ANNALES SOCIETATIS MATHEMATICAE POLONAE Series I: COMMENTATIONES MATHEMATICAE XXI (1979) ROCZNIKI POLSKIEGO TOWARZYSTWA MATEMATYCZNEGO
Séria I: PRACE MATEMATYCZNE XXI (1979)
A. La da (Bydgoszcz)
A two-point boundary value problem
for systems of ordinary non-linear differential equations and the lemma of Visik
1. Introduction. The lemma of Visik [4], [5] has been used for proving the existence of solutions of quasi-linear partial differential equations boundary value problems (see [1], [4], [5]). The aim of the paper is to apply the lemma for solving the two-point boundary value problem. A sufficient condition for existence of solution is given. The condition has an integral inequality form, which is convenient for verifying in a special case.
2. Preliminaries. Denote by C1 [a, b] the set of continuously differenti
able functions on the interval [a ,b ]. We have
b
(1.1) max (|w(t)|: t e [ a ,b ] } ^ y jb — a ( { \u' (s)|2 ds) 1 / 2 + \u (a)|,
a
where u e С1 [а, b].
Let H l [a, b] denote the completion of the linear space of the functions C1 [a, b] = { u e C1 [a, b~\: и (a) = u'(a) = u{b) = u'(b) = 0} in the norm
ь b
\\u\\ = ( j u2(s)ds+ j \u'(s)\2ds)112.
a a
It is known [6] that H1 [a, b] is a separable Hilbert space and the norm || • ||
is equivalent to
IMIi = ( j \u'{s)\2ds)112.
a
There exists a countable, linearly dense in H1 [a, b] sequence {щ}
a C1 [a, b] orthonormal under the scalar product
b
(u, rh = j u'(s)u'(s)ds,
where the derivative is understood in the generalized sense. We denote
n
by Hi [a, b] the Cartesian product Y[ [<T h] with the norm
i = 1
П
INL 2 =
YINI?’
u= (мь •••’
un)-i = 1
n
So, there exists a sequence {w*} <= C1 [a, b] orthonormal under the
i = 1
scalar product
П (u, v)„ = Y N vi)l
i — 1
and linearly dense in Hl[a, b~\.
Now we recall some results of [6] and [3].
Pr o p o s i t i o n 2.1. Let и: [a, b] -> R be continuous. Suppose that и has the generalized derivative u' e L2 [a, b] and is equal to zero at a and b. Then
u eH1 [a, b].
Th e o r e m 2.1. A closed sphere in a Hilbert space is weakly sequentially compact.
The following is due to Visik :
Le m m a 2.1. Let A: Rp -> Rp be a non-linear continuous mapping such that <Л(х), x ) ^ M \x\1+d — К for all x that |x| ^ L, where M > О, К ^ 0, d > 0, L ^ 0 are constants. Then there exists ùt least one solution of the equation A(x) = h, where h e R p is a constant vector.
By <•,■>, M we denote, respectively, the scalar product and the Eucli
dean norm in the space Rp. A generalization of Lemma 2.1 is given in [1].
3. A two-point boundary value problem for systems of non-linear ordinary differential equations. Let f : Rn+1^ R n be a continuous mapping and P, Q e R n. We will consider a problem of existence of the mapping x: [0, 7 ] -> Rn and a constant vector ceR" such that
(3.1) x(t) = f ( t , x ( t ) ) + c, x(0 ) = P, x{T) = Q.
Th e o r e m 3.1. Let the mapping f satisfy the condition
(V) I ( f ( t , u ( t ) + T - ‘f + t 9 Ù( t ) \ dt «: K JkIH + Kj,
where u e H l [ 0 , T] and Xj > 0, K2 > 0, 0 ^ r < 2 are constants. Then there П
exists at least one solution (x0, c0)e Y[ C1 [0> x X" satisfing (3.1). Moreover,
i = 1
{ T - t ) P + tQ
--- г --- ’ ^°e
1 i= 1
* o(0 = Уо(0 + - У о е П ^ Ч о . г ]
Two-point boundary value problem 159
and
1 T
c0 = — - - J / ( s , x0(s))ds.
* о P roof. Note that
, ч , ( T - t ) P + t Q \ P - Q F(t, y{t)) = . / ( t , y { t ) + --- --- l+-
T T
satisfies condition (V) with P = Q = 0. Thus problem (3.1) is equivalent to (3.2) y ( t ) = F(t,.v(r)) + c, y(0) = y ( T) = 0.
If (y0, c0) is a solution of (3.2), then
( T - t ) P + tQ ко (0-1---J ---» c0
is the solution of (3.1). Take the set {wfc} described in Section 2. We put
y j(0 = Z akWk(t)
k= 1
and consider the system
(3.3) J ( ÿ j ( t ) ~ F {t, yj{t)), wk(t)} dt = 0, 1 ^ к ^ j , о
of non-linear equations in relation to aj = (al 5..., aj ) e Rj. By orthonormality of {wk|, this system has the form
(3.4) Tak- $ ( F ( t , y j ( t ) ) , wk( 0 ) dt = 0, 1 ^ к ^ j.
о
We apply the lemma of ViSik for proving the existence of zero of the mapping A: RJ- + Rj, where
M a j) = Tak— J < F( t , y j ( t )), wk( t ) )dt , A = {Au ..., Aj).
о
From condition (V) we conclude that
< З Й У > = T\a*\2 — j ( F ( t , yj(t)), ÿj(t)y dt > Т \а > \* -К Л у,\\'п -И2
0
/ IS rf rl2 \
= г й 2- к, г '2И ' -к2 = И г ( г — j ^ p - ) - к 2.
Thus, according to the inequality 2 — r > 0, condition (2.1) is satisfied, from what it follows that for each j ^ 1 there exists at least one solution aj0
of equation (3.3). Therefore for
j y°j = Z ak w
fc = l к
we have by (3.3)
(3.5) \\y]\\2n - (t, yj{t)), ÿj(t )} dt = 0, 1 < oo.
о
Condition (V) implies \\у*}\\% ^ К г \\y<j\\rn + K2 and therefore from 2 — r > 0 we conclude that the sequence {у®} is bounded:
(3.6) WIL < K 3, 1 ^ 7 < 00.
Thus from inequalities (1.1) and (3.6) it follows that {у®} is jointly bounded and identically continuous on the interval [О, Т]. By the Ascoli-Arzeli’s theorem exists a subsequence (y® J uniformly convergent to a continuous y0.
From Theorem 2.1 and inequality (3.6) it follows that exists a subsequence n
{ÿyks} c= (y?) weakly convergent in the space Y[ L2[0, Г ]. Thus the weak i= 1
limit is equal to y0 in the generalized sense. It is obvious that To(0)
= y0(T) = 0. Therefore, according to Proposition 2.1, y0eHj, [О, Т ].
Putting 7 = j ks in (3.3) and letting j -> oo, we obtain
T
(Уо, wk)n- J ( F ( t , y0(t)), wk{ t ) }dt = 0, 1 ^ к < go, о
and consequently
(3.7) $ ( F ( t , y0( t ) ) - ÿ0( t ) , v{ t ) } dt = 0 о
for all v e H„ [0, Г ]. We recall that
T
(3.8) j ( c, v(t)) dt = 0
о
for a v e H ^ l 0, Г] and a constant vector c e R n. Notice that y0 ( t ) - F ( t , y0(tj)
d dt
t
yo(t ) ~ J F ( s , y0(s))ds + о
t
~T
T
f F ( s , y0(s))ds
0
- zr$- F( s, y0{s))ds
1 0
and put it into (3.7). Making use of (3.8) we obtain
<M
i d Уо( 0~ J F(s, yo(s))ds' +— J F ( s , y0(s))ds0 ^ 0
, v (t)\ dt = 0
Two-point boundary value problem 161
for all v e H l [ 0 , Т ]. Observe that
y o( t ) +- ^r $ F (s, y0 (s)) d s - $ F (s, у о (s)) ds e Я* [0, Г]
' 0 0
what in view of (3.9) gives
1 r
ÿo(t) = F (t, v0(0) — — $ F ( s , y0{s))ds, te\_0, Т ].
1 0
n
Finally, y 0e П С1 [0, Г] by continuity of F and y 0. The proof is complete.
i= 1
One ought to find such T, P, Q that inequality (V) remains true in a special problem.
Co r o l l a r y 3.1. Assume that f is periodic in the variable t with the period T and satisfies conditions of Theorem 3.1 with P = Q. Then there exists at least one solution (x0, c0) o f problem (3.1) and x0 is periodic with the period T.
References
[1] Yu. A. D u b in s k ii, Quasi-linear elliptic and parabolic equations of arbitrary degree, Usp.
Mat. Nauk, t. 23, 1 (1968), p. 45-90 (Russian).
[2] R. F a u re , Sur l'application d'un théorème de point fixé a l'existence de solutions pério
diques, C. R. Acad. Sci. Paris, t. 283, Sér. A (1976), p. 1295-1298.
[3] K. M a u rin , Methods of Hilbert spaces, Warszawa 1967.
[4] M. I. V isik, Solution of a quasi-linear system of the divergent form equations, under periodic boundary conditions, Dokl. Acad. Nauk SSSR, t. 137, 3 (1961), p. 502-505 (Russian).
[5] — Quasi-linear strong elliptic systems o f divergent form differential equations, Trudy Mosc.
Mat. Ob-wa, 12 (1963), p. 125-184 (Russian).
[6] L. P. V ole vie, В. P. N a n e y a k h , Some generalized function spaces and the embeddings theorems, Usp. Mat. Nauk, t. 20, 1 (1965) p. 3-74 (Russian).
INSTITUTE OF MATHEMATICS AND PHYSICS
TECHNICAL AND AGRICULTURAL ACADEMY. BYDGOSZCZ
II — Prace Matematyczne 21.1