4

I have edited the question to include what I think is an interesting property of this integral. I have noticed experimentally that:

$$\int_{0}^{1} \frac{1-x^2}{x^4+2x^3+2x^2-2x+1} dx=\frac{\pi}{4},\tag{1}$$

$$\int_{0}^{1} \frac{x}{x^4+2x^3+2x^2-2x+1} dx=\frac{\pi}{8},\tag{2}$$

$$\int_{0}^{1} \frac{1+x-x^2}{x^4+2x^3+2x^2-2x+1} dx=\frac{3\pi}{8}.\tag{3}$$

So a little variation in the numerator seems to produce always something like $n\pi$. Here is my new question: what is exactly the relation between $n$ and the numerator of the integrand?

The closest thing to my formula is formula $(34)$ in this list of formulas, but it is not exactly the same as mine.

  • 2
    Not to me - But is asking if something is known really a question? Is there something more specific you are trying to get at with this integral? – Snared Jul 16 '23 at 02:35
  • 1
    Approach Zero doesn't seem to have any results. Where did you get it from? – Accelerator Jul 16 '23 at 02:43
  • 5
    I am interested in references. What's wrong with my question? There is a tag for reference request for some reason. – Emmanuel José García Jul 16 '23 at 02:43
  • 2
    it's a really nice looking integral – Snared Jul 16 '23 at 02:50
  • @Accelerator: I came up with it from my imagination. ;) – Emmanuel José García Jul 16 '23 at 03:55
  • This can be solved by partial fraction decomposition, although the roots of the given denominator aren't very pleasurable. – Shivansh Jaiswal Jul 16 '23 at 04:35
  • 1
    Can someone explain to me why there are three votes to close this question when there is even a tag for requesting references? – Emmanuel José García Jul 16 '23 at 04:40
  • 2
    The denominator has a special Galois group, a transitive copy of $C_2 \times C_2 < S_4$. https://math.stackexchange.com/questions/3248128/finding-galois-group-over-rationals?noredirect=1 – Travis Willse Jul 16 '23 at 05:08
  • I see. As for the close votes, maybe it's because people just see the equality without any steps to try to prove it? – Accelerator Jul 16 '23 at 08:01
  • 1
    @Accelerator: Publishing a solution to this integral has no more weight than coming up with it. There are websites and apps that can solve this integral, but I don't think there are any websites or apps that will give you new pi formulas like this, do they? My interest is to know if it is new. It looks like it is. I will post my solution to my blog tomorrow or the day after tomorrow and I will put a link here. But again, I think that the integral itself is more interesting than any solution that even a computer can give. – Emmanuel José García Jul 16 '23 at 08:21
  • 4
    I think part of the issue here is that while this integral representation for $\pi$ is undoubtedly very nice, and you should feel proud of finding it (I have my own nice $\pi$ formula I found a while back), these sorts of results are considered simple enough that they are not usually published (other than in more recreational settings). Maybe the answers you're getting are somewhat trying to say "maybe this particular formula hasn't been published anywhere, but formulas of the sort are well understood, and aren't really novel; this simple proof I've provided is evidence of that fact". – Christian E. Ramirez Jul 16 '23 at 08:47
  • In any case, +1 for the cool identity. – Christian E. Ramirez Jul 16 '23 at 08:58
  • 2
    I agree with C-RAM. It's hard to say if this formula is new, because it's quite possible that someone might have found it before but not bothered to publish it. Such formulas are generally not considered to be interesting enough to be published "for their own sake"; there needs to be some further motivation, such as a fast algorithm for calculating $\pi$, or a better bound on the irrationality measure of $\pi$, or some connection to mathematics that is still not well understood. There are infinitely many rational functions that integrate to $\pi$, so which ones would one pick to publish? – Timothy Chow Jul 16 '23 at 12:30
  • Having said that, I think that if you have a new method for generating rational functions of low degree with small coefficients that integrate to $\pi$, that might be interesting. – Timothy Chow Jul 16 '23 at 12:32
  • @C-RAM: Thank you for the upvote, but when did I say that I wanted to publish this result anywhere other than on my blog? However, even if I decided to publish it, there are journals for elementary problems whose only requirement is that the problem be original and interesting. What I was telling Accelerator is that I find the integral itself more interesting than any solution we can provide. – Emmanuel José García Jul 16 '23 at 14:09
  • @C-RAM: Because solving elementary integrals is such a mechanical process that even a computer can do it, but the same cannot be said for Olympiad geometry problems; those require creativity. Do you know of any software that generates integrals that yield π in such a cool way as this one? I spent weeks carefully studying a certain type of integrals for another purpose, and coincidentally came across this result. – Emmanuel José García Jul 16 '23 at 14:09
  • @TimothyChow: When did I mention that I want to publish this result anywhere other than on my blog? I asked if this integral is known (or well-known), not if it is entirely new, which, as you say, is impossible to determine. I wanted to know if there was a reference, but it seems, judging by the 4 votes to close the question, that this question is off-limits for me (not for others), even when there is a tag for asking for references. – Emmanuel José García Jul 16 '23 at 14:21
  • @TimothyChow: On the other hand, if these π formulas have no interest at all as you suggest, how do you explain the existence of entire pages dedicated exclusively to them? Why would someone bother to compile them if they have no interest? – Emmanuel José García Jul 16 '23 at 14:22
  • Not trying to make a point here; I'm just trying to explain why the answerers might have been compelled to answer the way they did, with proofs rather than references. (I can't read their minds though, so perhaps I shouldn't be speculating) – Christian E. Ramirez Jul 16 '23 at 14:47
  • 1
    On the topic of software that generates formulas like this, I'm unaware of any in specific, but I know that such algorithms do exist for certain certain classes of expressions, and more broadly, experimental mathematics (it's an interesting field) often deals with the problem of finding interesting mathematical expressions. The famous BBP formula for $\pi$ was found using such an algorithm. Again though, none of this should diminish your accomplishment. I myself have found a few formulas like this one that I'm still quite fond of despite the fact that they're ultimately nothing groundbreaking. – Christian E. Ramirez Jul 16 '23 at 14:51
  • @C-RAM: I believe that the responses from Quanto and Claude should be deleted by the administrators since they do NOT answer my question. I added a tag: reference-request. Let's suppose that I have rediscovered one of the many formulas of Ramanujan and ask for references on this blog. Do you think the correct answer would be to prove the identity instead of providing a reference? If they are fair, those answers should be deleted by the administrators. Nonetheless, thank you for your comments. – Emmanuel José García Jul 16 '23 at 15:14
  • @EmmanuelJoséGarcía There is not a whole page dedicated exclusively to integrals of rational functions whose value is $\pi$. The formulas you refer to generally have some interest beyond the bare fact that they are formulas for $\pi$. For example, the formula you found which is closest to yours is the BBP formula, which is famous for giving an algorithm to find the $n$th binary (or hexadecimal) digit of $\pi$ without having to compute all the preceding digits. Does your formula have any interesting properties beyond the bare fact that it is a formula for $\pi$? – Timothy Chow Jul 16 '23 at 15:18
  • @TimothyChow: Look, for example, at the following page dedicated to the appearance of π in Pascal's Triangle. What intrigues us is the unexpected appearance of π in Pascal's Triangle, nothing else. If it were something else, the title would be different, most likely. Link: https://www.cut-the-knot.org/arithmetic/algebra/PiInPascal.shtml – Emmanuel José García Jul 16 '23 at 15:25
  • The integral I am presenting is certainly unexpected. Who would say at first glance that π would arise from that integral? I said there pages entirely dedicated to formulas for $\pi$, not just integrals. – Emmanuel José García Jul 16 '23 at 15:31
  • @EmmanuelJoséGarcía I'm not sure what you think you're achieving by being argumentative. I'm trying to explain to you how professional mathematicians think about this subject and why asking whether your formula is "new" may not have a definitive answer. Not being obvious at first glance is cute, but there are many similar formulas that are not obvious at first glance, such as $$\int_0^1 {x^8(1-x)^8(25+816x^2)\over 3164(1+x^2)},dx={355\over113}-\pi$$ but such formulas are not hard to come up with and they're not going to be published unless there's some further motivation. – Timothy Chow Jul 16 '23 at 15:43
  • @TimothyChow: See the following link too (in Spanish). The only interest have been the appearance of $\pi$ and $\phi$ in integrals: https://www.gaussianos.com/integrales-pi-aureas/ – Emmanuel José García Jul 16 '23 at 15:45
  • @EmmanuelJoséGarcía Let me try to put it another way. If your intent is to post to a blog, go for it. There's no need to ascertain whether the formula is "new." It has some intrinsic interest, and you found it by yourself. That's more than enough justification for a blog post. But your specific question about novelty opens up a can of worms that I think is irrelevant to your concerns. – Timothy Chow Jul 16 '23 at 15:48
  • @TimothyChow: You seem to be unfamiliar with elementary mathematics journals where this integral fits perfectly as a small challenge for students. If this integral has interesting properties that go beyond my own understanding, how will someone more knowledgeable and intelligent than me know about it if I don't publish it somewhere? – Emmanuel José García Jul 16 '23 at 15:53
  • @EmmanuelJoséGarcía I have published articles in elementary mathematics journals, so I'm quite familiar with them. You're drifting away from your original question, which was about novelty. If you now have a different question, then you may want to revise your question or post a separate question. – Timothy Chow Jul 16 '23 at 16:07
  • @TimothyChow: So, according to you, none of these people are professional mathematicians just because they are interested solely in the appearance of φ in integrals? A question with 167 upvotes! MathSE link: https://math.stackexchange.com/questions/1653979/is-there-any-integral-for-the-golden-ratio?page=1&tab=scoredesc#tab-top – Emmanuel José García Jul 16 '23 at 16:26
  • 5
    @EmmanuelJoséGarcía Again, I question what you think you're gaining by being argumentative. You continue to digress from your original question about novelty. – Timothy Chow Jul 16 '23 at 16:59
  • @TimothyChow: I am presenting evidence that you are mistaken. I am still waiting for references. – Emmanuel José García Jul 16 '23 at 17:01
  • @EmmanuelJoséGarcía You have not presented evidence that I am mistaken with regard to your original question about novelty. Please try to stay on topic. You are criticizing others for not answering your original question, but you yourself are failing to stay on topic. As I said, someone might well have come up with your integral before but not bothered to publish it. – Timothy Chow Jul 16 '23 at 17:02
  • @TimothyChow: You suggest that professional mathematicians are not interested in the mere appearance of $\pi$ in unexpected contexts. Your assertion should apply to other constants I guess, so I have shown you evidence that you are wrong with $\pi$ and also for $\phi$. – Emmanuel José García Jul 16 '23 at 17:06
  • 2
    @EmmanuelJoséGarcía That is not what I said. Again, I urge you to reflect on what you think you are gaining by being argumentative. https://xkcd.com/386/ – Timothy Chow Jul 16 '23 at 17:07
  • @TimothyChow: "As I said, someone might well have come up with your integral before but not bothered to publish it."

    The reference-request tag is for references in books, articles, notes, etc. Not for speculation... I am still waiting for the references.

    – Emmanuel José García Jul 16 '23 at 17:16
  • @EmmanuelJoséGarcía While I do see why you may be against the posting of proofs, this site is a community forum. I do not think your requests to delete the current answers (with a considerable amount of upvotes) will be fulfilled. Discussion is encouraged (especially in the comment section), and I think that your eristic tone does not reflect this quality. Clearly, nobody has found the reference you are looking for yet (I've done some of my own searching with no luck). Give it some time, maybe somebody will come forth with a satisfying answer. – Scene Jul 16 '23 at 21:16
  • @TimothyChow: I have edited my question. – Emmanuel José García Jul 17 '23 at 16:27
  • Integrals of rational functions are all "known" in principle. (It depends only on the problem of factoring a polynomial over $\mathbb C$.) Claude's answer shows your first one explicitly. – GEdgar Jul 17 '23 at 16:45
  • @TravisWillse: That is interesting observation! I wonder if it is possible to characterize the integrand $\frac{P(x)}{Q(x)}$ such that by inspection we can say: this is $n\pi$. – Emmanuel José García Jul 17 '23 at 16:52
  • @TravisWillse: Please, notice I have edited my question. – Emmanuel José García Jul 17 '23 at 16:57
  • 1
    Maybe a coincidence, but your third integral has the same value as $\operatorname B\left(\frac12,\frac52\right)=\frac{3\pi}8$ where $\rm B$ denotes the beta function. Perhaps there's a (chain of) substitutions that connects your integral to beta? – user170231 Jul 17 '23 at 17:51

3 Answers3

7

Substitute $x=\tan \frac t2$\begin{align} &\int_{0}^{1} \frac{1-x^2}{x^4+2x^3+2x^2-2x+1} dx\\ =& \int_0^{\pi/2} \frac{\cos t}{2-\sin 2t }dt \overset{x\to \frac\pi2 - x}= \frac12\int_0^{\pi/2} \frac{\cos t+\sin t}{2-\sin 2t }dt\\ =&\ \frac12 \int_0^{\pi/2} \frac{d(\sin t-\cos t)}{1+(\sin t -\cos t)^2} =\frac\pi4 \end{align} and, similarly \begin{align} \int_{0}^{1} \frac{x}{x^4+2x^3+2x^2-2x+1} dx = \frac12\int_0^{\pi/2} \frac{\sin t}{2-\sin 2t }dt =\frac\pi8 \end{align} In general \begin{align} \int_{0}^{1} \frac{ax^2 +b x + c}{x^4+2x^3+2x^2-2x+1} dx =\frac\pi8(c+b-a)+\frac\pi{3\sqrt3}(a+c) \end{align}

Quanto
  • 97,352
  • 2
    Somebody does not like our answers ! Cheers and (+1) for your which is simple and nice. Cheers :-) – Claude Leibovici Jul 16 '23 at 03:54
  • 3
    @ClaudeLeibovici: My actual question is whether this integral is known or not. – Emmanuel José García Jul 16 '23 at 04:54
  • I have chosen your answer because it gives a general formula for when the numerator is of degree 2. More generally, do you think it is possible to characterize $\frac{P(x)}{Q(x)}$ in such a way that by inspection we can say: this is $n\pi$? This would solve a large number of rational integrals that give $\pi$ in one shot. – Emmanuel José García Jul 17 '23 at 20:55
  • 1
    @EmmanuelJoséGarcía - It breaks down for the cubic power $$\int_{0}^{1} \frac{x^3}{x^4+2x^3+2x^2-2x+1} dx=\frac12\ln 2+\bigg(\frac18-\frac1{3\sqrt3}\bigg)\pi $$ – Quanto Jul 17 '23 at 21:20
  • Probably because you have left the denominator fixed. The question is more general: what relationship must there be between the coefficients of the numerator of degree $n$ and the denominator of degree $m$ so that the integral always produces $a\pi$? – Emmanuel José García Jul 17 '23 at 21:37
  • For example, this is a variant of formula $(34)$ in this list of $\pi$ formulas: $$\int_{0}^{1} \frac{x^2-x-1}{x^4-2x^3+4x-4},dx=\frac{3\pi}{16}.$$ The fact that the denominator is not the same suggests that a further generalization is possible. – Emmanuel José García Jul 18 '23 at 02:35
  • 1
    The following is a general formula for the BBP integral for $\pi$ (see eq. (34) in the list of $\pi$ formulas):

    $$\int_{0}^{1} \frac{ax^2+bx+c}{x^4-2x^3+4x-4},dx=\frac{\pi}{16}\color{red}{(2a-c)}+\frac{\ln{(3-\sqrt8)}}{\sqrt32}\color{red}{(b+c)}+\frac{\ln{(3-\sqrt8)}}{\sqrt8}\color{red}{a}.$$

    From this formula, we can deduce that the integral will yield rational multiples of $\pi$ when $b$ and $c$ are opposite to each other and $a=0$.

    – Emmanuel José García Jul 18 '23 at 15:45
  • 1
    @EmmanuelJoséGarcía - Glad you worked it out – Quanto Jul 18 '23 at 16:12
4

Rewrite $$x^4+2x^3+2x^2-2x+1=(x-a)(x-b)(x-c)(x-d)$$ where

$$a=-\frac{\sqrt{3}+1}{2} (1+i)\qquad\qquad b=-\frac{\sqrt{3}+1}{2} (1-i)$$ $$c=\frac{\sqrt{3}-1}{2} (1+i)\qquad\qquad d=\frac{\sqrt{3}-1}{2} (1-i)$$

Using partial fraction decomposition $$\frac{1-x^2}{x^4+2x^3+2x^2-2x+1}=\frac{1-a^2}{(a-b) (a-c) (a-d) (x-a)}+$$ $$\frac{b^2-1}{(a-b) (b-c) (b-d) (x-b)}+\frac{c^2-1}{(a-c) (c-b) (c-d) (x-c)}+$$ $$\frac{d^2-1}{(a-d) (d-b) (d-c) (x-d)}$$

And use, for $k$ being a complex $$\int_0^1\frac {dx}{x-k}=\log \left(\frac{k-1}{k}\right)$$

Replacing $(a,b,c,d)$ by their values and simplifying the complex numbers $$-\frac{\left(\frac{1}{144}-\frac{i}{144}\right) \left(3 i+(1+2 i) \sqrt{3}\right) (\pi +6 i \log (2))}{1+\sqrt{3}}+$$ $$\frac{\left(\frac{1}{4}+\frac{i}{4}\right) \left(\sqrt{3}+(2+i)\right) \log \left(\frac{\sqrt{3}+(2+i)}{1+\sqrt{3}}\right)}{3+\sqrt{3}}+$$ $$\frac{\left(\frac{1}{4}-\frac{i}{4}\right) \left(\sqrt{3}+(-2+i)\right) \log \left(\frac{\sqrt{3}+(-2+i)}{\sqrt{3}-1}\right)}{\sqrt{3}-3 }+$$ $$\frac{\left(\frac{1}{4}+\frac{i}{4}\right) \left(\sqrt{3}+(-2-i)\right) \log \left(\frac{\sqrt{3}+(-2-i)}{\sqrt{3}-1}\right)}{\sqrt{3}-3 }=\color{blue}{\huge\frac \pi 4}$$

3

More generally, let

$$ a_k = \int_0^1 \frac{x^k}{x^4 + 2 x^3 + 2 x^2 - 2 x + 1} \; dx $$

Of course $$a_{k+4} + 2 a_{k+3} + 2 a_{k+2} - 2 a_{k+1} + a_k = \int_0^1 x^k \; dx = \frac{1}{k+1} $$ so we can get $a_k$ for all natural numbers $k$ expressed as linear combinations over the integers of $a_0$ to $a_3$. Using Claude's methods: $$ \eqalign{ a_0 &= \frac{\pi \sqrt{3}}{9} + \frac{\pi}{8}\cr a_1 &= \frac{\pi}{8}\cr a_2 &= \frac{\pi\sqrt{3}}{9} - \frac{\pi}{8}\cr a_3 &= -\frac{\pi \sqrt{3}}{9}+\frac{\pi}{8}+\frac{\ln \! \left(2\right)}{2}\cr} $$ If you want a polynomial $p(x)$ of degree $\le 3$ with rational coefficients so $\int_0^1 \frac{p(x)}{x^4 + 2 x^3 + 2 x^2 - 2 x + 1}\; dx$ is a rational number times $\pi$, you want the coefficient of $x^3$ to be $0$ (because you can't handle $\ln(2)$), and the coefficients of $x^0$ and $x^2$ must add to $0$ (to get rid of the $\sqrt{3}$).

Robert Israel
  • 448,999