26

Is there any easy proof that $\int x^x dx$ can't be done in terms of elementary functions? I know this is true, because the Risch algorithm gives a decision process for integration in terms of elementary functions, Axiom provides a complete software implementation of the Risch algorithm, and Axiom can't do the integral. However, it would be nicer to have a human-readable proof. If it could be reduced to a standard special function such as a hypergeometric function, then we could reduce the proof to a proof that that function can't be expressed in terms of elementary functions. But neither Axiom nor Wolfram Alpha can reduce it to any other form.

  • 1
    Related: http://math.stackexchange.com/questions/21330/closed-form-for-sum-frac1nn – JavaMan Feb 10 '12 at 18:16
  • 2
    It's not difficult to manually apply the Liouville-Risch algorithm to this special case. For the few pages of theory needed see Rosenlicht's 1972 Monthly exposition Integration in finite terms. See also this answer. – Math Gems Feb 10 '12 at 20:03
  • Thank you ben crowell, for giving me a chance to answer. – IDOK Feb 11 '12 at 17:39
  • @MathGems : Even though your links are good, but I think that we need to use the consequence of the actual theorem instead of heading towards it directly ( for the sake of ease ). – IDOK Feb 11 '12 at 17:50
  • 4
    @Ben The accepted answer is not correct, nor can it be fixed - see my comment there. – Math Gems Feb 11 '12 at 18:11
  • This page gives infinite representation for your integral that uses incomplete Gamma functions: http://en.wikipedia.org/wiki/List_of_integrals_of_exponential_functions – Anixx Feb 28 '12 at 02:26

1 Answers1

7

See this sci.math article from 1993: http://groups.google.com/group/sci.math/msg/6b68b00362baf65f

Robert Israel
  • 448,999