6

In my post, I had proved $$I(a,b)=\int_{0}^{\infty} \frac{\sin (a x) \sin (b x)}{x^{2}} \, \mathrm d x = \frac{\min (a, b) \pi}{2}, $$ in 2021. Recently, I started to investigate the integral further to: $$I(a_1,a_2,\cdots,a_n)=\int_{0}^{\infty} \frac{\sin (a_1x) \sin (a_2 x)\cdots\sin(a_nx)}{x^{2}} \,\mathrm d x.$$ Trying with small number $n$, I found and conjectured that $$ \int_0^{\infty} \frac{\sin (a_1 x) \sin (a_2 x) \cdots \sin (a_n x)}{x^2} \, \mathrm d x = \begin{cases} \frac{\pi}{2^n} \sum_{k=0}^{2^{n-1}}(-1)^{k+1}\left|A_k\right|\quad \quad \text {if } n \text { is even } \\ \frac{1}{2^{n-1}} \sum_{k=0}^{2^{n-1}}(-1)^k A_k\ln|A_k|\quad \text {if } n \text { is odd }\end{cases} $$ where $A_k$ is a sum of $(n-k)$ $ a_i$’s and $k$ $-a_j$’s.


Starting from $n=3$, $$ \begin{aligned} &\quad \sin(ax)\sin(bx)\sin(cx) \\&= \frac{e^{a x i}-e^{-a x i}}{2 i} \cdot \frac{e^{b x i}-e^{-b x i}}{2 i} \cdot \frac{e^{c x i}-e^{-c x i}}{2 i} \\ &= \frac{1}{8 i^3}\left[\left(e^{(a+b+c) i x}-e^{-(a+b+c) i x}\right) -\left(e^{(a+b-c) i x}-e^{-(a+b-c) i}\right) -\left(e^{(a+c-b) i x}-e^{-(a+c-b) i x}\right) \right.\\ & \quad \left. -\left(e^{(b+c-a) i x}-e^{-(b+c-a) i x}\right)\right] \\ &= -\frac{1}{4} [ {\sin ((a+b+c) x)} - {\sin ((a+b-c)x)} -{\sin ((a+c-b) x)} -{\sin ((b+c-a)x)} ] \end{aligned} $$ Via integration by parts, we have $$ \begin{align} &\quad \int_0^{\infty} \frac{\sin (a x) \sin (b x) \sin (c x)}{x^2} \, \mathrm d x \\ &= -\frac{1}{4} \int_0^{\infty} \frac{\sin (A x)-\sin (B x)-\sin (C x)-\sin (D x)}{x^2} \, \mathrm d x \tag{*} \\ &= \frac{1}{4} \int_0^{\infty} \frac{A \cos (A x)-B \cos (B x)-C \cos (C x)-D \cos (D x)}{x} \, \mathrm dx\\ &= \frac{1}{4} \int_0^{\infty}\left[A \cos (A x)-B \cos (B x)-C \cos (C x)-D \cos (D x)\right]\int_0^{\infty} e^{-yx} \, \mathrm d y \,\mathrm d x \\ &= \frac{1}{4} \int_0^{\infty} \int_0^{\infty}\left[A \cos (A x)-B \cos (B x)-C \cos (C x)-D \cos (D x)\right] e^{-yx} \, \mathrm d x \, \mathrm d y \\ &= \frac{1}{4} \int_0^{\infty}\left[ -\frac{A y}{y^2+A^2}+\frac{B y}{y^2+B^2}+\frac{C y}{y^2+C^2}+\frac{D y}{y^2+D} \right] \, \mathrm d y \\ &= \frac{1}{8}\left[-A \ln \left(y^2+A^2\right)+B \ln \left(y^2+B^2\right)+C \ln \left(y^2+C^2\right)+D \ln \left(y^2+D^2 \right)\right]_0^\infty \\ &= \frac{1}{4} \ln \left(|A|^A|B|^{-B}|C|^{-C}|D|^{-D}\right)\end{align} $$ due to $A-B-C-D=0$, where $A=a+b+c, B=a+b-c,C=a+c-b$ and $D=b+c-a$.

In the step $(*)$, we keep non-zero capital letters only.

For examples,

A. When $n=3$, $$ \int_0^{\infty} \frac{\sin x \sin (2 x) \sin (3 x)}{x^2} \, \mathrm d x = \frac{1}{4} \ln \left(6^6 \cdot 2^{-2} \cdot 4^{-4}\right) =\ln \left(\frac{3 \sqrt{3}}{2}\right) $$

B. When $n=5$,

Following the same steps as $n=3$, we have

$$\sin x \sin (2 x) \sin (3 x)\sin(4x)\sin(5x)=\frac{1}{16} \left(\sin (15 x)-\sin (13 x)-\sin (11 x)+\sin (5 x)+\sin (3 x)+\sin x\right) $$ $$\int_0^{\infty} \frac{\sin x \sin (2 x) \sin (3 x) \sin (4 x) \sin (5 x)}{x^2} \, \mathrm d x = \frac{1}{16} \ln \left(\frac{13^{13} \cdot 11^{11}}{15^{15} \cdot 5^5 \cdot 3^3}\right) =\ln \left(\frac{13^{\frac{13}{16}} \cdot 11^{\frac{11}{16}}}{15 \cdot \sqrt[4]{5} \cdot \sqrt[8]{3}}\right)$$


Now if $n>2$ is even, say $n=4$ then \begin{aligned} &\quad \sin (a x) \sin (b x) \sin (e x) \sin (d x) \\ & = \frac{e^{axi} -e^{-a x i}}{2 i} \cdot \frac{e^{bxi}-e^{-b x i}}{2 i} \frac{e^{cx i} e^{-cxi}}{2 i} \cdot \frac{e^{dxi }-e^{-dxi}}{2 i} \\ & = \frac{1}{16}\left[e^{(a+b+c+d) xi}+e^{-(a+b+c+d) xi}\right. -\left(e^{(a+b+c-d) xi}+e^{-(a+b+c-d) xi}\right) \\ & \quad -\left(e^{(a+c+d-b) xi}+e^{-(a+c+d-b) xi}\right) -\left(e^{(a+b+d-c)xi}+e^{-(a+b+d-c) i x}\right) \\ & \quad -\left(e^{(a+c+d-a) xi}+e^{-(b+c+d-a) xi}\right) +\left(e^{(a+b-c-d) xi}+e^{-(a+b-c-d) xi}\right) \\ & \quad +\left(e^{(a+c-b-d) i x}+e^{-(a+c-b-d) xi}\right) \left.+\left(e^{(a+d-b-c) xi}+e^{-(a+d-b-c) xi}\right)\right] \\ & =\frac{1}{8}[\cos ((a+b+c+d) x)-\cos ((a+b+c-d) x) -\cos ((a+c+d-b) x(-\cos ((a+b+d-c) x )\\ & \quad -\cos ((b+c+d-a) x)+\cos ((a+b-c-d) x) +\cos ((a+c-b-d) x)+\cos ((a+d-b-c) x] \end{aligned}

Plugging back with integration by parts yields \begin{align} &\quad \int_0^{\infty} \frac{\sin (a x) \sin (bx) \sin (c x) \sin (d x) }{x^2} \, \mathrm d x \\&= \small{ -\frac{1}{8} \int_0^{\infty} \frac{A \sin (A x)-B \sin (B x)-C \sin (C x) -D \sin (D x)-E \sin (E x)+F \sin (F x)-G \sin (G x)-H \sin (H x)}{x}} \, \mathrm dx \\ &=\frac{\pi}{16}(-|A|+|B|+|C|+|D|+|E|-|F| -|G|-|H|) \end{align} where \begin{align} A &= a+b+c+d, & B &= a+b+c-d, \\ C &= a+c+d-a, & D &= a+b+d-c \\ E &= b+c+d-a, & F &= a+b-c-d, \\ G &= a+c-b-d, & H &= a+d-b-c. \end{align}

For example,

$$ \boxed{\int_0^{\infty} \frac{\sin x \sin (2 x) \sin (3 x) \sin (4 x)}{x^2} \, \mathrm d x = \frac{\pi}{16}(-10+2+6+4+8-4-2-0) = \frac{\pi}{4}}. $$


Conclusion

When $n$ is even, the product of sines can be expressed as a linear sum of cosine of multiple angles. Via integration by parts, it is changed to a sum sine of multiple angles and hence exact value of integral is $h\pi/2$.

When $n$ is odd, the product of sines can be expressed as a linear sum of sine of multiple angles. Via integration by parts, it is changed to a sum of cosine of multiple angles and hence exact value of integral is $\ln k$.


However, it is very hard to prove my conjecture rigorously.

Lai
  • 31,615
  • 1
    Have you tried using the residue theorem? Also, your conjecture is very non-specific if you do not specify $h$ and $k$ – Johannes Moerland Mar 26 '25 at 12:46
  • See Borwein integrals (especially, the paragraph “General formula”). – nejimban Mar 26 '25 at 12:47
  • 1
    @nejimban Borwein provides a formula for $\int_0^\infty x^{-n}\prod_{j=1}^n \sin(a_jx),dx$, but here the question is about $\int_0^\infty x^{-2}\prod_{j=1}^n \sin(a_jx),dx$ – Sassatelli Giulio Mar 26 '25 at 13:29
  • Looks to me that you have everything that you need, and it's nicely summed up in the conclusion. So what's your difficulty with proving this conjecture rigorously? EG Are you asking why "When $n$ is even/odd, the product of $n$ sines can be expressed as a linear sum of cosine/sine of multiple angles."? Or are you asking for what's the formula for $h, k$ (which follows from the previous question)? – Calvin Lin Mar 26 '25 at 19:01
  • 1
    @SassatelliGiulio You're right, but I think Borwein's technique can still work, at least in the even case. In his paper, Borwein computes integrals of the form $$\int_0^{+\infty}\prod_{k=1}^n\operatorname{sinc}(a_k x)\prod_{k=n+1}^{n+m}\cos(a_k x),\mathrm dx.$$ – nejimban Mar 26 '25 at 19:08
  • title should say rigorously! rather than vigorously! https://www.merriam-webster.com/dictionary/vigorously – Will Jagy Mar 26 '25 at 20:56
  • Yes, you are right! Fixed. Thank you very much. – Lai Mar 26 '25 at 22:58

0 Answers0