Buletin Stiintic - Universitatea din Piteti s Seria Matematic i Informatic, Nr. 16 (2010), pg.

1-N as a

Potential theoretics notions with respect to a multivalued operator I.

Abstract In the frame of a regular functional space, for a multivalued monotone operator G, that can be extended to the subdierential of a convex function with some supplementary properties (we call such a functional compact primitive-Denition 3.4) we extend some properties of potential. Also absolute (pure) potential is dened (denition 1.3), as an extension of the previous notion of (pure) potential. We also dene strongly and weakly operating contractions as a more precis notion adapted to the multivalud operator case and obtain some classical results in this new framework. A lower envelope principle (theorem 3.3) and a domination principle (theorem 3.4). Some more results, in a second part. 2010 Mathematics Subject Classication: Primary 31C45, 31A15 ; Secondary 31E99, 47H05. Key words and phrases: potential, convex functional, monotone operator, subdierential.

Introduction

Let X be a locally compact Hausdor space with a countably base and let be a positive (Radon) measure on X. Let X be a Banach space of

University of Pitesti

First AUTHOR, Second AUTHOR

real valued functions on X, locally -sumable. Let G : X X be a multivalued operator such that: a)
xX

G(x) = X .

b) for every u, v X , u1 G(u), v1 G(v), we have: < u1 v1 , u v >> 0. Condition a) means that G is surjective; condition b) means that G is strictly monotone. Here X stands for the dual of X and < u, v > is the value of u at v. Denition 1.1 The space X is called a functional space if (F) for every K compact subset of X, one can nd M (K) > 0, such that for every u X |u|d M (K)||u||.
K

Here ||.|| denotes the norm of the Banach space X . We will denote by M the space of all bounded - measurable function on X, with compact support. For all f in M, we dene Lf by: Lf (v) := f vd.

By axiom (F ), the functional Lf is in X . From the surjectivity of G, we get: Theorem 1.1 If X is a functional space and f M, then there exists a unique uf X such that one can nd u1 G(uf ) satisfying < u1 , v >= f vd,

for every v X . The unicity of uf comes from the property of being strictly monotone of the operator G. Next, we use the notation C := {f : X R, continuous with compact support}. Denition 1.2 A functional space X is called regular if:

The Title of the Article

(RF) C

X is dense in C and in X .

Denition 1.3 Let X be a regular functional space. An element u X is called potential (with respect to G), if there exists u1 G(u) such that one can nd a measure on X satisfying : < u1 , v >= vd, f or all v C X .

This measure satisfying the previous condition, is called the measure associated to u and we will use one of the notations u = u , u1 = u = u(u1 , ), (u, u1 ) to specify the relation between u, u1 and . If for some u1 G(u) the measure is positive, the potential u is said pure potential ( -pure or/and u1 -pure ). If u is u1 -pure for every u1 G(u), then u is said to be absolute pure. Let X be an open subset and let C() be the family of all real functions dened on with a compact support. On the space X , we will consider the order given by u v if u(x) v(x), - a.e. on X. Lemma 1.1 Let C1 be a dense linear subspace of C() such that v + is in C1 for every v C1 . Let L be a positive linear map on C1 . Then L can be extended to a linear positive functional on C(). As such, there exists a positive measure on with: L(v) =

vd f or all v C1 .

Normalized Contractions

Let X be a functional space. Denition 2.1 ([7, Calvert] )An application T : R R is called a normalized contraction on R i i) T 0 = 0. ii) |T r T r | |r r |, for all r, r R.

First AUTHOR, Second AUTHOR

If T is a normalized contraction, one denes: T : L1 L1 loc loc by (T u)(x) := T (u(x)), for all x X, with |u(x)| < . Here L1 := L1 (X, ) is the space of all real functions on X, which loc loc are locally integrable with respect to . Lemma 2.1 Let X be a reexive functional space and let T be a normalized contraction. Let (un )n be a sequence in X , such that T un X , n 1. If (un ) converges to u X with respect to the norm on X and T (un ) is bounded in (X , ||.||), then T u X and T un weakly converges to T u. Proof As X is reexive, (T un ) is relatively compact in the weak topology of X . Let (T un ) be a weakly convergent subsequence and let v be its limit. We have: | (T u v)f d| |T u T un ||f |d + | (T un v)f d|

for every f M and every n . Using Theorem 1.1, we nd u1 such that (T un v)f d =< u1 , T un v > 0 as n By axiom (F ), with M = M (suppf ), we have: |T un T u||f |d |un u||f |d (sup|f |)M ||un u|| 0, as n .

Using Lemma 1.1, we get (T u v)f d = 0 f or all f M. So, v = T u and T un T u weakly in X . Denition 2.2 Let 0 k . We dene: Tk : R [0, ), Tk (r) := min{max(r, 0), k}. We say that Tk is strongly operating on X (with respect to G), if :

The Title of the Article

(Ck ) Tk u X , u X . (GCk ) < u1 u2 , v Tk v > 0, u, v X and every u1 G(u + T v), u2 G(u). We say that Tk is weakly operating on X (with respect to G), if : (Ck ) Tk u X , u X . (GCk ) for all u and v X , there exists u1 G(u + Tk v), there exists u2 G(u), such that < u1 u2 , v Tk v > 0. Lemma 2.2 If condition (Ck0 ) is satised for some k0 , 0 < k0 < , then for every k, 0 < k < , we have Tk X , X and T X for all X C. Proof Let X and 0 < k < ; we have: Tk = ( k0 k ) X. )Tk0 ( k0 k

For k = , let be a function from C X . Let M > 0 be such that sup|| M . We get: T = TM X . Lemma 2.3 Let X be a regular functional space satisfying (Ck ) for some k, 0 < k < . Then, for every open D X, the set { X C, supp D} is dense in CD := { C|supp D}, with respect to the topology of C. Proof Let CD , 0 and let be > 0; by axiom (RF ) there exists X C such that | | < on X. Let := ( T )+ ; using Lema2.2, we get X C. Clearly (x) = 0 (x) = 0 and so CD X and | | < 2 on X. Proposition 2.1 Let X be a regular functional space and let r > 0 be such that G(u) = ||r1 sgnG(u) for arbitrary in R {0} and arbitrary u X .

First AUTHOR, Second AUTHOR

If for some k0 , 0 < k0 < , Tk0 is strongly (weakly) operating, then Tk is weakly (strongly) operating on X for all k, 0 < k < . Proof From Lemma 2.2, we see that condition (Ck ) is satised. We will prove that from (GCk0 ) we get (GCk ) . Let , X , 1 := k0 and k 1 := k0 . Clearly 1 , 1 X . Using (GCk0 ) we can nd u1 in G(1 + k Tk0 1 ) and u2 G(1 ) such that: < u1 u2 , 1 Tk0 1 > 0.
k We nd u G( k0 (1 + Tk0 1 )(= G( + Tk )) and u in G() such that 1 2 k k u = ( k0 )r1 u1 and u = ( k0 )r1 u2 . From these, we get 1 2

< u1 u2 , 1 TK0 >= (

k ? ) < u u , Tk > 0, 1 2 k0

that is (GCk ) . We skip the analogous proof for the case Tk0 is strongly operating.

Potentials

Let X be a regular functional space. We suppose that for every v X we have v + X and that the map v v + is continuous from X with the topology given by the norm, to X with the topology given by the duality with X . Theorem 3.1 ([1, Beurling and Deny]) Let u X and the statements: i) u is a pure potential. ii) there exists u1 in G(u), such that: < u1 , v > 0, v X + . i) u is an absolute pure potential. ii) for every u1 from G(u), we have < u1 , v > 0, v X + . .

The Title of the Article

Then statement i) is equivalent with statement ii) and statement i ) is equivalent with statement ii ). Proof. First, we prove that i) implies ii). From i), we nd u1 G(u) and a positive measure , such that: < u1 , v >= vd, v C X .

Let v X + ; by axiom (RF ) and i), we nd (vn ) a sequence in X + such that vn X C and (vn ) is weakly convergent to v. So, < u1 , v >= limn < u1 , vn >= limn vn d 0, that is statement ii) is true. Now, we prove that ii) implies i). Let be the map v < u1 , v > dened on C X with real values. By lemma 1.1 and axiom (RF ), we can nd a positive measure on X, such that < u1 , v >= d. We prove that i ) implies ii ). From i ), for every u1 G(u), there exists a positive measure on X, such that relation [] is veried. As above, we nd (vn ) a sequence in X C with vn v, in the weak topology, and for every u1 from G(u), we get: < u1 , v >= lim < u, vn >= lim vn d 0

and so ii ). Now, we prove ii ) implies i ). Let u1 be in G(u); we nd a positive measure such that < u1 , v >= vd, v C X and so ii ) implies i ).

From now on, we suppose that T is strongly operating on X . Lemma 3.1 KM (1974-7) Let and be measures associated with some potentials. If is a measure verifying , then is a measure associated with a potential. Proof Let (X C)+ . We have: d d =< u1 , >, u1 G(u )

First AUTHOR, Second AUTHOR

d =< u2 , v >, u2 G(u )

with u1 and u2 from denition 1.3. Each element from C X can be written as: = + . So, d = + d d < u1 , + > + < u2 , > .

Because the map + from X with the topology given by the norm, to X with the weak topology, was supposed continuous, we obtain the continuity with respect to the same topologies of the map . We conclude that the map | < u1 , + > |+| < u2 , > | is continuous from X with the norm topology, to R. We deduce that the map d : C X R

is continuous with respect to the norm topology on X . We nd now u in X such that < u , >= d, C X .

As G is surjective, one can nd u X , such that u G(u). From this relation, we see that is a measure associated with the potential u and this ends the proof. Corollary 3.1 Let , and be measures associated with potential such that and . Then is measure associated with a potential. Denition 3.1 An operator G from X to X , is said to be cyclic monotone if, for every n 1, ui X , u G(ui ), 0 i n, we have: i < u0 un , u > + + < u1 u0 , u > 0. n 0 Denition 3.2 Let the map : X = R {} be convex, lower semicontinuous and not identically . The operator : X = X , dened by := {u X |(u) (v) < u v, u >, v X }, is called the subdierential of . A classical result (Citat Barbu)

The Title of the Article

Theorem 3.2 Let E be a Banach space and A : E E be a multivalued operator. The operator A is the subdierential of a lower semicontinuous convex functional , not identically equal with if and only if A is cyclic maximal monotone operator. Such a functional is called a primitive of A. The maximality in this theorem is given by the order dened in the following denition Denition 3.3 Let E be a Banach space. For A1 , A2 : E E multivalued operators, we write A1 A2 A1 (u) A2 (u), u E. Denition 3.4 A calcic monotone operator G is said to be with (strict) compact primitive if there exists G1 a maximal cyclic monotone extension of G, which has a primitive with the following property: For every K X , closed convex, lower bounded , there exists a (unique) v0 K such that: (PC1) inf{(v); v K} = (v0 ). (PC2) there exists u0 such that < u0 , v v0 > 0, v K. The main theorem of this section is a principle of lower envelope. Theorem 3.3 Let , , be measures associated with potentials such that , and u is an absolute potential. If G(u ) has an element u1 such that B(.) := G(.) u1 has a compact primitive, then u u is a potential. Moreover, we can nd a measure associated with u u which also satises . Proof Let K := {v X |u u v} , let B1 be a cyclic maximal monotone extension of G(.) u1 and let be a compact primitive of B1 . Let v0 K with the property (v) (v0 ), v K and for u0 B1 (v0 ) we have < u0 , v v0 > 0, v K. For every v X + we have v0 + v K and so < u0 , v > 0, v X + . Because G is surjective, using equivalence i) ii) from Theorem 3.1, there exists u0 a pure potential such that u0 G(u0 ). Clearly 0 +

10

First AUTHOR, Second AUTHOR

is a measure associated with the potential v0 and 0 + , which proves the last statement of our theorem. We only need to prove that v0 = u u . As u u K, we get < u0 , v0 v0 u > 0. From (GC ), for every u2 B1 (u (u v0 )+ ), every u3 B1 (v0 u ), every u4 B(u ), such that u3 = u2 u4 , we have: < u3 , v0 v0 u >=< u2 , (u v0 ) >< u4 , (u v0 ) > . If u4 is chosen so that u4 = u5 u1 for some u5 G(u ), then, for 0, X C we have: < u5 , >= d d( ) 0.

As v v + is continuous, we get < u4 , v > 0, v X + . And so: < u4 , (u v0 ) > 0. Let u6 := u0 + u1 , u7 := u3 + u1 . We have: u6 G(v0 ), u7 G(v0 u ). Then < u6 u7 , v0 v0 u >=< u0 u3 , v0 v0 u > 0 and so v0 = v0 u . Because v0 = v0 u K, this implies < u0 , v0 v0 u > 0. As above, we nd u8 G(v0 ), u9 G(v0 u ) such that < u8 u9 , v0 v0 u > 0. From this, using the property of strict monotony of the operator G, we get v0 = v0 u . Because v0 K, we obtain v0 = u u . Corollary 3.2 Let = 0 be a measure associated with a potential, let u, v be pure potentials and let , respectively , positive measures associated with u, respectively v. Suppose that u is an absolute potential. Then u v is a pure potential. Proof From the previous theorem, we nd the measure associated with u v, such that and this ends the proof.

The Title of the Article

11

Theorem 3.4 ( a domination principle) Let , be measures associated to some potentials u , respectively u such that u is a pure potential. Suppose that is a measure associated with an absolute potential and satises , and < u1 u2 , (u u )+ > 0, u1 G(u ), u2 G(u ). Then u u . Proof As in Theorem 3.3, let be a measure associated with u u . As in Theorem 3.1 one can prove the existence of u3 G(u ), u4 G(u ), such that < u3 u4 , v > 0, v X + . From this relation we get: < u3 u4 , (u u )+ > 0. But u1 G(u , u3 G(u ) and G is strictly monotone and these implies u = u and so u u . Corollary 3.3 Let and be measures associated with potentials u , respectively u , such that u is an absolute potential and . If there exists u1 G(u ) such that B(.) := G(.) u1 has a compact primitive, then u u . Proof We take = in Theorem 3.4. Next denition is just a technical one. Denition 3.5 Two measures and satisfy condition () if they satisfy the following statements: (1 ) there exists u absolute potential associated with . (2 ) there exists u1 G(u ) such that B := G u1 has a compact primitive. (3 ) there exists u1 G(u ) such that B := G u1 has a primitive with an absolute minimum in v0 X . (4 ) there exists u1 G(u ) such that (2 ) and (3 ) are both veried by this u1 .

12

First AUTHOR, Second AUTHOR

Corollary 3.4 Let f M+ and let u be a pure potential such that and f d satisfy condition (). If u u on {x X|f (x) > 0}, then uf u . Proof We take = 0 in the theorem 3.4 and notice that there exists u1 G(uf ) such that < u1 , (uf u )+ >= 0. Theorem 3.5 Let k be a nonnegative number such that Tk is strongly operating on X . Let , and be measures associated with potentials, satisfying , . If and satisfy conditions (1 ) and (2 ), then u (u + k) is a potential associated with a measure , such that . Proof First, notice that u (v + k) X , u, v X . We make the notation: w := u (u + k), K := {v X |v w} and dene B : X X , B(v) := G(v) u1 , for u1 given by condition (2 ). As in the proof of the previous theorem, using condition (P C2) of the denition of a compact primitive, we nd v0 K such that there exists u0 B(v0 ) which satises the following condition: < u0 , v v0 >, v K, (1) and v0 is a potential with a measure associated, with and v0 = v0 u = v0 u . We have to prove only that v0 = v0 (u + k). We notice that < u0 , v0 v0 (u + k) > 0, as v0 K and u( + k) K. So v0 (u + k) K. We nd u3 B(v0 (u + k), u4 B(v0 u + Tk (v0 u )+ ) such that < u3 , v0 v0 (u + k) >=, u4 , (v0 u )+ Tk (v0 u )+ > . With (GCk ), we have < u4 , (v0 u )+ Tk (v0 u )+ >< v0 , v0 u0 (u + k) > 0, last inequality is obtained from condition (1).

The Title of the Article

13

From v0 = v0 u , we get u0 B(v0 u ). So, there exists u3 B(v0 (u + k)) such that < u3 , v0 v0 (u + k) > 0. Let be now G(v0 ) u6 = u0 + u1 or u0 = u6 u1 , G(v0 (v + k)) u7 = u3 + u1 or u3 = u7 u1 . From the previous inequality with (4.1) for v = v0 (u +k) we obtain: < u0 u3 , v0 v0 (u + k) > 0 or < u6 u7 , v0 (u + k) > 0. As G is strictly monotone, we get: v0 = v0 (u + k) and this with v0 = v0 (u + k) ends the proof. Corollary 3.5 Let k be such that Tk strongly operates on X . Let u and u be pure potentials such that (1 ) and (2 ) are veried. Then u (u + k) is a pure potential. Theorem 3.6 Let k be such that Tk strongly operates on X . Let and be measures satisfying () and , are associated with potentials.Let us suppose that there exists a measure associated with an absolute potential, such that , and < u1 u2 , (u u k)+ >= 0 for all u1 G(u and u2 G(u ). Then u u + k. Proof As G is strictly monotone, using Theorem T3.5 we follow the proof of Theorem T3.3 . Corollary 3.6 Let k be such that Tk strongly operates on X , f M+ and u a pure potential such that f d and satisfy conditions (1 ) and (2 ). If uf u + k on {x X|f (x) > 0}, then uf u + k.

14

First AUTHOR, Second AUTHOR

References
[1] A. Beurling, J. Deny Dirichlet spaces, Proc. Nat. Acad. Sci. U. S. A. , 45 (1959), 209-215. [2] F. E. Browder, Nonlinear variational inequalities and maximal monotone mappings in Banach spaces. Math. Ann. 183 (1969), 213-233. [3] N. Kenmochi, Y. Mizuta Potential Theoretic Properties of the Gradient of a Convex Function on a Functional Space, Nagoya Math. J. Vol. 59 (1975), 190-215. [4] N. Kenmochi, Y. Mizuta The gradient of a convex function on a regular functional space and its potential theoretic proprieties, Hiroshima Math. J. 4, (1974) 743-763. [5] Y. Mizuta, T. Nagai, Potential Theoretic Properties of the Subdierential of a Convex Function, Hiroshima Math. J. 7 (1977), 177-182. [6] Murali Rao, Zoran Vondracek, Nonlinear Potentials in Function Spaces Nagoya Math. J. Vol 165 (2002), 91-116 [7] B. Calvert, Potential theoretic properties for nonlinear monotone operators, Boll. Un. Ital. 5 (1972), 473-489. [8] R.E. Showalter, Monotone Operators in Banach Space and Nonlinear Partial Dierential Equations, American Mathematical Society, 1997. [9] R. Phelps, Convex functions, monotone operators and Dierentiability, Springer 1993 [10] R. S. Burachik, A. N. Iusem, Set valued mappins and enlargement of monotone operators, Springer 2008. [11] V. Barbu, Nonlinear Semigroups and Dierential Equations in Banach Spaces, Noordho, Leyden 1975.