I believe that (i.e., it would be convenient if, and visually appears that) the product of the two complementary error functions:


will have a solution, or can be approximated with a solution, of a Gaussian form (i.e., $c\operatorname{exp}\left[-\frac{x^2}{2d^2}\right]$, where $c$ and $d$ are functions of $a$ and $b$) when $a>0$ and $c>0$, however I cannot find a proof of this. Any help?

enter image description here

  • 2
    $\begingroup$ What do you mean by a Gaussian solution? $\endgroup$ – Lord Soth Oct 31 '13 at 23:12
  • 2
    $\begingroup$ "Compliment" (with an "i") and "complement" (with an "e") are two different words that mean two different things. (I changed the spelling in the question.) $\endgroup$ – Michael Hardy Oct 31 '13 at 23:15
  • $\begingroup$ @LordSoth I mean of the form $c \exp\left[-\frac{x^2}{2d^2}\right]$ where $c$ and $d$ are functions of $a$ and $b$. (edited the question for clarity) $\endgroup$ – mrsoltys Oct 31 '13 at 23:22
  • $\begingroup$ Compute. You will get a familiar integral of the erfc type, $\endgroup$ – André Nicolas Oct 31 '13 at 23:31

In this paper, the authors provide the following simple approximation of the error function:

$$\operatorname{erf}(z) = 1- \exp\{-c_1z-c_2z^2\},\; z\ge 0 $$


$$ c_1 = 1.09500814703333,\;\; c_2 = 0.75651138383854$$

Set $\frac{a-x}{b} \equiv z_1$, and $\frac{a+x}{b} =\equiv z_2$. Given also that $\operatorname{erfc} = 1- \operatorname{erf}$, and under the implied restrictions so that $\frac{a-x}{b}\ge 0,\;\frac{a+x}{b}\ge 0$, (for example, for $b>0,\; a>0,\; x\le |a|$) we have, using the approximation,

$$\operatorname{erfc}(z_1)\operatorname{erfc}(z_2) = [1-\operatorname{erf}(z_1)][1-\operatorname{erf}(z_1)] $$

$$\approx [1-1+ \exp\{-c_1z_1-c_2z_2^2\}][1-1+ \exp\{-c_1z_2-c_2z_2^2\}]$$

$$=\exp\{-c_1(z_1+z_2)-c_2(z_1^2+z_2^2)\} \\= \exp\left\{-c_1\left(\frac{a-x}{b}+\frac{a+x}{b}\right)-c_2\left[\left(\frac{a-x}{b}\right)^2+\left(\frac{a+x}{b}\right)^2\right]\right\}$$

$$ =\exp\left\{-\frac{2a}{b}c_1-c_2\left[\left(\frac{a-x}{b}+\frac{a+x}{b}\right)^2-2\frac{a-x}{b}\frac{a+x}{b}\right]\right\} $$

$$=\exp\left\{-\frac{2a}{b}c_1-c_2\left[4\left(\frac{a}{b}\right)^2-2\frac{a^2-x^2}{b^2}\right]\right\} $$

$$=\exp\left\{-\frac{2a}{b}c_1-c_2\left[2\left(\frac{a}{b}\right)^2+2\frac{x^2}{b^2}\right]\right\} $$

$$ = \exp\left\{-\frac{2a}{b}c_1-2c_2\left(\frac{a}{b}\right)^2\right\}\cdot\exp\left\{-2c_2\frac{x^2}{b^2}\right\}$$

$$=C\cdot\exp\left\{-\frac{x^2}{2d^2}\right\} $$

with $$ C\equiv \exp\left\{-\frac{2a}{b}c_1-2c_2\left(\frac{a}{b}\right)^2\right\},\;\; d^2 = \frac {b^2}{4c_2}$$

| cite | improve this answer | |

Let $x = a$. Then, the product becomes $\mathrm{erfc}(2a/b)$. If the product were equal to an exponential function, that would mean we could express $\mathrm{erfc}$ in terms of elementary functions, which, as far as I know, is impossible.

| cite | improve this answer | |

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.