5

In a recent post (Possible bug in hypergeometric function AppellF1) I reported a buggy behaviour of the function AppellF1[].

Now I would like to aks about the implementation which possibly can explain this behaviour, and specifically the appearance of the octogonal region in (x,y) where the implementation of the function is well-behaved. (I'm asking also because I have not received an answer to that post).

There are two definitions for AppellF1[] which could be used for an implementation ( http://mathworld.wolfram.com/AppellHypergeometricFunction.html )

Series representation

$$F1(a,b1,b2,c,x,y) = \sum _{m=0}^{\infty } \sum _{n=0}^{\infty }\frac{ (a)_{m+n} }{(c)_{m+n}} (\text{b1})_m (\text{b2})_n \frac {x^m}{m!} \frac {y^n}{n!} $$

Sum[
 Pochhammer[a, m + n]/
  Pochhammer[c, n + m] Pochhammer[b1, m] Pochhammer[b2, n] x^m/m! y^n/
  n!, {m, 0, \[Infinity]}, {n, 0, \[Infinity]}]

(* Out[2843]= AppellF1[a, b1, b2, c, x, y] *)

Integral representation

$$F1(a,b1,b2,c,x,y) = \frac{\Gamma (c)}{\Gamma (a) \Gamma (c-a)} \int_0^1 t^{a-1} (1-t)^{-a+c-1} (1-t x)^{-\text{b1}} (1-t y)^{-\text{b2}} \, dt$$

fctF1[a_, b1_, b2_, c_, x_, y_] := 
 Gamma[c]/(Gamma[a] Gamma[c - a])
   Integrate[
   t^(a - 1) (1 - t)^(c - a - 1) (1 - x t)^-b1 (1 - t y)^-b2, {t, 0, 
    1}]
Community
  • 1
Dr. Wolfgang Hintze
  • 13,039
  • 17
  • 47

1 Answers1

7

It seems the whole code for AppellF1 is implemented in the Wolfram Language, and can be viewed using PrintDefinitions from the GeneralUtilities` package:

<< GeneralUtilities`
PrintDefinitions[AppellF1]

For numerical evaluation, the third definition is used, which calls System`AppellF1Dump`NHypergeometricF1

You can see the source code by following the link, or by running

PrintDefinitions[System`AppellF1Dump`NHypergeometricF1]

Both representations mentioned by OP are being used. But one can see, by looking at the pattern conditions, that the function is not implemented for all cases.

QuantumDot
  • 19,601
  • 7
  • 45
  • 121