Download as pdf or txt
Download as pdf or txt
You are on page 1of 18

A dual principle for symmetric periodic solutions: Bistab

D. Núñeza,∗, L. Murciaa,1
a
Departamento de Ciencias Naturales y Matemáticas. Pontificia Universidad Javeriana Cali,
Facultad de Ingenierı́a y Ciencias, 760031, Cali, Colombia.

Abstract
In this work we introduce a new principle that provides a tool for searching odd
periodic responses of a general nonlinear oscillator with symmetries. This result
arises as the dual version of the variational principle first introduced in Ortega (2016),
because in this case the nonlinearity xD(t, x) verifies the reversed inequality, i.e.,
D(t, 0) < D(t, x) for all t ∈ R and x ̸= 0. Thus, our result reveals that under
certain conditions there exists a family of odd periodic responses with prescribed
nodal properties for the general oscillator. Indeed, contrary to that obtained in
Ortega (2016), the number of zeros of the solutions given by this dual principle is
at least bounded below. To illustrate the application of our principle we consider
a real example where the reversed inequality arises naturally: a noninterdigitated
Comb-drive MEMS modeled by a nonlinear version of the Mathieu equation. Then,
we prove the existence of a bi-stability operation regime for this example, since under
certain conditions the positive subharmonic of order 2 given by our principle and the
trivial solution are linearly stable. Our results are based on classical ODE tools and
the perturbation approach in Cen et al. (2020).
Keywords: MEMS, Comb-drive, Odd Periodic Solution, Ortega’s Principle, Nodal
properties


Funding: This work was supported by the Pontificia Universidad Javeriana Cali project PUJC-
CUT12

Corresponding Author
Email addresses: denunez@javerianacali.edu.co (D. Núñez),
larry061198@javerianacali.edu.co (L. Murcia)
1
Research assistant in the project PUJC-CUT12

Preprint submitted to Journal of Differential Equations August 3, 2023


1. Introduction
Let us consider the second order differential equation

ẍ + xD(t, x) = 0, (1)

where D ∈ C 0,1 (R/T Z × R) for some T > 0. Hereafter, we shall assume the following
hypotheses over the function D:

H1 D(−t, x) ≡ D(t, x) ≡ D(t, −x),

H2 D(t, 0) < D(t, x) for all t ∈ R and x ̸= 0.

The main motivation of this work comes from some beautiful ideas in the seminal
paper [1] (see Proposition 4 there) that solve a Dirichlet problem associated with (1)
by means of shootings from the origin with initial velocity v, counting the inner zeros
in a half period of time and detecting the minimal velocity such that this number
of zeros changes one unity. The hypotheses considered in that proposition, which
will hereafter be called Ortega’s principle, are simple: i) a bounded nonlinearity, ii)
D(t, x) < D(t, 0), if x is not zero. In our case the condition i) does not hold anymore
and the condition ii) is substituted by the reversed inequality H2. Contrary to that
reported in [1], the number of zeros of the solutions given by this dual principle is
bounded below and it possibly has no ceiling.

Recently, works like [2, 3, 4] have considered the ideas in [1] to introduce some
extensions and generalizations of the Ortega’s principle in different contexts. In
fact, the author in [2] obtained an extension of the variational principle to study
the existence of even periodic solutions for the generalized Sitnikov problem. Then,
the authors in [3] provided an extension of the principle for equations with finite
domain, and also studied the existence of odd periodic responses for a Comb-drive
finger MEMS model with a linear stiffness term and interdigitated fingers, and the
case with a cubic nonlinear stiffness was considered in [5]. Later, a generalization of
the Ortega’s principle was obtained in [4] by means of the truncation technique and
a priori bounds of period solutions. The general result in [4] was combined with the
perturbation technique in [6] to demonstrate the existence of bi-stability operation
regime for the Comb-drive finger MEMS model with cubic stiffness.

Thus, the main purpose of this work is to introduce a dual variational principle
for the existence of symmetric periodic solutions of (1) under the hypotheses H1-H2.
We remark that (1) encompasses nonlinear oscillators with symmetries, especially

2
those parametrically excited where H2 could naturally arise due to the parameters
setting, and which could have applications in multiple branches of physics and engi-
neering.

Perhaps, one of the most known types of parametrically excited ODEs is given
by the Mathieu equation and its nonlinear generalizations that have been considered
for different problems like the vibration of bodies (elliptical membranes, vibration
of columns, systems described by the Mathieu–Duffing equation), the angular oscil-
lations of a pendulum with mobile support point (see [7, 8], [9] and the references
therein), and the mechanical periodic responses of electrostatic microelectromechan-
ical systems (MEMS) [10, 11]. In fact, some of these general equations were studied
in literature (parametric resonance, existence of periodic responses and their stabil-
ity/instability analysis, among others aspects) by means of tools such as fixed points
theorems, normal forms, perturbation techniques and numerical strategies. Thus,
as a particular case of (1), here we also aim to study the existence of odd periodic
responses for a class of nonlinear Mathieu equations verifying H1-H2 (see Proposition
1 for equation (12)) through an elementary approach inspired by [1]. Additionally,
from a contextual point of view for (12), we shall consider a type of micro-scale de-
vice known as noninterdigitated Comb-drive finger MEMS which have applications
in sensing and signal filtering [12, 13, 11], to apply our result and prove the exis-
tence of odd periodic subharmonics with prescribed nodal properties (see section 3
for more details).

In section 2 we introduce some preliminaries, our main result and a general exam-
ple that shows its application. This example is motivated by the nonlinear version of
the Mathieu equation where a cubic nonlinear term is considered. Then, in section
3 we present a precise application of our existence result to microelectromechanical
systems and connect our approach with the stability properties in the linear sense.
Concretely, we prove that under certain conditions the odd positive subharmonic of
order 2 and the trivial solution are linearly stable. Finally, in section 4 we provide
the concluding remarks of this work and some open problems.-

2. The dual principle


Before we introduce the main result of this work, it is convenient to provide some
notations used throughout the document.

3
For m ∈ N we define the nonnegative integer number
0
νm := |{t ∈ ]0, mT /2[ | ϕ(t) = 0}|, (2)

where ϕ denotes the solution of the variational equation associated with (1) and
initial conditions (ϕ(0), ϕ̇(0)) = (0, 1). On the other hand, for each v ∈ R we denote
by x(t, v) the solution of equation (1) with initial conditions (x(0, v), ẋ(0, v)) = (0, v)
and thus define the following set

Σm := {v ∈ R | x(t, v) is defined on [0, mT /2]}. (3)

We observe that Σm is a nonempty open set because 0 ∈ Σm (trivial solution of


(1)) and the continuous dependence on initial conditions. Thus, we can define Σ0m
as the connected component of Σm containing v = 0. Then, Σ0m ∩ R+ = [0, v 0 [ for
some v 0 ∈ R = R+ ∪ {∞} (here the + superscript notation refers to the subset of
nonnegative elements). From the previous discussion, we are now able to introduce
the function νm : Σ0m ∩ R+ → Z+ such that

νm (v) := |{t ∈ ]0, mT /2[ | x(t, v) = 0}|. (4)

It follows from the definition above that for v ∈ Σ0m ∩ R+ , the number νm (v) < ∞
denotes the number of zeros (also called nodal properties) of solution x(t, v) on
the interval ]0, mT /2[. Indeed, those zeros are non-degenerate and therefore isolated.
Finally, for each m ∈ N let

ν̂m = ν̂m (m) := sup {νm (v) | v ∈ Σ0m ∩ R+ } ∈ R.

Henceforth, if νm is unbounded above, then we shall write ν̂m = ∞.


Remark 1. The case ν̂m < ∞ could happen. For instance, consider the following
oscillator
a(t, δ)x3
ẍ + x + = 0,
1 + x2
where a(t, δ) := 1 + δ cos (ωt) is a positive, T -periodic and continuous function for
2
ω > 0 and δ ∈ ]0, 1[. Notice that here D(t, x) = 1 + a(t.δ)x1+x2
verifies all conditions
of symmetry and comparison with x = 0. Consider now an initial velocity v ∈]0, v 0 [.
Since 1 ≤ D ≤ 2 + δ on its domain, then by the Sturm Comparison Theory and
considering the equation ü + Av (t)u = 0, Av (t) := D(t,√x(t, v)), we conclude that
number of zeros of x(t, v) on ]0, mT /2[ is not greater than 2+δmT

, because u = x(t.v)
satisfies the equation ü + Av (t)u = 0 and Av ≤ 2 + δ. This shows that ν̂m is finite
for this example.

4
Theorem 1. Let m > 0 and N ≥ 0 be two integer numbers. Consider the equation
(1) with D(t, x) verifying H1-H2 and assume that
0
νm + 2 ≤ ν̂m < ∞. (5)

Then there exists at least one odd mT -periodic solution of (1), xm,N (t), with exactly
N zeros on ]0, mT /2[ if
0
νm + 1 ≤ N ≤ ν̂m − 1. (6)
0
If ν̂m = ∞ the conclusion holds for each N ≥ νm + 1. Moreover, if the variational
equation associated with (1) at x ≡ 0 has not mth -unity roots as Floquet multipliers,
then the before conditions can be substituted by
0 0
νm + 1 ≤ ν̂m and νm ≤ N ≤ ν̂m − 1.

Proof. The proof shall be made in several steps and it relies on an existence result
of solutions with certain nodal properties of a null Dirichlet problem associated with
(1) on the interval [0, mT /2] (see Step 5 below). We recall the equivalency, under the
symmetries in H1, between the problem of finding odd mT -periodic solutions of (1)
and finding nontrivial solutions to the associated Dirichlet problem.
• Step 1. The number of zeros when v → 0+ .
Notice that fv (t) := x(t,v)
v
defined on [0, mT /2] and for v ∈ Σ0m ∩ R+ \ {0} is a
continuously differentiable function verifying that in the C 1 -topology fv (t) →
f (t) := ∂x
∂v
(t, 0) ≡ ϕ(t) uniformly on [0, mT /2] as v → 0+ . Since f (t) has non-
degenerate zeros and fv (0) = 0 for all v ∈ Σ0m ∩ R+ \ {0}, we conclude from
the C 1 -convergence lemma (see Lemma 1 in [1]) that there exists v ∗ > 0 such
that νm 0
≤ νm (v) ≤ νm 0
+ 1 for all v ∈ ]0, v ∗ [. Moreover, if ϕ (mT /2) ̸= 0,- then
the C 1 -convergence lemma implies that νm (v) = νm 0
for all v ∈ ]0, v ∗ [. Thus,
we have proved the following result.
Lemma 1. There exists v ∗ > 0 such that νm (v) = νm
0 0
or νm (v) = νm + 1 for
all v ∈ ]0, v ∗ [.

On the other hand, the variational equation associated with (1) at x ≡ 0 is


given by
ü + a0 (t)u = 0, (7)
where a0 (t) := D(t, 0) is a continuous, even and mT -periodic function (with
minimal period T ) as a consequence of the hypotheses. Therefore, the solu-
tion ϕ is an odd function (see Theorem 1.1 in [14]). This last assertion and

5
the equivalence previously mentioned show that ϕ(mT /2) = 0 if and only if ϕ
is a mT -periodic solution of (7)-. Because the spectrum of the correspond-
ing monodromy matrix Φ verifies sp(Φ(mT )) = sp(Φ(T )m ), we conclude that
ϕ(mT /2) ̸= 0 if and only if λm ̸= 1 for all Floquet multiplier λ of equation (7).
Hence, we obtain the following lemma.
Lemma 2. If λm ̸= 1 for all Floquet multiplier λ of (7), then νm (v) = νm
0
for
∗ ∗
all v ∈ ]0, v [ and v as in Lemma 1.

• Step 2. A lower bound for νm (v).


Let v ∈ ]0, v 0 [ for v 0 defined as in page 4, thus x(t, v) can be thought as a
solution of the linear equation
ü + av (t)u = 0, (8)
where av (t) := D(t, x(t, v)). Because a0 (t) ≪ av (t) and ϕ(0) = 0, we straight-
forwardly obtain from the Sturm Comparison theory applied to the equations
0
(7) and (8), that νm (v) ≥ νm .
• Step 3. The variational property.
0
Let N ∈ Z≥0 be such that νm + 1 ≤ N ≤ ν̂m − 1 and consider the following
nonempty set Ω̂m,N defined by
Ω̂m,N := {v ∈ 0, v 0 | νm (v) > N }.
 
(9)

We remark that Ω̂m,N ̸= ∅ as a consequence of the hypothesis and the fact that
{Ω̂m,N }N - is a decreasing sequence of sets. More precisely, from the definition
it is easy to check that Ω̂m,N2 ⊆ Ω̂m,N1 for all N1 , N2 ∈ Z≥0 - such that N2 >
N1 ≥ N -. Now, if ν̂m = ∞, then for N0 ∈ Z≥0 such that N0 > N there exists
v ∈ ]0, v 0 [ satisfying that νm (v) = N0 + 1 so v ∈ Ω̂m,N0 ⊆ Ω̂m,N . On the other
hand, if ν̂m = max νm < ∞ then there exists v ∈ ]0, v 0 [ such that νm (v) = ν̂m ,
thus v ∈ Ω̂m,r ⊆ Ω̂m,N , r := ν̂m − 1 ≥ N .
Additionally, by defining ω̂m,N := inf Ω̂m,N we conclude that ω̂m,N2 ≥ ω̂m,N1 for
N1 , N2 ∈ Z≥0 such that N2 > N1 ≥ N , and that ω̂m,N > 0. Indeed, the order
assertion follows straightforwardly from the decreasing sequence property, and
if we assume that ω̂m,N = 0, then there exists a decreasing sequence vn such
that vn ∈ Ω̂m,N for all n ∈ N and vn → 0 as n → ∞. This implies that for
0 0
n ∈ N large enough we obtain νm (vn ) = νm or νm (vn ) = νm + 1 (see Step 1),
nevertheless, this is a contradiction because by definition and the hypothesis
0
we know that νm + 2 ≤ N + 1 ≤ νm (vn )-.

6
• Step 4. The solution x(t, ω̂m,N ) has exactly N zeros in a half period.
For δ > 0 small enough consider v and w verifying that 0 < v − ω̂m,N < δ and
0 < ω̂m,N − w < δ, with w ∈
/ Ω̂m,N and v ∈ ω̂m,N . Then

νm (w) ≤ N and νm (v) ≥ N + 1. (10)

Moreover, from the C 1 -convergence lemma we have that

νm (ω̂m,N ) ≤ νm (w),

νm (v) ≤ νm (ω̂m,N ) + 1,
thus the first inequality and (10) imply that νm (ω̂m,N ) ≤ N , and from the
second inequality and (10) it follows that N + 1 ≤ νm (ω̂m,N ) + 1 so that
νm (ω̂m,N ) ≥ N . Therefore, νm (ω̂m,N ) = N . The lemma below summarizes the
previous discussion.

Lemma 3. Let N be as in the Theorem 1. Then ω̂m,N ∈ Ω̂m,N −1 and x(t, ω̂m,N )
has exactly N zeros on the interval ]0, mT /2[.

• Step 5. The solution x(t, ω̂m,N ) solves the Dirichlet problem.


To see that x (mT /2, ω̂m,N ) = 0 = x(0, ω̂m,N ). Suppose by contradiction that
x (mT /2, ω̂m,N ) ̸= 0, then the C 1 -convergence lemma implies that the number of
zeros must be locally constant, i.e., there exists δ > 0 : |ω̂m,N − v| < δ implies
νm (v) = νm (ω̂m,N ). This contradicts the jump proved above (see (10)).
0
To sum up, we have showed that for νm + 1 ≤ N ≤ ν̂m − 1 there exists an
odd mT -periodic solution of (1), xm,N (t) := x(t, ω̂m,N ), with exactly N zeros
in ]0, mT /2[.

The following example illustrates an application of Theorem 1.

Example (ν̂m = ∞). Next, we shall apply our main result to a family of periodically
forced oscillators modeled by a nonlinear version of the Mathieu equation. First, let
us consider the autonomous oscillator

ẍ + a0 x + c0 x3 = 0, (11)

7
where a0 and c0 are positive constants, and for Ω > 0 let us consider the correspond-
ing T = 2π

-periodic perturbation

ẍ + aδ (t)x + cδ (t)x3 = 0, (12)

where aδ and cδ are positive T -periodic functions given by aδ (t) := a0 + δp1 (t) and
cδ (t) := c0 + δp2 (t), for pi (i = 1, 2) nonnegative, even and smooth T -periodic func-
tions, and δ ≥ 0. We remark that the conclusion in this example can be also obtained
for nonconstant sign periodic functions pi (i = 1, 2) whenever proper adjustments to
the parameter δ and the subsequent ideas are considered.

We recognize here the structure of equation (1) with D(t, x) = aδ (t) + cδ (t)x2
verifying H1-H2. In the following we shall prove some facts regarding the nonlinear
oscillators above.

1. The equation (12) has its solutions defined for all t ∈ R:


In fact, for any trajectory (x, y) = (x(t), ẋ(t)) of (12) let us define the nonneg-
ative test function over the phase space xy:

y2 x2 x4
V (t) = V (t, x, y) := + aδ (t) + cδ (t) .
2 2 4

A straightforward computation shows that

2 x2 x4 x2 x4
V̇ (t) = y ẏ + xẋ(aδ (t) + cδ (t)x ) + ȧδ (t) + ċδ (t) = ȧδ (t) + ċδ (t) ,
2 4 2 4
δ max {∥ṗ ∥ ,∥ṗ ∥ }
1 ∞ 2 ∞
thus for any constant A ≥ min {a0 ,c0 }
we obtain that V̇ (t) ≤ AV (t).
This growth condition implies by the Grönwall’s inequality that for finite time
the test function cannot have an unbounded behavior, so the trajectory (x, y)
has no blow-up, i.e., v 0 = ∞.

2. The period function corresponding to the annulus period at x ≡ 0 for equation


(11) is monotone decreasing to zero:
It is not difficult to prove that (11) defines a nonlinear center at the origin in
the phase plane xy with a period T (v) (of the solution x(t, v), v > 0) that is
decreasing for all v > 0 andq verifies that limv→∞ T (v) = 0. Indeed, the simple
a0
change of variables x = c0
x̂ and t = √1a0 t̂ reveals that (11) is equivalent,
by using the same variable notations, to the second order differential equation

8
ẍ+x+x3 = 0. Thus, if h denotes the energy level√associated with the nontrivial
solution x(t, v) of this last oscillator, then v = 2h and from the Theorem A
(a.i) in [15] (see page 239) it follows that the corresponding period function
4
T (v) := T (h(v))- is monotone decreasing because n = 3 and Hn+1 (x, ẋ) = x4 ,
or equivalently T ′ (h) := dT
dh
(h) < 0 for all h > 0-. Moreover, from the Theorem
C (ii) in [15] (see page 240) we know that the origin is a global center, therefore
the characterization at h = ∞ given by Theorem C in [16] (see page 184)
implies that limv→∞ T (v) = 0 because the exponents are a = − 43 and b = − 12 .
In conclusion, for some C > 0 we have limv→∞ T (h(v)) = limh→∞ Cha+b+1 = 0.
Since the period annulus of the center at the origin in this case is proportional
to that of equation (11), the monotonic and asymptotic properties of the period
function are preserved under the change of variables. Finally, it is not difficult
to check that limv→0+ T (h(v)) = T0 := √2πa0 , thus T0 is an upper bound of T (h).

3. The conclusion is ν̂m = ∞:-


j√ k
a0 mT
It is easy to verify that for any integer N ≥ 2π
there exists vN > 0 such
that the nontrivial solution x(t, vN ) of (11) has exactly N zeros on ]0, mT /2[.
In fact, it is sufficient to choose vN such that T (vN ) = NmT +1
. On the other
hand, let us denote the flow for (12) by x(t, v, δ), i.e., the unique solution with
initial conditions x(0, v, δ) = 0 and ẋ(0, v, δ) = v. Thus, from the continuous
dependence on parameters we deduce that for small δ > 0 the solutions x(t, vN )
and x(t, vN , δ) are C 1 -close on [0, mT /2]. Therefore, the C 1 -convergence lemma
implies that x(t.vN , δ) has N or N + 1 zeros on ]0, mT /2[. So ν̂m = ∞ and we
can directly apply the Theorem 1 to equation (12).-
The following proposition summarizes the discussion of this example.
Proposition 1. Let m ∈ N. Then there exists at least one odd mT -periodic solution
of (12), mT
 √ xm,N (t), that  has exactly N zeros on the interval ]0, /2[ for each integer
a0 +δ∥p1 ∥∞ mT √ π
N ≥ 2π
+ 1 and whenever δ is small enough. Moreover, if a0 < mT
and δ is small enough then the conclusion is true for each integer N ≥ 0.
√ 
0 a0 +δ∥p1 ∥∞ mT
We highlight that in this case νm ≤ 2π
as a consequence of the
Sturm Comparison Theory, then we can apply the Theorem 1 with the hypothe-
sis ν̂m = ∞ to obtain the existence assertion. Furthermore, the second part of
Proposition 1 is a direct consequence of the Sturm Comparison Theory and the
Lyapunov-Zukovskii criterion. More precisely, under the second hypothesis of the

9
2
π −a0 (mT )2
proposition and for δ ≤ δ0 where δ0 := ∥p 1 ∥∞ (mT )
2 , we have that 0 < a0 ≤ D(t, 0) ≡

π 2

aδ (t) ≤ a0 + δ ∥p1 ∥∞ ≤ mT for all t ∈ R. Since this last condition implies that
0
the variational equation at x ≡ 0 is elliptic and that νm = 0, an application of the
last part of Theorem 1 with ν̂m = ∞ leads us to the result.-

3. Applications: Comb-drive finger MEMS with tuning


3.1. On the existence of odd periodic subharmonics via the dual principle-
In this section we shall prove that our main result (Theorem 1) provides an ele-
mentary and useful tool to find odd periodic solutions with prescribed nodal proper-
ties for a class of micro-scale systems known as Comb-drive finger MEMS. We notice
that these devices have noninterdigitated fingers and that a significant amount of
authors in literature have tackled the study of their dynamical behavior under para-
metric excitation, the existence of chaos, and the effects of nonlinearities on the
behavior of parametric resonance [17, 12, 13, 18, 11, 19, 20]. Moreover, some of them
could be employed as mass sensors and frequency filters by means of the parametric
resonance, and that the most common approaches for a mathematical analysis of the
dynamics are perturbation methods and numerical tools.-

Regarding the general configuration of the MEMS oscillators considered here, we


notice that they are composed of a movable backbone (mass shuttle) that is attached
to various flexures (fixed beams attached to the substrate that act like springs), and
a pair of noninterdigitated comb-drives that actuate the device whenever a voltage
signal is applied. Thus, the oscillator’s motion is described by the displacement of
the backbone (constrained to lie along the device axis), and the forces acting on it are
the elastic restoring force produced by the flexures, the electrostatic force from the
noninterdigitated comb-drives and the dissipative effects (which are generally small
[11]) primarily from the aerodynamic effects.

It is worth noting that the restoring force here is described by a cubic function
of the displacement, which introduces linear and cubic nonlinear mechanical stiff-
ness coefficients. Furthermore, because some of these parametrically excited devices
could even have two sets of independent comb-drives (one set utilizes a DC input
voltage for tuning purposes and the other set employs an AC voltage input), the
combined electrostatic driving force is also accurately described by a cubic function
of displacement. Therefore, we obtain a pair of linear and cubic nonlinear electro-
static stiffness, each associated with the corresponding set of comb-drives.- Indeed,

10
as mentioned in [20], the stiffness coefficients can be either negative or positive de-
pending on the comb-drive fingers alignment and the geometry, this is an interesting
and unique feature that produces different functioning configurations.

Next, we shall introduce the normalized nonlinear version of the Mathieu equa-
tion that models the mechanical response of the Comb-drive MEMS studied in this
section. Hereafter we shall neglect the damping effects and assume that the device
is operated in a vacuum. Let x denote the normalized displacement of the mass in
the normalized time t, thus the undamped model is given by [11, 20]-

ẍ + (β + γ + γ cos (Ωt))x + (ζ + η + η cos (Ωt))x3 = 0, (13)


where
r10 V02 x20 (k3 + r30 V02 ) r1A VA2 x20 r3A VA2
β =1+ , ζ= , γ= , η= ,
k1 k1 k1 k1
and x0 is the length of the shuttle mass, k1 and k3 are, respectively, the linear and
cubic nonlinear mechanical stiffness coefficients, r10 and r30 are, respectively, the lin-
ear and cubic nonlinear electrostatic coefficients due to the DC excited fingers when
a DC input voltage of amplitude V0 is supplied, r1A and r3A denote, respectively,
the linear and cubic nonlinear electrostatic coefficients due to the AC excited fingers
when a cosinep square-rooted AC input voltage of amplitude VA and frequency Ω
(V (t) := VAC 1 + cos (ωt)) is supplied.


We observe that (13) has the structure of general equation (1) with T = Ω
and
D(t, x) = aδ (t) + cδ (t)x2 ,
2
where for δ = VAC , and the proper redefinition of the parameters γ and η, we consider
aδ (t) := β +δγ(1+cos (Ωt)) and cδ (t) := ζ +δη(1+cos (Ωt)). Moreover, the condition
H1 holds in this case and if we assume that β, γ, ζ and η are positive then H2 is
easily verified. Hence,
j √ an application
k of Proposition 1 implies that for m ∈ N and
β+2δγmT
each integer N ≥ 2π
+ 1- there exists at least one mT -periodic solution
of (13) that has exactly N - zeros on the interval ]0, mT /2[ whenever
√ theπ amplitude
VAC of the AC input voltage is small enough. Additionally, if β < mT then the
conclusion is true for each integer N ≥ 0- and VAC small enough.

3.2. Bi-stability in noninterdigitated Comb-drive finger MEMS


In this section we shall prove that the odd positive subharmonic of order 2 ob-
tained in the previous section, x2,0 (t), is linearly stable whenever the amplitude VAC

11
of the AC input voltage is small enough. This result is a consequence of the perturba-
tion approach in [6] when the hypotheses of the previous section are satisfied.
√ More π
precisely, we shall assume that m > 0, N ≥ 0 are integer numbers and β < mT .
We recall that under these assumptions, the fact 2 in the Example (see page 8) re-
veals that for δ = 0 the origin is a global center and that the period function of the
corresponding annulus period verifies that for all h > 0

T ′ (h) < 0. (14)

For simplicity let us define F : R × R × R → R such that F (t, x, δ) = aδ (t)x + cδ (t)x3


as in the previous section. Thus, F (−t, x, δ) ≡ F (t, x, δ) ≡ −F (t, −x, δ), F (t, x, 0) ≡
f (x) := βx + ζx3 and (13) can be written as

ẍ + F (t, x, δ) = 0, (15)

with the following autonomous equation

ẍ + f (x) = 0. (16)

Now, for p = N + 1 let hm,p denote the level of energy such that

mT
T (hm,p ) = , (17)
p

which implies that there exists an odd mT -periodic solution of (16), φm,N (t) :=
x(t, vN ) (see the fact 3 in the Example, page 9), with exactly N zeros on the inter-
val ]0, mT /2[. We observe that φm,N coincides with the odd (m, p)-periodic solution
ϕm,p (t) defined in [6] for the corresponding autonomous equation, i.e., φm,N (t) ≡
ϕm,p (t) where ϕm,p (t) is an odd mT -periodic solution of the autonomous equation
with exactly 2p zeros on [0, mT [. Additionally, by following the arguments in the
discussion of the section 2.3 in [6] it is not difficult to conclude that (14) implies that
the solution φm,N is both parabolic unstable and Lyapunov unstable.

Furthermore, since (14) holds and ϕm,p (t) is a nontrivial solution of the null
Dirichlet problem associated with equation (16), we obtain as a direct consequence
of Theorem 3.1 in [6] that there exist δm,p > 0 and a unique smooth function ∆m,p (δ)
defined on [0, δm,p [ such that ∆m,p (0) = vN and ϕm,p (t, ∆m,p (δ), δ) for δ ∈ [0, δm,p [
is a nontrivial solution of the corresponding null Dirichlet problem associated with
(15) that emanates from ϕm,p (t). Moreover, because of the uniqueness of the branch
given above by the Implicit Function Theorem, we have that for δ > 0 small enough

12
ϕm,p (t, ∆m,p (δ), δ) ≡ xm,N (t), where xm,N (t) ≡ x(t, ω̂m,N , δ) is the solution of (15)
given by Proposition 1 in the former section. This reveals a direct connection be-
tween the odd periodic oscillations provided by dual principle and those obtained by
means of the corresponding perturbation of the nonlinear center at the origin.

We notice that the variational equation for(15) along the solution xm,N (t) is the
Hill’s equation
∂F (t, x, δ)
ü + q(t, δ)u = 0, q(t, δ) := , (18)
∂x (t,xm,N (t),δ)

and that for δ ≥ 0 small enough the corresponding trace τm,p (δ) is defined by

τm,p (δ) := u1 (mT, δ) + u̇2 (mT, δ),

where u1 (t, δ) and u2 (t, δ) denote the normalized solutions of the equation in (18).
Because τm,p (0) = 2, now we aim to determine the sign of the derivative of τm,p (δ)
at δ = 0.

A straightforward computation shows that


∂ 2F
≡ −Ω(γφm,N (t) + ηφ3m,N (t)) sin (Ωt), (19)
∂t∂δ (t,φm,N (t),0)

Thus, from the Theorem 3.4 in [6] (see formula (3.15) there) we obtain that
′ ′
τm,N +1 (0) = −(N + 1)T (hm,N +1 )Im,N , (20)

where
mT
∂ 2F
Z
Im,N := φ̇m,N (t)dt,
0 ∂t∂δ (t,φm,N (t),0)

and from a simple integration by parts of each term in the integrand, it is revealed
that Z mT
2
Im,N = Ω Hm,N (t) cos (Ωt)dt,
0
where
γ 2 η
Hm,N (t) := φm,N (t) + φ4m,N (t).
2 4

Therefore, sng(τm,N +1 (0)) = sng(Im,N ). Additionally, we know from the discussion
in [6] that the right-hand side of (20) is zero whenever m ̸= 2n(N + 1) for n ∈ N

13
and that φ2n(N +1),N (t) ≡ φ2n,0 (t). As a consequence of the aforementioned remarks,
here we shall only study the case m = 2, i.e., m = 2n(N + 1) with n = 1 and N = 0.

Next, we provide some basic identities and inequalities that are helpful through-
out this section. From the definition it follows that T (h2,1 ) = mT = 4π

, and because
T (h2,1 )
φ2,0 (t) is odd and 2 -antiperiodic we obtain that

φ2,0 (t) ≡ −φ2,0 (t − 2π/Ω) ≡ φ2,0 (2π/Ω − t). (21)

Moreover, we know that φ2,0 (t) > 0 and φ̇2,0 (t) > 0 for all t ∈ ]0, 2π/Ω[. Therefore, we
deduce that H2,0 (t) is an even and periodic function with minimal period 2π Ω
which
verifies
H2,0 (t) ≡ H2,0 (2π/Ω − t), (22)
and that for all t ∈ ]0, π/Ω[
Ḣ(t) > 0. (23)
Then, from the periodicity of H2,0 (t) and cos (Ωt) we obtain that
Z 2π

2
I2,0 = 2Ω H2,0 (t) cos (Ωt)dt
0
Z π Z 2π
!
Ω Ω
= 2Ω2 H2,0 (t) cos (Ωt)dt + H2,0 (t) cos (Ωt)dt .
π
0 Ω


It is not difficult to check that by using the change of variable u = Ω
− t and the
symmetry in (22) we obtain

Z

Z 0
H2,0 (t) cos (Ωt)dt = H2,0 (2π/Ω − u) cos (2π − Ωu)(−du)
π π
Ω Ω
Z π

= H2,0 (u) cos (Ωu)du,
0

so that
Z π

2
I2,0 = 4Ω H2,0 (t) cos (Ωt)dt
0
Z π Z π
! (24)
2Ω Ω
2
= 4Ω H2,0 (t) cos (Ωt)dt + H2,0 (t) cos (Ωt)dt .
π
0 2Ω

14
π
Additionally, by using the change of variable u = Ω
− t it follows that
π
Z

Z 0
H2,0 (t) cos (Ωt)dt = H2,0 (π/Ω − u) cos (π − Ωu)(−du)
π π
2Ω 2Ω
Z π
(25)
2Ω
=− H2,0 (π/Ω − u) cos (Ωu)du.
0

Hence Z π
2Ω
2
I2,0 = 4Ω (H2,0 (t) − H2,0 (π/Ω − t)) cos (Ωt)dt. (26)
0

Because Ωπ − t > t and cos (Ωt) > 0 for all√t ∈ ]0, π/2Ω[, we conclude from (23) and
′ π
(26) that τ2,1 (0) < 0. This implies that if β < 2T then the solution x2,0 (t) of (15)
that emanates from the solution φ2,0 (t) of (16) for δ > 0 small enough is linearly
stable. Finally, since this last condition also implies that the variational equation
along x ≡ 0 is elliptic we conclude that there exists a bi-stability operation regime
for a noninterdigitated Comb-drive MEMS modeled by (13) whenever a small VAC
load is supplied. The following theorem summarizes the discussions above.-
√ π
Theorem 2. Let m ∈ N and assume that β < mT . Then (13) has at least one
mT -periodic solution, xm,N (t), with exactly N zeros on ]0, mT /2[ for each N ∈ Z≥0
and VAC small enough. Moreover, the trivial solution x ≡ 0 is linearly stable, and
if m = 2 then the positive subharmonic of order 2, x2,0 (t), is also linearly stable for
VAC small enough.

4. Concluding remarks
The main result of this work is a novel and elementary variational principle that
provides sufficient conditions that guarantee the existence of odd periodic solutions
with prescribed nodal for a general class of nonlinear oscillators with symmetries.
This dual approach was inspired by the Ortega’s principle introduced in [1], and in
contrast to most of the works that study symmetric periodic solutions and which are
based on sophisticated mathematical tools like Degree theory, our result is elemen-
tary because only basic facts of the ODE Theory are employed.

On the other hand, we observe that though- Theorem 1 has a global character,
practical conditions on the nonlinearity to control the constant ν̂m could be diffi-
cult to attain as it’s illustrated by Proposition 1 where the assumption of small δ is
required. Hence, the crucial question here is: How can we obtain a global version

15
of Proposition 1? More precisely, this defect is a consequence of the difficult task
of estimating νm (v) = νm (v, δ) for large v-, specifically we need νm (v, δ) to be- ar-
bitrary large with v, and uniformly in δ. The argument in the example (see page
9) was to perturb a nonlinear global center of period function decreasing to zero
when v → ∞, with a small parameter. Perhaps, the strategy for figuring this out
consists in searching conditions to obtain x(mT /2, v, δ) ̸= 0 for large v and for all δ in
a compact interval, and also to get x(τ, v, δ) > 0 for some τ located before the zeros
of x(t, v, 0) in a half period of time.- Thus, by an argument of homotopy we have
that- the number of inner zeros is preserved for all δ > 0 (see Lemma 7.2 in [21]).
This seems to be a nontrivial and difficult matter to solve.

Another interesting remark is that the statement in Theorem 1 could be expected


to be an if and only if characterization, similar to that given by the Ortega’s principle
in [1]; nevertheless, we cannot obtain a result with sufficient and necessary conditions
here because the conclusion of Lemma 1 has two possibilities. Finally, we notice
that by following the ideas in [2], it could be possible to get a version of the dual
principle for the existence of even periodic solutions with prescribed nodal properties
in nonlinear oscillators like (1).

Acknowledgments
Authors are grateful to R. Ortega for his suggestion of considering the reversed
inequality H2 in some problems. This article has been financially supported by the
grant “Por una Universidad Transformadora (2023-2024)” project CUT12 020100824,
Pontifical Xavierian University.

References
[1] R. Ortega, Symmetric periodic solutions in the Sitnikov problem., Arch. Math.
107 (2016) 405–412. doi:10.1007/s00013-016-0931-1.

[2] M. Misquero, Resonance tongues in the linear sitnikov equation, Celestial


Mechanics and Dynamical Astronomy 130 (2018). URL: https://doi.org/10.
1007%2Fs10569-018-9825-9. doi:10.1007/s10569-018-9825-9.

[3] D. Núñez, O. Larreal, L. Murcia, Odd periodic oscillations in comb-drive fin-


ger actuators, Nonlinear Analysis: Real World Applications 61 (2021) 103347.
URL: https://doi.org/10.1016%2Fj.nonrwa.2021.103347. doi:10.1016/j.
nonrwa.2021.103347.

16
[4] D. Núñez, L. Murcia, On a bi-stability regime and the existence of odd subhar-
monics in a comb-drive MEMS model with cubic stiffness, Nonlinear Analysis:
Real World Applications 74 (2023) 103938. URL: https://doi.org/10.1016%
2Fj.nonrwa.2023.103938. doi:10.1016/j.nonrwa.2023.103938.
[5] O. Larreal, L. Murcia, D. Núñez, Odd periodic oscillations in comb-drive fin-
ger mems with cubic stiffness, Journal of Mathematical Control Science and
Applications 8 (2022) 185–197.
[6] X. Cen, X. Cheng, Z. Huang, M. Zhang, On the stability of symmetric periodic
orbits of the elliptic sitnikov problem, SIAM Journal on Applied Dynamical Sys-
tems 19 (2020) 1271–1290. URL: https://doi.org/10.1137%2F19m1258384.
doi:10.1137/19m1258384.
[7] M. Azimi, Stability and bifurcation of mathieu–duffing equation, Interna-
tional Journal of Non-Linear Mechanics 144 (2022) 104049. URL: https://doi.
org/10.1016%2Fj.ijnonlinmec.2022.104049. doi:10.1016/j.ijnonlinmec.
2022.104049.
[8] I. Kovacic, R. Rand, S. Mohamed Sah, Mathieu’s equation and its general-
izations: Overview of stability charts and their features, Applied Mechan-
ics Reviews 70 (2018) 020802. URL: https://doi.org/10.1115/1.4039144.
doi:10.1115/1.4039144.
[9] D. Younesian, E. Esmailzadeh, R. Sedaghati, Existence of periodic solu-
tions for the generalized form of mathieu equation, Nonlinear Dynamics
39 (2005) 335–348. URL: https://doi.org/10.1007%2Fs11071-005-4338-y.
doi:10.1007/s11071-005-4338-y.
[10] M. Younis, MEMS Linear and Nonlinear Statics and Dynamics, Microsystems,
Springer US, 2011.
[11] J. F. Rhoads, S. W. Shaw, K. L. Turner, J. Moehlis, B. E. DeMartini,
W. Zhang, Generalized parametric resonance in electrostatically actuated micro-
electromechanical oscillators, Journal of Sound and Vibration 296 (2006) 797–
829. URL: https://doi.org/10.1016%2Fj.jsv.2006.03.009. doi:10.1016/
j.jsv.2006.03.009.
[12] W. Zhang, R. Baskaran, K. L. Turner, Effect of cubic nonlinearity on
auto-parametrically amplified resonant MEMS mass sensor, Sensors and Ac-
tuators A: Physical 102 (2002) 139–150. URL: https://doi.org/10.1016%
2Fs0924-4247%2802%2900299-6. doi:10.1016/s0924-4247(02)00299-6.

17
[13] S. W. Shaw, K. L. Turner, J. F. Rhoads, R. Baskaran, Parametrically excited
MEMS-based filters, in: IUTAM Symposium on Chaotic Dynamics and Control
of Systems and Processes in Mechanics, Springer-Verlag, 2005, pp. 137–146.
doi:10.1007/1-4020-3268-4_13.

[14] W. Magnus, S. Winkler, Hill’s equation, Interscience Publishers (John Wiley),


1966.

[15] A. Gasull, A. Guillamon, V. Mañosa, F. Mañosas, The period function for


hamiltonian systems with homogeneous nonlinearities, Journal of Differen-
tial Equations 139 (1997) 237–260. URL: https://doi.org/10.1006%2Fjdeq.
1997.3296. doi:10.1006/jdeq.1997.3296.

[16] A. Cima, A. Gasull, F. Mañosas, Period function for a class of hamiltonian


systems, Journal of Differential Equations 168 (2000) 180–199. URL: https:
//doi.org/10.1006%2Fjdeq.2000.3912. doi:10.1006/jdeq.2000.3912.

[17] Y. Wang, S. Adams, J. Thorp, N. MacDonald, P. Hartwell, F. Bertsch, Chaos in


MEMS, parameter estimation and its potential application, IEEE Transactions
on Circuits and Systems I: Fundamental Theory and Applications 45 (1998)
1013–1020. doi:10.1109/81.728856.

[18] J. F. Rhoads, S. W. Shaw, K. L. Turner, R. Baskaran, Tunable microelectrome-


chanical filters that exploit parametric resonance, Journal of Vibration and
Acoustics (2005).

[19] B. DeMartini, J. Moehlis, K. Turner, J. Rhoads, S. Shaw, W. Zhang, Model-


ing of parametrically excited microelectromechanical oscillator dynamics with
application to filtering, in: SENSORS, 2005 IEEE, IEEE, 2005, pp. 4–pp.

[20] B. DeMartini, H. Butterfield, J. Moehlis, K. Turner, Chaos for a microelec-


tromechanical oscillator governed by the nonlinear mathieu equation, Journal
of Microelectromechanical Systems 16 (2007) 1314–1323. URL: https://doi.
org/10.1109%2Fjmems.2007.906757. doi:10.1109/jmems.2007.906757.

[21] J. Llibre, R. Ortega, On the families of periodic orbits of the sitnikov problem,
SIAM Journal on Applied Dynamical Systems 7 (2008) 561–576. URL: https:
//doi.org/10.1137%2F070695253. doi:10.1137/070695253.

18

You might also like