18.312 Algebraic Combinatorics: Mit Opencourseware

MIT OpenCourseWare http://ocw.mit.
edu
18.312 Algebraic Combinatorics

Spring 2009
For information about citing these materials or our Terms of Use, visit: http://ocw.mit.edu/terms.
Course 18.312: Algebraic Combinatorics
Lecture Notes # 1 Addendum by Gregg Musiker February 4th - 6th, 2009
Recurrence Relations and Generating Functions
Given an innite sequence of numbers, a generating function is a compact way of expressing this data. We begin with the notion of ordinary generating functions. To illustrate this denition, we start with the example of the Fibonacci numbers. {Fn } = {F0 , F1 , F2 , F3 , . . .} n=0 dened by F0 = 1, F1 = 1, and Fn = Fn1 + Fn2 for n 2. 1, 1, 2, 3, 5, 8, 13, 21, 34, . . . We dene F (X) := F0 + F1 x + F2 x2 + F3 x3 + F4 x4 + . . . = 1 + x + 2x2 + 3x3 + 5x4 + 8x5 + . . . . In other words, F (X) is the formal power series
k =0
Fk xk .
Remark. This is called a formal power series because we will consider x to be an indeterminate variable rather than a specic real number. In general, given a sequence of numbers {ai } = {a0 , a1 , a2 , a3 , . . .}, the associ i=0 ated formal power series is A(X) :=
k=0
ak xk = a0 + a1 x + a2 x2 + a3 x3 + . . . .
We will shortly write down F (X) in a compact form, but we begin with an easier example that you have already seen. 1
Recall that
n
k
n! . k!(nk)!
For example, = {1, 8, 28, 56, 70, 56, 28, 8, 1}.
88 k In fact if k > 8, sequence as 8

k
k=0
(e.g.
8 ) equals zero. Thus we can consider the entire innite 9 = {1, 8, 28, 56, 70, 56, 28, 8, 1, 0, 0, 0, . . .},
8 k
k=0
and then the associated formal power series 1 + 8x + 28x2 + 56x3 + 70x4 + 56x5 + 28x6 + 8x7 + x8 + 0x9 + 0x10 + . . . can be written compactly as (1 + x)8 . Generalizing this to any positive integer n, k has associated power series k=0 n n k n n (1 + x) , since (1 + x) = k=0 k x by the Binomial Theorem. This illustrates that from a formal power series, we can recover a sequence of numbers. We call these numbers the coecients of the formal power series. For example, we say that n is the coecient of xk in (1 + x)n . This is sometimes k n written as (1 + x) = n or [xk ](1 + x)n = n . k k
xk
1.1
More complicated formal power series
k We now want to write a similar expression for F (X) = k=0 Fk x , where Fk = Fk1 + Fk2 for k 2 and F0 = F1 = 1. Notice that k k ak x bk x = (ak bk )xk .
k=0 k=0 k=0
As a consequence, Fk = Fk1 + Fk2 for k 2 implies F (X) = 1 + F1 x + = 1 + F1 x + = 1 + F1 x +

Fk xk
(Fk1 + Fk2 )xk Fk1 x +
k k=2
k=2
k=2 k=2
Fk2 xk
= 1 + F1 x + xF (X) F0 x + x2 F (X).
Thus F (X)(1 x x2 ) = 1 + (F1 F0 )x = 1 + 0x and we obtain the rational expression F (X) = 1 . 1 x x2
If we look at the Taylor series for this rational function, we indeed obtain coef cients that are the Fibonacci numbers. Generating Functions are also helpful for obtaining closed formulas or asymptotic formulas. If we use partial fraction decomposition, we see that F (X) = B A + . 1 1 x 1 2 x
We know that (1 1 x)(1 2 x) = 1 x x2 so 1 2 = 1 1 2 = 1 Thus {1 , 2 } = { 1+2 5 , 12 5 }. Exercise 1: Solve for A and B and use this to obtain a closed form expression for Fk . Notice that as a consequence we can compute that {Fk+1 /Fk } = {1/1, 2/1, 3/2, 5/3, 8/5, 13/8, . . .} k converges to 1+2 5 since 12 5 0 as k , so Ak+1 Ak+1 + Bk+1 1 2 1 = 1 . k k A1 + B2 Ak 1

and
1.2
A Combinatorial Interpretation of the Fibonacci Num bers
Given a sequence of integers S = {s0 , s1 , s2 , . . .}, a combinatorial interpretation of S is a family F of objects (of various sizes) such that the number of objects in F of size k is exactly counted by sk . For example, a combinatorial interpretation of n is as the number of subsets k of an {1, 2, . . . , n} of size k. A domino tiling of a rectangular region R is a covering of R by horizontal (1-by-2) domino tiles and vertical (2-by-1) domino tiles such that every square of R is covered by exactly one domino.
For example, if we let R be a 2-by-2 grid, then there are two possible domino tilings. Either both tiles are vertical or both are horizontal. If we let R be a 2-by-3 grid, then there are three possible domino tilings, and a 2-by-4 grid would have ve such domino tilings. Proposition. The number of domino tilings of a 2-by-n grid is counted by the nth Fibonacci number, Fn for n 1. Proof. Let DTn denote the number of domino tilings of the 2-by-n grid. We rst check the initial conditions. There is one way to tile the 2-by-1 grid, and there are two ways to tile the 2-by-2 grid. Thus DT1 = 1 = F1 and DT2 = 2 = F2 . (Recall that F0 = 1, but we do not use this quantity in this combinatorial interpretation.) Domino tilings of the 2-by-n grid either look like a domino tiling of the 2-by (n 1) grid with a vertical tile tacked onto the end, or a domino tiling of the 2-by-(n 2) grid with two horizontal tiles tacked onto the end. Consequently, DTn = DTn1 + DTn2 , the same recurrence as the Fn s. Exercise 2: Show that this combinatorial interpretation can be rephrased as the statement Fn = The number of subsets S of {a1 , a2 , . . . , an1 } such that ai and ai+1 are not both contained in S.
1.3
Convolution Product Formula
In addition to adding formal power series together, we can also multiply them. If A(X) = ak xk and B(X) = bk xk , where ak (resp. bk ) counts the number k =0 k =0 of objects of type A (resp. B) and size k, then A(X)B(X) = C(X) =
n=0
cn xn
where cn = n ak bnk , and has a combinatorial interpretation as the number of k=0 objects of size n formed by taking an object of type A and concatenating it with an object of type B.
1.4
Connection between Linear Recurrences and Rational Generating Functions
The behavior we saw of the Fibonacci numbers and its generating function is an example of a more general theorem. Theorem. (Theorem 4.1.1 of Enumerative Combinatorics 1 by Richard Stanley) Let 1 , 2 , . . . , d be a xed sequence of complex numbers, d 1, and d = 0. The following conditions on a function f : N C are equivalent: i) The generating function F (X) equals
n=0
f (n)xn =
P (x) Q(x)
where Q(x) = 1 + 1 x + 2 x2 + . . . + d xd and P (x) is a polynomial of degree < d. ii) For all n 0, f (n) satises the linear recurrence relation f (n + d) + 1 f (n + d 1) + 2 f (n + d 2) + . . . + d f (n) = 0. iii) For all n 0, f (n) = where 1 + 1 x + 2 x + . . . + d x =
2 d k i=1
Pi (n)in
k i=1
(1 i x)ei
with the is distinct and each Pi (n) is a univariate polynomial (in n) of degree less than ei . Dention. A generating function of the form erating function.
P (x) Q(x)
is a called a rational gen-

18.312 Algebraic Combinatorics: Mit Opencourseware

Uploaded by

Copyright:

Available Formats

18.312 Algebraic Combinatorics: Mit Opencourseware

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

18.312 Algebraic Combinatorics: Mit Opencourseware

Uploaded by

Copyright:

Available Formats

MIT OpenCourseWare http://ocw.mit.

18.312 Algebraic Combinatorics

Course 18.312: Algebraic Combinatorics

Lecture Notes # 1 Addendum by Gregg Musiker February 4th - 6th, 2009

Recurrence Relations and Generating Functions

For example, = {1, 8, 28, 56, 70, 56, 28, 8, 1}.

88 k In fact if k > 8, sequence as 8

More complicated formal power series

As a consequence, Fk = Fk1 + Fk2 for k 2 implies F (X) = 1 + F1 x + = 1 + F1 x + = 1 + F1 x +

Thus F (X)(1 x x2 ) = 1 + (F1 F0 )x = 1 + 0x and we obtain the rational expression F (X) = 1 . 1 x x2

A Combinatorial Interpretation of the Fibonacci Num bers

Convolution Product Formula

Connection between Linear Recurrences and Rational Generating Functions

is a called a rational gen-

You might also like