## On the simpleness of zeros of Stokes multipliers

### November 27, 2006

Abstract
The aim of this paper is to discuss the simpleness of zeros of Stokes multipliers associated with the differential equation $-{\Phi }^{\prime \prime }\left(X\right)+W\left(X\right)\Phi \left(X\right)=0$  , where $W\left(X\right)={X}^{m}+{a}_{1}{X}^{m-1}+\cdots +{a}_{m}$  is a real monic polynomial. We show that, under a suitable hypothesis on the coefficients ${a}_{k}$  , all the zeros of the Stokes multipliers are simple.
Keywords: ODEs in the complex plane, Stokes multipliers, entire functions.
On the simpleness of zeros of Stokes multipliers

1. Introduction

We consider in the complex plane a second-order linear differential equation,
 $\begin{array}{c}-{\Phi }^{\prime \prime }\left(X\right)+W\left(X\right)\Phi \left(X\right)=0,\end{array}$ (1)
where $W\left(X\right)={X}^{m}+{a}_{1}{X}^{m-1}+\cdots +{a}_{m}$  is a monic polynomial of degree $m\in \mathbb{N}$  .
Equation ( 1 ) has an irregular singular point at $X=\infty$  , so that the asymptotic behaviors of the solutions at this point usually exhibit Stokes phenomena which are controlled by the Stokes multipliers, in relation with the so-called connection formulae.
These Stokes multipliers measure the lack of commutativity between the analytic continuations of the solutions and their asymptotics near infinity, when crossing the Stokes singular directions.
This justified their systematic study which has been initiated by Sibuya in his book [18and extended in many studies, in particular in the framework of the resurgent asymptotic analysis (see, e.g., [7, 8, 12, 15, 16, 19).
Since equation ( 1 ) depends on the parameter $\left({a}_{1},\cdots ,{a}_{m}\right)$  , this translates to the Stokes multipliers. For a convenient normalization, the Stokes multipliers are in fact holomorphic functions in $\left({a}_{1},\cdots ,{a}_{m}\right)$  , and the question of describing their zeros appears as a natural mathematical question. Specializing this question in the sole parameter ${a}_{m}$  , Sibuya has shown that all but a finite number of the zeros are simple (see [18, ch. 6), by exploring the asymptotic expansion at infinity of the Stokes multipliers with respect to ${a}_{m}$  .
In this text we would like to extend this result, considering the question of the simpleness of all these zeros. This question arises from the fact that the zeros of the Stokes multiplier are nothing but eigenvalues of a (complex) boundary value problem associated with equation ( 1 ). Such a problem has merged recently in the context of the physically well motivated study of the so-called $\mathcal{P}\mathcal{T}$  -symmetric models (see, e.g., [1, 4).
In the paper [21, we have shown that the simpleness of all the zeros of a convenient Stokes multiplier implies the non-degeneration of the eigenstates under the so-called $\mathcal{P}\mathcal{T}$  -pseudo-norm.
It is commonly believed that this property allows to define a conventional mathematical structure for a physically consistent $\mathcal{P}\mathcal{T}$  -symmetric quantum mechanic theory.
This explains the motivation of our present paper, that is to give a complete proof for the simpleness of all the zeros of the Stokes multipliers, under an appropriate condition (from [17) on the parameter $\left({a}_{1},\cdots ,{a}_{m-1}\right)$  so as to impose the reality of the zeros.
As a preparation for the necessary ingredients, the next section will serve to recall some important notions and facts from the theory of Sibuya for equation ( 1 ). The most crucial result is the existence and uniqueness theorem for the solutions characterized by an appropriate asymptotic behavior at infinity. Section 3 provides a detailed proof, in the spirit of [17, for the simpleness of zeros in an instructive special case, where all ${a}_{k}$  (except for the last coefficient ${a}_{m}$  ) are vanishing. A similar result, which holds for some more general cases, is then established by the same arguments. Finally, in the conclusion, we briefly discuss a model exhibiting non-simple real zeros.
On the simpleness of zeros of Stokes multipliers

2. Stokes multipliers

In this section, we briefly recall some classical results of Sibuya's theory on second-order linear differential equations with polynomial coefficients [18. The most fundamental fact is the following theorem, which asserts the existence and uniqueness of a solution characterized by its asymptotic behavior at infinity.
Theorem 1 (Sibuya). Equation ( 1 ) admits a unique solution ${\Phi }_{0}\left(X,a\right):={\Phi }_{0}\left(X,{a}_{1},{a}_{2},\dots ,{a}_{m}\right)$  such that:
• 1. ${\Phi }_{0}\left(X,a\right)$  is an entire function in $\left(X,{a}_{1},{a}_{2},\dots ,{a}_{m}\right)$  ,
• 2. ${\Phi }_{0}\left(X,a\right)$  and its derivative ${\Phi }_{0}^{\prime }\left(X,a\right)$  admit the following asymptotic behaviors  $\begin{array}{c}{\Phi }_{0}\left(X\right)\simeq {X}^{{r}_{m}}{e}^{-S\left(X,a\right)}\left[1+O\left({X}^{-1/2}\right)\right]\end{array}$ (2)
 $\begin{array}{c}{\Phi }_{0}^{\prime }\left(X\right)\simeq {X}^{\frac{m}{2}+{r}_{m}}{e}^{-S\left(X,a\right)}\left[-1+O\left({X}^{-1/2}\right)\right]\end{array}$ (3)
when $X\to \infty$  in each sub-sector strictly contained in the sector ${\Sigma }_{0}=\left\{|arg\left(X\right)|<\frac{3\pi }{m+2}\right\}$  and the asymptotic regimes occur uniformly with respect to $a=\left({a}_{1},{a}_{2},\dots ,{a}_{m}\right)$  in any compact of ${\mathbb{C}}^{m}$  .
In the above theorem ${r}_{m}$  and $S\left(X,a\right)$  can be determined explicitly from $W\left(X\right)$  . More concretely, as $X\to \infty$  , one can write
 $\begin{array}{c}\begin{array}{cc}\sqrt{W\left(X\right)}& ={X}^{\frac{m}{2}}{\left\{1+{a}_{1}{X}^{-1}+\cdots +{a}_{m}{X}^{-m}\right\}}^{1/2}\\ & ={X}^{\frac{m}{2}}\left\{1+{\sum }_{k=1}^{\infty }{b}_{k}\left(a\right){X}^{-k}\right\}\\ \end{array}\end{array}$ (4)
where, obviously, ${b}_{k}\left(a\right)$  are quasi-homogeneous polynomials in ${a}_{1},\dots ,{a}_{m}$  with real coefficients.
By integrating term-by-term the series in the right-hand side, we get a primitive of $\sqrt{W\left(X\right)}$  .
The function $S\left(X,a\right)$  is associated to the “principal part” of this primitive $S\left(X,a\right)=\frac{2}{m+2}{X}^{\frac{m+2}{2}}+\cdots$  that only contains terms with strictly positive powers of $X$  . And ${r}_{m}={r}_{m}\left(a\right)$  is defined by
 $\begin{array}{c}{r}_{m}\left(a\right)=\left\{\begin{array}{cc}-m/4& formodd\\ -m/4-{b}_{1+m/2}\left(a\right)& formeven\\ \end{array}\end{array}$ (5)
We should notice that for $m>2$  , ${r}_{m}\left(a\right)$  does not depend on the last coefficient ${a}_{m}$  and if all ${a}_{j}$  (possibly except ${a}_{m}$  ) are equal to zero then ${r}_{m}=-m/4$  .
On the simpleness of zeros of Stokes multipliers We shall define other solutions of ( 1 ) by introducing a rotation of the complex plan. Let us denote $\omega :={e}^{\frac{i2\pi }{m+2}}and{\omega }_{k}\left(a\right):=\left({\omega }^{k}{a}_{1},{\omega }^{2k}{a}_{2},\dots ,{\omega }^{km}{a}_{m}\right);\left(k\in \mathbb{Z}\right)$  For each $k\in \mathbb{Z}$  , we construct functions ${\Phi }_{k}\left(X,a\right)$  by setting
 $\begin{array}{c}{\Phi }_{k}\left(X,a\right):={\Phi }_{0}\left({\omega }^{-k}X,{\omega }_{-k}\left(a\right)\right).\end{array}$ (6)
It is not difficult to check that ${\Phi }_{k}\left(X,a\right)$  are indeed solutions of ( 1 ) and exponentially vanishing at infinity in the corresponding sector
 $\begin{array}{c}{S}_{k}:=\left\{|arg\left(X\right)-\frac{k2\pi }{m+2}|<\frac{\pi }{m+2}\right\}.\end{array}$ (7)
The following lemma, which can be verified in a straightforward way (see [12, 18), implies the linear independence of two consecutive solutions ${\Phi }_{k}$  and ${\Phi }_{k+1}$  .
Lemma 2. For any $k\in \mathbb{Z}$  , the Wronskian of ${\Phi }_{k}$  and ${\Phi }_{k+1}$  is given by the formula
 $\begin{array}{c}\mathsf{W}\mathsf{r}\left({\Phi }_{k},{\Phi }_{k+1}\right)=2\left(-1{\right)}^{k}{\omega }^{\frac{km}{2}-{r}_{m}\left({\omega }_{-k-1}\left(a\right)\right)}\end{array}$ (8)
From this observation, together with classical results on the structure of solutions of linear differential equations, we can infer that $\left\{{\Phi }_{k},{\Phi }_{k+1}\right\}$  constitutes a basis for the space of solutions of equation ( 1 ). Therefore, every solution can be expressed as a linear combination of ${\Phi }_{k},{\Phi }_{k+1}$  . In particular, for each $k\in \mathbb{Z}$  , we have
 $\begin{array}{c}{\Phi }_{k-1}={C}_{k}\left(a\right){\Phi }_{k}+{\stackrel{~}{C}}_{k}\left(a\right){\Phi }_{k+1}\end{array}$ (9)
The multipliers ${C}_{k}\left(a\right)$  and ${\stackrel{~}{C}}_{k}\left(a\right)$  are called the Stokes multipliers of ${\Phi }_{k-1}$  with respect to ${\Phi }_{k}$  and ${\Phi }_{k+1}$  . Further studies on these objects are addressed in [12, 15, 18. By definition, it is evident that ${C}_{k}\left(a\right)=\frac{\mathsf{W}\mathsf{r}\left({\Phi }_{k-1},{\Phi }_{k+1}\right)}{\mathsf{W}\mathsf{r}\left({\Phi }_{k},{\Phi }_{k+1}\right)}and{\stackrel{~}{C}}_{k}\left(a\right)=\frac{\mathsf{W}\mathsf{r}\left({\Phi }_{k-1},{\Phi }_{k}\right)}{\mathsf{W}\mathsf{r}\left({\Phi }_{k+1},{\Phi }_{k}\right)}$  Since ${\Phi }_{k}\left(X,a\right)$  are entire functions, it follows immediately from these equalities and Lemma  2 that ${C}_{k}\left(a\right)$  and ${\stackrel{~}{C}}_{k}\left(a\right)$  are also entire functions in $a$  . By definition, ${C}_{k}$  is closely related to ${C}_{0}$  in a “cyclic” way through the formula:
${C}_{k}\left(a\right)={C}_{0}\left({\omega }_{k}\left(a\right)\right),$  by which the information about ${C}_{k}$  can be derived from ${C}_{0}$  . Furthermore, we also get an explicit expression for ${\stackrel{~}{C}}_{k}\left(a\right)$  On the simpleness of zeros of Stokes multipliers ${\stackrel{~}{C}}_{k}\left(a\right)={\omega }^{-m-2{r}_{m}\left({\omega }_{-k}\left(a\right)\right)}.$  We emphasize that ${\stackrel{~}{C}}_{k}\left(a\right)$  is never vanishing. It thus can be reduced to 1 by a suitable renormalisation of the ${\Phi }_{k}$  's. To do this, it is sufficient to insert a simple factor in the right hand side of ( 6 ). For instance, when $k=0$  , by redefining
 $\begin{array}{c}\begin{array}{cc}& {\Phi }_{1}\left(X,a\right):={\omega }^{-m/2-{r}_{m}\left(a\right)}{\Phi }_{0}\left({\omega }^{-1}X,{\omega }_{-1}\left(a\right)\right)\\ and& {\Phi }_{-1}\left(X,a\right):={\omega }^{m/2+{r}_{m}\left(a\right)}{\Phi }_{0}\left(\omega X,\omega \left(a\right)\right)\\ \end{array}\end{array}$ (10)
we can write ( 9 ) under a slightly symmetric form,
 $\begin{array}{c}{\Phi }_{-1}=C\left(a\right){\Phi }_{0}+{\Phi }_{1},\end{array}$ (11)
where $C\left(a\right):={\omega }^{m/2+{r}_{m}\left(a\right)}{C}_{0}\left(a\right)$  is also called the Stokes multiplier of ${\Phi }_{-1}$  with respect to ${\Phi }_{0}$  .
Concerning this (sole) Stokes multiplier $C\left(a\right)$  , whose zeros are expected to be simple, we first have:
Proposition 3. For any $a\in {\mathbb{C}}^{m}$  ,
 $\begin{array}{c}\overline{C\left(a\right)}+C\left(\overline{a}\right)=0.\end{array}$ (12)
• Proof. By virtue of the quasi-homogeneity of equation ( 1 ), we can see that $\overline{{\Phi }_{0}\left(\overline{X},\overline{a}\right)}$  is also one of its solutions whose asymptotic behavior at infinity in the sector ${S}_{0}$  is the same as that of ${\Phi }_{0}\left(X,a\right)$  .
The uniqueness of the canonical solution in Theorem  1 implies immediately that  $\begin{array}{c}\overline{{\Phi }_{0}\left(X,a\right)}={\Phi }_{0}\left(\overline{X},\overline{a}\right).\end{array}$ (13)
Taking into account the above new definitions of ${\Phi }_{±1}$  in ( 10 ), we can check without difficulty that  $\begin{array}{c}{\Phi }_{-1}\left(\overline{X},\overline{a}\right)=\overline{{\Phi }_{1}\left(X,a\right)}\end{array}$ (14)
for any $X\in \mathbb{C}$  and any $a\in {\mathbb{C}}^{m}$  .
Putting these relations in ( 11 ) leads the desired identity.
Corollary 3.1. The zero set of $C\left(a\right)$  is invariant under the complex conjugation $a↦\overline{a}$  .
Corollary 3.2. Restricted on real coordinates $a=\left({a}_{1},...,{a}_{m}\right)\in {\mathbb{R}}^{m}$  , $-iC\left(a\right)$  is a real-valued function.
On the simpleness of zeros of Stokes multipliers

3. The simpleness of zeros of Stokes multiplier

In this section, we shall discuss the simpleness of zeros of $C\left(a\right)$  considered as a function of the last coefficient ${a}_{m}$  . For convenience, we consider $\lambda :={a}_{m}$  as variable of the entire function $C\left(a,\lambda \right):=C\left({a}_{1},\dots ,{a}_{m-1},\lambda \right)$  . We then show that under some hypotheses on ${a}_{1},{a}_{2},\dots ,{a}_{m-1}$  , the derivative $\frac{\partial }{\partial \lambda }C\left(a,\lambda \right)\ne 0$  if $C\left(a,\lambda \right)=0$  .
On the simpleness of zeros of Stokes multipliers

3.1. A special case: ${a}_{1}=0,\dots ,{a}_{m-1}=0$

We first concentrate on the case where all ${a}_{j}$  are vanishing except for ${a}_{m}=:\lambda$  . Equation ( 1 ) now simply reads
 $\begin{array}{c}-{\Phi }^{\prime \prime }\left(X\right)+\left({X}^{m}+\lambda \right)\Phi \left(X\right)=0\end{array}$ (15)
Even for this simple case, the study of the zeros of the Stokes multiplier is interesting because it has an intimate relation with the spectral analysis problem of Hamiltonians whose potentials are (possibly complex) homogeneous polynomials (see [3, 5, 22, 23).
We shall use all the notations of the previous section, with some minor modifications. By virtue of Sibuya's theorem, equation ( 15 ) possesses a unique solution ${\Phi }_{0}\left(X,\lambda \right)$  , which is an entire function in both $X$  and $\lambda$  . For each fixed $\lambda \in \mathbb{C}$  , this solution and its derivative with respect to $X$  satisfy the following asymptotic estimates
 $\begin{array}{c}\begin{array}{cc}& {\Phi }_{0}\left(X,\lambda \right)\simeq {X}^{-m/4}{e}^{-\frac{2}{m+2}{X}^{\frac{m+2}{2}}}\left[1+O\left({X}^{-1/2}\right)\right]\\ and& {\Phi }_{0}^{\prime }\left(X,\lambda \right)\simeq {X}^{m/4}{e}^{-\frac{2}{m+2}{X}^{\frac{m+2}{2}}}\left[-1+O\left({X}^{-1/2}\right)\right]\\ \end{array}as{S}_{0}\ni X\to \infty \end{array}$ (16)
We should notice that, both ${\Phi }_{0}$  and ${\Phi }_{0}^{\prime }$  vanish exponentially at infinity in the sector ${S}_{0}$  .
Besides, as an entire function of $X$  , ${\Phi }_{0}\left(X,\lambda \right)$  can be expanded in powers of $X$  ,
 $\begin{array}{c}{\Phi }_{0}\left(X,\lambda \right)={f}_{0}\left(\lambda \right)+{\sum }_{j=1}^{\infty }{f}_{j}\left(\lambda \right){X}^{j},\end{array}$ (17)
where ${f}_{j}\left(\lambda \right)$  are also entire functions of $\lambda$  and satisfy the symmetry property $\overline{{f}_{j}\left(\lambda \right)}={f}_{j}\left(\overline{\lambda }\right),j=0,1,2,\dots$  The companion solutions ${\Phi }_{±1}$  of ${\Phi }_{0}$  now read as follows:
 $\begin{array}{c}\begin{array}{c}{\Phi }_{1}\left(X,\lambda \right)={\omega }^{-m/4}{\Phi }_{0}\left({\omega }^{-1}X,{\omega }^{2}\lambda \right)={\omega }^{-m/4}\left\{{f}_{0}\left({\omega }^{2}\lambda \right)+{\sum }_{j=1}^{\infty }{f}_{j}\left({\omega }^{2}\lambda \right)\left({\omega }^{-1}X{\right)}^{j}\right\}\\ {\Phi }_{-1}\left(X,\lambda \right)={\omega }^{m/4}{\Phi }_{0}\left(\omega X,{\omega }^{-2}\lambda \right)={\omega }^{m/4}\left\{{f}_{0}\left({\omega }^{-2}\lambda \right)+{\sum }_{j=1}^{\infty }{f}_{j}\left({\omega }^{-2}\lambda \right)\left(\omega X{\right)}^{j}\right\}\\ \end{array}\end{array}$ (18)
Substituting these expressions into ( 11 ) and specializing that equality with $X=0$  lead the following relation between the Stokes multiplier $C\left(\lambda \right)$  and the entire function ${f}_{0}\left(\lambda \right)$  :
 $\begin{array}{c}C\left(\lambda \right){f}_{0}\left(\lambda \right)={\omega }^{m/4}{f}_{0}\left({\omega }^{-2}\lambda \right)-{\omega }^{-m/4}{f}_{0}\left({\omega }^{2}\lambda \right).\end{array}$ (19)
This intimate relation can serve to study the zeros of $C\left(\lambda \right)$  through those of ${f}_{0}\left(\lambda \right)$  . Before going further, we recall a consequence from Sibuya's asymptotic studies on these functions ([18, ch. $4,5$  ).
Proposition 4. The orders of entire functions ${f}_{0}\left(\lambda \right)$  and $C\left(\lambda \right)$  are both equal to $\frac{1}{2}+\frac{1}{m}$  .
Remark 5. For $m\ge 3$  , the order is not an integer. Therefore, ${f}_{0}\left(\lambda \right)$  and $C\left(\lambda \right)$  must have infinitely many zeros, whose accumulation point can only be infinity.
On the simpleness of zeros of Stokes multipliers
The following assertion locates the zeros of ${f}_{0}\left(\lambda \right)$  more concretely.
Proposition 6. All the zeros of ${f}_{0}\left(\lambda \right)$  are negative real numbers. Moreover, ${f}_{0}\left(\lambda \right)>0,\forall \lambda \ge 0$
Before proving the proposition, we need to remind a useful transform for a given second-order linear differential equation in the complex plane. Let $w\left(z\right)$  be a solution of the following equation,
 $\begin{array}{c}{w}^{\prime \prime }\left(z\right)-f\left(z\right)w\left(z\right)=0.\end{array}$ (20)
Then for any ${z}_{1},{z}_{2}\in \mathbb{C}$  , we have the following identity by multiplying ( 20 ) by $\overline{w\left(z\right)}$  and integrating it from ${z}_{1}$  to ${z}_{2}$  :
 $\begin{array}{c}\overline{w\left(z\right)}{w}^{\prime }\left(z\right){|}_{{z}_{1}}^{{z}_{2}}:=\overline{w\left({z}_{2}\right)}{w}^{\prime }\left({z}_{2}\right)-\overline{w\left({z}_{1}\right)}{w}^{\prime }\left({z}_{1}\right)={\int }_{{z}_{1}}^{{z}_{2}}|{w}^{\prime }\left(z\right){|}^{2}d\overline{z}+{\int }_{{z}_{1}}^{{z}_{2}}f\left(z\right)|w\left(z\right){|}^{2}dz\end{array}$ (21)
This identity, which is known as the Green's transform of ( 20 ), is true provided that the integrals in the right-hand side of ( 21 ) make sense.
In case where the integral path is a segment $\left[{z}_{1},{z}_{2}\right]=\left\{z\left(t\right)={z}_{1}+t{e}^{i\theta }/t\in \left[0,r\right]\right\}$  , with $\theta =arg\left({z}_{2}-{z}_{1}\right)$  and $r=|{z}_{2}-{z}_{1}|$  , then ( 21 ) turns into
 $\begin{array}{c}\overline{w\left(z\right)}{w}^{\prime }\left(z\right){|}_{{z}_{1}}^{{z}_{2}}={e}^{-i\theta }{\int }_{0}^{r}|{w}^{\prime }\left(z\left(t\right)\right){|}^{2}dt+{e}^{i\theta }{\int }_{0}^{r}|w\left(z\left(t\right)\right){|}^{2}f\left(z\left(t\right)\right)dt.\end{array}$ (22)
Besides, this equality holds true as $r\to +\infty$  , provided all limits exist.
Proof of Proposition  6 . Let ${\lambda }^{*}=\alpha +i\beta$  be a zero of ${f}_{0}\left(\lambda \right)={\Phi }_{0}\left(0,\lambda \right)$  . By applying the Green's transform ( 22 ) on the interval $\left[0,X\right]\subset \mathbb{R}$  to the solution ${\Phi }_{0}$  of ( 15 ) , we obtain
 $\begin{array}{c}\overline{{\Phi }_{0}\left(t,\lambda \right)}{\Phi }_{0}^{\prime }\left(t,\lambda \right){|}_{0}^{X}={\int }_{0}^{X}|{\Phi }_{0}^{\prime }\left(t,\lambda \right){|}^{2}dt+{\int }_{0}^{X}\left({t}^{m}+\lambda \right)|{\Phi }_{0}\left(t,\lambda \right){|}^{2}dt.\end{array}$ (23)
Substituting $\lambda ={\lambda }^{*}$  into ( 23 ) and letting $X\to +\infty$  yield $0={\int }_{0}^{+\infty }|{\Phi }_{0}^{\prime }\left(t,{\lambda }^{*}\right){|}^{2}dt+{\int }_{0}^{+\infty }\left({t}^{m}+\alpha +i\beta \right)|{\Phi }_{0}\left(t,{\lambda }^{*}\right){|}^{2}dt$  By separating the real and imaginary parts, we obtain $\beta =0and\alpha =-\frac{{\int }_{0}^{+\infty }|{\Phi }_{0}^{\prime }\left(t,{\lambda }^{*}\right){|}^{2}dt+{\int }_{0}^{+\infty }{t}^{m}|{\Phi }_{0}\left(t,{\lambda }^{*}\right){|}^{2}dt}{{\int }_{0}^{+\infty }|{\Phi }_{0}\left(t,{\lambda }^{*}\right){|}^{2}dt}<0\mathsf{O}\mathsf{n}\mathsf{t}\mathsf{h}\mathsf{e}\mathsf{s}\mathsf{i}\mathsf{m}\mathsf{p}\mathsf{l}\mathsf{e}\mathsf{n}\mathsf{e}\mathsf{s}\mathsf{s}\mathsf{o}\mathsf{f}\mathsf{z}\mathsf{e}\mathsf{r}\mathsf{o}\mathsf{s}\mathsf{o}\mathsf{f}\mathsf{S}\mathsf{t}\mathsf{o}\mathsf{k}\mathsf{e}\mathsf{s}\mathsf{m}\mathsf{u}\mathsf{l}\mathsf{t}\mathsf{i}\mathsf{p}\mathsf{l}\mathsf{i}\mathsf{e}\mathsf{r}\mathsf{s}$  For the rest of the proposition, we consider a fixed $\lambda \ge 0$  and treat $X$  as a real variable.
Hence, equality ( 13 ) implies that ${\Phi }_{0}\left(X,\lambda \right)$  is a real-valued function of $X\in \mathbb{R}$  , and so is also its derivative ${\Phi }_{0}^{\prime }\left(X,\lambda \right)$  . By taking the derivative of ( 23 ) with respect to $X$  , we get ${\left(\overline{{\Phi }_{0}\left(X,\lambda \right)}{\Phi }_{0}^{\prime }\left(X,\lambda \right)\right)}_{X}^{\prime }=|{\Phi }_{0}^{\prime }\left(X,\lambda \right){|}^{2}+\left({X}^{m}+\lambda \right)|{\Phi }_{0}\left(X,\lambda \right){|}^{2}.$  Since $\lambda \ge 0$  , the right-hand side of this equality is strictly positive on $\left[0,+\infty \right)$  . It follows that the real function ${\Phi }_{0}\left(X,\lambda \right){\Phi }_{0}^{\prime }\left(X,\lambda \right)$  is strictly increasing on $\left[0,+\infty \right)$  . Moreover, we deduce from ( 16 ) that ${lim}_{X\to +\infty }{\Phi }_{0}\left(X,\lambda \right){\Phi }_{0}^{\prime }\left(X,\lambda \right)=0$  Combining these facts implies that both ${\Phi }_{0}\left(X,\lambda \right)$  and ${\Phi }_{0}^{\prime }\left(X,\lambda \right)$  never vanish on $\left[0,+\infty \right)$  .
In particular, regarding its asymptotic asymptotic behavior in ( 16 ), we can conclude that ${\Phi }_{0}\left(X,\lambda \right)>0$  for all $X\ge 0$  . Putting $X=0$  completes the proof. $\square$  Concerning the zeros of $C\left(\lambda \right)$  , we have the following.
Theorem 7. All zeros of $C\left(\lambda \right)$  are real, positive and simple.
We should remind that the reality of all these zeros, in connection with $\mathcal{P}\mathcal{T}$  -symmetric quantum mechanics, has been studied in various ways by many authors [3, 10, 11, 13, 17, 20. The simpleness of all but a finite number of zeros has been indicated in [18(ch. $6$  ) by using some asymptotic estimates for the large zeros.
Next, we shall provide a rigorous proof for the simpleness of all zeros using ideas from the proof of Laguerre's theorem [6, after simply justifying the reality and positivity in our special case by the same way as in [17.
Proof of Theorem  7 . Let $\left\{{\lambda }_{n}{\right\}}_{n=\overline{0,\infty }}$  be zeros of the entire function ${f}_{0}\left(\lambda \right)$  . Applying Hadamard's factorization theorem ([6) to ${f}_{0}\left(\lambda \right)$  , whose order is smaller than $1$  for $m>2$  , we have, for all $\lambda \in \mathbb{C}$  ,
 $\begin{array}{c}{f}_{0}\left(\lambda \right)={A}^{\infty }{\prod }_{n=0}\left(1-\frac{\lambda }{{\lambda }_{n}}\right)\end{array}$ (24)
where $A={f}_{0}\left(0\right)>0$  .
Suppose that ${\lambda }^{*}$  is a zero of $C\left(\lambda \right)$  . We can deduce from ( 19 ) that
 $\begin{array}{c}A{{\omega }^{m/4}}^{\infty }{\prod }_{n=0}\left(1-\frac{{\omega }^{-2}{\lambda }^{*}}{{\lambda }_{n}}\right)=A{{\omega }^{-m/4}}^{\infty }{\prod }_{n=0}\left(1-\frac{{\omega }^{2}{\lambda }^{*}}{{\lambda }_{n}}\right)\end{array}$ (25)
By virtue of Proposition  6 , all the zeros ${\lambda }_{n}$  of ${f}_{0}\left(\lambda \right)$  are negative. Combining this with the fact that $0<|arg\left({\omega }^{±2}\right)|<\pi$  , we can conclude that ${f}_{0}\left({\omega }^{±2}{\lambda }^{*}\right)$  cannot be simultaneously vanishing. Therefore, both of sides of ( 25 ) are never vanishing.
By taking the absolute value of both sides and remarking $|z|=|\overline{z}|$  , we obtain
 $\begin{array}{c}{}^{\infty }{\prod }_{n=0}|\frac{{\omega }^{2}{\lambda }_{n}-{\lambda }^{*}}{{\omega }^{2}{\lambda }_{n}-\overline{{\lambda }^{*}}}|=1\mathsf{O}\mathsf{n}\mathsf{t}\mathsf{h}\mathsf{e}\mathsf{s}\mathsf{i}\mathsf{m}\mathsf{p}\mathsf{l}\mathsf{e}\mathsf{n}\mathsf{e}\mathsf{s}\mathsf{s}\mathsf{o}\mathsf{f}\mathsf{z}\mathsf{e}\mathsf{r}\mathsf{o}\mathsf{s}\mathsf{o}\mathsf{f}\mathsf{S}\mathsf{t}\mathsf{o}\mathsf{k}\mathsf{e}\mathsf{s}\mathsf{m}\mathsf{u}\mathsf{l}\mathsf{t}\mathsf{i}\mathsf{p}\mathsf{l}\mathsf{i}\mathsf{e}\mathsf{r}\mathsf{s}\end{array}$ (26)
Since ${\lambda }_{n}<0$  , we have $Im\left({\lambda }_{n}{\omega }^{2}\right)<0,\forall n$  . These inequalities imply that, unless ${\lambda }^{*}=\overline{{\lambda }^{*}}$  , the factors in the left hand side of ( 26 ) are all either strictly greater or less than $1$  . Consequently, the truth of the equality ( 26 ) requires that ${\lambda }^{*}\in \mathbb{R}$  .
To verify that ${\lambda }^{*}>0$  , we use the Green's transform again. Applying ( 22 ) to the solution ${\Phi }_{1}\left(X,{\lambda }^{*}\right)$  of the equation ( 15 ) on the ray $\left[0,\omega \infty \right)$  , we obtain
 $\begin{array}{c}\overline{{\Phi }_{1}\left(X,{\lambda }^{*}\right)}{\Phi }_{1}^{\prime }\left(X,{\lambda }^{*}\right){|}_{0}^{\omega \infty }={\omega }^{-1}{\int }_{0}^{\infty }|{\Phi }_{1}^{\prime }\left(\omega t,{\lambda }^{*}\right){|}^{2}dt+\omega {\int }_{0}^{\infty }|{\Phi }_{1}\left(\omega t,{\lambda }^{*}\right){|}^{2}\left(\left(\omega t{\right)}^{m}+{\lambda }^{*}\right)dt\end{array}$ (27)
Note that, by definition, ${\Phi }_{1}\left(X,{\lambda }^{*}\right)$  and its derivative ${\Phi }_{1}^{\prime }\left(X,{\lambda }^{*}\right)$  are exponentially vanishing at $\omega \infty$  . On the other hand, since ${\lambda }^{*}$  is a (real) zero of $C\left(\lambda \right)$  , we can deduce from ( 11 ) and ( 14 ) that ${\Phi }_{1}\left(X,{\lambda }^{*}\right)={\Phi }_{-1}\left(X,{\lambda }^{*}\right)=\overline{{\Phi }_{1}\left(\overline{X},\overline{{\lambda }^{*}}\right)}$  This implies that the left-hand side of ( 27 ), which now becomes $-\overline{{\Phi }_{1}\left(0,{\lambda }^{*}\right)}{\Phi }_{1}^{\prime }\left(0,{\lambda }^{*}\right)$  , is purely real.
Separating the imaginary part in ( 27 ), where $\omega ={e}^{i\theta }$  and $\theta =\frac{2\pi }{m+2}$  , we obtain $0=-sin\theta {\int }_{0}^{\infty }|{\Phi }_{1}^{\prime }\left(\omega t,{\lambda }^{*}\right){|}^{2}dt-sin\theta {\int }_{0}^{\infty }|{\Phi }_{1}\left(\omega t,{\lambda }^{*}\right){|}^{2}{t}^{m}dt+{\lambda }^{*}sin\theta {\int }_{0}^{\infty }|{\Phi }_{1}\left(\omega t,{\lambda }^{*}\right){|}^{2}dt$  This equality indicates the positivity of ${\lambda }^{*}$  .
To finish the demonstration, we have to show that ${C}^{\prime }\left(\lambda \right)\ne 0$  if $C\left(\lambda \right)=0$  . Since all zeros of $C\left(\lambda \right)$  are real according to the above proofs, it is sufficient to consider $\lambda$  as a real variable.
We now turn to ( 19 ). By setting $g\left(\lambda \right)={\omega }^{m/4}{f}_{0}\left({\omega }^{-2}\lambda \right)$  and regarding it as a complex-valued function of the real variable $\lambda$  , we can write $g\left(\lambda \right)={g}_{1}\left(\lambda \right)+i{g}_{2}\left(\lambda \right)$  where ${g}_{1,2}\left(\lambda \right)$  are real and differentiable functions on $\mathbb{R}$  .
On account of the relation ( 19 ), we see that a double zero ${\lambda }^{*}\in \mathbb{R}$  of $C\left(\lambda \right)$  must satisfy $g\left({\lambda }^{*}\right)=\overline{g\left({\lambda }^{*}\right)}and{g}^{\prime }\left({\lambda }^{*}\right)=\overline{{g}^{\prime }\left({\lambda }^{*}\right)}.$  This is equivalent to requiring that On the simpleness of zeros of Stokes multipliers ${g}_{2}\left({\lambda }^{*}\right)=0and{g}_{2}^{\prime }\left({\lambda }^{*}\right)=0,$  that is (since we know that $g\left({\lambda }^{*}\right)\ne 0$  ),
 $\begin{array}{c}\frac{{g}^{\prime }\left({\lambda }^{*}\right)}{g\left({\lambda }^{*}\right)}\in \mathbb{R}.\end{array}$ (28)
But, as indicated below, this is impossible.
Indeed, using ( 24 ), we can factorize $g\left(\lambda \right)$  as follows, for $\lambda \in \mathbb{R}$  , $g\left(\lambda \right)=A{{\omega }^{m/4}}^{\infty }{\prod }_{n=0}\left(1-\frac{{\omega }^{-2}\lambda }{{\lambda }_{n}}\right)$  This factorization also indicates that $g\left(\lambda \right)\ne 0$  on $\mathbb{R}$  . Taking the logarithmic derivative on both sides of the equality yields
 $\begin{array}{c}\frac{{g}^{\prime }\left(\lambda \right)}{g\left(\lambda \right)}={\sum }_{n=0}^{\infty }\frac{1}{\lambda -{\lambda }_{n}{\omega }^{2}}.\end{array}$ (29)
Moreover, we have ${\lambda }_{n}<0$  . Hence, for all $\lambda \in \mathbb{R}$  , the imaginary part of the right-hand side of ( 29 ) is $sin2\theta {\sum }_{n=0}^{\infty }\frac{{\lambda }_{n}}{\left(\lambda -{\lambda }_{n}cos2\theta {\right)}^{2}+\left({\lambda }_{n}sin2\theta {\right)}^{2}}$  and it is never vanishing; against ( 28 ). This completes the proof. $\square$
Remark 8. By following the above proof, we can conclude that all the zeros of ${f}_{0}\left(\lambda \right)$  are also simple, in connection with Proposition  6 affirming the reality of these zeros.
As a consequence of this theorem, the real-valued function $-i{C}^{\prime }\left(\lambda \right)$  changes its sign alternately at zeros of $C\left(\lambda \right)$  . This matter has been mentioned in [21as an attempt to justify the indefiniteness of $\mathcal{P}\mathcal{T}$  -pseudo-norm in $\mathcal{P}\mathcal{T}$  -symmetric quantum mechanics.
On the simpleness of zeros of Stokes multipliers

3.2. General case

In what follows, we proceed with the study of the zeros of the Stokes multiplier in the case where ${a}_{k}$  are not simultaneously equal to zero. Since we shall use again the previous arguments, a brief recall of the notions should be done.
For our goal, we consider the following equation:
 $\begin{array}{c}-{\Phi }^{\prime \prime }\left(X\right)+\left({X}^{m}+{a}_{1}{X}^{m-1}+\cdots +{a}_{m-1}X+\lambda \right)\Phi \left(X\right)=0\end{array}$ (30)
We start with one of its solutions ${\Phi }_{0}\left(X,a,\lambda \right)$  , whose existence and asymptotic behavior have been settled in the Sibuya's theorem  1 . As an entire function of $X$  , ${\Phi }_{0}$  can be written in the form
 $\begin{array}{c}{\Phi }_{0}\left(X,a,\lambda \right)={f}_{0}\left(a,\lambda \right)+{\sum }_{j=1}^{\infty }{f}_{j}\left(a,\lambda \right){X}^{j}\end{array}$ (31)
where ${f}_{0}\left(a,\lambda \right)$  is an entire function in both $a=\left({a}_{1},\dots ,{a}_{m-1}\right)$  and $\lambda$  . In particular, considered as an entire function in $\lambda$  , ${f}_{0}\left(a,\lambda \right)$  is of order $\frac{1}{2}+\frac{1}{m}$  (uniformly in $a$  for $a$  in a compact set). The companions solutions ${\Phi }_{±1}$  are defined by
 $\begin{array}{c}\begin{array}{c}{\Phi }_{1}\left(X,a,\lambda \right)={\omega }^{-m/2-{r}_{m}\left(a\right)}{\Phi }_{0}\left({\omega }^{-1}X,{\omega }_{-1}\left(a\right),{\omega }^{-m}\lambda \right)\\ {\Phi }_{-1}\left(X,a,\lambda \right)={\omega }^{m/2+{r}_{m}\left(a\right)}{\Phi }_{0}\left(\omega X,{\omega }_{1}\left(a\right),{\omega }^{m}\lambda \right)\\ \end{array}\end{array}$ (32)
The relation among these three solutions is realized by the Stokes multiplier $C\left(a,\lambda \right)$  :
${\Phi }_{-1}\left(X,a,\lambda \right)=C\left(a,\lambda \right){\Phi }_{0}\left(X,a,\lambda \right)+{\Phi }_{1}\left(X,a,\lambda \right).$  Putting $X=0$  in this equality, together with ( 31 ), we obtain
 $\begin{array}{c}C\left(a,\lambda \right){f}_{0}\left(a,\lambda \right)={\omega }^{m/2+{r}_{m}\left(a\right)}{f}_{0}\left({\omega }_{1}\left(a\right),{\omega }^{m}\lambda \right)-{\omega }^{-m/2-{r}_{m}\left(a\right)}{f}_{0}\left({\omega }_{-1}\left(a\right),{\omega }^{-m}\lambda \right).\end{array}$ (33)
For a fixed $a=\left({a}_{1},...,{a}_{m-1}\right)\in {\mathbb{R}}^{m-1}$  , we are examining the zeros of the first term $\text{1}$  in the right-hand side of ( 33 ) ${g}_{a}\left(\lambda \right):={\omega }^{m/2+{r}_{m}\left(a\right)}{f}_{0}\left({\omega }_{1}\left(a\right),{\omega }^{m}\lambda \right).$  Note that ${g}_{a}\left(\lambda \right)$  is also an entire function of order $\frac{1}{2}+\frac{1}{m}$  , so for $m>2$  , the order is non-integral. It follows that ${g}_{a}\left(\lambda \right)$  must have an infinite number of zeros.
Let ${\lambda }_{n}={\lambda }_{n}\left(a\right)$  , $n\in \mathbb{N}$  , be the zeros of ${g}_{a}\left(\lambda \right)$  . Then ${\Phi }_{0}\left(0,{\omega }_{1}\left(a\right),{\omega }^{m}{\lambda }_{n}\right)=0$  , where $Y\left(X\right):={\Phi }_{0}\left(X,{\omega }_{1}\left(a\right),{\omega }^{m}{\lambda }_{n}\right)$  verifies the following equation
 $\begin{array}{c}-{Y}^{\prime \prime }\left(X\right)+\left({X}^{m}+{a}_{1}{\omega }^{1}{X}^{m-1}+\cdots +{a}_{m-1}{\omega }^{m-1}X+{\lambda }_{n}{\omega }^{m}\right)Y\left(X\right)=0\end{array}$ (34)
We now briefly discuss a significant result of Shin on the reality of the zeros of $C\left(a,\lambda \right)$  in [17, where the author give a suitable hypothesis on the real coefficients ${a}_{k}$  under which all zeros of $C\left(a,\lambda \right)$  are showed to be real and positive. That hypothesis reads On the simpleness of zeros of Stokes multipliers
 $\begin{array}{c}\text{There exists an index}j,1\le j\le m/2\text{such that}\left(j-k\right){a}_{k}\ge 0\text{for all}k.\end{array}$ (H )
One of the crucial steps in his proof is to effectively apply the Green's transform to ( 34 ) on a suitably chosen ray, so that all the imaginary parts of ${\lambda }_{n}$  are non-positive. For our present purpose, we require that all $Im{\lambda }_{n}$  be strictly negative. This fact can actually be derived from the hypothesis $\left(H\right)$  thanks to Shin's proof itself in the cited article, the special case when $m=4$  and $j=2$  being apart. To overcome it, we only need to add a negligible supplement on $\left(H\right)$  as follows
 $\begin{array}{c}\text{If}\left(H\right)\text{occurs for}m=4\mathsf{a}\mathsf{n}\mathsf{d}j=2,\mathsf{t}\mathsf{h}\mathsf{e}\mathsf{n}{a}_{2}\le 0.\end{array}$ (s)
Indeed, for $m=4$  the hypothesis $\left(H\right)$  and its supplement $\left(s\right)$  imply that ${a}_{1}\ge 0$  and ${a}_{2,3}\le 0$  .
Applying ( 22 ) to ( 34 ) on the ray $\left[0,{e}^{i\theta }\infty \right)$  , where $\omega ={e}^{\frac{i\pi }{3}}$  and $|\theta |<\frac{\pi }{6}$  , we obtain
 $\begin{array}{ccc}0& =& {e}^{-i\theta }{\int }_{0}^{+\infty }|{Y}^{\prime }{|}^{2}dt+\end{array}$
 $\begin{array}{ccc}& & +{e}^{i\theta }{\int }_{0}^{+\infty }|Y{|}^{2}\left({e}^{i4\theta }{t}^{4}+{a}_{1}\omega {e}^{i3\theta }{t}^{3}+{a}_{2}{\omega }^{2}{e}^{i2\theta }{t}^{2}+{a}_{3}{\omega }^{3}{e}^{i\theta }t+{\lambda }_{n}{\omega }^{4}\right)dt\end{array}$
Taking the imaginary part in this equality, after multiplying it by ${e}^{i\left(-\theta +2\pi /3\right)}$  , we get
 $\begin{array}{c}\begin{array}{ccc}0& =& {b}_{0}sin\left(2\pi /3-2\theta \right)+{b}_{1}sin\left(2\pi /3+4\theta \right)+{b}_{2}{a}_{1}sin\left(\pi +3\theta \right)+\\ & & +{b}_{3}{a}_{2}sin\left(4\pi /3+2\theta \right)+{b}_{4}{a}_{3}sin\left(5\pi /3+\theta \right)+{b}_{5}Im{\lambda }_{n}\\ \end{array}\end{array}$ (35)
where the constants ${b}_{j}>0$  stand for the values of the integrals.
Following the above conditions, we can deduce that $\text{2}$  $Im{\lambda }_{n}<0$  by letting $\theta =0$  .
We now come to the following theorem, which is more general than theorem  7 .
Theorem 9. For a fixed $a=\left({a}_{1},\dots ,{a}_{m-1}\right)\in {\mathbb{R}}^{m-1}$  which satisfies the hypothesis $\left(H\right)$  and its supplement $\left(s\right)$  , all the zeros of the Stokes multiplier $C\left(a,\lambda \right)$  are real, positive and simple.
• Proof. The reality and positivity have been already proved by Shin [17. The simpleness can be handled in the same manner as for the proof of theorem  7 , thus allowing us to be sketchy.
Let ${\lambda }^{*}={\lambda }^{*}\left(a\right)$  be a double zero of $C\left(a,\lambda \right)$  . Because of the reality of the zeros, it is sufficient to consider $C\left(a,\lambda \right)$  as a function of $\lambda \in \mathbb{R}$  . We then deduce from ( 33 ) that  $\begin{array}{c}{g}_{a}\left({\lambda }^{*}\right)=\overline{{g}_{a}\left({\lambda }^{*}\right)}\text{and}\left({g}_{a}{\right)}_{\lambda }^{\prime }\left({\lambda }^{*}\right)=\overline{\left({g}_{a}{\right)}_{\lambda }^{\prime }\left({\lambda }^{*}\right)}\end{array}$ (36)
Note that ${g}_{a}\left(\lambda \right)$  is an entire function in $\lambda$  of order $\frac{1}{2}+\frac{1}{m}<1$  , whose zeros ${\lambda }_{n}$  now satisfy $Im{\lambda }_{n}<0$  . By the Hadamard's factorization theorem, we have On the simpleness of zeros of Stokes multipliers ${g}_{a}\left(\lambda \right)={g}_{a}\left(0{\right)}^{\infty }{\prod }_{n=0}\left(1-\frac{\lambda }{{\lambda }_{n}}\right).$  Obviously, ${g}_{a}\left(\lambda \right)\ne 0$  on $\mathbb{R}$  , so taking the logarithmic derivative this identity yields $\frac{\left({g}_{a}{\right)}_{\lambda }^{\prime }\left(\lambda \right)}{{g}_{a}\left(\lambda \right)}={\sum }_{n=0}^{\infty }\frac{1}{\lambda -{\lambda }_{n}}.$  The imaginary part of the right hand member is ${\sum }_{n=0}^{\infty }\frac{Im{\lambda }_{n}}{\left(\lambda -Re{\lambda }_{n}{\right)}^{2}+\left(Im{\lambda }_{n}{\right)}^{2}}<0.$  Thus, ( 36 ) cannot be realized at any ${\lambda }^{*}\in \mathbb{R}$  .
On the simpleness of zeros of Stokes multipliers

$\text{1}$  The other one is nothing but $\overline{{g}_{a}\left(\overline{\lambda }\right)}$  .

$\text{2}$  In fact, it holds for any real ${a}_{1}$  .

4. Conclusion

In this paper, we have shown that, under an appropriate hypothesis of signs on the real coefficients ${a}_{1},\dots ,{a}_{m-1}$  of equation ( 1 ), all the zeros of the Stokes multiplier $C\left(a\right)$  are simple.
On the one hand, our proof is partly based on the results of [17, in particular the hypothesis that we use to ensure the reality of these zeros is a sufficient condition first established by Shin. On the other hand, our main remaining arguments, essentially the Green's transform and the Hadamard's factorization theorem, are certainly natural in this context, being already used in various papers. In particular, whenever no parameter is concerned, our reasonings are quite simple.
Since our main theorem  9 makes use of a sufficient for the zeros to be real, we do not give any information about neither the simpleness of the zeros for the cases where some of them are complex, nor conditions on the parameter $a$  for the existence of multiple (real or complex) zeros. As an illustration of the case exhibiting double real zeros, we suggest a common paper with Delabaere [11, where the energy spectrum of the Hamiltonian $H={p}^{2}+i\left({q}^{3}+\alpha q\right)$  acting on ${L}^{2}\left(\mathbb{R}\right)$  was studied by semiclassical analysis. The differential equation associated with this Hamiltonian reads $-{\Phi }^{\prime \prime }\left(X\right)+\left({X}^{3}-\alpha X+E\right)\Phi \left(X\right)=0$  , up to a rotation (see also [20).
It has been shown that in this case some pairs of real zeros ${E}_{n}\left(\alpha \right)$  of the Stokes multiplier may coalesce before turning into complex conjugate at certain critical values of ${\alpha }_{crit}<0$  .
These degenerate values ${E}_{n}\left({\alpha }_{crit}\right)$  are nothing but common zeros of the Stokes multiplier and its derivative with respect to $E$  (see Fig.1 in [11). It seems to us that the question of characterizing these critical values of $\alpha$  is certainly an interesting but quite challenging problem.
From the mathematical viewpoint, this question is of course related to the fact that in general for $m\ge 3$  , neither special functions solution of ( 1 ) are known, nor their related Stokes multipliers (when $m=2$  these multipliers can be explicitly expressed in term of the Gamma function, while for $m=1$  they are constants). In this way, the present paper can be thought of as an attempt for exploring some hidden special functions in relation to their Stokes multipliers. But, as already said in the introduction, the main motivation for this paper was to add a new (small) stone for the mathematical foundation of $\mathcal{P}\mathcal{T}$  -symmetric models, in addition to our results in [21.
Acknowledgments. This work was supported by the International Centre for Theoretical Physics in the framework of Post-doctoral Fellowship. References

1. C.M. Bender, J. Brod, A. Refig, M. E Reuter, The operator $\mathcal{C}$  in $\mathcal{P}\mathcal{T}$  -symmetric quantum theories. J. Phys. A: Math. Gen. 37 No 43 (2004) 10139-10165.
2. C.M. Bender, P.N. Meisinger, Q.Wang, Calculation of the hidden symmetry operator in $\mathcal{P}\mathcal{T}$  -symmetric quantum mechanics. J. Phys. A 36 (2003), no. 7, 1973–1983.
3. On the simpleness of zeros of Stokes multipliers C.M. Bender, S. Boettcher, Real spectra in non-Hermitian hamiltonians having PT-symmetry. Phys. Rev. Lett. 80, 5243 (1998).
4. C.M. Bender, S. Boettcher, P.N. Meisinger, PT-symmetric quantum mechanics. J. Math. Phys. 40, 2201 (1999).
5. C.M. Bender, K.A. Milton, Model of supersymmetric quantum field theory with broken parity symmetry. Physical Review D, Vol 57, N ${}^{o}$  6, 3595-3608 (1998)
6. R. Ph. Jr. Boas, Entire functions. Academic Press Inc., New York, 1954.
7. E. Delabaere, H. Dillinger, F. Pham, Résurgence de Voros et périodes des courbes hyperelliptiques. Annales de l'Institut Fourier, Vol 43, Fasc. 1, 163-199 (1993).
8. E. Delabaere, H. Dillinger, F. Pham, Exact semi-classical expansions for one dimensional quantum oscillators. Journal Math. Phys. 38, 12, 6126-6184 (1997).
9. E. Delabaere, F. Pham, Resurgent methods in semi-classical asymptotics. Ann. Inst. Henri Poincaré, Sect. A, Vol. 71, N ${}^{o}$  1, 1-94 (1999).
10. E. Delabaere, F. Pham, Eigenvalues of complex hamiltonians with $\mathcal{P}\mathcal{T}$  -symmetry I. Phys. Lett. A 250, 25 (1998).
11. E. Delabaere, D.T. Trinh, Spectral analysis of the complex cubic oscillator. J.Phys. A: Math. Gen. 33 (2000), 8771-8796.
12. E. Delabaere, J.M.Rasoamanana, Resurgent deformations for an ordinary differential equation of order 2. To appear in Pacific Journal of Mathematics.
13. P. Dorey, C. Dunning, R. Tateo, Spectral equivalences, Bethe ansatz equations, and reality properties in $\mathcal{P}\mathcal{T}$  -symmetric quantum mechanics. J. Phys. A 34 (2001), no. 28, 5679–5704.
14. G. Lévai, M. Znojil, Conditions for complex spectra in a class of $\mathcal{P}\mathcal{T}$  -symmetric potentials. Modern Phys. Lett. A 16 (2001), no. 30, 1973–1981.
15. F. Pham, Confluence of turning points in exact WKB analysis. The Stokes phenomenon and Hilbert's 16th problem (Groningen, 1995), 215–235, World Sci. Publishing, River Edge, NJ (1996).
16. F. Pham, Multiple turning points in exact WKB analysis (variations on a theme of Stokes). Toward the exact WKB analysis of differential equations linear or non-linear (C.Howls, T.Kawai, Y.Takei ed.), Kyoto University Press (2000),71-85.
17. K.C. Shin, On the reality of the eigenvalues for a class of $\mathcal{P}\mathcal{T}$  -symmetric oscillators. Comm. Math. Phys. 229 (2002), no. 3, 543–564.
18. Y. Sibuya, Global Theory of a Second Order Linear Differential Equation with a Polynomial Coefficient. Mathematics Studies 18, North-Holland Publishing Company (1975).
19. D.T. Trinh, Coefficients de Stokes du modèle cubique: point de vue de la résurgence quantique. Ann. Fac. Sci. Toulouse Math. (6) 14 (2005), no. 1, 71-103.
20. On the simpleness of zeros of Stokes multipliers D.T. Trinh, On the Sturm-Liouville problem for the complex cubic oscillator. Asymptot. Anal. 40 (2004), no. 3-4, 211-234.
21. D.T. Trinh, Remarks on $\mathcal{P}\mathcal{T}$  -pseudo-norm in $\mathcal{P}\mathcal{T}$  -symmetric quantum mechanics. arXiv.org: math-ph/0502009. To appear in J. Phys. A.
22. A. Voros, The return of the quartic oscillator. The complex WKB method Ann. Inst. H.Poincaré, Physique Théorique, 39, 211-338 (1983).
23. A. Voros, Airy function – exact WKB results for potentials of odd degree. J. Phys. A 32 (1999), no. 7, 1301–1311.