[net.math] Infinite exponentiation

bobl (04/29/83)

To solve the problem

		...
	       x
              x
             x      = 2
				 ...
				x		
note that the exponent of x is x     itself, which we claim is 2,
so the equation becomes

	 2
	x  = 2

so x = (+ or -) sqrt(2).  (High school math WAS useful for something!)


				- Bob Lewis
				  ...tektronix!iddic!tekgds!bobl

ucbcad:moore (04/30/83)

#R:ucbvax:-48000:ucbcad:11900001:000:717
ucbcad!moore    Apr 29 02:55:00 1983

			...
			x
		       x
		      x = 2. 

	What is the value of x? 


since I don't have a picture mode editor, I will restate this as

x**x**x**x**..... = 2.

if we assume that x**x**x = x**(x**x) (right to left associative)

then x**x**x... = x**(x**x**x...) = x**2 => x**2 = 2. => x = sqrt(2).

    We have implicitly assumed that an answer exists here, in that we
are manipulating x**x**x... as though it was a well-behaved quantity.
As an example of the pitfalls, the problem x**x**x.. = 4 has the same
solution, i.e. x = sqrt(2)! Does anyone know what the largest n is
such that x**x**x.. = n has a solution? I suspect e, but have yet
to prove it.

    Peter Moore
    moore@Berkeley
    ...!ucbvax!ucbcad!moore 

halle1 (05/11/83)

x**x**x**...=n has a solution for all n >= 0.  (n=0 is an asymptotic
solution, i.e. lim n->0)
>From the earlier article, x**x**x**x...=x**(x**x**...)=n  ==> x**n=n.

Taking logs:  n ln(x)= ln(n)  or  ln(x)= ln(n) /n
	Therefore: x=exp( ln(n) /n) = (exp( ln(n))**(1/n)
	=>    x=n**(1/n)

                1/n
	   x = n      is defined for all n>o, equals 0 at n=0, and is
undefined (in the real plane) for n<0.  (Special cases excepted.)

The maximum value of x occurs at n=e (2.71828182859....)  x=1.4447  (rounded).