A NEW FORMULA FOR THE NATURAL LOGARITHM OF A

NATURAL NUMBER
SHAHAR NEVO
Abstract. For every natural number T, we write Ln T as a series, generalizing the
known series for Ln 2.
1. Introduction
The Euler-Mascheroni constant , [1, p. 18], is given by the limit
(1) = lim
n
A
n
,
where for every n 1, A
n
:= 1 +
1
2
+ +
1
n
Ln n. An elementary way to show the
convergence of {A
n
}

n=1
is to consider the series

n=0
(A
n+1
A
n
). (Here A
0
:= 0.) Indeed,
by Lagranges Mean Value Theorem, there exists for every n 1 a number
n
, 0 <
n
< 1
such that
A
n+1
A
n
=
1
n + 1
Ln(n + 1) + Ln n =
1
n + 1

1
n +
n
=

n
1
(n + 1)(n +
n
)
,
and thus 0 > A
n+1
A
n
>
1
n(n+1)
and the series converges to some limit .
2. The new formula
Let T 2 be an integer. We have
(2) A
nT
=
n1

k=0
T

j=1
1
kT + j
Ln(nT)
n
.
By subtracting (1) from (2) and using Ln(nT) = Ln n + Ln T, we get
n1

k=0
_
T

j=1
1
kT + j

1
k + 1
_

n
Ln T,
that is,
(3) Ln T =

k=0
_
1
kT + 1
+
1
kT + 2
+ +
1
kT + (T 1)

(T 1)
kT + T
_
.
We observe that (3) generalizes the formula Ln 2 = 1
1
2
+
1
3

1
4
+ . . . .
2010 Mathematics Subject Classication. 26A09, 40A05, 40A30.
This research is part of the European Science Foundation Networking Programme HCAA..
1
2 SHAHAR NEVO
We can write (3) also as
(4) Ln T =
_
1 +
1
2
+
1
3
+ +
1
T
1
_
+
_
1
T + 1
+
1
T + 2
+ +
1
2T

1
2
_
+ . . .
and this gives Ln T as a rearrangement of the conditionally convergent series 11+
1
2

1
2
+
1
3

1
3
+. . . . The formula (4) holds also for T = 1. Formulas (3) and (4) can be applied also
to introduce Ln Q as a series for any positive rational Q =
M
L
since Ln
M
L
= Ln MLn L.
Now, for any k 0, the nominators of the k-th element in (3) are the same and their
sum is 0. This fact is not random. For every constant a
1
, a
2
, . . . a
T
, the sum
(5) S
T
(a
1
, . . . , a
T
) :=

k=0
_
a
1
kT + 1
+
a
2
kT + 2
+ +
a
T
(k + 1)T
_
converges if and only a
1
+ a
2
+ + a
T
= 0. This follows by comparison to the series

k=1
1
k
2
< . By (3) and the notation (5), Ln T = S
T
(1, 1, . . . , 1, T 1).
For T 2, let us denote by (T) the collection of all sums of rational series of type
(5), i.e.,
(T) =
_
S
T
(a
1
, . . . , a
T
) : a
i
Q, 1 i T, a
1
+ +a
T
= 0
_
.
The collection (T) is a linear space of real numbers over Q (or over the eld of algebraic
numbers if we would dene (T) to be with algebraic coecients instead of rational
coecients), and dim(T) T 1. A spanning set of T 1 elements of (T) is
_
S
T
(1, 1, 0, 0, . . . , 0), S
T
(0, 1, 1, 0, 0, . . . , 0), . . . , S
T
(0, . . . , 0, 1, 1)
_
.
Also, if T is not a prime number, then dim(T) < T 1. If Q =
M
L
is a positive
rational number and P
1
, P
2
, . . . , P
k
are all the prime factors of M and L together, then
Ln Q (P
1
P
2
. . . P
k
).
We can get a non-trivial series for x = 0: Ln 4 = 2 Ln 2 = S
2
(2, 2) = S
4
(2, 2, 2, 2),
and also Ln(4) = S
4
(1, 1, 1, 3). Hence
0 = S
4
(2, 2, 2, 2) S
4
(1, 1, 1 3) = S
4
(1, 3, 1, 1)
=
_
1
1

3
2
+
1
3
+
1
4
_
+
_
1
5

3
6
+
1
7
+
1
8
_
+. . . .
THE NATURAL LOGARITHM OF A NATURAL NUMBER 3
3. The integral approach
The formula (3) can as well be deduced in the following way.
Ln T = lim
x1

Ln(1 + x + + x
T1
) = lim
x1

Ln
_
1 x
T
1 x
_
= lim
x1

(Ln(1 x
T
) Ln(1 x)) = lim
x1

_
x
0
_
Tu
T1
u
T
1
+
1
1 u
_
du
= lim
x1

_
x
0
Tu
T1
(1 + u + + u
T1
)
u
T
1
du
= lim
x1

_
x
0
1 u u
2
u
T2
+ (T 1)u
T1
u
T
1
du
= lim
x1

_
1
_
x
0
u 1
u
T
1
du + 2
_
x
0
u
2
u
u
T
1
du + 3
_
x
0
u
3
u
2
u
T
1
du + . . .
+ (T 2)
_
x
0
u
T2
u
T3
u
T
1
du + (T 1)
_
x
0
u
T1
u
T2
u
T
1
du
_
. (6)
For every 1 j T 1,
lim
x1

_
x
0
u
j
u
j1
u
T
1
du = lim
x1

_
x
0
_
u
j1

k=0
u
kT
u
j

k=0
u
kT
_
du
= lim
x1

_
x
0
_

k=0
u
kT+j1

k=0
u
kT+j
_
du = lim
x1

k=0
_
x
kT+j
kT +j

x
kT+j+1
kT + j + 1
_
.
The series in the last expression converges at x = 1, and thus it denes a continuous
function in [0, 1] and so the limit is
(7)
_
1
0
u
j
u
j1
u
T
1
du =

k=0
_
1
kT + j

1
kT + j + 1
_
= S
T
(0, . . . , 0,
j

1, 1, 0, . . . , 0).
By (6), we now get that
Ln T =

k=0
_
1
kT + 1

1
kT + 2
_
+ 2

k=0
_
1
kT + 2

1
kT + 3
_
+. . .
+ (T 2)

k=0
_
1
kT + T 1

1
kT + T 1
_
+ (T 1)

k=0
_
1
kT + T 1

1
(k + 1)T
_
=

k=0
_
1
kT + 1
+
1
kT + 2
+ +
1
kT + T 1

(T 1)
(k + 1)T
_
,
and this is formula (3).
If we put T = 3, j = 1 into (7), we get that
(8)
_
1
0
u 1
u
3
1
du =

k=0
_
1
3k + 1

1
3k + 2
_
= S
3
(1, 1, 0).
4 SHAHAR NEVO
On the other hand,
_
u 1
u
3
1
du =
2

3
arctan
_
2u + 1

3
_
,
and together with (8), this gives
2

3
_
arctan

3 arctan
1

3
_
= S
3
(1, 1, 0)
or
= 3

3 S
3
(1, 1, 0) = 3

3
__
1
1

1
2
_
+
_
1
4

1
5
_
+
_
1
7

1
8
_
+ . . .
_
.
This formula can also be deduced from Eulers formula [1, pp. 109,283].
We can also arrive at a formula of as follows:
arctan x =
_
dx
x
2
+ 1
=
_
x
2
1
x
4
1
dx =
_
x
2
x
x
4
1
dx +
_
x 1
x
4
1
dx.
Thus

4
=
_
1
0
arctan xdx =
_
1
0
x
2
x
x
4
1
dx +
_
1
0
x 1
x
4
1
dx
(7)
= S
4
(0, 1, 1, 0) + S
4
(1, 1, 0, 0) = S
4
(1, 0, 1, 0).
We then get the well-known formula
= 4
_
1
1
3
+
1
5

1
7
+ . . .
_
.
By exploiting similarly suitable integrals with denominator x
T
1 for larger values of T, we
can nd corresponding formulas for as a member of (T) (with algebraic coecients).
It is an interesting question whether belongs to (2). Since dim(2) = 1, it occurs if
and only if

Ln2
is an algebraic number.
References
1. P. Eymard and J.-P. Lafon, The Number , Amer. Math. Soc., Providence, RI, 2004.
Bar-Ilan University, Department of Mathematics, Ramat-Gan 52900, Israel
E-mail address: nevosh@macs.biu.ac.il