Integrable nonlinear oscillators with polynomial invariants: construction, Poincaré geometry, and an analytic stability boundary
Johannes Hagel · 2025
Read Integrable nonlinear oscillators with polynomial invariants: construction, Poincaré geometry, and an analytic stability boundary by Johannes Hagel (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2510-24080.txt.
Introduction
Johannes Hagel studies a family of nonlinear oscillators, systems whose motion is governed by an equation with a power term. He asks which cases in this family can be solved exactly, using invariants that are polynomial in position and speed. The usual energy does not serve here because the system changes with time.
Hagel shows that an invariant that is quadratic in speed exists exactly when a related function obeys a simple linear equation of third order. That condition gives a three parameter family of cases. For the squared power case, the result matches the known trigonometric form. The paper also maps the geometry behind the cases and finds a clear analytic border for stability.
Catalog record
| CREATOR | Johannes Hagel |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 13 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2510-24080
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Integrable nonlinear oscillators with polynomial invariants: construction, Poincaré geometry, and an analytic stability boundary Johannes Hagel Abstract Starting from the nonlinear ODE z ′′ + f ( t ) z + g ( t ) z m = 0 z^{\prime\prime}+f(t)\,z+g(t)\,z^{m}=0 with m > 1 m>1 , we show that after a suitable normal-form reduction of any Hill equation one may, without loss of generality, fix the linear part as f ( t ) ≡ ω 2 f(t)\equiv\omega^{2} (with ω > 0 \omega>0 constant). For the class z ′′ + ω 2 z + g ( t ) z m = 0 z^{\prime\prime}+\omega^{2}z+g(t)\,z^{m}=0 with m > 1 m>1 , our goal is to compile a catalogue of all possible integrable cases. We restrict attention to integrals that are polynomial in the variables z z and p = z ′ p=z^{\prime} . The Hamiltonian does not provide such an integral because it is explicitly time dependent. Instead, we search for invariants that are quadratic in p = z ′ p=z^{\prime} . We show that such invariants exist precisely when α 2 ( t ) := g ( t ) − 2 / ( m + 3 ) \alpha_{2}(t):=g(t)^{-2/(m+3)} satisfies the linear third-order ODE α 2 ′′′ + 4 ω 2 α 2 ′ = 0 \alpha_{2}^{\prime\prime\prime}+4\omega^{2}\alpha_{2}^{\prime}=0 . This yields the three-parameter solution g ( t ) = [ a 0 + a 1 cos ( 2 ω t ) + a 2 sin ( 2 ω t ) ] − ( m + 3 ) / 2 g(t)=[a_{0}+a_{1}\cos(2\omega t)+a_{2}\sin(2\omega t)]^{-(m+3)/2} . For m = 2 m=2 this reproduces the trigonometric structure with exponent − 5 / 2 -5/2 found in Hagel–Bouquet (1992). In addition we present a detailed stability analysis based on the invariant using Poincaré sections and find full agreement with numerical simulations. 1 Introduction The aim of this paper is to assemble a systematic catalogue of all integrable cases of the nonlinear equation z^{\prime\prime}+f(t)z+g(t)z^{m}=0,\quad m>1. We seek first integrals in polynomial form in the variables z , p=z^{\prime} , and t . The associated Hamiltonian is not suitable as a conserved integral, because it is explicitly time dependent. Instead, we search specifically for invariants that are quadratic in p=z^{\prime} , which then permit a classification of integrable cases. The nonlinear equation above can be analysed, whenever f(t) is periodic with period T , by applying the Courant–Snyder transformation [ 3 ] to the linear part z^{\prime\prime}+f(t)z=0 . In particular, y=\frac{z}{\sqrt{\beta(t)}},\qquad\Phi(t)=\int_{0}^{t}\frac{d\tau}{\beta(\tau)},\qquad\omega=\frac{\Phi(T)}{2\pi}, (where w(t)=\sqrt{\beta(t)} denotes the periodic envelope of the Hill equation z^{\prime\prime}+f(t)z=0 as long as z(t) remains bounded) maps the linear part to the harmonic oscillator \frac{d^{2}y}{d\Phi^{2}}+\omega^{2}y=0. The nonlinear contribution preserves its form under this change of variables, so that the full equation takes the normal form z^{\prime\prime}+\omega^{2}z+g(t)z^{m}=0. The structure of the paper is as follows. Section 2 defines the dynamical system in the normal form, introduces the quadratic ansatz for an invariant, and derives the conditions on g(t) under which integrability holds. Section 3 analyzes the invariant curve and establishes an exact analytical stability boundary, which is found to be in complete agreement with numerical simulations. Section 4 summarizes the results and outlines perspectives for further generalizations. 2 General formulation, integrability ansatz and example 2.1 Problem class and objectives We first consider the class of explicitly time-dependent nonlinear second-order differential equations z^{\prime\prime}+f(t)z+g(t)z^{m}=0,\quad m\in\mathbb{N},\,m\geq 2. (2.1) Although this equation can be derived from a Hamiltonian function H(z,p,t)=\frac{p^{2}}{2}+\frac{f(t)}{2}z^{2}+\frac{g(t)}{m+1}z^{m+1}, (2.2) it does not provide a conserved integral since \partial H/\partial t\neq 0 . Here f,g:\mathbb{R}\to\mathbb{R} are continuous and, without loss of generality, assumed 2\pi -periodic. Our aim is to identify within class (2.1) special integrable systems. We search for first integrals in quadratic polynomial form in p=z^{\prime} : I(z,p,t)=a_{0}(z,t)+a_{1}(z,t)p+a_{2}(z,t)p^{2},\qquad\frac{dI}{dt}=0. (2.3) As shown in Chapter 3 of [ 2 ] , condition (2.3) leads to a system of PDEs for the coefficients a_{0},a_{1},a_{2} of the form \displaystyle-a_{1}(z,t)\,[zf(t)+z^{m}g(t)]+\frac{\partial a_{0}}{\partial t} \displaystyle=0, (2.4) \displaystyle-2a_{2}(z,t)\,[zf(t)+z^{m}g(t)]+\frac{\partial a_{0}}{\partial z}+\frac{\partial a_{1}}{\partial t} \displaystyle=0, (2.5) \displaystyle\frac{\partial a_{1}}{\partial z}+\frac{\partial a_{2}}{\partial t} \displaystyle=0, (2.6) \displaystyle\frac{\partial a_{2}}{\partial z} \displaystyle=0. (2.7) The general solutions of (2.5)–(2.7) yield for a_{0},a_{1},a_{2} the following expressions: \displaystyle a_{2} \displaystyle=\alpha_{2}(t), (2.8) \displaystyle a_{1} \displaystyle=-\alpha_{2}^{\prime}(t)z+\alpha_{1}(t), (2.9) \displaystyle a_{0} \displaystyle=z^{2}\alpha_{2}(t)f(t)+\frac{2}{m+1}z^{m+1}\alpha_{2}(t)g(t)-z\alpha_{1}^{\prime}(t)+\tfrac{1}{2}z^{2}\alpha_{2}^{\prime\prime}(t)+\alpha_{0}(t). (2.10) Inserting these into (2.4) and comparing coefficients of z yields the following cases: (i) m=2 : \displaystyle\alpha_{1}^{\prime\prime}+f(t)\alpha_{1} \displaystyle=0, (2.11) \displaystyle\alpha_{2}^{\prime\prime\prime}+4f(t)\alpha_{2}^{\prime}+2\alpha_{2}f^{\prime}(t)-2\alpha_{1}g(t) \displaystyle=0, (2.12) \displaystyle 5g(t)\alpha_{2}^{\prime}+2\alpha_{2}g^{\prime}(t) \displaystyle=0. (2.13) (ii) m>2 : \displaystyle\alpha_{1}^{\prime\prime}+f(t)\alpha_{1} \displaystyle=0, (2.14) \displaystyle\alpha_{2}^{\prime\prime\prime}+4f(t)\alpha_{2}^{\prime}+2\alpha_{2}f^{\prime}(t)-2\alpha_{1}g(t) \displaystyle=0, (2.15) \displaystyle-\alpha_{1}g(t) \displaystyle=0, (2.16) \displaystyle(m+3)g(t)\alpha_{2}^{\prime}+2\alpha_{2}g^{\prime}(t) \displaystyle=0. (2.17) 2.2 The case f(t)=\omega^{2} From Chapter 1 it was established that in (2.1) we may set f(t)=\omega^{2} without loss of generality. Then (2.11)–(2.17) simplify to (i) m=2 : \displaystyle\alpha_{1}^{\prime\prime}+\omega^{2}\alpha_{1} \displaystyle=0, (2.18) \displaystyle\alpha_{2}^{\prime\prime\prime}+4\omega^{2}\alpha_{2}^{\prime}-2\alpha_{1}g(t) \displaystyle=0, (2.19) \displaystyle 5g(t)\alpha_{2}^{\prime}+2\alpha_{2}g^{\prime}(t) \displaystyle=0. (2.20) (ii) m>2 : \displaystyle\alpha_{1}^{\prime\prime}+\omega^{2}\alpha_{1} \displaystyle=0, (2.21) \displaystyle\alpha_{2}^{\prime\prime\prime}+4\omega^{2}\alpha_{2}^{\prime}-2\alpha_{1}g(t) \displaystyle=0, (2.22) \displaystyle-\alpha_{1}g(t) \displaystyle=0, (2.23) \displaystyle(m+3)g(t)\alpha_{2}^{\prime}+2\alpha_{2}g^{\prime}(t) \displaystyle=0. (2.24) For m=2 , (2.18) yields \alpha_{1}(t)=\tfrac{1}{2}C_{1}\cos(\omega t)+\tfrac{1}{2}C_{2}\sin(\omega t),\quad C_{1,2}\in\mathbb{R}. (2.25) Eq. (2.20) implies g(t)=\alpha_{2}(t)^{-5/2}. (2.26) Thus \alpha_{2}(t) satisfies \alpha_{2}^{\prime\prime\prime}+4\omega^{2}\alpha_{2}^{\prime}-[C_{1}\cos(\omega t)+C_{2}\sin(\omega t)]\alpha_{2}^{-5/2}=0. (2.27) Hence, in principle, there exists a five-parameter family of g(t) leading to integrable systems of the form z^{\prime\prime}+\omega^{2}z+g(t)z^{2}=0. (2.28) Since no closed form of (2.27) is known, we restrict to the case C_{1}=C_{2}=0 , which renders (2.27) linear with constant coefficients. Its general solution is three-parametric: \alpha_{2}(t)=A+B\cos(2\omega t)+C\sin(2\omega t). (2.29) Thus g(t)=[A+B\cos(2\omega t)+C\sin(2\omega t)]^{-5/2}. (2.30) For m>2 , equation (2.23) implies \alpha_{1}=0 (since g(t)\neq 0 ). Eq. (2.22) then yields \alpha_{2}(t)=A+B\cos(2\omega t)+C\sin(2\omega t). (2.31) From (2.24) we find after integration \alpha_{2}(t)=g(t)^{-\tfrac{2}{m+3}}. (2.32) Hence the integrable cases are given by g(t)=[A+B\cos(2\omega t)+C\sin(2\omega t)]^{-(m+3)/2}, (2.33) leading to integrable systems of the form z^{\prime\prime}+\omega^{2}z+g(t)z^{m}=0. (2.34) 2.3 Integral for m=2 and demonstration example For m=2 , a three-parameter form of \alpha_{2}(t) is given by \alpha_{2}(t)=A+B\cos(2\omega t)+C\sin(2\omega t),\qquad g(t)=\alpha_{2}(t)^{-5/2}. (2.35) Eq. (2.34) then admits the quadratic invariant \displaystyle I(z,p,t) \displaystyle=\alpha_{2}(t)p^{2}-\alpha_{2}^{\prime}(t)zp \displaystyle\quad+\Big(\omega^{2}\alpha_{2}(t)+\tfrac{1}{2}\alpha_{2}^{\prime\prime}(t)\Big)z^{2}+\tfrac{2}{3}\alpha_{2}(t)g(t)z^{3}. (2.36) A Poincaré section at times t_{k}=k\pi/\omega then yields an algebraic curve p=p(z) . As an example, choose A=1.3 , B=0.9 , C=0 , and \omega=1 . We integrate (2.34) with a Runge–Kutta method (step size h ), initial conditions z(0)=0.1 , z^{\prime}(0)=0 , over 0<t<600 . We then plot I(t)/I(0)-1 . Figure 1 shows the result in the range [-10^{-5},+10^{-5}] : the curve runs numerically flat at 0 , as expected, confirming integrability of (2.34) under assumption (2.35). Figure 1: Constancy of the invariant: plot of I(t)/I(0)-1 over 0<t<600 (parameters as in text). The range shown is [-10^{-5},+10^{-5}] . 2.4 Phase curve at the Poincaré section t_{k}=k\pi/\omega For C=0 , at t_{k}=k\pi/\omega we have \alpha_{2}(t_{k})=A+B , \alpha_{2}^{\prime}(t_{k})=0 , \alpha_{2}^{\prime\prime}(t_{k})=-4\omega^{2}B . With I_{0}=I(z(0),z^{\prime}(0),0) the phase curve becomes (A+B)p^{2}+\omega^{2}(A-B)z^{2}+\tfrac{2}{3}(A+B)^{-3/2}z^{3}=I_{0}. (2.37) This is quadratic in p , yielding p(z)=\pm\sqrt{\frac{I_{0}-\omega^{2}(A-B)z^{2}-\tfrac{2}{3}(A+B)^{-3/2}z^{3}}{A+B}}. (2.38) Figure 2: Poincaré phase section at t_{k}=k\pi/\omega : stroboscopic points (z(t_{k}),p(t_{k})) together with the analytic curve given by (2.38). 3 Stability analysis by means of the invariant curve We now investigate the stability boundary of equation (2.34) with specialization (2.35). For this purpose we consider the stroboscopic section at times t_{k}=k\pi/\omega , since at these points \dot{\alpha}_{2}(t_{k})=0 and the invariant assumes a particularly simple form. Set \alpha=A+R\cos(2\omega t-\varphi) with R=\sqrt{B^{2}+C^{2}} and suitable phase \varphi . The associated stroboscopic curve is then (A+R)p^{2}+\omega^{2}(A-R)z^{2}+\tfrac{2}{3}(A+R)^{-3/2}z^{3}=I_{0}, (3.1) where I_{0} is determined by the initial condition z(0)=z_{0},\,p(0)=0 : I_{0}=\omega^{2}(A-R)z_{0}^{2}+\tfrac{2}{3}(A+R)^{-3/2}z_{0}^{3}. (3.2) The shape of the admissible set p^{2}\geq 0 changes precisely when the discriminant of the cubic