On the sign of the Green’s function associated to

Hill’s equation with an indefinite potential∗

Alberto Cabada1 and José Ángel Cid2

1. Departamento de Análise Matemática, Facultade de Matemáticas,

Campus Sur, Universidade de Santiago de Compostela,
15782 Santiago de Compostela, Spain.
E-mail: [email protected]

2. Departamento de Matemáticas, Universidad de Jaén,

Campus Las Lagunillas, Ed. B3, 23071, Jaén, Spain.
E-mail: [email protected]
Abstract
p
In this note we give a L – criterium for the positiveness of the
Green’s function of the periodic boundary value problem

x00 + a(t) x = 0, x(0) = x(T ), x0 (0) = x0 (T ),

with and indefinite potential a(t). Moreover we prove that such Green’s
function is negative provided a(t) belongs to the image of a suitable
periodic Ricatti type operator.

Keywords. Green’s function; anti – maximum principle; Hill’s equation.

1 Introduction
Let us say that the linear problem

x00 + a(t) x = 0, x(0) = x(T ), x0 (0) = x0 (T ), (1.1)

∗
Partially supported by Ministerio de Educación y Ciencia, Spain, project MTM2007-
61724, and by Xunta de Galicia, Spain, project PGIDIT06PXIB207023PR.

1
is nonresonant when its unique solution is the trivial one. It is well known
that if (1.1) is nonresonant then, provided that h is a L1 – function, the
Fredholm’s alternative theorem implies that the non homogeneous problem

x00 + a(t) x = h(t), a. e. t ∈ [0, T ]; x(0) = x(T ), x0 (0) = x0 (T ),

always has a unique solution which, moreover, can be written as

Z T
x(t) = G(t, s)h(s)ds,
0

where G(t, s) is the Green’s function related to (1.1).

In recent years the condition

(H) Problem (1.1) is nonresonant and the corresponding Green’s function

G(t, s) is positive (nonnegative) on [0, T ] × [0, T ],

has become an standard assumption in the searching for positive solutions

of singular second order equations and systems (see for instance [3, 4, 5, 6,
12]). Moreover the positiveness of Green’s function implies that an anti –
maximum principle holds, which is a fundamental tool in the development
of the monotone iterative technique (see [2, 13]).
When a(t) ≡ k 2 condition (H) is equivalent to 0 < k 2 < (≤)λ1 :=
(π/T )2 , where λ1 is the first eigenvalue of the homogeneous equation x00 +
k 2 x = 0 with Dirichlet boundary conditions x(0) = 0 = x(T ).
For a non-constant function a(t) the best condition available in the liter-
ature implying (H) is a Lp – criterium proved in [10] (and based in an anti –
maximum principle given in [13]). For the sake of completeness let us recall
such result: define K(α, T ) as the best Sobolev constant in the inequality

Ckuk2α ≤ ku0 k22 for all u ∈ H01 (0, T ),

given explicitly by (see [9])

 ³ ´1−2/α ³ ´2

 2π 2 Γ(1/α)
αT 1+2/α 2+α Γ(1/2+1/α) , if 1 ≤ α < ∞,
K(α, T ) = (1.2)

 4
T, if α = ∞.

Through the paper a  0 means that a ∈ L1 (0, T ), a(t) ≥ 0 for a.a.

t ∈ [0, T ] and kak1 > 0, moreover a+ = max{a, 0} is the positive part of a
1 1
and for 1 ≤ p ≤ ∞ we denote by p∗ its conjugate (that is, + ∗ = 1). Now
p p
[10, Corollary 2.3] reads as follows.

2
Theorem 1.1. Assume that a ∈ Lp (0, T ) for some 1 ≤ p ≤ ∞, a  0 and
moreover
kakp < (≤)K(2p∗ , T ).
Then condition (H) holds.
Our main goal is to improve Theorem 1.1 by allowing a(t) to change
sign.
Z
In particular instead of a  0 we impose an integral condition, namely
T
a(t) > 0, which doesn’t prevent a(t) to be negative in a set of positive
0
measure. As far as we are aware this is the first anti – maximum principle for
problem (1.1) with an indefinite potential a(t) (compare with the previous
results obtained in [1, 10]). Moreover we notice that an improvement of
Theorem 1.1 immediately extends the applicability of those results available
on the literature which rely on condition (H) as, for instance, the validity
of the monotone iterative methods [13], or the existence of constant sign
periodic solutions for regular [10, 11], strong singular [4, 6] and weak singular
[3, 4, 5, 6, 12] second order boundary value problems.
This paper is organized as follows: in section 2 we present some known
results about the Dirichlet, periodic and anti – periodic eigenvalues of equa-
tion
x00 + (λ + a(t)) x = 0,
which are needed on section 3 to prove the positivity of the Green’s
Z
function
T
of (1.1) with an indefinite potential. In section 4, provided that a(t) dt <
0
0, we give a sufficient condition that ensures that the Green’s function related
to problem (1.1) is negative. Finally, in section 5, we conclude our paper
with some remarks referred to the general operator x00 + c(t) x0 + a(t) x with
c a L1 – function with mean value equals to zero.

2 Preliminaries
In this section we collect some known results (see [8]) for the eigenvalue
problem
x00 + (λ + a(t)) x = 0, (2.1)
where a ∈ L1 (0, T ), subject to periodic

x(0) = x(T ), x0 (0) = x0 (T ), (2.2)

anti – periodic
x(0) = −x(T ), x0 (0) = −x0 (T ), (2.3)

3
or Dirichlet boundary conditions

x(0) = 0 = x(T ). (2.4)

With respect to the periodic and anti – periodic eigenvalues there exist
sequences

λ0 (a) < λ1 (a) ≤ λ1 (a) < λ2 (a) ≤ λ2 (a) < . . . < λk (a) ≤ λk (a) < . . . (2.5)

such that
(i) λ is an eigenvalue of (2.1) – (2.2) if and only if λ = λk (a) or λk (a) for
k even.

(ii) λ is an eigenvalue of (2.1) – (2.3) if and only if λ = λk (a) or λk (a) for

k odd.
On the other hand the Dirichlet problem (2.1) – (2.4) has a sequence of
eigenvalues
λD D D
1 (a) < λ2 (a) < . . . < λk (a) < . . . ,

and the periodic and anti – periodic eigenvalues can be realized for k =
1, 2, . . . as

λk (a) = min{λD
k (as ) : s ∈ R}, λk (a) = max{λD
k (as ) : s ∈ R},

where as (t) := a(t + s) are translations.

In [15, Theorem 4] Zhang and Li established the following lower bound
for the first Dirichlet eigenvalue λD α
1 (a) in terms of the L -norm of a+ .

Theorem 2.1. Assume that a ∈ Lp (0, T ) for some 1 ≤ p ≤ ∞. If

ka+ kp ≤ K(2p∗ , T ),

where K is given by (1.2), then

µ ¶2 µ ¶
π ka+ kp
λD
1 (a) ≥ 1− ≥ 0.
T K(2p∗ , T )

Note that, since λ1 (a) = λD1 (as0 ) for some s0 ∈ R and, by considering
the T – periodic extension of the function a it is satisfied that k(as0 )+ kp =
k(a+ )kp , then under the assumptions of Theorem 2.1 we have
µ ¶2 µ ¶
π ka+ kp
λ1 (a) = λD
1 (as0 ) ≥ 1− ≥ 0. (2.6)
T K(2p∗ , T )

4
3 Positivity of the Green’s function
Firstly, we are going to give a sufficient condition for problem (1.1) to be
nonresonant which is equivalent to the existence of Green’s function.
Z T
Theorem 3.1. Assume that a ∈ Lp (0, T ) for some 1 ≤ p ≤ ∞, a(t)dt >
0
0 and moreover
ka+ kp ≤ K(2p∗ , T ).
Then problem (1.1) is nonresonant.
Proof. It is known (see [8]) that
Z T
λ0 (a) ≤ −1/T a(t)dt < 0.
0

On the other hand, since a satisfies the assumptions of Theorem 2.1, from
(2.6) it follows that λ1 (a) ≥ 0. Therefore, (2.5) implies that

λ0 (a) < 0 ≤ λ1 (a) < λ2 (a) ≤ λ2 (a),

which means that λ = 0 is not an eigenvalue of problem (2.1) – (2.2). u

t

Before to present our main result we need the following auxiliary result
(see [10, Theorem 2.1])
Lemma 3.1. Assume that (1.1) is nonresonant and that the distance between
two consecutive zeroes of a nontrivial solution of

x00 + a(t) x = 0

is strictly greater that T . Then the Green’s function G(t, s) doesn’t vanish
(and therefore has constant sign).
Now we are going to give a sufficient condition ensuring the positiveness
of the Green’s function of (1.1) with an indefinite potential a(t). To the best
of our knowledge this result is achieved for the first time for a non constant
sign potential a(t).
Z T
Theorem 3.2. Assume that a ∈ Lp (0, T ) for some 1 ≤ p ≤ ∞, a(t)dt >
0
0 and moreover
ka+ kp < K(2p∗ , T ).
Then G(t, s) > 0 for all (t, s) ∈ [0, T ] × [0, T ].

5
Proof. Claim.- The distance between two consecutive zeroes of a nontrivial
solution of x00 + a(t) x = 0 is strictly greater that T .
To the contrary assume that x is a nontrivial solution of the Dirichlet
problem
x00 (t) + ã(t) x(t) = 0, t ∈ [t1 , t2 ], x(t1 ) = 0 = x(t2 ), (3.1)
where 0 < t2 − t1 ≤ T and ã is the restriction of function a to the interval
[t1 , t2 ]. It is clear, from expression (1.2), that for any α fixed, the expression
K(α, T ) is strictly decreasing in T > 0. As consequence, since 0 < t2 − t1 ≤
T , we deduce the following properties:
kã+ kp ≤ ka+ kp < K(2p∗ , T ) ≤ K(2p∗ , t2 − t1 ).
From Theorem 2.1 it follows that

λD
1 (ã) > 0,

which contradicts that (3.1) has a nontrivial solution.

Now, Lemma 3.1 and the claim imply that G(t, s) doesn’t vanish. To
determinate its sign consider the periodic problem
x00 (t) + a(t) x(t) = 1, x(0) = x(T ), x0 (0) = x0 (T ).
It is clear that its unique solution is given by the expression
Z T
x(t) = G(t, s)ds.
0

Obviously x doesn’t vanish and has the same sign as G. Then, dividing
the equation by x and integrating over [0, T ] we obtain
Z T µ 0 ¶2 Z T Z T
x (t) dt
0< dt + a(t)dt = .
0 x(t) 0 0 x(t)
Hence x(t) > 0 on [0, T ] which implies G(t, s) > 0 on [0, T ] × [0, T ]. u
t

Example 3.1. As a direct consequence of the previous result, we deduce

that for any c > 0 and h ∈ L1 (0, 2 π), if k(c + b cos t)+ kp < K(2p∗ , 2 π), then
the following equation

x00 (t) + (c + b cos t) x(t) = h(t),

has a unique 2 π – periodic solution. Moreover, if h has constant sign then
x(t) h(t) ≥ 0 for all t ∈ [0, 2 π].

6
4 Negativeness of the Green function
When a ≺ 0 it is known that G(t, s) < 0. In this section we present a
sufficient condition that ensures us the negativeness of the Green’s function
even in the case of a(t) changes sign. As far as the authors are aware this is
the first result in this direction for an indefinite potential a(t).

Theorem 4.1. Assume that a ∈ L1 (0, T ) is of the form

a(t) = b0 (t) − b2 (t), b(0) = b(T ),

Z T
where b is an absolutely continuous function such that b(s)ds 6= 0.
0
Then G(t, s) < 0 for all (t, s) ∈ [0, T ] × [0, T ].

Proof. The key idea is to decompose the second order operator

L x = x00 + a(t) x,

as two first order operators L = L1 ◦ L2 , where

L1 x = x0 − b(t) x and L2 x = x0 + b(t) x.

The following claim is easily proved by direct integration.

Claim.- The problem x0 + b(t)x = h, x(0) = x(T ), has a unique solution for
Z T
all h ∈ L1 (0, T ) if and only if b(s)ds 6= 0. Moreover if h  (≺)0 then
Z T 0

x(t) b(s)ds > (<)0 for all t ∈ [0, T ].

0
Z T
Now, suppose that b(s)ds > 0 (the other case being analogous). If
0
Lx  0, x(0) = x(T ), x0 (0) = x0 (T ) then L1 (L2 x)  0 with L2 x(0) = L2 x(T )
and from the claim it follows that L2 x < 0. Now the claim implies again that
x < 0. This fact is equivalent to the negativeness of the Green’s function,
concluding the proof. u
t

Remark 4.1. Note that the previous result extends for the non constant
potential a(t) the classical one in which a(t) ≡ a < 0 is a strictly nega-
tive
Z
constant. Moreover we remark that assumptions of Theorem 4.1 imply
T
a(t)dt < 0.
0

7
Example 4.1. As a direct consequence of the previous result, we deduce
that for any n ∈ N, c ∈ R\{0} and h ∈ L1 (0, 2 π), the following equation

x00 (t) + (±n sin n t − (∓ cos n t + c)2 ) x(t) = h(t),

has a unique 2 π – periodic solution. Moreover, if h has constant sign then

x has the opposite one.

5 The general second order operator

In this section we extend Theorems 3.2 and 4.1 to the general second order
equation

u00 + c(t) u0 + a(t) u = h(t), u(0) = u(T ), u0 (0) = u0 (T ), (5.1)

where c ∈ L1 (0, T ) is a damping time – dependent coefficient with mean

Z T
value zero, i.e. c(s)ds = 0.
0
Let us define for all t ∈ [0, T ] the functions
Rt Z t
c(s)ds ds
ρ(t) = e 0 and w(t) = ,
0 ρ(s)

and denote R = w(T ).

Thus we obtain the following result.
Z T
Theorem 5.1. Assume that c ∈ L1 (0, T ) with c(s) ds = 0, a ∈ Lp (0, T )
Z T 0

for some 1 ≤ p ≤ ∞, ρ(t) a(t) dt > 0 and moreover

0
2p−1
kρ p a+ kLp [0,T ] < K(2p∗ , R).

Then G(t, s) > 0 for all (t, s) ∈ [0, T ]×[0, T ], where G is the Green’s function
of problem (5.1).

Proof. Making the change of variables x(r) = u(w−1 (r)) for all r ∈ [0, R] we
have that
x0 (r) = u0 (w−1 (r))ρ(w−1 (r)),
and

x00 (r) = u00 (w−1 (r))ρ2 (w−1 (r)) + u0 (w−1 (r))ρ2 (w−1 (r))c(w−1 (r)).

8
Thus, it is easy to check that if u is a solution of problem (5.1) then x(r) =
u(w−1 (r)) is a solution of problem

x00 (r) + ρ2 (w−1 (r))a(w−1 (r))x(r) = ρ2 (w−1 (r))h(w−1 (r)), r ∈ [0, R], (5.2)

x(0) = x(R), x0 (0) = x0 (R), (5.3)

and reciprocally, if x is a solution of (5.2) – (5.3) then u(t) = x(w(t)) is a
solution of (5.1).
On the other hand, the linear left-hand side of equation (5.2) is a Hill’s
equation of the form x00 (r) + ã(r) x(r) with

ã(r) = ρ2 (w−1 (r)) a(w−1 (r)).

From our assumptions it follows that

Z R Z T
ã(r)dr = ρ(s) a(s)ds > 0,
0 0

and 2p−1
kã+ kLp [0,R] = kρ p a+ kLp [0,T ] < K(2p∗ , R).
Therefore we can apply Theorem 3.2 to ensure that G̃(r, s) > 0 for all
(r, s) ∈ [0, R] × [0, R], where G̃ is the Green’s function related to problem
(5.2)-(5.3), and we also know that its unique solution is given by
Z R
x(r) = G̃(r, s) ρ2 (w−1 (s)) h(w−1 (s)) ds, for all r ∈ [0, R].
0

Thus the unique solution of (5.1) under our assumptions is given by

Z T
u(t) = x(w(t)) = G̃(w(t), w(s)) ρ(s) h(s) ds, for all t ∈ [0, T ].
0

This last equation implies that the Green’s function related to problem (5.1)
is equals to

G(t, s) = G̃(w(t), w(s)) ρ(s) for all (t, s) ∈ [0, T ] × [0, T ],

and hence G(t, s) > 0 for all (t, s) ∈ [0, T ] × [0, T ]. u

t

Remark 5.1. In Theorem 5.1 we have used

R a change of variables different
1 t
from the standard one, namely u(t) = e− 2 0 c(s)ds x(t), since it allows us to
impose less restrictive conditions over the function c(t).

9
 Z T
Although the assumption c(s)ds = 0 does not seem to have a physical
0
meaning, from the mathematical point of view our result complements [11,
Corollary 2.5], where the author established a Lp – maximum principle for
problem (5.1) with a constant positive coefficient c(t) ≡ c > 0. Moreover it
gives additional information to the one proved in [1] for the general operator
of second order coupled with different kinds of boundary conditions. There
two cases were considered: a < 0 or a positive and bounded with c bounded.
A related maximum principle for the general second order operator (5.1),
with a damped coefficient c(t) without necessarily mean value zero, was
proved in [14] and used in the recent paper [7] to prove the existence of a
periodic solution for a differential equation with a weak singularity.

In an similar way to the previous result, we deduce the following one as

a direct consequence of Theorem 4.1.
Z T
Theorem 5.2. Assume that c ∈ L1 (0, T ) with c(s)ds = 0, a ∈ L1 (0, T )
0
satisfies
ρ2 (t) a(t) = ρ(t) b0 (t) − b2 (t), b(0) = b(T ),
Z T
b(s)
with b an absolutely continuous function such that ds 6= 0.
ρ(s)
0
Then G(t, s) < 0 for all (t, s) ∈ [0, T ] × [0, T ], where G is the Green’s
function of problem (5.1).

Acknowledgments: We would like to thank the anonymous referee for

his/her valuable comments which improve a former version of this paper.

References
[1] I. V. Barteneva, A. Cabada and A. O. Ignatyev, Maximum and anti
– maximum principles for the general operator of second order with
variable coefficients, Appl. Math. Comput. 134 (2003) 173–184.

[2] A. Cabada, P. Habets and S. Lois, Monotone method for the Neumann
problem with lower and upper solutions in the reverse order, Appl.
Math. Comput. 117 (2001) 1–14.

[3] J. Chu and P. J. Torres, Applications of Schauder’s fixed point theorem

to singular differential equations, Bull. Lond. Math. Soc. 39 (2007) 653–
660.

10
 [4] J. Chu, P. J. Torres and M. Zhang, Periodic solutions of second order
non-autonomous singular dynamical systems, J. Differential Equations
239 (2007) 196–212.

[5] D. Franco and P. J. Torres, Periodic solutions of singular systems with-

out the strong force condition, Proc. Amer. Math. Soc. 136 (2008) 1229–
1236.

[6] D. Jiang, J. Chu and M. Zhang, Multiplicity of positive periodic solu-

tions to superlinear repulsive singular equations, J. Differential Equa-
tions 211 (2005) 282–302.

[7] X. Li, Z. Zhang, Periodic solutions for damped differential equa-

tions with a weak repulsive singularity, Nonlinear Analysis (2008),
doi:10.1016/j.na.2008.03.023.

[8] W. Magnus and S. Winkler, Hill’s equation (Corrected reprint of the

1966 edition, Dover Publications, Inc., New York, 1979).

[9] G. Talenti, Best constant in Sobolev inequality, Ann. Mat. Pura Appl.
4 (1976) 353–372.

[10] P. J. Torres, Existence of one-signed periodic solutions of some second-

order differential equations via a Krasnoselskii fixed point theorem, J.
Differential Equations 190 (2003) 643–662.

[11] P. J. Torres, Existence and stability of periodic solutions of a Duff-

ing equation by using a new maximum principle, Mediterr. J. Math. 1
(2004) 479-486.

[12] P. J. Torres, Weak singularities may help periodic solutions to exist, J.

Differential Equations 232 (2007) 277–284.

[13] P. J. Torres and M. Zhang, A monotone iterative scheme for a nonlinear

second order equation based on a generalized anti – maximum principle,
Math. Nachr. 251 (2003) 101–107.

[14] Y. Wang, H. Lian and W. Ge, Periodic solutions for a second order
nonlinear functional differential equation, Appl. Math. Lett. 20 (2007)
110115.

[15] M. Zhang and W. Li, A Lyapunov-type stability criterion using Lα

norms, Proc. Amer. Math. Soc. 130 (2002) 3325–3333.

11