1 Introduction
Let M be a (possibly noncompact) boundaryless smooth manifold embedded in ℝk. Consider the following parameterized delayed motion equation on M:
(1.1)
π
(
t
)
=
h
(
x
(
t
)
,
(
t
)
)
+
λ
f
(
t
,
x
(
t
)
,
(
t
)
,
x
(
t
-
τ
)
,
(
t
-
τ
)
)
,
where (t)π stands for the tangent part of the acceleration (t)∈ℝk at the point x(t)∈M, λ is a nonnegative real parameter, τ>0 is a constant time lag, and h,f are tangent vector fields on M. More precisely, the maps h:TM→ℝk and f:ℝ×TM×TM→ℝk are assumed to be continuous and such that h(p,v) and f(t,p,v,q,w) belong to TpM for any (t,p,v,q,w)∈ℝ×TM×TM. Here, given p∈M, TpM⊆ℝk is the tangent space of M at p, and
T
M
=
{
(
p
,
v
)
∈
ℝ
k
×
ℝ
k
:
p
∈
M
,
v
∈
T
p
M
}
is the tangent bundle of M. A second order delay differential equation (DDE for short) like (1.1) can be interpreted as a parameterized motion equation of a constrained mechanical system with configuration space M, acted on by a force which may as well depend on the state and velocity at a previous time instant. Indeed, for any fixed λ≥0, the right-hand side of (1.1) represents a delayed active force on M.
Assuming that the nonautonomuos vector field f is T-periodic in the first variable, we look for T-periodic solutions of (1.1). Firstly, we study the topological properties of the set of the so-called T-periodic pairs of (1.1) obtaining a global continuation result, Theorem 3.7 below. Then, in Section 4, we apply our result to get multiplicity of T-periodic solutions of (1.1) under additional assumptions.
To obtain our main result, the strategy essentially consists in writing (1.1) as a first order system on the tangent bundle TM which, in turn, is a boundaryless smooth submanifold of ℝk×ℝk. Thus, we take advantage of global continuation results for first order DDEs obtained by M. Furi and the third author in [19].
In the undelayed case, continuation results as well as multiplicity of T-periodic solutions for a second order ODE on M of the form
(1.2)
(
t
)
π
=
h
(
x
(
t
)
,
(
t
)
)
+
λ
Φ
(
t
,
x
(
t
)
,
(
t
)
)
,
where h is as in (1.1), and Φ:ℝ×TM→ℝk is T-periodic in the first variable and such that Φ(t,p,v) belongs to TpM for any (t,p,v)∈ℝ×TM, have been obtained in [18] and [15]. In the present paper, we aim to generalize all these results to the DDE case.
Recently, in collaboration with Benevieri and Furi, we devoted a series of papers to the study of first and second order DDEs on manifolds, focusing on continuation properties and existence of T-periodic solutions (see, e.g., [4] and references therein). In [4], we proved that, for any ε>0, the equation
(1.3)
π
(
t
)
=
Ψ
(
t
,
x
(
t
)
,
x
(
t
-
τ
)
)
-
ε
(
t
)
admits at least one forced oscillation (i.e., a T-periodic solution) for any continuous map Ψ:ℝ×M×M→ℝk which is T-periodic in the first variable and such that Ψ(t,p,q) belongs to TpM for any (t,p,q)∈ℝ×M×M, provided that the constraint M is compact with nonzero Euler–Poincaré characteristic. As far as we know, when the frictional coefficient ε is zero, the problem of the existence of forced oscillations of (1.3) is still open, even in the undelayed case. For another approach to the study of T-periodic solutions of DDEs on manifolds see, e.g., [2, 1].
More recently, we investigated a different and related class of equations, called retarded functional differential equations (RFDEs for short), in which the vector field(s) may depend on the whole history of the process. In [6], we studied, on a manifold M, a second order RFDE with infinite delay of the type
(1.4)
x
π
′′
(
t
)
=
F
(
t
,
x
t
)
,
where, with the standard notation in functional equations, the functional field F is a map
F
:
ℝ
×
B
U
(
(
-
∞
,
0
]
,
M
)
→
ℝ
k
such that F(t,φ)∈Tφ(0)M for all (t,φ) (here BU((-∞,0],M) denotes the set of bounded and uniformly continuous maps from (-∞,0] into M). In the particular case of the retarded spherical pendulum, i.e., when M=S2, in [6], we proved that (1.4) has a forced oscillation under the assumptions that F is bounded and locally Lipschitz in the second variable.
In principle, RFDEs represent a generalization of DDEs (as well as of ODEs). Nevertheless, due also to their broad generality, RFDEs require technical assumptions to guarantee basic properties of solutions (such as uniqueness and continuation, see, e.g., [5]), and we do not know how to overcome this difficulty. For instance, in [6], the functional field F is assumed to be locally Lipschitz in the second variable, unlike here, where the main results require the sole continuity assumption. Another characteristic of equations like (1.4) is that the right-hand side is not allowed to depend on the velocity (notice, however, that in [7] a different approach has been proposed to tackle this kind of problems). This can be a weak point in view of applications. The same drawbacks plague also the situation when, unlike in (1.4), the functional field F perturbs a nonzero vector field. In fact, in [10] for the spherical pendulum (i.e., when M=S2), as well as in [11] for the scalar case (i.e., M=ℝ), multiplicity results for T-periodic solutions of functional perturbations of autonomous vector fields have been obtained by the authors, even in the possible presence of friction, at the cost of imposing the same limitations as above on the forcing perturbations. This does not come as a surprise as also for first order RFDEs, like in, e.g., [16], it seems necessary to impose similar restrictions on the forcing term.
By contrast, continuation results as well as multiplicity of T-periodic solutions for periodically forced second order ODEs on manifolds can be obtained, under the sole continuity of the involved vector fields.
Roughly speaking, the problem that we address here could be considered as an “intermediate” situation between the ODE case of equation (1.2) and the RFDE one, where an equation similar to (1.4) is considered. In this middle ground, only minimal regularity assumptions are required, and the active forces may be allowed to depend explicitly on the velocity (which is not the case with (1.4)), thus providing greater flexibility. Besides filling a theoretical gap, our investigation is also justified by applications, since delay differential equations with constant time lag emerge naturally in applied mathematics (see, e.g., [3, 9, 13]).
To illustrate our results, we provide an application to the spherical pendulum, similar in spirit to that of [10], that besides permitting a more general form of friction, can involve a merely continuous periodic forcing term that depends on time and on present and past position and velocity (with a fixed time lag).
2 Preliminaries
2.1 The Degree of a Tangent Vector Field
Let M be a boundaryless smooth manifold embedded in ℝk, and let w:M→ℝk be a continuous tangent vector field on M, that is, w(p) belongs to TpM, the tangent space of M at p. Given an open subset U of M, we say that the pair (w,U) is admissible if w-1(0)∩U is compact. In this case, one can associate to the pair (w,U) an integer, deg(w,U), called the degree (or Euler characteristic) of w in U, which, roughly speaking, counts algebraically the number of zeros of w in U (for general reference see, e.g., [20, 21, 25]).
If w is (Fréchet) differentiable at p∈M and w(p)=0, then the differential dwp:TpM→ℝk maps TpM into itself, so that the determinant detdwp of dwp is defined. If, in addition, p is a nondegenerate zero (i.e., dwp:TpM→ℝk is injective), then p is an isolated zero and detdwp≠0. In fact, if w is admissible for the degree in U, when the zeros of w are all nondegenerate, then the set w-1(0)∩U is finite and
(2.1)
deg
(
w
,
U
)
=
∑
p
∈
w
-
1
(
0
)
∩
U
sign
det
d
w
p
.
Observe that in the flat case, i.e., when M=ℝk, deg(w,U) coincides with the Brouwer degree degB(w,V,0), where V is any bounded open neighborhood of w-1(0)∩U whose closure is contained in U. All the standard properties of the Brouwer degree on open subsets of Euclidean spaces, such as homotopy invariance, excision, additivity, existence, are still valid in the more general context of differentiable manifolds (see, e.g., [25]).
The excision property allows the introduction of the notion of index of an isolated zero of a tangent vector field. Indeed, let p∈M be an isolated zero of w. Clearly, deg(w,V) is well-defined for each open V⊆M such that V∩w-1(0)={p}. By the excision property, deg(w,V) is constant with respect to such V’s. This common value of deg(w,V) is, by definition, the index of w at p, and is denoted by i(w,p). With this notation, if (w,U) is admissible and all the zeros of w in U are isolated, the additivity property yields that
(2.2)
deg
(
w
,
U
)
=
∑
p
∈
w
-
1
(
0
)
∩
U
i
(
w
,
p
)
.
By formula (2.1), we have that if p is a nondegenerate zero of w, then
i
(
w
,
p
)
=
sign
det
d
w
p
.
Notice that (2.1) and (2.2) differ in the fact that, in the latter, the zeros of w are not necessarily nondegenerate as they have to be in the former. In fact, in (2.2), w need not be differentiable at its zeros.
In the case when M is a compact manifold, the celebrated Poincaré–Hopf theorem states that deg(w,M) coincides with the Euler–Poincaré characteristic χ(M) of M and, therefore, is independent of w. In particular, if all the zeros of w are isolated, it follows that
χ
(
M
)
=
∑
p
∈
w
-
1
(
0
)
i
(
w
,
p
)
.
2.2 Delay Differential Equations on Manifolds
In this paper, we will mainly be concerned with second order delay differential equations (DDEs for short) on a boundaryless smooth manifold M⊆ℝk. Observe that M need not be compact, nor closed as a subset of ℝk. Sometimes it will be convenient to write a second order DDE, in an equivalent way, as a first order DDE on the tangent bundle TM⊆ℝ2k:=ℝk×ℝk of M. For this reason, in this section, we also introduce a “general” first order DDE on a manifold N⊆ℝs. Then, in the sequel of the paper, we will consider the case in which N=TM⊆ℝ2k and, as we will show, the first order equation will descend from a second order one.
Let N⊆ℝs be a boundaryless smooth manifold. A continuous map G:ℝ×N×N→ℝs is said to be a (time-dependent) vector field tangent to N in the second variable if
G
(
t
,
ξ
,
)
∈
T
ξ
N
for all
(
t
,
ξ
,
)
∈
ℝ
×
N
×
N
.
Given τ≥0 and G as above, consider the delay differential equation
(2.3)
(
t
)
=
G
(
t
,
x
(
t
)
,
x
(
t
-
τ
)
)
.
By a solution of (2.3) we mean a continuous function x:J→N, defined on a (possibly unbounded) interval with length greater than τ, which is of class C1 on the subinterval (infJ+τ,supJ) of J and verifies (t)=G(t,x(t),x(t-τ)) for all t∈J with t>infJ+τ.
Let us now go back to the manifold M⊆ℝk and introduce a second order DDE on M. A continuous map φ:ℝ×TM×TM→ℝk such that φ(t,p,v,q,w)∈TpM for all (t,p,v,q,w)∈ℝ×TM×TM will be called a (time-dependent) delayed active force on M.
Given τ≥0 and a delayed active force φ on M, the motion equation associated with φ can be written in the form
(2.4)
(
t
)
π
=
φ
(
t
,
x
(
t
)
,
(
t
)
,
x
(
t
-
τ
)
,
(
t
-
τ
)
)
,
where (t)π∈Tx(t)M stands for the tangent component of the acceleration (t)∈ℝk at the point x(t). In fact, equation (2.4) can be interpreted as the motion equation of a constrained mechanical system with configuration space M, acted on by a force which may depend as well on the state and velocity at a previous time instant.
By a solution of (2.4), we mean a continuous function x:J→M, defined on a (possibly unbounded) interval, with length greater than τ, which is of class C2 on the subinterval (infJ+τ,supJ) of J and verifies
(
t
)
π
=
φ
(
t
,
x
(
t
)
,
(
t
)
,
x
(
t
-
τ
)
,
(
t
-
τ
)
)
for all t∈J with t>infJ+τ.
2.3 Parameterized DDEs and the Degree of a Second Order Vector Field
From now on, we will deal with a special case of (2.4), namely, we will consider a parameterized second order DDE of the form
(Eλ)
π
(
t
)
=
h
(
x
(
t
)
,
(
t
)
)
+
λ
f
(
t
,
x
(
t
)
,
(
t
)
,
x
(
t
-
τ
)
,
(
t
-
τ
)
)
,
where λ≥0, h:TM→ℝk is continuous and such that h(p,v)∈TpM for any (p,v)∈TM, and
f
:
ℝ
×
T
M
×
T
M
→
ℝ
k
is a time-dependent delayed active force.
It is known (see, e.g., [14]) that, by setting ξ(t)=(x(t),(t)), (Eλ) can be written as a first order equation on the tangent bundle TM in the form
(2.5)
(
t
)
=
(
ξ
(
t
)
)
+
λ
(
t
,
ξ
(
t
)
,
ξ
(
t
-
τ
)
)
,
where :TM→ℝk×ℝk and :ℝ×TM×TM→ℝk×ℝk are defined by
(
p
,
v
)
=
(
v
,
r
(
p
,
v
)
+
h
(
p
,
v
)
)
,
(
t
,
p
,
v
,
q
,
w
)
=
(
0
,
f
(
t
,
p
,
v
,
q
,
w
)
)
,
and r:TM→ℝk is a smooth map, quadratic in the second variable v∈TpM for any p∈M, with values in (TpM)⊥. Such a map r may be interpreted as the reactive force due to the constraint M. Actually, r(p,v) is the unique vector in ℝk which makes (v,r(p,v)) tangent to TM at (p,v). The map , called the second order vector field associated toh, is a tangent vector field on TM. It is readily verified that also is a vector field tangent to TM, even if it is not a second order vector field. Hence, for any fixed λ>0, the map
G
(
t
,
ξ
,
)
:
=
(
ξ
)
+
λ
(
t
,
ξ
,
)
,
ξ
=
(
p
,
v
)
,
=
(
q
,
w
)
,
is a time-dependent vector field tangent to TM in the second variable, and (2.5) is a first order DDE on TM of the form (2.3). According to the definition of solution of first and second order DDEs given in Section 2.2, one has that equations (Eλ) and (2.5) are equivalent, in the sense that x:J→M is a solution of (Eλ) if and only if ξ=(x,):J→TM is a solution of (2.5).
We close this section with a formula which expresses the degree of , which is a tangent vector field on TM, in terms of the degree of h|M:=h(⋅,0), which is a tangent vector field on M. This notation, as well as the one W∩M in Lemma 2.1 below, is motivated by the identification of M with the zero section of TM. Indeed, [18, Lemma 3.2 and Corollary 3.3] yield the following:
Lemma 2.1.
Let h and be as above, and let W be an open subset of TM. Then h|M is admissible on W∩M if and only if is admissible on W, and deg(,W)=deg(-h|M,W∩M). Furthermore, if M is compact, and W is such that W∩M=M, then deg(,W)=χ(M).
3 Branches of T-Periodic Pairs
In this section, assuming that the delayed active force f is periodic of period T>0 in t, we study the set of T-periodic solutions of the parameterized second order DDE (Eλ). Roughly speaking, we will consider the set of pairs (λ,x), with x a T-periodic solution of (Eλ) corresponding to λ≥0, which will be called T-periodic pairs (see Definition 3.4 below). We will prove a global continuation property of the set of T-periodic pairs. Namely, we will show that, under appropriate assumptions, there are connected sets of T-periodic pairs “emanating”, so to speak, from the set of zeros of h(⋅,0). In order to carry out this program, we will first specify the ambient space for the set of T-periodic pairs and then define a corresponding notion for general first order equations on manifolds for which a similar result holds. The latter, applied to equation (2.5), will be a major step in view of the main result of this section, Theorem 3.7.
Let us begin with some notation. Given a subset A of ℝn, we will denote by CTj(A), j∈{0,1}, the metric subspace of the Banach space (CTj(ℝn),∥⋅∥j) of all the T-periodic Cj maps x:ℝ→A with the usual Cj norm (when j=0 we will simply write CT(A)). Observe that CTj(A) is not complete unless A is complete (i.e. closed in ℝn). Nevertheless, if A is locally compact, then CTj(A) is locally complete.
As pointed out, it is convenient to preliminarily define the notion of T-periodic pair for parameterized first order DDEs. This is important because, as already mentioned, there is a very strict correlation between the set of T-periodic pairs of (Eλ) and the corresponding set for the associated first order equation (2.5).
Consider the following parameterized first order DDE on a (boundaryless smooth) manifold N⊆ℝs:
(3.1)
(
t
)
=
γ
(
x
(
t
)
)
+
λ
ψ
(
t
,
x
(
t
)
,
x
(
t
-
τ
)
)
,
where γ:N→ℝs is a tangent vector field on N, and ψ:ℝ×N×N→ℝs is a time-dependent vector field, T-periodic in the first variable and tangent to N in the second one.
Equation (3.1) is clearly a special case of (2.3). We say that (λ,x)∈[0,∞)×CT(N) is a T-periodic pair for (3.1) if x is a T-periodic solution of (3.1) corresponding to λ. If λ=0 and x is constant, then (λ,x) is said to be trivial. Indeed, it is not hard to see that trivial T-periodic pairs of (3.1) correspond to zeros of γ. More precisely, for any p∈γ-1(0), we have a trivial T-periodic pair (0,p¯) of (3.1), and, conversely, if (0,x)∈[0,∞)×CT(N) is a trivial T-periodic pair, then, by local uniqueness of the solutions of Cauchy problem, x(t)≡p for some p∈γ-1(0). Notice that (3.1) may have nonconstant T-periodic solutions for λ=0, as it happens, for example, in N=ℝ2, for T=2π and γ(x1,x2)=(-x2,x1).
Denote by Y the set of all the T-periodic pairs of (3.1). Known properties of the set of solutions of differential equations imply that Y is closed, hence it is locally complete, as a closed subset of a locally complete space. We will use the following fact, see [17, Lemma 3.1].
Lemma 3.1.
The set Y is locally compact. Moreover, if N is closed in Rs, any bounded subset of Y is actually totally bounded. As a consequence, in this case, closed and bounded sets of T-periodic pairs are compact.
The following result (see [19, Theorem 5.1]) concerns the structure of the set of T-periodic pairs of (3.1).
Let us introduce some further notation. Given N⊆ℝs, denote by i:N→[0,∞)×CT(N) the map that to any point ξ∈N associates the pair (0,ξ¯), where ξ¯ denotes the function constantly equal to ξ. It is not difficult to show that i is a closed map. Given an open subset 𝒪 of [0,∞)×CT(N), we introduce the handy notation 𝒪∩N for the open subset i-1(𝒪) of N.
Theorem 3.2 (see [19]).
Let γ:N→Rs and ψ:R×N×N→Rs be as in (3.1). Let O be an open subset of [0,∞)×CT(N), and assume that deg(γ,O∩N) is defined and nonzero. Then O contains a connected set of nontrivial T-periodic pairs for (3.1) whose closure in O meets the set {(0,ξ¯)∈O:γ(ξ)=0} and is not compact.
Our aim is to specialize Theorem 3.2 to the case of the first order DDE (2.5), i.e., when N=TM is the tangent bundle of M⊆ℝk, and the vector fields descend from a second order equation. In this situation, it is convenient to consider the map :M→[0,∞)×CT(TM) that associates p∈M to (0,(p,0)¯), where (p,0)¯ is the map in CT(TM) constantly equal to (p,0). Similarly to what we have done above, given an open subset Ω^ of [0,∞)×CT(TM), we denote by Ω^∩M the open set (Ω^)-1.
Theorem 3.2 combined with Lemma 2.1 yields the following fact concerning the T-periodic pairs of (2.5).
Lemma 3.3.
Let h, f, and be as in Section 2.3. Assume in addition that f, are T-periodic in t. Let Ω^⊆[0,∞)×CT(TM) be open. Assume that deg(h|M,Ω^∩M) is defined and nonzero. Then Ω^ contains a connected set of nontrivial T-periodic pairs for (2.5) whose closure in Ω^ meets the set {(0,(p,0)¯)∈Ω^:h|M(p)=0} and is not compact.
Proof.
Notice that with our notation, we have
(
0
)
-
1
∩
(
Ω
^
∩
T
M
)
=
{
(
p
,
0
)
∈
Ω
^
∩
T
M
:
h
(
p
,
0
)
=
0
}
=
{
p
∈
Ω
^
∩
M
:
h
|
M
(
p
)
=
0
}
=
h
|
M
-
1
(
0
)
∩
(
Ω
^
∩
M
)
.
Thus, by Lemma 2.1, we have deg(,Ω^∩TM)=deg(h|M,Ω^∩M). The assertion now follows from Theorem 3.2. ∎
We finally wish to apply Lemma 3.3 in order to obtain, for the second order DDE (Eλ), a global result in the spirit of Theorem 3.2. To this end, we need to introduce the notion of T-periodic pair for (Eλ), where we assume that f is T-periodic in the first variable.
Definition 3.4.
A pair (λ,x)∈[0,∞)×CT1(M) is called a T-periodic pair for the second order DDE (Eλ) if x is a T-periodic solution of (Eλ) corresponding to λ. In particular, we will say that (λ,x) is trivial if λ=0, and x is constant.
Observe that for λ=0, (Eλ) may also have nonconstant T-periodic solutions. This may happen when h=0, as in the case of the inertial motion on a sphere, but also for tangent vector fields with compact h-1(0) as, for instance, in M=ℝ and h(p)=-p. Thus one might have nontrivial T-periodic pairs also for λ=0. Denote by X⊆[0,∞)×CT1(M) and by ⊆[0,∞)×CT(TM) the set of all the T-periodic pairs of (Eλ) and (2.5), respectively.
As pointed out in Section 2.3, the equations (Eλ) and (2.5) are in some sense equivalent. So, also their T-periodic pairs correspond each other in the sense of the following remark.
Remark 3.5.
The map
[
0
,
∞
)
×
C
T
1
(
M
)
∋
(
λ
,
x
)
↦
𝑗
(
λ
,
x
,
)
∈
[
0
,
∞
)
×
C
T
(
T
M
)
establishes a correspondence between the sets X and that “preserves” the notion of triviality for a T-periodic pair. More precisely, if X*⊆X and ⊆* denote the sets of the trivial T-periodic pairs of (Eλ) and (2.5), respectively, the map j, when restricted to X, is a homeomorphism of X onto under which X* corresponds to *. Furthermore, j is a homeomorphism of [0,∞)×CT1(M) onto its image; thus, as a restriction of a linear map defined on ℝ×CT1(ℝk), it is actually a Lipschitz map with Lipschitz inverse. Consequently, under this homeomorphism, bounded sets correspond to bounded sets, and totally bounded sets correspond to totally bounded sets.
By this remark and Lemma 3.1, we get that X is locally compact. If, moreover, M is closed in ℝk, then the set X has in addition the following useful property.
Remark 3.6.
Assume that M is closed in ℝk. If A⊆X is bounded, by Remark 3.5, j(A) is bounded as well. Since TM is closed, Lemma 3.1 implies that j(A) is totally bounded, and thus, again from Remark 3.5, it follows that A=j-1(j(A)) is totally bounded. As a consequence, CT1(M) being complete, closed and bounded subsets of X are compact.
Consider the map i*:M→[0,∞)×CT1(M) that associates to any p∈M the constant pair (0,p¯), p¯∈CT1(M) being the function constantly equal to p. Notice that if N=TM, the map i:TM→[0,∞)×CT(TM), introduced before Theorem 3.2, is the composition of i* with the inclusion of [0,∞)×CT1(M) in [0,∞)×CT(M). In order to clarify the relations existing between the maps i, , i* and j introduced above, consider the following diagram:
Let Ω be an open subset of [0,+∞)×CT1(M), and denote, for simplicity, by Ω∩M the open subset (i*)-1(Ω) of M. Theorem 3.7 gives conditions which ensure the existence in Ω of a connected subset of nontrivial T-periodic pairs of equation (Eλ) (also called an Ω-global bifurcating branch), whose closure in Ω is noncompact and intersects the set of trivial T-periodic pairs contained in Ω.
We are now ready to state our main result, Theorem 3.7 below.
Theorem 3.7.
Let h:TM→Rk be continuous and such that h(p,v)∈TpM for any (p,v)∈TM, and let f:R×TM×TM→Rk be a T-periodic delayed active force on M. Let Ω⊆[0,∞)×CT1(M) be an open set, and assume that deg(h|M,Ω∩M) is defined and nonzero. Then Ω contains a connected set of nontrivial T-periodic pairs for (Eλ), whose closure in Ω meets the set {(0,p¯)∈Ω:h|M(p)=0} and is not compact.
Proof.
Let j be the map introduced in Remark 3.5. Since j(Ω) is relatively open in
j
(
[
0
,
∞
)
×
C
T
1
(
M
)
)
⊆
[
0
,
∞
)
×
C
T
(
T
M
)
,
there exists an open subset Ω^ of [0,∞)×CT(TM) such that
Ω
^
∩
j
(
[
0
,
∞
)
×
C
T
1
(
M
)
)
=
j
(
Ω
)
.
With our notation we have that Ω∩M=Ω^∩M.
As previously, denote by X and the set of T-periodic pairs of equations (Eλ) and (2.5), respectively. Let Γ⊆ be a connected component of T-periodic pairs for (2.5) as in Lemma 3.3. Then (j|)-1(Γ) is a connected set as in the assertion. ∎
We give now some consequences of Theorem 3.7 in the particular case when M is closed as a subset of ℝk, so that Remark 3.6 applies. The first corollary helps visualizing the notion of branch of T-periodic pairs, under this additional assumption.
Corollary 3.8.
Let M, h and f be as in Theorem 3.7, and assume in addition that M is closed in Rk. Let Ω⊆[0,∞)×CT1(M) be open, and assume that deg(h|M,Ω∩M) is defined and nonzero. Then a connected subset Γ of Ω, as in Theorem 3.7, is either unbounded or, if bounded, its closure cl(Γ) in cl(Ω) reaches the boundary ∂Ω of Ω.
Proof.
Assume that Γ is bounded, then so is its closure cl(Γ). By Remark 3.6, we get that cl(Γ) is compact. But Γ is not contained in any compact subset of Ω, hence cl(Γ)∩Ω≠∅. ∎
Another consequence of Theorem 3.7 is the following “geometric” property of the set of T-periodic pairs of (Eλ), which is in the spirit of a celebrated result due to P. H. Rabinowitz [26]. Its proof, that we provide here for completeness, is taken – almost verbatim – from [18, Corollary 4.4]
Corollary 3.9.
Let M, h and f be as in Theorem 3.7, and assume in addition that M is closed in Rk. Let U be an open subset of M such that deg(h|M,U)≠0. Then equation (Eλ) admits a connected set Γ of nontrivial T-periodic pairs whose closure cl(Γ) contains some (0,p¯0) with p0∈U and is either unbounded or goes back to some (0,p¯), where p∉U.
Proof.
Consider the following open subset of [0,∞)×CT1(M):
Ω
=
(
[
0
,
∞
)
×
C
T
1
(
M
)
)
∖
(
h
|
M
-
1
(
0
)
∖
U
)
.
Since Ω∩h|M-1(0)=U∩h|M-1(0), the excision property of the degree yields
deg
(
h
|
M
,
Ω
∩
M
)
=
deg
(
h
|
M
,
U
)
≠
0
.
Thus, by Theorem 3.7, we get the existence of a connected set Γ of nontrivial T-periodic pairs whose closure cl(Γ) in [0,∞)×CT1(M) is not contained in any compact subset of Ω.
Assume that
cl
(
Γ
)
∩
(
h
|
M
-
1
(
0
)
∖
U
)
=
∅
,
in this case cl(Γ)⊆Ω. Since M⊆ℝk is closed, Remark 3.6 implies that cl(Γ) cannot be both bounded and complete, otherwise it would be compact. Thus cl(Γ), being a closed subset of the complete metric space [0,∞)×CT1(M), must be unbounded. ∎
Finally, if M is compact with nonzero Euler–Poincaré characteristic, we get the following immediate consequence of Theorem 3.7.
Corollary 3.10.
Let M, h and f be as in Theorem 3.7. Assume in addition that M is compact with χ(M)≠0. Then there exists an unbounded connected set of nontrivial T-periodic pairs for (Eλ) whose closure in [0,∞)×CT1(M) intersects the set of trivial T-periodic pairs {(0,p¯):h|M(p)=0}.
4 Multiplicity Results
In this section, we apply our global continuation theorem to obtain multiplicity results for T-periodic solutions of (Eλ). Our results are related to those of [15], where the ODE case is considered, and on the other hand, to those of [10], in which we studied the equation of the retarded spherical pendulum. Key notions for our investigation will be those of T-resonance and of ejecting set (see, e.g., [8, 10, 15]).
Let us start by recalling the definition of ejecting set. Let Z be a metric space and X a subset of [0,+∞)×Z. Given λ≥0, we denote by Xλ the slice {z∈Z:(λ,z)∈X}. We recall the following definition from [15].
Definition 4.1.
We say that E⊆X0 is ejecting (for X) if it is relatively open in X0, and there exists a connected subset of X which meets {0}×E and is not contained in {0}×X0.
By inspection of the proof of [15, Theorem 3.3] one immediately sees that the following result holds.
Theorem 4.2 (see [15]).
Let Z be a metric space, and let X be a locally compact subset of [0,+∞)×Z. Assume that X0 contains n pairwise disjoint ejecting subsets E1,…,En. Suppose that n-1 of them are compact. Then there are open neighborhoods U1,…,Un in Z of E1,…,En, respectively, with pairwise disjoint closure, and a positive number λ* such that for λ∈[0,λ*)
X
λ
∩
U
i
≠
∅
,
i
=
1
,
…
,
n
.
In particular, we have that the cardinality of Xλ is greater than or equal to n for any λ∈[0,λ*).
A notion that is useful for the investigation of the set of T-periodic pairs in a neighborhood of a stationary point is that of T-resonance ([12], see also [8, 15]). Usually, this notion is introduced in the context of first order equations; however, for the sake of simplicity and given the focus of this paper, we confine ourselves to the case of equation (Eλ) for λ=0 (for this approach compare [10, Appendix]). Namely,
(4.1)
π
(
t
)
=
h
(
x
(
t
)
,
(
t
)
)
,
where h:TM→ℝk is as in (Eλ). Here, h is further assumed to be C1.
Definition 4.3.
Given T>0, a point p∈h|M-1(0) is said to be (second order) T-resonant for h if the linearized equation on TpM,
z
′′
=
∂
1
h
(
p
,
0
)
z
+
∂
2
h
(
p
,
0
)
z
′
,
admits T-periodic solutions other than the trivial one z(t)≡0. If this is not true, it is also customary to say informally that a point p∈h|M-1(0) is non-T-resonant for equation (4.1).
Remark 4.4.
The resonance of a zero p of h|M can be decided with a simple criterion. As already pointed out in Section 2.1, p∈h|M-1(0) implies that the differential d(h|M)p:TpM→ℝk maps TpM into itself. As shown, e.g., in [22] by a direct computation, one has that p is (second order) T-resonant for h if there exists ℓ∈ℤ such that
(4.2)
det
(
∂
1
h
(
p
,
0
)
+
2
π
ℓ
i
T
∂
2
h
(
p
,
0
)
+
(
2
π
ℓ
T
)
2
I
)
=
0
,
where I is the identity on TpM. In particular, if p is not T-resonant, taking ℓ=0, we have that
det
(
∂
1
h
(
p
,
0
)
)
≠
0
.
The connection between the notions of ejecting set and of T-resonant point is underlined by the following lemma that shows that, for the set of T-periodic pairs of the parameterized equation (Eλ), a zero of h|M which is not T-resonant corresponds to an ejecting set (or rather, ejecting point).
Lemma 4.5.
Let p be a zero of h|M which is not T-resonant. Then {p¯} is an ejecting set for the set X⊆[0,∞)×CT1(M) of T-periodic pairs of (Eλ).
Although there are important differences, the proof of this lemma broadly follows the same strategy as that of [10, Lemma 3.8]. For this reason, we only sketch it here.
Sketch of the proof of Lemma 4.5.
The proof is performed in three steps.
A similar argument to [10, Lemma 3.6] shows that since p is not T-resonant then, for any sufficiently small neighborhood V of p¯ in CT1(M), there exists a real number δV>0 such that [0,δV]×∂V does not contain any T-periodic pair of (Eλ). In particular, {p¯} is open in the slice X0.
By step (1), we have that h|M is admissible for the degree in any sufficiently small neighborhood of p. Furthermore, by Remark 4.4, we have that det(∂1h(p,0))≠0. Hence, the degree of h|M in a sufficiently small neighborhood of p is ±1.
By step (2), we have that in any sufficiently small neighborhood U of p the degree of h|M is nonzero. Take any neighborhood V as in step (1), and let Ω=[0,δV)×V. Restricting V if necessary, we can assume Ω∩M⊆U. Theorem 3.7 yields a connected set Γ of nontrivial T-periodic pairs for (Eλ), whose closure in Ω meets the set {(0,p¯)∈Ω:h|M(p)=0} and is not compact. By step (1), this set Γ cannot be contained in {0}×X0. Hence, taking into account that {p¯} is open in X0 by step (1), we conclude that {p¯} is an ejecting set for X. ∎
We are now in a position to prove the main result of this section, Theorem 4.6. This multiplicity result provides a generalization of [15, Theorem 3.7] to the delay differential equation (Eλ), see also [10, Theorem 3.9].
Theorem 4.6.
Let h:TM→Rk be continuous and such that h(p,v)∈TpM for any (p,v)∈TM, and let f:R×TM×TM→Rk be a T-periodic delayed active force on M. Assume further that M is closed in Rk and that h is C1 with h|M-1(0) compact. Let p1,…,pn-1 be zeros of h|M which are not T-resonant and such that
(4.3)
∑
i
=
1
n
-
1
i
(
h
|
M
,
p
i
)
≠
deg
(
h
|
M
,
M
)
.
Assume that the unperturbed equation
=
π
h
(
x
,
)
does not admit, in CT1(M), unbounded connected sets of T-periodic solutions that meet (h|M)-1(0). Then, for λ>0 sufficiently small, equation (Eλ) admits at least n solutions of period T.
Proof.
Observe first that the points p1,…,pn-1, being zeros of h|M which are not T-resonant, are isolated in h|M-1(0). As above, let X be the set of T-periodic pairs for (Eλ).
By Lemma 4.5, the sets {pi¯}, i=1,…,n-1, are ejecting for X; in particular, the constant functions pi¯, i=1,…,n-1, are isolated points of X0. Our aim is to prove the existence in the slice X0 of a further (possibly not compact) ejecting set. This will enable us to apply Theorem 4.2.
Let W1,…,Wn-1 be pairwise disjoint open neighborhoods of p1¯,…,pn-1¯ in CT1(M), respectively, with the property that
cl
(
W
i
)
∩
X
0
=
{
p
i
¯
}
for
i
=
1
,
…
,
n
-
1
.
Set
Ω
=
[
0
,
∞
)
×
(
C
T
1
(
M
)
∖
⋃
i
=
1
n
-
1
cl
(
W
i
)
)
.
By the additivity property of the degree, we get
deg
(
h
|
M
,
Ω
∩
M
)
=
deg
(
h
|
M
,
M
)
-
deg
(
h
|
M
,
⋃
i
=
1
n
-
1
(
W
i
∩
M
)
)
=
deg
(
h
|
M
,
M
)
-
∑
i
=
1
n
-
1
i
(
h
,
p
i
)
≠
0
.
Thus, Theorem 3.7 yields the existence of a connected set Γ⊆Ω of nontrivial T-periodic pairs for (Eλ) whose closure in Ω is not compact and meets the set of trivial T-periodic pairs. Let Υ be the connected component of X containing Γ.
We claim that Υ0, which is relatively open in X0, is an ejecting set. To see this, it is sufficient to show that Υ is not contained in {0}×X0. Assume by contradiction that this is not the case, that is Υ={0}×Υ0. Since [0,∞)×CT1(M) is complete, the closure cl(Υ) is complete as well. Hence, Υ is contained in the bounded complete subset cl(Υ) of Ω, which is a contradiction.
So far, we have proved that the subsets of X0,
E
1
:
=
{
p
1
¯
}
,
…
,
E
n
-
1
:
=
{
p
n
-
1
¯
}
,
E
n
:
=
Υ
0
,
are indeed n ejecting sets, the first n-1 of which are compact. The assertion now follows from Theorem 4.2. ∎
Remark 4.7.
Note that, when M is compact, by the Poincaré–Hopf theorem one gets deg(h|M,M)=χ(M), so that (4.3) becomes
∑
i
=
1
n
-
1
i
(
h
|
M
,
p
i
)
≠
χ
(
M
)
.
Observe that the multiple T-periodic solutions of (Eλ) given by Theorem 4.6 are distinct as elements of CT1(M). Nothing is said about their images. However, if we assume more about the manifold M or about the equation, a more precise statement is possible. An example of this fact is when M is a sphere in ℝ3, and h is a positional field with possible friction, see Section 5 below.
Another situation, where a more precise statement is possible, occurs when the map h of (Eλ) has a suitable form, as shown in the following proposition.
Proposition 4.8.
Let M be compact and g:M→Rk a tangent vector field on M. Let ν:[0,∞)→[0,∞) be continuous and such that
ν
(
s
)
>
0
for all
s
>
0
;
lim inf
s
→
∞
ν
(
s
)
=
ℓ
>
0
.
Let f be as in Theorem 4.6 and suppose, in addition, that f is bounded. Let p1,…,pn-1∈g-1(0) be non-T-resonant and such that
∑
i
=
1
n
-
1
i
(
g
,
p
i
)
≠
χ
(
M
)
.
Then, for λ>0 sufficiently small, the equation
(4.4)
π
(
t
)
=
g
(
x
(
t
)
)
-
ν
(
|
(
t
)
|
)
(
t
)
+
λ
f
(
t
,
x
(
t
)
,
(
t
)
,
x
(
t
-
τ
)
,
(
t
-
τ
)
)
admits n solutions of period T, whose images are pairwise not coincident.
Observe that the term -ν(|(t)|)(t) in equation (4.4) grants a quite generalized form of friction. In particular, for ν(s)=κ, κ a positive constant, we find the viscous friction formula -κ(t) so that equation (4.4) becomes
(
t
)
π
=
g
(
x
(
t
)
)
-
κ
(
t
)
+
λ
f
(
t
,
x
(
t
)
,
(
t
)
,
x
(
t
-
τ
)
,
(
t
-
τ
)
)
.
The proof of Proposition 4.8 is based on the following elementary technical lemma.
Lemma 4.9.
Let ν be a function as in Proposition 4.8, and let {ak}k∈N⊆[0,∞) be a sequence such that akν(ak)→0. Then ak→0.
Proof.
Observe first that, for any r>0, infs≥rν(s)>0, otherwise we could find an unbounded sequence {sk}k∈ℕ⊆[r,∞) with ν(sk)→0, against the assumption that lim infs→∞ν(s)>0.
Assume by contradiction that ak→0 as k→∞. Then, on one hand, ν(ak)→0. On the other hand, ak→0 means that there exists ε0>0 such that ak>ε0 for infinitely many k’s. By the first part of the proof, this means that
ν
(
a
k
)
≥
inf
s
≥
ε
0
ν
(
s
)
>
0
,
against ν(ak)→0. ∎
Proof of Proposition 4.8.
We claim that the unperturbed equation
=
π
g
(
x
)
-
ν
(
|
|
)
does not admit (in CT1(M)) unbounded connected sets of T-periodic solutions that meet g-1(0). We proceed by contradiction. Since lim infs→∞ν(s)=ℓ>0, there exists ζ>0 such that ν(s)>ℓ/2 for all s>ζ. By our contradictory assumption, as M is compact, there exists a T-periodic solution ξ of (4.4) such that
max
t
∈
[
0
,
T
]
|
(
t
)
|
>
max
{
ζ
,
2
G
ℓ
}
,
where G=maxp∈M|g(p)|. Put θ(t):=|(t)|2, and let σ∈ℝ be such that θ(σ)=max{θ(t):t∈ℝ}. Then
0
=
(
σ
)
=
2
〈
(
σ
)
,
(
σ
)
〉
=
2
〈
(
σ
)
,
π
(
σ
)
〉
=
2
〈
(
σ
)
,
g
(
ξ
(
σ
)
)
〉
-
2
ν
(
|
(
σ
)
|
)
|
(
σ
)
|
2
≤
2
|
(
σ
)
|
G
-
ℓ
|
(
σ
)
|
2
.
Hence maxt∈ℝ|(t)|=|(σ)|≤2G/ℓ. This contradiction proves the claim. Therefore, Theorem 4.6 with h(p,v)=g(p)-ν(|v|)v yields that (4.4) has, for λ>0 sufficiently small, at least n solutions x1λ,…,xnλ of period T.
We only need to prove that their images are pairwise not coincident. Let W1,…,Wn-1 be as in the proof of Theorem 4.6. By an argument as in step (1) of the proof of Lemma 4.5, reducing λ* if necessary, we can assume that, for i=1,…,n-1, Wi is the open ball in CT1(M) centered at pi¯ with radius equal to
𝔯
:
=
1
2
min
1
≤
i
<
j
≤
n
-
1
|
p
i
-
p
j
|
.
Thus, for all j=1,…,n-1, we have maxt∈ℝ|xjλ(t)-pj|<𝔯. That is, the image of xjλ is contained in the ball Wj. Hence, for i,j=1,…,n-1, i≠j, the images of xiλ and xjλ are confined into disjoint balls, hence they are disjoint.
To conclude the proof, it is sufficient to show that, reducing λ*>0 if necessary, we have that for λ∈[0,λ*), the image of xnλ cannot coincide with that of any of the other solutions x1λ,…,xn-1λ. Assume the contrary. Then there exists a sequence {λk}k∈ℕ, with λk↘0, such that the image of xnλk coincides with at least one of the images of x1λk,…,xn-1λk (not necessarily the same for all λk’s). Since the images of these solutions are disjoint, selecting a subsequence and reordering the solutions, we can assume that xnλk([0,T])=x1λk([0,T]) for all k. Letting k→∞, we have that the solution t↦xnλk(t) converges uniformly on [0,T] to the constant function t↦p1.
We claim that (t)nλk converges to zero uniformly on [0,T]. If we prove the claim, we get a contradiction (thus proving the assertion). To see this, let dk be the diameter of the orbit of xnλk, that is,
d
k
:
=
max
t
1
,
t
2
∈
[
0
,
T
]
|
x
n
λ
k
(
t
1
)
-
x
n
λ
k
(
t
2
)
|
.
Thus, dk→0 as k→∞.
Since g(p1)=0, there exist positive constants ρ and K such that |g(p)|≤K|p-p1| for all p∈M with |p-p1|<ρ. Thus, for sufficiently large k,
(4.5)
|
g
(
x
n
λ
k
(
t
)
)
|
≤
K
|
x
n
λ
k
(
t
)
-
p
1
|
for all t∈ℝ.
Now, for any sufficiently large k, let φk(t):=|nλk(t)|2, and let σk be such that φ(σk)=max{φk(t):t∈ℝ}. Taking the derivative of φk with respect to t at σk, we get, similarly to the first part of the proof,
(4.6)
0
≤
2
〈
n
λ
k
(
σ
k
)
,
g
(
x
n
λ
k
(
σ
k
)
)
〉
-
2
ν
(
|
n
λ
k
(
σ
k
)
|
)
|
n
λ
k
(
σ
k
)
|
2
+
λ
k
H
|
n
λ
k
(
σ
k
)
|
,
where H is an upper bound for the norm of the vector field f. Combining (4.6) with (4.5), we get
(4.7)
|
n
λ
k
(
σ
k
)
|
2
ν
(
|
n
λ
k
(
σ
k
)
|
)
≤
K
|
x
n
λ
k
(
σ
k
)
-
p
1
|
|
n
λ
k
(
σ
k
)
|
+
λ
k
H
|
n
λ
k
(
σ
k
)
|
.
We wish to prove that |nλk(σk)|→0 as k→∞. If |nλk(σk)|=0, there is nothing to prove. Therefore, we may assume |nλk(σk)|≠0. Inequality (4.7) becomes
|
n
λ
k
(
σ
k
)
|
ν
(
|
n
λ
k
(
σ
k
)
|
)
≤
K
|
x
n
λ
k
(
σ
k
)
-
p
1
|
+
λ
k
H
=
K
d
k
+
λ
k
H
.
Hence |nλk(σk)|ν(|nλk(σk)|)→0 as k→∞. Lemma 4.9 implies that |nλk(σk)|→0, whence the claim.
We conclude the proof by showing that the claim leads to a contradiction. Let Ω⊆[0,∞)×CT1(M) be as in the proof of Theorem 4.6, and put Wn=Ω0. We have Wj∩Wn=∅, j=1,…,n-1, and xnλk∈Wn for sufficiently large k. The claim, together with the fact that t↦xnλk(t) converges uniformly on [0,T] to the constant function t↦p1, implies that, eventually, xnλk∈W1. Hence xnλk∈W1∩Wn contradicting W1∩Wn=∅. ∎
Remark 4.10.
An inspection of the first part of the proof of Theorem 4.8 reveals that, in the hypotheses of Theorem 4.6, even without the additional assumptions of Theorem 4.8, the images of at least n-1 of the T-periodic solutions are actually disjoint (a stronger assertion than that of the theorem) for sufficiently small λ>0. Roughly speaking, these are the solutions branching away from the zeros of h|M which are not T-resonant.
5 An Application to the Spherical Pendulum
We illustrate here the results obtained in the previous sections by an application to multiplicity of forced oscillations for the spherical pendulum. Existence and multiplicity results for periodic orbits of pendulum and pendulum-like equations have been successfully pursued with a variety of techniques. We cannot provide here an exhaustive list of references, we only mention the survey papers [23, 24] for an historical perspective.
Let 𝑺⊆ℝ3 be the sphere centered at the origin with radius r>0, that is, 𝑺={x∈ℝ3:|x|=r}. Consider the following equation on 𝑺:
(5.1)
m
(
t
)
=
-
m
|
(
t
)
|
2
r
2
x
(
t
)
+
g
(
x
(
t
)
)
-
η
ν
(
|
(
t
)
|
)
(
t
)
+
λ
f
(
t
,
x
(
t
)
,
(
t
)
,
x
(
t
-
τ
)
,
(
t
-
τ
)
)
,
where
g
:
𝑺
→
ℝ
3
is a C1 tangent vector field on 𝑺;
ν
:
[
0
,
∞
)
→
[
0
,
∞
)
is as in Proposition 4.8;
f
:
ℝ
×
T
𝑺
×
T
𝑺
→
ℝ
3
, as in Theorem 4.6, is a T-periodic delayed active force on 𝑺.
Equation (5.1) represents the motion equation of a particle of mass m constrained to 𝑺 and acted on by three forces: the positional tangent vector field g, a possible friction and a T-periodic forcing term λf which depends on a past state of the system.
The coefficient η in equation (5.1) permits us to include, in contrast to Proposition 4.8, also the case of absence of friction. In fact, ν(s) is allowed to be zero only for s=0.
Observe that with an appropriate choice of units, we can assume without loss of generality that m=1. Moreover, the term R(p,v)=-m(|v|2/r2)p in equation (5.1) represents the reactive force due to the constraint. Hence, (5.1) can be rewritten, using our notation, as
(5.2)
π
(
t
)
=
g
(
x
(
t
)
)
-
η
ν
(
|
(
t
)
|
)
(
t
)
+
λ
f
(
t
,
x
(
t
)
,
(
t
)
,
x
(
t
-
τ
)
,
(
t
-
τ
)
)
.
Remark 5.1.
A physically relevant example is obtained when g is the tangential component of the gravitational force. That is, when g is the tangent vector field
g
𝔤
(
x
1
,
x
2
,
x
3
)
:
=
𝔤
r
2
(
x
3
x
1
,
x
3
x
2
,
-
(
r
2
-
x
3
2
)
)
,
where p=(x1,x2,x3)∈𝑺, and 𝔤 is a constant expressing the intensity of the gravitational field (considered uniform in this model).
As an application of Theorem 4.6 and Proposition 4.8, we obtain Proposition 5.2 below, yielding a multiplicity result for the T-periodic solutions of equation (5.2) (or, equivalently, (5.1)) when λ>0 is small. An analogous result for an RFDE (equation of the retarded spherical pendulum) has been obtained in [10, Theorem 3.1] (see also [16]). However, as pointed out in the introduction, Proposition 5.2 cannot be deduced as a direct consequence of [10, Theorem 3.1].
Proposition 5.2.
Let η, ν, g, λ and f be as above. Assume that g admits n-1 zeros, p1,…,pn-1, that are not T-resonant and such that
∑
i
=
1
n
-
1
i
(
g
,
p
i
)
≠
2
.
Then, for λ>0 sufficiently small, equation (5.2) (or (5.1), equivalently) admits n solutions of period T whose images are pairwise not coincident.
Proof.
The case η>0 follows immediately from Proposition 4.8 recalling that χ(𝑺)=2.
Consider the case η=0, that is consider on 𝑺 the equation
(5.3)
π
(
t
)
=
g
(
x
(
t
)
)
+
λ
f
(
t
,
x
(
t
)
,
(
t
)
,
x
(
t
-
τ
)
,
(
t
-
τ
)
)
that coincides with (5.2) when η=0.
Lemma 4.1 in [17] shows that any connected set of solutions in CT1(𝑺) of the second order equation (5.3) for λ=0 is unbounded. Thus, by Theorem 4.6 and Remark 4.7 (again recall that χ(𝑺)=2), we have that (5.3) admits at least n solutions of period T for λ>0 sufficiently small. The same argument of the second part of the proof of [10, Theorem 3.9] shows that the images of these solutions are pairwise not coincident. ∎
In the case when g represents the gravitational attraction, i.e., when g=g𝔤, there are two zeros of g𝔤 at the poles (0,0,r) and (0,0,-r), the first of which is necessarily not T-resonant, as it follows from (4.2). Hence, Proposition 5.2 yields the following consequence (compare [10, Theorem 3.2]).
Corollary 5.3.
When g=gg, for λ>0 sufficiently small, equation (5.1) admits at least two T-periodic solutions whose images are not coincident.
References
[1] P. Albers, U. Frauenfelder and F. Schlenk, An iterated graph construction and periodic orbits of Hamiltonian delay equations, preprint (2018), https://arxiv.org/abs/1802.07449v2. 10.1016/j.jde.2018.08.036Suche in Google Scholar
[2] P. Albers, U. Frauenfelder and F. Schlenk, What might a Hamiltonian delay equation be?, preprint (2018), https://arxiv.org/abs/1802.07453v2. Suche in Google Scholar
[3] B. Balachandran, T. Kalmár-Nagy and D. E. Gilsinn, Delay Differential Equations. Recent Advances and new Directions, Springer, New York, 2009. Suche in Google Scholar
[4] P. Benevieri, A. Calamai, M. Furi and M. P. Pera, Delay differential equations on manifolds and applications to motion problems for forced constrained systems, Z. Anal. Anwend. 28 (2009), no. 4, 451–474. 10.4171/ZAA/1393Suche in Google Scholar
[5] P. Benevieri, A. Calamai, M. Furi and M. P. Pera, On general properties of retarded functional differential equations on manifolds, Discrete Contin. Dyn. Syst. 33 (2013), no. 1, 27–46. 10.3934/dcds.2013.33.27Suche in Google Scholar
[6] P. Benevieri, A. Calamai, M. Furi and M. P. Pera, Global continuation of forced oscillations of retarded motion equations on manifolds, J. Fixed Point Theory Appl. 16 (2014), no. 1–2, 273–300. 10.1007/s11784-015-0215-6Suche in Google Scholar
[7] P. Benevieri, A. Calamai, M. Furi and M. P. Pera, On general properties of n-th order retarded functional differential equations, Rend. Istit. Mat. Univ. Trieste 49 (2017), 73–93. Suche in Google Scholar
[8] L. Bisconti and M. Spadini, About the notion of non-T-resonance and applications to topological multiplicity results for ODEs on differentiable manifolds, Math. Methods Appl. Sci. 38 (2015), no. 18, 4760–4773. 10.1002/mma.3390Suche in Google Scholar
[9] T. A. Burton, Stability and Periodic Solutions of Ordinary and Functional Differential Equations, Dover Publications, Mineola, 2005. Suche in Google Scholar
[10] A. Calamai, M. P. Pera and M. Spadini, Multiplicity of forced oscillations for the spherical pendulum acted on by a retarded periodic force, Nonlinear Anal. 151 (2017), 252–264. 10.1016/j.na.2016.12.006Suche in Google Scholar
[11] A. Calamai, M. P. Pera and M. Spadini, Multiplicity of forced oscillations for scalar retarded functional differential equations, Math. Methods Appl. Sci. 41 (2018), no. 5, 1944–1953. 10.1002/mma.4720Suche in Google Scholar
[12] J. Cronin, Differential Equations, 2nd ed., Monogr. Textb. Pure Appl. Math. 180, Marcel Dekker, New York, 1994. Suche in Google Scholar
[13] T. Erneux, Applied Delay Differential Equations, Surv. Tutor. Appl. Math. Sci. 3, Springer, New York, 2009. 10.1007/978-0-387-74372-1_8Suche in Google Scholar
[14] M. Furi, Second order differential equations on manifolds and forced oscillations, Topological Methods in Differential Equations and Inclusions (Montreal 1994), NATO Adv. Sci. Inst. Ser. C Math. Phys. Sci. 472, Kluwer Academic, Dordrecht (1995), 89–127. 10.1007/978-94-011-0339-8_3Suche in Google Scholar
[15] M. Furi, M. P. Pera and M. Spadini, Forced oscillations on manifolds and multiplicity results for periodically perturbed autonomous systems, J. Comput. Appl. Math. 113 (2000), no. 1–2, 241–254. 10.1016/S0377-0427(99)00259-9Suche in Google Scholar
[16] M. Furi, M. P. Pera and M. Spadini, Periodic solutions of retarded functional perturbations of autonomous differential equations on manifolds, Commun. Appl. Anal. 15 (2011), no. 2–4, 381–394. Suche in Google Scholar
[17] M. Furi and M. Spadini, Multiplicity of forced oscillations for the spherical pendulum, Topol. Methods Nonlinear Anal. 11 (1998), no. 1, 147–157. 10.12775/TMNA.1998.009Suche in Google Scholar
[18] M. Furi and M. Spadini, Branches of forced oscillations for periodically perturbed second order autonomous ODEs on manifolds, J. Differential Equations 154 (1999), no. 1, 96–106. 10.1006/jdeq.1998.3557Suche in Google Scholar
[19] M. Furi and M. Spadini, Periodic perturbations with delay of autonomous differential equations on manifolds, Adv. Nonlinear Stud. 9 (2009), no. 2, 263–276. 10.1515/ans-2009-0203Suche in Google Scholar
[20] V. Guillemin and A. Pollack, Differential Topology, Prentice-Hall, Englewood Cliffs, 1974. Suche in Google Scholar
[21] M. W. Hirsch, Differential Topology, Grad. Texts in Math. 33, Springer, New York, 1976. 10.1007/978-1-4684-9449-5Suche in Google Scholar
[22] M. Lewicka and M. Spadini, On the genericity of the multiplicity results for forced oscillations on compact manifolds, NoDEA Nonlinear Differential Equations Appl. 6 (1999), no. 4, 357–369. 10.1007/s000300050008Suche in Google Scholar
[23] J. Mawhin, Seventy-five years of global analysis around the forced pendulum equation, Proceedings of Equadiff 9, Masaryk University, Brno (1998), 115–145. Suche in Google Scholar
[24] J. Mawhin, Global results for the forced pendulum equation, Handbook of Differential Equations, Elsevier/North-Holland, Amsterdam (2004), 533–589. 10.1016/S1874-5725(00)80008-5Suche in Google Scholar
[25] J. W. Milnor, Topology from the Differentiable Viewpoint, The University Press of Virginia, Charlottesville, 1965. Suche in Google Scholar
[26] P. H. Rabinowitz, Some global results for nonlinear eigenvalue problems, J. Funct. Anal. 7 (1971), 487–513. 10.1016/0022-1236(71)90030-9Suche in Google Scholar
