SpringerOpen Newsletter

Receive periodic news and updates relating to SpringerOpen.

This article is part of the series Proceedings of the International Congress in Honour of Professor Hari M. Srivastava.

Open Access Research

An estimate of Sumudu transforms for Boehmians

Shrideh Khalaf Qasem Al-Omari1 and Adem Kılıçman2*

Author Affiliations

1 Department of Applied Sciences, Faculty of Engineering Technology, Al-Balqa Applied University, Amman, 11134, Jordan

2 Department of Mathematics and Institute of Mathematical Research, Universiti Putra Malaysia (UPM), 43400 UPM, Serdang, Selangor, Malaysia

For all author emails, please log on.

Advances in Difference Equations 2013, 2013:77  doi:10.1186/1687-1847-2013-77


The electronic version of this article is the complete one and can be found online at: http://www.advancesindifferenceequations.com/content/2013/1/77


Received:14 January 2013
Accepted:5 March 2013
Published:26 March 2013

© 2013 Al-Omari and Kılıçman; licensee Springer

This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

Abstract

The space of Boehmians is constructed using an algebraic approach that utilizes convolution and approximate identities or delta sequences. A proper subspace can be identified with the space of distributions. In this paper, we first construct a suitable Boehmian space on which the Sumudu transform can be defined and the function space S can be embedded. In addition to this, our definition extends the Sumudu transform to more general spaces and the definition remains consistent for S elements. We also discuss the operational properties of the Sumudu transform on Boehmians and finally end with certain theorems for continuity conditions of the extended Sumudu transform and its inverse with respect to δ- and Δ-convergence.

MSC: 54C40, 14E20, 46E25, 20C20.

Keywords:
Sumudu transforms; Boehmian spaces; the space H ( Y ) ; the space H ( Y s ) ; distributions

1 Introduction

The Sumudu transform of one variable function f ( x ) is introduced as a new integral transform by Watugala in [1] and is given by

S f ( t ) ( y ) = 1 y R + f ( t ) exp ( t y ) d t , y ( τ 1 , τ 2 )

over the set of functions

A = { f ( t ) : M , τ 1 , τ 2 > 0 , | f ( t ) | < M e t τ j , t ( 1 ) j × ( 0 , ) } ,

where f ( t ) is a function that can be expressed as a convergent series [2,3]. The Sumudu transform was applied to solve the ordinary differential equations in control engineering problems; see [3].

The Sumudu transform of the convolution product of f and u is given by

S ( f u ) ( y ) = y f s ( y ) u s ( y ) ,

where f s and u s are the Sumudu transforms of f and u, respectively.

Some of the properties were established by Weerakoon in [4,5]. In [6], further fundamental properties of this transform were also established by Asiru. Similarly, this transform was applied to a one-dimensional neutron transport equation in [7] by Kadem.

In [8], the Sumudu transform was extended to the distributions and some of their properties were also studied. Recently, this transform has been applied to solve the system of differential equations; see Kılıçman et al. in [9].

Note that a very interesting fact about Sumudu transform is that the original function and its Sumudu transform have the same Taylor coefficients except the factor n; see Zhang [10]. Similarly, the Sumudu transform sends combinations C ( m , n ) into permutations P ( m , n ) , and hence it will be useful in the discrete systems.

The following are the general properties of the Sumudu transform which are auxiliary from the substitution method and the properties of integral operators.

(i) If k 1 and k 2 are non-negative integers and S 1 and S 2 are the corresponding Sumudu transforms of f 1 and f 2 , respectively, then

S ( k 1 f 1 + k 2 f 2 ) ( y ) = k 1 S 1 ( y ) + k 2 S 2 ( y ) .

(ii) S f ( k t ) ( y ) = S ( k y ) , k R + .

(iii) lim t 0 f ( t ) = lim u 0 S ( y ) = f ( 0 ) , where S ( y ) is the Sumudu transform off.

More properties of the Sumudu transforms a long with a some of applications were given in [11] and [12].

2 Boehmian space

Boehmians were first constructed as a generalization of regular Mikusinski operators [13]. The minimal structure necessary for the construction of Boehmians consists of the following elements:

(i) a nonempty set A ;

(ii) a commutative semigroup ( B , ) ;

(iii) an operation : A × B A such that for each x A and s 1 , s 2 , B , x ( s 1 s 2 ) = ( x s 1 ) s 2 ;

(iv) a collection Δ B N such that

(a) If x , y A , ( s n ) Δ , x s n = y s n for all n, then x = y ;

(b) If ( s n ) , ( t n ) Δ , then ( s n t n ) Δ .

Elements of Δ are called delta sequences. Consider

Q = { ( x n , s n ) : x n A , ( s n ) Δ , x n s m = x m s n , m , n N } .

Now if ( x n , s n ) , ( y n , t n ) Q , x n t m = y m s n , m , n N , then we say ( x n , s n ) ( y n , t n ) . The relation ∼ is an equivalence relation in ℚ. The space of equivalence classes in ℚ is denoted by β. Elements of β are called Boehmians.

We note that between A and β there is a canonical embedding expressed as x x s n s n . The operation ⊙ can also be extended to β × A by x n s n t = x n t s n . The relationship between the notion of convergence and the product ⊙ is given by:

(i) If f n f as n in A and ϕ B is any fixed element, then f n ϕ f ϕ in A (as n );

(ii) If f n f as n in A and ( δ n ) Δ , then f n δ n f in A (as n ).

The operation ⊙ can be extended to β × B as follows: If [ f n s n ] β and ϕ B , then [ f n s n ] ϕ = [ f n ϕ s n ] . In β, there are two types of convergence as follows.

(1) A sequence ( h n ) inβis said to beδ-convergent tohinβ, denoted by h n δ h , if there exists ( s n ) Δ such that ( h n s n ) , ( h s n ) A , k , n N , and ( h n s k ) ( h s k ) as n in A for every k N .

(2) A sequence ( h n ) inβis said to be Δ-convergent tohinβ, denoted by h n Δ h , if there exists a ( s n ) Δ such that ( h n h ) s n A , n N , and ( h n h ) s n 0 as n in A .

For further discussion, see [14-16].

3 The Boehmian space H ( Y )

Denote by S + ( R ) and D + ( R ) the space of all rapidly decreasing functions over R + ( R + = ( 0 , ) ) and the space of all test functions of compact support, respectively. In what follows, we obtain preliminary results required to construct the Boehmian space H ( Y ) , where Y = ( S + , D + , Δ + ) .

Lemma 3.1

(1) If u 1 , u 2 D + ( R ) , then u 1 u 2 D + ( R ) .

(2) If f 1 , f 2 S + ( R ) and u 1 D + ( R ) , then ( f 1 + f 2 ) u 1 = f 1 u 1 + f 2 u 1 .

(3) u 1 u 2 = u 2 u 1 , u 1 , u 2 D + ( R ) .

(4) If f S + , u 1 , u 2 D + ( R ) , then ( f u 1 ) u 2 = f ( u 1 u 2 ) .

Proofs are analogous to those of classical cases and details are omitted.

Definition 3.2 A sequence ( s n ) of functions from D + ( R ) is said to be in Δ + if and only if

Δ + 1 R + s n ( x ) d x = 1 ; Δ + 2 R + | s n ( x ) | d x M , M  is a positive integer ; Δ + 3 supp s n ( x ) ( 0 , ε n ) , ε n 0  as  n .

This means that ( s n ) shrinks to zero as n . Each member of Δ + is called a delta sequence or an approximate identity or, sometimes, a summability kernel. Delta sequences, in general, appear in many branches of mathematics, but probably the most important applications are those in the theory of generalized functions. The basic use of delta sequences is the regularization of generalized functions, and further, they can be used to define the convolution product and the product of generalized functions.

Lemma 3.3If ( s n ) , ( t n ) Δ + , then supp ( s n t n ) supp s n + supp t n .

Lemma 3.4If u 1 , u 2 D + ( R ) , then so is u 1 u 2 and

R + | u 1 u 2 | R + | u 1 | R + | u 2 | .

Theorem 3.5Let f 1 , f 2 S + ( R ) and ( s n ) Δ + such that f 1 s n = f 2 s n , n = 1 , 2 , 3 ,  , then f 1 = f 2 in S + ( R ) .

Proof We show that f 1 s n = f 1 in S + ( R ) . Let K be a compact set containing the supp s n for every n N . Using Δ + 1 , we write

| x k D m ( f 1 s n f 1 ) ( x ) | K | s n ( t ) | | x k D m ( f 1 ( x t ) f 1 ( x ) ) | d t . (3.1)

The mapping t f 1 t , where f 1 t ( x ) = f 1 ( x t ) , is uniformly continuous from R + R + . From the hypothesis that supp s n 0 as n (by Δ + 3 ), we choose r > 0 such that supp s n [ 0 , r ] for large n and t < r . This implies

| f 1 ( x t ) f 1 ( x ) | = | f 1 t f 1 | < ε n M . (3.2)

Hence using Δ + 2 and (3.2), (3.1) becomes

| x k D m ( f 1 s n f 1 ) ( x ) | < ε n 0 as  n .

Thus f 1 s n f 1 in S + ( R ) . Similarly, we show that f 2 s n f 2 . This completes the proof of the theorem. □

Theorem 3.6If lim n f n = f in S + ( R ) and u D + ( R ) , then

lim n f n u = f u .

Proof

In view of the hypothesis of the theorem, we write

| x k D m ( f n u f u ) ( x ) | = | x k ( D m ( f n f ) u ) ( x ) | . (3.3)

The last equation follows from the fact that [17]

D m f u = D m f u = f D m u .

Hence, for each u D + ( R ) , we have

| x k D m ( f n u f u ) ( x ) | K x k | D m ( f n f ) ( x t ) | | u ( t ) | d t M γ k ( f n f ) for some constant  M 0 as  n .

The proof of the theorem is completed. □

Theorem 3.7If lim n f n = f in S + ( R ) and ( s n ) Δ + , then lim n f n s n = f .

Proof

In view of the analysis employed for Theorem 3.5, we get

lim n f n s n = f n f as  n .

Hence

lim n f n s n = f as  n .

This completes the proof. The Boehmian space H ( Y ) is therefore constructed. □

The canonical embedding between S + ( R ) and H ( Y ) is expressed as x [ x s n s n ] . The extension of ⋆ to H ( Y ) × S + is given by [ x n s n ] t = [ x n t s n ] . Convergence in H ( Y ) is defined in a natural way:

δ-convergence: A sequence ( h n ) in H ( Y ) is said to beδ-convergent tohin H ( Y ) , denoted by h n δ h , if there exists a delta sequence ( s n ) such that ( h n s n ) , ( h s n ) S + ( R ) , k , n N , and ( h n s k ) ( h s k ) as n in S + ( R ) for every k N .

Δ + -convergence: A sequence ( h n ) in H ( Y ) is said to be Δ + -convergent tohin H ( Y ) , denoted by h n Δ h , if there exists a ( s n ) Δ + such that ( h n h ) s n S + ( R ) , n N , and ( h n h ) s n 0 as n in S + ( R ) .

Theorem 3.8The mapping f [ f s n s n ] is a continuous embedding of S + ( R ) into H ( Y ) .

Proof The mapping is one-to-one. For detailed proof, let [ f 1 s n s n ] = [ f 2 t n t n ] , then ( f 1 s n ) t m = ( f 2 t m ) s n . Then since ( s n ) , ( t n ) Δ + , f 1 ( s m t n ) = f 2 ( t n s m ) = f 2 ( s m t n ) . Using Theorem 3.5, we get f 1 = f 2 . To show the mapping is continuous, let f n 0 as n in S + ( R ) . Then we show that

[ f n s m s m ] δ 0 as  n .

From Theorem 3.5, [ f n s m s m ] s m = f n s m 0 as n . This completes the proof of the theorem. □

Theorem 3.9Let f S + ( R ) and u D + ( R ) , then

S ( f u ) ( y ) = y f s ( y ) u s ( y ) .

4 The Boehmian space H ( Y s )

We describe another Boehmian space as follows. Let S + ( R ) be the space of rapidly decreasing functions [17]. Define

D + s ( R ) = { u s :  for all  u D + ( R ) } , (4.1)

where u s denotes the Sumudu transform of u. We also define f u s by

( f u s ) ( y ) = y f ( y ) u s ( y ) . (4.2)

Lemma 4.1Let f S + ( R ) and u s D + s ( R ) , then f u s S + ( R ) .

Proof If f S + ( R ) and u s D + s ( R ) , then using the topology of S + ( R ) and Leibnitz’ theorem, we get

| x k D x m ( f u s ) ( x ) | | x k j = 1 m D m j ( x f ( x ) ) D j u s ( x ) | j = 1 m | x k D m j ( x f ( x ) ) | | D j u s ( x ) | = j = 1 m | x k D m j f 1 ( x ) | | K u ( t ) D x j e t x x d t | ,

where f 1 ( x ) = x f ( x ) S + ( R ) and K is a compact subset containing the supp u ( t ) . Hence

| x k D x m ( f u s ) ( x ) | M γ k , m j ( f 1 ) <

for some positive constant M. This completes the proof of the lemma. □

Lemma 4.2The mapping

S + × D + s S + , ( f , u s ) f u s

satisfies the following properties:

(1) If u 1 s , u 2 s D + s ( R ) , then u 1 s u 2 s D + s ( R ) .

(2) If f 1 , f 2 S + ( R ) , u s D + s ( R ) , then ( f 1 + f 2 ) u s = f 1 u s + f 2 u s .

(3) For u 1 s , u 2 s D + s ( R ) , u 1 s u 2 s = u 2 s u 1 s .

(4) For f S + ( R ) , u 1 s , u 2 s D + s ( R ) , then ( f u 1 s ) u 2 s = f ( u 1 s u 2 s ) .

Proof The proof of the above lemma is straightforward. Detailed proof is as follows.

Proof of (1). Let u 1 , u 2 D + ( R ) , then u 1 u 2 D + ( R ) . Hence ( u 1 u 2 ) s D + s ( R ) by (4.1). Theorem 3.9 implies u 1 s u 2 s D + s ( R ) .

Proof of (2) is obvious.

Proof of (3). We have

( u 1 s u 2 s ) ( x ) = x u 1 s ( x ) u 2 s ( x ) = x u 2 s ( x ) u 1 s ( x ) = ( u 2 s u 1 s ) ( x ) .

Hence u 1 s u 2 s = u 2 s u 1 s .

Proof of (4). Let f S + ( R ) , u 1 s , u 2 s D + s ( R ) , then

( ( f u 1 s ) u 2 s ) ( x ) = x ( f u 1 s ) ( x ) u 2 s = x x f ( x ) u 1 s ( x ) u 2 s ( x ) = x f ( x ) x u 1 s ( x ) u 2 s ( x ) = x f ( x ) ( u 1 s u 2 s ) ( x ) = f ( u 1 s u 2 s ) ( x ) ,

that is,

( f u 1 s ) u 2 s = f ( u 1 s u 2 s ) .

This completes the proof of the theorem. □

Denote by Δ + s the set of all Sumudu transforms of delta sequences from Δ + . That is,

Δ + s = { ( s n s ) : ( s n ) Δ + , n N } . (4.3)

Lemma 4.3Let f 1 , f 2 S + ( R ) , ( s n s ) Δ + s be such that f 1 s n s = f 2 s n s , ∀n, then f 1 = f 2 in S + ( R ) .

Proof Let f 1 , f 2 S + ( R ) and ( s n s ) Δ + s . Since f 1 s n s = f 2 s n s , (4.2) implies x f 1 ( x ) s n s ( x ) = x f 2 ( x ) s n s ( x ) . Hence f 1 ( x ) = f 2 ( x ) for all x. The proof is completed. □

Lemma 4.4For each ( s n ) , ( t n ) Δ + , ( s n s t n s ) Δ + s .

Proof Since ( s n ) , ( t n ) Δ + , s n t n Δ + for all n. Hence, from Theorem 3.9, we get S ( s n t n ) ( x ) = x s n s ( x ) t n s ( x ) = s n s t n s Δ + s for every n. This completes the proof of the lemma. □

By aid of Lemma 4.3. and Lemma 4.4, Δ + s can be regarded as a delta sequence.

Lemma 4.5Let f n f in S + ( R ) , u s D + s ( R ) , then f n u s f u s in S + ( R ) .

Proof It is clear that u s is bounded in D + s ( R ) . Further,

( f n u s ) ( x ) x f ( x ) u s ( x ) ( f n u s ) ( x ) .

Hence ( f n u s ) f u s . □

Lemma 4.6Let f n f in S + ( R ) , ( s n s ) Δ + s , then f n s n s f in S + ( R ) .

Proof Let ( s n ) Δ + , then s n s ( x ) 1 x uniformly on compact subsets of R + . Hence

| x k D x m ( f n s n s f ) ( x ) | = | x k D x m ( x f n ( x ) s n s ( x ) f ( x ) ) | | x k D x m ( f n f ) ( x ) |

as n . Thus | x k D x m ( f n s n s f ) ( x ) | 0 as n . This yields f n s n s f in the topology of S + ( R ) . The proof is therefore completed. The space H ( Y s ) can be regarded as a Boehmian space, where Y s = ( S + , D + s , Δ + s ) . □

Lemma 4.7The mapping

f [ f s n s s n s ] (4.4)

is a continuous embedding of S + ( R ) into H ( Y s ) .

Proof For f S + ( R ) , s n s Δ + s , f s n s s n s is a quotient of sequences in the sense that ( f s n s ) s m s = f ( s m s s n s ) . We show that the map (4.4) is one-to-one. Let [ f 1 s n s s n s ] = [ f 2 t n s t n s ] , then ( f 1 s n s ) t m s = ( f 2 t m s ) s n s , m , n N . Using Lemma 4.2 and Lemma 4.3, we conclude f 1 = f 2 . □

To establish the continuity of (4.4), let f n 0 as n in S + ( R ) . Then f n s n s 0 as n by Lemma 4.6, and hence

[ f n s n s s n s ] 0

as n in H ( Y s ) . This completes the proof of the lemma.

5 The Sumudu transform of Boehmians

Let β = [ f n s n ] H ( Y ) , then we define the Sumudu transform of β by the relation

β 1 s = [ f n s s n s ] in  H ( Y s ) . (5.1)

Theorem 5.1 β 1 s : H ( Y ) H ( Y s ) is well defined.

Proof Let β 1 = β 2 H ( Y ) , where β 1 = [ f n s n ] , β 2 = [ g n t n ] . Then the concept of quotients yields f n t m = g m s n . Employing Theorem 3.9, we get x f n s ( x ) t m s ( x ) = x g m s ( x ) s n s ( x ) , i.e., f n s t m s = g m s s n s . Equivalently, f n s s n s g n s t n s . Thus β 1 s = β 2 s . This completes the proof of the theorem. □

Theorem 5.2 β s : H ( Y ) H ( Y s ) is continuous with respect toδ-convergence.

Proof Let β n 0 in H ( Y ) , then by [14], β n = [ f n , k s k ] and f n , k 0 as n in S + ( R ) . Applying the Sumudu transform to both sides yields f n , k s 0 as n . Hence

β n s = [ f n , k s s k s ] 0

as n in H ( Y s ) . This proves the theorem. □

Theorem 5.3 β s : H ( Y ) H ( Y s ) is a one-to-one mapping.

Proof Assume β 1 s = [ f n s s n s ] = [ g n s t n s ] = β 2 s , then f n s t m s = g m s s n s . Hence

( f n t m ) s = ( g m s n ) s .

Since the Sumudu transform is one-to-one, we get f n t m = g m s n . Thus

f n s n g n t n .

Hence

[ f n s n ] = β 1 = [ g n t n ] = β 2 .

This completes the proof of the theorem. □

Theorem 5.4Let β 1 , β 2 H ( Y ) , then

(1) ( β 1 + β 2 ) s = β 1 s + β 2 s ;

(2) ( k β ) s = k β s , λ C .

Proof is immediate from the definitions.

Theorem 5.5 β s : H ( Y ) H ( Y s ) is continuous with respect to Δ + -convergence.

Proof Let β n Δ β in H ( Y ) as n . Then there exist f n S + ( R ) and ( s n ) Δ + such that ( β n β ) s n = [ f n s k s k ] and f n 0 as n . Employing Eq. (5.1), we get

S ( ( β n β ) s n ) = [ S ( f n s k ) s k s ] .

Hence, we have S ( ( β n β ) s n ) = [ y f n s s k s s k s ] 0 as n in H ( Y s ) . Therefore

S ( ( β n β ) s n ) = y ( β n s β s ) s n s 0 as  n .

Hence, β n s Δ β s as n . □

Theorem 5.6 β s : H ( Y ) H ( Y s ) is onto.

Proof Let [ f n s s n s ] H ( Y s ) be arbitrary, then f n s s m s = f m s s n s for every m , n N . Then f n s m = f m s n . That is, f n s n is the corresponding quotient of sequences of f n s s n s . Thus [ f n s n ] H ( Y ) is such that S [ f n s n ] = [ f n s s n s ] in H ( Y s ) . This completes the proof of the lemma.

Let β s = [ f n s s n s ] H ( Y s ) , then we define the inverse Sumudu transform of β s by

β s 1 = [ f n s n ]

in the space H ( Y ) . □

Theorem 5.7Let [ f n s s n s ] H ( Y s ) and u D + ( R ) , u s D + s ( R )

β ( [ f n s n ] u ) = [ f n s s n s ] u and β s 1 ( [ f n s s n s ] u s ) = [ f n s n ] u .

Proof is immediate from the definitions.

Competing interests

The authors declare that they have no competing interests.

Authors’ contributions

Both of the authors contributed equally to the manuscript and read and approved the final draft.

Acknowledgements

Dedicated to Professor Hari M Srivastava.

The authors express their sincere thanks to the referee(s) for the careful and detailed reading of the manuscript and very helpful suggestions that improved the manuscript substantially.

References

  1. Watugala, GK: Sumudu transform: a new integral transform to solve differential equations and control engineering problems. Int. J. Math. Educ. Sci. Technol.. 24(1), 35–43 (1993). Publisher Full Text OpenURL

  2. Belgacem, FBM, Karaballi, AA, Kalla, LS: Analytical investigations of the Sumudu transform and applications to integral production equations. Math. Probl. Eng.. 3, 103–118 (2003)

  3. Weerakoon, S: Application of Sumudu transform to partial differential equations. Int. J. Math. Educ. Sci. Technol.. 25, 277–283 (1994). Publisher Full Text OpenURL

  4. Watugala, GK: Sumudu transform new integral transform to solve differential equations and control engineering problems. Math. Eng. Ind.. 6(4), 319–329 (1998)

  5. Watugala, GK: The Sumudu transform for functions of two variables. Math. Eng. Ind.. 8(4), 293–302 (2002)

  6. Asiru, MA: Further properties of the Sumudu transform and its applications. Int. J. Math. Educ. Sci. Technol.. 33(3), 441–449 (2002). Publisher Full Text OpenURL

  7. Kadem, A: Solving the one-dimensional neutron transport equation using Chebyshev polynomials and the Sumudu transform. An. Univ. Oradea, Fasc. Mat.. 12, 153–171 (2005)

  8. Eltayeb, H, Kılıçman, A, Fisher, B: A new integral transform and associated distributions. Integral Transforms Spec. Funct.. 21(5), 367–379 (2010). Publisher Full Text OpenURL

  9. Kılıçman, A, Eltayeb, H, Agarwal, PR: On Sumudu transform and system of differential equations. Abstr. Appl. Anal.. 2010, Article ID 598702. doi:10.1155/2010/598702 (2010)

  10. Zhang, J: A Sumudu based algorithm for solving differential equations. Comput. Sci. J. Mold.. 15(3(45)), 303–313 (2007)

  11. Weerakoon, S: Complex inversion formula for Sumudu transforms. Int. J. Math. Educ. Sci. Technol.. 29(4), 618–621 (1998)

  12. Kılıçman, A, Eltayeb, H: On the applications of Laplace and Sumudu transforms. J. Franklin Inst.. 347(5), 848–862 (2010). Publisher Full Text OpenURL

  13. Boehme, TK: The support of Mikusinski operators. Trans. Am. Math. Soc.. 176, 319–334 (1973)

  14. Mikusinski, P: Fourier transform for integrable Boehmians. Rocky Mt. J. Math.. 17(3), 577–582 (1987). Publisher Full Text OpenURL

  15. Mikusinski, P: Tempered Boehmians and ultradistributions. Proc. Am. Math. Soc.. 123(3), 813–817 (1995)

  16. Roopkumar, R: Mellin transform for Boehmians. Bull. Inst. Math. Acad. Sin.. 4(1), 75–96 (2009)

  17. Zemanian, AH: Generalized Integral Transformation, Dover, New York (1987) (First published by interscience publishers)