Generating Functions from Symmetric Functions Anthony Mendes And
Total Page:16
File Type:pdf, Size:1020Kb
Generating functions from symmetric functions Anthony Mendes and Jeffrey Remmel Abstract. This monograph introduces a method of finding and refining gen- erating functions. By manipulating combinatorial objects known as brick tabloids, we will show how many well known generating functions may be found and subsequently generalized. New results are given as well. The techniques described in this monograph originate from a thorough understanding of a connection between symmetric functions and the permu- tation enumeration of the symmetric group. Define a homomorphism ξ on the ring of symmetric functions by defining it on the elementary symmetric n−1 function en such that ξ(en) = (1 − x) /n!. Brenti showed that applying ξ to the homogeneous symmetric function gave a generating function for the Eulerian polynomials [14, 13]. Beck and Remmel reproved the results of Brenti combinatorially [6]. A handful of authors have tinkered with their proof to discover results about the permutation enumeration for signed permutations and multiples of permuta- tions [4, 5, 51, 52, 53, 58, 70, 71]. However, this monograph records the true power and adaptability of this relationship between symmetric functions and permutation enumeration. We will give versatile methods unifying a large number of results in the theory of permutation enumeration for the symmet- ric group, subsets of the symmetric group, and assorted Coxeter groups, and many other objects. Contents Chapter 1. Brick tabloids in permutation enumeration 1 1.1. The ring of formal power series 1 1.2. The ring of symmetric functions 7 1.3. Brenti’s homomorphism 21 1.4. Published uses of brick tabloids in permutation enumeration 30 1.5. First extensions of Brenti’s homomorphism 34 Chapter 2. Generating functions for permutations 51 2.1. Alternating permutations 51 2.2. Consecutive descents 56 2.3. Consecutive patterns 67 2.4. Descents, major indices, and inversions 91 Chapter 3. Generating functions for other objects 97 3.1. Wreath product statistics 97 3.2. Words 102 3.3. Fibonacci numbers 113 3.4. The exponential formula 113 Chapter 4. Conclusions 125 Appendix A. Permutation statistics 133 Appendix. Bibliography 135 iii CHAPTER 1 Brick tabloids in permutation enumeration We will begin with a discussion on the ring of formal power series in order to establish basic definitions and concepts needed understand the rest of this work. Generating functions and the benefits of their use are introduced here. The objec- tives of this monograph and a brief introduction to permutation statistics are also included in Section 1.1. Symmetric function theory will be used heavily throughout this work. In Sec- tion 1.2, the ring of symmetric functions is described completely from scratch. Our approach is not like that of any other published work in that we attempt to give combinatorial proofs for every basic symmetric function identity. It is in Section 1.2 where we introduce the notion of brick tabloids, a combinatorial object used extensively in the rest of this monograph. Section 1.3 contains the ideas at the heart of this monograph. Symmetric func- tion theory is used to create generating functions here. The rest of this monograph is devoted to further developing the methods in Section 1.3—this section is essential reading for those wanting a complete understanding of our building of generating functions. We systematically compare all previous published works relating brick tabloids and permutation statistics with the results in this monograph in Section 1.4. This should help to clarify the advances we will make in this work. Many proofs in the first three sections have been revamped and elements of the combinatorial approach to basic symmetric function identities included in this section have not been previously published, but this content is well known and none of it should be considered brand new. This is not the case for Section 1.5, however. Although some of the generating functions in this section are known, the methodology is original. It is in this section where we make our first significant advances in our development when we discuss ways to extend the homomorphism introduced in Section 1.3 to find and refine more results about descents in the symmetric group. 1.1. The ring of formal power series Let a0,a1,... be a sequence in a ring R and t an indeterminate. The formal power series for the sequence a0,a1,... is the expression a t0 + a t1 + a t2 + . 0 1 2 · · · n Such an object may be denoted n∞=0 ant . At face value, although we are using the plus symbol and summation notation, we are not performing the operation of P addition. We are simply presenting the sequence a0,a1,... in a specific way, using plus symbols to separate terms and using powers of t as placeholders. The set of 1 2 1. BRICK TABLOIDS IN PERMUTATION ENUMERATION all formal power series in t representing sequences in the ring R will be denoted by R[[t]]. For instance, consider the sequence in the ring of rational numbers defined by n(n 1)/2 an =2 − /n!. This sequence recorded as a formal power series is 8 64 ∞ 2n(n 1)/2 (1.1) 1+ t + t2 + t3 + t4 + = − tn. 6 24 · · · n! n=0 X If interpreted as a complex-valued function in the variable t, the above series would have a radius of convergence of 0 and thus most techniques from analysis would not apply. The object in (1.1) does, however, define a perfectly valid formal power series. In fact, this formal power series will be used in Section 3.4 to count the number of graphs on n nodes. If a0,a1,... is a sequence in R such that a0 = a1 = = aj 1 = 0 for some n n · · · − j 0, then we may denote ∞ a t as ∞ a t to reflect this fact. Along ≥ n=0 n n=j n j n similar lines, we write n=0 aPnt provided aPn = 0 for n > j. There are a handful of operations on R[[t]] which we now define. First, for each P nonnegative integer j, define a map j from R[[t]] to R such that ·|t ∞ n ant = aj . n=0 tj X j The element in R found by an application of j is called the coefficient of t . Two ·|t elements in R[[t]] are equal provided the coefficients of tj in each formal power series are equal for all j 0. Define the sum≥ of two formal power series by the rule ∞ n ∞ n ∞ n ant + bnt = (an + bn)t n=0 ! n=0 ! n=0 X X X where the plus symbol on the right hand side of the equation denotes the sum of two elements in the ring R. Define the product of two formal power series by the rule ∞ ∞ ∞ n n n ant bnt = (a0bn + a1bn 1 + + an 1b1 + anb0) t − · · · − n=0 ! n=0 ! n=0 X X X where the plus symbols and the adjacent elements on the right hand side of the above equation denote the sum and product of two elements in R, respectively. With these definitions, it is not difficult to show that R[[t]] is a ring; naturally, R[[t]] is called the ring of formal power series. Notice that R is commutative if and only if R[[t]] is commutative. For greater simplicity in our development of the ring of power series, we make the assumption that R is a commutative ring with unity from now on. If ∞ n ∞ n ant bnt =1 n=0 ! n=0 ! X X 1.1. THE RING OF FORMAL POWER SERIES 3 2 n where 1 represents 1 + 0t +0t + , then we say n∞=0 bnt is the reciprocal of n · · · ∞ ant and write n=0 P 1 P ∞ ∞ − n n 1 bnt = ant = . ∞ a tn n=0 n=0 ! n=0 n X X For example, by our definition of the product of twoP formal power series, (1 t)(1+ 2 2 1 − t + t + ) = 1 and therefore (1+ t + t + )=(1 t)− =1/(1 t). Formally, this is the· · · familiar formula for the sum of a· geometric · · − series. − n m Define the composition of n∞=0 ant and m∞=1 bmt as the formal power series P Pn ∞ ∞ m an bmt . n=0 m=1 ! X X A potential problem in this definition arises if any coefficient in the above formal power series is an infinite sum of elements in R. However, notice that n ∞ ∞ ∞ n a b tm = a b t1 + + b tj n m n 1 · · · j n=0 m=1 ! j n=0 j X X t X t j 1 j n = an b1t + + bj t · · · n=0 tj X j where in the last expression we are selecting the coefficient of t in a finite sum. j This shows that the stipulation b0 = 0 forces the coefficient of t in the composition as defined above to be a finite sum of elements in R for j 0. Thus, there are no problems with our definition of composition. ≥ The derivative of a formal power series is a map d/dt( ) from R[[t]] to R[[t]] defined by · d ∞ ∞ a tn = (n + 1)a tn dt n n+1 n=0 ! n=0 X X where n + 1 is the element 1 + + 1 (n + 1 times) in R. Similarly, the integral of a formal power series is a map · · · dt from R[[t]] to R[[t]] defined by · R ∞ n ∞ an 1 n a t dt = − t n n n=0 ! n=1 Z X X provided the multiplicative inverse of n exists in R for n 1.