How the substitutions x = a sinh t, x = a cosh t and t = tan(x/2) clear square roots and trigonometric fractions from integrals, and on which intervals each one is valid.
A square root like x2+1 or a fraction like 2+cosx1 looks stubborn until you find the right change of variable. Trigonometric substitution already handles many roots. Hyperbolic functions often do the same job with less fuss, and one clever substitution turns every rational expression in sinx and cosx into an ordinary rational function. Each trick comes with an interval where it is allowed, and ignoring that interval gives wrong answers.
Key terms
Hyperbolic substitutions
Definition Let a>0. The hyperbolic functions satisfy cosh2t−sinh2t=1, with cosht>0 for all real t and sinht≥0 for t≥0.
For an integrand containing a2+x2 on any interval, substitute x=asinht with t∈R. Then dx=acoshtdt and
a2+x2=a2(1+sinh2t)=acosht.
For an integrand containing x2−a2 on an interval where x≥a, substitute x=acosht with t≥0. Then dx=asinhtdt and x2−a2=asinht. On an interval where x≤−a, first put x=−u.
In plain words The identity cosh2t−sinh2t=1 is built to cancel the expressions under these square roots, so the root disappears and you integrate something simpler.
Example To find the integral
∫(1+x2)3/2dx,
put x=sinht. Then (1+x2)3/2=cosh3t and dx=coshtdt, so
∫cosh3tcoshtdt=∫sech2tdt=tanht+C=1+x2x+C,
using tanht=sinht/cosht and sinht=x. Differentiating the result gives back the integrand, which confirms it.
Inverse hyperbolic forms
Definition The inverse hyperbolic functions have the logarithmic forms
∫x2+a2dx=arsinhax+C on R,∫x2−a2dx=arcoshax+C on (a,∞),∫a2−x2dx=a1artanhax+C on (−a,a).
In plain words These are the answers that hyperbolic substitution produces, written either with inverse hyperbolic names or as plain logarithms. Both versions are correct, and they differ only by a constant when you expand the logarithm.
where the denominators equal 1. Dividing top and bottom by cos22x, which is not zero for −π<x<π, gives the expressions in t. The formula for dx comes from differentiating x=2arctant.
In plain words Sine, cosine and dx all become fractions in t with no trigonometry left, so the integral turns into an ordinary rational integral.
Example Over 0≤x≤π/2 the new limits are t=tan0=0 and t=tan4π=1. The denominator becomes
Definition A rational function of sinx and cosx is an expression R(sinx,cosx) where R(u,v) is a quotient of two polynomials in u and v. On any subinterval of (−π,π) where R(sinx,cosx) is defined, that is, where its denominator is not zero, the substitution t=tan2x gives
∫R(sinx,cosx)dx=∫R(1+t22t,1+t21−t2)1+t22dt,
and the new integrand is a rational function of t. So every such integral can be done with partial fractions.
In plain words No matter how messy the mix of sines and cosines, as long as it is built with only arithmetic, this substitution always leads to an integral you know how to finish.
Example In the integral of 1/(1+sinx+cosx) over [0,π/2] the denominator becomes
1+1+t22t+1+t21−t2=1+t22+2t.
So the integrand times dx is
2(1+t)1+t2⋅1+t22dt=1+tdt,
and
∫0π/21+sinx+cosxdx=∫011+tdt=ln2.
The substitution is a fallback, not a first choice. To integrate sin2xcosx the plain substitution u=sinx is much quicker.
Substitution intervals
Definition Substitution in a definite integral rests on this rule. If g has a continuous derivative on [α,β] and f is continuous on the range g([α,β]), then
∫αβf(g(t))g′(t)dt=∫g(α)g(β)f(x)dx.
When you substitute x=g(t) and want to return to x at the end, g must also be one to one on the interval, so that t can be written as a function of x. Every simplification of a root such as sinh2t+1=cosht or cosh2t−1=sinht must hold for the t values used. For t=tan2x this means working inside one interval (−π+2kπ,π+2kπ) with k an integer.
In plain words A substitution is only trustworthy on an interval where it behaves well. If the integration interval crosses a point where the substitution breaks, split the integral there.
Example The integrand 2+cosx1 is continuous and positive on [0,2π], and
∫02π2+cosxdx=32π≈3.6276.
The substitution suggests the antiderivative
F(x)=32arctan(3tan(x/2)).
Its derivative is 1/(2+cosx) wherever it is defined, but it is undefined at x=π and jumps there. Naively, F(2π)−F(0)=0−0=0, which is impossible for a positive integrand. Splitting at π fixes it. Since the integrand is continuous, ∫0π is the limit of F(b)−F(0) as b→π−, and ∫π2π is the limit of F(2π)−F(b) as b→π+. As x rises from 0 toward π, the values of F rise from 0 toward 32⋅2π. Just after π they start near −32⋅2π and rise to 0 at 2π. Each half contributes 3π, for a total of 32π.
Check yourself
Find ∫x2−9dx for x>3.
Show the answer
With x=3cosht and t>0 you get ∫dt=t+C, so the answer is arcosh3x+C, or equivalently ln(x+x2−9)+C. The two forms differ by the constant ln3.
Use t=tan2x to evaluate ∫0π/21+cosxdx.
Show the answer
The substitution gives 1+cosx=1+t22, so the integrand times dx is 21+t2⋅1+t22dt=dt. The limits are 0 and 1, so the integral is 1.
Why can you not apply t=tan2x directly to ∫03π/22+sinxdx in one piece?
Show the answer
The interval contains x=π, where tan2x is undefined. Split the integral at π. On each piece use the antiderivative from the substitution and take its one-sided limit at π, just as in the 2+cosx1 example above.
May we count page visits to understand which courses and learning features are useful? Cloudflare collects visit and performance statistics, without your account details or answers. Your choice does not affect access. Details