1970] A NOTE ON N!
65
Proof. { Log1o x } determines all of the digits of x and the integral part deter-
mines only the location of the decimal point. The terminating decimals are
dense in [1, 10] and log x = {logio x}, 1 <x<10 is continuous. Therefore
{ log,o x } is dense on [0, 1 ]. Since loga (x) = logb (x)loga b, the lemma follows for
any base.
LEMMA 2. {log N! } is dense on [0, 1].
Proof.
log(N + k)! = log N! + L log ( + N
= log N! + k log N + 1 log (I +
= logN!+ klogN + (N + 0 2
k k
= log N! + klogN + -N (k + 1) + 0
2N
We shall now construct an 1-net on [0, 1 ] with elements of { log NV
is small.
Let q >0. Since {log N} is dense on [0, 1], there is an infinite set of N, say
Na, such that 5'q/8 < { log Na } <7,q/8. Let l- 1 = [16/9,q], the greatest integer
in 16/9gq. Choose Na so large that
M(M + 1) + 4
2Naa_ 1
Then
{log(Na + k)!} = {log NQ! + k log NQ+k( + +- + (N-)} g(k).
2Na 2
Since Na is fixed, log Na,, and log Not! are fixed. For k
k(k+ 1) k )4 l
2a a 6
Therefore the principal part of the variation of g(k) as k varies is due to the
term k log Na. If g(k) < 1--, and since { log Na } < 7n/8, then
g(k + 1)-g(k) < {log N,, + 2 M2(M + ? (N2\
< 7-q/8 + 7/8 = -q.
If g(k) > 1 - (1/2),q, since { log(Na) } > (5/8)j,q it follows that g(k + 1) > (7/8)?
+(1/8)-7=n. Since M{Iog NJa>(l6/9n+l)(5/8)-7>(l6/9n)(5/8)n=lO/9, the
66 MATHEMATICS MAGAZINE [\4ar.-Apr.
whole unit interval is covered by an v net of points of { log N! } where the chosen
NareNaNa+1 . . . Na+M.
Proof of Theorem 1. If we take log to the base 10, we want there to exist an
integer t and an integer N such that AX1Ot<N!<(A+1)lOt or that t+log A
<log N! < t +log(A + 1). This can be done if
{log A} < flog NI} < {log(A + 1)}.
By Lemma 2 { log N! } is dense in [0, 1 ]. Thus, such an N can be chosen and the
theorem is proved.
Theorem 2 is a generalization of Lemma 2. We need the following definitions
and lemma.
Define log(j) m to be the jth iterant of log m, i.e., log(2) m =log log m.
Define 71 !) to be the jth iterant of n !, i.e., n 2n ! !
LEMMA 3. For j > 1,
\ log n!
Proof. By Stirling's formula,
log(2)(n + k)!(2) = log([(n + k) ! + 1/2] log(n + k)! - (n + k) ! log e + 0(1))
= log[(n + k)! log(n + k)!] [I + 0(lo( )
= log(n + k)! + log log(n + k)! + O(lg( k)
From the proof of Lemma 2, log(2) (n+k) !(2)
= log n! + k log n + log[log n! + k log n] + O(k2/n2)
k log n\
=logn!+loglogn!+klogn+O log n!
establishing a basis.
Induction hypothesis:
1k log n
log(j) (n + k)!(') = logn! + loglogn! + k logn + 0V- !
Then by Stirling's formula,
log (+l) (n + k) 1 (1+1)
= log(j)([(n + k) Stj) + 1/2] log(n + k) !(j) - (n + k) (j) log e + 0(1))
= log U)[(n + k) !i) 1 + 1/2 log(n + k) O ( 1
1970] A CATEGORICAL SYSTEM OF AXIOMS FOR THE COMPLEX NUMBERS 67
/ ~~~~log(n + k) !(j //log(n + k) !()2\
= log(rl) log(n + k) !(j) + 1/2 (n + k) ! + (n + k) !(i) ))
g [ g( +i) [ + /(n + 1k)=!(j) + ([(n
log(n + k) !+W]3]
+ k)!)\ +1
=log&'-) log(n + +)1j 1+120 I3
= log(U)(n + k) !U) ? 0Q +k) ! (D)
/k log n\
= log n! + log log n! + k log n + 01
log n!/
completing the proof.
From this, as in Lemma 2, one can construct an X net and prove:
THEOREM 2. { log(i) MU) } is dense on the unit interval.
Supported in part by NSF Grant NSF GP 5938.
A CATEGORICAL SYSTEM OF AXIOMS
FOR THE COMPLEX NUMBERS
W. BOSCH and P. KRAJKIEWICZ, University of Nebraska
In many elementary analysis texts the real numbers are introduced by means
of a categorical system of axioms. In contrast one finds in complex analysis texts
that the complex numbers are defined by means of some constructive process.
The purpose of this note is to remedy this omission by giving a brief exposition
of the complex numbers starting from a categorical system of axioms. In this
paper it is assumed that the elementary properties of a complete linearly ordered
field are known.
DEFINITION 1. A complex number system K is defined to be any commutative
field K with thefollowing property: There exists a homomorphism T: K-*K such that
(1) T(z) 5z z for at least one z E K,
(2) T(T(z)) = z for all z E K,
(3) the subset R = {z C K: T(z) = z} is a complete linearly ordered field.
The mapping T is called the conjugate operator. For any element zeK,
T(z) is called the conjugate of z and we denote it by s. The elements z in K are
called complex numbers.
THEOREM 1. For any complex number system K, the following properties of
the conjugate operator are valid:
(4) z 5 z, for at least one z C K,
(5) z=z, for all z E K,
