hours: Office Tuesdays from 2:15 to 2:45 p.m. (in B107 until further notice) Thursdays from 12:15 to 12:45 p.m. (in 813 Van Vleck) Theorem: If f(0) = 0, then 1/(1-f) = 1 + f + f^2 + f^3 + ... Proof: Let g = 1 + f + f^2 + f^3 + ... (check that this is well-defined: that's where we use f(0)=0). Then (by the continuity of addition and multiplication) (1-f)g = (1-f) + (f-f^2) + (f^2-f^3) + ... = 1. In a similar way, our proof that (1-x-x^2) (1+x+2x^2+3x^3+5x^4+...)=1. now makes rigorous sense. Likewise: 1/(1-x) = 1+x+x^2+x^3+... 1/(1-x)^2 = 1+2x+3x^2+4x^3+... 1/(1-x)^3 = 1+3x+6x^2+10x^4+... Recognize these numbers? 1/(1-x)^n = 1+nx+[n(n+1)/2]x^2+[n(n+1)(n+2)/6]x^3+... We can add more variables. 1/(1-x-y) = 1 + (x+y) + (x+y)^2 + (x+y)^3 + ... = 1 + (x+y) + (x^2 + 2xy + y^2) + (x^3 + 3 x^2 y+3 x y^2 + 3y^3) + ... (arrange as Pascal's triangle) = sum_{n=0}^infty sum_{k=0}^n {n choose k} x^k y^{n-k} On the other hand, let's divide the numerator and denominator by 1-x: (1/(1-x))/(1-y/(1-x)) = 1/(1-x)+y/(1-x)^2+y^2/(1-x)^3 = (1+x+x^2+x^3+...) + y(1+2x+3x^2+4x^3+...)+ y^2 (1+3x+6x^2+10x^3+...) + ... (Use the fingers method to verify that this is right.)
In the ring of formal power series, the binomial theorem tell us that if n is any non-negative integer, (1+f)^n is equal to the "infinite sum" 1 + [n ]f + [n(n-1)/2] f^2 + [n(n-1)(n-2)/6] f^3 + ... (which isn't so infinite, since all but finitely many terms vanish). But in fact this is true for negative integers n as well, as long as f(0)=0 (i.e., the constant term of f is 0): e.g., with n=-1, (1+f)^{-1} = 1 + [-1] f + [1] f^2 + [-1] f^3 + ... This requires proof, but I won't take the time to prove it now. In fact, we can use the formula (1+f)^c = 1 + [c/1!]f + [c(c-1)/2!]f^2 + [c(c-1)(c-2)/3!]f^3 + ... to define (1+f)^c for all real numbers c (or even complex numbers c). It can be shown that (1+f)^c (1+f)^d = (1+f)^(c+d) and ((1+f)^c)^d = (1+f)^(cd). Remark that ((1+f)^c)^d makes sense. "Did you remember to write down whom you worked with?" "Did you remember to write down how much time you spent on each of the two problems?" Collect homework. Reading for today: Intro to generating functions: Wilf (generatingfunctionology), sections 1.1-1.3 (pages 1-10). Questions? Remind students that for questions that I ask you, where the answer takes more than five or ten seconds, you should be looking at the answerer, not me. If you shouldn't be looking me, at I'll point my hands at the person you should be looking at. Motivate infrastructure of generating functions and recurrences: the role of sequences in combinatorics.
Problem: Find a formula for the sequence 1, 2, 5, 14, 41, ... with initial conditions f(0)=1 and f(1)=2, satisfying the recurrence (*) f(n+2)=4f(n+1)-3f(n) for all n 0. We set up a generating function F(x) = ... sum_{n=0}^{infty} f(n) x^n = 1 + 2x + 5x^2 + ... . Wilf's method: Multiply (*) by x^n for all ... n 0, and then sum. _{n 0} f(n+2) x^n = 4 _{n 0} f(n+1) x^n 3 _{n 0} f(n) x^n Re-index each term: _{n 2} f(n) x^{n-2} = 4 _{n 1} f(n) x^{n-1} 3 _{n 0} f(n) x^n (F(x)-1-2x)/x^2 = 4(F(x)-1)/x 3F(x) Multiply by x^2: F(x)-1-2x = 4 x (F(x)-1)x 3 x^2 F(x) (1 4x + 3x^2) F(x) = 1 + 2x 4x = 1 2x F(x) = (1 2x)/(1 4x + 3x^2) Check with Maple: taylor((1 - 2*x)/(1 - 4*x + 3*x^2),x,10); And we can proceed from there to find an exact formula. But rather than do it that way, let's start the problem over. My method: Even before we know what the initial conditions are, we can check that the recurrence relation gives (1-4x+3x^2) F(x) = (1-4x+3x^2)(f(0)+f(1)x+f(2)x^2+f(3)x^3+...) = [f(0)]+[f(1)-4f(0)]x+[f(2)-4f(1)+3f(0)]+[f(3)-4f(2)+3f(2)]+... =some linear polynomial in x. Conversely, any formal power series of the form (ax+b)/(1-4x+3x^2)
will have as its coefficients a sequence of numbers satisfying (*). So we have a two-parameter family of solutions to (*), and these are the only...