CC BY
7
MSC 43A35, 47B37
DOI: 10.14529/ mmp200202
POSITIVE SOLUTIONS TO SOBOLEV TYPE EQUATIONS WITH RELATIVELY p-SECTORIAL OPERATORS
J. Banasiak1,2, N.A. Manakova1, G.A. Sviridyuk1
1 South Ural State University, Chelyabinsk, Russian Federation
2University of Pretoria, Pretoria, South Africa
E-mails: jacek.banasiak@up.ac.za, manakovana@susu.ru, sviridyuk@susu.ru
The article describes sufficient conditions for the existence of positive solutions to both the Cauchy problem and the Showalter-Sidorov problem for an abstract linear Sobolev type equation. A distinctive feature of such equations is the phenomenon of non-existence and non-uniqueness of solutions. The research is based on the theory of positive semigroups of operators and the theory of degenerate holomorphic semigroups of operators. The merger of these theories leads to a new theory of degenerate positive holomorphic semigroups of operators. In spaces of sequences, which are analogues of Sobolev function spaces, the constructed abstract theory is used to study a mathematical model. The results can be used to study economic and engineering problems.
Keywords: Sobolev type equations; positive degenerate holomorphic semigroups of operators; positive solution; Sobolev sequence spaces.
Introduction
Let u, f be real Banach spaces, the operators L G L(u; f) (i.e. L is linear and continuous operator) and M G Cl(U; f) (i.e. M is linear, closed, and densely defined operator). Let us consider the nonhomogeneous linear Sobolev type equation
Lit = Mu + f. (1)
For the first time in history, the term "Sobolev type equations" appeared in the paper [1]. Currently, these equations occupy a very vast area among non-classical equations of mathematical physics, and are studied in very different aspects (see, for example, [2,3]). Besides that, there are many treatises in which Sobolev type equations have other names, for instance, "degenerate equations" [4] or "partial differential equations and systems not solvable with respect to the highest-order derivative" [5]. In this paper, the terms [4] and [5] are considered to be synonyms for the term [1-3]. We are interested in the conditions under which equation (1) (ker L = {0} is allowed) has a unique positive solution.
The main tool to find such conditions is the theory of positive degenerative holomorphic semigroups of operators. (Recall that the semigroup of operators V* = {V* : t G R} is called degenerate if s-lim V* = I [2, Ch. 2]). Therefore, in Section 1 of this paper, we
present the basic facts of the theory of positive holomorphic semigroups of operators given in [6, Ch. 1] and [7, Ch. 2 and Ch. 3]. Then, in Section 2, we extend the results of Section 1 to positive degenerate holomorphic semigroups of operators. The main result of Section 2 is the sufficient and necessary condition of the positivity of degenerate holomorphic semigroups of operators. Note that this result is published for the first time. Section 3 contains the main results of this paper. First, we consider the linear homogeneous Sobolev type equation
Lu = Mu. (2)
It turns out that if the operator M is strongly (L,p)-sectorial on the right for some p £ {0} U N [2, Ch. 2], then the resolving semigroup of equation (2) is a degenerate holomorphic semigroup of operators U* = {U* : t £ R}. If, in addition, the operator M is strongly (L,p)-sectorial for some p £ {0} U N [2, Ch. 2] and the semigroup U* is positive, then, for all u0 £ u1 ifu+, there exists the unique positive solution u = u(t) to the Cauchy problem
lim (u(t) - uo) = 0 (3)
t—o+
of equation (2) such that u(t) = U*u0. Here u1 is the phase space of equation (2), u+ is the proper generative cone, and u = (u, || ■ ||u, u+) is the Banach lattice. Then we state sufficient conditions for the existence of the unique positive solution to the Showalter-Sidorov problem
lim P(u(t) - u0) = 0 (4)
t—o+
for equation (1). Here the operator P = s-lim U* is the unit of the semigroup U*, which is
t—o
a projector by construction [2]. Note that, in this case, the positivity of the semigroup U* is not enough, and the Banach lattices u = (u, || ■ |u, u+) and f = (f, || ■ ||f, f+) should be matched.
Further, we use the abstract results in order to construct a positive solution to Showalter-Sidorov problem (4) for an interpretation of abstract equation (1), where the operators L = diag{Lk(A)}, M = diag{Mk(A)}, while Lk(A) and Mk(A) are polynomials with real coefficients such that
deg Lk(A) < deg Mk (A) Vk £ N.
In this case, it is necessary to switch classical Sobolev spaces to Sobolev spaces of sequences [8,9]
l™=\u= {uk} : J2 \J\uk\ < oo \ , m £ E, q £ [1, +oo),
k=1
which are analogues of Sobolev spaces W™. Here {Ak} C R+ is a monotonically increasing sequence such that lim Ak = The paper [10] was the first to find conditions for the
k—^^o
existence of a degenerate positive resolving group of operators of equation (2) in the case of the strongly positive relatively bounded operator M, as well as conditions for the existence of a positive solution to problems (2), (3) and (1), (4), where the operators L = L(A), M = M(A) are polynomials with real coefficients such that deg L > deg M.
The methods and approaches developed in the paper can be widely used to solve economic and technical problems in which the positivity of a solution is of practical importance (for example, studying the process of pressure of a filtered fluid). The proposed spaces of sequences allow to study problems in quasi-Sobolev spaces of sequences that are quasi-Banach spaces [8,9] in the case of q £ (0,1). The need to consider such non-classical spaces takes place in a number of technical problems [11]. Also, diagonal operators that can be investigated by the proposed methods arise in closed economic systems (that is, when the rate of production depends only on the output of the product itself) described by balance models (Leontief type models) [12].
1. Positive Holomorphic Semigroups of Operators
Consider the real Banach space v, the operator A : dom A C v ^ v such that A G Cl(v) (i.e., A is linear, closed, and densely defined). Denote by p(A) = {^ G C : (p! — A)-1 G L(v)} the resolvent set of the operator A, where L(b) is the space of linear continuous operators defined on b, a (A) = C \ p(A) is the spectrum of the operator A. Slightly departing from standard [6, Ch. 1], we define the sectorial operator.
Definition 1. The operator A G Cl(v) is called sectorial, if
(i) there exist the constants «gR and 6 G (f,vr) such that
Sa,& = {p G C : | arg(p — a)| < 6, p = a} C p(A);
(ii) there exists the constant K > 1 such that
K
for any p G Sa,@.
Definition 2. The mapping V* G C^(R+; L(v)) is called a semigroup of operators, if
VsVtv = Vs+tv Vs, t G R+ Vv G v. (5)
A semigroup is called holomorphic, if the semigroup can be continued in some sector containing the ray R+ with preservation of property (5).
Definition 3. The operator A G Cl(b) is called the infinitesimal generator of the semigroup V* = {V* : t G R+}, if Av = lim (V*v — v)t-1 for all v G dom A. If A is
the infinitesimal generator of the semigroup V*, then we write Vt = eAt.
Theorem 1. [6, Ch. 1] The following statements are equivalent.
(i) The operator A G Cl(b) is sectorial.
(ii) The operator A G Cl(b) is the infinitesimal generator of the holomorphic semigroup {etA : t G R+} having the form,
etA = — I (pi - A)~1eIItdfj,, t G R+, 2ni
where the contour r C Sa,© is such that | argp| ^ 6 for |p| ^ to, p G r.
Remark 1. If the operator A is sectorial, then there exists a unit of the semigroup {etA : t G R+} given by the formula I = s- lim et , where "s-lim" denotes strong (i.e.
pointwise) limit.
Corollary 1. [6, Ch. 1] Let A G Cl(b) be a sectorial operator. Then
eAt = s- lim (l- ¿¡-A) , t G R+. fciro V k
Note that Corollary 1 is true under more weak condition on the operator A.
Definition 4. The real Banach space b = (b, || ■ ||B) is called the ordered Banach space b = (b, || ■ ||B, >B) if there exists the order relation >B, which satisfies the axioms of reflexivity, transitivity, antisymmetry, and is consistent with the vector structure of the space b, i.e.
(i) (x >B y) ^ (ax >B ay) for any x, y £ b and for all a £ R+,
(ii) (x >B y) ^ (x + z >B y + z) for any x, y £ b and for all z £ b.
If, in addition, the order relation >B is consistent with the metric structure of the space b, i.e. for all x £ b there exist x+ ,x- £ b such that
(iii) (x+ >B 0) A (x- >B 0) A (x = x+ — x-),
(iv) (|x| >b |y|) ^ (||x||b > ||y||b) for any y £ b,
then the ordered Banach space b = (b, || ■ ||B, >B) is called the Banach lattice. (Here |x| = x+ + x-).
As an example, we note the following Banach lattices:
(i) the space of continuous functions C(Q; R) for any domain Q C Rn, if the order >
is defined by the formula (/ > g) ^ (/ (x) > g(x) for all x £ Q) with /+(x) = max{/(x), 0}, f-(x) = max{—f (x), 0}
x&l x&l
(ii) the Lebesgue space Lq(Q; R),q £ [1, to), Q C Rn, if the order > is defined by the formula (/ > g) ^ (/(x) > g(x) for a.e (almost everywhere) x £ Q) with /+ (x) = vraimax{/(x), 0}, /_(x) = vraimax{—/(x), 0},
(iii) the sequence space lq, q £ [1, to), if the order > is defined by the formula (x > y) ^ (xk > yk for all k £ N) with |x| = (|xk|),
(iv) the space Rn endowed with any norm, if the order > is defined by the formula (a > b) ^ (ak > 6k for all k = 1, n) with |a| = col(|ai|, |1, •••, \o>n\)-
Note that not every ordered Banach space is a Banach lattice. Indeed, let C1 ([0,1]; R) be a space of functions that are continuously differentiable on the interval [0,1] endowed with the norm
||x|| = max |x(t)| + max |x'(t)|. te[o,1] te[o,1]
Suppose that the order > is given by the formula (x > y) ^ (x(t) > y(t) for all t £ [0,1]). The order > satisfies the axioms of reflexivity, transitivity and antisymmetry. In addition, the order > is consistent with the vector structure of the space C1 ([0,1]; R). However, the order > is not consistent with the metric structure of the space C 1([0,1]; R). This fact is easy to establish by considering the functions x(t) = e and y(t) = e* for all t £ [0,1].
Definition 5. Let b be a Banach space. A convex set c C b such that ac + ^c C c for all a,^ £ R+ is called a cone. A cone c is called proper, if c if (—c) = {0}, and generative, if b = c — c.
Let b be a Banach lattice, then (it is easy to see) b+ = {x £ b : x > 0} is a proper generative cone. On the other hand, let b be a Banach space, and c C b be a proper generative cone. Let us introduce the order relation >B by the formula (x >B y) ^ (x — y £ c). Then (b, || ■ ||B, >B) is an ordered Banach space but, taking into account the
example of the functional space C 1([0,1]; R), we see that b may not be a Banach lattice. Further, we well distinguish the ordered Banach space (b, || ■ ||B, >B) and the Banach lattice (b, || ■ ||b, b+).
Definition 6. (i) Let (b = (b, || ■ ||B, b+) be a Banach lattice. The operator A G L(b) is called positive, if Ab+ C b+.
(ii) A semigroup of operators V* = (V4 : t G R+} is called positive, if C
v+ for all t G R+.
Theorem 2. Let b = (b, || ■ ||B, b+) be a Banach lattice, and A G C/(b) be a sectorial operator. Then the following statements are equivalent.
(i) The resolvent RM(A) = (pi — A)-1 is a positive operator for sufficiently large p G R+.
(ii) The holomorphic semigroup (etA : t G R+} is positive.
Proof. (i)^(ii) is true by virtue of Corollary 1. (ii)^(i) is true by virtue of the
oo
representation [6, Ch. 1] RM(A) = f etAe-Mtdt for all p > Re a(A).
0 □
2. Positive Degenerate Holomorphic Semigroups of Operators
Let us construct degenerate positive holomorphic semigroups of operators. Suppose that u, f are Banach spaces, the operators L G L(u;f) (i.e. L is linear and continuous), M G C/(u; f) (i.e. M is linear, closed, and densely defined). The foundation of our research is the theory of degenerate semigroups of operators and the phase space method described in [2, Ch. 3]. Let us give the necessary information on the theory of degenerate semigroups of operators in Banach spaces. Consider the L-resolvent set pL (M) = (p G C : (pL — M)-1 G L(f; u)} and the L-spectrum aL(M) = C \ pL(M) of the operator M, as well as the operator functions RL(m) = (pL — M)-1L and LL(M) = L(pL — M)-1, which are called the right and left L-resolvents of the operator M, respectively (see [2, Ch. 1]). Let pq G pL(M), q = 0,1,... ,p. The operator functions
(M) = П (M)> (M) = П (M)
5L RL(M), Ll
k=0 k=0
are called the right and left (L,p)-resolvents of the operator M, respectively.
Definition 7. [2, Ch. 3] The operator M is called p-sectorial with respect to the operator L with the number p G (0} U N (in short, (L,p)-sectorial), if
(i) there exist the constants a G R and 6 G (f,vr) such that
Sa,e(M) = (p G C : | arg(p — a)| < 8,p = a} C pL(M);
(ii) there exists the constant K G R+ such that
K
maxjlli^M)!!^), HLf^M)!!^} <
p
П - a|
q=0
for any pq G 5^е(М), q = 0,1,...
Lemma 1. [2, Ch. 3] Let the operator M be (L,p)-sectorial for some p G {0} U N. Then there exist the degenerate holomorphic semigroups of operators
U* = ±f RLu{M)e^d^i and Ff = — / LL(M)e^d^i.
Recall [2, Ch. 3] that the semigroup V* is called degenerate, if s- ^lim Vt = I. Here
t G R+, and the contour r C S^q(M) is such that | arg^ 6 for ^ ^ to, ^ G r. If V* G C^(R+; L(v)) is a degenerate holomorphic semigroup of operators, then we can define ker V• = {v G v : Vtv = 0 31 G R+}. Since the semigroups U* and F* are holomorphic, we set u0 = ker U•, f0 = ker F•. Let L0 be the restriction of the operator L to u0, and M0 be the restriction of the operator M to u0 fl dom M.
Theorem 3. [2, Ch. 3] Let the operator M be (L, p)-sectorial for some p G {0}U N. Then
(i) L0 G L(u0; f0) and M0 : u0 f dom M ^ f0;
(ii) there exists the inverse operator M—1 G L(f0; u0);
(iii) the operator H = M—1L0 G L(u0) (G = L0M—1 G L(f°)) is nilpotent of degree less than or equal to p.
If V* G G^(R+; L(v)) is a degenerate holomorphic semigroup of operators, then we can define im V* = {v G v : lim Vtv = v}. Since the semigroups U* and F* are
holomorphic, we
set u1 = im U*, f1 = im F*. Let L1 be the restriction of the operator L to u1, and M1 be the restriction of the operator M to u1 f dom M. Then u0 © u1 C u and
f0 © f1 c f.
Definition 8. [2, Ch. 3] The operator M is called strongly (L,p)-sectorial on the right (on the left) for some p G {0} U N, if M is (L,p)-sectorial for some p G {0} U N and
con st
\\Rhp)(M)(XL - M)~1Mu\\a <--- Vu G domM,
|A| EI k |
q=0
o
where const = const(u) (there exists the lineal f, which is dense in f and such that
con st o
||m(a£-m)-^p)(m)/||g< C°/ V/G f,
|A| II k|
q=0
where const = const(f)); G S^(M), q = 0,1,... ,p.
Remark 2. Without loss of generality, we can take a = 0 in Definition 7. Indeed, if we find the resolving semigroup of equation (2) {Ut : t G R+} for a = 0, then the semigroup {eatUt : t G R+} is resolving when a = 0.
Theorem 4. [2, Ch. 3] Let the operator M be (L,p)-sectorial on the right (on the left) for some p G {0} U N. Then there exists the projector P = s- lim Ut (the projector
Q = s- lim Ft).
Remark 3. It is easy to see that, under the conditions of Theorem 4, there exists the splitting of the space u and f, i.e.
u0 © u1 = u and f0 © f1 = f. (A1)
Definition 9. [2, Ch. 3] The operator M is called strongly (L,p)-sectorial for some p G {0} U N, if M is strongly (L,p)-sectorial on the left for some p G {0} U N and
con st
II(XL - My'L^M)^ <---
|A| II k|
q=0
for every A, G S^(M), q = 0,1,... ,p.
Theorem 5. [2, Ch. 3] Let the operator M be strongly (L, p)-sectorial for some p G {0}UN. Then there exists the inverse
operator L-1 G ¿(f1;u1). (A2)
Corollary 2. [2, Ch. 3] Let the conditions of Theorem 5 be satisfied, then
(i) the operators S = L-1 M1 G C/(u1) and T = M1L-1 G C/(f1) are sectorial, moreover, a(S) = a(T) = aL(M);
(ii) U
F4
etSP = s- lim
. -i
L~WW)M} L
k(p+1)
, for t G R+,
p, for t = 0, ...
k(p+i) (A3)
e Q = s- lim Q, for t = 0.
, for t G R+,
Therefore, if the operator M is strongly (L,p)-sectorial for some p G {0} U N then the degenerate holomorphic semigroups U• and F• have the forms (A3). Let us describe the conditions under which these semigroups are positive.
Theorem 6. Let the operator M be strongly (L,p)-sectorial for some p G {0} U N, and the Banach space u be a Banach lattice, u = (u, || ■ ||u, u+). Then the following statements are equivalent.
(i) The operator [R^(M)]p+1 is positive for all sufficiently large p G R+.
(ii) The degenerate holomorphic semigroup U is positive.
Proof. (i)^(ii) follows from (A3). (ii)^(i) can be obtained by considering the representation
p
(pL - M)-1 = pqM0-1 (I - Q) + (pI - S)-1L-1 Q
q=0
for all p G S@ (M). Hence
R(M)]p+1 = (pI - S)-p-1P.
Now we can use the arguments of Theorem 2, i.e.
(iul - S)-1P = J etSPe-Mt dt for all ^ > Re a(S) = Re aL(M).
0 / D
Remark 4. Theorem 6 remains true if we replace the space u (and the Banach lattice u = (u, || ■ ||u, u+)) by the space f (and the Banach lattice f = (f, || ■ ||f, f+)), the operator Rl (M) by the operator L^(M), and the semigroup U* by the semigroup F*. The proof of this fact is left to the reader.
Example 1. Let the operators L and M be represented by square matrices of the order n. The matrix M is called L-regular if there exists the number a G C such that det (aL—M) = 0. If the matrix M is L-regular, then there exist the non-degenerate matrices A and B of the order n such that [13, Ch. 12]
o o o
L = Bdiag {jp1, jp2,..., j, I„_m}A, M = B diag {Im, S }A,
o
where jpk is a Jordan box of the order pk with zeros on the main diagonal, Ik is the identity
i
matrix of the order k, S is a square matrix of the order n — m, pk = m. For the fixed
fc=1
p = maxpk, the L-regular matrix M is called (L,p)-regular. k=l,l
Therefore, let the matrix M be (L,p)-regular for some p G {0} U N. Since the L-spectrum aL(M) of the matrix M consists of the roots of the polynomial det (^L — M)-1 = 0, then the formula
Ut = J RL(Md^ = A-1 diag {Om, etS}A,
Y
where the contour 7 C C bounds the domain containing aL(M), Om is the zero matrix of the order m, which defines the degenerate holomorphic group U*. According to Theorem 6, the group U* is positive exactly when the matrix
[RL(M)]p+1 = A-1 diag {Om, (^I — S)-p-1}A
is positive for all sufficiently large ^ G R+.
Example 2. Let b = (b, || ■ ||B, c) be a Banach lattice, V* be a positive degenerate holomorphic semigroup of operators, im V* be an image of V*, and V0 be the unit of V*. It is easy to see that im V* f c = {0}. Let us show that the inverse proposition is not true. Let u = R3, the operators L and M be given by the matrices
1 1 0 0 1 0 L = 1 1 0 and M = 1 0 0 0 1 1 0 0 1
Since the L-spectrum of the matrix M is aL(M) = < 11, then the matrix M is (L, 0)-
regular and
2
1 1 0
RL (M ) = (2^ — 1)-1| 1 1
¡1 2^—1 1—¡1 /j.— 1
Hence
t /' 1 10 2
uf = e-1 1 10
-1 1 2
and im U* = {u G R3 : u1 = u2}. Consider the canonical cone c = (R+ )3 in R3. Obviously, im U* H c = {u G c : u1 = u2} = {0}, however, the group U* is not positive. In addition, the right L-resolvent R^(M) of the matrix M is not positive for all sufficiently large p G R+.
3. Positive Solutions to Abstract Equations
Let u, f be Banach spaces, the operators L G L(u;f) and M G C/(u;f). Consider linear homogeneous Sobolev type equation (2). The vector function u G C 1(R+; u) is called a solution to equation (2), if the function u satisfies this equation. The solution u = u(t) to equation (2) is called a solution to problem (2), (3) if condition (3) is satisfied for some u0 G u.
Definition 10. [2, Ch. 3] The set P is called the phase space of equation (2), if
(i) any solution u = u(t) to equation (2) belongs to P, i.e. u(t) G P for any t G R+,
(ii) there exists a unique solution to problem (2), (3) for any u0 G P.
Theorem 7. [2, Ch. 3] Let the operator M be strongly (L,p)-sectorial for somep G {0}UN. Then the phase space of equation (2) is the subspace u1.
Note that if u = u(t) is a solution to problem (2), (3) under the conditions of Theorem 7 and some u0 G u1, then the solution has the form u(t) = U4u0, where U* = {U4 : t G R} is a degenerate holomorphic semigroup of operators from Lemma 1. Further, if u = (u, || ■ ||u, u+) is a Banach lattice, and U* is a positive degenerate holomorphic semigroup of operators, then u+ H u1 = {0}. Indeed, (u G u1) ^ (Pu = u), where the projector P is the unit of the semigroup U*. Since the semigroup U* is positive, then we have U4u0 > 0 for any u0 G u+. If u0 G u+ H u1, then (U4u0 > 0) ^ (Pu0 > 0). Denote
u+ = u+ h u1.
Corollary 3. Let the conditions of Theorem 7 be satisfied and suppose that the degenerate holomorphic semigroup of operators U* is positive. Then there exists the unique positive solution to problem (2), (3) for any u0 G u+.
The proof of Corollary 3 is left to the reader.
Remark 5. Let us show that, for any u0 G u, there exists a solution to problem (2), (4), which has the form u(t) = U4u0. Since U4 = PU4 = U4P, then PU4u0 = U4Pu0 ^ Pu0 for t ^ 0+. The result follows from Theorem 7. If the degenerate holomorphic semigroup U* is positive, and u0 G u+, then the solution u(t) = U4u0 to problem (2), (4) is positive.
Now, consider linear unhomogeneous Sobolev type equation (1). The vector function u G C 1((0, t);u) is called a solution to equation (1), if the function u satisfies the equation for some t G R+ and f : (0,t) ^ f. The solution u = u(t) to equation (1) is called a solution to problem (1), (3) (problem (1), (4)), if the solution satisfies condition (3) (condition (4)) for some u0 G u.
Theorem 8. [2, Ch. 3] Let the operator M be strongly (L,p)-sectorial for some p G {0} U N. Then for any vector u0 G u and any vector function f : (0,t) ^ f such that f0 = (I-Q)f G Cp+1((0,r); f0) and f1 = Qf G C([0,t]; f1) there exists a unique solution u = u(t) to problem (1), (4), and the solution has the form,
p 0 t u(t) = -Y/Hm^1^r(t) + Utu0+ f Ut~sL-llf\s)ds) t G (0,t). (6)
q=0 0
In addition, if the initial vector u0 satisfies the relation
p dq f 0
(I - Q)u0 = - tlim J] HqMQl—J—{t), (7)
q=0
then there exists a unique solution u = u(t) to problem (1), (3), and the solution has the form (6).
Definition 11. Let the operator M be strongly (L,p)-sectorial for some p G {0} U N. The Banach lattices u = (u, || ■ ||u,u+)) and f = (f, II ■ ||f, f+)) are called concordant with respect to the pair (L, M) (briefly, (L, M)-concordant) if
(i) u+ = uk H u+ and f+ = fk H f+ are the proper generative cones, k = 0,1,
(ii) the operators L G L(u+;f+) and M G Cl(u+;f+), moreover, there exist the operators L-1 G L(f+; u+) and M0-1 G L(f+; u+).
Corollary 4. Let the operator M be strongly (L,p)-sectorial for some p G {0} U N, the Banach lattices u = (u, || ■ |u, u+)) and f = (f, || ■ ||f, f+)) be (L, M)-concordant, and the degenerate holomorphic semigroup of operators U* be positive. Then for any vector function f : (0, r) ->• F such that G C((0, r); F+), q = hp, and f1 = C([0, r]; F+), and for any vector u0 G u+ there exists a unique positive solution to problem (1), (4), and the solution has the form (6). If, in addition, the initial vector u0 G u+ satisfies condition (7), then there exists a unique positive solution to problem (1), (3), and the solution has the form (6).
In order to prove, we note that the operator H = M—1 L0 G L(u+) and, therefore, the operator H is positive.
Example 3. Let the spaces u = f = and the operators L, M G L(u; f) be given by the matrices
o o o
L = diag {jp1, jp2,..., , In-m}, M = diag {Im, S},
o
where jpk is a Jordan box of the order pk with zeros on the main diagonal, Ik is the
i
identity matrix of the order k, S is a square matrix of the order n — m, pk = m. Fix
k=1
p = max{p1,p2, ...,pk}, then the matrix M is (L,p)-regular according to Example 1.
Let us construct positive solutions to problem (1), (4). According to Corollary 4, it is necessary to show that the operator M is strongly (L,p)-sectorial for some p G {0} U N. To this end, we use equivalent conditions (A1), (A2), which are easier to verify in our case. Construct the projectors P = Q = diag {Om, In-m}. Define the subspaces u0 = span {e1, e2,..., em} and u1 = span {em+1, em+2 ,...,en}. Here ek G Rn, where the k-th
coordinate equals 1, while all the rest coordinates are zeroes. Then the space u can be represented as the direct sum of two subspaces u = uo © u1. Similarly, we can construct a splitting of the space f. Therefore, condition (A1) is satisfied. The actions of the operators L and M are splitted and their restrictions to the subspaces uk, k = 0,1 take the form
o o o
Lo = diag {jp1, jp2,..., J, On-m}, Li = diag {Om, In-m},
Mo = diag {Im, On-m}, Mi = diag {Om, S}. Then there exist the operators
L-1 = diag {Om, In-m}, Mo-1 = diag {Im, On-m}
and condition (A2) is satisfied. Therefore, we see that conditions (A1), (A2) are satisfied. Hence, the operator M is strongly (L,p)-sectorial. Construct the operator
o o o
H = Mo- Lo = diag {jp1, jp2,..., j, On-m}.
The formula
U4 = J RL(Mdp = diag {Om, etS}
Y
defines a degenerate holomorphic group of operators, where the contour y C C bounds the domain containing aL(M), and Om is the zero matrix of the order m. Moreover, Theorem 6 shows that the group U* is positive exactly when the matrix
[RL(M)]p+1 = diag {Om, (pI - S)-p-1}
is positive for all sufficiently large p G R+. Further, we assume that the matrix S satisfies this condition.
As a cone, we consider the proper cones u+ = f+ = {R+}n. Then the Banach lattices u = (u, || ■ ||u, u+)) and f = (f, || ■ if, f+)) are (L, M)-concordant. Therefore, all the conditions of Corollary 4 are satisfied. Hence, there exists a unique positive solution to problem (1), (4), which is given by formula (6) for any vector function f : (0, t) ^ f such that —^¿r G C((0,t);F+)> Q = 1>P> and /1 = C([0,t];F+)> and for any vector Uo G il+. Note that the conditions on the vector function f : (0,t) ^ f can be satisfied [10], for example, if we consider the vector function f (t) = eatfo, where a G R+, —fo G f+.
4. One Concrete Interpretation of Abstract Equation
Consider the monotonically increasing sequence {Ak} C R+ such that lim Ak = +to.
fc^TO
Construct Sobolev spaces of sequences Z^7", m G R, q G [1, +to), which are Banach spaces endowed with the norm
1 q
~2~
x—4
ev ki
vfc=l
Obviously, the embeddings Z^ ^ Z^ are dense and continuous for all m > n and q G [1, то). Suppose that the operator Ли = (А1м1, A2u2,...) acts in the space of sequences, then
A : /m+2 ^ l^7, is linear, continuous and continuous invertible for all m G R and q G [1, to)
rfc sk
[8,9]. Let Lk(Z) = akZj and Mk(Z) = jZj be polynomials of the degrees rk and sk, j=0 j=0 respectively, with real coefficients such that the roots of a finite number of the polynomials Lk (Z) are the numbers Z = Ak. Construct the operators
L = diag{Lk(Ak)}, M = diag{Mk(Ak)}
acting in the Banach spaces of sequences satisfying the conditions
rk < sk for all k G N. (8)
Suppose that there exists s = max{sk}, denote r = max{rk} and construct the operators
L G L(Zm+2r; /£*), M G L(Zm+2s; /£*). By virtue of condition (8), we obtain that s > r. Set u = ¿m+2r, f = m G R, q G [1, +to), then the operators L G L(u; f), M G CZ(u; f), domM = lm+2s.
Proposition 1. Suppose that condition (8) is satisfied and the polynomials Lk = Lk (Z) and Mk = Mk(Z) for all k G N have only real roots and do not have common roots, moreover, the roots of a finite number of the polynomials Lk(Z) are the numbers Z = Ak and the condition
akfc ■ 6kfc < 0 for all k G N such that Lk(Ak) = 0 (9)
is satisfied. Then the operator M is (L, 0)-sectorial.
Proof. Construct the L-spectrum of the operator M
aL(M) = {№6C:№ = ¥TTT> k : * 0>'
Lk (Ak )
where the numbers G aL(M) ordered taking into account their multiplicities are real
numbers and, by condition (8), tend to —to. Denote a = max. Denote by {ek : ek =
k
(0,..., 0,1, 0,...)} a family of vectors, where the unit is on the k-th position, and construct the operator
where ' at the sum means absence of the terms with the numbers k such that Lk(Ak) = 0. Next, consider
OO | ^ ^ 1,711 II q OO , , - m + 2r
\\-RL(M\„,\\q _ ^ /l<",efc>H|efc||^m+2r _ ^ ,\<u,ek>\"{\k)q^- _
\\nA1V1>U\\q,m+2r-l^ \H-Hk\q ~
k=1 k=1
CO / m + 2r \ CO / m + 2r \
- / / ki(afc)—5- \ < / / ki(afc)—5-
ti V / "S I sin%-a|
1 00 / ( m + 2r \ g 1
sin
EMKI(Afc) 2
sin
where the angle 9 = | + a, a G (0, Therefore, we see that the operator M is (L, 0) sectorial.
□
Construct the splitting of the space u into the direct sum of two subspaces:
no = zm+2r = i = { } G zm+2r ufc = 0, k : Lfc(Afc) = 0,
u = iq>o iu = {Uk}G : - =0, k : Lk(Ak) = 0
«k
u1 = /m+2r = u = («к} g /m+2r
ufc = 0, k : Lfc(Afc) = 0, ufc = 0, k : Lfc (Afc ) = 0
Similarly, we construct the splitting of the space f into the direct sum of two subspaces:
f0
jm
f1
m lq,1
f = Ufc} G / f = (fk} G /
m
q
fk = 0, k : Lfc (Afc) = 0, fk = 0, k : Lk(Ak) = 0
fk = 0, k : Lk (Afc) = 0, ^ • fk = 0, k : Lk (Ak) = 0
Then the operators Li G ¿(u1 ; f1), L-1 G ¿(f1; u1), M-1 G L(f0; u0) are defined as follows:
" <f1, ek >
L1U1 = 'Lk(Ak) <u1,ek > ek, L-1 f1 =
k=1
k=1
Lk (Ak)
-ek,
Mo-1f 0
E
</°,efc> Mfc(Afc)
-efc.
k: Lfc (Afc)=0
Therefore, we see that, under the conditions of Proposition 1, the operator M is (L, 0)-sectorial and conditions (A1), (A2) are satisfied.
In the space u, consider the family of vectors {ek} and the canonical cone u+ formed as a closure of a linear combination of the vectors {ek} with non-negative coefficients. Then the space (u, u+, || ■ ||q,m+2r) forms a Banach lattice. Similarly, in the space f, we consider the family of vectors {ek} and the canonical cone f+ as a closure of a linear combination of the vectors {ek} with non-negative coefficients. Then the space (f, f+, || ■ ||q,m) forms a Banach lattice. Here the sets are as follows:
u+
im+2r
fq,o
u
(u } G /m+2r : uk > 0, k : Lk(Ak) = 0,
(Uk} G 1q : « = 0, k : Lk (Ak) = 0
uk
u+ = с+2г 4 « = («k} G ¿m+2r : f = (fk} G II
f
40
m
lq
f+ = g
f = (fk} G l
q,1 = i f = (fk
m
q
uk = 0, k : Lk(Ak) = 0, Uk > 0, k : Lk (Ak) = 0
> 0, k : Lk (Ak) = 0, = 0, k : Lk(Ak) = 0
= 0, k : Lk(Ak) = 0,
> 0, k : Lk(Ak) = 0
Therefore, it follows from the construction of the operators L1, Lo , L- , Mo that the Banach lattices u = (u, || ■ ||u, u+)) and f = (f, || ■ if, f+)) are (L, M)-concordant.
Theorem 9. Suppose that the conditions of Proposition 1 are satisfied, the values of the polynomials Lk = Lk (Z) and Mk = Mk(Z) are positive for Z = Ak such that Lk (Ak) = 0. Then for any vector function f : (0,t) ^ f such that —fo G C1 ((0,t);f+), f1 G
C((0,t);f+ ) and for any u0 G u such that U) G u+, there exists the unique positive solution u = u(t) to problem (1), (4), and the solution can be represented as
u(
where
u
(t) = -Mo-1 f0(t) + U4 + L-1 f 1(s)ds,
rrt 1 /Mk (Ak Л 1
= ^ exP \ Lk{Xk) ) < и°,вк >вк
k=i ^Lfc (Afc )
Proof. Let us show that the operator M is positive and strongly (L, 0)-sectorial. Consider the L -resolvent of the operator M
ж ✓ < e > \ ж ✓ < e > \
{ilL ~ M)~1 = ^' (pLfc(Afc)'-Mfc(Afc)j 6fc = ^' (LMfr-K)) 6fc'
which is a positive operator by construction. Here ' at the sum means absence of the terms with the numbers k : L k(Ak) = 0. Therefore, we see that the operator M is (L , 0)-sectorial, conditions (A1), (A2) are satisfied, and the Banach lattices U = (U , || ■ ||u,U+)) and F = (F , || ■ if ,F+)) are (L , M)-concordant. The proof of the theorem is true by virtue of Corollary 4.
□
Conclusion
In this paper, we find conditions under which a resolving semigroup of operators is positive, and obtain sufficient conditions for the existence of positive solutions to both the Cauchy problem and the Showalter-Sidorov problem for an abstract linear Sobolev type equation in the case of a relatively sectorial operator. Abstract results are illustrated by finite-dimensional and infinite-dimensional problems.
Recently, the attention of many researchers is given to the search for positive solutions. Note an interesting approach based on the Stampacchia maximum principle [14], which is applied to stochastic partial differential equations presented in the Ito-Stratanovich-Skorokhod form. The obtained abstract results are applied to the stochastic Boussinesq temperature equation and the reaction-diffusion equations perturbed by non-Lipschitz nonlinear noise. As a result, the theorems on the stability of positive solutions to these equations are obtained. Another approach to the study of stochastic equations is based on the Nelson-Gliklikh derivative of stochastic processes [15-18]. In the future, we hope to extend this approach to the study of the stability of positive solutions to linear and nonlinear Sobolev type equations.
References / Литература
1. Showalter R.E. The Sobolev Type Equation I; II. Applicable Analysis, 1975, vol. 5, no. 1, pp. 5-22; vol. 5, no. 2, pp. 81-89. DOI: 10.1080/00036817508839103
2. Sviridyuk G.A., Fedorov V.E. Linear Sobolev Type Equations and Degenerate Semigroups of Operators. Utrecht, Boston, Koln, Tokyo, VSP, 2003. DOI: 10.1515/9783110915501
t
3. Alshin A.B., Korpusov M.O., Sveshnikov A.G. Blow-Up in Nonlinear Sobolev Type Equations. Berlin, Walter de Gruyter, 2011. DOI: 10.1515/9783110255294
4. Favini A., Yagi A. Degenerate Differential Equations in Banach Spaces. N.Y., Marcel Dekker Inc., 1999.
5. Demidenko G.V., Uspenskii S.V. Partial Differential Equations and Systems Not Solvable with Respect to the Highest Order Derivative. N.Y., Basel, Hong Kong, Marcel Dekker, Inc., 2003.
6. Henry D. Geometric Theory of Semilinear Parabolic Equations. Berlin, Springer, 1981.
7. Banasiak J., Arlotti L. Perturbations of Positive Semigroups with Applications. London, Springer, 2006. DOI: 10.1007/1-84628-153-9
8. Keller A.V., Zamyshlyaeva A.A., Sagadeeva M.A. On Integration in Quasi-Banach Spaces of Sequences. Journal of Computational and Engineering Mathematics, 2015, vol. 2, no. 1, pp. 52-56. DOI: 10.14529/jcem150106
9. Zamyshlyaeva A.A., Al-Isawi J.K.T. On Some Properties of Solutions to One Class of Evolution Sobolev Type Mathematical Models in Quasi-Sobolev Spaces. Bulletin of the South Ural State University. Series: Mathematical Modelling, Programming and Computer Software, 2015, vol. 8, no. 4, pp. 113-119. DOI: 10.14529/mmp150410
10. Solovyova N.N., Zagrebina S.A., Sviridyuk G.A. Sobolev Type Mathematical Models with Relatively Positive Operators in the Sequence Spaces. Bulletin of the South Ural State University. Series: Mathematics. Mechanics. Physics, 2017, vol. 9, no. 4, pp. 27-35. DOI: 10.14529/mmph170404
11. Vovk S.M., Borulko V.F. Statement of a Problem of Definition of Linear Signals Parameters in Quasinormed Space. Radioelectronics and Communications Systems, 2010, vol. 53, no. 7, pp. 367-375. DOI: 10.3103/S0735272710070046
12. Keller A.V. On the Computational Efficiency of the Algorithm of the Numerical Solution of Optimal Control Problems for Models of Leontieff Type. Journal of Computational and Engineering Mathematics, 2015, vol. 2, no. 2, pp. 39-59. DOI: 10.14529/jcem150205
13. Gantmacher F.R. The Theory of Matrices. AMS Chelsea Publishing, 2000.
14. Chekroun M.D., Park E., Temam R. The Stampacchia Maximum Principle for Stochastic Partial Equations and Applications. Journal of Differential Equations, 2016, vol. 260, no. 3, pp. 2926-972. DOI: 10.1016/j.jde.2015.10.022
15. Favini A., Sviridyuk G., Manakova N. Linear Sobolev Type Equations with Relatively p-Sectorial Operators in Space of "Noises". Abstract and Applied Analysis, 2015, vol. 2015, article ID: 697410, 8 p. DOI: 10.1155/2015/697410
16. Favini A., Sviridyuk G.A., Zamyshlyaeva A.A. One Class of Sobolev Type Equations of Higher Order with Additive "White Noise". Communications on Pure and Applied Analysis, 2016, vol. 15, no. 1, pp. 185-196. DOI: 10.3934/cpaa.2016.15.185
17. Favini A., Sviridyuk G., Sagadeeva M. Linear Sobolev Type Equations with Relatively p-Radial Operators in Space of "Noises". Mediterranean Journal of Mathematics, 2016, vol. 13, no. 6, pp. 4607-4621. DOI: 10.1007/s00009-016-0765-x
18. Favini A., Zagrebina S.A., Sviridyuk G.A. Multipoint Initial-Final Value Problems for Dynamical Sobolev-Type Equations in the Space of Noises. Electronic Journal of Differential Equations, 2018, vol. 2018, article ID: 128, 10 p.
Received May 21, 2019
УДК 517.9 Б01: 10.14529/шшр200202
ПОЗИТИВНЫЕ РЕШЕНИЯ УРАВНЕНИЙ СОБОЛЕВСКОГО ТИПА С ОТНОСИТЕЛЬНО р-СЕКТОРИАЛЬНЫМ ОПЕРАТОРОМ
Я. Банасяк1'2, Н.А. Манакова1, Г.А. Свиридюк1
1 Южно-Уральский государственный университет, г. Челябинск, Российская Федерация
2Университет Претории, г. Претория, Южно-Африканская Республика
В статье описаны условия, достаточные для существования позитивных решений задачи Коши и задачи Шоуолтера - Сидорова для абстрактного линейного уравнения соболевского типа. Отличительной чертой таких уравнений является феномен несуществования и неединственности решений. Фундаментом наших исследований стали теория позитивных полугрупп операторов и теория вырожденных голоморфных полугрупп операторов. В результате слияния этих теорий получилась новая теория вырожденных позитивных голоморфных полугрупп операторов. В пространствах последовательностей, являющихся аналогами функциональных пространств Соболева, построенная абстрактная теория применена для исследования одной математической модели. Полученные результаты могут быть применены для исследования экономических и инженерных задач.
Ключевые слова: уравнения соболевского типа; вырожденные позитивные голоморфные полугруппы операторов; позитивные 'решения; Соболевы пространства последовательностей.
Яцек Банасяк, доктор физико-математических наук, профессор, заведующий лабораторией, научно-исследовательская лаборатория прикладных полугрупповых исследований, Южно-Уральский государственный университет (г. Челябинск, Российская Федерация); кафедра математики и прикладной математики, Университет Претории (г. Претория, Южно-Африканская Республика), jacek.banasiak@up.ac.za.
Наталья Александровна Манакова, доктор физико-математических наук, кафедра уравнений математической физики, Южно-Уральский государственный университет (г. Челябинск, Российская Федерация), manakovana@susu.ru.
Георгий Анатольевич Свиридюк, доктор физико-математических наук, профессор, заведующий кафедрой, кафедра уравнений математической физики, ЮжноУральский государственный университет (г. Челябинск, Российская Федерация), sviridyuk@susu.ru.
Поступила в редакцию 21 мая 2019 г.
