betaFeedback

WorksPhysics and Astronomy

Bright Solitons in Non-Hermitian systems: Exploring the interplay of PT\mathcal{PT} Symmetry and Wadati Structure

Abstract

In this study, we revisit the subject of optical solitons in non-Hermitian potentials. For the underlying model of the focusing nonlinear Schrödinger equation with a complex localized potential, we offer a classification of such models into four canonical cases. These correspond to (a) PT\mathcal{PT}-symmetric Wadati potentials, (b) PT\mathcal{PT}-symmetric non-Wadati potentials, (c) Non-PT\mathcal{PT} but Wadati potentials, and (d) complex potentials with neither PT\mathcal{PT} nor Wadati symmetries. We study the existence and stability of optical solitons across prototypical model variations spanning all four relevant quadrants. We complement such an investigation with a low-dimensional effective description based on a non-conservative variational approximation method. The variational framework is assessed in terms of its ability to represent the structural properties and bifurcations of solutions. {Our results suggest that continuous families of bright solitons persist across sectors (a)–(c). For non-PT\mathcal{PT}-symmetric potentials, small departures from the Wadati limit, corresponding to sector (d), suggest the loss of continuous families of bright-solitons including, more specifically, the possibility of the generic absence of single-humped localized states. The corresponding variational analysis is broadly consistent with these observations. We also highlight the usefulness and limitations for such variational approaches in describing bright solitons in the presence of non-Hermitian potentials.

Introduction

The notion of PTPT-symmetry lies at the heart of non-Hermitian physics [3, 4] [1, 2], which seeks to describe nonconservative systems that interact with their external environment. Historically, the subject of non-Hermitian quantum physics attracted considerable interest following the seminal observation that certain non-Hermitian Hamiltonians possessing parity (P\mathcal{P}) and time-reversal (T\mathcal{T}) symmetry can exhibit an entirely real eigenspectrum [5]. Remarkably, this demonstrated that Hermiticity, or equivalently the self-adjointness of the associated Schrödinger operator, is not a necessary condition for the corresponding spectrum to be real. These developments subsequently motivated the study of broader classes of non-Hermitian operators, including pseudo-Hermitian operators admitting purely real spectra [6, 7].

The structural equivalence between the single-particle Schrödinger equation and the paraxial wave equation in classical optics opened the door to the rapidly growing field of PTPT-symmetric optics [8]. By mapping the quantum mechanical potential to a complex-valued refractive index profile, where the real part represents the waveguide’s refractive index and the imaginary part represents balanced gain and loss, this paradigm has sparked intense research into non-Hermitian optical beam propagation and the parallels between optics and quantum mechanics at both the linear and the nonlinear level [9, 10, 11, 12, 13]. Substantial experimental progress has since been made across a wide range of photonic platforms for realizing and probing PTPT-symmetric effects, including optical waveguides [14, 15], photonic lattices [16], optical metamaterials [17], optical microresonators [18], and coherently prepared warm atomic gases [19, 20]. Furthermore, the notion of non-Hermitian wave propagation was explored in various other disciplines, e.g., non-Hermitian acoustics [21], PTPT-symmetric metamaterials [22], exciton-polaritons and plasmonics [23], ultracold Fermi gases [24], and in electrical circuits [25] just to name a few; see Refs. [3] [1] for comprehensive reviews. Wadati introduced a systematic construction of PTPT-symmetric complex potentials by exploiting the correspondence between Dirac- and Schrödinger-type spectral problems arising in inverse-scattering theory [26]. Subsequent work showed that the same structural class can remain spectrally real and support continuous nonlinear-mode families even without PTPT symmetry [27, 28]. Since then, the so-called Wadati potentials have by themselves opened several avenues of investigation [29, 30, 31, 32, 33, 34]. A notable property of the nonlinear Schrödinger (NLS) equation with a non-Hermitian Wadati potential concerns the existence of continuous families of stable solitons for non-PTPT asymmetric potentials [28, 31]. Conversely, such systems can also support continuous families of asymmetric solitons within strictly PTPT-symmetric landscapes [35], as well as the existence of constant-intensity waves [33, 36]. Furthermore, hierarchies of integrable partial differential equations (PDEs) have been established by extending these non-Hermitian principles to nonlocal, PT\mathcal{PT}-symmetric variants of the NLS equation [37].

Self-localized states (solitons) in purely non-Hermitian settings typically exist as isolated solutions rather than as continuous families parameterized by a propagation constant [38, 39]. However, with additional symmetries—e.g., PT\mathcal{PT}-symmetry, i.e., balanced gain-loss distributions or underlying integrals of motion which exist for so called “Wadati” potentials—such continuous families could still be found (cf. Refs. [28, 40]). The existence and stability of such stationary modes are central to understanding various dynamical phenomena (such as scattering, cf. Refs. [41, 42]) in these non-Hermitian systems. Despite extensive previous studies addressing these questions in non-Hermitian systems possessing (or lacking) one or more of these properties, comparatively few studies have systematically examined soliton existence and stability across all possible combinations of PT\mathcal{PT} symmetry and Wadati structure. In this work, we address this question within a unified framework by considering the four “quadrants” of a classification of non-Hermitian systems that we introduce based on the presence or absence of these two properties:

  • (a) systems possessing both PT\mathcal{PT} symmetry and Wadati structure,

  • (b) systems possessing PT\mathcal{PT} symmetry but lacking Wadati structure,

  • (c) systems possessing Wadati structure but lacking PT\mathcal{PT} symmetry, and

  • (d) systems possessing neither PT\mathcal{PT} symmetry nor Wadati structure.

Furthermore, we examine the efficacy of low-dimensional descriptions within each of these cases in predicting soliton existence and stability. Quadrants (a)–(c) are found to possess continuous (in a canonical parameter, such as the solution frequency) families of standing bright solitary waves. The variational formulation begins to exhibit qualitative discrepancies in sector (c), most notably at the level of the linearization spectrum, underscoring the richness of the Wadati structure and the challenges associated with capturing it within a low-dimensional variational framework. We are unable to identify continuous such families in sector (d), where the isolated solutions we identify, in fact, also no longer preserve a single-humped spatial profile, in line with other generic non-Hermitian systems such as the complex potentials of Ref. [39].

Our presentation is organized as follows. In Sec. II, we present the NLS model with a complex external potential considered in this work. In Sec. III, we develop a low-dimensional description of its bright soliton solutions using a non-conservative variational approximation. In Sec. IV, we discuss the existence, stability, and dynamical properties of bright solitons across the four quadrants introduced above. Finally, Sec. V summarizes the main findings and outlines some outstanding challenges and directions for future work.

The nonlinear Schrödinger equation with complex potential

In the paraxial approximation, the evolution of optical beams in media featuring a focusing cubic (Kerr) nonlinearity and a spatially dependent refractive index and a gain-loss distribution is governed by the following dimensionless NLS equation [4, 43]:

iΨt+12Ψxx+[VR(x)+iVI(x)]Ψ+∣Ψ∣2Ψ=0,(1)i\Psi_t + \frac{1}{2}\Psi_{xx} + [V_R(x) + iV_I(x)]\Psi+ |\Psi|^2\Psi= 0, \tag*{(1)}

where the tt and xx subscripts denote partial derivatives. Here, the wave function Ψ(x,t)\Psi(x,t) represents the complex electric field envelope, the variables tt and xx denote the propagation and transverse directions, respectively, while VRV_R and VIV_I are real-valued functions representing the refractive index and the gain-loss distribution, respectively. In terms of Ψ(x,t)\Psi(x,t), the beam intensity ∣Ψ∣2|\Psi|^2 and the transverse power flow density 12i(Ψ∗Ψx−Ψx∗Ψ)\frac{1}{2i}(\Psi^*\Psi_x-\Psi_x^*\Psi) (asterisks denote complex conjugation) are generically not conserved in the presence of the imaginary part of the potential. In this work, we restrict ourselves to the scenario where Ψ\Psi evolves in the presence of a spatially localized inhomogeneity, i.e., when both VR(x)V_R(x) and VI(x)V_I(x) decay exponentially fast. The case of periodic potentials, e.g., in the spirit of Refs. [11, 12] is interesting in its own right as a subject of future study.

Stationary solitons or nonlinear eigenmodes, of the form Ψ(x,t)=Φ0(x)eiμt\Psi(x,t)=\Phi_0(x)e^{i\mu t}, where μ\mu is the propagation constant (or frequency), satisfy the following boundary value problem

−μΦ0+12Φ0′′+VΦ0+∣Φ0∣2Φ0=0,(2)-\mu\Phi_0+\frac{1}{2}\Phi_0''+V\Phi_0+|\Phi_0|^2\Phi_0=0, \tag*{(2)}

where the complex wavefunction Φ0(x)\Phi_0(x) satisfies the boundary conditions ∣Φ0(±∞)∣=0|\Phi_0(\pm\infty)|=0. We first review the properties of solitons along the two axes of PT\mathcal{PT} symmetry and Wadati structure, respectively. PT\mathcal{PT} symmetry entails invariance under the combined actions of parity (P\mathcal{P}) and time reversal (T\mathcal{T}), with the latter involving complex conjugation, i.e.,

Ψ(x,t)→PTΨ∗(−x,−t)\Psi(x,t) \xrightarrow{\mathcal{PT}} \Psi^{*}(-x,-t), which is permissible generally for VR(−x)=VR(x)V_{R}(-x)=V_{R}(x) (even) and VI(−x)=−VI(x)V_{I}(-x)=-V_{I}(x) (odd). Notably, the odd parity of the imaginary part VI(x)V_{I}(x) entails a “global” balance of gain and loss (between the left and right sides), rendering possible the existence of continuous families of stationary modes [3, 28]. For PT\mathcal{PT}-symmetric stationary modes with real μ\mu, this symmetry is inherited by the stationary profile according to Φ0(x)=Φ0∗(−x)\Phi_{0}(x)=\Phi_{0}^{*}(-x), i.e., the real (imaginary) parts possess even (odd) parity. Typically, the PT\mathcal{PT}-symmetry of the model encodes the same symmetry onto the soliton, except in the case of certain particular potentials; see, for instance, Ref. [35].

On another note, the family of Wadati potentials was first introduced by Wadati [26] in the context of constructing PT\mathcal{PT}-symmetric potentials for the Schrödinger operator using inverse scattering theory. The NLS with a Wadati potential possesses special properties that are not necessarily shared by the broader class of PT\mathcal{PT}-symmetric potentials [29]. In particular, Wadati potentials need not satisfy the PT\mathcal{PT}-symmetry condition. Instead, the real and imaginary parts of the potential are defined in terms of the function W(x)\mathcal{W}(x) as

VR(x)=12W2,VI(x)=−12Wx.V_{R}(x)=\frac{1}{2}\mathcal{W}^{2},\qquad V_{I}(x)=-\frac{1}{2}\mathcal{W}_{x}.

A crucial structural property of Wadati potentials is the existence of a first integral for the stationary spatial-dynamics system, which removes one of the independent solvability constraints and thereby permits continuous families of localized modes [44, 45]. Another important property of Wadati systems is the existence of constant-intensity waves [33], as well as the transformation Ψ=ψexp⁡(i∫xW(x′) dx′)\Psi=\psi\exp\left(i\int_{x}^{ } \mathcal{W}(x')\,\mathrm{d}x'\right), which renders Eq. (1) into the following inhomogeneous, advective model:

iψt+iW(x)ψx+12ψxx+∣ψ∣2ψ=0.(3)i\psi_{t}+i\mathcal{W}(x)\psi_{x}+\frac{1}{2}\psi_{xx}+|\psi|^{2}\psi=0. \tag*{(3)}

In this work, we explore the four cases defined by the axes of PT\mathcal{PT} symmetry and Wadati structure. Representative exponentially localized potentials considered in each quadrant are

  • (a) PT\mathcal{PT}-symmetric Wadati potential defined by the even-parity potential function

W(x)=αsech⁡2(x),with α>0.(4)\mathcal{W}(x)=\alpha\operatorname{sech}^{2}(x),\qquad\text{with }\alpha>0. \tag*{(4)}
  • (b) PT\mathcal{PT}-symmetric potentials lacking Wadati structure. Here, we consider a family of potentials with

VR(x)=α22sech⁡4(x),VI(x)=32αtanh⁡(x)sech⁡2(x).(5)V_{R}(x)=\frac{\alpha^{2}}{2}\operatorname{sech}^{4}(x),\qquad V_{I}(x)=\frac{3}{2}\alpha\tanh(x)\operatorname{sech}^{2}(x). \tag*{(5)}
  • (c) Wadati potentials lacking PT\mathcal{PT}-symmetry. Here, the potential VV is defined by the potential function W(x)\mathcal{W}(x). In this case, we consider the family of asymmetric single-hump potentials

W(x)=αsech⁡2(x)+Γαsech⁡2(x−x0),(6)\mathcal{W}(x)=\alpha\operatorname{sech}^{2}(x)+\Gamma\alpha\operatorname{sech}^{2}(x-x_{0}), \tag*{(6)}

and we choose a weak degree of asymmetry in the potential function, i.e., Γ≪1\Gamma\ll1 (Γ>0\Gamma>0), and 0<x0∼10<x_{0}\sim1.

  • (d) Potentials lacking both the PT\mathcal{PT}-symmetry and Wadati structure. Here, we consider a family of potentials

VR(x)=12[αsech⁡2(x)+Γαsech⁡2(x−x0)]2.(7)V_{R}(x)=\frac{1}{2}\left[\alpha\operatorname{sech}^{2}(x)+\Gamma\alpha\operatorname{sech}^{2}(x-x_{0})\right]^{2}. \tag*{(7)}
VI(x)=1−β2[2αtanh⁡(x)sech⁡2(x)+2Γαtanh⁡(x−x0)sech⁡2(x−x0)].(8)V_{I}(x)=\frac{1-\beta}{2}\left[2\alpha\tanh(x)\operatorname{sech}^{2}(x)+2\Gamma\alpha\tanh(x-x_{0})\operatorname{sech}^{2}(x-x_{0})\right]. \tag*{(8)}

wherein the potentials are perturbed away from an appropriate Wadati potential by tuning the proximity parameter β∼Γ≪1\beta\sim\Gamma\ll1. It is worth stressing again that, generally, such a non-PT\mathcal{PT} and non-Wadati system is expected to only have solutions at isolated values of μ\mu [38, 39].

We study the existence and stability of bright modes bifurcating from the bulk NLS limit (α=0\alpha=0). We employ variational approximations to derive a reduced-order description of the bright modes. Such approximations have been successfully applied in a variety of related contexts to non-conservative NLS models [46, 47]. In particular, we seek to assess the feasibility of a low-dimensional representation of the bright modes in capturing their existence, stability, and bifurcation structure.

Non-conservative variational approximation

We utilize a low-dimensional description of the bright soliton eigenmodes to the NLS with non-Hermitian potential. To this end, we reduce the NLS dynamics to an effective “particle” model, i.e., to a solitonic manifold described by a set of coupled ODEs governing the evolution of the bright-soliton parameters (amplitude, position, width, phase, velocity, and chirp). The underlying solitonic manifold construction is based on the non-conservative variational approximation (NCVA) for Schrödinger-type systems, as outlined in Refs. [46, 47]. The method draws inspiration from the Lagrangian (and Hamiltonian) formulation of non-conservative mechanical systems, as developed in Refs. [48, 49]. Inspired by these earlier works, the extension to non-conservative field theories (i.e., PDEs) was studied in Ref. [50] with the PT\mathcal{PT}-symmetric variants of the sine-Gordon and ϕ4\phi^4-models.

We set the stage by factoring out the linear phase accumulation associated with soliton propagation, through the substitution Ψ(x,t)=Φ(x,t)eiμt\Psi(x,t)=\Phi(x,t)e^{i\mu t}, yielding

iΦt−μΦ+V(x)Φ+12Φxx+∣Φ∣2Φ=0.(9)i\Phi_t-\mu\Phi+V(x)\Phi+\frac{1}{2}\Phi_{xx}+|\Phi|^2\Phi=0. \tag*{(9)}

Furthermore, the explicit contribution of the real (conservative) part of the potential VR(x)V_R(x) can be eliminated through a spatially dependent phase transformation. To this end, we introduce the auxiliary function W(x)W(x) such that W2/2=VRW^2/2=V_R and define Φ=ϕeiθ(x)\Phi=\phi e^{i\theta(x)}, with W(x)=θ′(x)W(x)=\theta'(x), to obtain the following perturbed NLS equation:

iϕt−μϕ+12ϕxx+∣ϕ∣2ϕ=R(ϕ).(10)i\phi_t-\mu\phi+\frac{1}{2}\phi_{xx}+|\phi|^2\phi=\mathcal{R}(\phi). \tag*{(10)}

Here, all non-conservative contributions R(ϕ)=−iWϕx−i(VI+Wx/2)ϕ\mathcal{R}(\phi)=-iW\phi_x-i(V_I+W_x/2)\phi have been collected on the right-hand side (RHS). Notably, for a Wadati potential, VI=−Wx/2V_I=-W_x/2, leading to the cancellation of the second term on the RHS: VI+Wx/2≡0V_I+W_x/2\equiv0.

Hereafter, we consider a six-parameter solitonic-ansatz for such bright modes:

ϕˉ(x,p⃗(t))=a sech⁡[w(x−ξ)]exp⁡[i(b(x−ξ)2+c(x−ξ)+χ)],\bar{\phi}(x,\vec{p}(t))=a\,\operatorname{sech}[w(x-\xi)]\exp\left[i\left(b(x-\xi)^2+c(x-\xi)+\chi\right)\right],

with p⃗(t)=[a(t),b(t),c(t),ξ(t),w(t),χ(t)]T\vec{p}(t)=[a(t),b(t),c(t),\xi(t),w(t),\chi(t)]^T, where a(t)a(t), b(t)b(t), c(t)c(t), ξ(t)\xi(t), w(t)w(t), and χ(t)\chi(t) are the soliton amplitude, chirp, velocity, position, width, and phase (time-dependent) parameters. The formulation incorporates a chirp bb, together with its variational conjugate, the (inverse) soliton width parameter ww. This pair of parameters is necessary to capture potential breathing dynamics [46], and to allow the soliton localization scale (set by ww) to adjust to, and potentially be dictated by, the underlying complex potential. Focusing on the conservative components of the potential, these define the Lagrangian density

Lcons=i2(ϕϕt∗−ϕ∗ϕt)+12∣ϕx∣2−12∣ϕ∣4+μ∣ϕ∣2,\mathcal{L}_{\mathrm{cons}}=\frac{i}{2}(\phi\phi_t^*-\phi^*\phi_t)+\frac{1}{2}|\phi_x|^2-\frac{1}{2}|\phi|^4+\mu|\phi|^2,

with an effective (averaged) Lagrangian given by

Lˉ=∫−∞∞Lcons[ϕˉ(x,p⃗(t))] dx=π2a2b˙6w3+a2w3+π2a2b23w3+c2a2w−2a43w+2μa2w+2a2w(χ˙−cξ˙).\bar{L}=\int_{-\infty}^{\infty}\mathcal{L}_{\mathrm{cons}}[\bar{\phi}(x,\vec{p}(t))]\,dx=\frac{\pi^2a^2\dot{b}}{6w^3}+\frac{a^2w}{3}+\frac{\pi^2a^2b^2}{3w^3}+\frac{c^2a^2}{w}-\frac{2a^4}{3w}+\frac{2\mu a^2}{w}+\frac{2a^2}{w}(\dot{\chi}-c\dot{\xi}).

Non-conservative variational calculus [46] yields the following effective, Euler-Lagrange (EL), evolution equations:

ddt(∂Lˉ∂p⃗˙)−∂Lˉ∂p⃗−I⃗p⃗=0,\frac{d}{dt}\left(\frac{\partial\bar{L}}{\partial\dot{\vec{p}}}\right)-\frac{\partial\bar{L}}{\partial\vec{p}}-\vec{I}_{\vec{p}}=0,

wherein the overlapping integrals are given by

I⃗p⃗=2Re⁡[∫−∞∞R(ϕˉ)∂ϕˉ∗∂p⃗ dx].\vec{I}_{\vec{p}}=2\operatorname{Re}\left[\int_{-\infty}^{\infty}\mathcal{R}(\bar{\phi})\frac{\partial\bar{\phi}^*}{\partial\vec{p}}\,dx\right].

These expressions arise from the non-conservative terms in the RHS of Eq. (10), and are computed numerically using trapezoidal quadrature (cf. the Appendix § 1.5.1 for details on their derivation).

The ODE system obtained from these EL equations can be explicitly represented in terms of the solitonic variables, in the form p⃗˙=F⃗(p⃗)\dot{\vec{p}}=\vec{F}(\vec{p}), as follows:

a˙=−ab+3w8aIχ−3w32aπ2Ib,b˙=−2b2−2a2w2π2+2w4π2+3w4π2a2Iw+3w32π2aIa,c˙=w2a2(−cIχ−Iξ),ξ˙=c+w2a2Ic,w˙=−2bw+w24a2Iχ−3w4a2π2Ib,χ˙=−3w8aIa−w23+c22+5a26+wc2a2Ic−μ−w24a2Iw.(11)\begin{aligned} \dot{a} &= -ab+\frac{3w}{8a}I_{\chi}-\frac{3w^{3}}{2a\pi^{2}}I_{b},\\ \dot{b} &= -2b^{2}-\frac{2a^{2}w^{2}}{\pi^{2}}+\frac{2w^{4}}{\pi^{2}}+\frac{3w^{4}}{\pi^{2}a^{2}}I_{w}+\frac{3w^{3}}{2\pi^{2}a}I_{a},\\ \dot{c} &= \frac{w}{2a^{2}}(-cI_{\chi}-I_{\xi}),\\ \dot{\xi} &= c+\frac{w}{2a^{2}}I_{c},\\ \dot{w} &= -2bw+\frac{w^{2}}{4a^{2}}I_{\chi}-\frac{3w^{4}}{a^{2}\pi^{2}}I_{b},\\ \dot{\chi} &= -\frac{3w}{8a}I_{a}-\frac{w^{2}}{3}+\frac{c^{2}}{2}+\frac{5a^{2}}{6}+\frac{wc}{2a^{2}}I_{c}-\mu-\frac{w^{2}}{4a^{2}}I_{w}. \tag*{(11)} \end{aligned}

The fixed points p⃗0\vec{p}_{0} of this ODE system, satisfying F⃗(p⃗0)=0⃗\vec{F}(\vec{p}_{0})=\vec{0}, represent reduced order representations (approximations) of stationary bright soliton solutions to Eq. (10). It is of interest to check if the bright solitonic ansatz captures the infinite-dimensional profile over a wide regime in parameter space.

It is useful to note that p⃗0\vec{p}_{0} may be recast by combining the conditions Fa=0F_{a}=0 and Fw=0F_{w}=0, which yield Iχ=0I_{\chi}=0. In particular, this condition provides the NCVA counterpart of the PDE power-balance constraint ∫R∣Φ0(x)∣2VI(x) dx=0\int_{\mathbb{R}}|\Phi_{0}(x)|^{2}V_{I}(x)\,dx=0, and thereby selects the NCVA soliton center ξ0\xi_{0}. Therefore, the six fixed-point equations for p⃗0\vec{p}_{0} can be written as

Iχ=0,(12)I_{\chi}=0, \tag*{(12)}
−2b02−2a02w02π2+2w04π2+3w04π2a02Iw+3w032π2a0Ia=0,(13)-2b_{0}^{2}-\frac{2a_{0}^{2}w_{0}^{2}}{\pi^{2}}+\frac{2w_{0}^{4}}{\pi^{2}}+\frac{3w_{0}^{4}}{\pi^{2}a_{0}^{2}}I_{w}+\frac{3w_{0}^{3}}{2\pi^{2}a_{0}}I_{a}=0, \tag*{(13)}
Iξ=0,(14)I_{\xi}=0, \tag*{(14)}
c0+w02a02Ic=0,(15)c_{0}+\frac{w_{0}}{2a_{0}^{2}}I_{c}=0, \tag*{(15)}
b0+3w032a02π2Ib=0,(16)b_{0}+\frac{3w_{0}^{3}}{2a_{0}^{2}\pi^{2}}I_{b}=0, \tag*{(16)}
−3w08a0Ia−w023+c022+5a026+w0c02a02Ic−μ−w024a02Iw=0.(17)-\frac{3w_{0}}{8a_{0}}I_{a}-\frac{w_{0}^{2}}{3}+\frac{c_{0}^{2}}{2}+\frac{5a_{0}^{2}}{6}+\frac{w_{0}c_{0}}{2a_{0}^{2}}I_{c}-\mu-\frac{w_{0}^{2}}{4a_{0}^{2}}I_{w}=0. \tag*{(17)}

Upon fixing the arbitrary global phase, e.g., by setting χ0≡0\chi_{0}\equiv0, the system comprises six fixed-point equations for the five unknowns {a0,b0,c0,ξ0,w0}\{a_{0},b_{0},c_{0},\xi_{0},w_{0}\}. Thus, for a generic non-Hermitian potential, the reduced stationary problem is overdetermined. For the phase-transformed NLS with a Wadati potential, however, we have Iξ=2Re⁡[∫−∞∞iW(x)∣∂xϕˉ(x,p⃗)∣2 dx]≡0I_{\xi}=2\operatorname{Re}\left[\int_{-\infty}^{\infty}iW(x)|\partial_{x}\bar{\phi}(x,\vec{p})|^{2}\,dx\right]\equiv0 identically on the variational manifold. Consequently, the condition Iξ=0I_{\xi}=0 imposes no additional constraint, leaving five independent fixed-point conditions for the five (nontrivial) variational parameters. Similarly, for a generic PT\mathcal{PT}-symmetric potential, a PT\mathcal{PT}-symmetric fixed point is characterized by ξ0=0\xi_{0}=0 (i.e., it is centered). Parity considerations then yield Iχ=0I_{\chi}=0 and Ib=0I_{b}=0, with the latter, through the corresponding fixed-point equation, implying b0=0b_{0}=0. In turn, the conditions ξ0=b0=0\xi_{0}=b_{0}=0 ensure that Iξ=0I_{\xi}=0 due to the underlying parity. Consequently, three of the six fixed-point conditions are satisfied identically, while the remaining three constitute independent constraints for the three nontrivial variational parameters {a0,c0,w0}\{a_{0},c_{0},w_{0}\}. Note that for generic PT\mathcal{PT}-symmetric potentials, symmetry breaking of PT\mathcal{PT}-symmetric solitons is anyway forbidden [51] with the exception of the Wadati potentials [35].

Besides existence, the linear stability of these fixed points (and hopefully, by extension, that of the bright solitons of the original NLS equation) can be probed by computing the eigenvalues {λ(j)}j=16\{\lambda^{(j)}\}_{j=1}^{6} (and associated eigenvectors {v⃗(j)}j=16\{\vec{v}^{(j)}\}_{j=1}^{6}) of the Jacobian matrix (dF⃗/dp⃗)∣p⃗0(d\vec{F}/d\vec{p})|_{\vec{p}_{0}}. To connect the NCVA-ODE eigenmodes to their PDE analogues, we consider perturbing on the solitonic manifold along the jjth NCVA eigendirection v⃗(j)\vec{v}^{(j)}, namely:

ϕˉ(x,p⃗0+ϵv⃗(j))≈ϕˉ(x,p⃗0)+ϵ∑i=16vi(j)∂piϕˉ(x,p⃗0),\bar{\phi}(x,\vec{p}_{0}+\epsilon\vec{v}^{(j)})\approx\bar{\phi}(x,\vec{p}_{0})+\epsilon\sum_{i=1}^{6}v_{i}^{(j)}\partial_{p_{i}}\bar{\phi}(x,\vec{p}_{0}),

where span⁡{∂piϕˉ(x,p⃗0)}i=16\operatorname{span}\{\partial_{p_{i}}\bar{\phi}(x,\vec{p}_{0})\}_{i=1}^{6} represents the local tangent space of the variational solitonic manifold at p⃗0\vec{p}_{0}, v⃗(j)∈C6\vec{v}^{(j)}\in\mathbb{C}^{6} and an approximate PDE eigenmode is given by ∑i=16vi(j)∂piϕˉ(x,p⃗0)\sum_{i=1}^{6}v_{i}^{(j)}\partial_{p_{i}}\bar{\phi}(x,\vec{p}_{0}). There are two aspects of interest when it comes to the fidelity of the NCVA stability picture: (i) Is the NCVA manifold tangent space span⁡{∂piϕ(x,p⃗0)}i=16\operatorname{span}\{\partial_{p_i}\phi(x,\vec{p}_0)\}_{i=1}^{6} “rich” enough to represent the PDE eigenmodes? and (ii) does each of the eigendirections {v⃗(j)}\{\vec{v}^{(j)}\} combine the basis elements ∂piϕ(x,p⃗0)\partial_{p_i}\phi(x,\vec{p}_0) appropriately so as to represent the PDE eigenmode? To partially address (i), we note that the NCVA can, at best, capture the discrete, spatially localized PDE eigenmodes represented in the six-parameter ansatz, whereas the continuous-spectrum modes (that satisfy radiation boundary conditions), cannot be represented within the present finite-dimensional variational framework. In what follows, we will probe both of the above aspects for each of the four PT\mathcal{PT}-Wadati quadrant sectors outlined in the Introduction.

Results

Wadati potentials with PT-symmetry

Here, we consider a one-parameter family of PT\mathcal{PT}-symmetric Wadati potentials V(x)V(x) as defined in Eq. (4). Furthermore, we specifically seek PT\mathcal{PT}-symmetric stationary solutions to such a family of potentials, i.e., solutions satisfying Φ0∗(−x)=Φ0(x)\Phi_0^*(-x)=\Phi_0(x). Assuming PT\mathcal{PT}-symmetry removes the generic overdetermination of the stationary problem (as discussed in § 1.1 and III) and thereby permits continuous families of localized modes parametrized by μ\mu; see Figure 1(a). In the present work, we display representative members of this families for μ=0.5\mu=0.5, for which the zero-potential limit corresponds to the unit-amplitude, unit-width NLS bright soliton. Our emphasis is therefore on the structural mechanisms governing stability, rather than on a systematic continuation of the solution families in μ\mu. An extension of the discussion can be made straightforwardly by introducing the rescaled coordinate y=2μ xy=\sqrt{2\mu}\,x together with ϕ0(x)=2μ u0(y)\phi_0(x)=\sqrt{2\mu}\,u_0(y) for stationary solutions in Eq. (10) allowing for the propagation constant μ\mu to be normalized, at the expense of a simultaneous rescaling of the strength and spatial scale of the complex potential. Thus, while the quantitative soliton profiles and spectra generally depend on μ\mu, the structural mechanisms leading to instability investigated here are not specific to the specific selection of the parameter μ\mu. Furthermore, we note that for the class of Wadati potentials considered here, with the associated function W(x)\mathcal{W}(x) possessing a single extremum, the corresponding linear Schrödinger operator remains in the unbroken PT\mathcal{PT}-symmetric phase [27].

Panels showing soliton amplitude as a function of soliton frequency for three potential functions

Figure 1. The panels illustrate continuous families of solitary waves as characterized by the soliton amplitude as a function of the soliton frequency μ\mu for: (a) the PT\mathcal{PT}-symmetric Wadati potential function [see Eq. (4)] with α=1\alpha=1, (b) the PT\mathcal{PT}-symmetric potential [see Eq. (5)] with α=1\alpha=1, and (c) the non-PT\mathcal{PT}-symmetric Wadati potential function [see Eq. (6)]. The other system parameters are fixed to α=1\alpha=1, x0=2.75x_0=2.75, and Γ=0.1\Gamma=0.1.

At the PDE level, we compute bright modes of the driven-dissipative system for 0<α≲1.740<\alpha\lesssim1.74 using numerical continuation based on the Newton conjugate-gradient method [52]. The (approximate) upper endpoint, α≈1.74\alpha\approx1.74, coincides with the linear limit, where the nonlinear bright mode bifurcates from an eigenmode of the underlying linear Schrödinger operator Llin=12d2dx2+V(x)\mathcal{L}_{\mathrm{lin}}=\frac{1}{2}\frac{d^2}{dx^2}+V(x). This algorithm can be initialized with the bright soliton solution of the NLS equation at α=0\alpha=0, given by the profile Φ0(x)=sech⁡(x)\Phi_0(x)=\operatorname{sech}(x), and then solutions are continued as one increases the value of the potential shape parameter α\alpha. We store the corresponding PDE bright soliton in steps of Δα=0.05\Delta\alpha=0.05.

Having computed this family of PDE modes, we analyze their linear stability by introducing perturbations to their profiles; that is Φ(x,t)=Φ0(x)+ϵ(x,t)\Phi(x,t)=\Phi_0(x)+\epsilon(x,t) with ∣ϵ∣≪1|\epsilon|\ll1. Substituting this, together with the expression for the complex-valued perturbation ϵ=ϕI(x,t)+iϕR(x,t)\epsilon=\phi^I(x,t)+i\phi^R(x,t) into Eq. (9), and collecting real and imaginary parts to leading order, yields the evolution equations governing the small-amplitude perturbations

∂t(ϕIϕR)=L(ϕIϕR),(18)\partial_t \begin{pmatrix} \phi^{I} \\ \phi^{R} \end{pmatrix} = \mathcal{L}\begin{pmatrix} \phi^{I} \\ \phi^{R} \end{pmatrix}, \tag*{(18)}

where the (linear) operator L\mathcal{L} is defined by

L=(−VI−2Φ0RΦ0Iμ−VR−12∂xx−[Φ0R]2−3[Φ0I]2−μ+VR+12∂xx+3[Φ0R]2+[Φ0I]2−VI(x)+2Φ0RΦ0I),(19)\mathcal{L} = \begin{pmatrix} -V_I - 2\Phi_0^R\Phi_0^I & \mu- V_R - \frac{1}{2}\partial_{xx} - \left[\Phi_0^R\right]^2 - 3\left[\Phi_0^I\right]^2 \\ -\mu+ V_R + \frac{1}{2}\partial_{xx} + 3\left[\Phi_0^R\right]^2 + \left[\Phi_0^I\right]^2 & -V_I(x) + 2\Phi_0^R\Phi_0^I \end{pmatrix}, \tag*{(19)}

and where we have decomposed the wavefunction into its real and imaginary components according to Φ0=Φ0R+iΦ0I\Phi_0 = \Phi_0^R + i\Phi_0^I.

We now proceed to compute the spectrum (λ\lambda) of this linear operator by using Fourier differentiation matrix approximations for the spatial derivative operators (cf. Ref. [53]), in conjunction with the generalized Schur decomposition implemented in Matlab’s eig function. It is worth pointing out that the linear operator L\mathcal{L} in Eq. (19) is unitarily equivalent to the operator M\mathcal{M} arising from the linearization about the bright mode of the phase-transformed model, Eq. (3). Specifically, the two operators are related by

M=(cos⁡θ(x)−sin⁡θ(x)sin⁡θ(x)cos⁡θ(x))−1L(cos⁡θ(x)−sin⁡θ(x)sin⁡θ(x)cos⁡θ(x)),(20)\mathcal{M} = \begin{pmatrix} \cos\theta(x) & -\sin\theta(x) \\ \sin\theta(x) & \cos\theta(x) \end{pmatrix}^{-1} \mathcal{L} \begin{pmatrix} \cos\theta(x) & -\sin\theta(x) \\ \sin\theta(x) & \cos\theta(x) \end{pmatrix}, \tag*{(20)}

where the Wadati phase function is θ(x)=∫−∞xW(x′) dx′\theta(x) = \int_{-\infty}^{x} \mathcal{W}(x')\,\mathrm{d}x'. Furthermore, the associated eigenmodes of M\mathcal{M} satisfy the relationship v=U−1d\mathbf{v} = U^{-1}\mathbf{d}, where UU is the rotation matrix and d\mathbf{d} are the eigenmodes of L\mathcal{L}. Thus, L\mathcal{L} and M\mathcal{M} possess identical eigenspectra. This unitary equivalence justifies the indirect comparison between the PDE stability results obtained from L\mathcal{L} and the NCVA stability predictions developed for the phase-transformed model (see Sec. III).

Concurrently to the PDE calculations, we also compute the fixed points p⃗0\vec{p}_0 of the NCVA ODE system Eq. (11) (see Sec. III) using standard fixed-point iterations. To assess the linear stability of these NCVA fixed points, we then evaluate the eigenvalues and eigenvectors of the corresponding 6×66 \times6 Jacobian matrix, (dF/dp⃗)∣p⃗0\left.(\mathrm{d}\mathbf{F}/\mathrm{d}\vec{p})\right|_{\vec{p}_0}. Naturally, this can only aspire to describe a small subset of point spectrum modes of the full PDE system.

Figure 2 depicts a representative bright soliton, for α=1\alpha= 1 and μ=0.5\mu= 0.5, obtained from the PDE together with its corresponding low-dimensional NCVA approximation. The amplitude profiles, shown in panel (a), exhibit good agreement, with only minor discrepancies near the peak. The NCVA phase profile, which approximates the quantity arg⁡(Φ0)−θ\arg(\Phi_0) - \theta, preserves the PT\mathcal{PT} symmetry of the PDE solution, with vanishing chirp b=0b = 0, and closely matches the PDE phase within the soliton core, as shown in panel (b). Noticeable differences arise only in spatial regions where the amplitude is negligible. The agreement between the PDE and NCVA amplitude profiles persists over most of the continuation interval in α\alpha. The linear limit predicted by the NCVA occurs at α≈1.88\alpha\approx1.88, compared to α≈1.74\alpha\approx1.74 for the PDE bright-soliton branch.

Representative bright mode for the NLS with a $\mathcal{PT}$-symmetric Wadati potential compared with its NCVA counterpart

Figure 2. Representative bright mode for the NLS with a PT\mathcal{PT}-symmetric Wadati potential (blue) compared with its NCVA counterpart (red) for α=1\alpha= 1 and μ=0.5\mu= 0.5. Panels (a) and (b) depict, respectively, the corresponding wavefunction amplitudes and phases

Next, we assess the fidelity of the NCVA in capturing the PDE eigenspectrum as a function of the potential strength parameter α\alpha; see Figure 3. A numerical computation of the eigenspectrum λ\lambda of L\mathcal{L} (or, equivalently, M\mathcal{M}) indicates linear stability, with Re⁡(λ)≡0\operatorname{Re}(\lambda) \equiv0 over the entire existence interval in α\alpha. Accordingly, Figure 3 shows the imaginary part of the eigenspectrum of L\mathcal{L} as a function of α\alpha (blue dots), displayed up to α=1.5\alpha= 1.5 for clarity, although the stability computations were performed over the entire existence interval of the PDE branch, up to its linear limit at α≈1.74\alpha\approx1.74. At α=0\alpha= 0, corresponding to the cubic NLS, four eigenvalues lie at the origin: one pair associated with translational invariance, and another one associated with phase (gauge) invariance. As α\alpha increases, the translational pair splits symmetrically from the origin, given the breaking of the symmetry due to the presence of the potential, and asymptotically approaches the edges of the continuous spectrum at ∣λ∣=μ\lvert\lambda\rvert= \mu. Thus, throughout the entire existence interval in α\alpha, both the direct linear stability analysis and the low-dimensional NCVA predict linearly stable nonlinear bright modes.

Dependence of the purely imaginary eigenvalues on the strength parameter $\alpha$

Figure 3. Dependence of the purely imaginary eigenvalues, Im⁡(λ)\operatorname{Im}(\lambda), on the strength parameter α\alpha of the PT\mathcal{PT}-symmetric Wadati potential. PDE eigenvalues are shown as blue dots and their NCVA counterparts as red circles. The translational eigenvalues, with arrows indicating their motion as α\alpha increases, asymptotically approach the edges of the continuous spectrum.

Interestingly, the NCVA description also yields an additional pair of eigenvalues (given the six-dimensional nature of the relevant representation) whose behavior appears to reflect, in a finite-dimensional sense, the influence of the PDE continuous spectrum. A caveat is warranted here: the low-dimensional approximation cannot genuinely capture the continuous spectrum or its associated delocalized eigenfunctions (cf. Sec. III). Nevertheless, it is noteworthy that this NCVA eigenvalue pair remains embedded within the continuous spectral band of the full PDE throughout the range of α\alpha considered, while progressively deviating from the NCVA approximation of the translational eigenvalue pair. This behavior (in conjunction with that of the pair showcasing the breaking of translational invariance) bears a suggestive resemblance to the asymptotic approach of the PDE translational eigenvalues toward the edges of the continuous spectrum.

It is relevant to mention that accurately computing the spectrum of a non-self-adjoint matrix differential operator on the entire real line is notoriously difficult, frequently resulting in clusters of spurious complex eigenvalues due to numerical artifacts [54, 55, 56, 57]. The strategy we employ to mitigate these spurious effects is to enlarge the computational domain and increase the resolution by employing more Fourier modes—in particular, this led us to observe the algebraic convergence to the true continuous spectrum. Specifically, for the computations reported here, we use the domain [−100,100][-100,100] with N=2048N = 2048 Fourier modes.

Non-Wadati PTPT-symmetric potentials

We now study a representative family of non-Hermitian optical systems possessing PTPT symmetry, but lacking the Wadati structure. To this end, we consider the one-parameter family of potentials defined in Eq. (5). A key difference from Sec. IV A is that, in the present case, the parameters defining the PTPT symmetry need not lie within the so-called “unbroken PTPT-symmetric phase.” In particular, the linear eigenvalue problem

(12d2dx2+V(x))φ=Λφ\left(\frac{1}{2}\frac{d^2}{dx^2}+V(x)\right)\varphi=\Lambda\varphi

may possess complex eigenvalues when the gain-loss component of the potential is sufficiently strong relative to its conservative counterpart (cf. [58] for a discussion of the related Scarf-II potential). At the outset, we remark that the PTPT-symmetry-breaking threshold of the underlying linear problem is numerically estimated to be αPT≈0.62\alpha_{\mathrm{PT}} \approx0.62. As Figure 1(b) evidences, the family of PTPT-symmetric soliton solutions exists for all values of the propagation constant μ\mu larger than the value corresponding to the linear limit, where the norm of the solution tends to zero.

With this in mind, we examine the stability properties of the nonlinear bright solitons for the representative propagation constant μ=0.5\mu=0.5. Notably, the PTPT-symmetry-breaking threshold of the underlying linear problem need not coincide with the onset of instability of the nonlinear bright-soliton branch.

We compute PTPT-symmetric bright solitons for 0<α≤0.60<\alpha\leq0.6 using the numerical continuation scheme discussed in Sec. IV A. Here, we restrict the continuation to a narrow interval beyond the instability threshold of the nonlinear bright solitons. Concurrent to the PDE solutions, we also study the fixed points (and stability) of the NCVA system (10). In Figure 4, we depict a representative case (α=0.55\alpha=0.55), lying in the unbroken PTPT-symmetric phase where the PDE and NCVA wavefunctions show good agreement in amplitude and reasonable agreement in phase [i.e., the phase of the soliton governed by Eq. (10)], particularly in the region of the soliton core.

Representative bright soliton for a non-Wadati, $\mathcal{PT}$-symmetric potential

Figure 4. Same as in Fig. 2 for a representative bright soliton for a non-Wadati, PT\mathcal{PT}-symmetric potential.

We now examine the linear stability spectrum λ(α)\lambda(\alpha) predicted by the PDE and assess the extent to which it is captured by the NCVA description. The resulting spectral properties as a function of the potential depth parameter α\alpha are shown in Figure 5, with the real and imaginary parts of the eigenvalues displayed in panels (a) and (b), respectively. As previously discussed in Sec. IV A, the finite-dimensional NCVA possesses only discrete eigenvalues; nevertheless, in the context of the Hopf bifurcation, one of these eigenvalues effectively tracks the relevant edge of the PDE continuous spectrum. For values of α<0.535\alpha<0.535, the spectrum is entirely imaginary, with the translational eigenmodes approaching the edges of the continuous spectrum as α\alpha increases. At α≈0.455\alpha\approx0.455, an additional discrete mode detaches from the edge of the continuous spectrum and subsequently approaches the translational eigenmodes with increasing α\alpha; see panel (b). At α≈0.535\alpha\approx0.535, these modes collide, giving rise to a complex quartet of eigenvalues through a Hopf-type bifurcation [59], thereby signaling the onset of oscillatory instability of the bright soliton. Remarkably, the NCVA accurately predicts the location of this bifurcation to within the second decimal place, besides respecting the quartet symmetry beyond. Beyond the bifurcation point, however, a small but progressively growing discrepancy between the NCVA and PDE growth rates, associated with the real part of the relevant eigenvalue, emerges as α\alpha increases.

Dependence of the spectrum on the depth parameter $\alpha$ of the $PT$-symmetric potential

Figure 5. Dependence of the spectrum [real part in panel (a) and imaginary part in panel (b)] on the depth parameter α\alpha of the PTPT-symmetric potential. PDE results are shown as blue dots and NCVA eigenvalues as red circles. For smaller α\alpha, the spectrum is purely imaginary, with translational eigenmodes approaching the edges of the continuous spectrum.

In Figure 6 we depict a typical evolution of perturbed, unstable bright solitons in the PDE [Eq. (10)] and NCVA [Eqs. (11)] within a PTPT-symmetric potential of the non-Wadati type, for α=0.55\alpha=0.55. The full PDE dynamics, governed by Eq. (10), is projected onto the solitonic manifold and shown in blue, while the corresponding NCVA predictions are depicted as solid red lines. At early times, the trajectories of both models closely follow the growth rates predicted by their respective linear stability analyses. The close agreement between the projected PDE and NCVA dynamics in this regime suggests that the tangent space of the solitonic manifold is sufficiently rich to capture the dominant unstable PDE eigenmode. As time advances, robust quasi-periodic dynamics emerge within the NCVA framework, which initially appear to qualitatively mimic the projected PDE dynamics. Closer inspection, however, reveals an important discrepancy: the PDE solution emits radiation, a feature that cannot be represented by the spatially localized NCVA ansatz. Consequently, while the NCVA predicts persistent quasi-periodic dynamics confined to the solitonic manifold (see Appendix § 1.5.2), the PDE dynamics are accompanied by weak cw radiation, reflecting degrees of freedom beyond those retained by the finite-dimensional variational approximation. Over long times (tracked up to t=1000t = 1000; not shown here), the projected PDE evolution exhibits qualitatively different behavior, with the amplitude and inverse-width parameters undergoing oscillations about a secularly increasing, phase-locked trend. On the other hand, the oscillations in the chirp intensify over longer times.

Destabilization dynamics of the bright soliton in the $\mathcal{PT}$-symmetric but non-Wadati case

Figure 6. Destabilization dynamics of the bright soliton in the PT\mathcal{PT}-symmetric but non-Wadati case for α=0.55\alpha= 0.55. Shown are the temporal evolution of (a) the amplitude a(t)a(t), (b) the chirp b(t)b(t), and (c) the center-of-mass position ξ(t)\xi(t) of the perturbed bright soliton. The PDE dynamics, projected onto the solitonic manifold, are shown in blue, while the corresponding NCVA dynamics are shown as solid red lines.

Non-PT\mathcal{PT}-symmetric, Wadati potentials

We now consider the case that exhibits Wadati structure without imposing PT\mathcal{PT} symmetry. In particular, we focus on real, single-humped Wadati potential functions W(x)\mathcal{W}(x), i.e., functions that vary monotonically up to a single extremum and monotonically in the opposite sense thereafter. Importantly, single-humped potentials within this class lack an even parity about their extremum. Through the correspondence between the associated Zakharov–Shabat and Schrödinger spectral problems [60], such single-humped functions are known to generate Wadati potentials whose discrete eigenvalues are purely real [27], in contrast to the PT-only case (Sec. IV B).

We note—from the outset—the existence of continuous families of solitons as parametrized by the soliton frequency (μ\mu), as shown in Figure 1(c). For concreteness, we fix Γ=0.1\Gamma= 0.1 and x0=2.75x_0 = 2.75 in Eq. (6). Starting from the cubic NLS bright soliton with propagation constant μ=0.5\mu= 0.5, we perform numerical continuation in the parameter α\alpha over the interval 0<α≲1.70 < \alpha\lesssim1.7 to compute the bright-soliton solutions of Eq. (2). The existence of continuous families of bright solitons in non-PT-symmetric Wadati potentials is a remarkable consequence of an additional conserved quantity associated with the stationary-mode equations [28, 44]. This conserved quantity effectively reduces the number of independent constraints governing smooth bright solitons, thereby closing the stationary problem for a prescribed μ\mu and potential (α,Γ)(\alpha, \Gamma) parameters.

The upper end of this continuation branch corresponds to the relevant linear limit. As expected, the resulting non-centered bright solitons are not PT-symmetric. Now, we also seek to assess whether the six-dimensional NCVA provides an appropriate minimal model. To this end, we concurrently compute the fixed points of Eq. (11) and examine the agreement between the resulting solitonic-manifold approximation and the corresponding PDE solution. As in Sec. IV A, the NCVA slightly overpredicts the soliton amplitude, which, in turn, leads to an approximate 9% extension of the predicted existence interval in α\alpha. A representative comparison for α=1\alpha= 1 is shown in Figure 7. The most notable discrepancies, both qualitative and quantitative, arise in the asymmetric phase profile of the PDE soliton; and thus cannot be adequately represented by the low-dimensional quadratic phase ansatz; see panel (b). Moreover, for the computations presented in this section, the NCVA bright-soliton fixed points satisfy b0<0b_0 < 0 and ξ0>0\xi_0 > 0, in contrast to the PT-symmetric cases, for which ξ0=b0=0\xi_0 = b_0 = 0. We probe the NCVA bright soliton parameters analytically, assuming

Comparison of a representative bright mode in a non-$\mathcal{PT}$-symmetric Wadati potential with its NCVA counterpart

Figure 7. Comparison of a representative bright mode in a non-PT\mathcal{PT}-symmetric Wadati potential with its NCVA counterpart, using parameters α=1\alpha= 1 and μ=0.5\mu= 0.5 (red: NCVA; blue: PDE): (a) The wavefunction amplitude demonstrates good agreement with the six-dimensional NCVA description. (b) The shifted phase of the wavefunction exhibits a substantially more pronounced, and even qualitative, disagreement with the PDE solution than in the case where PT\mathcal{PT} symmetry is also present. This discrepancy persists even within the spatial region where the soliton amplitude is appreciable (see thin vertical lines depicting the width of the bright soliton defined by ∣Φ0∣>1/e2|\Phi_0| > 1/e^2).

that w0(α)≈w0(0)w_0(\alpha) \approx w_0(0), i.e., that the soliton width remains approximately constant as α\alpha varies. This approximation is supported by the computed solutions over a broad range of x0x_0 corresponding to single-humped potentials, for which the variation of w0w_0 remains weak. This behavior can be understood from the fact that the dominant localization scale is set primarily by the underlying potential. Furthermore, for the discussion below, we focus on the case with μ=0.5\mu= 0.5, for which w0(0)=1w_0(0) = 1. We then examine how the weak asymmetry, characterized by Γ≪1\Gamma\ll1 in Eq. (6), induces a nonzero chirp, b0∼Γb1b_0 \sim\Gamma b_1, and a small displacement of the soliton center, ξ0∼Γξ1\xi_0 \sim\Gamma\xi_1. At O(1)\mathcal{O}(1) in the asymmetry parameter, the fixed-point condition Ix=0I_x = 0 is satisfied identically by parity considerations, since ∫−∞∞sech⁡4(x)tanh⁡(x) dx=0\int_{-\infty}^{\infty}\operatorname{sech}^{4}(x)\tanh(x)\,dx = 0. At O(Γ)\mathcal{O}(\Gamma), it yields the expression for the soliton center

ξ1=∫−∞∞sech⁡2(x−x0)sech⁡2(x)tanh⁡(x) dx∫−∞∞sech⁡4(x)[1−3tanh⁡2(x)] dx=30sech⁡2(x0)8tanh⁡4(x0)[x0(3−tanh⁡2(x0))−3tanh⁡(x0)],(21)\xi_1 = \frac{\int_{-\infty}^{\infty}\operatorname{sech}^{2}(x-x_0)\operatorname{sech}^{2}(x)\tanh(x)\,dx}{\int_{-\infty}^{\infty}\operatorname{sech}^{4}(x)\left[1-3\tanh^{2}(x)\right]\,dx} = \frac{30\operatorname{sech}^{2}(x_0)}{8\tanh^{4}(x_0)}\left[x_0\left(3-\tanh^{2}(x_0)\right)-3\tanh(x_0)\right], \tag*{(21)}

whose exact evaluation is standard, by leveraging the identity tanh⁡(x−x0)=[tanh⁡(x)−tanh⁡(x0)]/[1−tanh⁡(x)tanh⁡(x0)]\tanh(x-x_0)=[\tanh(x)-\tanh(x_0)]/[1-\tanh(x)\tanh(x_0)]. In the limit of weak asymmetry, x0≪1x_0 \ll1, the expansion ξ1(x0)≈x0−(10/21)x03+⋯\xi_1(x_0)\approx x_0-(10/21)x_0^3+\cdots reveals an approximately linear dependence of the soliton-center correction on x0x_0, and reduces to the result ξ1(0)=0\xi_1(0)=0 in the PTPT-symmetric limit. In particular, the formula in Eq. (21) indicates that the soliton center is shifted slightly right (since we are using x0>0x_0>0). This prediction agrees reasonably well with the bright-soliton center for α≤1\alpha\leq1, with deviations remaining within approximately 10%. At the same time, the center itself is found numerically to depend only weakly on α\alpha; see Figure 8(b), where the NCVA result is shown by the red solid line and its analytical approximation by the gray dashed line. Across the existence interval, the NCVA inverse-width parameter deviates from unity by as much as approximately 15%, which likely accounts for the increasing discrepancy of the analytical prediction for the center at larger α\alpha. Nevertheless, the weak overall dependence of ξ0\xi_0 on α\alpha can be understood directly from the mass-balance condition. In particular, owing to the weak dependence of w0w_0 on α\alpha, the condition Iχ(ξ0,w0)=0I_\chi(\xi_0,w_0)=0 itself depends only weakly on α\alpha, through its dependence on the inverse-width parameter. This, in turn, explains the weak dependence of the soliton center ξ0\xi_0 on α\alpha.

NCVA parameters compared with analytical approximations and PDE parameters

Figure 8. The NCVA parameters (red curves) are compared with their approximate analytical expressions (gray dashed lines) for the non-PT-symmetric Wadati potentials in Eq. (6), obtained under the assumption of an approximately constant soliton width. The corresponding PDE parameters are also overlaid (blue curves), obtained by projecting the PDE solutions onto the NCVA ansatz through a best-fit procedure. For the parameter set x0=2.75x_0 = 2.75, Γ=0.1\Gamma= 0.1, and μ=0.5\mu= 0.5, the soliton width remains nearly constant throughout the entire existence interval of the NCVA bright-soliton family. Shown here is the variation with α\alpha of (a) the chirp parameter and (b) the soliton center.

It is also useful to highlight the increasing discrepancy, with increasing α\alpha, between the NCVA and PDE soliton centers. The latter is estimated by performing a least-squares best fit of the NCVA solitonic ansatz to the PDE solution, with the global phase parameter fixed to χ0=0\chi_0=0, consistently with the phase convention adopted for the NCVA fixed point; see blue curve in Figure 8(b). On another note, the fixed-point relation b0≈−(3/(2a2π2))Ibb_0\approx-(3/(2a^2\pi^2))I_b yields, at O(Γ)O(\Gamma),

b1=−3απ2[∫−∞∞sech⁡2(x−x0)G(x) dx−2π245ξ1](22)b_1=-\frac{3\alpha}{\pi^2}\left[\int_{-\infty}^{\infty}\operatorname{sech}^2(x-x_0)G(x)\,\mathrm{d}x-\frac{2\pi^2}{45}\xi_1\right] \tag*{(22)}

where G(x)=x2sech⁡2(x)tanh⁡(x)G(x)=x^2\operatorname{sech}^2(x)\tanh(x), and whose evaluation involves a convergent Mellin-type integral [61]. Carrying out the integration, we obtain the approximate expression for the chirp, Γb1=−8Γαx0215π2ξ1<0\Gamma b_1=-\frac{8\Gamma\alpha x_0^2}{15\pi^2}\xi_1<0. This prediction is nearly coincident with the exact NCVA soliton chirp; see Figure 8(a). In contrast, as in the case of the soliton center, the chirp extracted from the PDE solution (blue line) exhibits an increasing discrepancy from the NCVA prediction as α\alpha is increased.

Next, as concerns the PDE stability spectrum, we find that iλα∈Ri\lambda_\alpha\in\mathbb{R} (besides spurious finite-domain numerical artifacts, progressively disappearing as the numerical domain gets larger) throughout the entire interval of 0<α≲1.70<\alpha\lesssim1.7 considered, indicating that the bright solitons remain spectrally stable. The resulting stability spectra is therefore qualitatively similar to that observed in Sec. IV A, and therefore not separately shown. We next examine, see Figure 9, the eigenvalues of the 6×66\times6 NCVA Jacobian as a function of the potential strength 0<α≲1.860<\alpha\lesssim1.86, noting at the outset that the continuous spectrum cannot be represented within this finite-dimensional framework. The arrows in the figure indicate the trajectories of the eigenvalues in the complex-λ\lambda plane as α\alpha is increased. In addition, the instability is exacerbated as the zero-amplitude limit is approached. Finally, the NCVA spectrum lacks the Hamiltonian-like quartet symmetry observed in the full linearization spectra of non-PTPT-symmetric Wadati systems [62]. We examine this issue analytically in the Appendix § 1.5.3, wherein it becomes apparent that the spurious instability is due to the limitations inherent to the chosen variational ansatz.

Trajectory of the NCVA eigenvalues in the complex-$\lambda$ plane

Figure 9. The trajectory of the NCVA eigenvalues λ\lambda as the potential strength α\alpha is increased (see arrows), for the no-PTPT-symmetric, Wadati potential case. The eigenvalues associated with the phase invariance stay fixed to the origin (blue and magenta circles), while the pair associated with the translational invariance of the α=0\alpha=0 case depart from the origin (green and magenta dots). This pair eventually predicts a maximum instability growth rate [Re⁡(λ)∼10−2][\operatorname{Re}(\lambda)\sim10^{-2}] corresponding to the zero-amplitude limit.

Non-Wadati and non-PT\mathcal{PT} symmetric potentials

In this section, we investigate non-Hermitian potentials lacking both PT\mathcal{PT} symmetry and Wadati structure. In particular, we consider the family of potentials defined in Eq. (7), where β≪1\beta\ll1 parametrizes the departure from the Wadati limit. Throughout this section, we additionally restrict the asymmetry parameter Γ\Gamma to be small, such that W(x)=2VR(x)W(x) = \sqrt{2V_R(x)} is a single humped potential function. The existence and properties of bright solitons in dissipative settings (besides the PT\mathcal{PT} and Wadati potentials) have attracted considerable interest, particularly in connection with the persistence, or breakdown, of continuous families of stationary solutions parametrized by μ\mu [39, 63]. Such investigations generally indicate that, for generic non-Hermitian potentials, bright solitons occur only as isolated solutions, since, for a given, prescribed μ\mu, the stationary-mode problem is effectively overdetermined. As in the PDE bright-soliton problem, the NCVA fixed-point system was shown in Sec. III to be generically overdetermined.

To compute bright solitons supported by generic non-Hermitian waveguides, we employ a standard shooting method; see, e.g., Ref. [28]. First, we express the stationary wavefunction in polar form, Φ0(x)=A(x)exp⁡[iΘ(x)]\Phi_0(x) = A(x)\exp[i\Theta(x)], where A(x)A(x) is the wave amplitude and u(x)=Θ′(x)u(x) = \Theta'(x) is the local hydrodynamic velocity (phase gradient). This yields the following system of coupled ODEs of hydrodynamic form,

A′′(x)=2μA(x)+A(x)u2(x)−2VR(x)A(x)−2A3(x),(23)A''(x) = 2\mu A(x) + A(x)u^2(x) - 2V_R(x)A(x) - 2A^3(x), \tag*{(23)}
u′(x)=−2u(x)A′(x)A(x)−2VI(x).(24)u'(x) = -2u(x)\frac{A'(x)}{A(x)} - 2V_I(x). \tag*{(24)}

Given the decay of the potential in the far field, the system can be solved exactly there, yielding A(x)→r±exp⁡(∓2μx)A(x) \to r_{\pm}\exp(\mp\sqrt{2\mu}x) as x→±∞x \to\pm\infty. The two far-field amplitudes r±r_{\pm}, together with the propagation constant μ\mu, constitute three unknown parameters, which are determined by requiring the solutions integrated from the two asymptotic boundaries to match smoothly at x=0x = 0. This gives the three matching conditions

A(0+)=A(0−),A′(0+)=A′(0−),u(0+)=u(0−).A(0^+) = A(0^-), \qquad A'(0^+) = A'(0^-), \qquad u(0^+) = u(0^-).

Having determined the unknown parameters r±r_{\pm} and μ\mu, the ODE system in Eq. (22) can be integrated forwards, after which the corresponding wavefunction is reconstructed as Φ0(x)=A(x)exp⁡[i∫−∞xu(x′) dx′]\Phi_0(x) = A(x)\exp[i\int_{-\infty}^{x}u(x')\,\mathrm{d}x']. We performed an extensive search for stationary solutions in the weakly non-Wadati regime, β≪1\beta\ll1, but were unable to identify single-humped bright solitons, i.e., solutions whose amplitude possesses a single local maximum. Instead, double-humped stationary modes were found generically. Moreover, the separation between the two amplitude maxima was observed to increase relative to the β=0\beta= 0 Wadati limit. A representative example of this double-humped localized solution and its stability spectrum are shown in Figure 10. We note that, as it is the case in the representative case presented in Figure 10, bright solitons in the non-PT\mathcal{PT} and non-Wadati case are generically unstable. Furthermore, unlike the PT\mathcal{PT}-symmetric or Wadati case(s), the quartet symmetry in the spectrum is not preserved; see panel (b) in Figure 10.

Generic bright soliton amplitude and phase profiles and linear stability spectrum

Figure 10. Generic bright solitons supported by non-PT\mathcal{PT}-symmetric and non-Wadati potentials [see Eq. (7)] with parameters α=0.1\alpha= 0.1, Γ=0.1\Gamma= 0.1 and β=0.05\beta= 0.05. The associated propagation constant is found to be μ≈0.694\mu\approx0.694. Panel (a) depicts the amplitude (blue) and phase (red) profiles of the bright soliton while panel (b) depicts its corresponding linear stability spectrum evidencing its unstable nature. Unlike the Wadati case, the spectrum does not display the quartet symmetry.

We next investigate the existence of bright solitons to the NCVA with non-conservative terms (where μ\mu is unknown) in Eq. (10) with R=−iWϕx−i(β/2)Wxϕ\mathcal{R} = -iW\phi_x - i(\beta/2)W_x\phi. In this vein, after suitable manipulation of the Iχ=0I_{\chi} = 0 equation (i.e., an integration by parts) we obtain

2(1−β)a02w0∫−∞∞W sech⁡2(w0(x−ξ0))tanh⁡(w0(x−ξ0)) dx=0.2(1-\beta)a_0^2w_0\int_{-\infty}^{\infty}\mathcal{W}\,\operatorname{sech}^2\left(w_0(x-\xi_0)\right)\tanh\left(w_0(x-\xi_0)\right)\,dx=0.

This yields the usual relationship ξ0∼Γξ1\xi_0\sim\Gamma\xi_1 as was derived in Eq. (21), for a small asymmetry parameter Γ\Gamma. Now we examine the integral

Iξ=2βa02b0∫−∞∞W′(x)(x−ξ0)sech⁡2(w0(x−ξ0)) dx=0.(25)I_{\xi}=2\beta a_0^2b_0\int_{-\infty}^{\infty}\mathcal{W}'(x)(x-\xi_0)\operatorname{sech}^2\left(w_0(x-\xi_0)\right)\,dx=0. \tag*{(25)}

The equation necessitates that either b0=0b_0=0 or the integral vanishes. Note that if W(x)\mathcal{W}(x) has an unique maximum at x=xmx=x_m (as a single-humped function), and utilizing Iχ=0I_{\chi}=0, we can show the integral is negative, since

∫−∞∞W′(x)(x−ξ0)sech⁡2(w0(x−ξ0)) dx+ξ0−xm(1−β)a02Iχ=∫−∞∞W′(x)(x−xm)sech⁡2(w0(x−ξ0)) dx<0.\int_{-\infty}^{\infty}\mathcal{W}'(x)(x-\xi_0)\operatorname{sech}^2\left(w_0(x-\xi_0)\right)\,dx+\frac{\xi_0-x_m}{(1-\beta)a_0^2}I_{\chi}=\int_{-\infty}^{\infty}\mathcal{W}'(x)(x-x_m)\operatorname{sech}^2\left(w_0(x-\xi_0)\right)\,dx<0.

This suggests that b0=0b_0=0. However, the condition Fw=0F_w=0 gives b0=−(3w03/(2a02π2))Ibb_0=-(3w_0^3/(2a_0^2\pi^2))I_b, where

Ib=2a02[(1−β)w0∫−∞∞W(x)(x−ξ0)2sech⁡2[w0(x−ξ0)]tanh⁡[w0(x−ξ0)] dx+β∫−∞∞W(x)(x−ξ0)sech⁡2[w0(x−ξ0)] dx].I_b=2a_0^2\left[(1-\beta)w_0\int_{-\infty}^{\infty}\mathcal{W}(x)(x-\xi_0)^2\operatorname{sech}^2\left[w_0(x-\xi_0)\right]\tanh\left[w_0(x-\xi_0)\right]\,dx +\beta\int_{-\infty}^{\infty}\mathcal{W}(x)(x-\xi_0)\operatorname{sech}^2\left[w_0(x-\xi_0)\right]\,dx\right].

Assuming β∼Γ≪1\beta\sim\Gamma\ll1, the leading-order contribution to IbI_b is inherited from the β=0\beta=0 Wadati limit and is generically nonzero, while the departure from the Wadati type of potential introduces only O(β)O(\beta) corrections to IbI_b. Since b0∼αΓb1b_0\sim\alpha\Gamma b_1 [see discussion of Eq. (22)] at β=0\beta=0, continuity for sufficiently small β\beta therefore suggests that IbI_b remains nonzero, and hence that b0≠0b_0\ne0. This is incompatible with the condition b0=0b_0=0 inferred from Fc=0F_c=0 [which can be deduced from Iξ=0I_{\xi}=0, see Eq. (25)], indicating a possible incompatibility between the Fc=0F_c=0 and Fw=0F_w=0 fixed-point equations and, consequently, the possible absence of NCVA fixed points in the regime β∼Γ≪1\beta\sim\Gamma\ll1.

Another intriguing aspect corresponds to the fate of the (single-humped) Wadati bright solitons (β=0\beta=0) when the Wadati structure of the underlying potential is perturbed. This of course has implications for their robustness, and thus, experimental observability. We generally observe that Wadati bright solitons (β=0\beta=0) appear to play an important role in the dynamics for β≪1\beta\ll1 as the mode executes small amplitude oscillations about its unperturbed center, while the envelope of oscillations is also increasing slightly in the projected soliton amplitude; see Figure 11 for a representative example (for α=0.1\alpha=0.1, Γ=0.1\Gamma=0.1, and β=0.05\beta=0.05). Similar behavior has also been reported for “pseudomodes” of the so-called W-dW potentials [39]. Notably, the projected soliton amplitude and width evolve nearly in phase and remain closely locked throughout the dynamics, while the oscillations of the projected soliton center are approximately π/2\pi/2 out of phase with them (see below).

Dynamics of Wadati bright solitons under a weak perturbation

Figure 11. The dynamics of Wadati bright solitons when the Wadati potential is weakly perturbed so as to no longer possess the Wadati property. The parameters characterizing the underlying Wadati potential are α=0.1\alpha=0.1, Γ=0.1\Gamma=0.1, x0=2.75x_0=2.75, and propagation constant μ=0.5\mu=0.5, while the perturbation parameter is β=0.05\beta=0.05. The PDE dynamics (blue curves) are projected onto the NCVA ansatz through a best-fit procedure and compared with the corresponding NCVA dynamics (red curves). (a) The peak amplitude of the bright soliton oscillates in phase with the inverse-width parameter, with the envelope of both oscillations exhibiting approximately linear growth. (b) The soliton center executes approximately periodic oscillations about the unperturbed equilibrium position ξ0\xi_0 corresponding to β=0\beta=0.

Interestingly, when the potential is perturbed away from the Wadati limit (β≠0\beta\ne0, Figure 11), the dynamics of the NCVA solitons initialized from the Wadati limit exhibit very good agreement with the PDE dynamics. Both the (projected) PDE and NCVA width parameters (not shown) oscillate in phase with the amplitude parameter and possess nearly the same temporal profile (i.e., a(t)≈w(t)a(t) \approx w(t)), consistent with the relationship ddt(∫R∣ϕ∣2dx)=ddt(2a(t)2/w(t))≈2a˙(t)\frac{d}{dt}\left(\int_{\mathbb{R}} \lvert\phi\rvert^2 dx\right) = \frac{d}{dt}\left(2a(t)^2/w(t)\right) \approx2\dot{a}(t). Moreover, the NCVA soliton center oscillates about its unperturbed (β=0\beta= 0) equilibrium position, with a phase difference of nearly π/2\pi/2 radians relative to the soliton amplitude, akin to the projected PDE dynamics in Figure 11(b).

Conclusions and scope for future work

In this work, we developed a unified classification of the non-Hermitian potentials based on their connection to the PT\mathcal{PT} and the Wadati property. This in turn yielded four distinct settings: systems possessing both properties, systems with PT\mathcal{PT} symmetry alone, systems with Wadati structure alone, and systems possessing neither. We examined the existence and stability of bright solitons in non-Hermitian optical systems. Within the above classification, we also assessed the extent to which a low six-dimensional variational reduction can faithfully capture the existence and spectral stability of the corresponding soliton families within each category.

When at least one of these structural properties is present, continuous families of bright solitons parametrized by their (temporal) frequency μ\mu are found to exist, a feature that is also captured by the variational approximation. At the NCVA level, this can be understood in part from the fact that the corresponding algebraic system governing the fixed points is exactly determined. For the PT\mathcal{PT}-symmetric families examined here, the NCVA reproduces the relevant PDE stability behavior with good accuracy. For the case in which Wadati structure is present in the absence of PT\mathcal{PT} symmetry, the NCVA successfully reproduces the bright-soliton intensity, but fails to accurately capture the asymmetric features of phase structure. A spurious instability in the NCVA reduced model is also witnessed, perhaps due to the inadequacy of the variational ansatz itself (see Appendix 3). It would be interesting to develop an improved solitonic ansatz that could incorporate a potential-adapted phase correction (while keeping the same number of parameters). Of interest would also be to check if an NCVA with such an improved ansatz could also preserve the quartet symmetry of the linearized stability spectrum as entailed by the Wadati structure.

For the near-Wadati family considered here, our numerical searches yield isolated, typically double-humped and unstable localized states, and we have not identified corresponding single-humped stationary modes in the parameter regime examined. This observation is consistent with the overdetermined system of equations governing the existence of NCVA fixed points. Moreover, weak departures from the Wadati limit lead, within the NCVA framework, to an incompatibility that (generically) precludes single-humped bright-soliton fixed points. Whether this obstruction at the variational level can be translated into a more concrete statement regarding the existence of single-humped NLS solitons under weak departures from Wadati structure remains an interesting question for future work. Yet another extension would be to generalize the NCVA ansatz so as to accommodate multi-humped localized states.

Further interesting directions include investigating how the NCVA framework may be extended to accommodate symmetry-broken branches in PT\mathcal{PT}-symmetric Wadati potentials [35]. Other natural avenues for future work include developing variational approximations for heteroclinic and homoclinic states of the defocusing NLS with Wadati potentials [64], and also carrying out a more comprehensive study of non-Hermitian scattering phenomena in both focusing and defocusing settings not only in one, but also in higher dimensions where associated studies are far more sparse.

Appendix

Non-conservative variational approximation

The non-conservative variational approximation (NCVA) represents a semi-analytical method [46] to dynamically reduce the NLS equation —in the presence of gain and loss terms— to an effective “particle” picture described by a set of coupled ODEs on the soliton parameters. To employ the NCVA, we consider two sets of coordinates ϕ1\phi_1 and ϕ2\phi_2, that are fixed at an initial time (tit_i), but are not fixed at the final time (tft_f). Upon application of variational calculus for a non-conservative system, both paths are set equal, ϕ1=ϕ2\phi_1 = \phi_2, and identified with the physical path ϕ\phi, i.e., “the physical limit” (PL). The action functional for ϕ1\phi_1 and ϕ2\phi_2 is defined as the total line integral of the difference of the Lagrangians (L\mathcal{L}) between the paths with the sum of the line integral of the functional R\mathcal{R} which describes the non-conservative, path-dependent forces:

S=∫titf(L(ϕ1,ϕ1,t,ϕ1,x,…,t)−L(ϕ2,ϕ2,t,ϕ2,x,…,t)+R) dt,(26)S=\int_{t_i}^{t_f}\left(\mathcal{L}(\phi_1,\phi_{1,t},\phi_{1,x},\ldots,t)-\mathcal{L}(\phi_2,\phi_{2,t},\phi_{2,x},\ldots,t)+\mathcal{R}\right)\,dt, \tag*{(26)}

Here, the tt and xx subscripts represent the partial derivatives with respect to the corresponding variables. The action in Eq. (26) defines a non-conservative Lagrangian density L~≡L1−L2+R\widetilde{\mathcal{L}}\equiv\mathcal{L}_1-\mathcal{L}_2+\mathcal{R}. Then, the equations of motion at the physical limit yield

∂∂t(δLδϕt∗)=δLδϕ∗+[δRδϕ−∗]PL,\frac{\partial}{\partial t}\left(\frac{\delta\mathcal{L}}{\delta\phi_t^*}\right)=\frac{\delta\mathcal{L}}{\delta\phi^*}+\left[\frac{\delta\mathcal{R}}{\delta\phi_-^*}\right]_{\mathrm{PL}},

where δ\delta denotes the Fréchet derivatives and ϕ−≡ϕ1−ϕ2\phi_-\equiv\phi_1-\phi_2. We thus obtain the Euler-Lagrange (EL) equation for the conservative terms and all non-conservative terms are lumped into [δR/δϕ−∗]PL[\delta\mathcal{R}/\delta\phi_-^*]_{\mathrm{PL}}. Notably, this field-theoretic formulation of the non-conservative problem has not deployed any approximation yet. Now, we proceed to solve the above EL equations using a suitable ansatz to approximately describe solitons in the non-conservative Lagrangian formulation. An effective method to study the dynamics of the solution is to restricted it on a solitonic manifold, i.e., assume an ansatz ϕ≡ϕˉ(x,p⃗(t))\phi\equiv\bar{\phi}(x,\vec{p}(t)) with time-dependent parameters p⃗(t)\vec{p}(t), for which the modified EL equations for the averaged Lagrangian Lˉ=∫−∞∞Lˉ dx\bar{L}=\int_{-\infty}^{\infty}\bar{\mathcal{L}}\,dx result in the ODE system

ddt(∂Lˉ∂p⃗)−∂Lˉ∂p⃗−Ip⃗=0,\frac{d}{dt}\left(\frac{\partial\bar{L}}{\partial\vec{p}}\right)-\frac{\partial\bar{L}}{\partial\vec{p}}-I_{\vec{p}}=0,

where the integrals Ip⃗I_{\vec{p}} are given by Ip⃗≡∫−∞∞[∂R/∂p⃗−]PLdxI_{\vec{p}}\equiv\int_{-\infty}^{\infty}\left[\partial\mathcal{R}/\partial\vec{p}_{-}\right]_{\mathrm{PL}}dx and the overline indicates evaluation on the solitonic ansatz.

Periodic Orbits and Bifurcations in the NCVA Dynamics of the PT-Symmetric, non-Wadati case

We comment further on the long-time dynamics of the finite-dimensional NCVA description observed in Figure 6, particularly the emergence of multiscale, quasi-periodic breather dynamics. To elucidate their origin, we compute the branch of periodic breather solutions bifurcating from the Hopf bifurcation of the stationary bright soliton identified in Figure 5, and subsequently examine the stability and secondary bifurcations of these periodic orbits. At α=0.55\alpha=0.55, the stationary NCVA solution lies beyond the Hopf bifurcation point. The resulting complex quartet {λ,λ∗,−λ,−λ∗}\{\lambda,\lambda^*,-\lambda,-\lambda^*\} contains the eigenvalue λ≈0.0472+0.5110i\lambda\approx0.0472+0.5110i. We denote the corresponding complex eigenvector by V⃗=a⃗+ib⃗\vec{V}=\vec{a}+i\vec{b}, where a⃗,b⃗∈R6\vec{a},\vec{b}\in\mathbb{R}^6. Periodic orbits of the NCVA system (10) are then computed as fixed points of the time-TT map, χT[P⃗(0)]=P⃗(0)\chi^T[\vec{P}(0)]=\vec{P}(0), where χT[⋅]\chi^T[\cdot] denotes the flow map of the dynamical system over a time interval TT and P⃗(0)\vec{P}(0) denotes the initial point on the periodic orbit. An additional phase-pinning condition, a(0)=a0a(0)=a_0, is appended to this shooting problem, thereby selecting the a priori unknown breather propagation constant μ\mu. We perform continuation in the period TT, with the initial period informed by the imaginary part of the Hopf eigenvalue, T0=2π/Im⁡(λ)T_0=2\pi/\operatorname{Im}(\lambda). The shooting computation is initialized using P⃗breather(0)≈p⃗0+ϵa⃗\vec{P}_{\mathrm{breather}}(0)\approx\vec{p}_0+\epsilon\vec{a}, where ϵ≪1\epsilon\ll1 and p⃗0\vec{p}_0 denotes the fixed point of the NCVA ODE system. To determine the stability of the resulting periodic breather orbits, we compute their Floquet multipliers Λ\Lambda, i.e., the eigenvalues of the monodromy matrix obtained by linearizing the time-TT map about each periodic orbit.

Stability is determined by the location of these multipliers relative to the unit circle in the complex plane. Importantly, the periodic breather branch is itself found to lose stability through a Neimark–Sacker bifurcation of the associated Poincaré map [65], at T≈12.37T\approx12.37. This secondary instability of the periodic breather thus provides a natural mechanism (and a qualitative explanation) for the observed beating and multiscale NCVA quasi-periodic dynamics in Figure 6. Thus, the long-time NCVA dynamics should be interpreted as dynamics associated with the instability of the breather branch beyond its Neimark–Sacker bifurcation.

NCVA solitons in non-PT-symmetric, Wadati potentials

In this section, we investigate the origin of the spurious NCVA instability identified in Sec. IV C, and discuss how it may arise from limitations inherent to the variational ansatz considered. To begin our discussion, we recall the identity Tr⁡(J)=∇p⃗⋅F⃗(p⃗;α)∣p⃗0=∑i=16λi\operatorname{Tr}(J)=\left.\nabla_{\vec{p}}\cdot\vec{F}(\vec{p};\alpha)\right|_{\vec{p}_0}=\sum_{i=1}^{6}\lambda_i. For a neutrally stable fixed point with a purely imaginary spectrum, the trace must therefore vanish, whereas asymptotic stability requires Tr⁡(J)<0\operatorname{Tr}(J)<0. More generally, the Hamiltonian-like quartet symmetry {λ,−λ,λ∗,−λ∗}\{\lambda,-\lambda,\lambda^{*},-\lambda^{*}\} likewise requires Tr⁡(J)=∇p⃗⋅F⃗(p⃗;α)∣p⃗0=0\operatorname{Tr}(J)=\left.\nabla_{\vec{p}}\cdot\vec{F}(\vec{p};\alpha)\right|_{\vec{p}_0}=0. To estimate this quantity analytically, we note at the outset, that gauge invariance implies ∂Fχ/∂χ∣p⃗0=0\left.\partial F_{\chi}/\partial\chi\right|_{\vec{p}_0}=0. Furthermore, evaluation of the NCVA equations at the fixed point yields ∂Fa/∂a∣p⃗0≡0\left.\partial F_a/\partial a\right|_{\vec{p}_0}\equiv0. Finally, for Wadati potentials, the fixed-point condition Fc(p⃗0)=0F_c(\vec{p}_0)=0, together with Iξ≡0I_{\xi}\equiv0 [cf. the discussion below Eq. (11)], yields ∂Fc/∂c∣p⃗0≡0\left.\partial F_c/\partial c\right|_{\vec{p}_0}\equiv0. Thus, the trace of the 6×66\times6 Jacobian is given by

∇p⃗⋅F⃗(p⃗;α)∣p⃗0=[∂Fb∂b+∂Fξ∂ξ+∂Fw∂w]p⃗0\left.\nabla_{\vec{p}}\cdot\vec{F}(\vec{p};\alpha)\right|_{\vec{p}_0} = \left[ \frac{\partial F_b}{\partial b} + \frac{\partial F_{\xi}}{\partial\xi} + \frac{\partial F_w}{\partial w} \right]_{\vec{p}_0}

As the localization scale is dominantly set by the complex potential (Sec. IV C), we assume that w0(α)≈w0(0)w_0(\alpha)\approx w_0(0), i.e., that the soliton width remains approximately constant as α\alpha varies. Furthermore, we focus on the parameter μ=0.5\mu=0.5, for which w0(0)=1w_0(0)=1.

With the expressions for the chirp Γb1=−8Γαx215π2ξ1<0\Gamma_{b1}=-\frac{8\Gamma\alpha x_2}{15\pi^2}\xi_1<0 and soliton center in Eq. (21), we obtain the three contributions to the Jacobian trace, which, upon retaining terms up to O(Γ)\mathcal{O}(\Gamma), yields

∂Fb∂b∼6αΓπ2[∫−∞∞sech⁡2(x−x0)xsech⁡2(x) dx−23ξ1]=6αΓπ2[2x0csch⁡2(x0)(x0coth⁡(x0)−1)−23ξ1],(27)\frac{\partial F_b}{\partial b} \sim\frac{6\alpha\Gamma}{\pi^2}\left[\int_{-\infty}^{\infty}\operatorname{sech}^2(x-x_0)x\operatorname{sech}^2(x)\,\mathrm{d}x-\frac{2}{3}\xi_1\right] = \frac{6\alpha\Gamma}{\pi^2}\left[2x_0\operatorname{csch}^2(x_0)\left(x_0\operatorname{coth}(x_0)-1\right)-\frac{2}{3}\xi_1\right], \tag*{(27)}
∂Fξ∂ξ∼αΓ[∫−∞∞ddx[sech⁡2(x−x0)]xsech⁡2(x)tanh⁡(x) dx+415ξ1],(28)\frac{\partial F_\xi}{\partial\xi} \sim\alpha\Gamma\left[\int_{-\infty}^{\infty}\frac{\mathrm{d}}{\mathrm{d}x}\left[\operatorname{sech}^2(x-x_0)\right]x\operatorname{sech}^2(x)\tanh(x)\,\mathrm{d}x+\frac{4}{15}\xi_1\right], \tag*{(28)}
∂Fw∂w∼αΓ[∫−∞∞sech⁡2(x−x0)H(x) dx+45ξ1],(29)\frac{\partial F_w}{\partial w} \sim\alpha\Gamma\left[\int_{-\infty}^{\infty}\operatorname{sech}^2(x-x_0)H(x)\,\mathrm{d}x+\frac{4}{5}\xi_1\right], \tag*{(29)}

where

H(x)=12xsech⁡2(x)[1−3tanh⁡2(x)]−24π2x2sech⁡2(x)tanh⁡(x)−6π2x3sech⁡2(x)[1−3tanh⁡2(x)].H(x)=\frac{1}{2}x\operatorname{sech}^2(x)\left[1-3\tanh^2(x)\right]-\frac{24}{\pi^2}x^2\operatorname{sech}^2(x)\tanh(x)-\frac{6}{\pi^2}x^3\operatorname{sech}^2(x)\left[1-3\tanh^2(x)\right].

The integral on the RHS of Eq. (28) can be evaluated exactly as 4x0csch⁡2(x0)[x0coth⁡(x0)(1+3csch⁡2(x0))−2−3csch⁡2(x0)]4x_0\operatorname{csch}^2(x_0)\left[x_0\operatorname{coth}(x_0)\left(1+3\operatorname{csch}^2(x_0)\right)-2-3\operatorname{csch}^2(x_0)\right], while the integral on the RHS of Eq. (29) yields

∫−∞∞sech⁡2(x−x0)H(x) dx=1π2csch⁡2(x0){4x0(3x02+π2)[x0coth⁡(x0)(1+3csch⁡2(x0))−2−3csch⁡2(x0)]+(16x02+2π2)[3coth⁡(x0)−x0(2+3csch⁡2(x0))]}.\int_{-\infty}^{\infty}\operatorname{sech}^2(x-x_0)H(x)\,\mathrm{d}x=\frac{1}{\pi^2}\operatorname{csch}^2(x_0)\left\{4x_0(3x_0^2+\pi^2)\left[x_0\operatorname{coth}(x_0)\left(1+3\operatorname{csch}^2(x_0)\right)-2-3\operatorname{csch}^2(x_0)\right]+(16x_0^2+2\pi^2)\left[3\operatorname{coth}(x_0)-x_0\left(2+3\operatorname{csch}^2(x_0)\right)\right]\right\}.

For x0=2.75x_0=2.75 and Γ=0.1\Gamma=0.1, we compare the numerically extracted Jacobian trace (positive across the entire range in α\alpha) with the constant-ww approximation and find agreement to within 11%. This level of agreement is reasonable given that w(α)w(\alpha) itself varies by up to 16.7% over the full range of α\alpha considered. Figure 12, depicting the trace of the NCVA Jacobian, indicates that, in the vicinity of the fixed point p⃗0\vec{p}_0, the reduced flow is locally volume-expanding on the solitonic manifold.

Trace of the NCVA Jacobian for the non-$\mathcal{PT}$-symmetric Wadati potential

Figure 12. The trace of the NCVA Jacobian for the non-PT\mathcal{PT}-symmetric Wadati potential (red dots: numerical evaluation of Jacobian trace, blue line: analytical prediction) shown for the entire existence regime 0<α≲1.860 < \alpha\lesssim1.86 for μ=0.5\mu= 0.5.

Beyond the spectral discrepancies discussed above, we also examine the differences between the NCVA and PDE representations of the eigenmodes associated with the eigenvalue pair that moves away from the origin upon the breaking of translational invariance for α>0\alpha>0. A representative comparison at α=0.6\alpha=0.6 is shown in Figure 13. At this parameter value, the PDE eigenvalue is λ≈0.35i\lambda\approx0.35i, while the corresponding NCVA eigenvalue is λ≈−0.0011+0.3781i\lambda\approx-0.0011+0.3781i. Thus, although the NCVA eigenvalue has already acquired a small nonzero real part, it has not yet crossed into the right-half plane. The NCVA approximation to the corresponding two-component eigenmode [ζ(1)(x),ζ(2)(x)]T[\zeta^{(1)}(x),\zeta^{(2)}(x)]^{T} of the linear stability operator Eq. (20) is constructed as

Components of the eigenmode

Figure 13. The components of the eigenmode [ζ(1)(x),ζ(2)(x)]T[\zeta^{(1)}(x),\zeta^{(2)}(x)]^{T} associated with the eigenvalue λ≈0.35i\lambda\approx0.35i of the linear stability operator M\mathcal{M} corresponding to Eq. (3). The PDE eigenmode (blue) is compared with its NCVA counterpart (red).

[∑i=16vi(j)∂piRe⁡(ϕ(x,p⃗0)),∑i=16vi(j)∂piIm⁡(ϕ(x,p⃗0))]T.\left[\sum_{i=1}^{6}v_i^{(j)}\partial_{p_i}\operatorname{Re}\left(\phi(x,\vec{p}_0)\right),\sum_{i=1}^{6}v_i^{(j)}\partial_{p_i}\operatorname{Im}\left(\phi(x,\vec{p}_0)\right)\right]^{T}.

where v⃗(j)\vec{v}^{(j)} denotes the associated NCVA (normalized) eigendirection. As shown in Figure 13, appreciable quantitative discrepancies are present in all four components, while panel (d) additionally exhibits a clear qualitative mismatch between the NCVA and PDE eigenmodes. In particular, this indicates that the tangent space spanned by the NCVA manifold may be insufficient to accurately represent the corresponding PDE eigenmodes.

Acknowledgments

We acknowledge the support from the US National Science Foundation under Grants No. PHY-2408989 (R.C.G.) and No. PHY-2408988 (P.G.K.).

References

References

  1. [3]V. V. Konotop, J. Yang, and D. A. Zezyulin, Nonlinear waves in PT-symmetric systems, Rev. Mod. Phys. 88, 035002 (2016).DOI
  2. [4]D. Christodoulides and J. Yang, eds., Parity-time Symmetry and Its Applications, Springer Tracts in Modern Physics, Vol. 280 (Springer Singapore, Singapore, 2018).DOI
  3. [5]C. M. Bender and S. Boettcher, Real spectra in non-Hermitian Hamiltonians having PT symmetry, Phys. Rev. Lett. 80, 5243 (1998).DOI
  4. [6]A. Mostafazadeh, Pseudo-Hermiticity versus PT-symmetry II: A complete characterization of non-Hermitian Hamiltonians with a real spectrum, arXiv preprint math-ph/0110016 (2001).DOI
  5. [7]A. Mostafazadeh, Pseudo-Hermiticity versus PT symmetry: The necessary condition for the reality of the spectrum of a non-Hermitian Hamiltonian, J. Math. Phys. 43, 205 (2002).DOI
  6. [8]L. Feng, R. El-Ganainy, and L. Ge, Non-Hermitian photonics based on parity-time symmetry, Nature Photonics 11, 752 (2017).DOI
  7. [9]K. G. Makris, R. El-Ganainy, D. Christodoulides, and Z. H. Musslimani, Beam dynamics in PT symmetric optical lattices, Phys. Rev. Lett. 100, 03904 (2008).pubmed.ncbi.nlm.nih.gov/18352189
  8. [10]R. El-Ganainy, K. Makris, D. Christodoulides, and Z. H. Musslimani, Theory of coupled optical PT-symmetric structures, Opt. Lett. 32, 2632 (2007).
  9. [11]Z. H. Musslimani, K. G. Makris, R. El-Ganainy, and D. N. Christodoulides, Optical solitons in PT periodic potentials, Phys. Rev. Lett. 100, 030402 (2008).
  10. [12]K. G. Makris, R. El-Ganainy, D. N. Christodoulides, and Z. H. Musslimani, PT-symmetric optical lattices, Phys. Rev. A 81, 063807 (2010).DOI
  11. [13]A. Ruschhaupt, F. Delgado, and J. Muga, Physical realization of PT-symmetric potential scattering in a planar slab waveguide, Journal of Physics A: Mathematical and General 38, L171 (2005).DOI
  12. [14]A. Guo, G. J. Salamo, D. Duchesne, R. Morandotti, M. Volatier-Ravat, V. Aimez, G. A. Siviloglou, and D. N. Christodoulides, Observation of PT-symmetry breaking in complex optical potentials, Phys. Rev. Lett. 103, 093902 (2009).pubmed.ncbi.nlm.nih.gov/19792798
  13. [15]C. E. Rüter, K. G. Makris, R. El-Ganainy, D. N. Christodoulides, M. Segev, and D. Kip, Observation of parity-time symmetry in optics, Nature Photonics 6, 192 (2010).DOI
  14. [16]A. Regensburger, C. Bersch, M.-A. Miri, G. Onishchukov, D. N. Christodoulides, and U. Peschel, Parity–time synthetic photonic lattices, Nature 488, 167 (2012).DOI
  15. [17]L. Feng, Y.-L. Xu, W. S. Fegadolli, M.-H. Lu, J. E. Oliveira, V. R. Almeida, Y.-F. Chen, and A. Scherer, Experimental demonstration of a unidirectional reflectionless parity-time metamaterial at optical frequencies, Nature Materials 12, 108 (2013).DOI
  16. [18]L. Chang, X. Jiang, S. Hua, C. Yang, J. Wen, L. Jiang, G. Li, G. Wang, and M. Xiao, Parity–time symmetry and variable optical isolation in active–passive-coupled microrresonators, Nature Photonics 8, 524 (2014).
  17. [19]Z. Zhang, Y. Zhang, J. Sheng, L. Yang, M.-A. Miri, D. N. Christodoulides, B. He, Y. Zhang, and M. Xiao, Observation of parity-time symmetry in optically induced atomic lattices, Phys. Rev. Lett. 117, 123601 (2016).DOI
  18. [20]C. Hang and G. Huang, Parity-time symmetry with coherent atomic gases, Advances in Physics: X 2, 737 (2017).
  19. [21]L. Huang, S. Huang, C. Shen, S. Yves, A. S. Pilipchuk, X. Ni, S. Kim, Y. K. Chiang, D. A. Powell, J. Zhu, et al., Acoustic resonances in non-Hermitian open systems, Nature Reviews Physics 6, 11 (2024).DOI
  20. [22]P. Charumathi and K. Senthilnathan, Unravelling PT symmetry: Applications in metamaterials, Plasmonics 20, 1767 (2025).DOI
  21. [23]T. Gao, E. Estrecho, K. Blio kh, T. Liew, M. Fraser, S. Brodbeck, M. Kamp, C. Schneider, S. Höfling, Y. Yamamoto, et al., Observation of non-Hermitian degeneracies in a chaotic exciton-polariton billiard, Nature 526, 554 (2015).DOI
  22. [24]J. Li, A. K. Harter, J. Liu, L. de Melo, Y. N. Joglkar, and L. Luo, Observation of parity-time symmetry breaking transitions in a dissipative Floquet system of ultracold atoms, Nature Communications 10, 855 (2019).DOI
  23. [25]J. Schindler, A. Li, M. C. Zheng, F. M. Ellis, and T. Kottos, Experimental study of active LRC circuits with PT symmetries, Phys. Rev. A 84, 040101 (2011).DOI
  24. [26]M. Wadati, Construction of parity-time symmetric potential through the soliton theory, Journal of the Physical Society of Japan 77, 074005 (2008).DOI
  25. [27]E. N. Tsoy, I. M. Allayarov, and F. K. Abdullaev, Stable localized modes in asymmetric waveguides with gain and loss, Opt. Lett. 39, 4215 (2014).DOI
  26. [28]V. V. Konotop and D. A. Zezyulin, Families of stationary modes in complex potentials, Opt. Lett. 39, 5535 (2014).DOI
  27. [29]I. Barashenkov, D. Zezyulin, and V. V. Konotop, Exactly solvable Wadati potentials in the PT-symmetric Gross-Pitaevskii equation, in Non-Hermitian Hamiltonians in Quantum Physics: Selected Contributions from the 15th International Conference on Non-Hermitian Hamiltonians in Quantum Physics, Palermo, Italy, 18-23 May 2015, Springer Proceedings in Physics, Vol. 184 (Springer, 2016) pp. 143–155.DOI
  28. [30]D. A. Zezyulin, Nonlinear Schrödinger equations with amplitude-dependent Wadati potentials, Phys. Rev. E 106, 054209 (2022).DOI
  29. [31]J. Yang, Analytical construction of soliton families in one-and two-dimensional nonlinear Schrödinger equations with nonparity-time-symmetric complex potentials, Stud. Appl. Math. 147, 4 (2021).DOI
  30. [32]B. Bagchi and J. Yang, New families of non-parity-time-symmetric complex potentials with all-real spectra, J. Math. Phys. 61 (2020).DOI
  31. [33]K. G. Makris, Z. H. Musslimani, D. N. Christodoulides, and S. Rotter, Constant-intensity waves and their modulation instability in non-Hermitian potentials, Nature Communications 6, 7257 (2015).DOI
  32. [34]K. G. Makris, A. Brandstötter, P. Ambichl, Z. H. Musslimani, and S. Rotter, Wave propagation through disordered media without backscattering and intensity variations, Light: Science & Applications 6, e17035 (2017).
  33. [35]J. Yang, Symmetry breaking of solitons in one-dimensional parity-time-symmetric optical potentials, Opt. Lett. 39, 5547 (2014).DOI
  34. [36]K. G. Makris, Z. H. Musslimani, D. N. Christodoulides, and S. Rotter, Constant intensity supermodes in non-Hermitian lattices, IEEE Journal of Selected Topics in Quantum Electronics 22, 42 (2016).DOI
  35. [37]M. J. Ablowitz and Z. H. Musslimani, Inverse scattering transform for the integrable nonlocal nonlinear Schrödinger equation, Nonlinearity 29, 915 (2016).DOI
  36. [38]J. Yang, Necessity of PT symmetry for soliton families in one-dimensional complex potentials, Phys. Lett. A 378, 367 (2014).DOI
  37. [39]D. A. Zyuzulin, A. O. Slobodenyuk, and G. L. Alfimov, On nonexistence of continuous families of stationary nonlinear modes for a class of complex potentials, Stud. Appl. Math. 148, 99 (2022).
  38. [40]M. Wimmer, A. Regensburger, M.-A. Miri, C. Bersch, D. N. Christodoulides, and U. Peschel, Observation of optical solitons in PT-symmetric lattices, Nature Communications 6, 7782 (2015).DOI
  39. [41]U. A. Khawaja, S. Al-Marzouqi, H. Bahlouli, and Y. S. Kivshar, Unidirectional soliton flows in PT-symmetric potentials, Phys. Rev. A 88, 023830 (2013).
  40. [42]F. K. Abdullaev, V. Brazhnyi, and M. Salerno, Scattering of gap solitons by PT-symmetric defects, Phys. Rev. A 88, 043829 (2013).
  41. [43]Y. Kivshar and G. Agrawal, Optical Solitons: From Fibers to Photonic Crystals (Elsevier Science, 2003).
  42. [44]S. D. Nixon and J. Yang, Bifurcation of soliton families from linear modes in non-PT-symmetric complex potentials, Stud. Appl. Math. 136, 459 (2016).
  43. [45]S. Nixon and J. Yang, Nonlinear light behaviors near phase transition in non-parity-time-symmetric complex waveguides, Opt. Lett. 41, 2747 (2016).DOI
  44. [46]J. Rossi, R. Carretero-González, and P. Kevrekidis, Non-conservative variational approximation for nonlinear Schrödinger equations, The European Physical Journal Plus 135, 1 (2020).DOI
  45. [47]J. Rossi, S. Chandramouli, R. Carretero-González, and P. G. Kevrekidis, On the temporal tweezing of cavity solitons, Journal of Nonlinear Mathematical Physics 31, 43 (2024).arxiv.org/abs/2304.05925
  46. [48]C. R. Galley, Classical mechanics of nonconservative systems, Phys. Rev. Lett. 110, 174301 (2013).pubmed.ncbi.nlm.nih.gov/23679733
  47. [49]C. R. Galley, D. Tsang, and L. C. Stein, The principle of stationary nonconservative action for classical mechanics and field theories, arXiv preprint arXiv:1412.3082 (2014).arxiv.org/abs/1412.3082
  48. [50]P. Kevrekidis, Variational method for nonconservative field theories: Formulation and two PT-symmetric case examples, Phys. Rev. A 89, 010102 (2014).DOI
  49. [51]J. Yang, Can parity-time-symmetric potentials support families of non-parity-time-symmetric solitons?, Stud. Appl. Math. 132, 332 (2014).
  50. [52]J. Yang, Newton-conjugate-gradient methods for solitary wave computations, J. Comput. Phys. 228, 7007 (2009).DOI
  51. [53]L. N. Trefethen, Spectral Methods in MATLAB (SIAM, 2000).DOI
  52. [54]O. Guba and J. Lorenz, Continuous spectra and numerical eigenvalues, Mathematical and Computer Modelling 54, 2616 (2011).DOI
  53. [55]B. Sandstede and A. Scheel, On the stability of periodic travelling waves with large spatial period, Journal of Differential Equations 172, 134 (2001).DOI
  54. [56]B. Deconinck and J. N. Kutz, Computing spectra of linear operators using the Floquet-Fourier-Hill method, J. Comput. Phys. 219, 296 (2006).DOI
  55. [57]B. Deconinck, D. E. Pelinovsky, and J. D. Carter, Transverse instabilities of deep-water solitary waves, Proceedings of the Royal Society A: Mathematical, Physical and Engineering Sciences 462, 2039 (2006).DOI
  56. [58]Z. Ahmed, Real and complex discrete eigenvalues in an exactly solvable one-dimensional complex PT-invariant potential, Phys. Lett. A 282, 343 (2001).DOI
  57. [59]A. Chernyavsky, P. G. Kevrekidis, and D. E. Pelinovsky, Krein signature in Hamiltonian and PT-symmetric systems, in Parity-time Symmetry and Its Applications, Springer Tracts in Modern Physics, Vol. 280 (Springer, 2018) pp. 465–491.arxiv.org/abs/1711.02191
  58. [60]M. Klaus and J. K. Shaw, Purely imaginary eigenvalues of Zakharov-Shabat systems, Phys. Rev. E 65, 036607 (2002).pubmed.ncbi.nlm.nih.gov/11909285
  59. [61]I. S. Gradshteyn and I. M. Ryzhik, Table of Integrals, Series, and Products (1980).
  60. [62]J. Yang and S. Nixon, Stability of soliton families in nonlinear Schrödinger equations with non-parity-time-symmetric complex potentials, Phys. Lett. A 380, 3803 (2016).DOI
  61. [63]Y. Kominis, J. Cuevas-Maraver, P. G. Kevrekidis, D. J. Frantzeskakis, and A. Bountis, Continuous families of solitary waves in non-symmetric complex potentials: A Melnikov theory approach, Chaos, Solitons & Fractals 118, 222 (2019).DOI
  62. [64]S. Chandramouli, P. Sprenger, and M. A. Hoefer, Homoclinic and heteroclinic solutions of the nonlinear Schrödinger equation with a complex Wadati potential, arXiv preprint arXiv:2604.08794 (2026).arxiv.org/abs/2604.08794
  63. [65]Y. A. Kuznetsov, Elements of Applied Bifurcation Theory (Springer, 1998).

Paper details

vlnc:W_bexff3m5
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents