5

How do I prove that indefinite integral of $\sec x$ is equal to $\ln(\sec x + \tan x) + C$?

I tried to substitute $t = \cos x$ but that didn't help. I have no idea how to integrate it any other way, and my textbook doesn't offer a derivation.

Travis Willse
  • 99,363
Sk Dayn
  • 61
  • Note that this is an historically famous integral, and for that reason you can find PLENTY of information about it on the net. – David H Oct 15 '14 at 06:57
  • This particular indefinite integral is of significant historical interest too---by construction an antiderivative of $\sec x$ is the function that maps a latitude to its corresponding vertical coordinate in the Mercator projection. Thanks to its conformality and another basic feature, this projection has the property that constant compass headings correspond to straight lines on the projection, which makes it especially useful for navigation. – Travis Willse Oct 15 '14 at 06:58
  • This may be a cheap way out, but: As you already know the supposed answer, simply calculate the derivative of that and prove that you get $\sec x$ as the answer. – PhoemueX Oct 15 '14 at 07:41

4 Answers4

6

$$\sec x=\frac1{\cos x}=\frac{\cos x}{\cos^2x}=\frac{\cos x}{1-\sin^2x}$$ Substitute $u=\sin x$ and use the method of partial fractions.

Since $$\frac1{1-u^2}=\frac12\left(\frac1{1+u}+\frac1{1-u}\right)$$ we get $$\int\sec xdx=\int\frac{du}{1-u^2}=\frac12(\ln(1+u)-\ln(1-u))=\ln\sqrt{\frac{1+u}{1-u}}.$$ Now $$\frac{1+u}{1-u}=\frac{1+\sin x}{1-\sin x}=\frac{(1+\sin x)^2}{1-\sin^2x}=\frac{(1+\sin x)^2}{\cos^2x}=\left(\frac{1+\sin x}{\cos x}\right)^2=(\sec x+\tan x)^2$$ so $$\int\sec xdx=\ln\sqrt{\frac{1+u}{1-u}}=\ln\sqrt{(\sec x+\tan x)^2}=\ln|\sec x+\tan x|.$$

bof
  • 78,265
5

As Travis noted, differentiation is the best way to do this.

But strangely enough, this very question was an important open question in the mid-17th century (before the invention of Calculus). See e.g. this web page of mine

Robert Israel
  • 448,999
4

It's typically very easy to verify whether a function $F(x)$ is an antiderivative of a function $f(x)$; by definition, you need only check that $$F'(x) = f(x).$$

In this case, on, say, $\left(-\frac{\pi}{2}, \frac{\pi}{2}\right)$, \begin{multline}\frac{d}{dx} \log(\sec x + \tan x) = \frac{\frac{d}{dx}(\sec x + \tan x)}{\sec x + \tan x} \\= \frac{\sec x \tan x + \sec^2 x}{\sec x + \tan x} = \frac{(\sec x + \tan x) \sec x}{\sec x + \tan x} = \sec x,\end{multline} and so (by definition) $$\int \sec x \, dx = \log |\sec x + \tan x| + C.$$ Here the absolute value signs simply account for the possibility that you're integrating over an interval on which $\sec x + \tan x < 0$, and you can justify this formula using the identities $\sec(x + \pi) = -\sec x$ and $\tan (x + \pi) = \tan x$.

Travis Willse
  • 99,363
3

Hint: Write $$\sec x = \sec x \left(\frac{\sec x + \tan x}{\sec x + \tan x}\right)$$ and remember what the derivatives of $\sec x$ and $\tan x$ are.

hjhjhj57
  • 4,125
  • 1
    I hate this so-called derivation. The only motivation for multiplying and dividing by $\sec x+\tan x$ is that you know how it's going to work out, in which case you might as well just start with $\ln(\sec x+\tan x)$ and differentiate. – bof Oct 15 '14 at 07:04
  • Well, it's just a matter of opinion. Each non-trivial integral requires some trick which will seem obvious once we know the answer. I like your method by the way +1. – hjhjhj57 Oct 15 '14 at 07:13