AAA Abstract and Applied Analysis 1687-0409 1085-3375 Hindawi Publishing Corporation 10.1155/2014/745479 745479 Research Article Perturbation of m -Isometries by Nilpotent Operators http://orcid.org/0000-0002-4034-6060 Bermúdez Teresa 1 Martinón Antonio 1 Müller Vladimir 2 Noda Juan Agustín 1 Başar Feyzi 1 Departamento de Análisis Matemático Universidad de La Laguna La Laguna, 38271 Tenerife Spain ull.es 2 Mathematical Institute, Czech Academy of Sciences, 115 67 Prague Czech Republic 2014 992014 2014 08 04 2014 18 07 2014 9 9 2014 2014 Copyright © 2014 Teresa Bermúdez et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

We prove that if T is an m -isometry on a Hilbert space and Q an n -nilpotent operator commuting with T , then T + Q is a 2 n + m - 2 -isometry. Moreover, we show that a similar result for m ,   q -isometries on Banach spaces is not true.

1. Introduction

The notion of m -isometric operators on Hilbert spaces was introduced by Agler . See also . Recently Sid Ahmed  has defined m -isometries on Banach spaces, Bayart  introduced ( m , q ) -isometries on Banach spaces, and ( m , q ) -isometries on metric spaces were considered in . Moreover, Hoffman et al.  have studied the role of the second parameter q . Recall the main definitions.

A map T : E E ( m 1 integer and q > 0 real), defined on a metric space E with distance d , is called an ( m , q ) -isometry if, for all x , y E , (1) k = 0 m - 1 m - k m k d T k x , T k y q = 0 . We say that T is a strict ( m , q ) -isometry if either m = 1 or T is an ( m , q ) -isometry with m > 1 but is not an ( m - 1 , q ) -isometry. Note that ( 1 , q ) -isometries are isometries.

The above notion of an ( m , q ) -isometry can be adapted to Banach spaces in the following way: a bounded linear operator T : X X , where X is a Banach space with norm · , is an ( m , q ) -isometry if and only if, for all x X , (2) k = 0 m - 1 m - k m k T k x q = 0 .

In the setting of Hilbert spaces, the case q = 2 can be expressed in a special way. Agler  gives the following definition: a linear bounded operator T : H H acting on a Hilbert space H is an ( m , 2 ) -isometry if (3) k = 0 m - 1 m - k m k T * k T k = 0 . ( m , 2 ) -isometries on Hilbert spaces will be called for short m -isometries.

The paper is organized as follows. In the next section we collect some results about applications of arithmetic progressions to m -isometric operators.

In Section 3 we prove that, in the setting of Hilbert spaces, if T is an m -isometry, Q is an n -nilpotent operator, and they commute, and then T + Q is a ( 2 n + m - 2 ) -isometry. This is a partial generalization of the following result obtained in [10, Theorem 2.2]: if T is an isometry and Q is a nilpotent operator of order n commuting with T , then T + Q is a strict ( 2 n - 1 ) -isometry.

In the last section we give some examples of operators on Banach spaces which are of the form identity plus nilpotent, but they are not ( m , q ) -isometries, for any positive integer m and any positive real number q .

Notation. Throughout this paper H denotes a Hilbert space and B ( H ) the algebra of all linear bounded operators on H . Given T B ( H ) , T * denotes its adjoint. Moreover, m 1 is an integer and q > 0 a real number.

2. Preliminaries: Arithmetic Progressions and <inline-formula> <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML" id="M68"> <mml:mo mathvariant="bold">(</mml:mo> <mml:mi>m</mml:mi> <mml:mo mathvariant="bold">,</mml:mo> <mml:mi>q</mml:mi> <mml:mo mathvariant="bold">)</mml:mo></mml:math> </inline-formula>-Isometries

In this section we give some basic properties of m -isometries. We need some preliminaries about arithmetic progressions and their applications to m -isometries. In , some results about this topic are recollected.

Let G be a commutative group and denote its operation by +. Given a sequence a = a n n 0 in G , the difference sequence D a = ( D a ) n 0 is defined by ( D a ) n a n + 1 - a n . The powers of D are defined recursively by D 0 a : = a , D k + 1 a = D ( D k a ) . It is easy to show that (4) D k a n = i = 0 k - 1 k - i k i a i + n , for all k 0 and n 0 integers.

A sequence a in a group G is called an arithmetic progression of order h = 0,1 , 2 , if D h + 1 a = 0 . Equivalently, (5) i = 0 h + 1 - 1 h + 1 - i h + 1 i a i + j = 0 for j = 0,1 , 2 , . It is well known that the sequence a in G is an arithmetic progression of order h if and only if there exists a polynomial p ( n ) in n , with coefficients in G and of degree less than or equal to h , such that p ( n ) = a n , for every n = 0,1 , 2 ; that is, there are γ h , γ h - 1 , , γ 1 , γ 0 G , which depend only on a , such that, for every n = 0,1 , 2 , , (6) a n = p n = i = 0 h γ i n i . We say that the sequence a is an arithmetic progression of strict order h = 0,1 , 2 , if h = 0 or if it is of order h > 0 but is not of order h - 1 ; that is, the polynomial p of (6) has degree h .

Moreover, a sequence a in a group G is an arithmetic progression of order h if and only if, for all n 0 , (7) a n = k = 0 h - 1 h - k n n - 1 n - k n - h k ! h - k ! a k ; that is, (8) a n = k = 0 h - 1 h - k n k n - k - 1 h - k a k .

Now we give a basic result about m -isometries.

Theorem 1.

Let H be a Hilbert space. An operator T B ( H ) is a strict m -isometry if and only if there are A m - 1 0 , A m - 2 , , A 1 , A 0 in B ( H ) , which depend only on T , such that, for every n = 0,1 , 2 , (9) T * n T n = i = 0 m - 1 A i n i ; that is, the sequence T * n T n n 0 is an arithmetic progression of strict order m - 1 in B ( H ) .

Proof.

If T B ( H ) is a strict m -isometry, then it satisfies (3). Hence, for each integer i 0 , (10) k = 0 m - 1 m - k m k T * i T * k T k T i = k = 0 m - 1 m - k m k T * k + i T k + i = 0 , but (11) k = 0 m - 1 - 1 m - 1 - k m - 1 k T * k T k 0 . By (5), the operator sequence ( T * n T n ) n 0 is an arithmetic progression of strict order m - 1 . Therefore, from (6) we obtain that there is a polynomial p ( n ) of degree m - 1 in n , with coefficients in B ( H ) satisfying p ( n ) = T * n T n ; that is, there are operators A m - 1 0 , A m - 2 , , A 1 , A 0 in B ( H ) , such that, for every n = 0,1 , 2 , (12) T * n T n = A m - 1 n m - 1 + A m - 2 n m - 2 + + A 1 n + A 0 .

Conversely, if ( T * n T n ) n 0 is an arithmetic progression of strict order m - 1 , then (10) and (11) hold. Taking i = 0 we obtain (3), so T is a strict m -isometry.

Now we recall an elementary property of ( m , q ) -isometries on metric spaces which will be used in the next sections.

Proposition 2 (see [<xref ref-type="bibr" rid="B7">8</xref>, Proposition 3.11]).

Let E be a metric space and let T : E E be an ( m , q ) -isometry. If T is an invertible strict ( m , q ) -isometry, then m is odd.

3. <inline-formula> <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML" id="M154"> <mml:mrow> <mml:mi>m</mml:mi></mml:mrow> </mml:math></inline-formula>-Isometry Plus <inline-formula> <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML" id="M155"> <mml:mrow> <mml:mi>n</mml:mi></mml:mrow> </mml:math></inline-formula>-Nilpotent

Recall that an operator Q B ( H ) is nilpotent of order n ( n 1 integer), or n -nilpotent, if Q n = 0 and Q n - 1 0 .

In any finite dimensional Hilbert space H , strict m -isometries can be characterized in a very simple way: a linear operator T B ( H ) is a strict m -isometry if and only if m is odd and T = A + Q , where A and Q are commuting operators on H and A is unitary and Q a nilpotent operator of order m + 1 / 2 ([12, page 134] and [10, Theorem 2.7]).

It was proved in [10, Theorem 2.2] that if A B ( H ) is an isometry and Q B ( H ) is an n -nilpotent operator such that T Q = Q T , then T + Q is a strict ( 2 n - 1 ) -isometry. Now we obtain a partial generalization of this result: if T B ( H ) is an m -isometry and Q B ( H ) is an n -nilpotent operator commuting with T , then T + Q is a ( 2 n + m - 2 ) -isometry. However, T + Q is not necessarily a strict ( 2 n + m - 2 ) -isometry. For example, if T is an isometry and Q any n -nilpotent operator ( n > 1 ) such that T Q = Q T , then T = T + Q + ( - Q ) is not a strict ( 4 n - 3 ) -isometry.

Theorem 3.

Let H be a Hilbert space. Let T B ( H ) be an m -isometry and Q B ( H ) an n -nilpotent operator ( n 1 integer) such that T Q = Q T . Then T + Q is ( 2 n + m - 2 ) -isometry.

Proof.

Fix an integer k 0 and denote h min { k , n - 1 } . Then we have (13) T + Q * k T + Q k = i = 0 h k i Q * i T * k - i j = 0 h k j T k - j Q j = i , j = 0 h k i k j Q * i T * k - i T k - j Q j = 0 i < j h k i k j Q * i T * j - i T * k - j T k - j Q j WW + 0 j i h k i k j Q * i T * k - i T k - i T i - j Q j . From (9) we obtain, for certain A m - 1 , , A 0 B ( H ) , (14) T + Q * k T + Q k = 0 i < j h k i k j Q * i T * j - i r = 0 m - 1 A r k - j r Q j + 0 j i h k i k j Q * i r = 0 m - 1 A r k - i r T i - j Q j . Write (15) B r , i , j Q * i T * j - i A r Q j B H , C r , i , j Q * i A r T i - j Q j B H , q r , i , j k i k j k - j r , p r , i , j k i k j k - i r . Note that k i and k j are real polynomials in k of degree less than or equal to h n - 1 , and ( k - j ) r and ( k - i ) r have degree r m - 1 . Hence q r , i , j and p r , i , j are real polynomials of degree less than or equal to m - 1 + 2 ( n - 1 ) = 2 n + m - 3 . Consequently we can write (16) T + Q * k T + Q k = r = 0 m - 1 0 i < j h B r , i , j q r , i , j + r = 0 m - 1 0 j i h C r , i , j p r , i , j , which is a polynomial in k , of degree less than or equal to 2 n + m - 3 with coefficients in B ( H ) . By Theorem 1, the operator T + Q is an ( 2 n + m - 2 ) -isometry.

For isometries it is possible to say more [10, Theorem 2.2].

Theorem 4.

Let H be a Hilbert space. Let T B ( H ) be an isometry and let Q B ( H ) be an n -nilpotent operator ( n 1 integer) such that T Q = Q T . Then T + Q is a strict ( 2 n - 1 ) -isometry.

Proof.

By Theorem 3 we obtain that T + Q is a ( 2 n - 1 ) -isometry; that is, ( ( T + Q ) * k ( T + Q ) k ) k 0 is an arithmetic progression of order less than or equal to 2 n - 2 . Now we prove that it is an arithmetic progression of strict order 2 n - 2 , or equivalently the polynomial (9) has degree 2 n - 2 . Note that as T is an isometry we have T * k T k = I , for every positive integer k .

As in the proof of Theorem 3, for any integer k 0 , we have that (17) T + Q * k T + Q k = i , j = 0 h k i k j Q * i T * k - i T k - j Q j = 0 i < j h k i k j Q * i T * j - i Q j + 0 j i h k i k j Q * i T i - j Q j , where h min { k , n - 1 } .

The coefficient of k 2 n - 2 in the polynomial ( T + Q ) * k ( T + Q ) k is (18) 1 n - 1 ! 2 Q * n - 1 Q n - 1 , which is null if and only if Q * n - 1 Q n - 1 = 0 , that is, if and only if Q n - 1 = 0 . Therefore, if Q is nilpotent of order n , then ( T + Q ) * k ( T + Q ) k can be written as a polynomial in k , of degree 2 n - 2 and coefficients in B ( H ) . Consequently T + Q is a strict ( 2 n - 1 ) -isometry.

Now we obtain the following corollary of Theorem 4.

Corollary 5.

Let H be a Hilbert space. Let Q B ( H ) be an n -nilpotent operator ( n 1 integer). Then I + Q is a strict ( 2 n - 1 ) -isometry.

Recall that an operator T B ( H ) is N -supercyclic ( N 1 integer) if there exists a subspace F H of dimension N such that its orbit { T n x : n 0 , x F } is dense in H . Moreover, T is called supercyclic if it is 1 -supercyclic. See [13, 14].

Bayart [7, Theorem 3.3] proved that on an infinite dimensional Banach space an ( m , q ) -isometry is never N -supercyclic, for any N 1 . In the setting of Banach spaces, Yarmahmoodi et al. [15, Theorem 2.2] showed that any sum of an isometry and a commuting nilpotent operator is never supercyclic. For Hilbert space operators we extend the result [15, Theorem 2.2] to m -isometries plus commuting nilpotent operators.

Corollary 6.

Let H be an infinite dimensional Hilbert space. If T B ( H ) is an m -isometry that commutes with a nilpotent operator Q , then T + Q is never N -supercyclic for any N .

4. Some Examples in the Setting of Banach Spaces

Theorem 4 is not true for finite-dimensional Banach spaces even for m = 1 .

Denote l p d ( C d , · p ) .

Example 1.

Let Q : C 2 C 2 be defined by Q ( x , y ) ( y , 0 ) ; hence Q is a 2 -nilpotent operator. The following assertions hold:

I + Q is not a ( 3 , p ) -isometry on l p 2 for any 1 p < and p 2 ;

I + Q is not a ( 3 , p ) -isometry on l 2 for any p > 0 ;

I + Q is a strict ( 2 k + 1,2 k ) -isometry on ( C 2 , · 2 k ) for any k = 1,2 , 3 , .

Proof.

For ( x , y ) C 2 we have (19) I + Q x , y = x + y , y , I + Q 2 x , y = x + 2 y , y , I + Q 3 x , y = x + 3 y , y . Write (20) A x , y ; p , q I + Q 3 x , y p q - 3 I + Q 2 x , y p q + 3 I + Q x , y p q - x , y p q .

(1) We consider two cases: 1 < p < and p = 1 .

Case 1 < p < . For x = 0 , y = 1 , and q = p , we have (21) A 0,1 ; p , p = 3 p + 1 - 3 · 2 p - 3 + 6 - 1 = 3 p - 3 · 2 p + 3 .

So A ( 0,1 ; p , p ) = 0 if and only if 3 p - 1 + 1 = 2 p , which is true only when p = 2 or p = 1 since the function f ( t ) = 3 t - 1 + 1 - 2 t is null only for t = 1 and t = 2 .

Consequently I + Q is not a ( 3 , p ) -isometry on l p 2 if p 2 and 1 < p < .

Case p = 1 . In order to prove that I + Q is not a ( 3,1 ) -isometry on l 1 2 , we take the vector ( 1 , - 1 ) and obtain that (22) A 1 , - 1 ; 1,1 = I + Q 3 1 , - 1 1 - 3 I + Q 2 1 , - 1 1 + 3 I + Q 1 , - 1 1 - 1 , - 1 1 0 .

(2) For ( x , y ) C 2 we have (23) A x , y ; , p I + Q 3 x , y p - 3 I + Q 2 x , y p + 3 I + Q x , y p - x , y p = max x + 3 y , y p - 3 max x + 2 y , y p + 3 max x + y , y p - max x , y p . In particular, for x : = 1 and y : = - 1 , (24) A 1 , - 1 ; , p = 2 p - 1 0 . Therefore I + Q is not a ( 3 , p ) -isometry on l 2 for any p > 0 .

(3) First we prove by induction on k that I + Q is a ( 2 k + 1,2 k ) -isometry on l 2 k 2 for any k = 1 , 2 , 3 . Note that, for ( x , y ) C 2 , (25) I + Q s x , y = x + s y , y , s = 0,1 , 2 . By Corollary 5, the operator I + Q is a strict ( 3,2 ) -isometry on l 2 2 . Hence I + Q is a strict ( 2 k + 1,2 k ) -isometry on l 2 2 for all k = 1 , 2 , 3 [9, Corollary 4.6]. Thus for ( x , y ) C 2 , (26) s = 0 2 k + 1 - 1 2 k + 1 - s 2 k + 1 s x + s y 2 + y 2 k = 0 . Suppose that I + Q is a ( 2 i - 1,2 i - 2 ) -isometry on l 2 i - 2 2 for every i = 2,3 , , k . Hence I + Q is also a ( 2 k + 1,2 i - 2 ) -isometry on l 2 i - 2 2 . Then, for ( x , y ) C 2 , (27) s = 0 2 k + 1 - 1 2 k + 1 - s 2 k + 1 s x + s y 2 i - 2 + y 2 i - 2 = 0 , m m m m m m m m m m m m m m m m m i m m 2 i k . Therefore, (28) s = 0 2 k + 1 - 1 2 k + 1 - s 2 k + 1 s x + s y 2 i - 2 = 0 , m m m m m m m m m m m m m m 2 i k . Taking into account equality (28) we can write (26) in the following way: (29) 0 = s = 0 2 k + 1 - 1 2 k + 1 - s 2 k + 1 s i = 0 k k i x + s y 2 i y 2 k - i = i = 0 k - 1 k i y 2 k - i s = 0 2 k + 1 - 1 2 k + 1 - s 2 k + 1 s x + s y 2 i + s = 0 2 k + 1 - 1 2 k + 1 - s 2 k + 1 s x + s y 2 k = s = 0 2 k + 1 - 1 2 k + 1 - s 2 k + 1 s x + s y 2 k + y 2 k . Therefore I + Q is a ( 2 k + 1,2 k ) -isometry on l 2 k 2 .

Now we prove that I + Q is a strict ( 2 k + 1,2 k ) -isometry on l 2 k 2 . Suppose on the contrary that I + Q is a ( 2 k , 2 k ) -isometry on l 2 k 2 . Then, (30) s = 0 2 k - 1 - 1 2 k - 1 - s 2 k - 1 s x + s y 2 k + y 2 k = 0 for all ( x , y ) C 2 . So (31) s = 0 2 k - 1 - 1 2 k - 1 - s 2 k - 1 s x + s y 2 k = 0 for all ( x , y ) C 2 . In particular, for y = 1 and x = 0,1 , 2 , , we have (32) s = 0 2 k - 1 - 1 2 k - 1 - s 2 k - 1 s x + s 2 k = 0 . So ( s 2 k ) s = 0 is an arithmetic progression of order 2 k - 2 , which is a contradiction with (6).

Remark 7.

Notice that, in any Hilbert space of dimension n , there are strict m -isometries only for any m 2 n - 1 . However, as the above example shows, there are strict ( 2 k + 1,2 k ) -isometries for any integer k in a Banach space of dimension 2.

The following example gives an operator of the form I + Q with Q a nilpotent operator such that I + Q is not an ( m , q ) -isometry for any integer m and any q > 0 .

Example 2.

Let X be the Banach space of all real continuous functions f on [ 0,1 ] such that f ( 1 ) = 0 endowed with the supremun norm. Define Q : X X by (33) Q f t f t + 1 2 , if 0 t 1 2 , 0 , if 1 2 < t 1 . Then Q B ( X ) is 2 -nilpotent operator. Moreover, I + Q is not an ( m , q ) -isometry for any m = 1,2 , 3 , and any q > 0 .

Proof.

It is clear that I + Q is not an isometry since the function f X given by (34) f t 1 , if 0 t 1 2 , - 2 t + 2 , if 1 2 < t 1 satisfies f = 1 and ( I + Q ) f = 2 .

For m = 2,3 , 4 , consider the function f m X defined by (35) f m t - 4 t + 1 , if 0 t 1 4 , 0 , if 1 4 < t 1 2 , - 4 m - 1 t + 2 m - 1 , if 1 2 < t 3 4 , 4 m - 1 t - 4 m - 1 , if 3 4 < t 1 .

Note that f m ( 3 / 4 ) = 1 / 1 - m = min 0 t 1 f m ( t ) (Figure 1).

Graphics of functions f 3 , f 5 , and f 7 .

Fix q > 0 . For k = 0,1 , 2 , , we have (36) I + Q k f m q = I + k Q f m q = sup 0 t 1 f m t + k Q f m t q .

If 0 k m - 1 , then (37) I + Q k f m q = f m 0 + k f m 1 2 q = 1 , since k 1 / m - 1 1 . But as m 1 / m - 1 > 1 we obtain (38) I + Q m f m q = f m 1 4 + m f m 3 4 q = m m - 1 q > 1 . Consequently, (39) k = 0 m - 1 m - k m k I + Q k f m q = k = 0 m - 1 - 1 m - l m k + I + Q m f m q = - 1 + m m - 1 q 0 . Therefore I + Q is not an ( m , q ) -isometry for any m = 1,2 , 3 and any q > 0 .

Disclosure

After submitting this paper for publication we received from Le and Gu et al. the papers [16, 17], in which they obtained (independently) Theorem 3. Their arguments are different from ours, using the Hereditary Functional Calculus.

Conflict of Interests

The authors declare that there is no conflict of interests regarding the publication of this paper.

Acknowledgments

The first author is partially supported by Grant of Ministerio de Ciencia e Innovación, Spain, Project no. MTM2011-26538. The third author was supported by Grant no. 14-07880S of GA ČR and RVO:67985840.

Agler J. A disconjugacy theorem for Toeplitz operators The American Journal of Mathematics 1990 112 1 1 14 10.2307/2374849 MR1037599 Richter S. Invariant subspaces of the Dirichlet shift Journal für die Reine und Angewandte Mathematik 1988 386 205 220 10.1515/crll.1988.386.205 MR936999 Agler J. Stankus M. m -isometric transformations of Hilbert space Integral Equations and Operator Theory 1995 21 4 383 429 10.1007/BF01222016 MR1321694 2-s2.0-0000816441 Agler J. Stankus M. m -isometric transformations of Hilbert space. II Integral Equations and Operator Theory 1995 23 1 1 48 10.1007/BF01261201 MR1346617 2-s2.0-0039144682 Agler J. Stankus M. m -isometric transformations of Hilbert space. III Integral Equations and Operator Theory 1996 24 4 379 421 10.1007/BF01191619 MR1382018 Sid Ahmed O. A. M. m -isometric operators on Banach spaces Asian-European Journal of Mathematics 2010 3 1 1 19 10.1142/S1793557110000027 MR2654806 ZBL1197.47008 2-s2.0-80255130177 Bayart F. m -isometries on Banach spaces Mathematische Nachrichten 2011 284 17-18 2141 2147 10.1002/mana.200910029 MR2859754 ZBL1230.47018 2-s2.0-81155124944 Bermúdez T. Martinón A. Müller V. ( m , q ) -isometries on metric spaces The Journal of Operator Theory 2014 Hoffmann P. Mackey M. Searcóid M. On the second parameter of an ( m , p ) -isometry Integral Equations and Operator Theory 2011 71 3 389 405 10.1007/s00020-011-1905-0 MR2852193 2-s2.0-80255139955 Bermúdez T. Martinón A. Noda J. An isometry plus a nilpotent operator is an m -isometry. Applications Journal of Mathematical Analysis and Applications 2013 407 2 505 512 10.1016/j.jmaa.2013.05.043 MR3071119 2-s2.0-84879788779 Bermúdez T. Martinón A. Noda J. Arithmetic progressions and its applications to (m, q)-isometries: a survey Preprint Agler J. Helton J. W. Stankus M. Classification of hereditary matrices Linear Algebra and Its Applications 1998 274 125 160 10.1016/S0024-3795(97)00307-8 MR1612001 2-s2.0-0042355460 Hilden H. M. Wallen L. J. Some cyclic and non-cyclic vectors of certain operators Indiana University Mathematics Journal 1973-1974 23 557 565 MR0326452 Feldman N. S. n -supercyclic operators Studia Mathematica 2002 151 2 141 159 10.4064/sm151-2-3 MR1917229 ZBL1006.47008 2-s2.0-0036330471 Yarmahmoodi S. Hedayatian K. Yousefi B. Supercyclicity and hypercyclicity of an isometry plus a nilpotent Abstract and Applied Analysis 2011 2011 11 686832 10.1155/2011/686832 MR2800079 2-s2.0-80052657543 Gu C. Stankus M. m-isometries and n-symmetries: products and sums with a nilpotent operators preprint Le T. Algebraic properties of operators roots of polynomials The Journal of Mathematical Analysis and Applications. In press