Applied Mathematics, 2013, 4, 12-17 http://dx.doi.org/10.4236/am.2013.47A004 Published Online July 2013 (http://www.scirp.org/journal/am)
Generalized Powers of Substitution with Pre-Function Operators
Laurent Poinsot University Paris 13, Paris Sorbonne Cité, LIPN, CNRS, Villetaneuse, France Email: [email protected]
Received April 13, 2013; revised May 13, 2013; accepted May 20, 2013
Copyright © 2013 Laurent Poinsot. 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.
ABSTRACT An operator on formal power series of the form SS , where is an invertible power series, and is a series of the form tt 2 is called a unipotent substitution with pre-function. Such operators, denoted by a pair , , form a group. The objective of this contribution is to show that it is possible to define a generalized powers for a a such operators, as for instance fractional powers , b for every . b
Keywords: Formal Power Series; Formal Substitution; Riordan Group; Generalized Powers; Sheffer Sequences; Umbral Calculus
1. Substitution of Formal Power Series inverse of is then denoted by 1 and satisfies 11 In this contribution we let denote any field of t . Finally, it is possible to define a characteristic zero. We recall some basic definitions from semi-direct product of groups by considering pairs [1,2]. The algebra of formal power series in the variable , where is a unipotent series, and is a t is denoted by t . In what follows we sometimes unipotent substitution, and the operation 11, use the notation S t for S t to mean that S 22,, 1 2 212 . The identity element is is a formal power series of the variable t. We recall that (1, t) . This group has been previously studied in [3-5], any formal power series of the form tS for and is called the group of (unipotent) substitutions with and S t is invertible with respect to pre-function. These substitutions with pre-function act on the usual product of series. Its inverse is denoted by 1 t as follows: ,SS for every 1 series S . In [3] is associated a doubly-infinite matrix and has the form tV for some V t . In particular, the set of all series of the form 1t T M , to each such operator which defines a matrix forms a group under multiplication, called the group of representation of the group of substitutions with pre- unipotent series. For a series of the form tt2T , function, and it is proved that there exists a one- n parameter sub-group M , . Therefore, it sa- we may define for any other series S n t an n0 n tisfies MM,, M, for every , , and operation of substitution given by S n0n . A 2 unipotent substitution is a series of the form ttT . M , is the usual -th power of M , whenever Such series form a group under the operation of is an integer. It amounts that for every , M , is substitution, called the group of unipotent substitutions the matrix representation of a substitution with pre- (whenever 0 , a series tt2T is invertible function say , so that MM . The under substitution, and the totality of such series forms a , , group under the operation of substitution called the authors of [3] then define ,, . Actually group of substutions, and it is clear that the group of in [3] no formal proof is given for the existence of such unipotent substitutions is a sub-group of this one). The generalized powers for matrices or unipotent substitu-
Copyright © 2013 SciRes. AM L. POINSOT 13 tions with pre-function. n the form n L where n for every n 0 . In this contribution, we provide a combinatorial proof n0 The series Ln converges to an operator of for the existence of these generalized powers for unipo- n0 n tent substitutions with pre-function, and we show that x in the topology of simple convergence (when this even forms a one-parameter sub-group. To achieve has the discrete topology) since for every p x , this objective we use some ingredients well-known in there exists np such that for all nn p , combinatorics such as delta operators, Sheffer sequences Lpn 0 , so that we may define and umbral composition which are briefly presented in np what follows (Sections 2, 3, 4 and 5). The Section 6 con- nn nnLp Lp . tains the proof of our result. nn00 According to [6], if D is a differential operator, then 2. Differential and Delta Operators, and if, and only if, commutes with D , i.e., Their Associated Polynomial Sequences D DD . Moreover, if DDn , then By operator we mean a linear endomorphism of the n0 n -vector space of polynomials x (in one indeter- is also a differential operator if, and only if, 0 0 minate x ). The composition of operators is denoted by a and 1 0 . simple juxtaposition. If p x , then we sometimes Following [1], let us define a sequence of polynomials write px to mean that p is a polynomial in the va- x x x 1 n by 0 and x i riable x . n 0 n 1 n 1!i0 n0 Let p x be a sequence of polynomials. n n0 for every integer n . For , we denote by It is called a polynomial sequence if deg pn n for n every n 0 (in particular, p0 ). It is clear that a polynomial sequence is thus a basis for x . x the value of the polynomial for x . Let L be An operator D is called a differential operator (see n [6]) if a lowering operator, and let UidL L be its 1) D 0 for every . unipotent part. Then we may consider generalized power 2) degDp deg p 1 for every non-constant poly- n nomial p . UL L (in particular, this explains n0 n For instance, the usual derivation of polynomials is a differential operator. Moreover, let , and let the notation E for the shift operator). We observe that k us define the shift-invariant operator E as the unique for every integer k , U really coincides to the k -th n n linear map such that E xx for every power UU of U . Moreover, U UU for n 0 . Then, E1 id is also a differential operator. k factors A polynomial sequence p is said to be a normal every , . We may also form n n0 family if n1 1 n 1) p0 1 . logUidL log LL n1 n 2) pn 0 0 for every n 0 . Let D be a differential operator. A normal family in such a way that for every , p is said to be a basic family for D if n n0 U exp logU Dpnn1 n1 p where for every LLn with 0 , for every n 0 . It is proved in [6] that for any differen- n0 n 0 tial operator admits is one and only one basic family, and, 1 exp n (it is a well-defined operator). This conversely, any normal family is the basic family of a n0 n! unique differential operator. As an example, the normal kind of generalized powers may be used to compute frac- n family x is the basic family of . a n0 Let L be an operator such that for every non-zero tional power of the form U b for every a , b polynomial p x , degLp deg p (in particu- 1 n n lar, L 0 for every constant ). Such an op- (for instance, UU ). They satisfy the usual proper- erator is called a lowering operator (see [7]). For in- ties of powers: Uid0 , UUU . The objective stance any differential operator is a lowering operator. of this contribution is to provide a proof of the existence Then given a lowering operator L , we may consider the of such generalized powers for unipotent substitutions algebra of formal power series L of operators of with pre-function.
Copyright © 2013 SciRes. AM 14 L. POINSOT
Following [8], we may consider the following sub-set , of formal power series in t with in- of differential operators, called delta operators. A poly- vertible, and 0 0 , 1 0 , such that nomial sequence p is said to be of binomial-type xt n n0 EGF s ;t t e . if for every n 0 , n n n Remark 1. The basic set of a delta operat or D is a pppnknkxy x y x,y. k =0 Sheffer sequence. Let D be a delta-operator with basic set p . An operator is a shift-invariant operator if for n n0 Following [8], the following result holds. every , E E . Now, a delta operator D Propos ition 1. A polynomial sequence s is a is a shift inva riant operator such that Dx . For in- n n0 Sheffer sequence if, and only if, there exists an invertible stance, the usual derivation of polynomials i s a delta 1 shift-invariant operator S such that snn Sp for each operator. It can be proved that a delta operator is a dif- n 0 . Moreover, let S be an invertible shift-invariant ferential operator. The basic family (uniquely) associated operator. Let t be the unique form al power to a delta operator is called its basic set. Moreover, the basic set of a delta operator is of binomial-type, and to series such that S . Then, is invertible, and any polynomial sequence of binomial-type is uniquely 1 1 EGF s ;t 1 ex associated a delta operator. If D is a delta operator, n n then there exists a unique -algeb ra isomorphism from where s is the Sheffer se quence define d by n n t to the ring of shift-in variant operators D 1 n n t D snn Sp for each n 0 , and t is the that maps Ss n to SD sn . In [8] n0 n! n0 n! unique formal power series su ch that D . Finally is proved that given a delta op erator D , and a series we also have the following ch aracterization. n Proposition 2. Let sn be a polynomial seq uence. It n1 n t 0 0 with 1 0 , then D is n also a delta operator. Conversely, if is a shift- in- is a Sheffer sequence if, and only if, there exists a delta operator D with basic set p such that variant operator (so that D ), then if it is a n n delta operator, the unique series n n n ssnkxy xpnk y x,y. t k t k 0 n0 n n! 4. Umbral Composition such that D satisfies 0 0 and 1 0 . This section is based on [11]. 3. Sheffer Sequences Let p be a fixed polynomial sequ ence. Let us n n n In this section, we also briefly recall some definitions define an operator by x pn for each n 0 . Since p is a basis of x , this means that is and results from [8]. n n Let pn be a sequence of polynomials in x . a linear isomorphism of x . When p is th e n0 n n We define the exponential generating function of basic set of a delta operator, then is referred to as an pn as umbral operator, while if p is a Sheffer se quence, n0 n n tn then is said to be a Sheffer operat or. An umbral op- EGF pnn;t p x t . n n! erator maps basic sets to basic se ts, while a Sheffer op- n0 erator m aps Sheffer sequences to Sheffer sequences. Let D be a delta operator and pn be its basic Let p be a polynomial sequence. For every n , n0 n n set. Let t with 0 and 0 such n 0 1 pp xxkk where p xk is the coefficient that D . Then from [8], nnk 0 of xk in the polynomial p x . Let p and 1 n n EGF p ;t ext . q be two polynomial sequences. T heir umbral com- n n n n position is defined as the polynomial sequence r A polynomial sequence s is said to be a Sheffer n n n n0 p# q defined by nnnn sequence (also called a polynomial sequence of type zero n k in [9] or a poweroid in [10]) if there exists a delta opera- rpqn nx k tor D such that k 0 for each n 0 . By simple computations, it may be 1) s0 , proved that rpxxkk n qx. The se t of 2) Dsnn1 n 1 s for every n . nnk all polynomial sequences becomes a (non-commutative) Following [9], a polynomial sequence sn is a n0 monoid under # with xn as identity. We observe Sheffer sequence if, and only if, there exists a pair n
Copyright © 2013 SciRes. AM L. POINSOT 15
n that if T is the operator defined by Tqx n for each Sheffer sequence is also unipotent. n 0 , then pq# T p. More generally, we It is clear from theorem 4 that the group of basic sets nnnn n n under umbral composition is isomorphic to the group of #k #k have Tknx q where q is the substitutions. Moreover, the group of unipotent basic sets n n n n n also is isomorphic to the group of unipotent substitutions. k -th power of q for the umbral compositi on (it is n n Likewise, the group of (unipotent) Sheffer sequences is equal to a sequence say rk and we denote rk n n n isomorphic to the group of (unipotent) substitutions with #k by qn ). Under umbral comp osition, the set of a l l pre-function (see also [12]). Sheffer sequences is a (non-com mutative) group, called Lemma 1. Let , be a substitution with pre- the Sheffer group ([12]), and the set of all basic se- function, and let spnn , be the Sheffer sequence quences is a sub-group of the Sheffer group. nn and the basic set associated to the delta operator From [8] we have the following result that combines and the invertible shift-inv a r iant operator (this delta operators, basis sets, Sheffer sequences and umbral means that p is the basic set of , and composition. n n Theorem 1. Let Q and P be two delta operators 1 s p for each n ). Then, , is a uni- with respective basic sets q and p . Let S nn n n n n and T be two inverti ble shift -invariant operators. Let potent substitution with pre-function if, and only if, s and t be the Sheffer sequences de fined by psx1nn x for eve ry n . n n n n nn s S1q and tTp 1 for each n . Let , be nn nn Proof. Let us first assume that , is a unipotent two invertible s eries such that S , T . substitution with prefunction. We have p0 1 for every Let , be two form al po wer series with 00 0 , basic set, so that p0 11 . Let n . We have 110 such that Q and P . Then, k . Then, pnp1 is RPT is a shift-invariant opera - k2 k nn1 equivalent to tor, is a delta operator with basic sequence nn1 pnpxk1 xk. vqp # . Finally, let r be the Sheffer kk0=nn1 0 nnnnn n n n sequence given by rs # t. Then, rRv 1 n nnnn nn nnBy identification of the coefficient of x on both sides, for each n . we obtain np1x nn1 np 1x It may be proved that if s is the Sheffer sequence nn1 n n (since is assumed to be a unipotent su bstitution), and, obtained from the delta operator D with basic n1 by induction, pn1 x1 . Besides, we have set pn and the invertible shift-invariant operator n 1 1 1 1 snn p for each n . But S , i.e., snn Sp for each n , then the inverse (because there is a ring isomorphism between t qn of pn with respect to the um bral composition 1 n n and ), and 1t , where t . 1 is the basic set of the delta operator , the inverse Then, by identification of the coefficient of xn , we have t of s with respect to the umbral com position nn n n n n spnxx1 n for every n . Conversely, let u s 1 is the Sheffer sequence tqnn . assume that sp, is the Sheffer seq uence and nnnn 5. Unipo tent Sequences the basic set associated to the delta operator and the invertible shift-invariant operator with The basic set p of a delta operator D is said to be nn n n psx1 x for every n . By constructio n we unipotent if the unique series such that D nn n1 n satisfies 1 1 (and, obviously, 0 0 ), i.e., is a have 11np1x nn np 1x so that 1 1. unipotent substitution. A Sheffer sequence s asso- n n nn Likewise, 0 spnnx x, so that 0 1. □ ciated to a delta operator D (with 0 0 , 1 0 ) and an invertible shift-inva riant ope rator 1 S (with invertible), i.e., snn Sp for 6. Generalized Powers of Unipotent every n where p is the basic set o f D , is said to n n Substitutions with Pre-Functio n be unipotent if p is unipotent, and if is unipo- n n The purpose of this section is to define , for any tent, i.e., 0 1. It is also clear from the previous sec- and any unipotent substitution with pre-function tion (theorem 4) that the (umbral) inverse of a unipotent , , and to prove that it is also a unipotent substitu- basic set is unipotent, and the (umbral) inverse of a tion with pre-function. Moreover we show that
Copyright © 2013 SciRes. AM 16 L. POINSOT
, is a one-parameter sub-group, i.e., n n sknixy sk x pni k y i0 i ,,, for every , , and (2) n 0 n ki kni ,1, t. TRxy, i0 i Let , be a unipotent substitution with pre- Now, let z be a variable commuting with x and y , function of t . Let p be the unipotent basic n n and let us define set of the (unipotent ) delta operator . Let s n n zzn be the unipotent Sheffer sequence associated to in in sLn zxxii00 Lx,z and the (unipotent) invertible shift-invariant operator . ii n Let R be the umbral operator given by Rpx n for and similarly, all n , and let T be the Sheffer operator defined by n z n z Tsx n for all n . It is easily checked that for every in in pNn zx Nxx,z, #k kn #k kn ii00ii intege r k , pRn x and sTn x . In par- nnn1 nkfor each n . As polynomials in the variable z , their ticular, for each n , Tsxxnn s xx k0 degrees are at most n . As polynomials in the va riable (by Lemm a 5). Therefore, Tid L, where z , sn zx y , pn zxy , nnkn1 Lsxx n x for each n . The operator L is k 0 n n ppzx zy actually a lowering operator. Then according to section 2, i0 ini i it is possible to define TLL k for k0 and k n n every . Moreover, we have TT exp log . spinzxi zy i0 i n For each , let us define sTn x for have also a degree at most n . Because the equations (1) every n . When k , we have sn k and (2) hold for every integer k , the polynomials (in the kn # k variable z ) Tsx . So that in this case, s k is the n n n k n n unipotent She ffer sequen ce associ ated to , . This ppnizxy i0 zxpni zy i means that if ,, k , and q is the uni- kk n n and potent basic set of the (unipotent) delta operator k , n n snzxy si zxpni zy #k 1 i0 i then snk qn for each n . Similarly , let nnn1 kk are identically zero, and the above equations hold for Rpxxnn k 0 p xx for every n . There- n1 every . Therefore, p is a polynomial nkk n n fore, RidN, where Npxxx n is a k 0 sequence of binomial-type, and s is a Sheffer n n lowering operator. Again for every , we define sequence for every . Moreover , for every R exp log RNk N. For each , , we have k0 k pR xxnnRR n n , we define pRn x for each n . In particular n kk #k Rpnn k=0 pxRx for k , pknn p , so that it is the basic set k n ppxk of the unipotent delta op erator . Clearly, k=0 n k 1 snkpkn k for each n . Thus for every so that ppq # . Similarly, k , we have nnn nn n sss # for every , . n n nnnnn n Moreover, pknixy pkx pni k y i0 i (1) pR0xxx00nn T n q0. n nnnn n n ki kni nn RRxy, i0 i Therefore, p and s n n n n
Copyright © 2013 SciRes. AM L. POINSOT 17 are one-parameter sub-groups. It follows that REFERENCES #1 [1] A. Benhissi, “ Ri ngs of Formal Power Series,” Queen’s pp nnnn Papers in Pure and Applied Mathematics, Queen’s Uni- versity, Kingsone, 2003. and [2] R. P. Stanley, “Enumerative Combinatorics—Volume 1, #1 ss Volume 49 of Cambridge Studies in Advanced Mathe- nnnn matics,” Cambridge University Press, Cambridge, 2000. (inverses under umbral operation). [3] G. H. E. Duchamp, K. A. Penson, A. I. Solomon, A. Hor- We define , as the pair of formal power series zeal and P. Blasiak, “One-Parameter Groups and Combi- , such that is the substitution that defines natorial Physics,” Proceedings of the Symposium CO- PROMAPH3: Contemporary Problems in Mathematical the delta operator with basic sequence p , n n Physics, Cotonou, 2004, pp. 436-449. and is the invertible series such that [4] L. Poinsot and G. H. E. Duchamp, “A Formal Calculus on 1 the Riordan near Algebra,” Advances and Applications in spn n Discrete Mathematics, Vol. 6, No. 1, 2010, pp. 11-44. for each . Since p and s are n n n n [5] L. W. Shapiro, S. Getu, W.-J. Woan and L. C. Woodson, unipotent sequences, it is clear that is unipotent, and “The Riordan Group,” Discrete Applied Mathematics, Vol. 34, No. 1-3, 1991, pp. 229-239. is a unipotent substitution. It is also clear that doi:10.1016/0166-218X(91)90088-E whenever k , then ,, k . Let us kk [6] G. Markowsky, “Differential Operators and the Theory of check that , is a one-parameter sub- Binomial Enumeration,” Journal of Mathematical Analy- group of the group of unipotent substitutio ns with pre- sis and Application, Vol. 63, No. 1, 1978, pp. 145-155. function. This means that for every , , doi:10.1016/0022-247X(78)90111-7 [7] G. H. E. Duchamp, L. Poinsot, A. I. Solomon, K. A. Pen- ,,, son, P. Blasiak and A. Horzela, “Ladder Operators and Endomorphisms in Combinatorial Physics,” Discrete Ma- , thematics and Theoretical Computer Science, Vol. 12, No. 2, 2010, pp. 23-46. First of all, by definition, is the unipotent sub- stitution associated to the basic set [8] G.-C. Rota, D. Kahaner and A. Odlyzko, “Finite Operator Calculus,” Journal of Mathematical Analysis and Its Ap- ppp # , plications, Vol. 42, No. 3, 1973, pp. 684-760. nnnnn n [9] I. M. Sheffer, “Some Properties of Polynomial Sets of and therefore . In a similar way, the se- Type Zero,” Duke Mathematical Journal, Vol. 5, No. 3, ries is uniquely associated t o the Sheffer sequence 1939, pp. 590-622. doi:10.1215/S0012-7094-39-00549-1 sss # and to the basic set [10] J. F. Steffensen, “The Poweroid, an Extension of the Ma- nnnnnn thematical Notion of Power,” Acta Mathematica, Vol. 73, ppp # . Again this means nnnnn n No. 1, 1941, pp. 333-366. doi:10.1007/BF02392231 that . Therefore, we obtain the ex- [11] S. Roman, “The Umbral Calculus,” Dover Publications, New York, 1984. pected res ult. It is also clear that 00,1,t . [12] T.-X. He, L.C. Hsu and P.J.-S. Shiue, “The Sheffer Group Remark 2. In particular, since is a field of and the Riordan Group,” Discrete Applied Mathematics, characteristic zero, for every q , we may define Vol. 155, No. 15, 2007, pp. 1895-1909. q fractional powers , such as for instance n , doi:10.1016/j.dam.2007.04.006 for each integer n .
Copyright © 2013 SciRes. AM