L 1 -maximal regularity for quasilinear second order differential equation with damped term
Kordan Ospanov
BL. N. Gumilyov Eurasian National University, 2 Satpaev Street, Astana, 010008, Kazakhstan Received 7 March 2015, appeared 13 July 2015
Communicated by Michal Feˇckan
Abstract. We investigated a quasilinear second order equation with damped term on the real axis. We gave some suitable conditions for existence of the L1-maximal regular solutions of this equation.
Keywords: differential equation, nonlinear damped term, separation of operator, fixed point theorem.
2010 Mathematics Subject Classification: 34A34.
1 Introduction and statement of main result
LetR:= (−∞,+∞),L1:= L1(R)andk · k1 be the norm ofL1. We consider the equation Ly:= −y00+r(x,y)y0+q(x,y)y= f(x), x∈R, (1.1) where ris continuously differentiable andqis a continuous function, f ∈ L1. This is a useful equation in mathematical physics (see [5,26]).
ByC0(k)(R)(k=1, 2, . . .) we denote the set ofktimes continuously differentiable functions with compact support. Let Cloc(j)(R) =y :ψy∈C(0j)(R),∀ψ∈C(0∞)(R) (j=1, 2).
Definition 1.1. Lety ∈L1, if there is a sequence {yn}∞n=1 ⊂Cloc(2)(R)such that
nlim→∞kψ(yn−y)k1=0 and lim
n→∞kψ(Lyn− f)k1 =0,
∀ψ∈C0(∞)(R). Thenyis called a solution of (1.1).
The purpose of this work is to find some conditions forr andqsuch that for every f ∈L1, the equation (1.1) has a solutionywhich satisfies
ky00k1+kr(·,y)y0k1+kq(·,y)yk1 <∞.
BEmail: ospanov_kn@enu.kz
The separability of differential operators introduced by Everitt and Giertz in [7, 8] plays an important role in the study of second order differential equations. Recall that the Sturm–
Liouville operator
eLy:=−y00+q1(x)y,
acting inL2(R)is separable, if there is a constant c>0 such that
k −y00k2+kq1(·)yk2 ≤c(keLyk2+kyk2), ∀ y∈ D(eL).
Everitt and Giertz [7,8] proved that if q1 and its derivatives satisfy some conditions, then eL is separable in L2(R). In the case q1 is not differentiable function, the separability of eL in L2(R) was discussed in [3, 23]. In [9], Everitt, Giertz and Weidmann give an example of non-separable Sturm–Liouville operator in L2(R) with strongly oscillating and infinitely smooth coefficientq1. The separability of linear partial differential operators was studied in [4,15,17,21,24]. Some sufficient conditions of separability of operators on Riemann manifolds are obtained in [1,2,12,13].
The separability is also an important tool when dealing with quasilinear equations. In [16], Muratbekov and Otelbaev used the separability to discuss the solvability of the nonlinear equation
−y00+q0(x,y)y= f(x), (1.2) where f ∈ L2(R). Grinshpun and Otelbaev showed that the solvability of the equation (1.2) in L1 implies q0 ≥ 1 (see [6]). This method is useful for the multidimensional (Schrödinger) equation−∆u+q(x,u)u= F(x), x ∈Rn(see [17,20] for details).
In general, the expression (1.1) can be converted neither to (1.2) nor to the form
− p2(x,y)y00
+q2(x,y)y= f(x).
In [22], we considered the equation−y00+r(x,y)y0 = f(x), f ∈ L2(R), and found some conditions forr such that this equation is solvable. In the present paper, we discuss the more general equation (1.1), in the case f ∈ L1. Under weaker conditions on r than in [22] the existence and regularity of solutions of (1.1) are established.
Schauder’s fixed-point theorem is used to prove our main result (see [10]).
Let gandh be some functions onRand let αg,h(t) =
Z t
0
|g(η)|dηess sup
ξ∈(t,+∞)
|h(ξ)|−1, t>0, βg,h(τ) =
Z 0
τ
|g(η)|dη ess sup
ξ∈(−∞,τ)
|h(ξ)|−1, τ<0, γg,h=max
sup
t>0
αg,h(t), sup
τ<0
βg,h(τ)
. The main result of this paper is the following.
Theorem 1.2. Let r be a continuously differentiable function and q a continuous function satisfying r≥δ1
p1+x2 (δ1 >0) (1.3)
and
sup
c0∈R
γq(·,c0),r(·,c0)< ∞. (1.4)
Then for any f ∈ L1, the equation(1.1)has a solution y such that
ky00k1+kr(·,y)y0k1+kq(·,y)yk1 <∞. (1.5) Example 1.3. Letr =10+x10+5y4,q=x3+cos4x+2y. Thenr andqsatisfy the conditions of Theorem1.2.
2 Auxiliary statements
By Muckenhoupt’s theorem (Theorem 2 in [14]), we obtain the following lemma.
Lemma 2.1. Let g and h be continuous functions onRsuch thatγg,h< ∞. Then
Z
R|g(x)y(x)|dx ≤γg,h Z
R|h(x)y0(x)|dx, ∀y∈C0(1)(R). (2.1) Moreover,γg,his the smallest constant which satisfies(2.1).
Letrbe a continuously differentiable function and
l0y =−y00+r(x)y0, D(l0) =C0(2)(R). Denote bylthe closure ofl0 in L1.
Lemma 2.2. Let r be continuously differentiable and satisfy(1.3). Then l is invertible, R(l) = L1and ky00k1+kry0k1 ≤3klyk1, ∀y∈ D(l). (2.2) Proof. We use the method of [25] to prove (2.2). Let y ∈ C(02)(R) be a real function, and γ> −1. Then by integration by parts, we get
Z
R(ly)y0
(y0)2γ/2 dx =
Z
Rr
(y0)2γ/2+1 dx.
By Hölder’s inequality, we have that Z
Rr
(y0)2γ/2+1 dx≤ Z
R|r−αly|pdx 1/pZ
R|rα(y0)γ+1|qdx 1/q
, (2.3)
where 1 < p < ∞, q = p/(p−1). We choose αandγ as follows: (γ+1)q = γ+2, αq= 1.
This implies γ+2 = p. So from (2.3) it follows that kr1/py0kp ≤ kr−1/qlykp. Taking the limit p →1, we get
kry0k1≤ klyk1, y∈C0(2)(R). (2.4) Sinceky00k1≤ klyk1+kry0k1 ≤2klyk1, we haveky00k1+kry0k1 ≤3klyk1,∀y∈C0(2)(R). Since lis a closed operator, the last inequality holds for anyy∈ D(l).
From Lemma2.1, (1.3) and (2.4) follows that
kyk1≤c1klyk1, y∈ D(l). (2.5) So the inversel−1 oflexists.
Next, we show thatR(l) =L1. Let R(l)6=L1. Sincelis closed, (2.5) impliesR(l)is closed.
Hence there exists a nonzero elementz0 ∈ ⊥R(l)such thatl∗z0 =−z000−(r(x)z0)0 =0, where l∗ is adjoint operator ofl. Then
z0exp
Z x
a r(η)dη 0
=c2exp Z x
a r(η)dη
. Letc26=0. Without loss of generality, we can assume that c2=−1. Then
z0exp
Z x
a r(η)dη 0
<0,
i.e. z0(x) is a monotonically decreasing function, moreoverz0(x−k) > exp(k)z0(x), x ∈ R, k=1, 2, . . ., which implies thatz0 ∈/ L∞(R).
Letc2=0. Then
z0=c3exp
−
Z x
a r(η)dη
, c3 6=0.
Thereforez0 ∈/ L∞(R). We obtain a contradiction, hence R(l) =L1. We consider the following linear equation
ly:=−y00+r1(x)y0+q1(x)y= f(x), x∈R. (2.6) The functiony ∈ L1 is called a solution of (2.6), if there exists a sequence{yn}∞n=1 ⊂ C0(2)(R) such thatkyn−yk1→0 andklyn− fk1 →0 asn→∞.
Lemma 2.3. Let r1 be a continuously differentiable function such that r1 ≥δ1
√1+x2.Assume q1is a continuous function andγq1,r1 <∞. Then for every f ∈ L1, the equation(2.6)has a unique solution y such that
ky00k1+kr1y0k1+kq1yk1≤ c4kfk1, (2.7) where c4depends only onγq1,r1.
Proof. Letx =at(a>0), then (2.6) becomes that
−y˜00+a−1r˜1(t)y˜0+a−2q˜1(t)y˜= f˜, (2.8) where ˜y(t) =y(at), ˜r1(t) =r1(at), ˜q1(t) =q1(at), ˜f(t) = a−2f(at). Letl0ay˜ = −y˜00tt+a−1r˜1y˜0,
˜
y∈ C(02)(R). Byla we denote the closure in L1of l0a. Since a−1r˜1(t)satisfies the conditions of Lemma2.2, it follows that the operatorla is continuously invertible and
ky˜00k1+ka−1r˜1y˜0k1 ≤3klay˜k1, ∀y˜∈ D(la). (2.9) Leta=4 1+γq˜1,˜r1
. By2.1, we obtain that ka−2q˜1y˜k1 ≤ γq˜1,˜r1
a2 kr˜1y˜0k1≤ 3
4klay˜k1, ∀y˜ ∈ D(la). (2.10) Hence, by a well-known theorem (see [11, Chapter 4, Theorem 1.16]), we find that the operator la+a−2q˜1(t)Ecorresponding to (2.8) is invertible andR la+a−2q˜1E
=L1. Let ˜ybe a solution of the equation (2.8), by (2.9) and (2.10), we obtain that
ky˜00k1+ka−1r˜1y˜0k1+ka−2q˜1y˜k1 ≤4klay˜k1. (2.11)
From (2.10) it follows that
ka−2q˜1y˜k1≤3k la+a−2q˜1E
˜ yk1. So
klay˜k1≤ k la+a−2q˜1E
y˜k1+ka−2q˜1y˜k1 ≤4k la+a−2q˜1E
y˜k1. (2.12) The inequalities (2.11) and (2.12) imply that for the solution ˜yof (2.8), the following inequality holds:
ky˜00k1+ka−1r˜1y˜0k1+ka−2q˜1y˜k1≤16kf˜k1. Takingt =a−1x, we to obtain (2.7).
Remark 2.4. Letr1(x)be continuously differentiable. Assumeq1(x)is a continuous function.
L will denote the closure in L1 of the operator L0y := −y00+r1y0+q1y, D(L0) = C(02)(R). If there is a constantc5>0 such thatk −y00k1+kr1(·)y0k1+kq1(·)yk1≤c5(kLyk1+kyk1),∀ y∈ D(L), thenL is called separable inL1.
If the conditions of Lemma2.3 hold, then the operator Lis separable inL1.
3 Proof of the main theorem
Let C(R) be the space of bounded continuous functions on R with the norm kykC(R) = supx∈R|y(x)|. Letεand Abe positive numbers. Set
SA= (
z∈C(R): sup
x∈R
|z(x)| ≤ A )
.
Letv∈SA. Lv,ε denote the closure inL1of the following linear differential expression L0,v,εy:=−y00+r(x,v(x)) +ε 1+x2
y0+q(x,v(x))y, ∀y∈ C0(2)(R). We consider the equation
Lv,εy= f(x). (3.1)
˜
r1,ε,v(x) := r(x, v(x)) +ε 1+x2
and ˜qv(x) := q(x,v(x)) satisfy all of the conditions of Lemma 2.3. Indeed, by (1.3), ˜r1,ε,v(x) ≥ δ1
√
1+x2, δ1 > 0. Hence γ1,˜r1,ε,v < ∞. From (1.4) it follows thatγq˜v(x),˜r1,ε,v(x) ≤C1supt∈Rγq(x,t),r(x,t) <∞. Therefore, for any f ∈ L1, the equation (3.1) has a unique solution yand
ky00k1+
r(·, v(·)) +ε(1+x2)y0
1 +kq(·, v(·))yk1 ≤C2kfk1, (3.2) whereC2 does not depend on A. By2.1, we have that
kyk1 ≤C3kr˜1,ε,vy0k1,
p1+x2y
1 ≤C4
1+x2 y0
1. (3.3)
By using (3.2), (3.3) and Theorem 1 given in Chapter 3 of [18], we obtain that kykW :=ky00k1+
r(·, v(·)) + (1+x2)y0 1+
h|q(·, v(·))|+p1+x2i y
1
+sup
x∈R
(1+x2)3/8y(x)
≤C5kfk1, y∈ D(Lv,ε), (3.4)
whereC5 also does not depend on A.
Let A = C5kfk1+1. SetPε(v) = L−v,ε1f, wherev ∈ SA, ε >0, f ∈ L1 and L−v,ε1 is inverse to Lv,ε. According to (3.4), the operator Pε maps SA into itself. Moreover, the operator Pε maps SAto the set
QA:= {y:kykW ≤ C5kfk1}.
QA is compact inC(R). Indeed, letγ > 0, then by (3.4) there exists l ∈ Nsuch that for any z∈ QA
kz00kL
1(R\[−l,l])+
r(·, v(·)) + (1+x2)z0
L1(R\[−l,l])
+
h|q(·, v(·))|+p1+x2i z
L1(R\[−l,l]) <γ/2, (3.5) and
sup
x:|x|≥l
C5kfk1+1
(1+x2)3/8 <γ/2. (3.6)
Let ϕl ∈C0(∞)(−l−1, l+1) (l=1, 2, . . .) such that ϕl(x) =1 for x∈[−l, l],ϕl(x) =0 for x ∈/ [−l−1, l+1]and 0 ≤ ϕl ≤1. We denote Tl = {ϕlz:z∈ QA}. By (3.5) and (3.6),Tl is a γ-net ofQA. On the other hand,Tl is a subset of the Sobolev space
◦
W21(−l−1, l+1) =θ∈W12(−l−1, l+1):θ(x) =0, as|x| ≥l+1 .
Notice that the embedding of W◦ 21(−l −1,l+1) in C0[−l−1,l+1] is compact, where C0[−l−1,l+1] = {η∈ C[−l−1, l+1]:η(x) =0, as|x| ≥l+1} (see [19, 27]). So Tl is a compactγ-net ofQA. By Hausdorff’s theorem (see [10, Chapter 1]),QA is compact inC(R).
Next, we show that the operator Pε is continuous on SA. Let{vn}∞n=1 ⊂ SAbe a sequence such that supx∈R|vn(x)−v(x)| →0 asn→+∞. Ifyn (n=1, 2, . . .) andysatisfy
Lv,εy= f, Lvn,εyn= f. (3.7)
ThenPε(vn)−Pε(v) =yn−y. So it suffices to show that supx∈R|yn(x)−y(x)| →0 asn→∞.
By (3.7), we deduce that
y−yn =L−v,ε1[r(x,vn(x))−r(x,v(x))]y0n+ [q(x,vn(x))−q(x,v(x))]yn . (3.8) Since functions v and vn (n = 1, 2, . . .) are continuous, we see that r(x,vn(x))−r(x,v(x)) and q(x,vn(x))−q(x,v(x))are continuous functions. Therefore, from (3.8) it follows that
kyn−ykL1(−a,a)≤c max
x∈[−a,a]
r(x,vn(x))−r(x,v(x))|,|q(x,vn(x))−q(x,v(x))|
×hky0nkL1(−a,a)+kynkL1(−a,a)i→0
(3.9)
as n → ∞, for everya > 0. On the other hand, by (3.4), we have {yn}+n=∞1 ⊂ QA, kynkW ≤ A(n = 1, 2, . . .), y ∈ QA, kykW ≤ A. Since the set QA is compact in C(R), without loss of generality, we can assume that the sequence{yn}+n=∞1converges to somez∈ C(R). By (3.4)
|xlim|→∞y(x) =0, lim
|x|→∞z(x) =0. (3.10)
Since the operatorL−v,ε1 is closed, by (3.9) and (3.10), we obtain thatz =y. ThusPε is continu- ous.
SoPε is a completely continuous operator inC(R)and it maps the ball SA into itself. By Schauder’s theorem (see [10, Chapter XVI]), Pε has a fixed pointy inSA, i.e.Pε(y) = y. And ysatisfies the equality
−y00+r(x, y) +ε(1+x2)y0+q(x,y)y= f(x). By Lemma2.3, we obtain that
ky00k1+kr(·, y) +ε(1+x2)y0k1+kq(·, y)y0k1 ≤C5kfk1. Now, let
εj ∞j=1 be a sequence of positive numbers such that limj→∞εj = 0. Recall that Pεj(v) =L−v,ε1jf. Ifyj ∈SAis the fixed point of the operatorPεj, then
−y00j +r(x, yj) +εj(1+x2)y0j+q(x,yj)yj = f(x). Then according to Lemma2.3, we have
y00j
1+
r ·, yj(·)+εj 1+x2 y0j
1 +q ·, yj(·)yj
1 ≤C5kfk1. (3.11) Let (a,b) be an arbitrary finite interval. It is known that the spaceW12(a, b)is compactly embedded to L1(a, b). Therefore, by virtue of (3.11), we can select a subsequence
˜
yj ∞j=1 of yj ∞j=1⊂W12(a, b)such thatky˜j−ykL1(a,b)→0 asj→∞. By Definition1.1,yis a solution of equation (1.1). By Lemma2.3, we obtain that forythe estimate (1.5) holds.
Remark 3.1. The condition (1.3) is natural. If (1.3) does not hold, from Lemma2.1 it follows that the domain D(L)ofLis not included inL1.
Acknowledgements
This work is supported by project 5132/GF4 of Science Committee of Ministry of Education and Science of Republic of Kazakhstan.
References
[1] H. A. Atia, Separation problem for second order elliptic differential operators on Rie- mannian manifolds,J. Comput. Anal. Appl.19(2015), No. 2, 229–240.MR3309954
[2] H. A. Atia, R. S. Alsaedi, A. Ramady, Separation of bi-harmonic differential operators on Riemannian manifolds,Forum Math.26(2014), 953–966.MR3200356
[3] K. Kh. Boimatov, Separability theorems for the Sturm–Liouville operator (in Russian), Mat. Zametki 14(1973), 349–359, [English translation in: Math. Notes 14(1973), 761–767].
MR0338835
[4] K. Kh. Boimatov, Separability theorems, weighted spaces and their applications (in Rus- sian),Trudy Mat. Inst. Steklov. 170(1984), 37–76, [English translation in: Proc. Steklov Inst.
Math.170(1987), 39–81].MR0790327
[5] S. A. Gabov, G. Yu. Malysheva, A. G. Sveshnikov, A. K. Shatov, On equations arising in the dynamics of a rotating stratified and compressible fluid (in Russian), Zh. Vychisl.
Mat. i Mat. Fiz.24(1984), No. 12, 1850–1863.MR0776596
[6] E. Z. Grinshpun, M. Otelbaev, Smoothness of solutions of a nonlinear Sturm–Liouville equation in L1(−∞,+∞), Izv. Akad. Nauk Kazakh SSR Ser. Fiz.-Mat. 5(1984), 26–29.
MR0774312
[7] W. N. Everitt, M. Giertz, Some properties of the domains of certain differential opera- tors,Proc. Lond. Math. Soc. (3)23(1971), 301–324.MR0289840
[8] W. N. Everitt, M. Giertz, Some inequalities associated with certain differential opera- tors,Math. Z.126(1972), 308–326.MR0303001
[9] W. N. Everitt, M. Giertz, J. Weidmann, Some remarks on the separation and limit-point criterion of second order ordinary differential expressions,Math. Ann.200(1973), 335–346.
MR0326047
[10] L. V. Kantorovich, G. P. Akilov, Functional analysis, Pergamon Press, Oxford, 1982.
MR664597
[11] T. Kato,Perturbation theory for linear operators, Springer Science & Business Media, 1995.
MR1335452
[12] O. Milatovic, A separation property for magnetic Schrödinger operators on Riemannian manifolds,J. Geom. Phys.61(2011), 1–7.MR2746975
[13] O. Milatovic, Separation property for Schrödinger operators in Lp-spaces on non- compact manifolds, Complex Var. Elliptic Equ.58(2013), No. 6, 853–864.MR3170665 [14] B. Muckenhoupt, Hardy’s inequality with weights, Stud. Math. 44(1972), 31–38.
MR0311856
[15] M. B. Muratbekov, K. Ospanov, S. Igisinov, Solvability of a class of mixed type second order equations and nonlocal estimates, Appl. Math. Lett. 25(2012), 1661–1665.
MR2957731;url
[16] M. B. Muratbekov, M. Otelbaev, On the smoothness of solutions of nonlinear Sturm–
Liouville equation, in: VII-th Kazakhstan interuniversity conf. on math. and mech. 1981, Kara- ganda, 1981, 34–35.
[17] M. B. Muratbekov, M. Otelbaev, Smoothness and approximation properties for solu- tions of a class of nonlinear equations of Schrödinger type,Izv. Vyssh. Uchebn. Zaved. Mat.
33(1989), No. 3, 68–74.MR1006572
[18] K. T. Mynbaev, M. Otelbaev, Vesovye funktsional’nye prostranstva i spektr differentsial’nykh operatorov (in Russian) [Weighted functional spaces and spectrum of differential opera- tors], Nauka, Moscow, 1988.MR950172
[19] S. M. Nikol’sky, Approximation of functions of several variables and embedding theorems, Springer, New York, 1974.MR0374877
[20] R. Oinarov, Separability of the Schrödinger operator in the space of summable functions, Dokl. Akad. Nauk SSSR285(1985), No. 5, 1062–1064.MR820597
[21] K. Ospanov, Coercive estimates for a degenerate elliptic system of equations with spec- tral applications,Appl. Math. Lett.24(2011), 1594–1598.MR2803716;url
[22] K. N. Ospanov, R. D. Akhmetkaliyeva, Separation and the existence theorem for second order nonlinear differential equation, Electron. J. Qual. Theory Differ. Equ. 2012, No. 66, 1–12.MR2966808
[23] M. Otelbaev, The summability with weight of the solution of a Sturm–Liouville equation (in Russian),Mat. Zametki16(1974), 969–980, [English translation in: Math. Notes16(1974), No. 6, 1172–1179].MR0369798
[24] M. Otelbaev, Coercive estimates and separation theorems for elliptic equations in Rn, Proc. Steklov Inst. Math.161(1984), 213–239.
[25] M. Otelbaev, Coercive estimates for the solutions of difference equations, Proc. Steklov Inst. Math.181(1989), 265–274.
[26] E. I. Shemyakin, Propagation of nonstationary perturbations in a visco-elastic medium, Dokl. Akad. Nauk. SSSR104(1955), No. 1, 34–37.
[27] H. Triebel, Interpolation theory, function spaces, differential operators, J. A. Barth Verlag, Heidelberg, 1995.MR1328645