0

I've just reviewed this answer: partial fraction for complex roots

Whilst I understand most of it, there is one component I am having some trouble understanding. in particular

$$\frac {2s^2+5s+12} {(s^2+2s+10)(s+2)}=\frac a{s+2}+\frac {b+cs}{s^2+2s+10}$$

Can someone take me through where the cs term comes from? I assume it something related to simplifying

$$\frac {A} {s+1+3i} + \frac {B}{s+1-3i}$$

but I cannot seem to get to the bottom of it. I end up with

$$\frac {A(s+1-3i)+B(s+1+3i)} {s^2+2+10}$$

and unsure where to go from here to get something that looks like the term

$$\frac {b+cs}{s^2+2s+10}$$

Any help would be greatly appreciated. Thanks.

snuh
  • 1

1 Answers1

1

The $cs$ terms is present in real partial fraction decomposition. You are decomposing into proper rational functions, but when the denominator has degree $2$, your numerator must have degree $1$ to represent all possible proper rational functions. The correct form for a general factor $a x^n + \cdots + b$ of the denominator is $$ \frac{A x^{n-1} + B x^{n-2} + \cdots + C x + D}{a x^n + \cdots + b} \text{,} $$ that is, the numerator is always one degree less then the factor in the denominator. (If the factor appears to a power, the numerator still gets its degree from the factor; it ignores the power.)

This is demonstrated on the partial fractions page as example 2.

More pointedly, say one needs to decompose $$ \frac{x^2}{x(x^2-4x+8)} $$ There is no solution in reals $A$ and $B$ such that $$ \frac{x^2}{x(x^2-4x+8)} = \frac{A}{x} + \frac{B}{x^2-4x+8} \text{.} $$ When you clear the denominators, to match the quadratic coefficients, $A = 1$, but there is no choice of $B$ such that $-4x+8+Bx = 0$. For a solution to exist, you need $$ \frac{x^2}{x(x^2-4x+8)} = \frac{A}{x} + \frac{Bx + C}{x^2-4x+8} $$ to arrive at $-4x+8+Bx + C = 0$, which does have a solution.

Now, as to your particular problem ... Clearing denominators, we have \begin{align*} 2s^2+5s+12 &= \frac{a (s^2+2s+10)(s+2)}{s+2}+\frac {(b+cs)(s^2+2s+10)(s+2)}{s^2+2s+10} \\ &= a (s^2+2s+10) + (b+cs)(s+2) \\ &= (as^2 + 2as + 10a) + (b(s+2)+cs(s+2)) \\ &= (as^2 + 2as + 10a) + (bs+2b+cs^2 + 2cs) \\ &= (a+c) s^2 + (2a + b + 2c)s + (10a + 2b) \text{.} \end{align*} For two polynomials (real or complex) to agree, they must agree in each coefficient, so we have the system \begin{align*} a + c &= 2 \\ 2a+b+2c &= 5 \\ 10a+2b &= 12 \end{align*} Subtracting two copies of the second from the first, $b = 1$. Using that in the third, $a = 1$. Using that in the first, $c = 1$. Therefore, $$ \frac {2s^2+5s+12}{(s+2)(s^2+2s+10)} = \frac{1}{s+2} + \frac{s+1}{s^2+2s+10} \text{.} $$

That's the real decomposition. Continuing, using the complex factoring $$ s^2+2s+10 = (s - (-1-3\mathrm{i}))(s - (-1+3\mathrm{i})) \text{,} $$ $$ \frac{s+1}{s^2+2s+10} = \frac{d}{s - (-1-3\mathrm{i}))} + \frac{e}{s - (-1+3\mathrm{i}))} \text{.} $$ Clearing denominators, we obtain \begin{align*} s+1 &= \frac{d(s^2+2s+10)}{s - (-1-3\mathrm{i}))} + \frac{e(s^2+2s+10)}{s - (-1+3\mathrm{i}))} \\ &= d(s - (-1+3\mathrm{i})) + e(s - (-1-3\mathrm{i})) \\ &= ds - d(-1+3\mathrm{i})) + es - e(-1-3\mathrm{i})) \\ &= (d+e)s +d -3d\mathrm{i} + es + e +3e\mathrm{i} \\ \end{align*} Again equating coefficients, \begin{align*} d+e &= 1 \\ d+e + 3(-d+e)\mathrm{i} &= 1 + 0 \mathrm{i} \text{.} \end{align*} Subtracting the first of these from the second, $3(-d+e)\mathrm{i} = 0$, forcing $d = e$. Then the first gives $d = e = 1/2$.

The development above first expanded in factors of the denominator that were reducible over the reals, then factored further over the complexes. These can be done all at once.

$$ \frac{2s^2+5s+12}{(s^2+2s+10)(s+2)} = \frac{a }{s+2}+\frac{d}{s - (-1-3\mathrm{i}))} + \frac{e}{s - (-1+3\mathrm{i}))} \text{.} $$ Now all the denominators are (powers of) factors of degree $1$, so their numerators have degree $0$.

Eric Towers
  • 67,037
  • oof this is more complicated than I bargained for. Okay thanks for the answer, I think I understand, not sure if this is the kind of answer my prof is expecting from me though – snuh May 09 '20 at 01:11