2
$\begingroup$

This question is related to my question here such that i want to find a closed form of $\int_{-1}^1 x^{2k} (\operatorname{erf}(x))^k \,dx $ , for $k$ is even integer because for odd integer is $0$ as we have integrand of odd function , Really for $k=2$ it gives a nice closed form as shown here in wolfram alpha, Now my question here is :

Question:
Is it possible to get a Closed form of :$\int_{-1}^1 x^{2k} \operatorname{erf}(x)^k \,dx $ for $ k$ is even integer?

Note: The Motivation of this question is to find a Series representation of :$\int _{0}^{t}\exp(-x^2 \operatorname{erf}(x))dx$

$\endgroup$
  • 2
    $\begingroup$ As a general rule of thumb: since $1$ and $-1$ are not spacial points for the integrand, his definite integral is likely just as difficult as the indefinite integral. So: $\int x^4 \text{erf}(x)^2 dx$ has a known closed form, plug in and get your definite integral. But for $\int x^8 \text{erf}(x)^4 dx$ you are probably out of luck. $\endgroup$ – Gerald Edgar May 31 '18 at 23:57
  • $\begingroup$ What for? Isn't it art for art's sake? $\endgroup$ – user64494 Jun 1 '18 at 9:03
  • $\begingroup$ No, it is not art for art's sake. $\endgroup$ – Liviu Nicolaescu Jun 1 '18 at 13:15
  • $\begingroup$ @Liviu Nikolaescu: Please, base your statement. $\endgroup$ – user64494 Jun 1 '18 at 17:50
  • $\begingroup$ @user64494, Probably Liviu Nicolaescu think that is easy for evaluation $\endgroup$ – zeraoulia rafik Jun 1 '18 at 17:52
3
$\begingroup$

I understand from the OP that the motivation for this question is to find a series expansion in powers of $t$ of $$I(t)=\int _{0}^{t}\exp(-x^2 \operatorname{erf}(x))dx=\sum_{p=1}^\infty c_p t^p.$$ The coefficients $c_p=p^{-1}d_{p-1}$ follow from the series expansion $e^{-x^2\,{\rm erf}\,x}=\sum_{p=0}^\infty d_p x^p$, resulting in $$I(t)=\sum_{p=1}^\infty c_p t^p=t-\frac{t^4}{2 \sqrt{\pi }}+\frac{t^6}{9 \sqrt{\pi }}+\frac{2 t^7}{7 \pi }-\frac{t^8}{40 \sqrt{\pi }}-\frac{4 t^9}{27 \pi }+\frac{(\pi -28) t^{10}}{210 \pi ^{3/2}}+O\left(t^{11}\right).$$ The series $I(1)=\sum_{p=1}^\infty c_p$ seems to converge:

Plot of $I_N=\sum_{p=0}^{N} c_p $ as a function of $N$ up to $N=25$. The value of $I_{25}=0.8162$ agrees with $I(1)=0.816377$ to three decimal places. For $N=50$ the agreement is up to six decimal places.

$\endgroup$
  • $\begingroup$ But is it possible to get it for n th term and it's seems that is a formel power series ? $\endgroup$ – zeraoulia rafik Jun 1 '18 at 13:05
  • $\begingroup$ The value of I(50) from wolfram alpha is : 0.972107 , wolframalpha.com/input/… $\endgroup$ – zeraoulia rafik Jun 1 '18 at 17:02
  • $\begingroup$ @zeraouliarafik -- when I say $I_{50}=0.816377$ I am referring to the sum $I_{N}=\sum_{p=0}^N c_p$, which is the series expansion of $I(t)$ for $t=1$ to order 50; what you have evaluated with Wolfram alpha is $I(t)$ for $t=50$, which is a different thing. $\endgroup$ – Carlo Beenakker Jun 1 '18 at 18:08
  • $\begingroup$ I'm confused how you calculated the coeffecients of series using error function ? $\endgroup$ – zeraoulia rafik Jun 2 '18 at 2:50

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.