6

Rational Functions & Partial Fraction Decomposition

x 2 4x + 1 . We call a rational function x 3 + 5x 2 3 proper if the degree of the numerator is less than the degree of the denominator, and improper A rational function is a ratio of polynomials, such as otherwise1 . You should be quite comfortable with the idea of combining rational functions into one through nding a common denominator, but in calculus we will often nd it necessary to reverse this procedure. Fortunately, this can always be done... provided that we can manage to factor the denominator. Fact: Any proper rational function can be expressed as the sum of simpler rational functions, whose denominators are either linear or irreducibly quadratic. Example: 5x2 5x + 4 3x 1 2 = 2 + 3 2 x x x2 x +x+1 x2 How do we do this? The Method of Partial Fractions essentially consists of guessing the form of the decomposition on the right by taking advantage of the experience we have in working in the other direction. If we think about all of the various things that can happen when we combine rational functions together, we arrive at the following procedure. We begin by factoring the denominator as far as possible, into linear and irreducibly quadratic factors (the Fundamental Theorem of Algebra guarantees that this is always possible in theory, although in practice it can be dicult). We then predict the form of the partial fraction decomposition using three rules (which are admittedly dicult to explain clearly, but which should become clearer through the examples): 1. For any linear factor (c1 x + c0 ) in the denominator, the decomposition will contain a A term of the form , for some constant A. c1 x + c0 2. For any quadratic factor (c2 x2 + c1 x + c0 ) in the denominator, the decomposition will Ax + B contain a term of the form , for some constants A and B . c 2 x2 + c 1 x + c 0
Some authors use the term marginally proper if the numerator and denominator are of the same degree, in which case our term proper needs to be replaced by strictly proper.
1

3. For any factor which is repeated, n times, we need n terms of the forms given by Rules 1 & 2, but distinguished by the exponents 1 through n. If youre not sure why these rules work the way they do, just try doing some of the following examples in reverse (that is, go through the procedure of putting the results over the common denominator), and you should begin to see the logic behind them! Examples: 1. Consider x+2 x+2 . Factoring the denominator gives , so we only need x 2 + 5x + 4 (x + 4)(x + 1) x+2 A B Rule 1: 2 = + . Now, to determine the values of A and B , the x + 5x + 4 x+4 x+1 idea is to put these expressions over a common denominator again and match up the coecients: x+2 A B A(x + 1) + B (x + 4) = + = x 2 + 5x + 4 x+4 x+1 x 2 + 5x + 4 We can cancel the denominators, and this leaves us with x + 2 = A(x + 1) + B (x + 4) = (A + B )x + (A + 4B ). The only way these two polynomials can be equal is if the coecients are equal, so we have the pair of equations nd A = 2/3, and B = 1/3, and so we have our result: x+2 1 2 1 = + . x 2 + 5x + 4 3 x+4 x+1 Note: Once we get to the point marked *, we could nd A & B more quickly by substituting values for x. This cover-up trick doesnt work so well when we have quadratic factors, so youll still need the concept of matching coecients, but it does give us a useful shortcut when the factors are linear:2 set x = 1 to get 1 = 3B (so B = 1/3) set x = 4 to get 2 = 3A (so A = 2/3). 2. Now consider x . We need Rules 1 and 2 here: (x + 1)(x2 + x + 1) x A Bx + C = + 2 2 (x + 1)(x + x + 1) x+1 x +x+1
It might occur to you that these are precisely the values of x at which our original function is undened, so we shouldnt be allowed to do this! However, we could get exactly the same results by taking limits as x approaches 1 and 4, so the problem isnt really a problem after all (to use some terminology well introduce properly later on, these points are removable discontinuities).
2

1= A+B 2 = A + 4B

. Solving this, we

x = A(x2 + x + 1) + (Bx + C )(x + 1).

Since we have one quadratic factor, the cover-up method can be used to get one of the constants quickly: by setting x = 1 we immediately nd that A = 1. To get the remaining two values, though, its quickest to match the coecients: Comparing x2 terms, we see that we must have 0 = A + B , while comparing constant terms, we nd that we must have 0 = A + C .3 Therefore B = 1 and C = 1, so x x+1 1 = 2 . (x + 1)(x2 + x + 1) x +x+1 x+1

3. Consider

(x + 2) (x2 + 1)3 2x 5 x 3 2) (x2

2x5 x3

. We need all three rules here:

(x +

+ 1)

A Bx + C Dx + E Fx + G + 2 + 2 + 2 x+2 x +1 (x + 1) (x2 + 1)3

These problems get very tedious, so we wont nish this one! Fact: If f (x) is an improper rational function, then it can always be written as the sum of a polynomial and a proper rational function. How? One option is long division (you may have seen synthetic division, but this only works for linear denominators). Example: Rewrite f (x) = x3 1 . x 2 + 2x + 1

Solution: The long division calculution should look something like this:
3 To see why this is more ecient than relying exclusively on the cover-up method, consider trying to complete this example that way. Setting x = 0 does at least look helpful; it gives us 0 = A + C , but this is entirely equivalent to comparing the constant terms. After that, though, there are really no more useful values of x; the best we can do is pick a nice round number like x = 1. This gives us the equation 1 = 3A + 2(B + C ). This is certainly sucient for us to determine the values of all three constants, but this last equation is denitely more complicated than the equation we obtained from comparing the coecients of x2 . The lesson is that the most ecient way to proceed is to use a sensible combination of the two techniques.

x2 x 2 + 2x + 1 x3 1 x 3 + 2x 2 + x 2x 2 x 1 2x2 4x 2 3x + 1 1 3x + 1 = x2+ 2 , and now we are in a position to + 2x + 1 x + 2x + 1 factor the denominator and proceed with the partial fraction decomposition. This tells us that x2 Another option is to include the polynomial terms in our partial fraction decomposition procedure directly: Since deg (num) deg (denom) = 1, and x2 + 2x + 1 = (x + 1)2 , we know that x3 1 C D = Ax + B + + 2 x + 2x + 1 x + 1 (x + 1)2 x3 1 Ax (x + 1)2 + B (x + 1)2 + C (x + 1) + D = (x + 1)2 (x + 1)2 x3 1 = A x3 + 2x2 + x + B x2 + 2x + 1 + C (x + 1) + D. x3

= =

Setting x = 1 gives 2 = D.

Setting x = 0 gives 1 = B + C + D, i.e. B + C = 1. Comparing x3 terms tells us that 1 = A, while the x2 terms tell us that 0 = 2A + B , so B = 2 and then C = 3. Thus we nd that x2 x3 1 3 2 =x2+ . + 2x + 1 x + 1 (x + 1)2

Note that we might have realized at the beginning that A = 1, by doing the rst step of long division in our heads.

Application to Curve Sketching (if time permits)

 the curve passes through (1, 0) and through (0, 1) f lies above the line y = x 2 when x 1 (because f (x) x 2 + f lies below the line y = x 2 when x 1 (for the same reason) as x 1, f , since the
2 (x+1)2 3 x+1 )

term dominates.

This is all we need to determine that the graph looks like this:

Figure 1:

Example: Sketch the graph of f (x) = Solution: Use long division:

x3 x2 x + 3 . x1

x2 x1

x3 x2 x + 3 x3 x2 x + 3 x + 1 2 2 . x1

f ( x) = x2 1 +

From this we can see that f has a vertical asymptote at x = 1, and approaches the parabola y = x2 1 asymptotically as x . The y -intercept is (0, 3). The other part of the graph doesnt have any intercepts, so to anchor the graph lets just nd one point: notice that f (2) = 5. Heres our graph: 5

Figure 2: