2
$\begingroup$

I am trying to make this into a partial fractions form but i can't seem to find a way to do it.

The question is here:

Change into a partial fractions form.

\begin{align} \frac{2s}{(s+1)^2(s^2 + 1)} \\\\ \end{align}

i change to \begin{align} \frac{A}{(s+1)}+\frac{B}{(s+1)^2}+\frac{Cs+D}{(s^2+1)} \end{align}

I expand it and group like terms tgt but i am stump at this point \begin{align} 2s = (A+C)s^3 + (B+A+2C+D)s^2 + (A+C+2D)s + (B+A+D) \end{align}

Anyone can help?

$\endgroup$

4 Answers 4

9
$\begingroup$

Such partial fraction decomposition is very easy using the Heaviside cover up method. As I show in that answer, the method generalizes to quadratic denominators. Let's apply it to your problem.

$$\dfrac{a}{x+1} + \dfrac{\color{#c00}b}{(x+1)^2} + \dfrac{cx+d}{x^2+1}\, =\, \dfrac{2x}{(x+1)^2(x^2+1)}\tag{E}$$

Scaling $\rm\,(E)\,$ by $\,(x+1)^2,\,$ then evaluating at its root $\,x = -1\,$ yields

$$\ \ \ \ \color{#c00}b\, =\, \dfrac{2x}{x^2+1}\:\!\Bigg|_{\:\!\large x\,=\,-1} =\, \color{#c00}{ -1}$$

Subtracting out this known summand leaves a fraction in partial form, so we are done:

$$\require{cancel}\!\!\! \dfrac{\color{#0a0}{2x}}{(x+1)^2(x^2+1)}\, -\, \dfrac{\!\!\!\!\!\color{#c00}{-1}}{(x+1)^2}\: =\: \dfrac{\cancel{\color{#0a0}{2x}+x^2+1}}{\cancel{(x+1)^2}(x^2+1)}\, =\, \dfrac{1}{x^2+1}\quad {\bf\small QED}$$


Remark $ $ It's unneeded - but instructive - to use this quadratic Heaviside method for all terms:

Scaling $\rm\,(E)\,$ by $\,x^2+1\,$ then evaluating at its root $\,x = i\,$ yields

$$ \color{#0a0}c\, i + \color{#c00}d\, =\, \dfrac{2i}{(i\!+\!1)^2} \, =\, \dfrac{2i}{2i} \,=\, \color{#c00}1\ \Rightarrow\ \color{#0a0}{c = 0},\ \color{#c00}{d = 1} $$

Scaling $\rm\,(E)\,$ by $\,(x\!+\!1)^2\,$ then evaluating at its root $\,x = w,\,$ so $\,\color{#c0d}{w^2\!+\!1 = -2w}$

$$a(w+1)+ b\, =\, \dfrac{2w}{\color{#c0d}{w^2\!+\!1}}\,=\,\dfrac{\ \ \,2w}{\color{#c0d}{-2w}} \,=\, -1\,\Rightarrow\ a=0,\ b=-1 $$


Below are some further examples asked about in comments.

$$ \dfrac{f}{(x+1)^2} + \dfrac{cx+d}{x^2+1}\, =\, \dfrac{2x+1}{(x+1)^2(x^2+1)}\tag{$\rm \dot E$}$$

Scaling $\rm\,(\dot E)\,$ by $\,(x\!+\!1)^2\,$ then evaluating at its root $\,x = w,\,$

$$f\, =\, \dfrac{\color{#0a0}{2w+1}}{\color{#c0d}{w^2\!+\!1}}\,=\,\dfrac{\color{#0a0}{-w^2}}{\color{#c0d}{-2w}} \,=\, \dfrac{w}2 $$

by $\,w^2\!+2w\!+\!1=0\ \Rightarrow\ \color{#0a0}{2w\!+\!1=-w^2},\,\ \color{#c0d}{w^2\!+\!1 = -2w}$


$$ \dfrac{f}{(x-2)^2} + \dfrac{c}{x-1}\, =\, \dfrac{x}{(x-2)^2(x-1)}\tag{$\rm \ddot E$}$$

Scaling $\rm\,(\ddot E)\,$ by $\,(x\!-\!2)^2\,$ then evaluating at its root $\,x = w,\,$

$$f\, =\, \dfrac{w}{\color{c0d}{w\!-\!1}}\,=\,1+\color{#c00}{\dfrac{1}{w\!-\!1}} = 4\!-\!w $$

$\!\underbrace{{\rm since\ we've }\,\ 0=(w\!-\!2)^2\! = (w\!-\!3)(w\!-\!1) + 1}_{\textstyle \text{polynomial}\ \color{darkorange}{\rm divide}\,\ (w\!-\!2)^2 \:\! \ {\rm by}\ \ (w\!-\!1)\qquad\ \ \ }\,\Rightarrow\, \color{#c00}{\dfrac{1}{w\!-\!1}= 3\!-\!w}$

OR $\bmod w\!-\!1\!:\ (w\!-\!2)^2 \equiv 1^2\,$ so $\,w\!-\!1\mid (w\!-\!2)^2\!\color{#c00}{-\!1^2} = \color{#c00}{(w\!-\!3)(w\!-\!1)}$


Generally given coprime polynomials $\,p,q\,$ over a field $K$ (e.g. $\Bbb Q, \Bbb R, \Bbb C),\,$ we can use the Euclidean algorithm to compute partial fraction decompositions as follows

$$\begin{align} \dfrac{f}p + \dfrac{g}q \,&=\, \dfrac{h}{pq}\\[.4em] \iff\ q\:\!f + p\:\!g\, &=\, h\end{align}\qquad$$

We can solve the latter equation for $\,f,g\,$ using the same methods that we do for integers - by the extended) Euclidean algorithm and closely related methods, e.g. evaluating it mod $p\,$ we deduce $\,qf\equiv h\iff f\equiv h/q\,\pmod{\!p},\,$ which is exactly the same as the Heaviside method above. When $\,q\,$ is linear we can compute $\,1/q\bmod p\,$ simply by (long hand) polynomial $\color{darkorange}{\text{dividing}}$ $\,p\,$ by $\,q\,$ (as we did in the final example above), which amounts to optimizing the case when the extended Euclidean algorithm terminates in a single step.

Similar ideas were employed by Hermite in general partial fraction decomposition algorithms used for integrating rational functions, e.g. see here.

$\endgroup$
7
  • $\begingroup$ How do you get $B$ and $C$ in $\frac{x}{(x-1)(x-2)^2} = \frac{A}{x-1}+\frac{B}{x-2}+\frac{C}{(x-2)^2}$ using the "quadratic Heaviside method" you mentioned? Here you have $\frac{w}{w-1}$ where the denominator is linear, not quadratic. $\endgroup$
    – Neat Math
    May 4 at 14:34
  • $\begingroup$ Also in the original question if the numerator of the RHS is $2x+1$ instead of $2x$ then your approach to get $a$ and $b$ doesn't seem to work. $\endgroup$
    – Neat Math
    May 4 at 14:37
  • $\begingroup$ @NeatMath Please be more precise about what "doesn't seem to work" for you? Where are you stuck? $\endgroup$ May 4 at 14:50
  • $\begingroup$ It's not clear to me how you'd arrive at $\frac{2w+1}{w^2+1}=\frac 12 (w+1) - \frac 12$ given $w^2+1=-2w$ for $\frac{2x+1}{(x+1)^2(x^2+1)}$; it's also not clear to me in the first example I mentioned above ($\frac {x}{(x-1)(x-2)^2}$): given $w^2-4w+4=0$ how do you get $\frac{w}{w-1}=-(w-2)+2$? $\endgroup$
    – Neat Math
    May 6 at 15:47
  • $\begingroup$ @NeatMath I appended text explaining your examples. $\endgroup$ May 12 at 10:08
2
$\begingroup$

Since you arrived to $$\begin{align} 2s = (A+C)s^3 + (B+A+2C+D)s^2 + (A+C+2D)s + (B+A+D) \end{align}$$ identification of the coefficients of the different powers of $s$ gives $$A+C=0$$ $$B+A+2C+D=0$$ $$A+C+2D=2$$ $$B+A+D=0$$ for which the solutions are easily obtained either using matrix or elimination and the final result is : $A=0$,$B=-1$,$C=0$ and $D=1$

$\endgroup$
4
  • $\begingroup$ You are very welcome ! $\endgroup$ May 21, 2014 at 9:43
  • $\begingroup$ @user152399 It's much easier to use Heaviside - see my answer. $\endgroup$ May 21, 2014 at 15:09
  • 1
    $\begingroup$ @BillDubuque. You are right but, by the end, Jyrki Lahtonen's solution is the fastest. Cheers. $\endgroup$ May 22, 2014 at 3:46
  • 1
    $\begingroup$ @Claude I meant easiest in general (ad-hoc methods don't generally apply). $\endgroup$ May 22, 2014 at 4:39
2
$\begingroup$

Hint: It is probably easiest to observe that $$ 2s=(s+1)^2-(s^2+1). $$ Use that in the numerator and see what you can cancel from the two terms you get.

$\endgroup$
2
  • $\begingroup$ Sorry, I just love spotting ad hoc tricks. $\endgroup$ May 21, 2014 at 9:31
  • $\begingroup$ We can discover this (vs. pull it out of a hat) if we use the Heaviside cover-up method - see the 3rd displayed equation in my answer - where the numerators are $\,\color{#0a0}{2x} + \color{#c00}{x^2+1} = (x+1)^2\,$ when over the common denominator $(x+1)^2(x^2+1).\,$ That way it's no longer ad-hoc. $\endgroup$ May 12 at 15:20
0
$\begingroup$

HINT:

Using Partial Fraction Decomposition formula, $$\frac{2s}{(s+1)^2(s^2+1)}=\frac A{s+1}+\frac B{(s+1)^2}+\frac{Cs+D}{s^2+1}$$

$\endgroup$
4
  • $\begingroup$ I tried it but i only am able to find B with Cover up rule. $\endgroup$
    – Alex
    May 21, 2014 at 8:38
  • $\begingroup$ @user152399, If you have four unknowns with four linear equations, right? $\endgroup$ May 21, 2014 at 9:18
  • $\begingroup$ I don't understand what you mean. $\endgroup$
    – Alex
    May 21, 2014 at 9:23
  • $\begingroup$ @lab Using quadratic Heaviside is usually simpler than solving such systems of equations, see the Remark in my answer. $\endgroup$ May 21, 2014 at 15:44

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.