12
$\begingroup$

Is it possible to solve $\int \sec^2 x ~dx$ without knowing that $\frac{d}{dx}\tan x = \sec^2 x$?

$\endgroup$
3
  • 1
    $\begingroup$ Why would you not know that? It's not magic, it's very easy to derive. $\endgroup$
    – Javier
    Commented Sep 21, 2013 at 20:33
  • $\begingroup$ @JavierBadia Mostly out of curiosity, and also because I'm brushing up on integration by parts and integration by substitution. The textbook I was reading seemed to be asking me to integrate $\sec^2 x$ without having introduced the standard trig integrals. $\endgroup$ Commented Sep 22, 2013 at 8:32
  • $\begingroup$ I would interpret this question as asking if one can use a limit of Riemann sums directly to find the antiderivative. The accepted answer doesn't strike me as more "first principles" than knowing the derivative of $\tan$. $\endgroup$
    – 2'5 9'2
    Commented May 27, 2014 at 20:51

5 Answers 5

15
$\begingroup$

Yes, this can be done, but the method is far longer than if you know the derivative of $\tan x$ already. The point is to convert this trigonometric integral into the integral of a rational function, integrate that with partial fractions, and then convert back. We use the substitution $u = \tan(x/2)$, for which $$ \sin x = \frac{2u}{1+u^2}, \ \ \ \cos x = \frac{1-u^2}{1+u^2}, \ \ \ dx = \frac{2\,du}{1+u^2}. $$ (This is called the $\tan(x/2)$-substitution, and is discovered naturally by comparing the trigonometric parametrization of the unit circle and a rational parametrization of the unit circle. See http://en.wikipedia.org/wiki/Tangent_half-angle_substitution.)

[EDIT: Let me show how to derive those transformation formulas above, especially the one for $dx$ since there is a comment on my answer that it depends on knowing the derivative of $\tan x$ is $\sec^2 x$ already, but that's not really the case. We do need to know that $1 + \tan^2 t = \sec^2 t = 1/\cos^2 t$. Then $\cos^2(x/2) = 1/(1+\tan^2(x/2)) = 1/(1+u^2)$, so from the double-angle formulas for sine and cosine we get
$$ \cos x = 2\cos^2(x/2) - 1 = 2\sec^2(x/2) - 1 = \frac{2}{1+u^2} - 1 = \frac{1-u^2}{1+u^2} $$ and $$ \sin x = 2\sin(x/2)\cos(x/2) = 2\tan(x/2)\cos^2(x/2) = \frac{2u}{1+u^2}. $$ Then differentiate both sides of the equation for $\sin x$ using the quotient rule on the right: $$ \cos x\,dx = \frac{2(1+u^2)-(2u)(2u)}{(1+u^2)^2}\,du = \frac{2(1-u^2)}{(1+u^2)^2}\,du \Rightarrow dx = \frac{2\,du}{1+u^2}. $$ ]

From the formula for $\cos x$ in terms of $u$ we have $\sec^2 x = 1/\cos^2 x = (1+u^2)^2/(1-u^2)^2$, so $$ \int \sec^2 x \,dx = \int \frac{(1+u^2)^2}{(1-u^2)^2}\frac{2}{1+u^2}\,du = \int \frac{2(1+u^2)}{(1-u^2)^2}\,du. $$ The partial fraction decomposition of that last integrand is $$ \frac{2(1+u^2)}{(1-u^2)^2} = \frac{2(1+u^2)}{(1-u)^2(1+u)^2} = \frac{1}{(1-u)^2} + \frac{1}{(1+u)^2}, $$ so \begin{eqnarray*} \int \sec^2 x\,dx & = & \int\left(\frac{1}{(1-u)^2} + \frac{1}{(1+u)^2}\right)\,du \\ & = & \frac{1}{1-u} - \frac{1}{1+u} + C \\ & = & \frac{2u}{1-u^2} + C \\ & = & \frac{2\tan(x/2)}{1 - \tan^2(x/2)} + C. \end{eqnarray*} Recall now the addition formula for the tangent function: $$\tan(a+b) = \frac{\tan a + \tan b}{1-(\tan a)(\tan b)}.$$ Thus the double-angle formula is $\tan(2a) = 2(\tan a)/(1 - \tan^2 a)$. Therefore, using $a = x/2$, we get $$ \int \sec^2 x\,dx = \tan\left(2\frac{x}{2}\right) + C = \tan x + C. $$

$\endgroup$
6
  • 4
    $\begingroup$ I was thinking the same, but then I thought: we can only do this if we already know that the derivative of $\tan x$ is $\sec^2 x$ in order to do the $dx$ part of the substitution ... (the derivation in the Wikipedia page relies on knowing the derivative of $\tan x$). $\endgroup$
    – Old John
    Commented Sep 21, 2013 at 20:57
  • $\begingroup$ @OldJohn: Strictly speaking we don't need to know the derivative of tangent already. See my edit to the answer. We do need to know the derivatives of sine and cosine, which would give us the derivative of tangent if we were omniscient enough to compute that. Nevertheless, we don't need to recognize what $\sec^2 x$ is the derivative of to develop this solution. $\endgroup$
    – KCd
    Commented Sep 21, 2013 at 21:15
  • $\begingroup$ You are probably right. I just feel that the original question seems a bit like "how can we integrate a trig function without knowing derivatives of trig functions". I am probably being over-critical :) $\endgroup$
    – Old John
    Commented Sep 21, 2013 at 21:18
  • 3
    $\begingroup$ I agree that the question is not a practical one, but I am sure it was asked just out of curiosity. If this inspires someone to look into the $\tan(x/2)$-substitution and figure out why it works, then it's a good question. $\endgroup$
    – KCd
    Commented Sep 21, 2013 at 21:21
  • $\begingroup$ Agreed. I wish I had know as a student that all the trig integrals I had to struggle with could all have been reduced to integrals of rational functions - ought to be better known! $\endgroup$
    – Old John
    Commented Sep 21, 2013 at 21:24
12
$\begingroup$

Use the trig identity for $\text{sec}^2x$ and write the expression in terms of $\text{sin }x$ and $\text{cos }x$: $$ \int{\text{sec}^2x\text{ }dx}=\int{(\text{tan}^2x+1)\text{ }dx}=\int{\text{tan}^2x\text{ }dx}+\int{dx}=\int{\frac{\text{sin}^2x}{\text{cos}^2x}\text{ }dx}+\int{dx} $$ Prepare for integration by parts: $$ u=\text{sin }x~~~~~~~~~~du=\text{cos }x\text{ }dx~~~~~~~~~~dv=\frac{\text{sin }x}{\text{cos}^2x}\text{ }dx~~~~~~~~~~v=\frac{1}{\text{cos }x} $$ Integrate by parts and simplify: $$ \int{\frac{\text{sin}^2x}{\text{cos}^2x}\text{ }dx}+\int{dx}=\frac{\text{sin }x}{\text{cos }x}-\int{\frac{\text{cos }x}{\text{cos }x}\text{ }dx}+\int{dx}=\text{tan }x+C $$

$\endgroup$
1
  • $\begingroup$ Nice and simple. $\endgroup$
    – JDG
    Commented Mar 2, 2022 at 18:38
5
$\begingroup$

Here is an approach. First, note that

$$ \sec(x)=\frac{2}{e^{ix}+e^{-ix}}, $$

then use the substitution $u=e^{ix}$ to evaluate the integral.

$\endgroup$
2
$\begingroup$

Show that for all $t_i \in x_i-x_{i-1}$ where the $x_i$ are partition points in $J=[a,b]$, $\forall \epsilon > 0$, there exists a $\delta$ such that while $\max{\left(x_i-x_{i-1}\right)}<\delta$ we have, $$\left|\sum_{i=0}^{\infty} \sec^2(t_i)(x_i-x_{i-1})-(\tan(b)-\tan(a))\right| < \epsilon $$

EDIT
As Kcd points out, $J$ must be chosen appropriately, so that no where in $J$ does $\sec^{2}{x}$ "blow up", otherwise the FTC would not hold. How, you might ask, would one come to the conclusion that one should ought to prove this? Well for an appropriately small mesh, $\max{(x_i-x_{i-1})}$, one might notice that $$ \sum_{i=1}^{N}\sec^{2}{t}(x_i-x_{i-1}) \approx \tan{b} - \tan{a} $$

$\endgroup$
4
  • $\begingroup$ At first I think that your answer was very "cynical", but now I think that you really are answering the question as the OP wishes: by using first principles ;-). $\endgroup$ Commented Sep 21, 2013 at 21:22
  • $\begingroup$ @MatemáticosChibchas Thank you! $\endgroup$
    – Rustyn
    Commented Sep 21, 2013 at 21:23
  • $\begingroup$ @RustynYazdanpour: What I meant was "how would you show this"? Your type of solution appears to require that you already know what the answer is supposed to be. Of course all of us proposing solutions know the answer in advance by the elementary method, but your method doesn't seem to derive the answer "as if" it weren't known already. You also should be more careful about the intervals you're using: if they contain a point where $\sec^2 x$ blows up then the fundamental theorem of calculus is invalid and your inequality at the end is incorrect. $\endgroup$
    – KCd
    Commented Sep 21, 2013 at 21:29
  • $\begingroup$ @Kcd Of course, the intervals needed to be chosen correctly--and no we do not know apriori what the answer is supposed to be, we have a guess, and if one wanted to prove this guess to be true, one could attempt to prove this statement over an appropriate $J$, and, if successful, at the very least, one would have an Riemann number $I$ over a domain where $\sec^{2}{x}$ is integrable... $\endgroup$
    – Rustyn
    Commented Sep 21, 2013 at 21:33
0
$\begingroup$

What if you write the numerator as $\sin^{2}(x) + \cos^{2}(x)$ and split the fraction? To integrate $\tan^{2}(x)$ you can add and subtract 1 to obtain a tangent as part of your anti derivative. The "subtraction" of 1 cancels against the second fraction after the split.

$\endgroup$

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .