Acta Mathematica Siniea, New Series 1987, Voi 3, No. 4, pp. 351--361 tt A Note on Nonlinear Oscillations at Resonance Pierpaolo Omari Dipartimento di Scienze Matematiche Uaiversi~ di Trieste 34127-Trieste Italy Fabio Zanolin Dipartimento di Svienze Matematiche UniversitZt di Trieste 34127-Trieste Italy Received February 9, 1987 Abotrac*t. Existence of 2n-periodic solutions to the equation ~ + g(x)=p(t) is proved, under sharp nonresonance conditions on the interaction of g(x)/x with two consecutive eigenvalues m 2 and (m + 1)2: touch with the eigenvalues is allowed. Introduction and the Main Result We are concerned with the solvability of the periodic boundary value problem (1) ~ + g(x) = p(t), (~ = dx/dt) (2) x(0) - a(2~) = ~(0) - ~(2~) = 0, where g: R--) R is continuous and p: [0, 2g]-* R is (Lebesgue) integrable. Solutions to (1)-(2) are intended in the Carathbodory sense and will be referred to as 2g- periodic. We investigate the existence of solutions to (1)-(2) assuming that g verifies,for some integer m > 1, the condition (3) m z <=g(x)/x <= (m + 1) z, for [xl ~ d > 0. This kind of study was initiated by Loud [16] in 1967 and, since then on, it has received attention from several authors like Lazer and Leach [13], I.e.ach [14], Fabry and Franchetti [10], Cesari [5], Mawhin [17], [18], Reissig [23], Ding [81 [9], Omari and Zanolin [19], Iannacci and Nkashama [12], Deng [7]. Since condition (3) is not sufficient, by itself, to ensure the existence of 2rt-periodic solutions to (1), for a general peLt(O, 2:t), as trivial linear counterexamples show, some additional restrictions have been put on the range of g(x)/x (and even on g'(x)), or on the function p. Here, in
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Acta Mathematica Siniea, New Series
1987, Voi 3, No. 4, pp. 351--361 tt A Note on Nonlinear Oscillations at Resonance
Pierpaolo Omari
Dipartimento di Scienze Matematiche
Uaiversi~ di Trieste
34127-Trieste
Italy
Fabio Zanolin
Dipartimento di Svienze Matematiche
UniversitZt di Trieste
34127-Trieste
Italy
Received February 9, 1987
Abotrac*t. Existence of 2n-periodic solutions to the equation ~ + g(x)=p(t) is proved, under sharp
nonresonance conditions on the interaction of g(x)/x with two consecutive eigenvalues m 2 and (m + 1)2:
touch with the eigenvalues is allowed.
Introduction and the Main Result
We are concerned with the solvability of the periodic boundary value problem
(1) ~ + g(x) = p(t) , (~ = dx/dt)
(2) x(0) - a(2~) = ~(0) - ~(2~) = 0,
where g: R - - ) R is continuous and p: [0, 2 g ] - * R is (Lebesgue) integrable.
Solutions to (1)-(2) are intended in the Carathbodory sense and will be referred to as 2 g -
periodic.
We investigate the existence of solutions to (1)-(2) assuming that g verifies,for some integer
m > 1, the condition
(3) m z <= g(x)/x <= (m + 1) z, for [xl ~ d > 0.
This kind of study was initiated by Loud [16] in 1967 and, since then on, it has received
attention from several authors like Lazer and Leach [13], I.e.ach [14], Fabry and Franchetti [10],
Cesari [5], Mawhin [17], [18], Reissig [23], Ding [81 [9], Omari and Zanolin [19], Iannacci and
Nkashama [12], Deng [7].
Since condition (3) is not sufficient, by itself, to ensure the existence of 2rt-periodic solutions
to (1), for a general peLt(O, 2:t), as trivial linear counterexamples show, some additional
restrictions have been put on the range of g(x)/x (and even on g'(x)), or on the function p. Here, in
352 P ie rpao lo O m a r i & F a b i o Zano l in
order to control the closeness of g(x)/x to the constants m 2 and (m + 1) 2, we introduce two functions
a• R - * R, with possibly lim a• = 0, such that
(4) m 2 + ~-(x) <= g(x)/x ~_ (m + 1) 2 - ~+(x), for I~1 >-- d.
Moreover, we define, for each integer k >-0, the spaces
P t : = fp~Ll(O, 2n): fi 'p(s)cos(ks)ds ffi fi'p(s)~in(~)as = 0 }.
Then the following result holds true.
Theorem. The periodic BVP (1)-(2) admits at least a solution i f (4) holds and one of the fol lowing conditions is ful f i l led
(i) p ~ P . ( 3 v .+l and ltm ;~(~) +o0,
(ii) p e Pm x2~-W'" = ltm " "-';~i=+(x)= + oo, a n d lim Is t - + oo ~ + z o
(iii) P~Pm+I and lim Ixla-(x)---- lira xZct+(x)= + m, lst-- + o0 Ixl-- + oo
(iv) p is arbitrary and lim Ixla• + oo. lai~ + oo
Remark 1. We point out that the condition
lira Ix l~+(x)= + ~ (resp. lira I x l ~ - ( x ) = + oo), 1 ~ + ~ M ~ + r
cannot be relaxed, if Pr (resp. pCPm). Indeed, a particular counterexample (already proposed in [8]) is provided by the differential equation
(5) ~ + ~(x)= ~(,),
with g(x): = (m + 1)2x - arctg x and p(t): = 4 cos((m + 1)t). Obviously, ~ satisfies (4), with a-(x): = 2m (a constant function) and rt+(x): = x -1 arctg x, for x r 0, ~+(0): = 1. Clearly,
lim xZct-(x)-- + oo, but lim Ixla+(x)--1t/2 < +oo.
Moreover, p ~ P~,, but p r Pm+ l- Therefore, condition (4)-(ii) of our theorem is fulfilled, except for
the requirement lim Ixla+(x) -- + oo, yet equation (5) has no 27t-periodic solution (see [8, Sect. Ixl~ 4-0o
III], or [13, Th. 2.11). Rema-k 2. We obvserve that our theorem constitutes a refinement of some results we proved
in [19, p. 147] and [20, p. 2921, where conditions (4)-(ii), (4)--(iii) and (4)-(iv) were permitted but (4)-(i) was not. However, we stress the fact that the statements in [191, [201, by themselves, are improvements both of the results in [171, [181, where (generalizing the existence theorem of Loud [16]) (4) with a+: positive constants is assumed, and of the results which can be deduced from [3],
A Note on Nonl inear Oscillations at Resonance 353
[2], [6], [12], where (on the lines of Lazer and Leach's paper [13]) it is essentially supposed that
(4) holds with lim [xl0t-(x) = + oo and 0t + a positive constant (resp. ~ - a positive constant and Ixl~+oo
lim [ x [ ~ + ( x ) = + co). Actually, in [3], [23, [6], [183, [123, g is assumed also to depend on the t -
variable, g: = g(t, x) and some kinds of non-uniform conditions (with respect to t) are considered.
However, they reduce to the ones just explained when g is independent of t. Furthermore, in [3],
[2], [63, [12], to control the interaction of g(x)/x with m z, the" so--called Landesman-Lazer
condition is assumed [131 yet it becomes the previous hypothesis on ct-, if one is concerned with
the solvability of (1)-(2) for every pr 2rt) (see also the final section).
As regards the paper [8], the author gives therein a rather complete treatment of problem
(1)-(2), under the condition
(6) m2<=g'(x)<(m+l) 2, for all x ~ R .
He proves that (6) is sufficient for the existence of 27r-periodic solutions to (1), i f p ~ P,, A Pro+ 1. Otherwise, he assumes, besides (6), condition (4)-(iv). It is clear that our result improves this one,
whenp r Pm N P,,+ 1, and is independent of it in the opposite case. Observe that anyhow we do not
even suppose the existence of g'. Finally, we remark that the recent result [7], when applied to the scalar equation (1),
essentially corresponds to our theorem, under condition (4)-(iv). Namely, in [7], the author on the
one side assumes a monotonicity condition on (m + 1)2x - g(x) we do not need, but on the other
t side he replaces the hypothesis lim Ixl~-(x) = + oo, with the weaker one lim Ixl- ' (g(u) I . ~ + Qo Ixl~ + Qo 0
-- rn2u)du = + oo.
Remark 3. As concerns the cases in which (3) is violated, we refer to [11], when lim g(x)/x
= + c~, to [223, when g(x)/x 5 0, for Ixl large, and to [21], when 0 _-< g(x)/x <-_ 1, for Ixl large. In the last two cases, the results in [22], [21] work as well for the (more general) Li6nard equation
(7) + Ax) + g(x) = p(t),
withf . 'R ~ R being an arbitrary continuous function. More precisely, in [21], we proved that the
periodic BVP (7)-(2) has at least one solution, for each p ~LI(0, 2rr), if there are two functions
such that
(8)
a + : R ~ R , with lim ~+(x) = 0, Ixl~ + oo
oC(x) <= g(x)/x =< 1 - ct+(x), for Ix[ > d > 0,
with lim Ixt0~-(x)= =lAin txi0t+(=)= + oo and x0~+(x) bounded for x _-> 0. Note that the
condition on 0c + allows g(x)/x to cross the eigenvalue 1 on the whole positive semiaxis. Finally, we
wish to observe that, combining the approach of this paper with some devices of [21], a previous
statement of Reissig [24], for problem (7)-(2), could be refined on the lines of our theorem, by
assuming (8) with 0t-: = 0 (a constant function) and lim x20t+(x)= + oo, if p ~Po 0 P1. Ixl--, + oo
354 Pierpaolo Omari & Fabio Zanolin
Proof of the Theorem. In what follows, we shall use the classical Banach spaces C~ 27t) and
i f (0 , 21t) (1 ~ p < ~ ) and the Sobolev spaces W2'P(0, 27t)(1 ~ p < oo), whose respective norms
will be indicated by I'lc o, I'1, P, I-IFr2.J" (see e.g. [1]). The L2-bilinear pairing will be denoted by (., %.
According to Remark 2, we prove the theorem only when (4)-(i) is assumed; for the other
cases we refer to [19] and [20J. Moreover, we replace condition (4)-(i) with the more general one:
there are two functions f l - : R ~ R + and fl+: R + ~ R + (with R + being the set of nonnegative
There is no restriction as to supposing also if+ to be non-decreasing functions. Let us fix a constant 0 < v < 2m + 1 and consider the equation (equivalent to (1))
(12) ~ + (m 2 + v)x = p(t) - g(x) + (m 2 + v)x.
Observe that, for each ueL l (0 , 2~), there exists exactly one function x = K u e Wa'I(0, 27t)
verifying (2) and
~i + (m 2 + v)x = u, for a.e. t z [0, 21t].
Moreover, define the Nemytzkii operator
N: C~ 2~) ~ L'(0, 2~), Nx : = p(') - d x ) + (m" + ,)x.
Then it is easily checked that the Hammerstein operator
KN: C~ 2~) ~ CO(0, 21t)
is completely continuous and its (possible) fixed points are the solutions of (12)-(2) in Wa'I(0, 2~).
Therefore, for solving (12)-(2), that is (1)-(2), we can apply the Leray-Schauder continuation
theorem [15] to the operator equation
x = KNx, x e C~ 2~).
Accordingly, we are looking for a constant R > 0 such that
(13) Ixlco < R,
for every function x satisfying, for some ),6 [0, lJ ,
for Ix] -_> d' and 4 ~ [0, 1"1. Finally, using the continuity of h: R x [0, 1"] ~ R and the boundedness
356 Pierpaolo Cmari & Fabio Zanolin
from below of fl+, relation (19) is obtained, for a suitable choice of C 1 > C2( > 0). Moreover, since fl+ is non-negative and (10) and (11) hold, it is possible to find a concave function 7+: R + --* R +, with 7+(0)= (0), such that
.
7+(~) =< fl+((2m + 1)-~) - M +, for every ~ ' R +, with M + > 0, (23)
(24)
and (25)
lim 7+(~)= + a3
lim ~-~7+(~) = 0.
Therefore, owing to (19) and (23), the function h verifies
where 0:--- ~ ~Pk , ~ : = Z (rn2--k2)-ltPk k=O k=O
k#m,m+l kC:m, m+l
,m2_k2,2) k~m, m+l
A Note on Nonlinear Oscillations at Resonance 357
Let us multiply both sides of equation (15~) by q~(') and integrate between 0 and 2=. Integration by parts (with the use of the boundary conditions (2)), the hypothesis p e P . , N P.,+I and the continuity of the L2-bilinear pairing give
(x + r~x, Ch = (~, q')z. = (x, h(x, 2))2 ,
(h(:, ~), r (~o, "h I~ (~ 1,2)-,,,_ = = - , ~ , L 2 k ~ O
kC:m
k = O
k ,~ m,m + l
Hence, by (28), HSlder inequality and 2 < 1, we find
(29) (~.hO,, ,~-)h-< - ~ (~-k~)-t{o,,,{[~ + ~-~Lt{'~I o k=O C
k#m
<= ~.(m z kZ) - , z - - [e.JL~ + "=~{LtI~iL2"
k=O
k~m
Now, compare (29) with
(x,h(x, 2))z _>_ (2m + 1)- ' ~ {(Pkl~z- 2n(C, + M+), k = O k ~ m
which follows from (26), by putting v = x and using }7+(]r 1 ~ 0. By easy calculations, w~ derive that
(4(., + 1)(2,. + 1))-'(2.~ + z)l~lh =< ~ ((2~ + a)-' + (m 2 - k2) -')l~0k{~, L k = O
k # m , m + l
< K[PILII~IL: + 2~(C1 + M+).
So that, there exists a constant C3 > 0 (independent of x and 2) such that
(30 ) {~{L2 <= C 3 .
Next, compare (29) with (26), where we put again t ' = x, By means of (30), we get
Hence, as :~+ is subadditive (being concave and non-negative), and, by (25), verifies 7+(~) _< {z + C5, for every ~ e R +, with C 5 > 0 being a suitable constant, we have
;: f: ~,+({~",~.,(s)i)es = ~,+Oo,.-~,(s)+ r ~,){~s
j:- f:- =< ~+({~0)0~s + ,,,+O~,(s)Oas
=< ;,+(lo,(~){)d~ + {~,(~){~ ,+ 2~C,,
358 Pierpaolo Cmari & Fabio l~nolin
and then, by (30) and (31),
c , .
Moreover, as ?+ verifies the condition
(32) ~+(~:) _~ #V+(~), for every ~ E a + and #~[0, 1], (a consequence of ),+ being concave and y+(0)= 0), we obtain
Moreover, since 1~- is non-negative and (10) and (11) hold, it is possible to find a function 7 R +, with 7-(~)= 0, for every ~ < 0, 7- being concave for ~ _-> 0, such that
(37) 7-(~) < / ~ - ( ~ ) - M - , for every ~ R +, with M - => 0,
- : R
A Note on Nonl inear Oscillations at Resonance 359
and
(38) lim ~ - ( ~ ) = + oo.
Then, form (37), (16), (34) and (35), we get, for some constant C12 > O,
Now let us suppose that Ixmlco > 2Ix - x,,Ico (otherwise, lXlc o < 3Ix - x,,,IcO < 3Clo). Then,
�9 (t) = x.(t) + 4 0 - x.(t) _>_ x.(t) - I x - X.,cO
_-> Ix,,lcO #xml x~,(t) - = Ixmlco s i n ( a t + 0) - ~ ,
for every t E I'0, 2zr]. Hence, using y - non-decreasing and the counterpart of (32), we obtain
[sin(,~ + 0) > 2 l
>f -{ ( = 1 y r o sin(ms + 0)-- _~ ds [sin(ms + O) > ~ ]
>=7-('X.lcO"f[ 1 . ( s i n ( m s + O ' - ~ ) d s = ( x / ~ - x / 4 " - ( l X . ' c O ' �9 sin(.~ + 0) > 2 ]
Hence, by (39), we get
~'-(Ix,,}cO ) _-< (.v/2 - at/4)- l(Clx + C,2).
Thus, by (38), we can find a constant C13 > 0 such that
and then, by (35),
lx,,,Ico < C13,
Ixlco < Clo + Cls : = C14,
with C14 > 0 independent of x and 2. Hence (13) follows for any R > C14. This concludes the
proof.
Final Remarks Let us consider the equation
+ g(t , ~ ) = p(t) ,
where an explicit dependence of g on the t-variable is assumed. In this context, we suppose that g :
I-0, 2rt] • R.---, R verifies the following (Carat6odory) conditions: g(t~') is continuous for a.e.
t ~ [0, 2rt], g(., x) is measurable for every x ~ R , and, for each r > 0, there is a function 3, e L2(0,
2r~) such that Ig(t, x)l < cS,(t), for a.e. t~ [0, 2~] and lxf < r. Then replacing g(x) with g(t, x) in condition (4), supposed to hold for a.e. t ~ [0, 2rt], all the conclusions stated in our theorem are still true.
Moreover, from the proof of the theorem (see step (9)), it is clear that condition (4)-0) can be
360 Pierpaolo Cmari & Fabio Zanolin
sharpened by repalacing the assumption
lim
I,d-* +cr with the weaker one
xZa-(x) = + oo (resp. lira x2~t+(x) = + oo)
r
o r
a- (x ) ~ O, for x < O, and lim x2ot-(x) = +
(resp. a+(x) >= O, for x =< O, and lira x2ct+(x) = + oo) X"* § r
(j') *t-(x) > 0, for x >= 0, and lira x2a-(x) = + oo x oo
(resp. ~t+(x)>= O, for x >-O, and lim xZa+(x )= + o r ) x
Conditions (j) and (j') say that At, x ) = m2x, on a x-semiaxis, is allowed.
Such a remark seems of interest, when g explicitly depends on t, in the light of an example
proposed in [121 which shows that, even if p ~ P . , for a nonlinear g, the condition (alone)
m2_-- < g (t, x)/x, for a.e. t~ [0, 27t] and every x # 0,
is not sufficient for the existence of 27t-periodic solutions. Note that g(t, x) = mZx, for x _-< 0, was
already considered in [4], but under further monotonicity conditions on g.
Extension of our theorem can be obtained for the differential-delay equation
+ =p(t),
with r ~ ]0, 2~r[ and x(t - T) defined as usual when dealing with the periodic problem for DDEs.
Such a result follows from some arguments proposed in [19], [20 1 together with a careful reading
of the above proof.
Finally, the existence of classical (i.e. of class C 2) 2~t-periodic solutions is provided by our
theorem, whenever p : R --* R and ~ : R x R ~ R are continuous and 27t-periodic in the t -
variable.
References
[ 1 ] Adams, B.A., Sobolev Spaces, Academic Press, New York, 1975. [ 2 ] Amann, H. Mancini, G., Some applications of monotone operator theory to resonance problems, J. Nonlinear
Anal., T.M.A., 3(1979), 815-830 [ 3 ] BtE'zis, H. Nirenber 8, L., Characterizations of the ranges of some nonlinear operators and applications to boundary
value problems, Ann. Scuola Norm. Sup. Pisa (Rend. CI. Sol.), 5(1978), 225-326.
I" 4 ] Caristi, G., Monotone perturbations of linear operators having nulbpace made of oscillating functions, J. Nonlinear
Anal., T.M.A., (to appear). [ 5 ] Cesarl, L., Functional analysis, nonlinear differential equations, and the alternative method, in "Nonlinear
Functional Analysis and Differential equations" (L. Cesari, R. Kannan and J. D. Schuur, ods.), Dekker, New York
(1976), 1-197.
A Note on Nonlinear Oscillations at Resonance 361
[ 6 ] de Figueiredo, D.G., Semilinear elliptic equations at resonance: higher eigenvalues and unbounded nonlinearities,
in "Recent Advances in Differential Equations" (R. Conti, ed.), Academic Press, New York (1981), 89-99.
[ 7 ] Deng, S., On the periodic solutions of Hamiltonian systems of second order, Acta Math. Siniea (New Series,),
I(1985), 319-326.
[ 8 ] Ding, T.R., Nonlinear oscillations at a point of resonance, Scientia Siniea (Series A), 25(1982), 918-931.
1-9] Ding, T.R., Unbounded perturbations of forced harmonic oscillations at resonance, Proc. Amer. Math. Soc.,
88(1983), 59-66.
1-10] Fabry, C. Franchetti, C., Nonlinear equations with growth restrictions on the nonlinear term, J. Differential
Equations, 20(1976), 2X3-291.
['11] Fucik, S., Lovicar, V., Periodic solutions of the equation x"(t) + g(x (t)) =p(t), ~as. P~st. Mat., 100(1975), 160-
175.
[12] Iannacci, R. Nkashama, M.N., Unbounded perturbations of forced second order ordinary differential equations at
resonance, 1985. (preprint).
1-13] Lazer, A.C. Leach D.E.', Bounded perturbations of forced harmonic oscillators at resonance, Ann. Mat. PuraAppl.,
82(1969), 49--68.
[14] Leach, D.E., On Poincar6's perturbation theorem and a theorem of W.S. Loud, J. Differential Equations 7(1970),
34-53.
1-15] Leray, J. Schauder, J., Topologie et 6quations fonctionnelles, Ann Sci. Ecole .Norm. Sup. Paris, 51(1934), 4,5-78.
[16] Loud, W.S., Periodic solutions of nonlinear differential equations of Duffing type, in "Differential and Functional
Equations" (W.A.Harris and Y. Sibuya, eds.), Benjamin, New York (1967), 199-224.
[17] Mawhin, J., Recent trends in nonlinear boundary value problems, in "Proc. VII Int. Conf. Nonlinear Oscillations
(Berlin)" (G. Schmidt, ed.), Akademie-Verlag, Berlin (1977), 1(2), 51-70.
[18] Mawhin, J., Compacit6, monotonie et convexit6 dans 1' 6tude des probl6mes aux limites semi-lin6aires, S6m. Anal.
Moderne 19, Universit6 de Sherbrooke, 1981.
[19] Omarin, P. Zanolin, F., Existence results for forced nonlinear periodic BVPs at resonance, Ann. Mat. Pura Appl.,
141(1985), 127-157.
[20] Omari, P. Zanolin, F., Boundary value problems for forced nonlinear equations at resonance, in "Ordinary and
Differential Equations" (B.D. Sleeman and R.J. Jarvis, eds.), L.N.M. 1151, Springer-Verlag, Berlin (1985), 285-
294.
[21] Omari, P. Zanolin, F., On the existence of periodic solutions of forced Li6nard differential equations, J. Nonlinear
Analysis, T.M.A., 10(1986). (to appear).
[22] Reissig, R., Extensions of some results concerning the generalized Li6nard equation, Ann. Mat. Pura Appl.,
104(1975), 269-281.
['23] Reissig, R., Contraction mappings and periodically perturbed noneonservative systems, Atti Accad. Naz. Lincei
(Rend. Cl. Sci.), 58(1975), 696-702.
[24] Reissig, R., Continua of pefiokic solutions of the Li6nard equation, in "Constructive Methods for Nonlinear
Boundary Value Problems and Nonlinear Oscillations" (J. Albrecht, L. Collatz and K. lGrchg~issner, eds.), I.S.N.M.