2
$\begingroup$

I am wondering if anyone knows a trick on how to solve this integral $$f(a)=\int_{-\infty}^{\infty} \text{sech}^2(x-a)\text{sech}^2(x+a)dx.$$ The answer should be a function of $a$.

Basically I am trying to reproduce some results from a paper. I've tried Matlab/Wolfram/Mathematica to no avail.

$\endgroup$
3
  • 2
    $\begingroup$ We can put the integral into more a familiar form as $f(a)=16\int_{-\infty}^{+\infty}\frac{1}{\left(e^{2x}+e^{-2x}+e^{2a}+e^{-2a}\right)^2}\ \mathrm{d}x$. Wolfram Alpha manages to evaluate this as $4(2a\coth(2a)-1)\operatorname{csch}^2(2a)$. $\endgroup$ Commented Jul 19, 2019 at 18:22
  • 1
    $\begingroup$ A good exercise is verifying the $a\to0$ limit is $\frac43$, for WA's result as well as the integral. $\endgroup$ Commented Jul 19, 2019 at 21:08
  • $\begingroup$ Without limits this can be found: math.stackexchange.com/questions/792160/int-textsech4-x-dx $\endgroup$ Commented Jul 29, 2019 at 15:16

2 Answers 2

4
$\begingroup$

Use the exponential definition, $\operatorname{sech}(x)=\frac{2}{e^{x}+e^{-x}}$, to rearrange the integrand.

$$\begin{aligned}\operatorname{sech}^2(x-a)\operatorname{sech}^2(x+a)&=\left(\frac{2}{e^{x-a}+e^{-(x-a)}}\right)^2\left(\frac{2}{e^{x+a}+e^{-(x+a)}}\right)^2 \\ &=\frac{16}{\left(e^{2x}+e^{-2x}+c\right)^2} \end{aligned}$$

where $c=e^{2a}+e^{-2a}=\frac{2}{\operatorname{sech}(2a)}$. The integrand is an even function, so $f(a)=32\int_{0}^{+\infty}\frac{1}{\left(e^{2x}+e^{-2x}+c\right)^2}\ \mathrm{d}x$. Substitute $t=e^{2x}$, $\mathrm{d}x=\frac1{2t}\mathrm{d}t$.

$$\begin{aligned}f(a)&=32\int_1^{+\infty}\frac{1}{(t+1/t+c)^2}\cdot\frac{1}{2t}\ \mathrm{d}t \\ &=16\int_1^{+\infty}\frac{t}{(t^2+ct+1)^2}\ \mathrm{d}t \end{aligned}$$

It then suffices to use the standard integrals of $\frac{x}{(Ax^2+Bx+C)^n}$ and $\frac{1}{Ax^2+Bx+C}$. To avoid confusion (since unfortunately we've chosen the same letters), I'll only refer to the constants from the integral table in capitals.

$$\begin{aligned}f(a)&=16\left[-\frac{ct+2}{(4-c^2)(t^2+ct+1)}-\frac{c}{4-c^2}\int\frac{1}{(t^2+ct+1)}\ \mathrm{d}t\right]_1^{+\infty} \\ &=16\left[-\frac{ct+2}{(4-c^2)(t^2+ct+1)}-\frac{c}{4-c^2}\left(\frac{1}{\sqrt{c^2-4}}\ln\left|\frac{2t+c-\sqrt{c^2-4}}{2t+c+\sqrt{c^2-4}}\right|\right)\right]_1^{+\infty} \end{aligned} $$

Note that $4-c^2<0$ except when $a=0$. Hence, we used the integral in the table for the case $4AC-B^2<0$.

As $t\to+\infty$, the $t^2$ in the denominator of the first term dominates, so the term tends to $0$. Likewise, the argument of the logarithm tends to $1$, so the second term tends to $0$. Hence, we just substitute $t=1$ and remember to take the negative sign since it's the integral's lower bound.

$$ \begin{aligned} f(a)&=-16\left[-\frac{c+2}{(4-c^2)(1+c+1)}-\frac{c}{4-c^2}\left(\frac{1}{\sqrt{c^2-4}}\ln\left|\frac{2+c-\sqrt{c^2-4}}{2+c+\sqrt{c^2-4}}\right|\right)\right] \\ &= \frac{16}{(4-c^2)}\left(1+\frac{c}{\sqrt{c^2-4}}\ln\left(\frac{c-\sqrt{c^2-4}}{2}\right)\right) \\ \end{aligned}$$

Note that $c>0$, which allowed us to simplify the absolute value in the logarithm.

Now, to find $f(a)$ in terms of $a$, we substitute back $c=e^{2a}+e^{-2a}$ and find

$$ \begin{aligned} f(a)&=\frac{16}{-(e^{4a}-2+e^{-4a})}\left(1+\frac{e^{2a}+e^{-2a}}{\sqrt{(e^{4a}-2+e^{-4a})}}\ln\left(\frac{e^{2a}+e^{-2a}-\sqrt{(e^{4a}-2+e^{-4a})}}{2}\right)\right) \\ &=\frac{-16}{(e^{2a}-e^{-2a})^2}\left(1+\frac{e^{2a}+e^{-2a}}{e^{2a}-e^{-2a}}\ln\left(\frac{e^{2a}+e^{-2a}-e^{2a}+e^{-2a}}{2}\right)\right) \\ &=\frac{16}{(e^{2a}-e^{-2a})^2}\left(2a\frac{e^{2a}+e^{-2a}}{e^{2a}-e^{-2a}}-1\right) \\ &=4\left(2a\frac{e^{2a}+e^{-2a}}{e^{2a}-e^{-2a}}-1\right)\left(\frac{2}{e^{2a}-e^{-2a}}\right)^2 \\ &=4\left(2a\coth(2a)-1\right)\operatorname{csch}^2(2a) \end{aligned}$$

Where, in the second line, we have used $(e^{4a}-2+e^{-4a})=(e^{2a}-e^{-2a})^2$. Hence, the solution to the integral agrees with the one given by Wolfram Alpha.

$\endgroup$
2
$\begingroup$

Using the identity $\cosh \left( x\pm a \right)=\cosh x\cosh a\pm \sinh x\sinh a$ the integrand can be written as: $$ \begin{align} & ={{\left( \frac{1}{\cosh \left( x-a \right)}\frac{1}{\cosh \left( x+a \right)} \right)}^{2}} \\ & =\frac{1}{{{\left( {{\cosh }^{2}}x{{\cosh }^{2}}a-{{\sinh }^{2}}x{{\sinh }^{2}}a \right)}^{2}}} \\ & =\frac{{{\operatorname{sech}}^{4}}x\ {{\operatorname{sech}}^{4}}a}{{{\left( 1-{{\left( \tanh x\ \tanh a \right)}^{2}} \right)}^{2}}} \\ \end{align} $$ Enforcing $u=\tanh x\ \tanh a$ where $a\ne 0$ and the identity ${{\operatorname{sech}}^{2}}x=1-{{\tanh }^{2}}x$

$$ f\left( a \right)=2\frac{{{\operatorname{sech}}^{4}}a}{{{\tanh }^{3}}a}\int_{0}^{\tanh a}{\frac{{{\tanh }^{2}}a-{{u}^{2}}\ }{{{\left( 1-{{u}^{2}} \right)}^{2}}}du} $$ For the last integral(##)

$$\begin{align} & ={{\tanh }^{2}}a\ \int{\frac{\ du}{\left( 1-{{u}^{2}} \right)}}+\left( {{\tanh }^{2}}a\ -1 \right)\int{\frac{{{u}^{2}}\ }{{{\left( 1-{{u}^{2}} \right)}^{2}}}du} \\ & ={{\tanh }^{2}}a\ {{\tanh }^{-1}}u+\left( {{\tanh }^{2}}a\ -1 \right)\ \times -\frac{1}{2}\left( \frac{u}{{{u}^{2}}-1}+{{\tanh }^{-1}}u \right) \\ \end{align}$$ Finally $$f\left( a \right)=2\frac{{{\operatorname{sech}}^{4}}a}{{{\tanh }^{3}}a}\left( \frac{1}{2}\left( a\left( {{\tanh }^{2}}a+1 \right)-\tanh a \right) \right)$$.

(##) I have used the following integration rules: $$\begin{align} & \int{\frac{\ du}{\left( 1-{{u}^{2}} \right)}}={{\tanh }^{-1}}u+C \\ & \int{\frac{{{u}^{2}}\ }{{{\left( 1-{{u}^{2}} \right)}^{2}}}du}=-\frac{1}{2}\left( \frac{u}{{{u}^{2}}-1}+{{\tanh }^{-1}}u \right)+C \\ \end{align}$$ Notice that the second one is easily verified using integration by parts.

$\endgroup$
8
  • $\begingroup$ You seem to get a different solution to me and Wolfram Alpha: desmos.com/calculator/4qk6u5mzvd $\endgroup$ Commented Jul 19, 2019 at 23:50
  • $\begingroup$ i think i made a mistake some where?!? $\endgroup$ Commented Jul 19, 2019 at 23:51
  • $\begingroup$ It seems to be at $f\left( a \right)=2\frac{{{\operatorname{sech}}^{4}}a}{{{\tanh }^{3}}a}\int_{0}^{\tanh a}{\frac{{{\tanh }^{2}}a-{{u}^{2}}\ }{{{\left( 1-{{u}^{2}} \right)}^{2}}}du}$. $\endgroup$ Commented Jul 19, 2019 at 23:56
  • $\begingroup$ let me check my calculation $\endgroup$ Commented Jul 19, 2019 at 23:57
  • $\begingroup$ i think it is true?1?! $\endgroup$ Commented Jul 20, 2019 at 0:03

You must log in to answer this question.

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.