The previous lesson ended with fractions such as (s + 3)/((s + 1)(s + 2)) and handled them with one trick: multiply by a factor and set s to its root. That trick covers one case. This lesson takes the cases in order of difficulty, starting from the rule that decides which terms to write.
Definition · The partial-fraction formLet F be a fraction in s whose numerator has lower degree than its denominator. Each factor (s − a)^m of the denominator contributes the m terms A₁/(s − a) + A₂/(s − a)² + ⋯ + Aₘ/(s − a)^m, and each irreducible quadratic factor q contributes (B·s + C)/q.
Here A₁, …, Aₘ, B and C are the numbers to find. The lab refuses a fraction whose numerator is not lower in degree, because the table has no entry for a polynomial in s.
Distinct linear factors: substitute the root
For (s + 3)/((s + 1)(s + 2)) the form is A/(s + 1) + B/(s + 2). Multiply both sides by the whole denominator, so that s + 3 = A·(s + 2) + B·(s + 1). This is an identity in s, which holds at every value. At s = −1 the term with B vanishes and 2 = A. At s = −2 the term with A vanishes and 1 = −B. The lab prints 2/(s + 1) − 1/(s + 2), the same coefficients the previous lesson found.
A repeated factor: every power is needed
For 1/(s·(s + 1)²) the form has three terms, and the worked example below finds them. The reason the middle one cannot be dropped is arithmetic. Suppose the form were only A/s + C/(s + 1)². Multiplying out gives 1 = A·(s + 1)² + C·s. The s² coefficient on the right is A, which must be 0 to match the left side; yet at s = 0 the identity gives A = 1. The two demands contradict each other, so no such A and C exist. The term B/(s + 1) supplies the missing s².
Substituting roots finds only some coefficients: s = 0 gives A and s = −1 gives C, but no value of s isolates B. The remaining coefficient comes from comparing the powers of s on both sides of the identity, here the s² terms, which gives 0 = A + B.
An irreducible quadratic: a two-part numerator
For 1/(s·(s² + 2s + 5)) the quadratic has no real roots, since 2² − 4·5 = −16 is negative. The form is A/s + (B·s + C)/(s² + 2s + 5). At s = 0 the identity 1 = A·(s² + 2s + 5) + (B·s + C)·s gives 1 = 5A, so A = 1/5. The s² terms give 0 = A + B, so B = −1/5, and the s terms give 0 = 2A + C, so C = −2/5.
Completing the square, s² + 2s + 5 = (s + 1)² + 4, shows how this inverts. Write the numerator s + 2 as (s + 1) + 1. The part with s + 1 over (s + 1)² + 4 is the cosine entry shifted by e^(−t). The leftover 1 over (s + 1)² + 4 is half of the sine entry, since that entry has 2 on top. With the factor −1/5 in front, the inverse is 1/5 − (1/5)·e^(−t)·cos(2t) − (1/10)·e^(−t)·sin(2t), which is what the lab prints, with the fraction written as (1/5)/s − (s/5 + 2/5)/((s + 1)² + 4).
(s + 2)/((s + 1)² + 4) = (s + 1)/((s + 1)² + 4) + (1/2)·2/((s + 1)² + 4)
The lab does not trust any of this arithmetic. After it finds the coefficients and writes the inverse, it transforms the inverse back and compares the result with the original fraction exactly. If they differ it reports an error and prints no answer; for the three presets they agree. A decomposition is unique, so a set of coefficients that reproduces the fraction is the set.
Two limits of the lab, named here because they decide what you can type. A quadratic factor whose frequency is irrational, such as s² + 2, is refused with the factor named, since its inverse would hold cos(√2·t). And an irreducible factor of degree three or more is refused. Neither makes the method wrong; they bound what the lab prints.