# Estimate $\int^1_0 e^{-x^2}\, dx$

Estimate $\int^1_0 e^{-x^2}\, dx$

This is in a section on Taylor series so I would assume that is how it should be solved.

I started by using the Taylor series formula for $e^x$ replacing $x$ with $x^2$. I ended up with this:

$1+(-x^2)+\frac{(-x^2)^2}{2!}+\frac{(-x^2)^3}{3!}+\frac{(-x^2)^4}{4!}+...$

It looks to me like from here I should get some Fundamental Theorem of Calculus going, but I'm not sure how. The integral evaluated at $0$ is just $1$, however the integral evaluated at $1$ is not so easy. Is there some summation formula that would apply here that I could use?

• Have you tried integrating $1-x^2 + \frac{x^4}{2}-\frac{x^6}{6}$ between $0$ and $1$? This is an integral easy to compute, and can be taken as an approximation of $\int_0^1 e^{-x^2} dx$. (The more polynomial terms you get, the better the approximation will be) – Clement C. Apr 27 '15 at 2:10
• That's what I figure you have to do, I thought there would be a way to sum up all the terms though so it is a really good estimate. – Señor Sandia Apr 27 '15 at 2:13
• Given the function itself, it is equal to the sum of all the terms of the series, as $e^y = \sum_{n=0}^\infty \frac{y^n}{n!}$ (for all $y\in\mathbb{R}$). Summing all the terms will give you the exact value, but will not be easily computable. – Clement C. Apr 27 '15 at 2:15
• Ah, well since this is a homework question I guess I will just do a few values and maybe make a note about more values giving a better estimate. Thanks for the help! – Señor Sandia Apr 27 '15 at 2:25
• You are not evaluating the integral, you are evaluating the function itself. – Yves Daoust Apr 27 '15 at 9:48

The exponential function is an entire function, hence: $$e^x = \sum_{n\geq 0}\frac{x^n}{n!} \tag{1}$$ leads to: $$\forall x\in[0,1],\quad e^{-x^2}=\sum_{n\geq 0}\frac{(-1)^n}{n!}\,x^{2n}\tag{2}$$ and by integrating termwise $(2)$ over $[0,1]$ we get: $$\int_{0}^{1}e^{-x^2}\,dx = \sum_{n\geq 0}\frac{(-1)^n}{(2n+1)\,n!}\tag{3}$$ where the RHS of $(3)$ is a series that converges pretty fast; for instance: $$\left|\sum_{n\geq N}\frac{(-1)^n}{(2n+1)n!}\right|\leq\frac{1}{(2N+1)N!}\tag{4}$$ and: $$\int_{0}^{1}e^{-x^2}\,dx = \sum_{n=0}^{6}\frac{(-1)^n}{(2n+1)n!}+\theta = \frac{1614779}{2162160}+\theta = 0.7468360\ldots+\theta\tag{5}$$ where $|\theta|\leq\frac{1}{15\cdot 7!}$ gives: $$\int_{0}^{1}e^{-x^2}\,dx = \color{red}{0.7468\ldots}\tag{6}$$ Other (tight) approximations are possible by using continued fractions, since $$\int_{0}^{1}e^{-x^2}\,dx = \frac{\sqrt{\pi}}{2}\,\operatorname{Erf}(1)\approx\color{blue}{\frac{3969}{1955 e}}\tag{7}$$ where the Gauss continued fraction (look for "error function" here) provides good bounds.

$\bf{My\; Solution}$ We Know that in $$x\in (0,1)\;\;, x^2<x\Rightarrow -x^2>-x\Rightarrow e^{-x^2}>e^{-x}$$

So $$\displaystyle \int_{0}^{1}e^{-x^2}dx > \int_{0}^{1}e^{-x}dx = \left(1-\frac{1}{e}\right)$$

and in $$\displaystyle x\in (0,1)\;, x^2>0\Rightarrow e^{-x^2}<e^{-0}$$

So $$\displaystyle \int_{0}^{1}e^{-x^2}dx<\int_{0}^{1}e^{-0}dx = 1$$

So $$\displaystyle \left(1-\frac{1}{e}\right)< \int_{0}^{1}e^{-x^2}dx < 1$$

• That's an interesting way to think about it, but it seems like it leaves a large window for the result. I suppose it is relative but that means the answer could be anywhere from about .63 to 1. – Señor Sandia Apr 27 '15 at 3:35

Why not use a numerical method such as Simpson's, Trapezoidal or Midpoint?

Using Taylor or a Series also work as you see above.

A very crude option is to realise that you can calculate

$$\int_0^1e^{-ax}\,dx,$$

for a constant $a$. Therefore approximate $x^2$ by some $a$ on $[0,1]$. I found the minimum of

$$\int_0^1(x^2-ax)^2\,dx$$

at $a=\frac34$ and this yields

$$\int_0^1e^{-x^2}\,dx\approx\int_0^1e^{-\frac{3}{4}x}\,dx=\frac43\left(1-e^{-3/4}\right)$$

Use the standard normal distribution table.

Consider $$\frac{1}{\sqrt{2π}}\int e^{-x^2/2}\,dx$$

Let $u=\frac{x}{\sqrt{2}}$. Then $\frac{du}{dx}=\frac{1}{\sqrt{2}}$.

Then $$\frac{1}{\sqrt{2π}}\int e^{-x^2/2}\,dx=\frac{1}{\sqrt{2π}}\int e^{-u^2}\,\sqrt{2}du=\frac{1}{\sqrt{π}}\int e^{-u^2}\,du$$

So

$$\frac{1}{\sqrt{π}}\int_{0}^1 e^{-u^2}\,du=\frac{1}{\sqrt{2π}}\int_{0}^\sqrt{2} e^{-x^2/2}\,dx$$

$$\int_{0}^1 e^{-u^2}\,du=\sqrt{π}*\frac{1}{\sqrt{2π}}\int_{0}^\sqrt{2} e^{-x^2/2}\,dx\approx\sqrt{π}*0.4207\approx0.7457$$

By integrating term-wise the Taylor polynomial $$\sum_{k=0}^\infty\frac{(-1)^kx^{2k+1}}{(2k+1)k!}$$and evaluating in $(0,1)$: $$\color{green}{\sum_{k=0}^\infty\frac{(-1)^k}{(2k+1)k!}}.$$

Also, integrating by parts,

\begin{align} \int e^{-x^2}\,dx&=xe^{-x^2}+2\int x^2e^{-x^2}\,dx\\ &=xe^{-x^2}+\frac23x^3e^{-x^2}+\frac{2^2}3\int x^4e^{-x^2}\,dx\\ &=xe^{-x^2}+\frac23x^3e^{-x^2}+\frac{2^2}{3\cdot5}x^5e^{-x^2}+\frac{2^3}{3\cdot5}\int x^6e^{-x^2}\,dx\\ &=\cdots\\ &=\sum_{k=0}^\infty\frac{2^kx^{2k+1}}{(2k+1)!!}e^{-x^2}. \end{align} So in the range $(0,1)$,

$$\color{green}{\sum_{k=0}^\infty\frac{2^k}{(2k+1)!!}e^{-1}}.$$

Here are the $16$ first approximations, using both formulas \begin{align} & 0.666666666667 & 0.613132401952 \\ & 0.766666666667 & 0.711233586265 \\ & 0.742857142857 & 0.739262496068 \\ & 0.747486772487 & 0.745491142691 \\ & 0.746729196729 & 0.746623623896 \\ & 0.746836034336 & 0.746797851773 \\ & 0.746822806823 & 0.746821082157 \\ & 0.74682426574 & 0.746823815143 \\ & 0.746824120701 & 0.746824102826 \\ & 0.746824133824 & 0.746824130224 \\ & 0.746824132734 & 0.746824132607 \\ & 0.746824132818 & 0.746824132797 \\ & 0.746824132812 & 0.746824132811 \\ & 0.746824132812 & 0.746824132812 \\ & 0.746824132812 & 0.746824132812 \\ \end{align}

Simpson's rule requires to compute more costly terms \begin{align} \frac16\left(1+4e^{-1/4}+e^{-1}\right)&=0.747180\\ \frac1{12}\left(1+4e^{-1/16}+2e^{-1/4}+4e^{-9/4}+e^{-1}\right)&=0.746855\\ \frac1{18}\left(1+4e^{-1/36}+2e^{-1/9}+4e^{-1/4}+2e^{-4/9}+4e^{-25/36}+e^{-1}\right)&=0.746830 \end{align}