Theoretical Partial Fractions

We will follow the iterative procedure to partial fraction decomposition as described in this Wikipedia article.

Problem 1. Given polynomials f, g, prove that there exist unique polynomials q, r with \deg(r) < \deg(g) such that

\displaystyle \frac{f(x)}{g(x)} = q(x) + \frac{r(x)}{g(x)},\quad g(x) \neq 0

(Click for Solution)

Solution. By polynomial long division, there exist unique polynomials q,r with \deg(r) < \deg(g) such that

\displaystyle f = q \cdot g + r.

Hence, any x \in \mathbb R such that g(x) \neq 0,

\displaystyle \frac{f(x)}{g(x)} = \frac{q(x) \cdot g(x) + r(x)}{g(x)} = q(x) + \frac{r(x)}{g(x)}.

Henceforth, we assume that \deg(f) < \deg(g).

Definition 1. For polynomials f, g, denote f \mid g if there exists a polynomial d such that g = d \cdot f. Denote \gcd(f, g) = 1 if for any polynomial h, h \mid f and h \mid g implies that h is a nonzero constant polynomial.

Problem 2. For non-constant polynomials g_1, g_2 such that \gcd(g_1,g_2) = 1, prove that there exists unique polynomials r_1, r_2 with \deg(r_i) < \deg(g_i) such that

\displaystyle \frac{f}{g_1 \cdot g_2} = \frac{r_1}{g_1} + \frac{r_2}{g_2}.

(Click for Solution)

Solution. The claim is equivalent to ensuring the polynomials r_1,r_2 satisfy the equation

f = r_1 \cdot g_2 + r_2 \cdot g_1,

which resembles Bézout’s lemma. In what follows, we adopt the proof therein. Consider the collection of polynomials

\mathcal K := \{s_1 \cdot g_2 + s_2 \cdot g_1 : s_1,s_2 \in \mathbb R[x]\backslash \{0\}\},

where we denote the set of all polynomials with real coefficients by \mathbb R[x], and the subset of \mathbb N defined by

K := \{\deg(s) : s \in \mathcal K\} \equiv \deg(\mathcal K).

We claim that \mathcal K is non-empty. Use polynomial long division to obtain suitable polynomials q,r such that

g_2 = q \cdot g_1 + r,\quad \deg(r) < \deg(g_1).

If r = 0, then g_1 \mid g_2. Since g_1 \mid g_1 as well, we have g_1 is a nonzero constant, a contradiction. Therefore, r \neq 0 and r = 1 \cdot g_2 + (-q) \cdot g_1 \in \mathcal K. Therefore, \mathcal K is non-empty, so that K \subseteq \mathbb N is non-empty. By the well-ordering principle, K = \deg(\mathcal K) has a smallest element d.

Find polynomials s_1,s_2 such that

\deg(\underbrace{s_2 \cdot g_1 + s_1 \cdot g_2}_t) = d.

We claim that t \mid g_1. Use polynomial long division again to obtain suitable polynomials \tilde q, \tilde r such that

g_1 = \tilde q \cdot t + \tilde r,\quad \deg(\tilde r) < \deg(t) = d.

If \tilde r \neq 0, then

\tilde r = g_1 - \tilde q \cdot t = (1 - s_2 \cdot t) \cdot g_1 + (- s_1 \cdot t) \cdot g_2 \in \mathcal K,

and d \leq \deg(\tilde r) < d, a contradiction. Therefore, \tilde r = 0 so that t \mid g_1. Symmetrically, t \mid g_2, so that t is a nonzero constant. Without loss of generality, we can assume t = 1, so that

f = f \cdot 1 = f \cdot (s_2 \cdot g_1 + s_1 \cdot g_2) = (f \cdot s_2) \cdot g_1 + (f \cdot s_1) \cdot g_2.

Dividing both sides by g_1 \cdot g_2 yields

\displaystyle \frac{f}{g_1 \cdot g_2} = \frac{f \cdot s_1}{g_1} + \frac{f \cdot s_2}{g_2}.

Use Problem 1 to obtain polynomials q_1, r_1 such that

\displaystyle \frac{f}{g_1 \cdot g_2} = q_1 + \frac{r_1}{g_1} + \frac{f \cdot s_2}{g_2} = \frac{r_1}{g_1} + \frac{q_1 \cdot g_2 + f \cdot s_2}{g_2}.

Defining r_2 := q_1 \cdot g_2 + f \cdot s_2, it remains to prove that \deg(r_2) < \deg(g_2). Indeed, since

f = r_1 \cdot g_2 + r_2 \cdot g_1 \quad \Rightarrow \quad r_2 \cdot g_1 = f - r_1 \cdot g_2,

we have

\begin{aligned}\deg(r_2) &= \deg(f - r_1 \cdot g_2) - \deg(g_1) \\ &\leq \max\{ \deg(f), \deg(r_1 \cdot g_2)\}  - \deg(g_1) \\ &< \max\{ \deg(g_1 \cdot g_2), \deg(g_1 \cdot g_2)\}  - \deg(g_1) \\ &= \deg(g_1 \cdot g_2) - \deg(g_1) \\ &= \deg(g_1) + \deg(g_2) - \deg(g_1) = \deg(g_2), \end{aligned}

as required.

Problem 3. For non-constant polynomials g_1, \dots, g_n such that \gcd(g_i,g_j) = 1 for any i \neq j, prove that there exists unique polynomials r_1, \dots, r_n with \deg(r_i) < \deg(g_i) such that

\displaystyle \frac{f}{g_1\cdot \dots \cdot g_n} = \sum_{i=1}^n \frac{r_i}{g_i}.

(Click for Solution)

Solution. We prove by strong induction on n. Problem 2 establishes the base case n = 2. For the general case n = k+1, observe that \gcd(g_{k+1}, g_1 \cdot \dots \cdot g_k) = 1. Apply the n = 2 case and the n = k case to obtain the unique decomposition

\begin{aligned}\frac{f}{g_1\cdot \dots \cdot g_k g_{k+1}} &= \frac{r_0}{g_1\cdot \cdots \cdot g_k} + \frac{r_{k+1}}{g_{k+1}},\quad \deg(r_0) < \deg(g_1\cdot \cdots \cdot g_k), \\ &= \sum_{i=1}^k \frac{r_i}{g_i} + \frac{r_{k+1}}{g_{k+1}} = \sum_{i=1}^{k+1} \frac{r_i}{g_i}.\end{aligned}

Definition 2. Call a polynomial f irreducible if for polynomials g,h \neq f, g \cdot h = f implies that either g or h is constant.

Problem 4. Let f, g be polynomials such that \deg(g) = \deg(f)-1. Prove that if f is irreducible, then \gcd(f, g) = 1.

(Click for Solution)

Solution. Let h be a polynomial such that h \mid f and h \mid g. Find polynomials s, t such that f = h \cdot s, g = h \cdot t. If h is non-constant, then s \neq 0 is, so that g = \frac 1s \cdot f \cdot t, which implies

\deg(g) = \deg(f) + \deg(t) \geq \deg(f) > \deg(g),

a contradiction. Therefore, h is constant.

Problem 5. Suppose g is irreducible. Prove that there exists unique polynomials s_1, \dots, s_n with \deg(s_i) < \deg(g) such that

\displaystyle \frac{f}{g^n} = \sum_{i=1}^n \frac{s_i}{g^i}.

(Click for Solution)

Solution. For the polynomial g(x) = \sum_{i=0}^n a_ix^i, define its formal derivative by g'(x) = \sum_{i=1}^n i a_i x^{i-1}. Since g is irreducible and \deg(g') = \deg(g)-1, by Problem 4, \gcd(g, g') = 1. By the proof in Problem 2, there exists polynomials s, t such that g \cdot s + g' \cdot t = 1. We prove the result by induction. The case n = 1 is obvious. For the general case n = k+1,

\displaystyle \frac{f}{g^{k+1}} = \frac{f \cdot 1}{g^{k+1}} = \frac{f \cdot g \cdot s + f \cdot g' \cdot t}{g^{k+1}}.

Use polynomial long division to obtain polynomials q, s_{k+1} such that

f \cdot g' \cdot t = q \cdot g + s_{k+1},\quad \deg(s_{k+1}) < \deg(g).

Substituting,

\displaystyle \frac{f}{g^{k+1}} = \frac{f \cdot g \cdot s + q \cdot g + s_{k+1}}{g^{k+1}} = \frac{f \cdot s + q}{g^k} + \frac{s_{k+1}}{g^{k+1}}.

All that remains is to verify that \deg(f \cdot s + q) < k \cdot \deg(g):

\begin{aligned} \deg(f \cdot s + q) &\leq \max \{\deg(f ), \deg(s_{k+1})\}-\deg(g) \\ &< (k+1) \cdot \deg(g) - \deg(g) = k \cdot \deg(g). \end{aligned}

By the induction hypothesis,

\displaystyle \frac{f}{g^{k+1}} = \frac{f \cdot s + q}{g^k} + \frac{s_{k+1}}{g^{k+1}} = \sum_{i=1}^k \frac{s_i}{g^i} + \frac{s_{k+1}}{g^{k+1}} = \sum_{i=1}^{k+1} \frac{s_i}{g^i}.

Problem 6. Given irreducible polynomials g_1,\dots, g_n such that \gcd(g_i,g_j) = 1 for any i \neq j, prove that there exist unique polynomials s_{k,\ell} with \deg(s_{k,\ell}) < \deg(g_k) such that

\displaystyle \frac{f}{\prod_{k=1}^n g_k^{\alpha_k}} = \sum_{k=1}^n \sum_{\ell=1}^{\alpha_k} \frac{s_{k,\ell}}{g_k^\ell},\quad \deg(s_{k,\ell}) < \deg(g_k).

(Click for Solution)

Solution. We note that for any i \neq j, \gcd(g_i^{\alpha_i}, g_j^{\alpha_j}) = 1. By Problem 3, there exist unique polynomials r_1,\dots, r_n such that

\displaystyle \frac{f}{\prod_{k=1}^n g_k^{\alpha_k}} = \sum_{k=1}^n \frac{r_k}{g_k^{\alpha_k}},\quad \deg(r_k) < \deg(g_k^{\alpha_k}).

By Problem 5, for each k, there exist unique polynomials s_{k,\ell} such that

\displaystyle \frac{f}{\prod_{k=1}^n g_k^{\alpha_k}} = \sum_{k=1}^n \frac{r_k}{g_k^{\alpha_k}} = \sum_{k=1}^n \sum_{\ell=1}^{\alpha_k} \frac{s_{k,\ell}}{g_k^{\ell}},\quad \deg(s_{k,\ell}) < \deg(g_k).

Problem 7. Given distinct real numbers \alpha_i and irreducible quadratic polynomials q_j that have no common factors, determine the partial fraction decomposition of

\displaystyle \frac{f(x)}{\prod_{i=1}^m (x-\alpha_i)^{\beta_i} \cdot \prod_{j=1}^n q_j^{\gamma_j}(x)} ,\quad \deg(f) < \sum_{i=1}^m \beta_i + 2 \cdot \sum_{j=1}^n \gamma_j.

(Click for Solution)

Solution. By Problem 6,

\displaystyle \frac{f(x)}{\prod_{i=1}^m (x-\alpha_i)^{\beta_i} \cdot \prod_{j=1}^n q_j^{\gamma_j}(x)} = \sum_{i=1}^m \sum_{k=1}^{\beta_i} \frac{A_{i,k}}{(x-\alpha_i)^{k}} + \sum_{j=1}^n \sum_{\ell=1}^{\alpha_j} \frac{B_{j,\ell} \cdot x + C_{j,\ell}}{q_j^{\ell}(x)}.

Finally, by the fundamental theorem of algebra, every polynomial can be factorised into a product of linear and quadratic factors. By Problem 7, therefore, any ratio of polynomials can be written as a sum of partial fractions.

—Joel Kindiak, 16 Sept 25, 1713H

,

Published by


Leave a comment