56

So far, I know and can use a reasonable number of 'tricks' or techniques when I solve integrals. Below are the tricks/techniques that I know for indefinite and definite integrals separately.


Indefinite integrals

  • Standard integrals, such as those of polynomial, trigonometric, logarithmic and exponential functions, including usage of trig identies.
  • Basic substitution.
  • Weierstrass and Euler substitutions.
  • Integration by parts.
  • $$\int\frac{1}{x+x^n}dx=\int\frac{x^{-n}}{1+x^{1-n}}dx=\frac{1}{1-n}\ln\lvert 1+x^{1-n}\rvert+C$$
  • $$\int\frac{1}{x^{\frac{a+b}{a+b}}\cdot x^{\frac{a}{a+b}}+x^{\frac{b}{a+b}}}dx=\int \frac{x^{-\frac{b}{a+b}}}{\left(x^{\frac{a}{a+b}}\right)^2+1}dx=\arctan x^{\frac{a}{a+b}}+C$$
  • Substitution $u=\frac{1-x}{1+x}$ for integrals involving $\ln$ and/or the bounds $0$ and $1$.
  • Reduction formulae.
  • $$\int e^x(f(x)+f'(x))dx=e^xf(x)+C$$
  • Writing $\sin$'s and $\cos$'s as complex exponentials.
  • $$\int\frac{a\sin x+b\cos x}{c\sin x+d\cos x}dx=Ax+B\ln\lvert c\sin x+d\cos x\rvert+C$$ where $$A=\frac{ac+bd}{c^2+d^2}~~~B=\frac{bc-ad}{c^2+d^2}$$ which can be found using simultaneous equations.

Definite integrals

  • Differentiation under the integral sign ('Feynman's technique')
  • $$\int_a^b f(x)dx=\int_a^bf(a+b-x)dx$$
  • Usage of power series to evaluate integrals such as $\int_0^1\frac{\ln(1-x)}{x}dx$ and the like.
  • Making use of even or odd function properties.
  • (My newest personal favourite) For even functions $f(x)$ and $g(x)$, and an odd function $h(x)$: $$\int_{-a}^a\frac{f(x)}{1\pm g(x)^{h(x)}}dx=\int_{0}^a f(x)~dx$$ which allows us to evaluate wonderful things like $$\int_{-\infty}^{\infty}\frac{e^{-x^2}}{1+\pi^{\sin x} }dx=\frac{\sqrt{\pi}}{2}$$

Question:

Do you know any other integration techniques or tricks that I can use whose usage don't rely on anything beyond high school calculus* or perhaps the first year of a Mathematics degree course?

I know that a similar question has been asked here and here but I've looked through them and nothing beyond what I have written above was mentioned, apart from some techniques I couldn't understand such as residue calculus and contour integrals.

Many thanks for your help.


*Roughly what I mean by high school level calculus:

INCLUDED

  • Integration of polynomials and the basic trigonometric functions, such as $\sin x$, $\cos x$, $\tan x$, $\sec x$, $\operatorname{cosec} x$, $\cot x$, $\sec^2 x$, $\sec x\tan x$, $\operatorname{cosec} x\cot x$, $\operatorname{cosec}^2 x.$
  • Integration of all $x^n$ including $n=1$. Integration of exponentials.
  • Integration by parts.
  • Integration using substitution, such as using trigonometric/hyperbolic substitutions, and Weierstrass and Euler substitutions (this also includes integration by 'inspection' which is really just substitution but when the individual doesn't need to substitute anything).
  • Integration using partial fractions and logarithms, such as $\int\frac{f'(x)}{f(x)}dx$.
  • Reduction formulae. Ability to understand and use the concepts of even and odd functions in integration. Improper integrals.
  • Integrating which results in elementary functions.

NOT INCLUDED

  • Fourier, Laplace and Mellin transforms.
  • Indefinite integrals that include non-elementary functions in the solution.
  • Contour integration.
  • Residue calculus and similar methods.
  • Comments are not for extended discussion; this conversation has been moved to chat. – Pedro Apr 26 '21 at 15:52
  • If you're looking for more examples than those in the two posts you linked to (which are arguably duplicates of this post) why not issue a bounty there instead of creating yet another post of "integration techniques"? The post is already too broad and not exactly suitable for the site, and the fact there are already two very active and rich posts with the same purpose does not help the case of leaving this open. – Pedro Apr 26 '21 at 15:54
  • 1
    Note that the answer you got here is the most upvoted answer in one of the posts you already mentioned. – Pedro Apr 26 '21 at 16:05
  • @PedroTamaroff I'm sorry, I didn't stress it enough: I'm only looking for integration techniques that are high school level or perhaps the level of the first year of a mathematics degree course. The other questions contain methods which are far too advanced for me currently, and placing a bounty on the other questions I think would only attract more advanced techniques which I wouldn't understand. That's why I asked this question: to attract answers of a specific level. It is true that the answer I have received is in the other post, ...(continued) – A-Level Student Apr 26 '21 at 17:46
  • ...@PedroTamaroff but I don't think that that warrants closure of my question or multiple downvotes! Please note, I am not angry at you at all, I do understand what you are saying. I am slightly irritated by the downvoters though: they have not given their reasons as you have. – A-Level Student Apr 26 '21 at 17:46
  • 1
    interesting examples of differentiating under integral sign. It helps! https://arxiv.org/ftp/arxiv/papers/1901/1901.01249.pdf – Svyatoslav May 02 '21 at 08:17
  • Hello, where can I find proof of your "My newest personal favorite"? – Zaragosa Dec 23 '23 at 06:46
  • @Zaragosa Most of it can be found on blackpenredpen's channel here: https://www.youtube.com/watch?v=gK8Wf9ZkZW4&list=PLj7p5OoL6vGxXRSqPkn9dV5PuXIQh7P1O – A-Level Student Apr 14 '24 at 15:25
  • @FelixMarin what about those that don't use Instagram? – Integreek Nov 21 '24 at 09:50
  • @Zaragosa use the property $$\int_a^bf(x)\mathrm dx=\int_a^bf(a+b-x)\mathrm dx$$ – Integreek Nov 21 '24 at 09:52

9 Answers9

33

As a high school student, most of the tricks I'm aware of were already stated by you, or in the comments. However, there's one more trick that I don't think anyone has mentioned: Integrating an inverse function.

$$\int\!f^{-1}(x)\ dx = x\cdot\!f^{-1}(x)\ - F(f^{-1}(x))\ + c$$where $$F(x) = \int\!f(x)\ dx$$

So for instance, if you wish to find $\int\cos^{-1}(x)\ dx,$ you will have $$f(x)= \cos x$$ and $$F(x) = \int\cos x\ dx = \sin x\ (+c)$$

So to find $\int\cos^{-1}(x)\ dx,$ use the formula as the follows:

$$\int\cos^{-1}(x)\ dx = x\cdot\cos^{-1}(x)\ - \sin(\cos^{-1}(x))\ + c$$ $$= x\cdot\cos^{-1}(x)\ - \sqrt {1-x^2}\ + C$$

I personally like this trick as it can be generalized to any inverse function. A simple way to prove it would be using the Chain Rule but it's a really nice formula that avoids working things out from scratch every time.

Hope that helped to add to your list :)

devam_04
  • 828
25

Since you listed Feynman's trick as one of the methods you know, I'll assume you're at least a bit familiar with multivariable integrals. If you allow this, then some techniques you can use are the following

  • One technique is to work with double integrals to evaluate a single integral. A great example is found here, where this technique is used to evaluate $\int_0^{\infty} \frac{\sin(x)}{x} \, \mathrm{d}x$. In the linked answer, the OP shows that you can start with the equation: $$ \int_{0}^{\infty} \left(\int_{0}^{\infty} e^{-xy} \sin x \, \mathrm{d}y\right)\, \mathrm{d}x = \int_{0}^{\infty} \left(\int_{0}^{\infty} e^{-xy} \sin x \, \mathrm{d}x\right)\, \mathrm{d}y $$ and afterward, integrating the L.H.S. first with respect to $y$ and then $x$, but on the R.H.S integrating first with respect to $x$ and then $y$ you get $$ \int_{0}^{\infty} \frac{\sin x}{x} \, \mathrm{d}x= \int_{0}^{\infty}\frac{1}{1+y^2}\, \mathrm{d}y = \lim_{x\to \infty}\arctan(x) - 0 = \frac{\pi}{2} $$
  • Also in the multivariable tricks is to use a change of coordinate system to evaluate an integral. This is a standard way of evaluating $\int_{-\infty}^{\infty} e^{-x^2} \, \mathrm{d}x$ by doing $$ \left(\int_{-\infty}^{\infty} e^{-x^2} \, \mathrm{d}x\right)^2 = \left(\int_{-\infty}^{\infty} e^{-x^2} \, \mathrm{d}x\right)\left(\int_{-\infty}^{\infty} e^{-y^2} \, \mathrm{d}y\right) = \int_{-\infty}^{\infty} \int_{-\infty}^{\infty} e^{-\left(x^2 + y^2 \right)} \, \mathrm{d}y \, \mathrm{d}x $$ and here noticing that we're integrating over all the cartesian plane (since we're going from $- \infty$ to $\infty$ in both $x$ and $y$ directions) we can transform to polar coordinates remembering that $x ^2 + y^2 = r^2$ and that the area differential in polar coordinates is $\mathrm{d}A = \mathrm{d}y\, \mathrm{d}x = r\, \mathrm{d}r \, \mathrm{d}\theta$. We get $$ \int_{-\infty}^{\infty} \int_{-\infty}^{\infty} e^{-\left(x^2 + y^2 \right)} \, \mathrm{d}y \, \mathrm{d}x = \int_0^{2 \pi} \int_{0}^{\infty} e^{-r^2} r\, \mathrm{d}r \, \mathrm{d}\theta = \int_0^{2 \pi} \frac{1}{2} \, \mathrm{d}\theta = \pi $$ and hence we conclude that $$ \int_{-\infty}^{\infty} e^{-x^2} \, \mathrm{d}x = \sqrt{\pi} $$

Another approach you can take is to try and convert an integration question into equations where the variable is the integral itself. One example of this is the trick used to evaluate $I =\int e^x \sin(x)\, \mathrm{d}x$. Applying integration by parts twice we get that \begin{align*} &\underbrace{\int e^x \sin(x) \, \mathrm{d}x}_{\color{blue}{I}} = \sin(x) e^x- \cos(x)e^x - \underbrace{\int e^x \sin(x)\, \mathrm{d}x}_{\color{blue}{I}}\\ \implies& \int e^x \sin(x)\, \mathrm{d}x = I = \frac{e^x (\sin(x) - \cos(x))}{2} + C \end{align*} where we see that we transform our integral question into solving a linear equation with the one unknown being $I$.


Moreover, if you are a bit familiar with differential equations, you may find that sometimes you can go from an integral question to a differential equations question.

One example of this is the evaluation of $\int_{-\infty}^{\infty} e^{-x^2}\cos(\sqrt{2}x) \mathrm{d}x$. We thus define the function $I(\xi) = \int_{-\infty}^{\infty} e^{- x^2}\cos(2 \xi x)\, \mathrm{d}x $. Using Feynman's trick we see that \begin{align} I'(\xi) &=\int_{-\infty}^{\infty} \left[-2xe^{- x^2}\right]\sin(2 \xi x)\, \mathrm{d}x\overset{\text{I.B.P.}}{=} -2\xi \underbrace{\int_{-\infty}^{\infty} e^{- x^2}\cos(2 \xi x)\, \mathrm{d}x}_{\color{blue}{I(\xi)}} \end{align} So we arrive at $I' = -2\xi I$, which is a differential equation that we can attempt to solve for the function $I(\xi)$. Rewriting the differential equation (dividing both sides by $I$) we get $\frac{1}{I}\frac{\mathrm{d} I}{\mathrm{d}\xi} = -2\xi$ from which we can integrate both sides and get $$ \frac{1}{I}\frac{\mathrm{d} I}{\mathrm{d}\xi} = -2\xi \mathbin{\color{purple}{\implies}} \int_0^{\xi}\frac{1}{I}\frac{\mathrm{d} I}{\mathrm{d}\widetilde{\xi}} \mathrm{d} \widetilde{\xi} = \int_0^{\xi}-2 \widetilde{\xi}\, \mathrm{d}\widetilde{\xi} \mathbin{\color{purple}{\implies}} \ln\Bigg|\frac{I(\xi)}{I(0)}\Bigg| = -\xi^2 \mathbin{\color{purple}{\implies}} I(\xi) = I(0)e^{-\xi^2} $$ But since from our original defintion we know $I(0) =\int_{-\infty}^{\infty} e^{- x^2}\, \mathrm{d}x = \sqrt{\pi}$ (noticing this last integral is the same one evaluated in the "change of coordinates" section!) we can conclude $$ \int_{-\infty}^{\infty} e^{- x^2}\cos(2 \xi x)\, \mathrm{d}x = \sqrt{\pi} e^{-\xi^2} $$ and substituting $\xi = \frac{1}{\sqrt{2}}$ we get $$ \int_{-\infty}^{\infty} e^{-x^2}\cos(\sqrt{2}x) \mathrm{d}x= \sqrt{\frac{\pi}{e}} $$


Not necessarily techniques, but there are several integral formulas that may be helpful in simplifying an integral. Note the following formulas are valid whenever the integral experessions makes sense (i.e. they converge and are smooth enough):

  1. $$ \int_{0}^{\infty} \frac{\ln(x)}{ax^2 + bx +c}\, \mathrm{d}x \overset{\color{blue}{x\to 1/x}}{=} -\int_{0}^{\infty} \frac{\ln(x)}{cx^2 + bx +a} \, \mathrm{d}x\color{blue}{\implies} \int_{0}^{\infty} \frac{\ln(x)}{ax^2 + bx +a}\, \mathrm{d}x=0 $$
  2. As shown in this answer $$ \int_{0}^{\pi} x f \left( \sin(x)\right)\, \mathrm{d}x = \frac{\pi}{2}\int_{0}^{\pi} f \left( \sin(x)\right) \, \mathrm{d}x $$ Similarly, for a symmetric function with two inputs (i.e. $f(x,y) = f(y,x)$): $$ \int_{0}^{\frac{\pi}{2}} xf\left(\sin(x), \cos(x)\right)\, \mathrm{d}x = \frac{\pi}{4} \int_{0}^{\frac{\pi}{2}}f\left(\sin(x), \cos(x)\right)\, \mathrm{d}x $$
  3. A special case of Glasser's master theorem is $$ \int_{-\infty}^{\infty} f(x) \, \mathrm{d}x = \int_{-\infty}^{\infty} f\left(x- \frac{1}{x}\right) \, \mathrm{d}x $$ Sometimes the general statement $ \int_{-\infty}^{\infty} f(x) \, \mathrm{d}x = \int_{-\infty}^{\infty} f\left(x- \frac{a}{x}\right) \, \mathrm{d}x $ for $a >0$ is also useful. A proof is shown here.
  4. We also have the definite integral result for the integral of inverse functions. For $y = f(x)$, $f(a) = c$ and $f(b) = d$ a nice result is the identity $$ \int_{a}^{b} f(x) \, \mathrm{d}x + \int_c^d f^{-1}(y) \, \mathrm{d}y = bd - ac $$
  5. We have the Frullani integral which gives: $$ \int _{0}^{\infty }{\frac {f(ax)-f(bx)}{x}} \, \mathrm{d}x =\left(f(\infty)-f(0)\right)\ln\left(\frac {a}{b}\right) $$ where $f(\infty) = \lim_{x \to \infty} f(x)$. There's also a periodic version of the Frullani integral which says that if $f(x)$ is periodic with period $p$ (i.e. $f(x+p) = f(x)$) and $\int_0^{A} \frac{f(x)}{x}\, \mathrm{d}x <\infty$ for all $A >0$ then $$ \int _{0}^{\infty }{\frac {f(ax)-f(bx)}{x}} \, \mathrm{d}x = \frac{\ln\left(\frac{a}{b}\right)}{p} \int_{u}^{u+p} f(x)\, \mathrm{d}x, \qquad \forall u \in \mathbb{R} $$
  6. For $f$ with a bounded antiderivative on $[0, \infty)$, then $$ \int_{0}^{\infty} f(x) \, \mathrm{d}x \overset{\color{blue}{x\to 1/x}}{=} \frac{1}{2} \int_{0}^{\infty}f(x) + \frac{f\left(\frac{1}{x}\right)}{x^2}\, \mathrm{d}x $$
  7. For a function $f$ that's $\pi$-periodic (i.e. $f(\pi\pm x) = f(x)$) Lobachevsky's integral formula gives \begin{align} \int_{0}^{\infty} \frac{\sin^2(x)}{x^2}f(x) \, \mathrm{d}x = \int_{0}^{\infty} \frac{\sin(x)}{x}f(x) \, \mathrm{d}x =\int_{0}^\frac\pi2 f(x) \, \mathrm{d}x\\ \int_{0}^{\infty} \frac{\sin^4(x)}{x^4}f(x) \, \mathrm{d}x =\int_{0}^\frac\pi2 f(x) \, \mathrm{d}x - \frac23 \int_{0}^\frac\pi2 \sin^2(x)f(x) \, \mathrm{d}x \end{align}
Robert Lee
  • 7,654
  • WOAH! Your answer is AMAZING, thanks so much! In the differential equation section, how did you get to that solution for $f(t)$ though? Also, how does $$\left(\int_{-\infty}^{\infty} e^{-x^2} \ dx\right)\left(\int_{-\infty}^{\infty} e^{-y^2} \ dy\right) = \int_{-\infty}^{\infty} \int_{-\infty}^{\infty} e^{-\left(x^2 + y^2 \right)} \ dy \ dx$$? I'm not that familiar with multivariable calculus yet. – A-Level Student Apr 29 '21 at 16:57
  • 1
    To solve the differential equation I used a Laplace transform, but since you mentioned you weren't familiar with those types of transforms I omitted those details. That said, I tried to pose the solution as an ansatz that you could “guess and check” to make it easier to follow. – Robert Lee Apr 30 '21 at 08:07
  • 1
    As for the second part, the simple answer is that each one of the integrals is constant with respect to the other since $x$ doesn't depend on $y$, so you can "factor it out" just as you do with any constant.

    If you want a more formal answer, you can find some more detailed explanations as to why this is valid in the answers to this question.

    – Robert Lee Apr 30 '21 at 08:13
  • 1
    Thanks so much. By the way, it's really nice to see $e$ turn up in an integral's solution when it isn't explicitly in the integrand or the integral's limits, it's quite rare :) – A-Level Student May 20 '21 at 22:10
  • Perhaps you'd like to answer this question? https://math.stackexchange.com/questions/4192890/curious-and-short-question-about-int-infty-0-frac-sin2xx2x21dx You'd almost certainly get the accepted answer. – A-Level Student Jul 07 '21 at 21:51
17

You can add binomial integrals (Chebyschev integrals) which are those of the form $\int x^m(a+bx^n)^{\frac pq}dx$ where $a,b$ are real, $p,q$ integer and $m,n$ rational. Chebyschev proved that these integrals are elementary functions only when at least one of $\dfrac pq,\dfrac{m+1}{n}$ or $\dfrac{m+1}{n}+\dfrac pq$ are integers. For example $\int x^4(1+x^4)^{\frac 12}dx$ is not calculable by elementary methodes.

More precisely for the three above cases of elementary solubility we have:

$\dfrac pq$ is an integer: Apply Newton's binomial.

$\dfrac{m+1}{n}$ is an integer: change variable $u=(a+bx^n)^{\frac 1q}$.

$\dfrac{m+1}{n}+\dfrac pq$ is an integer: change variable $u=\left(\dfrac{a+bx^n}{x^n}\right)^{\frac 1q}$

EXAMPLE. If $\int\frac{1}{x^4\sqrt{1+x^2}}dx$ then putting $u=\sqrt{1+\frac{1}{x^2}}$ leads to the integral $\int(1-u^2)du$

Ataulfo
  • 32,657
  • Thank you very much!! I'd never seen this before and I find it fascinating. – A-Level Student Apr 26 '21 at 21:56
  • 1
    You are welcome. Let me, please, to show the following example: $\int\dfrac{dx}{x^2\sqrt[4]{(4-x^4)^3}}$. Here $m=-2, n=4$ and $p=-\dfrac 34$ so the change $u=\left(\dfrac{4-x^4}{x^4}\right)^{\frac14}$ which leads to the easy integral $-\frac14\int du$. But could be useful for you to note that for example $\int\dfrac{dx}{x^2\sqrt[4]{(4-x^4)^2}}$ is not elementary anymore. And this is so for many changes of the integral keeping its condition of binomial. – Ataulfo Apr 27 '21 at 14:37
10

This one's pretty obvious, but extremely underrated and unused: if you think you can manipulate an integral to fit the form of the quotient rule of differentiation, do it! I can't tell you how many integrals I've evaluated with this technique that, at first glance, are seemingly impossible to express in terms of elementary functions. One example is

$$\int\frac{1}{\ln(x)}-\frac{1}{\ln^2(x)}\text{ }dx$$

An elementary expression for this integral seems hopelessly out of reach: it is well known that $\int\frac{1}{\ln(x)}dx$ and $\int\frac{1}{\ln^2(x)}dx$ are non-elementary integrals, and the one above is a linear combination of the two. What do we do? Well, some algebraic manipulation never hurts, even if it seems to make the integrand more messy, so let's try that. In particular, let's combine the terms by making them have the same denominator; the easiest way to do this is to multiply and divide $\frac{1}{\ln(x)}$ by $\ln(x)$.

\begin{align*} \int\frac{1}{\ln(x)}-\frac{1}{\ln^2(x)}\text{ }dx &= \int\frac{\ln(x)}{\ln^2(x)}-\frac{1}{\ln^2(x)}\text{ }dx\\ &= \int\frac{\ln(x)-1}{\ln^2(x)}dx \end{align*}

Now for the key manipulation: replace $1$ with $\frac{x}{x}$. This is perfectly acceptable because the original integrand was undefined at $0$ ($\ln x$ is undefined for $x=0$).

\begin{align*} \int\frac{\ln(x)-1}{\ln^2(x)}dx &= \int\frac{1\cdot\ln(x)-\frac{x}{x}}{\ln^2(x)}dx\\ &= \int\frac{1\cdot\ln(x)-\frac{1}{x}\cdot x}{\ln^2(x)}dx \end{align*}

You probably see it by now. If we let $f(x)=x$ and $g(x)=\ln(x)$, then the quotient rule gives

$$\left(\frac{f}{g}\right)'(x)=\frac{f'(x)g(x)-g'(x)f(x)}{[g(x)]^2}=\frac{1\cdot\ln(x)-x\cdot\frac{1}{x}}{\ln^2(x)}$$

Thus, our integrand is simply the derivative of $\frac{x}{\ln(x)}$. This immediately gives

$$\int\frac{1}{\ln(x)}-\frac{1}{\ln^2(x)}\text{ }dx=\frac{x}{\ln(x)}+C$$

Alann Rosas
  • 6,553
  • Thank you, I'll remember your advice :) Nice example! – A-Level Student May 01 '21 at 21:36
  • 1
    Your answer has already been useful: see here: https://math.stackexchange.com/questions/4123919/got-stuck-trying-to-do-the-integration-of-int-frac-ex-cos-x-sin-x/4124202#4124202 :) – A-Level Student May 02 '21 at 12:17
  • 4
    This is effectively the equivalent of the standard integration by parts formula, but using the quotient rule instead of the product rule for the derivation. One can prove that in any case where this works, one can equivalently use the standard product rule based version of the integration by parts formula. Similar results hold for triple products and so on. (I did a proof of this on my own time back in High School after figuring out that one could produce an alternative formula to the usual integration by parts by using the quotient rule instead of the product rule). – Justin Benfield Feb 24 '22 at 03:02
  • An alternative way is to substitute $t=\ln x$ and use $\int e^x\left(f(x)+f'(x)\right)\mathrm dx=e^xf(x)+C$. – Integreek Nov 22 '24 at 06:58
8

Here are a couple more:

  1. A common trick also is to let the integral in question be $I$, and add it to itself to simplify things. For example, try computing:

$$I=\int_{0}^{2014}{\frac{\sqrt{2014-x}}{\sqrt{x}+\sqrt{2014-x}}dx}$$

Doing a substitution $u = 2014-x$ and adding the resulting integral to original gives $2I = \int_0^{2014} 1 \,dx$.

  1. Another trick is geometry. The most common one students encounter is a circle, but here's a different one (from the MIT Integration Bee 2019):

$$\lim_{n \to \infty}\int_{-\infty}^{\infty} e^{-x^{2n}} \,dx.$$If you draw the picture, you just get a rectangle of height $1$ and length $2$, so the answer is $2$.

  1. Riemann sums. Try computing (without residues)

$$\int_0^{\pi} \ln(1-2a \cos x + a^2) \,dx$$

If you split the interval $[0,\pi]$ into $n$ equally sized intervals, you get a nice factorization that evaluates to $\frac{\pi}{n} \ln \frac{a^{2n}-1}{a^2-1}$, so all you need to do is take the limit.

Alvin Jin
  • 3,129
  • Thank you for your answer! Thanks in particular for the second one, I'd actually tried that question from the Bee and hadn't thought of your approach. – A-Level Student May 01 '21 at 21:34
5

Here is one: if $f(a)=f^{-1}(a)$ and $f(b)=f^{-1}(b)$ then $$\int_a^bf(x)+f^{-1}(x)dx=b^2-a^2$$This is how you should solve $$\int_1^2\left(e^{1-\frac{1}{(1-x)^2}}+1\right)+\left(1+\frac{1}{\sqrt{1-\ln(x-1)}}\right)dx$$Which was the second integral in the Berkeley 2020 integration bee. Using the above trick, the grand integrator solved this in a few seconds.

Kamal Saleh
  • 6,973
5

Here is another trick and is almost fading into oblivion. It is called the Ostrogradsky method. It states that, if you have an integral of the form $$\int\frac{P(x)}{Q(x)}dx$$ Where $Q(x)$ has more than one root, then the integral is equivalent to $$\frac{P_1(x)}{Q_1(x)}+\int\frac{P_2(x)}{Q_2(x)}dx$$Where $Q_1(x)$ is $\gcd(Q(x),Q'(x))$ and $$Q_2(x)=Q(x)/Q_1(x)$$And the polynomials $P_1$ and $P_2$, which have one less degree than $Q_1$ and $Q_2$ respectively, have undetermined coefficients which you could find by differentiating both sides. Its strength is equal to that of the Weierestrass Substitution in my opinion.

Kamal Saleh
  • 6,973
3

As a high school student myself, I would like to share some techniques:

$1.$ For integrals of the type

$$\int\frac{ae^x+be^{-x}}{pe^x+qe^{-x}}\mathrm dx$$

determine suitable constants $l$ and $m$ such that

$$pe^x+qe^{-x}=l(ae^x+be^{-x})+m(ae^x-be^{-x})$$

$2.$ (I find this particularly time-saving)For integrals involving $\ln x$ like

$$\int x\ln(\ln x)+\frac{x}{\ln^2x}\mathrm dx$$ $$\int\frac1{\ln x}-\frac1{\ln^2x}\mathrm dx$$

Substitute $\ln x=t\implies \mathrm dx=e^t\mathrm dt$, then apply the IBP formula that you stated to get the answer in a single step:

$$\int e^x\left(f(x)+f’(x)\right)\mathrm dx=e^xf(x)+C$$

$3.$ To evaluate $$\int(x-a)^m(x-b)^n\mathrm dx$$ where $m+n=-2$, take either of $(x-a)^2$ or $(x-b)^2$ common and substitute $\frac{x-b}{x-a}$ or $\frac{x-a}{x-b}$ respectively as $t$.

$4.$ To evaluate integrals of the following types, use $$\int P’(x)\mathrm dx=P(x)+C$$ as stated:

$$\begin{array}{c|cc}\text{Integrand}&{P(x)}\\\hline \frac1{(a+b\sin x)^2}&\frac{\cos x}{a+b\sin x}\\ \frac1{(a+b\cos x)^2}&\frac{\sin x}{a+b\cos x}\end{array}$$

$5.$ To evaluate integrals involving terms $(x\cos x+n\sin x)$ in the denominator of the integrand like

$$\int\frac{x^2+n(n-1)}{(x\sin x+n\cos x)^2}\mathrm dx$$ $$\int\frac{(1+x^2)(2+x^2)}{(x\cos x+\sin x)^4}\mathrm dx$$

the compound angle formula proves to be very useful: $$x\cos x+n\sin x=\sqrt{x^2+n^2}\sin\left(x+\cot^{-1}\frac{x}{n}\right)$$

$6.$ Algebraic twins(as it’s known in my country):

To evaluate $$\int\frac{\mathrm dx}{x^4+kx^2+1}, \int\frac{x^2\mathrm dx}{x^4+kx^2+1}$$ write the numerator as $(x^2+1)\pm(x^2-1)$, then split into two integrals, divide by $x^2$ in the numerator and the denominator and recognize $\mathrm d\left(x\pm\frac1{x}\right)$.

This technique is also used to evaluate the trigonometric twins $$\int\sqrt{\tan x}\mathrm dx, \int\sqrt{\cot x}\mathrm dx $$

$7.$ For evaluating integrals that involve $\sqrt{x-\alpha}$ and $\sqrt{\beta-x}$, substitute $x=\alpha\cos^2\theta+\beta\sin^2\theta$ or $x=\alpha\tanh^2\theta+\beta\text{sech}^2\theta$ or $x=\alpha\coth^2\theta+\beta\text{csch}^2\theta$.

$8.$ For evaluating integrals that involve $\sqrt{x-\alpha}$ and $\sqrt{x-\beta}$ where $\beta>\alpha$, substitute $x=\beta\cosh^2\theta-\alpha\sinh^2\theta$ or $x=\alpha\tan^2\theta-\beta\sec^2\theta$ for $x\ge\beta$ and $x=\alpha\cosh^2\theta-\beta\sinh^2\theta$ or $x=\alpha\sec^2\theta-\beta\tan^2\theta$ for $x\le\alpha$.

$9.$ To evaluate integrals of the type $$\int e^{ax}\sin bx\mathrm dx, \int e^{ax}\cos bx\mathrm dx $$ $$\int_0^{2\pi}e^{\cos x}\cos(\sin x)\mathrm dx$$ use $$\cos\theta=\Re e^{\imath\theta}, \sin\theta=\Im e^{\imath\theta} $$ $10.$ $$\int_0^{2a}f(x)\mathrm dx=\int_0^a f(x)+f(2a-x)\mathrm dx$$ $11.$ Here is a list of general strategies to integrate irrational functions involving linear and quadratic polynomials:

$$\begin{array}{c|cc}\text{Integrand}&\text{Substitution}\\ \hline\int\frac{\mathrm dx}{L_1\sqrt{L_2}}&t=\sqrt L_2\\ \int\frac{\mathrm dx}{L\sqrt{Q}}&L=\frac1{t} \\ \int\frac{\mathrm dx}{Q\sqrt{L}}&t=\sqrt L\\ \int\frac{\mathrm dx}{Q_1\sqrt{Q_2}}&x=\frac1{t}\end{array}$$

$12.$ To evaluate integrals of the form $\int\sin^mx\cos^nx\mathrm dx$: $$\begin{array}{c|cc}\text{Condition}&\text{Technique}\\ \hline \text{m and n both are odd}&\text{Substitute the term with the higher power as }t.\\ \text{One is odd, the other even.}&\text{Substitute the term with the odd power as }t.\\ \text{m and n both are even}&\text{Reduce the powers by using multiple angle formulae or De-Moivre’s theorem.}\\ m,n\in\mathbb Q^-\text{ such that }m+n\in\mathbb Z^-& \text{Take }\sec^{-m-n}x\text{ common and substitute }t=\tan x.\end{array}$$

Integreek
  • 8,530
  • 2
    For your interest, another obscure technique--probably for competitive math or recreational math journals like AMM--for certain integrals with an inverse (hyperbolic) trig expression in the numerator and a polynomial in the denominator is to introduce a substitution that, when adding the original integral and the new integral together, "gets rid" of the inverse (hyperbolic) trig function. Example: $$\int_{0}^{\frac{1}{2}}\frac{\tanh^{-1}x}{x^{2}-ax+2a-1}dx = \frac{1}{2}\tanh^{-1}\left(\frac{1}{2}\right)\int_{0}^{\frac{1}{2}}\frac{1}{x^{2}-ax+2a-1}dx$$ for all real numbers $a$. – Accelerator Nov 21 '24 at 11:30
  • On second thought, the equality doesn't work when $a \in \left[\frac{1}{2},4-2\sqrt{3}\right]$ due to the non-integrable singularities the function would have. My bad.‍♂️ – Accelerator Nov 26 '24 at 11:10
  • @Accelerator it's okay...still I am not able to find a substitution to make the $\tanh^{-1}$ disappear no matter how much I try. Can you give a hint? – Integreek Nov 26 '24 at 13:17
  • A little hint.... I wouldn't phrase it as "make the $\operatorname{tanh}^{-1}$ disappear," but more of like finding a substitution $x=g(u)$ such that $\operatorname{tanh}^{-1}(x) + \operatorname{tanh}^{-1}(g(u))$ equals a constant when you add the original and new integrals together. That new substitution keeps the bounds and denominator the same. – Accelerator Nov 26 '24 at 20:04
  • 1
    @Accelerator I finally got it: $x\to\frac{2x-1}{x-2}$, but through reverse engineering by considering $\tanh^{-1}\frac12-\tanh^{-1}x=\tanh^{-1}\left(\frac{2x-1}{x-2}\right)$. Is there an intuitive way to arrive at this substitution? I could've never thought of this on my own as I'm familiar with substitutions like $x\to ax+b$ and $x\to\frac1{x}$ in definite integrals such that the bounds remain the same. – Integreek Nov 27 '24 at 03:28
  • 1
    IDK if there's an intuitive way to come up with it. I made up generalizations of that integral through trial and error. I saw a solution for Problem 2121 of this magazine that used a similar technique which is what inspired me to make up that problem. None of the solutions I post on this website, esp. my most recent answer, is intuitive, so I suppose the answer to your question is seeing someone else solve a problem first then you apply it afterward. Nice job BTW :) – Accelerator Dec 01 '24 at 04:40
1

There is one elegant method to solve integrals using matrice transformations which I saw in an old book. I don't think anyone has mentioned this technique.

Eg:

To evaluate $\int e^{ax}\cos(bx+c)\mathrm dx$, we define a matrix transformation to perform differentiation on a given vector

$$ \frac{\mathrm d(V)}{\mathrm dx} = \begin{bmatrix} a & -b\\ b & a \end{bmatrix} V $$ Such that $$ \frac{\mathrm d}{\mathrm dx} \begin{bmatrix} e^{ax}\cos(bx+c)\\e^{ax}\sin(bx+c) \end{bmatrix} = \begin{bmatrix} a & -b\\ b & a \end{bmatrix} \begin{bmatrix} e^{ax}\cos(bx+c)\\e^{ax}\sin(bx+c) \end{bmatrix} = \begin{bmatrix} ae^{ax}\cos(bx+c)-be^{ax}\sin(bx+c)\\ be^{ax}\cos(bx+c)+ae^{ax}\sin(bx+c) \end{bmatrix} $$

Now, integration will be the inverse of this matrix $$ \therefore \int V\mathrm dx = \begin{bmatrix}a & -b\\b & a\end{bmatrix}^{-1} = \frac1{a^2+b^2}\begin{bmatrix} a & b\\-b & a \end{bmatrix} $$

Now, putting our original vector back, we get:

$$ \int \begin{bmatrix} e^{ax}\cos(bx+c) \\ e^{ax}\sin(bx+c) \end{bmatrix}\mathrm dx = \frac1{a^2+b^2}\begin{bmatrix} a & b \\ -b & a \end{bmatrix} \begin{bmatrix} e^{ax}\cos(bx+c) \\ e^{ax}\sin(bx+c) \end{bmatrix} = \frac1{a^2+b^2}\begin{bmatrix} ae^{ax}\cos(bx+c)+be^{ax}\sin(bx+c)\\ -be^{ax}\cos(bx+c)+ae^{ax}\sin(bx+c) \end{bmatrix}$$ $$\therefore \begin{bmatrix} \displaystyle\int e^{ax}\cos(bx+c)\mathrm dx\\\displaystyle\int e^{ax}\sin(bx+c)\mathrm dx \end{bmatrix} = \begin{bmatrix} \frac1{a^2+b^2}(ae^{ax}\cos(bx+c)+be^{ax}\sin(bx+c))\\ \frac1{a^2+b^2}(-be^{ax}\cos(bx+c)+ae^{ax}\sin(bx+c)) \end{bmatrix} $$

Equating the corresponding elements, we get:

$$\int e^{ax}\cos(bx+c)\mathrm dx=\frac{e^{ax}}{a^2+b^2}(a\cos(bx+c)+b\sin(bx+c))+const. \\ \int e^{ax}\sin(bx+c)=\frac{e^{ax}}{a^2+b^2}(-b\cos(bx+c)+a\sin(bx+c))+const.$$

However, I haven't seen any other examples where you can use this technique; would love to know if there are more.

Integreek
  • 8,530
Ionex
  • 36