3

I have some difficulty with the following exercise:

Calculate $\int_{0}^{\infty}\frac{\sin(xy)}{1+x^2+y^2}dx$.

I tried to differentiate $\frac{\sin(xy)}{1+x^2+y^2}$ to simplify the integral, but the result seems to be more difficult.

I don't have any idea on how to solve this. Can anyone help me or give me a hint?

amWhy
  • 209,954
Jaytone
  • 145
  • The integral depends on $y$. Is it really your willingness ? – mathcounterexamples.net Aug 22 '21 at 07:58
  • 1
    @mathcounterexamples.net oh, still wish you a good evening =)) And yes, I think the answer is depended on y – Jaytone Aug 22 '21 at 08:02
  • 1
    The solution involves the hyperbolic sine integral and hyperbolic cosine integral. Are you familiar with these functions? If not, then perhaps your integral should be $$\int_{0}^{\infty}\frac{\color{red}{\cos(xy)}}{1+x^2+y^2}dx$$ which has an elementary solution. – projectilemotion Aug 22 '21 at 08:14
  • 1
    Some calculations show that $$\int_{0}^{\infty} \frac{\sin(xy)}{1+x^2+y^2} \text{d}x =\mathcal{P.V.}\int_{0}^{\infty} \frac{e^{-xy}}{1-x^2+y^2} \text{d}x$$The Right hand side is Exponential Integral $\text{Ei}(z)$.To arrive the final result,you should make some substitutions and take limits to $(y^2+1){-}$ and $(y^2+1)+$. – Setness Ramesory Aug 22 '21 at 08:18
  • 1
    @projectilemotion I have know hyperbolic sine and hyperbolic cosine but have never heard about shi and chi (their integral) before. I think I have to spare some minutes to look up and explore what it is. Can you help me if the integral is cos(xy) ? – Jaytone Aug 22 '21 at 08:22
  • 2
    If we have $\cos(xy)$ instead, one could use differentiation under the integral sign twice on the integral $$I(a)=\int_0^{\infty} \frac{\cos(ax)}{x^2+b^2}~dx$$ and solve the ODE $I''(a)=b^2 I(a)$. Alternatively, see How to solve the following integral? 4 or Evaluating an integral with Laplace. – projectilemotion Aug 22 '21 at 08:26
  • As the resulting integral I get $\frac{\sin \left(y \sqrt{-y^2-1}\right) \left(\text{Ci}\left(y \left(x-\sqrt{-y^2-1}\right)\right)+\text{Ci}\left(y \left(x+\sqrt{-y^2-1}\right)\right)\right)+\cos \left(y \sqrt{-y^2-1}\right) \left(\text{Si}\left(y \left(x-\sqrt{-y^2-1}\right)\right)-\text{Si}\left(y \left(x+\sqrt{-y^2-1}\right)\right)\right)}{2 \sqrt{-y^2-1}}$ where Ci(x) is the Cosine integral and Si(x) is the Sine integral function. –  Aug 22 '21 at 09:06
  • Although the resulting integral function doesn't look very nice, I hope it helps a little. With the help of substitutions, one can certainly simplify the term. –  Aug 22 '21 at 09:13
  • 1
    The definite integral involves Meijer G function. – Claude Leibovici Aug 22 '21 at 09:36
  • Correct - my flaw is that I considered the wrong range from $-\infty$ to $\infty$. –  Aug 22 '21 at 10:07
  • Considering the correct range from $0$ to $\infty$, then I obtain the following function (and here as @ClaudeLeibovici mentioned the Meijer G function is involved): $\frac{\sqrt{\pi } G_{1,3}^{2,1}\left(\frac{1}{4} \left(y^4+y^2\right)| \begin{array}{c} 1 \ 1,1,\frac{1}{2} \ \end{array} \right)}{y^3+y}$. The following prerequisite condition must be true: $\Im(y)\leq 0\land \left(\Im\left(\sqrt{-y^2-1}\right)\neq 0\lor \left(\Re(y)=0\land \Re\left(y^2\right)+1>0\land \Im(y)+1>0\land \Re\left(\sqrt{-y^2-1}\right)\leq 0\right)\right)$. –  Aug 22 '21 at 10:29
  • @Zacky: I ran Mathematica and used the wrong range in the first trial. After changing it to the correct one I obtained a more compact term that involves the Meijer G function. In order to get a rough idea how these functions look like, I added some plots (the answer section allowed me to add figures). I hope these may help and support the investigations. –  Aug 22 '21 at 11:06

2 Answers2

5

Consider the more general integral for variable $a>0$ and fixed $b>0$: $$I(a)=\int_0^{\infty} \frac{\sin(ax)}{x^2+b^2}~dx. \tag{1}$$ We will show that $I$ satisfies a certain ODE, and then solve it. First, note by integration by parts that $$\begin{align*} I(a)&=\left[-\frac{\cos(ax)}{a(x^2+b^2)}\right]_0^{\infty}-\frac{2}{a}\int_0^{\infty} \frac{x\cos(ax)}{(x^2+b^2)^2}~dx\\&=\frac{1}{ab^2}-\frac{2}{a}\int_0^{\infty} \frac{x\cos(ax)}{(x^2+b^2)^2}~dx. \tag{2} \end{align*}$$ Thus it follows that $$\int_0^{\infty} \frac{x\cos(ax)}{(x^2+b^2)^2}~dx=\frac{1-ab^2I(a)}{2b^2}. \tag{3}$$ Note that $1/a$ is a common factor in equation $(2)$, so multiplying by $a$ and differentiating w.r.t. $a$ gives $$\begin{align*} I(a)+aI'(a)&=2\int_0^{\infty} \frac{x^2\sin(ax)}{(x^2+b^2)^2}~dx\\&=2\int_0^{\infty} \frac{\sin(ax)}{x^2+b^2}~dx-2b^2\int_0^{\infty} \frac{\sin(ax)}{(x^2+b^2)^2}~dx\\&=2I(a)-2b^2\int_0^{\infty} \frac{\sin(ax)}{(x^2+b^2)^2}~dx. \end{align*}$$ Differentiating with respect to $a$ one last time and using equation $(3)$ gives $$\begin{align*} I'(a)+I'(a)+aI''(a)&=2I'(a)-2b^2\int_0^{\infty} \frac{x\cos(ax)}{(x^2+b^2)^2}~dx\\&=2I'(a)-1+ab^2 I(a). \end{align*}$$ Simplifying gives the ODE $$I''(a)-b^2 I(a)=-1/a. \tag{4}$$ We now proceed to solve this differential equation. We can do this using variation of parameters. A fundamental set of solutions for the homogeneous equation is given by $\{e^{ab},e^{-ab}\}$. The method results in the need to consider the exponential integral $\operatorname{Ei}$ defined for nonzero real values of $x$ by $$\operatorname{Ei}(x):=-\int_{-x}^{\infty} \frac{e^{-t}}{t}~dt.$$ The general solution can then be written as $$I(a)=C_1 e^{ab}+C_2 e^{-ab}+\frac{e^{-ab}\operatorname{Ei}(ab)}{2b}-\frac{e^{ab} \operatorname{Ei}(-ab)}{2b}.$$ We now claim that $C_1=C_2=0$. Note that $$|I(a)|\leq \int_0^{\infty} \frac{1}{x^2+b^2}~dx=\frac{\pi}{2b},$$ hence for fixed $b$ we have that $I$ is bounded. Taking the limit as $a\to \infty$ and using Proof of $\lim_{x\to\infty}\frac{\operatorname{Ei}(x)}{e^x}=0$ and showing that $\lim_{x\to\infty}\frac{\operatorname{Ei}(-x)}{e^{-x}}=0$ proves that $C_1=0$. Finally, taking the limit as $a\to 0^+$ shows that $C_2=0$ since $\lim_{a\to 0^+} I(a)=0$. Therefore, the solution to the integral is (if you wish, you can generalize to $a,b\neq 0$ using the symmetry of the integrand) $$I(a)=\frac{e^{-ab}\operatorname{Ei}(ab)}{2b}-\frac{e^{ab} \operatorname{Ei}(-ab)}{2b}.$$ Hence, it follows that for $y\geq 0$ $$\bbox[5px,border:2px solid #C0A000]{\int_{0}^{\infty}\frac{\sin(xy)}{1+x^2+y^2}dx=\frac{e^{-y\sqrt{1+y^2}}\operatorname{Ei}(y\sqrt{1+y^2})}{2\sqrt{1+y^2}}-\frac{e^{y\sqrt{1+y^2}} \operatorname{Ei}(-y\sqrt{1+y^2})}{2\sqrt{1+y^2}}.}$$


Note: An alternative form for $I(a)$ is $$I(a)=\frac{\operatorname{Shi}(ab)\cosh(ab)-\operatorname{Chi}(ab)\sinh(ab)}{b},$$ where $\operatorname{Shi}$ is the hyperbolic sine integral and $\operatorname{Chi}$ is the hyperbolic cosine integral. Alternatively, one can write the result in terms of the Meijer G function as written by @EldarSultanow.

1

The searched integral function involves the Meijer G function and it is given by:

$\int_{0}^{\infty} \frac{\sin(xy)}{1+x^2+y^2}dx=\frac{\sqrt{\pi } G_{1,3}^{2,1}\left(\frac{1}{4} \left(y^4+y^2\right)| \begin{array}{c} 1 \\ 1,1,\frac{1}{2} \\ \end{array} \right)}{y^3+y}$

where the following condition must be met:

$\Im(y)\leq 0\land \left(\Im\left(\sqrt{-y^2-1}\right)\neq 0\lor \left(\Re(y)=0\land \Re\left(y^2\right)+1>0\land \Im(y)+1>0\land \Re\left(\sqrt{-y^2-1}\right)\leq 0\right)\right)$

The plot of original function for $x,y\in[0,10]$ is:

enter image description here

and the plot of the resulting integrated function for $y\in[9,9.5]$ is:

enter image description here