Problem set 7: Formal Power Series

1 Formal Power Series

1.1 Definition: Formal Power Series

Consider the set of infinite sequences

\[ (a_0,a_1,\ldots,a_n,\ldots), \]

where the coefficients belong to \(\mathbb{Q}\), \(\mathbb{R}\), or \(\mathbb{C}\).

We define addition componentwise:

\[ (a_0,a_1,\ldots)+(b_0,b_1,\ldots) = (a_0+b_0,a_1+b_1,\ldots). \]

We define multiplication by

\[ (a_0,a_1,\ldots)\cdot(b_0,b_1,\ldots) = (c_0,c_1,\ldots), \]

where

\[ c_n=\sum_{j=0}^{n}a_jb_{n-j}. \]

Let

\[ 1=(1,0,0,\ldots), \qquad t=(0,1,0,\ldots). \]

(a) Find

\[ t^2,t^3,\ldots,t^n. \]

(b) Prove that every sequence

\[ (a_0,a_1,\ldots,a_n,\ldots) \]

can be represented uniquely in the form

\[ a_0\cdot 1+a_1t+a_2t^2+\cdots+a_nt^n+\cdots. \]

Expressions of the form

\[ a_0+a_1t+a_2t^2+\cdots \]

with the addition and multiplication operations defined above are called formal power series.

The coefficient \(a_0\) is called the constant term of the series.

(c) Prove that the usual algebraic properties of polynomials extend to formal power series:

  • Commutativity of addition and multiplication.
  • Associativity of addition and multiplication.
  • Distributivity of multiplication over addition.

(d) Let \(f(t)\) be a formal power series.

Is the substitution

\[ f(g(t)) \]

well-defined for an arbitrary formal power series \(g(t)\)?

What if \(g(t)\) has zero constant term?

1.2 Formal Differentiation

Let

\[ A=a_0+a_1t+a_2t^2+\cdots+a_nt^n+\cdots \]

be a formal power series.

Its formal derivative is defined by

\[ A' = a_1+2a_2t+3a_3t^2+\cdots+na_nt^{n-1}+\cdots. \]

Higher derivatives are defined by repeated differentiation, and \(A^{(n)}\) denotes the \(n\)-th derivative.

Prove the following properties of formal differentiation.

(a) Linearity

For any formal power series \(A,B\) and constants \(\alpha,\beta\in\mathbb R\),

\[ (\alpha A+\beta B)' = \alpha A'+\beta B'. \]

(b) Leibniz Rule

Prove that

\[ (AB)'=A'B+AB'. \]

(c) Taylor Coefficients

Prove that the coefficient \(a_n\) of a formal power series

\[ A=\sum_{k=0}^{\infty}a_kt^k \]

is the constant term of the series

\[ \frac{A^{(n)}}{n!}. \]

(a) Let

\[ F=1+t+t^2+t^3+\cdots \]

and

\[ G=1-t+t^2-t^3+\cdots. \]

Find

\[ F+G \qquad\text{and}\qquad F\cdot G. \]

(b) Let

\[ F=\sum_{k=0}^{\infty}\frac{1}{k!}t^k, \qquad G=\sum_{k=0}^{\infty}\frac{(-1)^k}{k!}t^k. \]

Find

\[ F\cdot G \qquad\text{and}\qquad F^2. \]

Do there exist two nonzero formal power series \(F\) and \(G\) such that

\[ F\cdot G=F+G? \]

Prove your answer or provide an example.

1.3 Invertible Formal Power Series

A formal power series \(G\) is called invertible if there exists a formal power series \(F\) such that

\[ F\cdot G=1. \]

In this case, \(F\) is called the multiplicative inverse of \(G\) and is denoted by

\[ F=G^{-1}. \]

Let

\[ G=a_0+a_1t+a_2t^2+a_3t^3+\cdots. \]

Prove that \(G\) is invertible if and only if

\[ a_0\neq 0. \]

Moreover, prove that the inverse series \(G^{-1}\) is unique.

Find the coefficients of the following formal power series.

(a)

\[ \frac{1}{1-t}. \]

(b)

\[ \frac{1}{2-t}. \]

(c)

\[ \frac{1}{3+4t}. \]

(d)

\[ \frac{1}{(1-t)^2}. \]

(e)

\[ \frac{1}{(1-t)^m}, \qquad m\in\mathbb N. \]

(a) Is the following identity valid for formal power series?

\[ \frac{1}{1-t^2}\cdot\frac{1}{1+t^2} = \frac{1}{1-t^4}. \]

(b) By comparing the coefficients of the formal power series on both sides, derive a combinatorial identity.

(a) Under what conditions on formal power series \(F\) and \(G\) is it possible to divide \(F\) by \(G\)?

In other words, determine when the equation

\[ G\cdot X=F \]

has a solution \(X\) in the ring of formal power series.

(b) Under what conditions on a formal power series \(F\) does the equation

\[ X^2=F \]

have a solution?

(c) Let

\[ X=\sum_{k=0}^{\infty}\alpha_kt^k \]

be the formal power series determined by the equation

\[ X^2=1+t. \]

Find the coefficients

\[ \alpha_0,\alpha_1,\alpha_2,\alpha_3. \]

1.4 Generalized Binomial Coefficients

Let \(\alpha\in\mathbb R\).

Define the generalized binomial coefficient by

\[ \binom{\alpha}{n} = \frac{\alpha(\alpha-1)\cdots(\alpha-n+1)}{n!}, \qquad n\geq 1, \]

with

\[ \binom{\alpha}{0}=1. \]

(a) Prove that if \(\alpha\in\mathbb N\) and \(\alpha\geq n\), then

\[ \binom{\alpha}{n} \]

coincides with the usual binomial coefficient.

(b) Prove the generalized Vandermonde identity:

\[ \boxed{ \binom{\alpha+\beta}{n} = \sum_{j=0}^{n} \binom{\alpha}{j} \binom{\beta}{n-j}. } \]

(c) Define the formal power series

\[ (1+t)^\alpha = \sum_{n=0}^{\infty} \binom{\alpha}{n}t^n. \]

Prove that when \(\alpha\in\mathbb N\), this definition agrees with the usual polynomial expression for \((1+t)^\alpha\).

(d) Prove that

\[ \boxed{ (1+t)^\alpha(1+t)^\beta = (1+t)^{\alpha+\beta}. } \]

(e) Using the definition above, find the coefficients of the formal power series

\[ (1+t)^{1/2}. \]

Compare your answer with the result obtained in Problem 8(c).