Trigonometric Substitutions are especially useful when we want to get rid of x2−a2, x2+a2 and a2−x2 under integral sign.
Recall that trignomeric identity states cos2(x)+sin2(x)=1.
Multiplying both sides by a2 gives a2cos2(x)+a2sin2(x)=a2.
Dividing both sides of equation by cos2(x) yields: a2+a2tan2(x)=a2sec2(x).
From these equations we have that cos(x)=a1a2−(asin(x))2 and tan(x)=a1(asec(x))−a2 (since integral is indefinite, we can drop absolute value bars, but this can't be done in definite integral). Notice similarity with above roots.
In general:
If you have a2−x2, make substitution x=asin(u).
If you have x2−a2, make substitution x=asec(u).
If you have x2+a2, make substitution x=atan(u).
We already made first substitution when discussed Substitution Rule.
However, carefully examine integral: maybe you don't need none of the above substitutions: for example, for integral ∫xx2+a2dx there is a simpler substitution u=x2+a2.
Example 1. Find ∫x4x2−16dx
We see similar to above root in numerator, just need to rewrite it a bit: 4x2−16=2x2−4.
Now, we are ready to make substitution x=2sec(u), then dx=2sec(u)tan(u)du.
So, 2∫xx2−4dx=2∫2sec(u)4sec2(u)−42sec(u)tan(u)du=
=4∫∣tan(u)∣tan(u)du
Since integral is indefinite we will drop absolute value bars: 4∫tan2(u)du. We know how to do such integrals (see Trigonometric Integrals note). So, 4∫tan2(u)du=4∫(sec2(u)−1)du=4(tan(u)−u)+C.
To return to old variable we will need to find tan(u) in terms of x. From our substitution we see that sec(u)=2x.
Since tan(u)=sec2(u)−1 then tan(u)=4x2−1=21x2−4.
Since sec(u)=2x then cos(u)=x2 and u=arccos(x2) (note that we could express u in terms of x through inverse secant: u=arcsec(2x), people just more familiar with cosines therefore we use inverse cosine).
Finally, ∫x4x2−16dx=4(21x2−4−arccos(x2))+C=
=2(x2−4−2arccos(x2))+C.
Now, let's see how to handle absolute value bars in the definite integral.
Example 2. Calculate ∫24x4x2−16dx.
It is same integral as in example 1. First, simplify expression under square root: 2∫24xx2−4dx.
Let x=2sec(u) then dx=2sec(u)tan(u)du.
x is changing from 2 to 4, so sec(u) is changing from 22=1 to 24=2, that's why cos(u) is changing from 21 to 1 and u is changing from 0 to 3π. Tangent on this interval is positive, so x2−4=4sec2(u)−4=2∣tan(u)∣=2tan(u).
Note that in determining value of u we took the smallest positive value.
Now, integral becomes 4∫03πtan2(u)du=4∫03π(sec2(u)−1)du=4(tan(u)−u)∣02π=
=4((tan(0)−0)−(tan(3π)−3π))=−4(0−0−3+3π)=
=43−34π.
Example 3. Calculate ∫−4−2x4x2−16dx.
It is same integral as in example 1. First, simplify expression under square root: 2∫−4−2xx2−4dx.
Let x=2sec(u) then dx=2sec(u)tan(u)du.
x is changing from -4 to -2, so sec(u) is changing from 2−4=−2 to 2−2=−1, that's why cos(u) is changing from −21 to −1 and u is changing from 32π to π.
This means that tangent on this interval is negative, so x2−4=4sec2(u)−4=2∣tan(u)∣=−2tan(u).
Note that in determining value of u we took the smallest positive value.
Now, integral becomes −4∫32ππtan2(u)du=−4∫32ππ(sec2(u)−1)du=4(tan(u)−u)∣32ππ=
=−4((tan(π)−π)−(tan(32π)−32π))=−4(0−π+3+32π)=
=34π−43.
Example 4. Calculate ∫x24−x21dx.
We will need substitution 2 here: x=2sin(u) then dx=2cos(u)du and integral can be rewritten as:
Since integral is indefinite, we can drop absolute value bars (we assume that cosine is positive):
41∫sin2(u)1du=41∫csc2(u)du=−41cot(u)+C.
Now, we will need to return to old variables.
Since x=2sin(u) then sin(u)=2x and cos(u)=1−4x2=214−x2, so cot(u)=sin(u)cos(u)=21x214−x2=x4−x2.
Finally, ∫x24−x21dx=−4x4−x2+C.
Example 5. Calculate ∫02(x2+4)25x2dx
Since (x2+4)25=(x2+4)5 then make substitution number 3: x=2tan(u). In this case dx=2sec2(u)du.
So, integral can be rewritten as ∫02(4tan2(u)+4)54tan2(u)2sec2(u)du=∫02(2∣sec(u)∣)54tan2(u)2sec2(u)du.
Since x is changing from 0 to 2 then tan(u) is changing from 20=0 to 22=1. This means that u is changing from 0 to 4π. This means that sec(u) is positive on this interval, so ∣sec(u)∣=sec(u).
So, integral becomes 41∫04πsec3(u)tan2(u)du.
To calculate this integral it is better to convert it to sines and cosines: 41∫04πsec3(u)tan2(u)du=41∫04πsin2(u)cos(u)du.
Let t=sin(u) then dt=cos(u)du. Since u is changing from 0 to 4π then t is changing from sin(0)=0 to sin(4π)=21.
And integral becomes 41∫021t2dt=121t3∣021=121((2)31−03)=2421.
Example 6. Evaluate ∫x2+4x+13dx.
It seems that this integral can't be evaluated using above substitution. However, recall that any quadratic function ax2+bx+c can be transformed into a(x+d)2+e by completing a square.
So, x2+4x+13=x2+4x+4+9=(x+2)2+9.
This is similar to x2+a2 except x+2 term. But this doesn't matter. Let x+2=3tan(u). Then dx=3sec2(u)du.
Again expression under the square root is not like the above 3 expressions. However, note that ex−1=(e2x)2−1. And it is similar to substitution 2. So, let e2x=sec(u) then 21e2xdx=sec(u)tan(u)du. We need to express dx in terms of u only.
Since e2x=sec(u) then 21sec(u)dx=sec(u)tan(u)du or dx=2tan(u)du.
Integral now becomes ∫sec2(u)−1⋅2tan(u)du=2∫∣tan(u)∣tan(u)du.
Since integral is indefinite, we drop absolute value bars: 2∫tan2(u)du=2∫(sec2(u)−1)du=2(tan(u)−u)+C.
Since sec(u)=e2x then tan(u)=(e2x)2−1=ex−1.
And u is either arcsec(e2x) or arctan(ex−1) (we choose second option).
So, ∫ex−1dx=2(ex−1−arctan(ex−1))+C.
Note, that this integral can be solved another way: with double substitution; first substitution is u=ex and second is t=u−1.
We have seen (last two examples) that some integrals can be converted into integrals that can be solved using trigonometric substitution described above.