1. Introduction
We analytically study the persistence of periodic oscillations for certain three-dimensional systems of ordinary differential equations (ODEs) with periodic perturbations and a slowly-varying variable. The considered ODEs are derived from a model of a finite lattice of particles coupled by linear springs under distributed harmonic excitation, which is described in detail in
Section 2. This model presents a low-energy nonlinear acoustic vacuum. We refer the reader for more motivations, further details and applications to [
1,
2]. Following the computations of [
1], we consider just two modes in
Section 3, and we postpone higher modes investigation to our future paper, since another approach will be used. Melnikov analysis is demonstrated in
Section 4 for finding conditions for the existence of periodic solutions for the perturbed ODEs corresponding to two modes. More precisely, following [
1], we derive a three-dimensional periodically-perturbed system of ODEs with a slowly-varying variable. Then, we analyze an unperturbed autonomous system of ODEs to compute its family of periodic solutions by revising the results of [
1] in more detail. Since we are interested in the persistence of periodic solutions for the perturbed ODEs, we compute the corresponding Melnikov functions by [
3]. Due to the difficulty of finding simple roots of these Melnikov functions explicitly, we outline an asymptotic approach for the location of some of them. Note that the simple roots of Melnikov functions predict the persistence and location of periodic solutions for perturbed ODEs. This is a novelty and a contribution of our paper.
Section 5 outlines possible future research along with summarizing our achievements in this paper.
2. The Model
We consider a lattice consisting of
N identical particles coupled by identical linear springs (they are un-stretched when the lattice is in the horizontal position) and executing in-plane oscillations (see
Figure 1). Fixed boundary conditions and dissipative terms are imposed, and the transverse harmonic forces are also applied. The equations of motion can be expressed as follows,
with
being the longitudinal and transversal displacements of the
i-th particle, respectively,
the angle between the
i-th spring and the horizontal direction,
the damping coefficient,
the deformation of the
i-th spring,
the exciting transverse force and
m the mass of each particle of the lattice. The tensile forces are proportional to the deformations of the springs, and considering the geometry of the deformed state of the lattice (see
Figure 1), one may write:
with
being the equilibrium length of the
i-th spring (each spring has the same length) and
k the linear stiffness coefficient of each coupling spring. Introducing
,
,
,
, where
and
are the normalized axial and transverse displacements, and the “slow” time scale
, Equation (
1) can be rewritten in normalized form:
where:
and:
The normalized system (
2) is referred to as the “exact lattice” in the following sections.
According to the previous research [
1], when we introduce this system (
2) without extra transverse force and damping terms, an interesting feature is that in the low energy limit and under the assumption that the axial displacements are assumed to be an order of magnitude smaller compared to the transverse ones, it was shown that, correct to the leading order of approximation, the transverse oscillations decouple from the axial ones and are governed by the following reduced system of equations for predominantly transverse oscillations of the particles:
Then, the nearly linear axial oscillations are driven by the transverse responses (see [
1,
2] for more details). Therefore, we focus our analysis like in [
1] just on Equation (
3), which presents
a low-energy nonlinear acoustic vacuum, because in the absence of linear terms, it possesses zero speed of sound in the context of classical linear acoustics. What is more, it is notable that the existence of the strongly nonlocal multiplier
indicates that the response of each particle is dependent (and hence, it is coupled) on the responses of all other particles. Equation (
3) admits
N exact nonlinear standing waves, or nonlinear normal modes (NNMs), in the form:
for the
p-th NNM,
, where
denotes the
p-th modal amplitude. These, by construction, are mutually orthogonal, and there are no other NNMs in this system, nor any NNM bifurcations [
1]. Substituting this NNM ansatz into Equation (
3) yields a set of
N uncoupled nonlinear equations governing the time-dependent amplitudes of the NNMs:
with:
which is the
p-th natural frequency of the corresponding linear system Equation (
3) and the prime denoting differentiation with respect to
. The exciting force, which is applied on each particle in the transverse direction, is expressed as:
where
, for the
p-th NNM,
, and this exciting force includes NNMs in the form
for the
p-th NNM,
and the
p-th natural linear frequency
.
The frequency of the p-th NNM is tunable with the force and energy, and it also paves the way for nonlinear resonances between NNMs widely separated in the nonlinear spectrum, given that their energies tune their frequencies to satisfy certain rational relationships.
Summarizing, we can write (
3) as:
where
,
M is a symmetric matrix given by
and
is the standard scalar product on
. The eigenvectors of
M are
with the corresponding eigenvalues
,
. Moreover, it holds (see [
4], p. 37):
The forced (
3) has the form:
Therefore, considering the basis
of
, we take
in (
4) to get:
Next, applying the coordinate transformation to (
5):
we get:
3. Two-Mode System
In this paper, we consider just two modes:
k and
p, so we study the system:
for
small and parameters
,
. Using the transformation:
and performing an averaging approach like in [
1], Equation (
6) is modified to the form:
Introducing
and
, we get:
where we consider
as a constant parameter. Now, by introducing the coordinate transformations
and
into Equation (
7), we get:
In the rest of the paper, we assume
and
for a natural number
k, so we study the periodically-perturbed system:
where we scaled
,
. We may suppose:
First, consider the unperturbed case where
, so the system:
By introducing the temporal variable
in Equation (
9), we get:
which is fully integrable and gives us the first integral
of the degenerate slow flow. If we consider the following initial conditions:
, where
and
, then we get
. To find exact solutions of (
10), we first derive:
for
small, since (
10) gives
,
(see
Figure 2) and
, so for
small, we have
.
By using the formula in [
4] ((2.599.4) p. 205), we obtain:
which gives:
recalling
. Of course, Formula (
11) holds for all
, not just for small positive ones. Next, using (
10) and (
11), we have:
which can be easily solved to arrive at:
Hence, the exact solution of the system (
10) is given as:
where the period is:
It is easy to verify the following symmetry property (see
Figure 3):
Summarizing, the exact solution of the unperturbed (
9) is the following:
with the period:
Consequently, for any:
taking:
Equation (
9) has the
-periodic solution:
4. Melnikov Analysis for Periodic Oscillations
Writing (
8) as:
and using [
5], (3.5.11), p. 111, with
, and [
6], Lemma 2.5, p. 283, we compute the Melnikov function:
as:
since differentiating by
the identity:
we get:
which is independent of
, and:
Formulas (
13) and (
14) are similar to [
3], (2.7). We are looking for a simple zero of
M, which is equivalent to considering:
To solve:
we first solve the scalar equation:
to get its root
and
. Then, we look for
and
with
such that:
Summarizing, we have the following result.
Theorem 1. If there are , satisfying (15), and with solving (16), then for any near and near with and small, there are near and near such that (8) with and has a -periodic solution near (12) with . Next, taking
in (
12), we obtain:
Then, (
15) gives as
,
with asymptotic solutions
satisfying either
or
. The asymptotic equation of (
16) is as follows:
which has solutions:
and
when
or
and
when
. However,
for
, so Theorem 1 cannot be applied directly. Note we consider just the first asymptotic equation of (
16), since the second one is a scalar multiple of the first one.
On the other hand, following the method of [
5], p. 111, we get the following result.
Corollary 1. For any sufficiently large and sufficiently small, there is , with and such that (8) has a T-periodic solution near (17) for small and either , , or , , . Proof. The bifurcation equation has the form (see [
5], p. 111):
i.e.,
for
and
. For
and
, (
18) takes the form:
The determinant of (
19) is
, so (
19) has the only solutions:
The determinants of Jacobians of (
19) at these zeros are as follows:
and:
respectively. Hence, the zeroes
and
are simple, so we can apply the implicit function theorem to get the result. The proof is complete. ☐
5. Discussion
Melnikov analysis is applied for the persistence of periodic oscillations for periodically-perturbed systems of ODEs with a slowly-varying variable. The ODEs are obtained from a model of a finite lattice of particles coupled by linear springs under distributed harmonic excitation, which presents a low-energy nonlinear acoustic vacuum (see [
1]), but we consider just two modes. We extend the study of [
1] to a problem with small exciting harmonic forces. We apply an analytical method based on derivation of Melnikov functions and then on the location of their simple roots. Melnikov functions are derived by using the approach of [
3,
5] developed for slowly-varying ODEs. Since the Melnikov functions are rather complicated, we follow an asymptotic way for solving the corresponding Melnikov equations. It would be nice to solve these Melnikov equations numerically for finding other simple roots, which is postponed to our next research. These roots determine and locate periodic solutions of the periodically-perturbed systems of ODEs derived from the two-mode low-energy nonlinear acoustic vacuum system. Moreover, our next investigation will be also to consider higher numbers of modes represented by a system of ODEs in (
5). The method used will be different from that in this paper, since it will be based on the results of Section 3.3 of [
5]. Note that higher modes of (
2) were numerically studied in [
2], which is another challenge for our study.