]
M0 W0,4 = −F2,0,0 W1,3 , W1,3 − 2 k 2 D̂ W0 ,

[6] [5] [5] [4] [4]

M0 W0,5 = −a1 Ju f1,0 (0) W0,3 − b1 Ju f0,1 (0) W0,3 − a2 Ju f1,0 (0) W0,3 − b2 Ju f0,1 (0) W0,3
 
[4] [4] [4] [1] [5]
− a21 Ju f2,0 (0) W0,3 − a1 b1 Ju f1,1 (0) W0,3 − b21 Ju f0,2 (0) W0,3 − 2 F2,0,0 ϕ1 , W1,2
     
[2] [4] [3] [3] [1] [4]
− 2 F2,0,0 W1 , W1,3 − 2 F2,0,0 W1 , W1,3 − 2 a1 F2,1,0 ϕ1 , W1,3
     
[1] [4] [2] [3] [2] [3]
− 2 b1 F2,0,1 ϕ1 , W1,3 − 2 a1 F2,1,0 W1 , W1,3 − 2 b1 F2,0,1 W1 , W1,3
     
[1] [3] [1] [3] [1] [3]
− 2 a2 F2,1,0 ϕ1 , W1,3 − 2 b2 F2,0,1 ϕ1 , W1,3 − 2 a21 F2,2,0 ϕ1 , W1,3
   
[1] [3] [1] [3]
− 2 a1 b1 F2,1,1 ϕ1 , W1,3 − 2 b21 F2,0,2 ϕ1 , W1,3 ,

     
[6] [1] [5] [5] [2] [4] [2] [4]
M0 W0,6 = −2 F2,0,0 ϕ1 , W1,3 − W1,4 − 4 F2,0,0 W0 , W0,3 − 2 F2,0,0 W2 , W2,3
   
[3] [3] [1] [1] [4] [4]
− 2 F2,0,0 W1,2 , W1,3 − 3 F3,0,0 ϕ1 , ϕ1 , 2 W0,3 + W2,3
   
[1] [3] [2] [2] [1] [1] [1] [3]
− 6 F3,0,0 ϕ1 , W1,3 , 2 W0 + W2 − 12 F4,0,0 ϕ1 , ϕ1 , ϕ1 , W1,3 ,

and
 
[6] [1] [5] [2]
Ju f0,0 (0) W0,7 = −2 F2,0,0 ϕ1 , W1 − 2 k 2 D̂ W0 .

Furthermore, the determination of the vectors that are a multiple of eix leads to a solvability condition given
by
2
α1 A2XX + i α2 A2X + i α3 |A1 | A2X + i α3 A1 A1X Ā2 + i α3 Ā1 A1X A2 + i α4 A21 Ā2X + 2 i α4 A1 Ā1X A2
2 2 4
+ α5 A2 + α6 A21 Ā2 + 2 α6 |A1 | A2 + 2 α7 |A1 | A21 Ā2 + 3 α7 |A1 | A2
2 2 4
+ α1,2 A1XX + i α2,2 A1X + i α3,2 |A1 | A1X + i α4,2 A21 Ā1X + α5,2 A1 + α6,2 |A1 | A1 + α7,2 |A1 | A1 = 0,
(103)

79
 D E
[6]
αp,2 = Qp , ψ , for p = 1, . . . , 7, with

[6] [5] [5] [4] [2]

Q1 = a1 Ju f1,0 (0) W1 + b1 Ju f0,1 (0) W1 − 2 k 2 D̂ W1,3 + k 2 D̂ W1 ,
[6] [5] [5] [4] [4]
Q2 = a1 Ju f1,0 (0) W1,2 + b1 Ju f0,1 (0) W1,2 + a2 Ju f1,0 (0) W1,3 + b2 Ju f0,1 (0) W1,3
[3] [3] [4] [4]
+ a3 Ju f1,0 (0) W1,3 + b3 Ju f0,1 (0) W1,3 + a21 Ju f2,0 (0) W1,3 + a1 b1 Ju f1,1 (0) W1,3
[4] [3] [3]
+ b21 Ju f0,2 (0) W1,3 + 2 a1 a2 Ju f2,0 (0) W1,3 + (a1 b2 + a2 b1 ) Ju f1,1 (0) W1,3
[3] [3] [3] [3]
+ 2 b1 b2 Ju f0,2 (0) W1,3 + a31 Ju f3,0 (0) W1,3 + a21 b1 Ju f2,1 (0) W1,3 + a1 b21 Ju f1,2 (0) W1,3
[3] [4]
+ b31 Ju f0,3 (0) W1,3 + 2 k 2 D̂ W1 ,
 
[6] [5] [5] [1] [5] [5]
Q3 = a1 Ju f1,0 (0) W1,3 + b1 Ju f0,1 (0) W1,3 + 2 F2,0,0 ϕ1 , W0,3 + W2,3
     
[2] [4] [2] [4] [4] [3] [3]
+ 4 F2,0,0 W0 , W1,3 + 2 F2,0,0 W1 , W0,3 + W2,3 + 4 F2,0,0 W0 , W1,3
     
[1] [4] [4] [1] [4] [4] [2] [3]
+ 2 a1 F2,1,0 ϕ1 , W0,3 + W2,3 + 2 b1 F2,0,1 ϕ1 , W0,3 + W2,3 + 4 a1 F2,1,0 W0 , W1,3
     
[2] [3] [1] [1] [4] [1] [2] [3]
+ 4 b1 F2,0,1 W0 , W1,3 + 6 F3,0,0 ϕ1 , ϕ1 , W1,3 + 12 F3,0,0 ϕ1 , W1 , W1,3
   
[1] [1] [3] [1] [1] [3] [4]
+ 6 a1 F3,1,0 ϕ1 , ϕ1 , W1,3 + 6 b1 F3,0,1 ϕ1 , ϕ1 , W1,3 + 4 k 2 D̂ W1,2 ,
   
[6] [5] [5] [1] [5] [2] [4]
Q4 = a1 Ju f1,0 (0) W1,4 + b1 Ju f0,1 (0) W1,4 − 2 F2,0,0 ϕ1 , W0,3 − 2 F2,0,0 W1 , W0,3
     
[2] [4] [3] [3] [1] [4]
− 2 F2,0,0 W2 , W1,3 − 2 F2,0,0 W2 , W1,3 − 2 a1 F2,1,0 ϕ1 , W0,3
     
[1] [4] [2] [3] [2] [3]
− 2 b1 F2,0,1 ϕ1 , W0,3 − 2 a1 F2,1,0 W2 , W1,3 − 2 b1 F2,0,1 W2 , W1,3
     
[1] [1] [4] [1] [2] [3] [1] [1] [3]
− 3 F3,0,0 ϕ1 , ϕ1 , W1,3 − 6 F3,0,0 ϕ1 , W1 , W1,3 − 3 a1 F3,1,0 ϕ1 , ϕ1 , W1,3
 
[1] [1] [3] [4]
− 3 b1 F3,0,1 ϕ1 , ϕ1 , W1,3 + 2 k 2 D̂ W1,2 ,
[6] [5] [5] [4] [4]
Q5 = a1 Ju f1,0 (0) W1,5 + b1 Ju f0,1 (0) W1,5 + a2 Ju f1,0 (0) W1 + b2 Ju f0,1 (0) W1
[3] [3] [2] [2]
+ a3 Ju f1,0 (0) W1 + b3 Ju f0,1 (0) W1 + a4 Ju f1,0 (0) W1 + b4 Ju f0,1 (0) W1
[1] [1] [4] [4]
+ a5 Ju f1,0 (0) ϕ1 + b5 Ju f0,1 (0) ϕ1 + a21 Ju f2,0 (0) W1 + a1 b1 Ju f1,1 (0) W1
[4] [3] [3]
+ b21 Ju f0,2 (0) W1 + 2 a1 a2 Ju f2,0 (0) W1 + (a1 b2 + a2 b1 ) Ju f1,1 (0) W1
[3] [2] [2]
+ 2 b1 b2 Ju f0,2 (0) W1 + 2 a1 a3 Ju f2,0 (0) W1 + (a1 b3 + a3 b1 ) Ju f1,1 (0) W1
[2] [1] [1]
+ 2 b1 b3 Ju f0,2 (0) W1 + 2 a1 a4 Ju f2,0 (0) ϕ1 + (a1 b4 + a4 b1 ) Ju f1,1 (0) ϕ1
[1] [2] [2] [2]
+ 2 b1 b4 Ju f0,2 (0) ϕ1 + a22 Ju f2,0 (0) W1 + a2 b2 Ju f1,1 (0) W1 + b22 Ju f0,2 (0) W1
[1] [1] [1]
+ 2 a2 a3 Ju f2,0 (0) ϕ1 + (a2 b3 + a3 b2 ) Ju f1,1 (0) ϕ1 + 2 b2 b3 Ju f0,2 (0) ϕ1
[3] [3] [3] [3]
+ a31 Ju f3,0 (0) W1 + a21 b1 Ju f2,1 (0) W1 + a1 b21 Ju f1,2 (0) W1 + b31 Ju f0,3 (0) W1
[2] [2] [2] [2]
+ 3 a21 a2 Ju f3,0 (0) W1 + 2 a1 a2 b1 Ju f2,1 (0) W1 + a2 b21 Ju f1,2 (0) W1 + a21 b2 Ju f2,1 (0) W1
[2] [2] [1]
+ 2 a1 b1 b2 Ju f1,2 (0) W1 + 3 b21 b2 Ju f0,3 (0) W1 + 3 a21 a3 Ju f3,0 (0) ϕ1
[1] [1] [1] [1]
+ 2 a1 a3 b1 Ju f2,1 (0) ϕ1 + a3 b21 Ju f1,2 (0) ϕ1 + a21 b3 Ju f2,1 (0) ϕ1 + 2 a1 b1 b3 Ju f1,2 (0) ϕ1
[1] [1] [1] [1]
+ 3 b21 b3 Ju f0,3 (0) ϕ1 + 3 a1 a22 Ju f3,0 (0) ϕ1 + 2 a1 a2 b2 Ju f2,1 (0) ϕ1 + a1 b22 Ju f1,2 (0) ϕ1
[1] [1] [1] [2]
+ a22 b1 Ju f2,1 (0) ϕ1 + 2 a2 b1 b2 Ju f1,2 (0) ϕ1 + 3 b1 b22 Ju f0,3 (0) ϕ1 + a41 Ju f4,0 (0) W1
[2] [2] [2] [2]
+ a31 b1 Ju f3,1 (0) W1 + a21 b21 Ju f2,2 (0) W1 + a1 b31 Ju f1,3 (0) W1 + b41 Ju f0,4 (0) W1
[1] [1] [1] [1]
+ 4 a31 a2 Ju f4,0 (0) ϕ1 + 3 a21 a2 b1 Ju f3,1 (0) ϕ1 + 2 a1 a2 b21 Ju f2,2 (0) ϕ1 + a2 b31 Ju f1,3 (0) ϕ1

80
 [1] [1] [1] [1]
+ a31 b2 Ju f3,1 (0) ϕ1 + 2 a21 b1 b2 Ju f2,2 (0) ϕ1 + 3 a1 b21 b2 Ju f1,3 (0) ϕ1 + 4 b31 b2 Ju f0,4 (0) ϕ1
[1] [1] [1] [1]
+ a51 Ju f5,0 (0) ϕ1 + a41 b1 Ju f4,1 (0) ϕ1 + a31 b21 Ju f3,2 (0) ϕ1 + a21 b31 Ju f2,3 (0) ϕ1
[1] [1]
+ a1 b41 Ju f1,4 (0) ϕ1 + b51 Ju f0,5 (0) ϕ1 ,
[6] [5] [5] [4] [4]
Q6 = a1 Ju f1,0 (0) W1,6 + b1 Ju f0,1 (0) W1,6 + a2 Ju f1,0 (0) W1,2 + b2 Ju f0,1 (0) W1,2
[3] [3] [4] [4]
+ a3 Ju f1,0 (0) W1,2 + b3 Ju f0,1 (0) W1,2 + a21 Ju f2,0 (0) W1,2 + a1 b1 Ju f1,1 (0) W1,2
[4] [3] [3]
+ b21 Ju f0,2 (0) W1,2 + 2 a1 a2 Ju f2,0 (0) W1,2 + (a1 b2 + a2 b1 ) Ju f1,1 (0) W1,2
[3] [3] [3] [3]
+ 2 b1 b2 Ju f0,2 (0) W1,2 + a31 Ju f3,0 (0) W1,2 + a21 b1 Ju f2,1 (0) W1,2 + a1 b21 Ju f1,2 (0) W1,2
   
[3] [1] [5] [5] [2] [4] [4]
+ b31 Ju f0,3 (0) W1,2 + 2 F2,0,0 ϕ1 , 2 W0 + W2 + 2 F2.0,0 W1 , 2 W0 + W2
   
[3] [3] [3] [4] [2] [2]
+ 2 F2,0,0 W1 , 2 W0 + W2 + 2 F2,0,0 W1 , 2 W0 + W2
   
[1] [4] [4] [1] [4] [4]
+ 2 a1 F2,1,0 ϕ1 , 2 W0 + W2 + 2 b1 F2,0,1 ϕ1 , 2 W0 + W2
   
[2] [3] [3] [2] [3] [3]
+ 2 a1 F2,1,0 W1 , 2 W0 + W2 + 2 b1 F2,0,1 W1 , 2 W0 + W2
   
[3] [2] [2] [3] [2] [2]
+ 2 a1 F2,1,0 W1 , 2 W0 + W2 + 2 b1 F2,0,1 W1 , 2 W0 + W2
   
[1] [3] [3] [1] [3] [3]
+ 2 a2 F2,1,0 ϕ1 , 2 W0 + W2 + 2 b2 F2,0,1 ϕ1 , 2 W0 + W2
   
[2] [2] [2] [2] [2] [2]
+ 2 a2 F2,1,0 W1 , 2 W0 + W2 + 2 b2 F2,0,1 W1 , 2 W0 + W2
   
[1] [2] [2] [1] [2] [2]
+ 2 a3 F2,1,0 ϕ1 , 2 W0 + W2 + 2 b3 F2,0,1 ϕ1 , 2 W0 + W2
   
[1] [3] [3] [1] [3] [3]
+ 2 a21 F2,2,0 ϕ1 , 2 W0 + W2 + 2 a1 b1 F2,1,1 ϕ1 , 2 W0 + W2
   
[1] [3] [3] [2] [2] [2]
+ 2 b21 F2,0,2 ϕ1 , 2 W0 + W2 + 2 a21 F2,2,0 W1 , 2 W0 + W2
   
[2] [2] [2] [2] [2] [2]
+ 2 a1 b1 F2,1,1 W1 , 2 W0 + W2 + 2 b21 F2,0,2 W1 , 2 W0 + W2
   
[1] [2] [2] [1] [2] [2]
+ 4 a1 a2 F2,2,0 ϕ1 , 2 W0 + W2 + 2 (a1 b2 + a2 b1 ) F2,1,1 ϕ1 , 2 W0 + W2
   
[1] [2] [2] [1] [2] [2]
+ 4 b1 b2 F2,0,2 ϕ1 , 2 W0 + W2 + 2 a31 F2,3,0 ϕ1 , 2 W0 + W2
   
[1] [2] [2] [1] [2] [2]
+ 2 a21 b1 F2,2,1 ϕ1 , 2 W0 + W2 + 2 a1 b21 F2,1,2 ϕ1 , 2 W0 + W2
   
[1] [2] [2] [1] [1] [4]
+ 2 b31 F2,0,3 ϕ1 , 2 W0 + W2 + 9 F3,0,0 ϕ1 , ϕ1 , W1
   
[1] [2] [3] [2] [2] [2]
+ 18 F3,0,0 ϕ1 , W1 , W1 + 3 F3,0,0 W1 , W1 , W1
   
[1] [1] [3] [1] [1] [3]
+ 9 a1 F3,1,0 ϕ1 , ϕ1 , W1 + 9 b1 F3,0,1 ϕ1 , ϕ1 , W1
   
[1] [2] [2] [1] [2] [2]
+ 9 a1 F3,1,0 ϕ1 , W1 , W1 + 9 b1 F3,0,1 ϕ1 , W1 , W1
   
[1] [1] [2] [1] [1] [2]
+ 9 a2 F3,1,0 ϕ1 , ϕ1 , W1 + 9 b2 F3,0,1 ϕ1 , ϕ1 , W1
   
[1] [1] [1] [1] [1] [1]
+ 3 a3 F3,1,0 ϕ1 , ϕ1 , ϕ1 + 3 b3 F3,0,1 ϕ1 , ϕ1 , ϕ1
   
[1] [1] [2] [1] [1] [2]
+ 9 a21 F3,2,0 ϕ1 , ϕ1 , W1 + 9 a1 b1 F3,1,1 ϕ1 , ϕ1 , W1
   
[1] [1] [2] [1] [1] [1]
+ 9 b21 F3,0,2 ϕ1 , ϕ1 , W1 + 6 a1 a2 F3,2,0 ϕ1 , ϕ1 , ϕ1
   
[1] [1] [1] [1] [1] [1]
+ 3 (a1 b2 + a2 b1 ) F3,1,1 ϕ1 , ϕ1 , ϕ1 + 6 b1 b2 F3,0,2 ϕ1 , ϕ1 , ϕ1
   
[1] [1] [1] [1] [1] [1]
+ 3 a31 F3,0,3 ϕ1 , ϕ1 , ϕ1 + 3 a21 b1 F3,2,1 ϕ1 , ϕ1 , ϕ1

81
    
[1] [1] [1] [1] [1] [1]
+ 3 a1 b21 F3,1,2 ϕ1 , ϕ1 , ϕ1 + 3 b31 F3,0,3 ϕ1 , ϕ1 , ϕ1 ,
 
[6] [5] [5] [1] [5] [5]
Q7 = a1 Ju f1,0 (0) W1,7 + b1 Ju f0,1 (0) W1,7 + 2 F2,0,0 ϕ1 , 2 W0,2 + W2,2
     
[2] [4] [4] [2] [4] [3] [3] [3]
+ 2 F2,0,0 W1 , 2 W0,2 + W2,2 + 2 F2,0,0 W2 , W3 + 2 F2,0,0 W1,2 , 2 W0 + W2
   
[3] [3] [4] [2] [2]
+ 2 F2,0,0 W2 , W3 + 2 F2,0,0 W1,2 , 2 W0 + W2
   
[1] [4] [4] [1] [4] [4]
+ 2 a1 F2,1,0 ϕ1 , 2 W0,2 + W2,2 + 2 b1 F2,0,1 ϕ1 , 2 W0,2 + W2,2
   
[2] [3] [2] [3]
+ 2 a1 F2,1,0 W2 , W3 + 2 b1 F2,0,1 W2 , W3
   
[3] [2] [2] [3] [2] [2]
+ 2 a1 F2,1,0 W1,2 , 2 W0 + W2 + 2 b1 F2,0,1 W1,2 , 2 W0 + W2
   
[1] [1] [4] [4] [1] [2] [3] [3]
+ 3 F3,0,0 ϕ1 , ϕ1 , 3 W1,2 + W3 + 6 F3,0,0 ϕ1 , W1 , 3 W1,2 + W3
   
[1] [3] [2] [2] [1] [3] [2] [2]
+ 12 F3,0,0 ϕ1 , W0 , 2 W0 + W2 + 12 F3,0,0 ϕ1 , W2 , W0 + W2
   
[2] [2] [2] [2] [2] [2] [2]
+ 12 F3,0,0 W1 , W0 , W0 + W2 + 6 F3,0,0 W1 , W2 , W2
   
[1] [1] [3] [3] [1] [1] [3] [3]
+ 3 a1 F3,1,0 ϕ1 , ϕ1 , 3 W1,2 + W3 + 3 b1 F3,0,1 ϕ1 , ϕ1 , 3 W1,2 + W3
   
[1] [2] [2] [1] [2] [2]
+ 12 a1 F3,1,0 ϕ1 , W0 , W0 + 12 b1 F3,0,1 ϕ1 , W0 , W0
   
[1] [2] [2] [2] [1] [2] [2] [2]
+ 6 a1 F3,1,0 ϕ1 , W2 , 2 W0 + W2 + 6 b1 F3,0,1 ϕ1 , W2 , 2 W0 + W2
   
[1] [1] [1] [3] [3] [1] [1] [2] [2] [2]
+ 8 F4,0,0 ϕ1 , ϕ1 , ϕ1 , 3 W0 + 2 W2 + 24 F4,0,0 ϕ1 , ϕ1 , W1 , 3 W0 + 2 W2
   
[1] [1] [1] [2] [2] [1] [1] [1] [2] [2]
+ 8 a1 F4,1,0 ϕ1 , ϕ1 , ϕ1 , 3 W0 + 2 W2 + 8 b1 F4,0,1 ϕ1 , ϕ1 , ϕ1 , 3 W0 + 2 W2
   
[1] [1] [1] [1] [2] [1] [1] [1] [1] [1]
+ 50 F5,0,0 ϕ1 , ϕ1 , ϕ1 , ϕ1 , W1 + 10 a1 F5,1,0 ϕ1 , ϕ1 , ϕ1 , ϕ1 , ϕ1
 
[1] [1] [1] [1] [1]
+ 10 b1 F5,0,1 ϕ1 , ϕ1 , ϕ1 , ϕ1 , ϕ1 .

Here, we highlight that αp,2 = 0 for all p = 1, . . . , 7 when a2 q+1 = b2 q+1 = 0 for q = 0, 1, 2, which implies
that A2 = 0. If this was not the case, then the change A2 = (R2 + i R1 φ2 ) ei φ1 can be used to solve (103).
To simplify calculations and close up with key clarifications, we assume that A2 = 0.
After finding a solution for (103), using the same idea developed after Equation (102), we note that
[6] [6] α1,2
M1 W1 = −Q1 + ψ,
⟨ψ, ψ⟩

[6] αp,2
M1 W1,p = −Q[6]
p + ψ for p = 2, 3, 4, 5, 6, 7,
⟨ψ, ψ⟩

82
Moreover,
[6] [5] [5] [4]
M2 W2 = −a1 Ju f1,0 (0) W2 − b1 Ju f0,1 (0) W2 − a2 Ju f1,0 (0) W2
[4] [3] [3] [2]
− b2 Ju f0,1 (0) W2 − a3 Ju f1,0 (0) W2 − b3 Ju f0,1 (0) W2 − a4 Ju f1,0 (0) W2
[2] [4] [4] [4]
− b4 Ju f0,1 (0) W2 − a21 Ju f2,0 (0) W2 − a1 b1 Ju f1,1 (0) W2 − b21 Ju f0,2 (0) W2
[3] [3] [3]
− 2 a1 a2 Ju f2,0 (0) W2 − (a1 b2 + a2 b1 ) Ju f1,1 (0) W2 − 2 b1 b2 Ju f0,2 (0) W2
[2] [2] [2]
− 2 a1 a3 Ju f2,0 (0) W2 − (a1 b3 + a3 b1 ) Ju f1,1 (0) W2 − 2 b1 b3 Ju f0,2 (0) W2
[2] [2] [2] [3]
− a22 Ju f2,0 (0) W2 − a2 b2 Ju f1,1 (0) W2 − b22 Ju f0,2 (0) W2 − a31 Ju f3,0 (0) W2
[3] [3] [3] [2]
− a21 b1 Ju f2,1 (0) W2 − a1 b21 Ju f1,2 (0) W2 − b31 Ju f0,3 (0) W2 − 3 a21 a2 Ju f3,0 (0) W2
[2] [2] [2]
− 2 a1 a2 b1 Ju f2,1 (0) W2 − a2 b21 Ju f1,2 (0) W2 − a21 b2 Ju f2,1 (0) W2
[2] [2] [2] [2]
− 2 a1 b1 b2 Ju f1,2 (0) W2 − 3 b21 b2 Ju f0,3 (0) W2 − a41 Ju f4,0 (0) W2 − a31 b1 Ju f3,1 (0) W2
 
[2] [2] [2] [1] [5]
− a21 b21 Ju f2,2 (0) W2 − a1 b31 Ju f1,3 (0) W2 − b41 Ju f0,4 (0) W2 − 2 F2,0,0 ϕ1 , W1,5
     
[2] [4] [3] [3] [1] [4]
− 2 F2,0,0 W1 , W1 − F2,0,0 W1 , W1 − 2 a1 F2,1,0 ϕ1 , W1
     
[1] [4] [2] [3] [2] [3]
− 2 b1 F2,0,1 ϕ1 , W1 − 2 a1 F2,1,0 W1 , W1 − 2 b1 F2,0,1 W1 , W1
     
[1] [3] [1] [3] [2] [2]
− 2 a2 F2,1,0 ϕ1 , W1 − 2 b2 F2,0,1 ϕ1 , W1 − a2 F2,1,0 W1 , W1
     
[2] [2] [1] [2] [1] [2]
− b2 F2,0,1 W1 , W1 − 2 a3 F2,1,0 ϕ1 , W1 − 2 b3 F2,0,1 ϕ1 , W1
     
[1] [1] [1] [1] [1] [3]
− a4 F2,1,0 ϕ1 , ϕ1 − b4 F2,0,1 ϕ1 , ϕ1 − 2 a21 F2,2,0 ϕ1 , W1
     
[1] [3] [1] [3] [2] [2]
− 2 a1 b1 F2,1,1 ϕ1 , W1 − 2 b21 F2,0,2 ϕ1 , W1 − a21 F2,2,0 W1 , W1
     
[2] [2] [2] [2] [1] [2]
− a1 b1 F2,1,1 W1 , W1 − b21 F2,0,2 W1 , W1 − 4 a1 a2 F2,2,0 ϕ1 , W1
     
[1] [2] [1] [2] [1] [1]
− 2 (a1 b2 + a2 b1 ) F2,1,1 ϕ1 , W1 − 4 b1 b2 F2,0,2 ϕ1 , W1 − 2 a1 a3 F2,2,0 ϕ1 , ϕ1
     
[1] [1] [1] [1] [1] [1]
− (a1 b3 + a3 b1 ) F2,1,1 ϕ1 , ϕ1 − 2 b1 b3 F2,0,2 ϕ1 , ϕ1 − a22 F2,2,0 ϕ1 , ϕ1
     
[1] [1] [1] [1] [1] [2]
− a2 b2 F2,1,1 ϕ1 , ϕ1 − b22 F2,0,2 ϕ1 , ϕ1 − 2 a31 F2,3,0 ϕ1 , W1
     
[1] [2] [1] [2] [1] [2]
− 2 a21 b1 F2,2,1 ϕ1 , W1 − 2 a1 b21 F2,1,2 ϕ1 , W1 − 2 b31 F2,0,3 ϕ1 , W1
     
[1] [1] [1] [1] [1] [1]
− 3 a21 a2 F2,3,0 ϕ1 , ϕ1 − 2 a1 a2 b1 F2,2,1 ϕ1 , ϕ1 − a2 b21 F2,1,2 ϕ1 , ϕ1
     
[1] [1] [1] [1] [1] [1]
− a21 b2 F2,2,1 ϕ1 , ϕ1 − 2 a1 b1 b2 F2,1,2 ϕ1 , ϕ1 − 3 b21 b2 F2,0,3 ϕ1 , ϕ1
     
[1] [1] [1] [1] [1] [1]
− a41 F2,4,0 ϕ1 , ϕ1 − a31 b1 F2,3,1 ϕ1 , ϕ1 − a21 b21 F2,2,2 ϕ1 , ϕ1
   
[1] [1] [1] [1]
− a1 b31 F2,1,3 ϕ1 , ϕ1 − b41 F2,0,4 ϕ1 , ϕ1 ,

83
 [6] [5] [5] [4] [4]
M2 W2,2 = −a1 Ju f1,0 (0) W2,2 − b1 Ju f0,1 (0) W2,2 − a2 Ju f1,0 (0) W2,2 − b2 Ju f0,1 (0) W2,2
 
[4] [4] [4] [1] [5] [5]
− a21 Ju f2,0 (0) W2,2 − a1 b1 Ju f1,1 (0) W2,2 − b21 Ju f0,2 (0) W2,2 − 2 F2,0,0 ϕ1 , W1,6 + W3
     
[2] [4] [2] [4] [4] [2] [4]
− 4 F2,0,0 W0 , W2 − 2 F2,0,0 W1 , W1,2 + W3 − 4 F2,0,0 W2 , W0
     
[3] [3] [3] [3] [3] [1] [4] [4]
− 4 F2,0,0 W0 , W2 − 2 F2,0,0 W1 , W1,2 + W3 − 2 a1 F2,1,0 ϕ1 , W1,2 + W3
     
[1] [4] [4] [2] [3] [2] [3]
− 2 b1 F2,0,1 ϕ1 , W1,2 + W3 − 4 a1 F2,1,0 W0 , W2 − 4 b1 F2,0,1 W0 , W2
     
[2] [3] [3] [2] [3] [3] [2] [3]
− 2 a1 F2,1,0 W1 , W1,2 + W3 − 2 b1 F2,0,1 W1 , W1,2 + W3 − 4 a1 F2,1,0 W2 , W0
     
[2] [3] [1] [3] [3] [1] [3] [3]
− 4 b1 F2,0,1 W2 , W0 − 2 a2 F2,1,0 ϕ1 , W1,2 + W3 − 2 b2 F2,0,1 ϕ1 , W1,2 + W3
     
[2] [2] [2] [2] [1] [3] [3]
− 4 a2 F2,1,0 W0 , W2 − 4 b2 F2,0,1 W0 , W2 − 2 a21 F2,2,0 ϕ1 , W1,2 + W3
   
[1] [3] [3] [1] [3] [3]
− 2 a1 b1 F2,1,1 ϕ1 , W1,2 + W3 − 2 b21 F2,0,2 ϕ1 , W1,2 + W3
   
[2] [2] [2] [2]
− 4 a21 F2,2,0 W0 , W2 − 4 a1 b1 F2,1,1 W0 , W2
   
[2] [2] [1] [1] [4] [4]
− 4 b21 F2,0,2 W0 , W2 − 6 F3,0,0 ϕ1 , ϕ1 , W0 + W2
   
[1] [2] [3] [3] [1] [3] [2] [2]
− 12 F3,0,0 ϕ1 , W1 , W0 + W2 − 12 F3,0,0 ϕ1 , W1 , W0 + W2
   
[2] [2] [2] [2] [1] [1] [3] [3]
− 6 F3,0,0 W1 , W1 , W0 + W2 − 6 a1 F3,1,0 ϕ1 , ϕ1 , W0 + W2
   
[1] [1] [3] [3] [1] [2] [2] [2]
− 6 b1 F3,0,1 ϕ1 , ϕ1 , W0 + W2 − 12 a1 F3,1,0 ϕ1 , W1 , W0 + W2
   
[1] [2] [2] [2] [1] [1] [2] [2]
− 12 b1 F3,0,1 ϕ1 , W1 , W0 + W2 − 6 a2 F3,1,0 ϕ1 , ϕ1 , W0 + W2
   
[1] [1] [2] [2] [1] [1] [2] [2]
− 6 b2 F3,0,1 ϕ1 , ϕ1 , W0 + W2 − 6 a21 F3,2,0 ϕ1 , ϕ1 , W0 + W2
   
[1] [1] [2] [2] [1] [1] [2] [2]
− 6 a1 b1 F3,1,1 ϕ1 , ϕ1 , W0 + W2 − 6 b21 F3,0,2 ϕ1 , ϕ1 , W0 + W2
   
[1] [1] [1] [3] [1] [1] [2] [2]
− 16 F4,0,0 ϕ1 , ϕ1 , ϕ1 , W1 − 24 F4,0,0 ϕ1 , ϕ1 , W1 , W1
   
[1] [1] [1] [2] [1] [1] [1] [2]
− 16 a1 F4,1,0 ϕ1 , ϕ1 , ϕ1 , W1 − 16 b1 F4,0,1 ϕ1 , ϕ1 , ϕ1 , W1
   
[1] [1] [1] [1] [1] [1] [1] [1]
− 4 a2 F4,1,0 ϕ1 , ϕ1 , ϕ1 , ϕ1 − 4 b2 F4,0,1 ϕ1 , ϕ1 , ϕ1 , ϕ1
   
[1] [1] [1] [1] [1] [1] [1] [1]
− 4 a21 F4,2,0 ϕ1 , ϕ1 , ϕ1 , ϕ1 − 4 a1 b1 F4,1,1 ϕ1 , ϕ1 , ϕ1 , ϕ1
 
[1] [1] [1] [1]
− 4 b21 F4,0,2 ϕ1 , ϕ1 , ϕ1 , ϕ1 ,

   
[6] [1] [5] [5] [2] [4]
M2 W2,3 = −2 F2,0,0 ϕ1 , W1,7 + W3,2 − 4 F2,0,0 W0 , W2,2
     
[2] [4] [4] [3] [3] [3] [3]
− 2 F2,0,0 W2 , 2 W0,2 + W4 − F2,0,0 W1,2 , W1,2 − 2 F2,0,0 W1,3 , W3
   
[1] [1] [4] [4] [4] [1] [2] [3] [3]
− 3 F3,0,0 ϕ1 , ϕ1 , 2 W0,2 + 2 W2,2 + W4 − 12 F3,0,0 ϕ1 , W0 , W1,2 + W3
   
[1] [2] [3] [3] [2] [2] [2]
− 6 F3,0,0 ϕ1 , W2 , 2 W1,2 + W3 − 12 F3,0,0 W0 , W0 , W2
   
[2] [2] [2] [1] [1] [1] [3] [3]
− 3 F3,0,0 W2 , W2 , W2 − 4 F4,0,0 ϕ1 , ϕ1 , ϕ1 , 4 W1,2 + 3 W3
   
[1] [1] [2] [2] [1] [1] [2] [2] [2]
− 24 F4,0,0 ϕ1 , ϕ1 , W0 , W0 − 6 F4,0,0 ϕ1 , ϕ1 , W2 , 8 W0 + 3 W2
   
[1] [1] [1] [1] [2] [2] [1] [1] [1] [1] [1] [1]
− 5 F5,0,0 ϕ1 , ϕ1 , ϕ1 , ϕ1 , 8 W0 + 7 W2 − 15 F6,0,0 ϕ1 , ϕ1 , ϕ1 , ϕ1 , ϕ1 , ϕ1 ,

84
  
[6] [3] [3] [4] [2]
M2 W2,4 = F2,0,0 W1,3 , W1,3 + 4 k 2 D̂ W2,3 − 2 k 2 D̂ W2 ,

[6] [5] [5] [4]

M2 W2,5 = −a1 Ju f1,0 (0) W2,3 − b1 Ju f0,1 (0) W2,3 − a2 Ju f1,0 (0) W2,3
[4] [4] [4] [4]
− b2 Ju f0,1 (0) W2,3 − a21 Ju f2,0 (0) W2,3 − a1 b1 Ju f1,1 (0) W2,3 − b21 Ju f0,2 (0) W2,3
     
[1] [5] [2] [4] [3] [3]
− 2 F2,0,0 ϕ1 , W1,2 − 2 F2,0,0 W1 , W1,3 − 2 F2,0,0 W1 , W1,3
     
[1] [4] [1] [4] [2] [3]
− 2 a1 F2,1,0 ϕ1 , W1,3 − 2 b1 F2,0,1 ϕ1 , W1,3 − 2 a1 F2,1,0 W1 , W1,3
     
[2] [3] [1] [3] [1] [3]
− 2 b1 F2,0,1 W1 , W1,3 − 2 a2 F2,1,0 ϕ1 , W1,3 − 2 b2 F2,0,1 ϕ1 , W1,3
     
[1] [3] [1] [3] [1] [3] [4]
− 2 a21 F2,2,0 ϕ1 , W1,3 − 2 a1 b1 F2,1,1 ϕ1 , W1,3 − 2 b21 F2,0,2 ϕ1 , W1,3 − 8 k 2 D̂ W2 ,

     
[6] [1] [5] [5] [2] [4] [2] [4]
M2 W2,6 = −2 F2,0,0 ϕ1 , W1,3 + W3,3 − 4 F2,0,0 W0 , W2,3 − 2 F2,0,0 W2 , W0,3
   
[3] [3] [1] [1] [4] [4]
− 2 F2,0,0 W1,2 , W1,3 − 3 F3,0,0 ϕ1 , ϕ1 , W0,3 + 2 W2,3
   
[1] [3] [2] [2] [1] [1] [1] [3] [4]
− 6 F3,0,0 ϕ1 , W1,3 , 2 W0 + W2 − 12 F4,0,0 ϕ1 , ϕ1 , ϕ1 , W1,3 − 12 k 2 D̂ W2,2 ,

     
[6] [1] [5] [2] [4] [3] [3]
M2 W2,7 = −2 F2,0,0 ϕ1 , W1,4 + 2 F2,0,0 W2 , W0,3 + 2 F2,0,0 W1,3 , W3
   
[1] [1] [4] [1] [2] [3]
+ 3 F3,0,0 ϕ1 , ϕ1 , W0,3 + 6 F3,0,0 ϕ1 , W2 , W1,3
 
[1] [1] [1] [3] [4]
+ 4 F4,0,0 ϕ1 , ϕ1 , ϕ1 , W1,3 − 4 k 2 D̂ W2,2 ,

and
 
[6] [1] [5] [4] [2]
M2 W2,8 = −2 F2,0,0 ϕ1 , W1 + 4 k 2 D̂ W2,3 − 2 k 2 D̂ W2 .

Order O ε7 . Finally, at order seven, (4) becomes

0 = Ju f0,0 (0) u[7] + a1 Ju f1,0 (0) u[6] + b1 Ju f0,1 (0) u[6] + a2 Ju f1,0 (0) u[5] + b2 Ju f0,1 (0) u[5]
+ a3 Ju f1,0 (0) u[4] + b3 Ju f0,1 (0) u[4] + a4 Ju f1,0 (0) u[3] + b4 Ju f0,1 (0) u[3]
+ a5 Ju f1,0 (0) u[2] + b5 Ju f0,1 (0) u[2] + a6 Ju f1,0 (0) u[1] + b6 Ju f0,1 (0) u[1]
+ a21 Ju f2,0 (0) u[5] + a1 b1 Ju f1,1 (0) u[5] + b21 Ju f0,2 (0) u[5] + 2 a1 a2 Ju f2,0 (0) u[4]
+ (a1 b2 + a2 b1 ) Ju f1,1 (0) u[4] + 2 b1 b2 Ju f0,2 (0) u[4] + 2 a1 a3 Ju f2,0 (0) u[3]
+ (a1 b3 + a3 b1 ) Ju f1,1 (0) u[3] + 2 b1 b3 Ju f0,2 (0) u[3] + 2 a1 a4 Ju f2,0 (0) u[2]
+ (a1 b4 + a4 b1 ) Ju f1,1 (0) u[2] + 2 b1 b4 Ju f0,2 (0) u[2] + 2 a1 a5 Ju f2,0 (0) u[1]
+ (a1 b5 + a5 b1 ) Ju f1,1 (0) u[1] + 2 b1 b5 Ju f0,2 (0) u[1] + a22 Ju f2,0 (0) u[3]
+ a2 b2 Ju f1,1 (0) u[3] + b22 Ju f0,2 (0) u[3] + 2 a2 a3 Ju f2,0 (0) u[2] + (a2 b3 + a3 b2 ) Ju f1,1 (0) u[2]
+ 2 b2 b3 Ju f0,2 (0) u[2] + 2 a2 a4 Ju f2,0 (0) u[1] + (a2 b4 + a4 b2 ) Ju f1,1 (0) u[1] + 2 b2 b4 Ju f0,2 (0) u[1]
+ a23 Ju f2,0 (0) u[1] + a3 b3 Ju f1,1 (0) u[1] + b23 Ju f0,2 (0) u[1] + a31 Ju f3,0 (0) u[4]
+ a21 b1 Ju f2,1 (0) u[4] + a1 b21 Ju f1,2 (0) u[4] + b31 Ju f0,3 (0) u[4] + 3 a21 a2 Ju f3,0 (0) u[3]
+ 2 a1 a2 b1 Ju f2,1 (0) u[3] + a2 b21 Ju f1,2 (0) u[3] + a21 b2 Ju f2,1 (0) u[3] + 2 a1 b1 b2 Ju f1,2 (0) u[3]
+ 3 b21 b2 Ju f0,3 (0) u[3] + 3 a21 a3 Ju f3,0 (0) u[2] + 2 a1 a3 b1 Ju f2,1 (0) u[2] + a3 b21 Ju f1,2 (0) u[2]
+ a21 b3 Ju f2,1 (0) u[2] + 2 a1 b1 b3 Ju f1,2 (0) u[2] + 3 b21 b3 Ju f0,3 (0) u[2] + 3 a21 a4 Ju f3,0 (0) u[1]

85
+ 2 a1 a4 b1 Ju f2,1 (0) u[1] + a4 b21 Ju f1,2 (0) u[1] + a21 b4 Ju f2,1 (0) u[1] + 2 a1 b1 b4 Ju f1,2 (0) u[1]
+ 3 b21 b4 Ju f0,3 (0) u[1] + 3 a1 a22 Ju f3,0 (0) u[2] + 2 a1 a2 b2 Ju f2,1 (0) u[2] + a1 b22 Ju f1,2 (0) u[2]
+ a22 b1 Ju f2,1 (0) u[2] + 2 a2 b1 b2 Ju f1,2 (0) u[2] + 3 b1 b22 Ju f0,3 (0) u[2] + a32 Ju f3,0 (0) u[1]
+ a22 b2 Ju f2,1 (0) u[1] + a2 b22 Ju f1,2 (0) u[1] + b32 Ju f0,3 (0) u[1] + 6 a1 a2 a3 Ju f3,0 (0) u[1]
+ 2 (a2 a3 b1 + a1 a3 b2 + a1 a2 b3 ) Ju f2,1 (0) u[1] + 2 (a1 b2 b3 + a2 b1 b3 + a3 b1 b2 ) Ju f1,2 (0) u[1]
+ 6 b1 b2 b3 Ju f0,3 (0) u[1] + a41 Ju f4,0 (0) u[3] + a31 b1 Ju f3,1 (0) u[3] + a21 b21 Ju f2,2 (0) u[3]
+ a1 b31 Ju f1,3 (0) u[3] + b41 Ju f0,4 (0) u[3] + 4 a31 a2 Ju f4,0 (0) u[2] + 3 a21 a2 b1 Ju f3,1 (0) u[2]
+ 2 a1 a2 b21 Ju f2,2 (0) u[2] + a2 b31 Ju f1,3 (0) u[2] + a31 b2 Ju f3,1 (0) u[2] + 2 a21 b1 b2 Ju f2,2 (0) u[2]
+ 3 a1 b21 b2 Ju f1,3 (0) u[2] + 4 b31 b2 Ju f0,4 (0) u[2] + 4 a31 a3 Ju f4,0 (0) u[1] + 3 a21 a3 b1 Ju f3,1 (0) u[1]
+ 2 a1 a3 b21 Ju f2,2 (0) u[1] + a3 b31 Ju f (0) u[1] + a31 b3 Ju f3,1 (0) u[1] + 2 a21 b1 b3 Ju f2,2 (0) u[1]
+ 3 a1 b21 b3 Ju f1,3 (0) u[1] + 4 b31 b3 Ju f0,4 (0) u[1] + 6 a21 a22 Ju f4,0 (0) u[1] + 3 a1 a22 b1 Ju f3,1 (0) u[1]
+ a22 b21 Ju f2,2 (0) u[1] + 3 a21 a2 b2 Ju f3,1 (0) u[1] + 4 a1 a2 b1 b2 Ju f2,2 (0) u[1] + 3 a2 b21 b2 Ju f1,3 (0) u[1]
+ a21 b22 Ju f2,2 (0) u[1] + 3 a1 b1 b22 Ju f1,3 (0) u[1] + 6 b21 b22 Ju f0,4 (0) u[1] + a51 Ju f5,0 (0) u[2]
+ a41 b1 Ju f4,1 (0) u[2] + a31 b21 Ju f3,2 (0) u[2] + a21 b31 Ju f2,3 (0) u[2] + a1 b41 Ju f1,4 (0) u[2]
+ b51 Ju f0,5 (0) u[2] + 5 a41 a2 Ju f5,0 (0) u[1] + 4 a31 a2 b1 Ju f4,1 (0) u[1] + 3 a21 a2 b21 Ju f3,2 (0) u[1]
+ 2 a1 a2 b31 Ju f2,3 (0) u[1] + a2 b41 Ju f1,4 (0) u[1] + a41 b2 Ju f4,1 (0) u[1] + 2 a31 b1 b2 Ju f3,2 (0) u[1]
+ 3 a21 b21 b2 Ju f2,3 (0) u[1] + 4 a1 b31 b2 Ju f1,4 (0) u[1] + 5 b41 b2 Ju f0,5 (0) u[1] + a61 Ju f6,0 (0) u[1]
+ a51 b1 Ju f5,1 (0) u[1] + a41 b21 Ju f4,2 (0) u[1] + a31 b31 Ju f3,3 (0) u[1] + a21 b41 Ju f2,4 (0) u[1]
   
+ a1 b51 Ju f1,5 (0) u[1] + b61 Ju f0,6 (0) u[1] + 2 F2,0,0 u[1] , u[6] + 2 F2,0,0 u[2] , u[5]
       
+ 2 F2,0,0 u[3] , u[4] + 2 a1 F2,1,0 u[1] , u[5] + 2 b1 F2,0,1 u[1] , u[5] + 2 a1 F2,1,0 u[2] , u[4]
       
+ 2 b1 F2,0,1 u[2] , u[4] + a1 F2,1,0 u[3] , u[3] + b1 F2,0,1 u[3] , u[3] + 2 a2 F2,1,0 u[1] , u[4]
       
+ 2 b2 F2,0,1 u[1] , u[4] + 2 a2 F2,1,0 u[2] , u[3] + 2 b2 F2,0,1 u[2] , u[3] + 2 a3 F2,1,0 u[1] , u[3]
       
+ 2 b3 F2,0,1 u[1] , u[3] + a3 F2,1,0 u[2] , u[2] + b3 F2,0,1 u[2] , u[2] + 2 a4 F2,1,0 u[1] , u[2]
       
+ 2 b4 F2,0,1 u[1] , u[2] + a5 F2,1,0 u[1] , u[1] + b5 F2,0,1 u[1] , u[1] + 2 a21 F2,2,0 u[1] , u[4]
       
+ 2 a1 b1 F2,1,1 u[1] , u[4] + 2 b21 F2,0,2 u[1] , u[4] + 2 a21 F2,2,0 u[2] , u[3] + 2 a1 b1 F2,1,1 u[2] , u[3]
     
+ 2 b21 F2,0,2 u[2] , u[3] + 4 a1 a2 F2,2,0 u[1] , u[3] + 2 (a1 b2 + a2 b1 ) F2,1,1 u[1] , u[3]
     
+ 4 b1 b2 F2,0,2 u[1] , u[3] + 2 a1 a2 F2,2,0 u[2] , u[2] + (a1 b2 + a2 b1 ) F2,1,1 u[2] , u[2]
     
+ 2 b1 b2 F2,0,2 u[2] , u[2] + 4 a1 a3 F2,2,0 u[1] , u[2] + 2 (a1 b3 + a3 b1 ) F2,1,1 u[1] , u[2]
     
+ 4 b1 b3 F2,0,2 u[1] , u[2] + 2 a1 a4 F2,2,0 u[1] , u[1] + (a1 b4 + a4 b1 ) F2,1,1 u[1] , u[1]
     
+ 2 b1 b4 F2,0,2 u[1] , u[1] + 2 a22 F2,2,0 u[1] , u[2] + 2 a2 b2 F2,1,1 u[1] , u[2]
     
+ 2 b22 F2,0,2 u[1] , u[2] + 2 a2 a3 F2,2,0 u[1] , u[1] + (a2 b3 + a3 b2 ) F2,1,1 u[1] , u[1]
     
+ 2 b2 b3 F2,0,2 u[1] , u[1] + 2 a31 F2,3,0 u[1] , u[3] + 2 a21 b1 F2,2,1 u[1] , u[3]
     
+ 2 a1 b21 F2,1,2 u[1] , u[3] + 2 b31 F2,0,3 u[1] , u[3] + a31 F2,3,0 u[2] , u[2]
       
+ a21 b1 F2,2,1 u[2] , u[2] + a1 b21 F2,1,2 u[2] , u[2] + b31 F2,0,3 u[2] , u[2] + 6 a21 a2 F2,3,0 u[1] , u[2]

86
      
+ 4 a1 a2 b1 F2,2,1 u[1] , u[2] + 2 a2 b21 F2,1,2 u[1] , u[2] + 2 a21 b2 F2,2,1 u[1] , u[2]
     
+ 4 a1 b1 b2 F2,1,2 u[1] , u[2] + 6 b21 b2 F2,0,3 u[1] , u[2] + 3 a21 a3 F2,3,0 u[1] , u[1]
     
+ 2 a1 a3 b1 F2,2,1 u[1] , u[1] + a3 b21 F2,1,2 u[1] , u[1] + a21 b3 F2,2,1 u[1] , u[1]
     
+ 2 a1 b1 b3 F2,1,2 u[1] , u[1] + 3 b21 b3 F2,0,3 u[1] , u[1] + 3 a1 a22 F2,3,0 u[1] , u[1]
     
+ 2 a1 a2 b2 F2,2,1 u[1] , u[1] + a1 b22 F2,1,2 u[1] , u[1] + a22 b1 F2,2,1 u[1] , u[1]
     
+ 2 a2 b1 b2 F2,1,2 u[1] , u[1] + 3 b1 b22 F2,0,3 u[1] , u[1] + 2 a41 F2,4,0 u[1] , u[2]
     
+ 2 a31 b1 F2,3,1 u[1] , u[2] + 2 a21 b21 F2,2,2 u[1] , u[2] + 2 a1 b31 F2,1,3 u[1] , u[2]
     
+ 2 b41 F2,0,4 u[1] , u[2] + 4 a31 a2 F2,4,0 u[1] , u[1] + 3 a21 a2 b1 F2,3,1 u[1] , u[1]
     
+ 2 a1 a2 b21 F2,2,2 u[1] , u[1] + a2 b31 F2,1,3 u[1] , u[1] + a31 b2 F2,3,1 u[1] , u[1]
     
+ 2 a21 b1 b2 F2,2,2 u[1] , u[1] + 3 a1 b21 b2 F2,1,3 u[1] , u[1] + 4 b31 b2 F2,0,4 u[1] , u[1]
     
+ a51 F2,5,0 u[1] , u[1] + a41 b1 F2,4,1 u[1] , u[1] + a31 b21 F2,3,2 u[1] , u[1]
     
+ a21 b31 F2,2,3 u[1] , u[1] + a1 b41 F2,1,4 u[1] , u[1] + b51 F2,0,5 u[1] , u[1]
     
+ 3 F3,0,0 u[1] , u[1] , u[5] + 6 F3,0,0 u[1] , u[2] , u[4] + 3 F3,0,0 u[1] , u[3] , u[3]
     
+ 3 F3,0,0 u[2] , u[2] , u[3] + 3 a1 F3,1,0 u[1] , u[1] , u[4] + 3 b1 F3,0,1 u[1] , u[1] , u[4]
     
+ 6 a1 F3,1,0 u[1] , u[2] , u[3] + 6 b1 F3,0,1 u[1] , u[2] , u[3] + a1 F3,1,0 u[2] , u[2] , u[2]
     
+ b1 F3,0,1 u[2] , u[2] , u[2] + 3 a2 F3,1,0 u[1] , u[1] , u[3] + 3 b2 F3,0,1 u[1] , u[1] , u[3]
     
+ 3 a2 F3,1,0 u[1] , u[2] , u[2] + 3 b2 F3,0,1 u[1] , u[2] , u[2] + 3 a3 F3,1,0 u[1] , u[1] , u[2]
     
+ 3 b3 F3,0,1 u[1] , u[1] , u[2] + a4 F3,1,0 u[1] , u[1] , u[1] + b4 F3,0,1 u[1] , u[1] , u[1]
     
+ 3 a21 F3,2,0 u[1] , u[1] , u[3] + 3 a1 b1 F3,1,1 u[1] , u[1] , u[3] + 3 b21 F3,0,2 u[1] , u[1] , u[3]
     
+ 3 a21 F3,2,0 u[1] , u[2] , u[2] + 3 a1 b1 F3,1,1 u[1] , u[2] , u[2] + 3 b21 F3,0,2 u[1] , u[2] , u[2]
   
+ 6 a1 a2 F3,2,0 u[1] , u[1] , u[2] + 3 (a1 b2 + a2 b1 ) F3,1,1 u[1] , u[1] , u[2]
     
+ 6 b1 b2 F3,0,2 u[1] , u[1] , u[2] + 2 a1 a3 F3,2,0 u[1] , u[1] , u[1] + (a1 b3 + a3 b1 ) F3,1,1 u[1] , u[1] , u[1]
     
+ 2 b1 b3 F3,0,2 u[1] , u[1] , u[1] + a22 F3,2,0 u[1] , u[1] , u[1] + a2 b2 F3,1,1 u[1] , u[1] , u[1]
     
+ b22 F3,0,2 u[1] , u[1] , u[1] + 3 a31 F3,3,0 u[1] , u[1] , u[2] + 3 a21 b1 F3,2,1 u[1] , u[1] , u[2]
     
+ 3 a1 b21 F3,1,2 u[1] , u[1] , u[2] + 3 b31 F3,0,3 u[1] , u[1] , u[2] + 3 a21 a2 F3,3,0 u[1] , u[1] , u[1]
     
+ 2 a1 a2 b1 F3,2,1 u[1] , u[1] , u[1] + a2 b21 F3,1,2 u[1] , u[1] , u[1] + a21 b2 F3,2,1 u[1] , u[1] , u[1]
     
+ 2 a1 b1 b2 F3,1,2 u[1] , u[1] , u[1] + 3 b21 b2 F3,0,3 u[1] , u[1] , u[1] + a41 F3,4,0 u[1] , u[1] , u[1]
     
+ a31 b1 F3,3,1 u[1] , u[1] , u[1] + a21 b21 F3,2,2 u[1] , u[1] , u[1] + a1 b31 F3,1,3 u[1] , u[1] , u[1]
     
+ b41 F3,0,4 u[1] , u[1] , u[1] + 4 F4,0,0 u[1] , u[1] , u[1] , u[4] + 12 F4,0,0 u[1] , u[1] , u[2] , u[3]

87
      
+ 4 F4,0,0 u[1] , u[2] , u[2] , u[2] + 4 a1 F4,1,0 u[1] , u[1] , u[1] , u[3] + 4 b1 F4,0,1 u[1] , u[1] , u[1] , u[3]
     
+ 6 a1 F4,1,0 u[1] , u[1] , u[2] , u[2] + 6 b1 F4,0,1 u[1] , u[1] , u[2] , u[2] + 4 a2 F4,1,0 u[1] , u[1] , u[1] , u[2]
     
+ 4 b2 F4,0,1 u[1] , u[1] , u[1] , u[2] + a3 F4,1,0 u[1] , u[1] , u[1] , u[1] + b3 F4,0,1 u[1] , u[1] , u[1] , u[1]
     
+ 4 a21 F4,2,0 u[1] , u[1] , u[1] , u[2] + 4 a1 b1 F4,1,1 u[1] , u[1] , u[1] , u[2] + 4 b21 F4,0,2 u[1] , u[1] , u[1] , u[2]
   
+ 2 a1 a2 F4,2,0 u[1] , u[1] , u[1] , u[1] + (a1 b2 + a2 b1 ) F4,1,1 u[1] , u[1] , u[1] , u[1]
   
+ 2 b1 b2 F4,0,2 u[1] , u[1] , u[1] , u[1] + a31 F4,3,0 u[1] , u[1] , u[1] , u[1]
     
+ a21 b1 F4,2,1 u[1] , u[1] , u[1] , u[1] + a1 b21 F4,1,2 u[1] , u[1] , u[1] , u[1] + b31 F4,0,3 u[1] , u[1] , u[1] , u[1]
   
+ 5 F5,0,0 u[1] , u[1] , u[1] , u[1] , u[3] + 10 F5,0,0 u[1] , u[1] , u[1] , u[2] , u[2]
   
+ 5 a1 F5,1,0 u[1] , u[1] , u[1] , u[1] , u[2] + 5 b1 F5,0,1 u[1] , u[1] , u[1] , u[1] , u[2]
   
+ a2 F5,1,0 u[1] , u[1] , u[1] , u[1] , u[1] + b2 F5,0,1 u[1] , u[1] , u[1] , u[1] , u[1]
   
+ a21 F5,2,0 u[1] , u[1] , u[1] , u[1] , u[1] + a1 b1 F5,1,1 u[1] , u[1] , u[1] , u[1] , u[1]
   
+ b21 F5,0,2 u[1] , u[1] , u[1] , u[1] , u[1] + 6 F6,0,0 u[1] , u[1] , u[1] , u[1] , u[1] , u[2]
   
+ a1 F6,1,0 u[1] , u[1] , u[1] , u[1] , u[1] , u[1] + b1 F6,0,1 u[1] , u[1] , u[1] , u[1] , u[1] , u[1]
 2 [7]
∂ 2 u[5] ∂ 2 u[3] ∂ 2 u[3] ∂ 2 u[1]


[1] [1] [1] [1] [1] [1] [1]

2 ∂ u
+ F7,0,0 u , u , u , u , u , u , u + k D̂ +2 +2 + +2 .
∂x2 ∂x∂X ∂x∂Ξ ∂X 2 ∂X∂Ξ
Here, we are only interested in the solvability condition coming out of this equation, which is given by
2
α1 A3XX + i α2 A3X + i α3 |A1 | A3X + i α3 A1 A1X Ā3 + i α3 Ā1 A1X A3 + i α4 A21 Ā3X + 2 i α4 A1 Ā1X A3
2 2 4
+ α5 A3 + α6 A21 Ā3 + 2 α6 |A1 | A3 + 2 α7 |A1 | A21 Ā3 + 3 α7 |A1 | A3
2
+ α1,2 A2XX + i α2,2 A2X + i α3,2 |A1 | A2X + i α3,2 A1 A1X Ā2 + i α3,2 Ā1 A1X A2 + i α4,2 A21 Ā2X
2 2 4
+ 2 i α4,2 A1 Ā1X A2 + α5,2 A2 + α6,2 A21 Ā2 + 2 α6,2 |A1 | A2 + 2 α7,2 |A1 | A21 Ā2 + 3 α7,2 |A1 | A2
2
+ i α3 Ā1 A2 A2X + i α3 A1 Ā2 A2X + i α3 A1X |A2 | + i α4 Ā1X A22 + 2 i α4 A1 A2 Ā2X
2 2 2 2
+ α6 Ā1 A22 + 2 α6 A1 |A2 | + α7 A31 Ā22 + 3 α7 |A1 | Ā1 A22 + 6 α7 |A1 | A1 |A2 |
2 2
+ i α1,3 A1XXX + α2,3 A1XX + α3,3 |A1 | A1XX + α4,3 A21 Ā1XX + i α5,3 A1X + i α6,3 |A1 | A1X
4 2 2 2
+ i α7,3 |A1 | A1X + i α8,3 A21 Ā1X + i α9,3 |A1 | A21 Ā1X + α10,3 Ā1 (A1X ) + α11,3 A1 |A1X |
2 4 6 2
+ α12,3 A1 + α13,3 |A1 | A1 + α14,3 |A1 | A1 + α15,3 |A1 | A1 + 2 α1 A1XΞ + i α2 A1Ξ + i α3 |A1 | A1Ξ
+ i α4 A21 Ā1Ξ = 0, (104)
D E
[7]
where αp,3 = Qp , ψ for p = 1, . . . , 15, with

[7] [5] [3]

Q1 = 2 k 2 D̂ W1 + k 2 D̂ W1,3 ,
[7] [6] [6] [5] [5] [5]
Q2 = a1 Ju f1,0 (0) W1 + b1 Ju f0,1 (0) W1 + a2 Ju f1,0 (0) W1 + b2 Ju f0,1 (0) W1 + a21 Ju f2,0 (0) W1
[5] [5] [5] [3]
+ a1 b1 Ju f1,1 (0) W1 + b21 Ju f0,2 (0) W1 − 2 k 2 D̂ W1,2 + k 2 D̂ W1 ,
     
[7] [1] [6] [6] [2] [5] [1] [1] [5] [5]
Q3 = 2 F2,0,0 ϕ1 , W0,7 + W2,8 + 4 F2,0,0 W0 , W1 + 6 F3,0,0 ϕ1 , ϕ1 , W1 − 2 k 2 D̂ W1,3
[3]
+ 2 k 2 D̂ W1,2 ,
     
[7] [1] [6] [2] [5] [1] [1] [5] [5] [3]
Q4 = 2 F2,0,0 ϕ1 , W0,7 + 2 F2,0,0 W2 , W1 + 3 F3,0,0 ϕ1 , ϕ1 , W1 − 2 k 2 D̂ W1,4 + k 2 D̂ W1,2 ,

88
 [7] [6] [6] [5] [5]
Q5 = a1 Ju f1,0 (0) W1,2 + b1 Ju f0,1 (0) W1,2 + a2 Ju f1,0 (0) W1,2 + b2 Ju f0,1 (0) W1,2
[4] [4] [3] [3]
+ a3 Ju f1,0 (0) W1,3 + b3 Ju f0,1 (0) W1,3 + a4 Ju f1,0 (0) W1,3 + b4 Ju f0,1 (0) W1,3
[5] [5] [5] [4]
+ a21 Ju f2,0 (0) W1,2 + a1 b1 Ju f1,1 (0) W1,2 + b21 Ju f0,2 (0) W1,2 + 2 a1 a2 Ju f2,0 (0) W1,3
[4] [4] [3]
+ (a1 b2 + a2 b1 ) Ju f1,1 (0) W1,3 + 2 b1 b2 Ju f0,2 (0) W1,3 + 2 a1 a3 Ju f2,0 (0) W1,3
[3] [3] [3]
+ (a1 b3 + a3 b1 ) Ju f1,1 (0) W1,3 + 2 b1 b3 Ju f0,2 (0) W1,3 + a22 Ju f2,0 (0) W1,3
[3] [3] [4] [4]
+ a2 b2 Ju f1,1 (0) W1,3 + b22 Ju f0,2 (0) W1,3 + a31 Ju f3,0 (0) W1,3 + a21 b1 Ju f2,1 (0) W1,3
[4] [4] [3] [3]
+ a1 b21 Ju f1,2 (0) W1,3 + b31 Ju f0,3 (0) W1,3 + 3 a21 a2 Ju f3,0 (0) W1,3 + 2 a1 a2 b1 Ju f2,1 (0) W1,3
[3] [3] [3] [3]
+ a2 b21 Ju f1,2 (0) W1,3 + a21 b2 Ju f2,1 (0) W1,3 + 2 a1 b1 b2 Ju f1,2 (0) W1,3 + 3 b21 b2 Ju f0,3 (0) W1,3
[3] [3] [3] [3]
+ a41 Ju f4,0 (0) W1,3 + a31 b1 Ju f3,1 (0) W1,3 + a21 b21 Ju f2,2 (0) W1,3 + a1 b31 Ju f1,3 (0) W1,3
[3] [5]
+ b41 Ju f0,4 (0) W1,3 + 2 k 2 D̂ W1,5 ,
[7] [6] [6] [5] [5]
Q6 = a1 Ju f1,0 (0) W1,3 + b1 Ju f0,1 (0) W1,3 + a2 Ju f1,0 (0) W1,3 + b2 Ju f0,1 (0) W1,3
 
[5] [5] [5] [1] [6] [6]
+ a21 Ju f2,0 (0) W1,3 + a1 b1 Ju f1,1 (0) W1,3 + b21 Ju f0,2 (0) W1,3 + 2 F2,0,0 ϕ1 , W0,5 + W2,5
     
[2] [5] [2] [5] [5] [3] [4]
+ 4 F2,0,0 W0 , W1,2 + 2 F2,0,0 W1 , W0,3 + W2,3 + 4 F2,0,0 W0 , W1,3
     
[3] [4] [4] [3] [4] [1] [5] [5]
+ 2 F2,0,0 W1 , W0,3 + W2,3 + 4 F2,0,0 W1,3 , W0 + 2 a1 F2,1,0 ϕ1 , W0,3 + W2,3
     
[1] [5] [5] [2] [4] [2] [4]
+ 2 b1 F2,0,1 ϕ1 , W0,3 + W2,3 + 4 a1 F2,1,0 W0 , W1,3 + 4 b1 F2,0,1 W0 , W1,3
     
[2] [4] [4] [2] [4] [4] [3] [3]
+ 2 a1 F2,1,0 W1 , W0,3 + W2,3 + 2 b1 F2,0,1 W1 , W0,3 + W2,3 + 4 a1 F2,1,0 W0 , W1,3
     
[3] [3] [1] [4] [4] [1] [4] [4]
+ 4 b1 F2,0,1 W0 , W1,3 + 2 a2 F2,1,0 ϕ1 , W0,3 + W2,3 + 2 b2 F2,0,1 ϕ1 , W0,3 + W2,3
     
[2] [3] [2] [3] [1] [4] [4]
+ 4 a2 F2,1,0 W0 , W1,3 + 4 b2 F2,0,1 W0 , W1,3 + 2 a21 F2,2,0 ϕ1 , W0,3 + W2,3
     
[1] [4] [4] [1] [4] [4] [2] [3]
+ 2 a1 b1 F2,1,1 ϕ1 , W0,3 + W2,3 + 2 b21 F2,0,2 ϕ1 , W0,3 + W2,3 + 4 a21 F2,2,0 W0 , W1,3
     
[2] [3] [2] [3] [1] [1] [5]
+ 4 a1 b1 F2,1,1 W0 , W1,3 + 4 b21 F2,0,2 W0 , W1,3 + 6 F3,0,0 ϕ1 , ϕ1 , W1,2
     
[1] [2] [4] [1] [3] [3] [2] [2] [3]
+ 12 F3,0,0 ϕ1 , W1 , W1,3 + 12 F3,0,0 ϕ1 , W1 , W1,3 + 6 F3,0,0 W1 , W1 , W1,3
     
[1] [1] [4] [1] [1] [4] [1] [2] [3]
+ 6 a1 F3,1,0 ϕ1 , ϕ1 , W1,3 + 6 b1 F3,0,1 ϕ1 , ϕ1 , W1,3 + 12 a1 F3,1,0 ϕ1 , W1 , W1,3
     
[1] [2] [3] [1] [1] [3] [1] [1] [3]
+ 12 b1 F3,0,1 ϕ1 , W1 , W1,3 + 6 a2 F3,1,0 ϕ1 , ϕ1 , W1,3 + 6 b2 F3,0,1 ϕ1 , ϕ1 , W1,3
     
[1] [1] [3] [1] [1] [3] [1] [1] [3]
+ 6 a21 F3,2,0 ϕ1 , ϕ1 , W1,3 + 6 a1 b1 F3,1,1 ϕ1 , ϕ1 , W1,3 + 6 b21 F3,0,2 ϕ1 , ϕ1 , W1,3
[5]
+ 4 k 2 D̂ W1,6 ,
     
[7] [1] [6] [6] [2] [5] [2] [5] [5]
Q7 = 2 F2,0,0 ϕ1 , W0,6 + W2,6 + 4 F2,0,0 W0 , W1,3 − 2 F2,0,0 W2 , W1,4 − W3,3
   
[3] [4] [4] [3] [4]
+ 2 F2,0,0 W1,2 , W0,3 + W2,3 + 4 F2,0,0 W1,3 , W0,2
 
[1] [1] [5] [5] [5]
+ 3 F3,0,0 ϕ1 , ϕ1 , 2 W1,3 − W1,4 + W3,3
   
[1] [2] [2] [4] [4] [1] [3] [3]
+ 6 F3,0,0 ϕ1 , 2 W0 + W2 , W0,3 + W2,3 + 12 F3,0,0 ϕ1 , W1,2 , W1,3
   
[2] [2] [3] [2] [2] [3]
+ 12 F3,0,0 W0 , W0 , W1,3 + 6 F3,0,0 W2 , W2 , W1,3
   
[1] [1] [1] [4] [4] [1] [1] [3] [2] [2]
+ 12 F4,0,0 ϕ1 , ϕ1 , ϕ1 , W0,3 + W2,3 + 24 F4,0,0 ϕ1 , ϕ1 , W1,3 , 2 W0 + W2

89
  
[1] [1] [1] [1] [3] [5]
+ 30 F5,0,0 ϕ1 , ϕ1 , ϕ1 , ϕ1 , W1,3 + 6 k 2 D̂ W1,7 ,
[7] [6] [6] [5] [5]
Q8 = a1 Ju f1,0 (0) W1,4 + b1 Ju f0,1 (0) W1,4 + a2 Ju f1,0 (0) W1,4 + b2 Ju f0,1 (0) W1,4
 
[5] [5] [5] [1] [6]
+ a21 Ju f2,0 (0) W1,4 + a1 b1 Ju f1,1 (0) W1,4 + b21 Ju f0,2 (0) W1,4 − 2 F2,0,0 ϕ1 , W0,5
       
[2] [5] [2] [5] [3] [4] [3] [4]
− 2 F2,0,0 W1 , W0,3 − 2 F2,0,0 W2 , W1,2 − 2 F2,0,0 W1 , W0,3 − 2 F2,0,0 W2 , W1,3
     
[4] [3] [1] [5] [1] [5]
− 2 F2,0,0 W2 , W1,3 − 2 a1 F2,1,0 ϕ1 , W0,3 − 2 b1 F2,0,1 ϕ1 , W0,3
     
[2] [4] [2] [4] [2] [4]
− 2 a1 F2,1,0 W1 , W0,3 − 2 b1 F2,0,1 W1 , W0,3 − 2 a1 F2,1,0 W2 , W1,3
     
[2] [4] [3] [3] [3] [3]
− 2 b1 F2,0,1 W2 , W1,3 − 2 a1 F2,1,0 W1,3 , W2 − 2 b1 F2,0,1 W1,3 , W2
     
[1] [4] [1] [4] [2] [3]
− 2 a2 F2,1,0 ϕ1 , W0,3 − 2 b2 F2,0,1 ϕ1 , W0,3 − 2 a2 F2,1,0 W2 , W1,3
     
[2] [3] [1] [4] [1] [4]
− 2 b2 F2,0,1 W2 , W1,3 − 2 a21 F2,2,0 ϕ1 , W0,3 − 2 a1 b1 F2,1,1 ϕ1 , W0,3
     
[1] [4] [2] [3] [2] [3]
− 2 b21 F2,0,2 ϕ1 , W0,3 − 2 a21 F2,2,0 W2 , W1,3 − 2 a1 b1 F2,1,1 W2 , W1,3
     
[2] [3] [1] [1] [5] [1] [2] [4]
− 2 b21 F2,0,2 W2 , W1,3 − 3 F3,0,0 ϕ1 , ϕ1 , W1,2 − 6 F3,0,0 ϕ1 , W1 , W1,3
     
[1] [3] [3] [2] [2] [3] [1] [1] [4]
− 6 F3,0,0 ϕ1 , W1 , W1,3 − 3 F3,0,0 W1 , W1 , W1,3 − 3 a1 F3,1,0 ϕ1 , ϕ1 , W1,3
     
[1] [1] [4] [1] [2] [3] [1] [2] [3]
− 3 b1 F3,0,1 ϕ1 , ϕ1 , W1,3 − 6 a1 F3,1,0 ϕ1 , W1 , W1,3 − 6 b1 F3,0,1 ϕ1 , W1 , W1,3
     
[1] [1] [3] [1] [1] [3] [1] [1] [3]
− 3 a2 F3,1,0 ϕ1 , ϕ1 , W1,3 − 3 b2 F3,0,1 ϕ1 , ϕ1 , W1,3 − 3 a21 F3,2,0 ϕ1 , ϕ1 , W1,3
   
[1] [1] [3] [1] [1] [3] [5]
− 3 a1 b1 F3,1,1 ϕ1 , ϕ1 , W1,3 − 3 b21 F3,0,2 ϕ1 , ϕ1 , W1,3 + 2 k 2 D̂ W1,6 ,
     
[7] [1] [6] [6] [2] [5] [2] [5]
Q9 = −2 F2,0,0 ϕ1 , W0,6 − W2,7 + 4 F2,0,0 W0 , W1,4 − 2 F2,0,0 W2 , W1,3
     
[3] [4] [3] [4] [3] [4]
− 2 F2,0,0 W1,2 , W0,3 − 2 F2,0,0 W1,3 , W2,2 − 2 F2,0,0 W3 , W2,3
   
[1] [1] [5] [5] [1] [2] [4]
− 3 F3,0,0 ϕ1 , ϕ1 , W1,3 − 2 W1,4 − 12 F3,0,0 ϕ1 , W0 , W0,3
   
[1] [2] [4] [4] [1] [3] [3] [3]
− 6 F3,0,0 ϕ1 , W2 , W0,3 + W2,3 − 6 F3,0,0 ϕ1 , W1,3 , W1,2 + W3
   
[2] [2] [3] [1] [1] [1] [4] [4]
− 12 F3,0,0 W0 , W2 , W1,3 − 4 F4,0,0 ϕ1 , ϕ1 , ϕ1 , 3 W0,3 + W2,3
   
[1] [1] [3] [2] [2] [1] [1] [1] [1] [3]
− 24 F4,0,0 ϕ1 , ϕ1 , W1,3 , W0 + W2 − 20 F5,0,0 ϕ1 , ϕ1 , ϕ1 , ϕ1 , W1,3
[5]
+ 4 k 2 D̂ W1,7 ,
     
[7] [1] [6] [3] [4] [1] [3] [3]
Q10 = 2 F2,0,0 ϕ1 , W2,4 − 2 F2,0,0 W1,3 , W0,3 − 3 F3,0,0 ϕ1 , W1,3 , W1,3
[5] [3]
− 2 k 2 D̂ W1,3 + 2 k 2 D̂ W1,2 ,
     
[7] [1] [6] [3] [4] [4] [1] [3] [3]
Q11 = 4 F2,0,0 ϕ1 , W0,4 + 2 F2,0,0 W1,3 , W0,3 + W2,3 + 6 F3,0,0 ϕ1 , W1,3 , W1,3
[5] [5] [3]
− 2 k 2 D̂ W1,3 − 4 k 2 D̂ W1,4 + 4 k 2 D̂ W1,2
[7] [6] [6] [5] [5]
Q12 = a1 Ju f1,0 (0) W1,5 + b1 Ju f0,1 (0) W1,5 + a2 Ju f1,0 (0) W1,5 + b2 Ju f0,1 (0) W1,5
[4] [4] [3] [3]
+ a3 Ju f1,0 (0) W1 + b3 Ju f0,1 (0) W1 + a4 Ju f1,0 (0) W1 + b4 Ju f0,1 (0) W1
[2] [2] [1] [1]
+ a5 Ju f1,0 (0) W1 + b5 Ju f0,1 (0) W1 + a6 Ju f1,0 (0) ϕ1 + b6 Ju f0,1 (0) ϕ1
[5] [5] [5] [4]
+ a21 Ju f2,0 (0) W1,5 + a1 b1 Ju f1,1 (0) W1,5 + b21 Ju f0,2 (0) W1,5 + 2 a1 a2 Ju f2,0 (0) W1
[4] [4] [3]
+ (a1 b2 + a2 b1 ) Ju f1,1 (0) W1 + 2 b1 b2 Ju f0,2 (0) W1 + 2 a1 a3 Ju f2,0 (0) W1

90
 [3] [3] [2]
+ (a1 b3 + a3 b1 ) Ju f1,1 (0) W1 + 2 b1 b3 Ju f0,2 (0) W1 + 2 a1 a4 Ju f2,0 (0) W1
[2] [2] [1]
+ (a1 b4 + a4 b1 ) Ju f1,1 (0) W1 + 2 b1 b4 Ju f0,2 (0) W1 + 2 a1 a5 Ju f2,0 (0) ϕ1
[1] [1] [3]
+ (a1 b5 + a5 b1 ) Ju f1,1 (0) ϕ1 + 2 b1 b5 Ju f0,2 (0) ϕ1 + a22 Ju f2,0 (0) W1
[3] [3] [2] [2]
+ a2 b2 Ju f1,1 (0) W1 + b22 Ju f0,2 (0) W1 + 2 a2 a3 Ju f2,0 (0) W1 + (a2 b3 + a3 b2 ) Ju f1,1 (0) W1
[2] [1] [1]
+ 2 b2 b3 Ju f0,2 (0) W1 + 2 a2 a4 Ju f2,0 (0) ϕ1 + (a2 b4 + a4 b2 ) Ju f1,1 (0) ϕ1
[1] [1] [1] [1]
+ 2 b2 b4 Ju f0,2 (0) ϕ1 + a23 Ju f2,0 (0) ϕ1 + a3 b3 Ju f1,1 (0) ϕ1 + b23 Ju f0,2 (0) ϕ1
[4] [4] [4] [4]
+ a31 Ju f3,0 (0) W1 + a21 b1 Ju f2,1 (0) W1 + a1 b21 Ju f1,2 (0) W1 + b31 Ju f0,3 (0) W1
[3] [3] [3] [3]
+ 3 a21 a2 Ju f3,0 (0) W1 + 2 a1 a2 b1 Ju f2,1 (0) W1 + a2 b21 Ju f1,2 (0) W1 + a21 b2 Ju f2,1 (0) W1
[3] [3] [2] [2]
+ 2 a1 b1 b2 Ju f1,2 (0) W1 + 3 b21 b2 Ju f0,3 (0) W1 + 3 a21 a3 Ju f3,0 (0) W1 + 2 a1 a3 b1 Ju f2,1 (0) W1
[2] [2] [2] [2]
+ a3 b21 Ju f1,2 (0) W1 + a21 b3 Ju f2,1 (0) W1 + 2 a1 b1 b3 Ju f1,2 (0) W1 + 3 b21 b3 Ju f0,3 (0) W1
[1] [1] [1] [1]
+ 3 a21 a4 Ju f3,0 (0) ϕ1 + 2 a1 a4 b1 Ju f2,1 (0) ϕ1 + a4 b21 Ju f1,2 (0) ϕ1 + a21 b4 Ju f2,1 (0) ϕ1
[1] [1] [2] [2]
+ 2 a1 b1 b4 Ju f1,2 (0) ϕ1 + 3 b21 b4 Ju f0,3 (0) ϕ1 + 3 a1 a22 Ju f3,0 (0) W1 + 2 a1 a2 b2 Ju f2,1 (0) W1
[2] [2] [2] [2]
+ a1 b22 Ju f1,2 (0) W1 + a22 b1 Ju f2,1 (0) W1 + 2 a2 b1 b2 Ju f1,2 (0) W1 + 3 b1 b22 Ju f0,3 (0) W1
[1] [1] [1] [1]
+ a32 Ju f3,0 (0) ϕ1 + a22 b2 Ju f2,1 (0) ϕ1 + a2 b22 Ju f1,2 (0) ϕ1 + b32 Ju f0,3 (0) ϕ1
[1] [1]
+ 6 a1 a2 a3 Ju f3,0 (0) ϕ1 + 2 (a2 a3 b1 + a1 a3 b2 + a1 a2 b3 ) Ju f2,1 (0) ϕ1
[1] [1]
+ 2 (a1 b2 b3 + a2 b1 b3 + a3 b1 b2 ) Ju f1,2 (0) ϕ1 + 6 b1 b2 b3 Ju f0,3 (0) ϕ1
[3] [3] [3] [3]
+ a41 Ju f4,0 (0) W1 + a31 b1 Ju f3,1 (0) W1 + a21 b21 Ju f2,2 (0) W1 + a1 b31 Ju f1,3 (0) W1
[3] [2] [2] [2]
+ b41 Ju f0,4 (0) W1 + 4 a31 a2 Ju f4,0 (0) W1 + 3 a21 a2 b1 Ju f3,1 (0) W1 + 2 a1 a2 b21 Ju f2,2 (0) W1
[2] [2] [2] [2]
+ a2 b31 Ju f1,3 (0) W1 + a31 b2 Ju f3,1 (0) W1 + 2 a21 b1 b2 Ju f2,2 (0) W1 + 3 a1 b21 b2 Ju f1,3 (0) W1
[2] [1] [1] [1]
+ 4 b31 b2 Ju f0,4 (0) W1 + 4 a31 a3 Ju f4,0 (0) ϕ1 + 3 a21 a3 b1 Ju f3,1 (0) ϕ1 + 2 a1 a3 b21 Ju f2,2 (0) ϕ1
[1] [1] [1] [1]
+ a3 b31 Ju f1,3 (0) ϕ1 + a31 b3 Ju f3,1 (0) ϕ1 + 2 a21 b1 b3 Ju f2,2 (0) ϕ1 + 3 a1 b21 b3 Ju f1,3 (0) ϕ1
[1] [1] [1] [1]
+ 4 b31 b3 Ju f0,4 (0) ϕ1 + 6 a21 a22 Ju f4,0 (0) ϕ1 + 3 a1 a22 b1 Ju f3,1 (0) ϕ1 + a22 b21 Ju f2,2 (0) ϕ1
[1] [1] [1] [1]
+ 3 a21 a2 b2 Ju f3,1 (0) ϕ1 + 4 a1 a2 b1 b2 Ju f2,2 (0) ϕ1 + 3 a2 b21 b2 Ju f1,3 (0) ϕ1 + a21 b22 Ju f2,2 (0) ϕ1
[1] [1] [2] [2]
+ 3 a1 b1 b22 Ju f1,3 (0) ϕ1 + 6 b21 b22 Ju f0,4 (0) ϕ1 + a51 Ju f5,0 (0) W1 + a41 b1 Ju f4,1 (0) W1
[2] [2] [2] [2]
+ a31 b21 Ju f3,2 (0) W1 + a21 b31 Ju f2,3 (0) W1 + a1 b41 Ju f1,4 (0) W1 + b51 Ju f0,5 (0) W1
[1] [1] [1] [1]
+ 5 a41 a2 Ju f5,0 (0) ϕ1 + 4 a31 a2 b1 Ju f4,1 (0) ϕ1 + 3 a21 a2 b21 Ju f3,2 (0) ϕ1 + 2 a1 a2 b31 Ju f2,3 (0) ϕ1
[1] [1] [1] [1]
+ a2 b41 Ju f1,4 (0) ϕ1 + a41 b2 Ju f4,1 (0) ϕ1 + 2 a31 b1 b2 Ju f3,2 (0) ϕ1 + 3 a21 b21 b2 Ju f2,3 (0) ϕ1
[1] [1] [1] [1]
+ 4 a1 b31 b2 Ju f1,4 (0) ϕ1 + 5 b41 b2 Ju f0,5 (0) ϕ1 + a61 Ju f6,0 (0) ϕ1 + a51 b1 Ju f5,1 (0) ϕ1
[1] [1] [1] [1]
+ a41 b21 Ju f4,2 (0) ϕ1 + a31 b31 Ju f3,3 (0) ϕ1 + a21 b41 Ju f2,4 (0) ϕ1 + a1 b51 Ju f1,5 (0) ϕ1
[1]
+ b61 Ju f0,6 (0) ϕ1 ,
[7] [6] [6] [5] [5]
Q13 = a1 Ju f1,0 (0) W1,6 + b1 Ju f0,1 (0) W1,6 + a2 Ju f1,0 (0) W1,6 + b2 Ju f0,1 (0) W1,6
[4] [4] [3] [3]
+ a3 Ju f1,0 (0) W1,2 + b3 Ju f0,1 (0) W1,2 + a4 Ju f1,0 (0) W1,2 + b4 Ju f0,1 (0) W1,2
[5] [5] [5] [4]
+ a21 Ju f2,0 (0) W1,6 + a1 b1 Ju f1,1 (0) W1,6 + b21 Ju f0,2 (0) W1,6 + 2 a1 a2 Ju f2,0 (0) W1,2
[4] [4] [3]
+ (a1 b2 + a2 b1 ) Ju f1,1 (0) W1,2 + 2 b1 b2 Ju f0,2 (0) W1,2 + 2 a1 a3 Ju f2,0 (0) W1,2
[3] [3] [3]
+ (a1 b3 + a3 b1 ) Ju f1,1 (0) W1,2 + 2 b1 b3 Ju f0,2 (0) W1,2 + a22 Ju f2,0 (0) W1,2
[3] [3] [4] [4]
+ a2 b2 Ju f1,1 (0) W1,2 + b22 Ju f0,2 (0) W1,2 + a31 Ju f3,0 (0) W1,2 + a21 b1 Ju f2,1 (0) W1,2

91
 [4] [4] [3] [3]
+ a1 b21 Ju f1,2 (0) W1,2 + b31 Ju f0,3 (0) W1,2 + 3 a21 a2 Ju f3,0 (0) W1,2 + 2 a1 a2 b1 Ju f2,1 (0) W1,2
[3] [3] [3] [3]
+ a2 b21 Ju f1,2 (0) W1,2 + a21 b2 Ju f2,1 (0) W1,2 + 2 a1 b1 b2 Ju f1,2 (0) W1,2 + 3 b21 b2 Ju f0,3 (0) W1,2
[3] [3] [3] [3]
+ a41 Ju f4,0 (0) W1,2 + a31 b1 Ju f3,1 (0) W1,2 + a21 b21 Ju f2,2 (0) W1,2 + a1 b31 Ju f1,3 (0) W1,2
   
[3] [1] [6] [6] [2] [5] [5]
+ b41 Ju f0,4 (0) W1,2 + 2 F2,0,0 ϕ1 , 2 W0 + W2 + 2 F2,0,0 W1 , 2 W0 + W2
     
[3] [4] [4] [4] [3] [3] [5] [2] [2]
+ 2 F2,0,0 W1 , 2 W0 + W2 + 2 F2,0,0 W1 , 2 W0 + W2 + 2 F2,0,0 W1,5 , 2 W0 + W2
   
[1] [5] [5] [1] [5] [5]
+ 2 a1 F2,1,0 ϕ1 , 2 W0 + W2 + 2 b1 F2,0,1 ϕ1 , 2 W0 + W2
   
[2] [4] [4] [2] [4] [4]
+ 2 a1 F2,1,0 W1 , 2 W0 + W2 + 2 b1 F2,0,1 W1 , 2 W0 + W2
   
[3] [3] [3] [3] [3] [3]
+ 2 a1 F2,1,0 W1 , 2 W0 + W2 + 2 b1 F2,0,1 W1 , 2 W0 + W2
   
[4] [2] [2] [4] [2] [2]
+ 2 a1 F2,1,0 W1 , 2 W0 + W2 + 2 b1 F2,0,1 W1 , 2 W0 + W2
   
[1] [4] [4] [1] [4] [4]
+ 2 a2 F2,1,0 ϕ1 , 2 W0 + W2 + 2 b2 F2,0,1 ϕ1 , 2 W0 + W2
   
[2] [3] [3] [2] [3] [3]
+ 2 a2 F2,1,0 W1 , 2 W0 + W2 + 2 b2 F2,0,1 W1 , 2 W0 + W2
   
[3] [2] [2] [3] [2] [2]
+ 2 a2 F2,1,0 W1 , 2 W0 + W2 + 2 b2 F2,0,1 W1 , 2 W0 + W2
   
[1] [3] [3] [1] [3] [3]
+ 2 a3 F2,1,0 ϕ1 , 2 W0 + W2 + 2 b3 F2,0,1 ϕ1 , 2 W0 + W2
   
[2] [2] [2] [2] [2] [2]
+ 2 a3 F2,1,0 W1 , 2 W0 + W2 + 2 b3 F2,0,1 W1 , 2 W0 + W2
   
[1] [2] [2] [1] [2] [2]
+ 2 a4 F2,1,0 ϕ1 , 2 W0 + W2 + 2 b4 F2,0,1 ϕ1 , 2 W0 + W2
   
[1] [4] [4] [1] [4] [4]
+ 2 a21 F2,2,0 ϕ1 , 2 W0 + W2 + 2 a1 b1 F2,1,1 ϕ1 , 2 W0 + W2
   
[1] [4] [4] [2] [3] [3]
+ 2 b21 F2,0,2 ϕ1 , 2 W0 + W2 + 2 a21 F2,2,0 W1 , 2 W0 + W2
   
[2] [3] [3] [2] [3] [3]
+ 2 a1 b1 F2,1,1 W1 , 2 W0 + W2 + 2 b21 F2,0,2 W1 , 2 W0 + W2
   
[3] [2] [2] [3] [2] [2]
+ 2 a21 F2,2,0 W1 , 2 W0 + W2 + 2 a1 b1 F2,1,1 W1 , 2 W0 + W2
   
[3] [2] [2] [1] [3] [3]
+ 2 b21 F2,0,2 W1 , 2 W0 + W2 + 4 a1 a2 F2,2,0 ϕ1 , 2 W0 + W2
   
[1] [3] [3] [1] [3] [3]
+ 2 (a1 b2 + a2 b1 ) F2,1,1 ϕ1 , 2 W0 + W2 + 4 b1 b2 F2,0,2 ϕ1 , 2 W0 + W2
   
[2] [2] [2] [2] [2] [2]
+ 4 a1 a2 F2,2,0 W1 , 2 W0 + W2 + 2 (a1 b2 + a2 b1 ) F2,1,1 W1 , 2 W0 + W2
   
[2] [2] [2] [1] [2] [2]
+ 4 b1 b2 F2,0,2 W1 , 2 W0 + W2 + 4 a1 a3 F2,2,0 ϕ1 , 2 W0 + W2
   
[1] [2] [2] [1] [2] [2]
+ 2 (a1 b3 + a3 b1 ) F2,1,1 ϕ1 , 2 W0 + W2 + 4 b1 b3 F2,0,2 ϕ1 , 2 W0 + W2
   
[1] [2] [2] [1] [2] [2]
+ 2 a22 F2,2,0 ϕ1 , 2 W0 + W2 + 2 a2 b2 F2,1,1 ϕ1 , 2 W0 + W2
   
[1] [2] [2] [1] [3] [3]
+ 2 b22 F2,0,2 ϕ1 , 2 W0 + W2 + 2 a31 F2,3,0 ϕ1 , 2 W0 + W2
   
[1] [3] [3] [1] [3] [3]
+ 2 a21 b1 F2,2,1 ϕ1 , 2 W0 + W2 + 2 a1 b21 F2,1,2 ϕ1 , 2 W0 + W2
   
[1] [3] [3] [2] [2] [2]
+ 2 b31 F2,0,3 ϕ1 , 2 W0 + W2 + 2 a31 F2,3,0 W1 , 2 W0 + W2
   
[2] [2] [2] [2] [2] [2]
+ 2 a21 b1 F2,2,1 W1 , 2 W0 + W2 + 2 a1 b21 F2,1,2 W1 , 2 W0 + W2
   
[2] [2] [2] [1] [2] [2]
+ 2 b31 F2,0,3 W1 , 2 W0 + W2 + 6 a21 a2 F2,3,0 ϕ1 , 2 W0 + W2

92
    
[1] [2] [2] [1] [2] [2]
+ 4 a1 a2 b1 F2,2,1 ϕ1 , 2 W0 + W2 + 2 a2 b21 F2,1,2 ϕ1 , 2 W0 + W2
   
[1] [2] [2] [1] [2] [2]
+ 2 a21 b2 F2,2,1 ϕ1 , 2 W0 + W2 + 4 a1 b1 b2 F2,1,2 ϕ1 , 2 W0 + W2
   
[1] [2] [2] [1] [2] [2]
+ 6 b21 b2 F2,0,3 ϕ1 , 2 W0 + W2 + 2 a41 F2,4,0 ϕ1 , 2 W0 + W2
   
[1] [2] [2] [1] [2] [2]
+ 2 a31 b1 F2,3,1 ϕ1 , 2 W0 + W2 + 2 a21 b21 F2,2,2 ϕ1 , 2 W0 + W2
   
[1] [2] [2] [1] [2] [2]
+ 2 a1 b31 F2,1,3 ϕ1 , 2 W0 + W2 + 2 b41 F2,0,4 ϕ1 , 2 W0 + W2
     
[1] [1] [5] [1] [2] [4] [1] [3] [3]
+ 9 F3,0,0 ϕ1 , ϕ1 , W1,5 + 18 F3,0,0 ϕ1 , W1 , W1 + 9 F3,0,0 ϕ1 , W1 , W1
     
[2] [2] [3] [1] [1] [4] [1] [1] [4]
+ 9 F3,0,0 W1 , W1 , W1 + 9 a1 F3,1,0 ϕ1 , ϕ1 , W1 + 9 b1 F3,0,1 ϕ1 , ϕ1 , W1
     
[1] [2] [3] [1] [2] [3] [2] [2] [2]
+ 18 a1 F3,1,0 ϕ1 , W1 , W1 + 18 b1 F3,0,1 ϕ1 , W1 , W1 + 3 a1 F3,1,0 W1 , W1 , W1
     
[2] [2] [2] [1] [1] [3] [1] [1] [3]
+ 3 b1 F3,0,1 W1 , W1 , W1 + 9 a2 F3,1,0 ϕ1 , ϕ1 , W1 + 9 b2 F3,0,1 ϕ1 , ϕ1 , W1
     
[1] [2] [2] [1] [2] [2] [1] [1] [2]
+ 9 a2 F3,1,0 ϕ1 , W1 , W1 + 9 b2 F3,0,1 ϕ1 , W1 , W1 + 9 a3 F3,1,0 ϕ1 , ϕ1 , W1
     
[1] [1] [2] [1] [1] [1] [1] [1] [1]
+ 9 b3 F3,0,1 ϕ1 , ϕ1 , W1 + 3 a4 F3,1,0 ϕ1 , ϕ1 , ϕ1 + 3 b4 F3,0,1 ϕ1 , ϕ1 , ϕ1
     
[1] [1] [3] [1] [1] [3] [1] [1] [3]
+ 9 a21 F3,2,0 ϕ1 , ϕ1 , W1 + 9 a1 b1 F3,1,1 ϕ1 , ϕ1 , W1 + 9 b21 F3,0,2 ϕ1 , ϕ1 , W1
     
[1] [2] [2] [1] [2] [2] [1] [2] [2]
+ 9 a21 F3,2,0 ϕ1 , W1 , W1 + 9 a1 b1 F3,1,1 ϕ1 , W1 , W1 + 9 b21 F3,0,2 ϕ1 , W1 , W1
   
[1] [1] [2] [1] [1] [2]
+ 18 a1 a2 F3,2,0 ϕ1 , ϕ1 , W1 + 9 (a1 b2 + a2 b1 ) F3,1,1 ϕ1 , ϕ1 , W1
   
[1] [1] [2] [1] [1] [1]
+ 18 b1 b2 F3,0,2 ϕ1 , ϕ1 , W1 + 6 a1 a3 F3,2,0 ϕ1 , ϕ1 , ϕ1
   
[1] [1] [1] [1] [1] [1]
+ 3 (a1 b3 + a3 b1 ) F3,1,1 ϕ1 , ϕ1 , ϕ1 + 6 b1 b3 F3,0,2 ϕ1 , ϕ1 , ϕ1
     
[1] [1] [1] [1] [1] [1] [1] [1] [1]
+ 3 a22 F3,2,0 ϕ1 , ϕ1 , ϕ1 + 3 a2 b2 F3,1,1 ϕ1 , ϕ1 , ϕ1 + 3 b22 F3,0,2 ϕ1 , ϕ1 , ϕ1
     
[1] [1] [2] [1] [1] [2] [1] [1] [2]
+ 9 a31 F3,3,0 ϕ1 , ϕ1 , W1 + 9 a21 b1 F3,2,1 ϕ1 , ϕ1 , W1 + 9 a1 b21 F3,1,2 ϕ1 , ϕ1 , W1
     
[1] [1] [2] [1] [1] [1] [1] [1] [1]
+ 9 b31 F3,0,3 ϕ1 , ϕ1 , W1 + 9 a21 a2 F3,3,0 ϕ1 , ϕ1 , ϕ1 + 6 a1 a2 b1 F3,2,1 ϕ1 , ϕ1 , ϕ1
     
[1] [1] [1] [1] [1] [1] [1] [1] [1]
+ 3 a2 b21 F3,1,2 ϕ1 , ϕ1 , ϕ1 + 3 a21 b2 F3,2,1 ϕ1 , ϕ1 , ϕ1 + 6 a1 b1 b2 F3,1,2 ϕ1 , ϕ1 , ϕ1
     
[1] [1] [1] [1] [1] [1] [1] [1] [1]
+ 9 b21 b2 F3,0,3 ϕ1 , ϕ1 , ϕ1 + 3 a41 F3,4,0 ϕ1 , ϕ1 , ϕ1 + 3 a31 b1 F3,3,1 ϕ1 , ϕ1 , ϕ1
     
[1] [1] [1] [1] [1] [1] [1] [1] [1]
+ 3 a21 b21 F3,2,2 ϕ1 , ϕ1 , ϕ1 + 3 a1 b31 F3,1,3 ϕ1 , ϕ1 , ϕ1 + 3 b41 F3,0,4 ϕ1 , ϕ1 , ϕ1 ,
[7] [6] [6] [5] [5]
Q14 = a1 Ju f1,0 (0) W1,7 + b1 Ju f0,1 (0) W1,7 + a2 Ju f1,0 (0) W1,7 + b2 Ju f0,1 (0) W1,7
 
[5] [5] [5] [1] [6] [6]
+ a21 Ju f2,0 (0) W1,7 + a1 b1 Ju f1,1 (0) W1,7 + b21 Ju f0,2 (0) W1,7 + 2 F2,0,0 ϕ1 , 2 W0,2 + W2,2
     
[2] [5] [5] [2] [5] [3] [4]
+ 2 F2,0,0 W1 , 2 W0,2 + W2,2 + 2 F2,0,0 W2 , W3 + 4 F2,0,0 W0 , W1,2
     
[3] [4] [4] [3] [4] [4] [3] [4] [4]
+ 2 F2,0,0 W1 , 2 W0,2 + W2,2 + 2 F2,0,0 W1,2 , 2 W0 + W2 + 2 F2,0,0 W2 , W1,2 + W3
     
[3] [4] [5] [2] [2] [1] [5] [5]
+ 2 F2,0,0 W3 , W2 + 2 F2,0,0 W1,6 , 2 W0 + W2 + 2 a1 F2,1,0 ϕ1 , 2 W0,2 + W2,2
   
[1] [5] [5] [2] [4] [4]
+ 2 b1 F2,0,1 ϕ1 , 2 W0,2 + W2,2 + 2 a1 F2,1,0 W1 , 2 W0,2 + W2,2
   
[2] [4] [4] [3] [3] [3]
+ 2 b1 F2,0,1 W1 , 2 W0,2 + W2,2 + 2 a1 F2,1,0 W1,2 , 2 W0 + W2
   
[3] [3] [3] [4] [2] [2]
+ 2 b1 F2,0,1 W1,2 , 2 W0 + W2 + 2 a1 F2,1,0 W1,2 , 2 W0 + W2

93
      
[4] [2] [2] [2] [4] [2] [4]
+ 2 b1 F2,0,1 W1,2 , 2 W0 + W2 + 2 a1 F2,1,0 W2 , W3 + 2 b1 F2,0,1 W2 , W3
     
[3] [3] [3] [3] [1] [4] [4]
+ 2 a1 F2,1,0 W2 , W3 + 2 b1 F2,0,1 W2 , W3 + 2 a2 F2,1,0 ϕ1 , 2 W0,2 + W2,2
     
[1] [4] [4] [2] [3] [2] [3]
+ 2 b2 F2,0,1 ϕ1 , 2 W0,2 + W2,2 + 2 a2 F2,1,0 W2 , W3 + 2 b2 F2,0,1 W2 , W3
   
[3] [2] [2] [3] [2] [2]
+ 2 a2 F2,1,0 W1,2 , 2 W0 + W2 + 2 b2 F2,0,1 W1,2 , 2 W0 + W2
   
[1] [4] [4] [1] [4] [4]
+ 2 a21 F2,2,0 ϕ1 , 2 W0,2 + W2,2 + 2 a1 b1 F2,1,1 ϕ1 , 2 W0,2 + W2,2
     
[1] [4] [4] [2] [3] [2] [3]
+ 2 b21 F2,0,2 ϕ1 , 2 W0,2 + W2,2 + 2 a21 F2,2,0 W2 , W3 + 2 a1 b1 F2,1,1 W2 , W3
   
[2] [3] [3] [2] [2]
+ 2 b21 F2,0,2 W2 , W3 + 2 a21 F2,2,0 W1,2 , 2 W0 + W2
   
[3] [2] [2] [3] [2] [2]
+ 2 a1 b1 F2,1,1 W1,2 , 2 W0 + W2 + 2 b21 F2,0,2 W1,2 , 2 W0 + W2
   
[1] [1] [5] [5] [1] [2] [4] [4]
+ 3 F3,0,0 ϕ1 , ϕ1 , 3 W1,6 + W3 + 12 F3,0,0 ϕ1 , W0 , 2 W0 + W2
   
[1] [2] [4] [4] [1] [2] [4] [4]
+ 6 F3,0,0 ϕ1 , W1 , 3 W1,2 + W3 + 12 F3,0,0 ϕ1 , W2 , W0 + W2
   
[1] [3] [3] [3] [1] [3] [3] [3]
+ 12 F3,0,0 ϕ1 , W0 , W0 + W2 + 6 F3,0,0 ϕ1 , W1 , 3 W1,2 + W3
   
[1] [3] [3] [2] [3] [2] [2]
+ 6 F3,0,0 ϕ1 , W2 , W2 + 12 F3,0,0 W0 , W1 , W0 + W2
   
[2] [2] [3] [3] [2] [2] [3] [3]
+ 12 F3,0,0 W0 , W1 , 2 W0 + W2 + 3 F3,0,0 W1 , W1 , 3 W1,2 + W3
   
[2] [2] [3] [3] [2] [2] [3]
+ 12 F3,0,0 W1 , W2 , W0 + W2 + 6 F3,0,0 W2 , W2 , W1
   
[1] [1] [4] [4] [1] [1] [4] [4]
+ 3 a1 F3,1,0 ϕ1 , ϕ1 , 3 W1,2 + W3 + 3 b1 F3,0,1 ϕ1 , ϕ1 , 3 W1,2 + W3
   
[1] [2] [3] [3] [1] [2] [3] [3]
+ 12 a1 F3,1,0 ϕ1 , W0 , 2 W0 + W2 + 12 b1 F3,0,1 ϕ1 , W0 , 2 W0 + W2
   
[1] [2] [3] [3] [1] [2] [3] [3]
+ 6 a1 F3,1,0 ϕ1 , W1 , 3 W1,2 + W3 + 6 b1 F3,0,1 ϕ1 , W1 , 3 W1,2 + W3
   
[1] [2] [3] [3] [1] [2] [3] [3]
+ 12 a1 F3,1,0 ϕ1 , W2 , W0 + W2 + 12 b1 F3,0,1 ϕ1 , W2 , W0 + W2
   
[2] [2] [2] [2] [2] [2]
+ 12 a1 F3,1,0 W0 , W0 , W1 + 12 b1 F3,0,1 W0 , W0 , W1
   
[2] [2] [2] [2] [2] [2] [2] [2]
+ 6 a1 F3,1,0 W1 , W2 , 2 W0 + W2 + 6 b1 F3,0,1 W1 , W2 , 2 W0 + W2
   
[1] [1] [3] [3] [1] [1] [3] [3]
+ 3 a2 F3,1,0 ϕ1 , ϕ1 , 3 W1,2 + W3 + 3 b2 F3,0,1 ϕ1 , ϕ1 , 3 W1,2 + W3
   
[1] [2] [2] [1] [2] [2]
+ 12 a2 F3,1,0 ϕ1 , W0 , W0 + 12 b2 F3,0,1 ϕ1 , W0 , W0
   
[1] [2] [2] [2] [1] [2] [2] [2]
+ 6 a2 F3,1,0 ϕ1 , W2 , 2 W0 + W2 + 6 b2 F3,0,1 ϕ1 , W2 , 2 W0 + W2
   
[1] [1] [3] [3] [1] [1] [3] [3]
+ 3 a21 F3,2,0 ϕ1 , ϕ1 , 3 W1,2 + W3 + 3 a1 b1 F3,1,1 ϕ1 , ϕ1 , 3 W1,2 + W3
   
[1] [1] [3] [3] [1] [2] [2]
+ 3 b21 F3,0,2 ϕ1 , ϕ1 , 3 W1,2 + W3 + 12 a21 F3,2,0 ϕ1 , W0 , W0
   
[1] [2] [2] [1] [2] [2]
+ 12 a1 b1 F3,1,1 ϕ1 , W0 , W0 + 12 b21 F3,0,2 ϕ1 , W0 , W0
   
[1] [2] [2] [2] [1] [2] [2] [2]
+ 6 a21 F3,2,0 ϕ1 , W2 , 2 W0 + W2 + 6 a1 b1 F3,1,1 ϕ1 , W2 , 2 W0 + W2
   
[1] [2] [2] [2] [1] [1] [1] [4] [4]
+ 6 b21 F3,0,2 ϕ1 , W2 , 2 W0 + W2 + 8 F4,0,0 ϕ1 , ϕ1 , ϕ1 , 3 W0 + 2 W2
   
[1] [1] [2] [3] [3] [1] [1] [3] [2] [2]
+ 24 F4,0,0 ϕ1 , ϕ1 , W1 , 3 W0 + 2 W2 + 24 F4,0,0 ϕ1 , ϕ1 , W1 , 3 W0 + 2 W2

94
    
[1] [2] [2] [2] [2] [1] [1] [1] [3] [3]
+ 24 F4,0,0 ϕ1 , W1 , W1 , 3 W0 + 2 W2 + 8 a1 F4,1,0 ϕ1 , ϕ1 , ϕ1 , 3 W0 + 2 W2
   
[1] [1] [1] [3] [3] [1] [1] [2] [2] [2]
+ 8 b1 F4,0,1 ϕ1 , ϕ1 , ϕ1 , 3 W0 + 2 W2 + 24 a1 F4,1,0 ϕ1 , ϕ1 , W1 , 3 W0 + 2 W2
   
[1] [1] [2] [2] [2] [1] [1] [1] [2] [2]
+ 24 b1 F4,0,1 ϕ1 , ϕ1 , W1 , 3 W0 + 2 W2 + 8 a2 F4,1,0 ϕ1 , ϕ1 , ϕ1 , 3 W0 + 2 W2
   
[1] [1] [1] [2] [2] [1] [1] [1] [2] [2]
+ 8 b2 F4,0,1 ϕ1 , ϕ1 , ϕ1 , 3 W0 + 2 W2 + 8 a21 F4,2,0 ϕ1 , ϕ1 , ϕ1 , 3 W0 + 2 W2
   
[1] [1] [1] [2] [2] [1] [1] [1] [2] [2]
+ 8 a1 b1 F4,1,1 ϕ1 , ϕ1 , ϕ1 , 3 W0 + 2 W2 + 8 b21 F4,0,2 ϕ1 , ϕ1 , ϕ1 , 3 W0 + 2 W2
   
[1] [1] [1] [1] [3] [1] [1] [1] [2] [2]
+ 50 F5,0,0 ϕ1 , ϕ1 , ϕ1 , ϕ1 , W1 + 100 F5,0,0 ϕ1 , ϕ1 , ϕ1 , W1 , W1
   
[1] [1] [1] [1] [2] [1] [1] [1] [1] [2]
+ 50 a1 F5,1,0 ϕ1 , ϕ1 , ϕ1 , ϕ1 , W1 + 50 b1 F5,0,1 ϕ1 , ϕ1 , ϕ1 , ϕ1 , W1
   
[1] [1] [1] [1] [1] [1] [1] [1] [1] [1]
+ 10 a2 F5,1,0 ϕ1 , ϕ1 , ϕ1 , ϕ1 , ϕ1 + 10 b2 F5,0,1 ϕ1 , ϕ1 , ϕ1 , ϕ1 , ϕ1
   
[1] [1] [1] [1] [1] [1] [1] [1] [1] [1]
+ 10 a21 F5,2,0 ϕ1 , ϕ1 , ϕ1 , ϕ1 , ϕ1 + 10 a1 b1 F5,1,1 ϕ1 , ϕ1 , ϕ1 , ϕ1 , ϕ1
 
[1] [1] [1] [1] [1]
+ 10 b21 F5,0,2 ϕ1 , ϕ1 , ϕ1 , ϕ1 , ϕ1 ,
     
[7] [1] [6] [6] [2] [5] [3] [4] [4]
Q15 = 2 F2,0,0 ϕ1 , 2 W0,3 + W2,3 + 2 F2,0,0 W2 , W3,2 + 2 F2,0,0 W1,2 , 2 W0,2 + W2,2
   
[3] [4] [4] [5] [2] [2]
+ 2 F2,0,0 W3 , W2,2 + W4 + 2 F2,0,0 W1,7 , 2 W0 + W2
   
[1] [1] [5] [5] [1] [2] [4] [4]
+ 3 F3,0,0 ϕ1 , ϕ1 , 3 W1,7 + W3,2 + 12 F3,0,0 ϕ1 , W0 , 2 W0,2 + W2,2
   
[1] [2] [4] [4] [4] [1] [3] [3] [3]
+ 6 F3,0,0 ϕ1 , W2 , 2 W0,2 + 2 W2,2 + W4 + 3 F3,0,0 ϕ1 , W1,2 , 3 W1,2 + 2 W3
   
[1] [3] [3] [2] [2] [3]
+ 6 F3,0,0 ϕ1 , W3 , W3 + 12 F3,0,0 W0 , W0 , W1,2
   
[2] [2] [3] [3] [2] [2] [3] [3]
+ 12 F3,0,0 W0 , W2 , W1,2 + W3 + 3 F3,0,0 W2 , W2 , 2 W1,2 + W3
   
[1] [1] [1] [4] [4] [4] [1] [1] [2] [3] [3]
+ 4 F4,0,0 ϕ1 , ϕ1 , ϕ1 , 6 W0,2 + 4 W2,2 + W4 + 24 F4,0,0 ϕ1 , ϕ1 , W0 , 3 W1,2 + W3
   
[1] [1] [2] [3] [3] [1] [2] [2] [2] [2]
+ 12 F4,0,0 ϕ1 , ϕ1 , W2 , 4 W1,2 + 3 W3 + 16 F4,0,0 ϕ1 , W0 , W0 , 2 W0 + 3 W2
   
[1] [2] [2] [2] [2] [1] [1] [1] [1] [3] [3]
+ 12 F4,0,0 ϕ1 , W2 , W2 , 4 W0 + W2 + 25 F5,0,0 ϕ1 , ϕ1 , ϕ1 , ϕ1 , 2 W1,2 + W3
   
[1] [1] [1] [2] [2] [2] [1] [1] [1] [2] [2]
+ 40 F5,0,0 ϕ1 , ϕ1 , ϕ1 , W0 , 3 W0 + 4 W2 + 70 F5,0,0 ϕ1 , ϕ1 , ϕ1 , W2 , W2
 
[1] [1] [1] [1] [1] [2] [2]
+ 30 F6,0,0 ϕ1 , ϕ1 , ϕ1 , ϕ1 , ϕ1 , 4 W0 + 3 W2
 
[1] [1] [1] [1] [1] [1] [1]
+ 35 F7,0,0 ϕ1 , ϕ1 , ϕ1 , ϕ1 , ϕ1 , ϕ1 , ϕ1 .

Finally, as we stated at the previous order, we assume that A2 = 0, which implies that (104) becomes
2
α1 A3XX + i α2 A3X + i α3 |A1 | A3X + i α3 A1 A1X Ā3 + i α3 Ā1 A1X A3 + i α4 A21 Ā3X + 2 i α4 A1 Ā1X A3
2 2 4
+ α5 A3 + α6 A21 Ā3 + 2 α6 |A1 | A3 + 2 α7 |A1 | A21 Ā3 + 3 α7 |A1 | A3
2 2
+ i α1,3 A1XXX + α2,3 A1XX + α3,3 |A1 | A1XX + α4,3 A21 Ā1XX + i α5,3 A1X + i α6,3 |A1 | A1X
4 2 2 2
+ i α7,3 |A1 | A1X + i α8,3 A21 Ā1X + i α9,3 |A1 | A21 Ā1X + α10,3 Ā1 (A1X ) + α11,3 A1 |A1X |
2 4 6 2
+ α12,3 A1 + α13,3 |A1 | A1 + α14,3 |A1 | A1 + α15,3 |A1 | A1 + 2 α1 A1XΞ + i α2 A1Ξ + i α3 |A1 | A1Ξ
+ i α4 A21 Ā1Ξ = 0. (105)
Now, from the solvability condition (6), we note that
1  2 2 4

A′′1 = − i α2 A′1 + i α3 |A1 | A′1 + i α4 A21 Ā′1 + α5 A1 + α6 |A1 | A1 + α7 |A1 | A1 ,
α1

95
which can be used to simplify (105) into
2 2 4
α1 A3XX + i α2 A3X + i α3 |A1 | A3X + i α4 A21 Ā3X + α5 A3 + 2 α6 |A1 | A3 + 3 α7 |A1 | A3
2
+ i α3 Ā1 A1X A3 + 2 i α4 A1 Ā1X A3 + α6 A21 Ā3 + 2 α7 |A1 | A21 Ā3 + i α3 A1 A1X Ā3
2 4 2
+ i α1,4 A1X + i α2,4 |A1 | A1X + i α3,4 |A1 | A1X + i α4,4 A21 Ā1X + i α5,4 |A1 | A21 Ā1X
2 2 2 4 6
+ α6,4 Ā1 (A1X ) + α7,4 A1 |A1X | + α8,4 A1 + α9,4 |A1 | A1 + α10,4 |A1 | A1 + α11,4 |A1 | A1
2
+ 2 α1 A1XΞ + i α2 A1Ξ + i α3 |A1 | A1Ξ + i α4 A21 Ā1Ξ = 0, (106)
where
 
α5 α2 α2
α1,4 = α5,3 − α1,3 − α2,3 + α1,3 ,
α1 α1 α1
 
α2 α6 α3 α2
α2,4 = α6,3 − α3,3 − 2 α1,3 − α2,3 + 2 α1,3 ,
α1 α1 α1 α1
   
α7 α3 α3 α4 α4
α3,4 = α7,3 − 3 α1,3 − α3,3 + α1,3 + α4,3 + α1,3 ,
α1 α1 α1 α1 α1
α2 α4 α6
α4,4 = α8,3 + α4,3 − α2,3 − α1,3 ,
α1 α1 α1
α3 α4 α7 α1,3
α5,4 = α9,3 + α4,3 − α3,3 − 2 ,
α1 α1 α1
α3 α1,3
α6,4 = α10,3 +
α1
(α3 + 2 α4 ) α1,3
α7,4 = α11,3 + ,
α
 1 
α5 α2 α1,3
α8,4 = α12,3 − α2,3 + ,
α1 α1
   
α5 (α3 + α4 ) α1,3 α6 α2 α1,3
α9,4 = α13,3 − α3,3 + α4,3 + − α2,3 + ,
α1 α1 α1 α1
   
α6 (α3 + α4 ) α1,3 α7 α2 α1,3
α10,4 = α14,3 − α3,3 + α4,3 + − α2,3 + ,
α1 α1 α1 α1
 
α7 (α3 + α4 ) α1,3
α11,4 = α15,3 − α3,3 + α4,3 + .
α1 α1
Now, we proceed to solve equation (106). First, we set A3 = (R3 + i R1 φ3 ) ei φ1 , where R3 , and φ3 are real
functions. This implies that we obtain two equations, one corresponding to the real and another for the
imaginary part of the resulting equation. The first one, after using equations (26) and (25) to simplify it,
becomes


(α3 + α4 ) R13 R3 X
 
2 α1,4 R1 R1X α2 α6,4 (α2,4 + α4,4 )
R1 φ3XX + 2 R1 R1X φ3X = − − + − R13 R1X
α1 2 α1 α12 α1
 
(α3 + α4 ) α6,4 (α3,4 + α5,4 )
+ − R15 R1X ,
2 α12 α1
which implies
 
α2 α1,4 (α3 + α4 ) R1 R3 α2 α6,4 α2,4 + α4,4
φ3X =− − − ζΞ − + − R12
2 α1 2 α1 2 α1 4 α12 4 α1
 
(α3 + α4 ) α6,4 α3,4 + α5,4 4 α1 ω3
+ 2 − R14 + .
12 α1 6 α1 R12
where ω3 ∈ R is a constant of integration. With this, the second equation becomes
dp 2
α1 R3XX − α1 β1 + 6 β3 R12 + 15 β5 R14 R3 = 2 α1 β†,3 R1 + α1 (R1 ) + α1 β∗,3 R1 (R1′ ) ,


(107)
dR1

96
where
α2 (α2 + α1,4 ) α8,4 α2
β†,3 = (α3 − 3 α4 ) ω3 − − − ζΞ ,
4 α12 2 α1 2 α1
dp
α1 (R1 ) = −β3,3 R13 − β5,3 R15 − β7,3 R17 ,
dR1
α2 (α3 − α4 + α2,4 − α4,4 ) α1,4 (α3 − α4 ) α22 (α7,4 − α6,4 )
β3,3 = + + + α9,4 ,
2 α1 2 α1 4 α12
α2 (α3,4 − α5,4 ) α3 (3 α2,4 − α4,4 ) α4 (α2,4 + 5 α4,4 ) α2 α3 (2 α7,4 − 3 α6,4 )
β5,3 = + − +
2 α1 8 α1 8 α1 8 α12
α2 α4 (α6,4 + 2 α7,4 )
+ + α10,4 ,
8 α12
α3 (2 α3,4 − α5,4 ) α4 α5,4 α2 (3 α7,4 − 5 α6,4 ) α3 α4 (3 α7,4 − α6,4 )
β7,3 = − + 3 +
6 α1 2 α1 48 α12 24 α12
α2 (α6,4 + α7,4 )
+ 4 + α11,4 ,
16 α12
1
β∗,3 = − (α6,4 + α7,4 ) .
α1
Now, note that (107) is a linear equation with respect to R3 that has a homogeneous solution given by
Z X !
′ ds
R3,h = R1 ω1 + ω2 2 ,
(R1′ )
which implies that a particular solution is given by
! !
X X X X
R12
Z Z Z Z
ds ds p (R1 )
R3,p = R1′ −β†,3 R12 2 − 2 dσ − p (R1 ) 2 − 2 dσ
(R1′ ) (R1′ ) (R1′ ) (R1′ )
Z X !
Z σ
3 ds
− β∗,3 R1 (R1′ ) 2 dσ
(R1′ )
Z X !Z !
X
′ 3 ds
+ β†,3 R12 + p (R1 ) + β∗,3 R1 (R1 ) dσ 2
(R1′ )
X X
R12
Z Z
p (R1 )
= R1′ β†,3 2 dσ + 2 dσ
(R1′ ) (R1′ )
! ! ! !
Z X Z X Z X Z σ
3 ds 3 ds
+β∗,3 R1 (R1′ ) dσ 2 − β ∗,3 R1 (R1′ ) 2 dσ .
(R1′ ) (R1′ )
Therefore,
R3 = R3,h + R3,p
X


ω2 + β†,3 R12 + p (R1 )
Z
= R1′ ω1 + 2 dσ
(R1′ )
! ! ! !
Z X Z X Z X Z σ
3 ds 3 ds
+β∗,3 R1 (R1′ ) dσ 2 − β∗,3 R1 (R1′ ) 2 dσ ,
(R1′ ) (R1′ )
which lets us note that, as X → −∞, we have
3/2 √

−β13 β⋆,3 + 6 β32 ω2 + 24 i α1 β1 β32 ω3 −√β1 X i φ1

β3
A3 ∼ − √ e e ,
β1 12 2 β33

97
implying that we must set

β13 (α6,4 + α7,4 )

ω2 = − , and ω3 = 0,
6 α1 β32

to keep A3 bounded as X → −∞. On the other hand, using these variables, as X → ∞ we have
 3/2
− ββ31  p 
A3 ∼ √ 2 6 2 α1 β3 + i β1 (α3 + α4 ) −6 α1 β1 β33 β†,3 + 3 α1 β34 ω2 − 3 β12 β32 β3,3
24 2 α1 β3

√
+4 β13 β3 β5,3 − 6 β14 β7,3 e2 β1 X ei φ1
 3/2
β12 − ββ31
√ 3 6 3 α2 β33 4 α1 β1 α6,4 + (α3 + α4 ) α1,4 + 4 α12 β3


−i
12 2 α1 β3
+ 2 α1 2 α1 β32 4 β12 (α3,4 + α5,4 ) − 3 β1 β3 (α2,4 + α4,4 ) + 3 β32 α1,4

+ (α3 + α4 ) −β12 β3 (β3 (5 α6,4 + α7,4 ) + 4 β5,3 ) + 3 β33 α8,4 + 12 β13 β7,3



√
+3 α22 (α3 + α4 ) β33 + 6 α1 β33 α2 (α3 + α4 ) + 4 α12 β3 ζΞ X e2 β1 X ei φ1 .

Once again, this expression must equal zero for A3 to remain finite as X → ∞. In particular, to obtain
explicit conditions that make this quantity equal to zero, note that if α2 = 0, then we can choose


β3 β12 (β3 (α6,4 + α7,4 ) − 8 β5,3 ) − 6 β32 α8,4 + 6 β1 β3 β3,3
β7,3 = − , (108)
12 β13
1
(α3 + α4 ) 2 β12 (β3 (3 α6,4 + α7,4 ) − 2 β5,3 ) − 9 β32 α8,4 + 6 β1 β3 β3,3


ζΞ = 2 3
12 α1 β3
−2 α1 β3 4 β12 (α3,4 + α5,4 ) − 3 β1 β3 (α2,4 + α4,4 ) + 3 β32 α1,4 ,



(109)

whilst, if α2 ̸= 0, then we choose

1
α2,4 = 3 α2 β33 (2 α1 (2 β1 α6,4 + 2 α1 β3 + (α3 + α4 ) ζΞ ) + (α3 + α4 ) α1,4 )
12 α12 β1 β33
+ 2 α1 2 α1 β32 4 β12 (α3,4 + α5,4 ) − 3 β3 β1 α4,4 + 3 β32 α1,4 + 3 α22 (α3 + α4 ) β33

+ (α3 + α4 ) −β12 β3 (β3 (5 α6,4 + α7,4 ) + 4 β5,3 ) + 3 β33 α8,4 + 12 β13 β7,3 + 12 α12 β34 ζΞ , (110)

1 β12 β3 (β3 (α6,4 + α7,4 ) − 8 β5,3 ) − 6 β33 α8,4 + 12 β13 β7,3 + 6 β1 β32 β3,3
 
3 (α2 + α1,4 )
ζΞ = − (,111)
6 α2 β33 α1

where (108) and (110) correspond to equations that must be solved in terms of the parameters of the
expansion in order to obtain a correction to the Maxwell point, and (109) and (111) correspond to the values
of ζΞ in these two cases, where the equalities (108) and (110) have been used to explicitly state the value of
ζΞ in each instance. We highlight that in both cases, ζΞ ∈ R is a constant, which implies that ζ is a linear
function on Ξ. Z X
As a final remark, we highlight that the integrals in this appendix were defined as ds just for
Z X
simplicity, as the integrals for the homogeneous solution do converge when computing ds, but the same
X0
integrals for the particular solution do not. In any case, we highlight that this does not change the value of

98
the leading-order coefficient of A3 at X = X0 , which is given by

eπ ξ+ζ i
− q
5/4 9/2 2−η i
192 α13 β1 (−β3 ) − X01−X (X0 − X)
  p 
× −3 (2 π + 3 i) α2 β33 (α3 + α4 ) β1 + 2 i α1 β3 (α1,4 + 2 α1 ζΞ )
  
5/2
+ i α1 2 α1 β3 β3 −16 β1 (α3,4 + α5,4 ) + i β12 (5 β3 (α6,4 + α7,4 ) + 8 β5,3 ) − 6 (2 π + 3 i) β32 α8,4
p
+6 i β1 β3 β3,3 ) + 12 (2 π − 3 i) β13 β7,3 + (α3 + α4 ) β1 β12 β3 (β3 (15 α6,4 − α7,4 ) + 8 β5,3 )

+6 (−3 + 2 π i) β33 α8,4 + 12 (−5 − 2 π i) β13 β7,3 + 6 β32 β1 β3,3

 p 
−3 (2 π + 3 i) α22 β33 (α3 + α4 ) β1 + 2 α1 β3 i .

C Equation for the remainder up to order 7

In this section, just in order to be complete, we state the solutions to each equation when expanding the
remainder to solve its corresponding equation (see Section 5). In particular, we have that
[5] [5] 2 [5] [5] [5] [5]
RN = 2 A1 B̄1 W0 + 4 |A1 | A1 B̄1 W0,2 + i A1X B̄1 W0,3 + i Ā1 B1X W0,3 + B1XX eix W1
[5] 2 [5] [5] [5]
+ i B1X eix W1,2 + i |A1 | B1X eix W1,3 + i A1 A1X B̄1 eix W1,3 + i Ā1 A1X B1 eix W1,3
[5] [5] [5] [5]
+ i A21 B̄1X eix W1,4 + 2 i A1 Ā1X B1 eix W1,4 + B1 eix W1,5 + A21 B̄1 eix W1,6
2 [5] 2 [5] 4 [5] [3]
+ 2 |A1 | B1 eix W1,6 + 2 |A1 | A21 B̄1 eix W1,7 + 3 |A1 | B1 eix W1,7 + i B1Ξ eix W1,3
[5] [5] 2 [5] [5]
+ 2 A1 B1 e2ix W2 + A31 B̄1 e2ix W2,2 + 3 |A1 | A1 B1 e2ix W2,2 + i A1X B1 e2ix W2,3
[5] [5] [5] 2 [5]
+ i A1 B1X e2ix W2,3 + 3 A21 B1 e3ix W3 + A41 B̄1 e3ix W3,2 + 4 |A1 | A21 B1 e3ix W3,2
[5] [5] [4] [4]
+ i A21 B1X e3ix W3,3 + 2 i A1 A1X B1 e3ix W3,3 + 2 A2 B̄1 W0 + 2 A1 B̄2 W0
2 [4] [4] 2 [4] [4]
+ 4 |A1 | A1 B̄2 W0,2 + 4 A21 Ā2 B̄1 W0,2 + 8 |A1 | A2 B̄1 W0,2 + i A1X B̄2 W0,3
[4] [4] [4] [3] [3]
+ i A2X B̄1 W0,3 + i Ā2 B1X W0,3 + i Ā1 B2X W0,3 + 2 A2 B̄2 W0 + 2 A3 B̄1 W0
[3] [2] [2] [2] [2] [4]
+ 2 A1 B̄3 W0 + 2 A1 B̄4 W0 + 2 A3 B̄2 W0 + 2 A4 B̄1 W0 + 2 A2 B̄3 W0 + B2 eix W1
[4] 2 [4] [4] [4]
+ A21 B̄2 eix W1,2 + 2 |A1 | B2 eix W1,2 + 2 Ā1 A2 B1 eix W1,2 + 2 A1 A2 B̄1 eix W1,2
[4] [4] [3] [3]
+ 2 A1 Ā2 B1 eix W1,2 + i B2X eix W1,3 + B3 eix W1 + A22 B̄1 eix W1,2
[3] [3] [3] [3]
+ A21 B̄3 eix W1,2 + 2 A1 Ā2 B2 eix W1,2 + 2 Ā1 A2 B2 eix W1,2 + 2 A1 A2 B̄2 eix W1,2
2 [3] 2 [3] [3] [3]
+ 2 |A2 | B1 eix W1,2 + 2 |A1 | B3 eix W1,2 + 2 A1 Ā3 B1 eix W1,2 + 2 Ā1 A3 B1 eix W1,2
[3] [3] [2] [1] [4]
+ 2 A1 A3 B̄1 eix W1,2 + i B3X eix W1,3 + B4 eix W1 + B5 eix ϕ1 + 2 A1 B2 e2ix W2
[4] [4] 2 [4] [4]
+ 2 A2 B1 e2ix W2 + A31 B̄2 e2ix W2,2 + 3 |A1 | A1 B2 e2ix W2,2 + 3 A21 Ā2 B1 e2ix W2,2
[4] 2 [4] [4] [4]
+ 3 A21 A2 B̄1 e2ix W2,2 + 6 |A1 | A2 B1 e2ix W2,2 + i A1 B2X e2ix W2,3 + i A1X B2 e2ix W2,3
[4] [4] [3] [3]
+ i A2 B1X e2ix W2,3 + i A2X B1 e2ix W2,3 + 2 A2 B2 e2ix W2 + 2 A1 B3 e2ix W2
[3] [2] [2] [2]
+ 2 A3 B1 e2ix W2 + 2 A3 B2 e2ix W2 + 2 A2 B3 e2ix W2 + 2 A1 B4 e2ix W2
[2] [4] [4] [3]
+ 2 A4 B1 e2ix W2 + 3 A21 B2 e3ix W3 + 6 A1 A2 B1 e3ix W3 + 3 A22 B1 e3ix W3
[3] [3] [3]
+ 3 A21 B3 e3ix W3 + 6 A1 A2 B2 e3ix W3 + 6 A1 A3 B1 e3ix W3 + . . . + c.c.,

where ’. . .’ represents terms that are multiples of e4ix or e5ix . Moreover,

[6] [6] 2 [6] 4 [6] [6] [6]
RN = 2 A1 B̄1 W0 + 4 |A1 | A1 B̄1 W0,2 + 6 |A1 | A1 B̄1 W0,3 + 2 A1X B̄1X W0,4 + i Ā1 B1X W0,5

99
 [6] 2 [6] [6] 2 [6]
+ i A1X B̄1 W0,5 + i |A1 | Ā1 B1X W0,6 + i Ā21 A1X B1 W0,6 + 2 i |A1 | A1X B̄1 W0,6
[6] [6] [4] [5] [5]
+ A1 B̄1XX W0,7 + A1XX B̄1 W0,7 + i Ā1 B1Ξ W0,3 + 2 A1 B̄2 W0 + 2 A2 B̄1 W0
2 [5] [5] 2 [5] [5]
+ 4 |A1 | A1 B̄2 W0,2 + 4 A21 Ā2 B̄1 W0,2 + 8 |A1 | A2 B̄1 W0,2 + i Ā1 B2X W0,3
[5] [5] [5] [4] [4]
+ i A1X B̄2 W0,3 + i Ā2 B1X W0,3 + i A2X B̄1 W0,3 + 2 A2 B̄2 W0 + 2 A1 B̄3 W0
[4] [4] [4] 2 [4]
+ 2 A3 B̄1 W0 + 4 A21 Ā2 B̄2 W0,2 + 4 A1 Ā22 B1 W0,2 + 4 |A1 | A1 B̄3 W0,2
[4] 2 [4] 2 [4] 2 [4]
+ 4 A21 Ā3 B̄1 W0,2 + 8 |A1 | A2 B̄2 W0,2 + 8 |A2 | A1 B̄1 W0,2 + 8 |A1 | A3 B̄1 W0,2
[4] [4] [4] [4] [4]
+ i Ā2 B2X W0,3 + i A2X B̄2 W0,3 + i Ā1 B3X W0,3 + i Ā3 B1X W0,3 + i A1X B̄3 W0,3
[4] [4] [3] [3] [3]
+ i A1Ξ B̄1 W0,3 + i A3X B̄1 W0,3 + 2 A3 B̄2 W0 + 2 A2 B̄3 W0 + 2 A1 B̄4 W0
[3] [2] [2] [2] [2] [2]
+ 2 A4 B̄1 W0 + 2 A4 B̄2 W0 + 2 A3 B̄3 W0 + 2 A2 B̄4 W0 + 2 A1 B̄5 W0 + 2 A5 B̄1 W0
[6] [6] 2 [6] [6]
+ B1XX eix W1 + i B1X eix W1,2 + i |A1 | B1X eix W1,3 + i A1 A1X B̄1 eix W1,3
[6] [6] [6] [6]
+ i Ā1 A1X B1 eix W1,3 + i A21 B̄1X eix W1,4 + 2 i A1 Ā1X B1 eix W1,4 + B1 eix W1,5
[6] 2 [6] 2 [6] 4 [6]
+ A21 B̄1 eix W1,6 + 2 |A1 | B1 eix W1,6 + 2 |A1 | A21 B̄1 eix W1,7 + 3 |A1 | B1 eix W1,7
[4] [3] [5] [5] 2 [5]
+ i B1Ξ eix W1,3 + i B2Ξ eix W1,3 + B2XX eix W1 + i B2X eix W1,2 + i |A1 | B2X eix W1,3
[5] [5] [5] [5]
+ i Ā1 A1X B2 eix W1,3 + i A1 A1X B̄2 eix W1,3 + i Ā1 A2X B1 eix W1,3 + i A1X Ā2 B1 eix W1,3
[5] [5] [5] [5]
+ i A1 A2X B̄1 eix W1,3 + i Ā1 A2 B1X eix W1,3 + i A1 Ā2 B1X eix W1,3 + i A1X A2 B̄1 eix W1,3
[5] [5] [5] [5]
+ i A21 B̄2X eix W1,4 + 2 i A1 Ā1X B2 eix W1,4 + 2 i A1 A2 B̄1X eix W1,4 + 2 i A1 Ā2X B1 eix W1,4
[5] [5] [5] 2 [5]
+ 2 i Ā1X A2 B1 eix W1,4 + B2 eix W1,5 + A21 B̄2 eix W1,6 + 2 |A1 | B2 eix W1,6
[5] [5] [5] 2 [5]
+ 2 Ā1 A2 B1 eix W1,6 + 2 A1 Ā2 B1 eix W1,6 + 2 A1 A2 B̄1 eix W1,6 + 2 |A1 | A21 B̄2 eix W1,7
[5] 4 [5] 2 [5] 2 [5]
+ 2 A31 Ā2 B̄1 eix W1,7 + 3 |A1 | B2 eix W1,7 + 6 |A1 | Ā1 A2 B1 eix W1,7 + 6 |A1 | A1 Ā2 B1 eix W1,7
2 [5] [4] [4] [4] [4]
+ 6 |A1 | A1 A2 B̄1 eix W1,7 + B3 eix W1 + A21 B̄3 eix W1,2 + A22 B̄1 eix W1,2 + 2 Ā1 A2 B2 eix W1,2
[4] [4] 2 [4] 2 [4]
+ 2 A1 Ā2 B2 eix W1,2 + 2 A1 A2 B̄2 eix W1,2 + 2 |A1 | B3 eix W1,2 + 2 |A2 | B1 eix W1,2
[4] [4] [4] [4]
+ 2 A1 Ā3 B1 eix W1,2 + 2 A1 A3 B̄1 eix W1,2 + 2 Ā1 A3 B1 eix W1,2 + i B3X eix W1,3
[3] [3] [3] 2 [3] [3]
+ B4 eix W1 + A22 B̄2 eix W1,2 + A21 B̄4 eix W1,2 + 2 |A2 | B2 eix W1,2 + 2 A1 A3 B̄2 eix W1,2
[3] [3] [3] [3]
+ 2 A1 Ā3 B2 eix W1,2 + 2 Ā1 A3 B2 eix W1,2 + 2 Ā1 A2 B3 eix W1,2 + 2 A1 Ā2 B3 eix W1,2
[3] 2 [3] [3] [3]
+ 2 A1 A2 B̄3 eix W1,2 + 2 |A1 | B4 eix W1,2 + 2 Ā1 A4 B1 eix W1,2 + 2 A1 Ā4 B1 eix W1,2
[3] [3] [3] [3]
+ 2 A2 Ā3 B1 eix W1,2 + 2 Ā2 A3 B1 eix W1,2 + 2 A2 A3 B̄1 eix W1,2 + 2 A1 A4 B̄1 eix W1,2
[3] [2] [1] [6] [6]
+ i B4X eix W1,3 + B5 eix W1 + B6 eix ϕ1 + 2 A1 B1 e2ix W2 + A31 B̄1 e2ix W2,2
2 [6] 2 [6] 4 [6] [6]
+ 3 |A1 | A1 B1 e2ix W2,2 + 2 |A1 | A31 B̄1 e2ix W2,3 + 4 |A1 | A1 B1 e2ix W2,3 + 2 A1X B1X e2ix W2,4
[6] [6] 2 [6] [6]
+ i A1 B1X e2ix W2,5 + i A1X B1 e2ix W2,5 + i |A1 | A1 B1X e2ix W2,6 + i A21 A1X B̄1 e2ix W2,6
2 [6] [6] [6] [6]
+ 2 i |A1 | A1X B1 e2ix W2,6 + i A31 B̄1X e2ix W2,7 + 3 i A21 Ā1X B1 e2ix W2,7 + A1 B1XX e2ix W2,8
[6] [4] [4] [5] [5]
+ A1XX B1 e2ix W2,8 + i A1 B1Ξ e2ix W2,3 + i A1Ξ B1 e2ix W2,3 + 2 A1 B2 e2ix W2 + 2 A2 B1 e2ix W2
[5] 2 [5] [5] [5]
+ A31 B̄2 e2ix W2,2 + 3 |A1 | A1 B2 e2ix W2,2 + 3 A21 A2 B̄1 e2ix W2,2 + 3 A21 Ā2 B1 e2ix W2,2
2 [5] [5] [5] [5]
+ 6 |A1 | A2 B1 e2ix W2,2 + i A1 B2X e2ix W2,3 + i A1X B2 e2ix W2,3 + i A2 B1X e2ix W2,3
[5] [4] [4] [4] [4]
+ i A2X B1 e2ix W2,3 + 2 A2 B2 e2ix W2 + 2 A1 B3 e2ix W2 + 2 A3 B1 e2ix W2 + A31 B̄3 e2ix W2,2
[4] [4] 2 [4] [4]
+ 3 A21 Ā2 B2 e2ix W2,2 + 3 A21 A2 B̄2 e2ix W2,2 + 3 |A1 | A1 B3 e2ix W2,2 + 3 Ā1 A22 B1 e2ix W2,2

100
 [4] [4] [4] 2 [4]
+ 3 A1 A22 B̄1 e2ix W2,2 + 3 A21 Ā3 B1 e2ix W2,2 + 3 A21 A3 B̄1 e2ix W2,2 + 6 |A1 | A2 B2 e2ix W2,2
2 [4] 2 [4] [4] [4]
+ 6 |A2 | A1 B1 e2ix W2,2 + 6 |A1 | A3 B1 e2ix W2,2 + i A2 B2X e2ix W2,3 + i A2X B2 e2ix W2,3
[4] [4] [4] [4]
+ i A1 B3X e2ix W2,3 + i A1X B3 e2ix W2,3 + i A3 B1X e2ix W2,3 + i A3X B1 e2ix W2,3
[3] [3] [3] [3] [2]
+ 2 A3 B2 e2ix W2 + 2 A2 B3 e2ix W2 + 2 A1 B4 e2ix W2 + 2 A4 B1 e2ix W2 + 2 A4 B2 e2ix W2
[2] [2] [2] [2]
+ 2 A3 B3 e2ix W2 + 2 A2 B4 e2ix W2 + 2 A1 B5 e2ix W2 + 2 A5 B1 e2ix W2 + . . . + c.c.

where ’. . .’ represents terms that are multiples of e3ix , e4ix , e
5ix , or e6ix .
Last but not least, the solvability condition at order O ε7 is given by

2
α1 B3XX + i α2 B3X + i α3 |A1 | B3X + i α3 Ā1 A1X B3 + i α3 A1 A1X B̄3 + i α4 A21 B̄3X
2 2 4
+ 2 i α4 A1 Ā1X B3 + α5 B3 + α6 A21 B̄3 + 2 α6 |A1 | B3 + 2 α7 |A1 | A21 B̄3 + 3 α7 |A1 | B3
2
+ α1,2 B2XX + i α2,2 B2X + i α3,2 |A1 | B2X + i α3,2 Ā1 A1X B2 + i α3,2 A1 A1X B̄2 + i α4,2 A21 B̄2X
2 2 2
+ 2 i α4,2 A1 Ā1X B2 + α5,2 B2 + α6,2 A21 B̄2 + 2 α6,2 |A1 | B2 + 2 α7,2 |A1 | A21 B̄2 + 3 α7,2 |A1 | A1 B2
+ i α3 Ā1 A2 B2X + i α3 A1 Ā2 B2X + i α3 Ā1 A2X B2 + i α3 A1X Ā2 B2 + i α3 A1 A2X B̄2
+ i α3 A1X A2 B̄2 + 2 i α4 Ā1X A2 B2 + 2 i α4 A1 A2 B̄2X + 2 i α4 A1 Ā2X B2 + 2 α6 Ā1 A2 B2
2 2
+ 2 α6 A1 Ā2 B2 + 2 α6 A1 A2 B̄2 + 2 α7 A31 Ā2 B̄2 + 6 α7 |A1 | Ā1 A2 B2 + 6 α7 |A1 | A1 Ā2 B2
2 2
+ 6 α7 |A1 | A1 A2 B̄2 + i α1,3 B1XXX + α2,3 B1XX + α3,3 |A1 | B1XX + α3,3 Ā1 A1XX B1 + α3,3 A1 A1XX B̄1
2
+ α4,3 A21 B̄1XX + 2 α4,3 A1 Ā1XX B1 + i α5,3 B1X + i α6,3 |A1 | B1X + i α6,3 Ā1 A1X B1
4 2 2
+ i α6,3 A1 A1X B̄1 + i α7,3 |A1 | B1X + 2 i α7,3 |A1 | Ā1 A1X B1 + 2 i α7,3 |A1 | A1 A1X B̄1
2
+ i α8,3 A21 B̄1X + 2 i α8,3 A1 Ā1X B1 + i α9,3 |A1 | A21 B̄1X + i α9,3 A31 Ā1X B̄1
2
+ 3 i α9,3 |A1 | A1 Ā1X B1 + α10,3 A21X B̄1 + 2 α10,3 Ā1 A1X B1X + α11,3 A1 Ā1X B1X
2 2
+ α11,3 A1 A1X B̄1X + α11,3 |A1X | B1 + α12,3 B1 + α13,3 A21 B̄1 + 2 α13,3 |A1 | B1
2 4 4 6
+ 2 α14,3 |A1 | A21 B̄1 + 3 α14,3 |A1 | B1 + 3 α15,3 |A1 | A21 B̄1 + 4 α15,3 |A1 | B1
2
+ 2 α1 B1XΞ + i α2 B1Ξ + i α3 |A1 | B1Ξ + i α3 Ā1 A1Ξ B1 + i α3 A1 A1Ξ B̄1 + i α4 A21 B̄1Ξ
2
+ 2 i α4 A1 Ā1Ξ B1 + i α3 Ā1 A3 B1X + i α3 A1 Ā3 B1X + i α3 |A2 | B1X + i α3 Ā1 A3X B1 + i α3 A1 A3X B̄1
+ i α3 A1X Ā3 B1 + i α3 Ā2 A2X B1 + i α3 A1X A3 B̄1 + i α3 A2 A2X B̄1 + i α4 A22 B̄1X
+ 2 i α4 A1 A3 B̄1X + 2 i α4 A1 Ā3X B1 + 2 i α4 Ā1X A3 B1 + 2 i α4 A2 Ā2X B1 + α6 A22 B̄1
2
+ 2 α6 Ā1 A3 B1 + 2 α6 A1 Ā3 B1 + 2 α6 A1 A3 B̄1 + 2 α6 |A2 | B1 + 2 α7 A31 Ā3 B̄1
2 2
+ 3 α7 Ā21 A22 B1 + 3 α7 A21 Ā22 B1 + 6 α7 |A1 | A1 Ā3 B1 + 6 α7 |A1 | Ā1 A3 B1
2 2 2 2 2
+ 6 α7 |A1 | A1 A3 B̄1 + 6 α7 |A1 | A22 B̄1 + 6 α7 A21 |A2 | B̄1 + 12 α7 |A1 | |A2 | B1
+ i α3,2 Ā1 A2 B1X + i α3,2 A1 Ā2 B1X + i α3,2 Ā1 A2X B1 + i α3,2 A1X Ā2 B1 + i α3,2 A1 A2X B̄1
+ i α3,2 A1X A2 B̄1 + 2 i α4,2 A1 A2 B̄1X + 2 i α4,2 A1 Ā2X B1 + 2 i α4,2 Ā1X A2 B1 + 2 α6,2 Ā1 A2 B1
2
+ 2 α6,2 A1 Ā2 B1 + 2 α6,2 A1 A2 B̄1 + 2 α7,2 A31 Ā2 B̄1 + 6 α7,2 |A1 | Ā1 A2 B1
2 2
+ 6 α7,2 |A1 | A1 Ā2 B1 + 6 α7,2 |A1 | A1 A2 B̄1 = 0. (112)

101