This post explores the value of two general infinite series I was curious
about:
_
i =0
o
(i +1)
u
x
i
and _
i =1
o i
u
(i ÷1)!
, x Ε (÷1, 1).
These are the infinite series we're used to geometric and "reciprocal of factorial" only with polyno
mials multiplying them. For u = 1, the first sum is just a geometric series: as u increases, we're multi
plying the geometric series terms by squares, cubes, and so on: u = 2 gives 1 +4 x +9 x
2
+..., u = 3
gives 1 +8 x +27 x
2
+64 x
3
+..., etc. The second sum only has one variable of freedom the expo
nent u whereas the first has two: the exponent on the "constant" term, and the geometric series
constant. (The first sum has an (i +1)
u
instead of an i because otherwise the constant term is 0,
which we don't want; the second sum would have had an i ! in the denominator instead of (i ÷1)!,
but then the i ' s cancel out and we have an identical sum, except for the u ÷1, which we can just
replace by u.) We begin by finding the value of the first sum in general: our method is observed
best by considering some example cases.
1 +x +x
2
+x
3
+... +x
n÷1
= S
n,0
; here we denote S
n,u
= _
i =1
n
i
u
x
i ÷1
. What is the value of this
series? We first write
x S
n,0
= x +x
2
+x
3
+... +x
n
so that
(1 ÷x ) S
n,0
= 1 ÷x
n
, and finally that
S
n,0
=
1÷x
n
1÷x
, or as n ÷o, S
o,0
=
1
1÷x
, the well known value of this series.
S
n,1
= 1 +2 x +3 x
2
+4 x
3
+... +n x
n÷1
; the same procedure gives
x S
n,1
= x +2 x
2
+3 x
3
+4 x
4
+... +(n ÷1) x
n÷1
+n x
n
, and
(1 ÷x ) S
n,1
= 1 +x +x
2
+x
3
+... +(n ÷(n ÷1)) x
n÷1
÷n x
n
; dropping this last term (the (n +1) x
n+1
)
as n ÷o, we have that
S
n,1
=
1
1÷x

1
1÷x
=
1
(1÷x )
2
, so that 1 +
2
2
+
3
4
+
4
8
+
5
16
+
6
32
+... =
1
1l4
= 4, or
1 ÷2 +3 ÷4 +5 ÷6 +... =
1
4
= 1 +1 +1 +1 +... =
(÷1) +(÷1) +(÷1) +... = 2 ÷5 +8 ÷11 +14 ÷17 +... = (÷3) +(÷3) +(÷3) +...
The general method for this doesn't really become apparent until u = 2 :
S
n,2
= 1 +4 x +9 x
2
+16 x
3
+... +n
2
x
n÷1
x S
n,2
= x +4 x
2
+9 x
3
+16 x
4
+... +(n ÷1)
2
x
n÷1
+n
2
x
n
(1 ÷x ) S
n,2
=
1 +3 x +5 x
2
+7 x
3
+... +n
2
÷(n ÷1)
2
] x
n÷1
÷n
2
x
n
= 1 +3 x +... +(2 n ÷1) x
n
÷n
2
x
n
.
Here, we notice that this sum is really equal to 2 S
n,1
÷S
n,0
as n gets large (dropping the last term):
you can check for yourself that the terms do sum properly, but it is clear from the n
th
term that this
is so. Thus, we have that
(1 ÷x ) S
o,2
= 2 S
o,1
÷S
o,0
=
2
(1÷x )
2
÷
1
1÷x
=
2+x ÷1
(1÷x )
2
=S
o,2
=
1+x
(1÷x )
3
; so, for example, we have that
1 +
4
2
+
9
2
+
16
2 2
+
25
4
+
36
4 2
+... =
1+1¡ 2
1÷1¡ 2 ]
3
= 2 17 +12 2 ] = 2 1 + 2 ]
4
.
Generally, we have
S
n,u
= 1 +2
u
x +3
u
x
2
+4
u
x
3
+... +n
u
x
n÷1
x S
n,u
= x +2
u
x
2
+3
u
x
3
+... +(n ÷1)
u
x
n
÷n
u
x
n
so
that
(1 ÷x ) S
n,u
= 1 +(2
u
÷1) x +(3
u
÷1) x
2
+... +(n
u
÷(n ÷1)
u
) x
n÷1
÷n x
n
. Combined with the bino
mial theorm (applied to (n +1)
u
÷n
u
) gives us the desired recursive definition of S
o,u
:
S
o,u
= S
o,0
_
k =1
u

u
k
 (÷1)
k
S
o,u÷k
.
This is entered below, where the function S is the value of this series at the given values of
u and x : it also varies n, but we can look at n ÷oif we want (it is given after the string of values in
the plot below). SExp is also given: it is the explicit value of the infinite sum as a function of x and
u.
This post explores the value of two general infinite series I was curious
about:
_
i =0
o
(i +1)
u
x
i
and _
i =1
o i
u
(i ÷1)!
, x Ε (÷1, 1).
These are the infinite series we're used to geometric and "reciprocal of factorial" only with polyno
mials multiplying them. For u = 1, the first sum is just a geometric series: as u increases, we're multi
plying the geometric series terms by squares, cubes, and so on: u = 2 gives 1 +4 x +9 x
2
+..., u = 3
gives 1 +8 x +27 x
2
+64 x
3
+..., etc. The second sum only has one variable of freedom the expo
nent u whereas the first has two: the exponent on the "constant" term, and the geometric series
constant. (The first sum has an (i +1)
u
instead of an i because otherwise the constant term is 0,
which we don't want; the second sum would have had an i ! in the denominator instead of (i ÷1)!,
but then the i ' s cancel out and we have an identical sum, except for the u ÷1, which we can just
replace by u.) We begin by finding the value of the first sum in general: our method is observed
best by considering some example cases.
1 +x +x
2
+x
3
+... +x
n÷1
= S
n,0
; here we denote S
n,u
= _
i =1
n
i
u
x
i ÷1
. What is the value of this
series? We first write
x S
n,0
= x +x
2
+x
3
+... +x
n
so that
(1 ÷x ) S
n,0
= 1 ÷x
n
, and finally that
S
n,0
=
1÷x
n
1÷x
, or as n ÷o, S
o,0
=
1
1÷x
, the well known value of this series.
S
n,1
= 1 +2 x +3 x
2
+4 x
3
+... +n x
n÷1
; the same procedure gives
x S
n,1
= x +2 x
2
+3 x
3
+4 x
4
+... +(n ÷1) x
n÷1
+n x
n
, and
(1 ÷x ) S
n,1
= 1 +x +x
2
+x
3
+... +(n ÷(n ÷1)) x
n÷1
÷n x
n
; dropping this last term (the (n +1) x
n+1
)
as n ÷o, we have that
S
n,1
=
1
1÷x

1
1÷x
=
1
(1÷x )
2
, so that 1 +
2
2
+
3
4
+
4
8
+
5
16
+
6
32
+... =
1
1l4
= 4, or
1 ÷2 +3 ÷4 +5 ÷6 +... =
1
4
= 1 +1 +1 +1 +... =
(÷1) +(÷1) +(÷1) +... = 2 ÷5 +8 ÷11 +14 ÷17 +... = (÷3) +(÷3) +(÷3) +...
The general method for this doesn't really become apparent until u = 2 :
S
n,2
= 1 +4 x +9 x
2
+16 x
3
+... +n
2
x
n÷1
x S
n,2
= x +4 x
2
+9 x
3
+16 x
4
+... +(n ÷1)
2
x
n÷1
+n
2
x
n
(1 ÷x ) S
n,2
=
1 +3 x +5 x
2
+7 x
3
+... +n
2
÷(n ÷1)
2
] x
n÷1
÷n
2
x
n
= 1 +3 x +... +(2 n ÷1) x
n
÷n
2
x
n
.
Here, we notice that this sum is really equal to 2 S
n,1
÷S
n,0
as n gets large (dropping the last term):
you can check for yourself that the terms do sum properly, but it is clear from the n
th
term that this
is so. Thus, we have that
(1 ÷x ) S
o,2
= 2 S
o,1
÷S
o,0
=
2
(1÷x )
2
÷
1
1÷x
=
2+x ÷1
(1÷x )
2
=S
o,2
=
1+x
(1÷x )
3
; so, for example, we have that
1 +
4
2
+
9
2
+
16
2 2
+
25
4
+
36
4 2
+... =
1+1¡ 2
1÷1¡ 2 ]
3
= 2 17 +12 2 ] = 2 1 + 2 ]
4
.
Generally, we have
S
n,u
= 1 +2
u
x +3
u
x
2
+4
u
x
3
+... +n
u
x
n÷1
x S
n,u
= x +2
u
x
2
+3
u
x
3
+... +(n ÷1)
u
x
n
÷n
u
x
n
so
that
(1 ÷x ) S
n,u
= 1 +(2
u
÷1) x +(3
u
÷1) x
2
+... +(n
u
÷(n ÷1)
u
) x
n÷1
÷n x
n
. Combined with the bino
mial theorm (applied to (n +1)
u
÷n
u
) gives us the desired recursive definition of S
o,u
:
S
o,u
= S
o,0
_
k =1
u

u
k
 (÷1)
k
S
o,u÷k
.
This is entered below, where the function S is the value of this series at the given values of
u and x : it also varies n, but we can look at n ÷oif we want (it is given after the string of values in
the plot below). SExp is also given: it is the explicit value of the infinite sum as a function of x and
u.
2 Im port ant Infinit e S um s . nb
This post explores the value of two general infinite series I was curious
about:
_
i =0
o
(i +1)
u
x
i
and _
i =1
o i
u
(i ÷1)!
, x Ε (÷1, 1).
These are the infinite series we're used to geometric and "reciprocal of factorial" only with polyno
mials multiplying them. For u = 1, the first sum is just a geometric series: as u increases, we're multi
plying the geometric series terms by squares, cubes, and so on: u = 2 gives 1 +4 x +9 x
2
+..., u = 3
gives 1 +8 x +27 x
2
+64 x
3
+..., etc. The second sum only has one variable of freedom the expo
nent u whereas the first has two: the exponent on the "constant" term, and the geometric series
constant. (The first sum has an (i +1)
u
instead of an i because otherwise the constant term is 0,
which we don't want; the second sum would have had an i ! in the denominator instead of (i ÷1)!,
but then the i ' s cancel out and we have an identical sum, except for the u ÷1, which we can just
replace by u.) We begin by finding the value of the first sum in general: our method is observed
best by considering some example cases.
1 +x +x
2
+x
3
+... +x
n÷1
= S
n,0
; here we denote S
n,u
= _
i =1
n
i
u
x
i ÷1
. What is the value of this
series? We first write
x S
n,0
= x +x
2
+x
3
+... +x
n
so that
(1 ÷x ) S
n,0
= 1 ÷x
n
, and finally that
S
n,0
=
1÷x
n
1÷x
, or as n ÷o, S
o,0
=
1
1÷x
, the well known value of this series.
S
n,1
= 1 +2 x +3 x
2
+4 x
3
+... +n x
n÷1
; the same procedure gives
x S
n,1
= x +2 x
2
+3 x
3
+4 x
4
+... +(n ÷1) x
n÷1
+n x
n
, and
(1 ÷x ) S
n,1
= 1 +x +x
2
+x
3
+... +(n ÷(n ÷1)) x
n÷1
÷n x
n
; dropping this last term (the (n +1) x
n+1
)
as n ÷o, we have that
S
n,1
=
1
1÷x

1
1÷x
=
1
(1÷x )
2
, so that 1 +
2
2
+
3
4
+
4
8
+
5
16
+
6
32
+... =
1
1l4
= 4, or
1 ÷2 +3 ÷4 +5 ÷6 +... =
1
4
= 1 +1 +1 +1 +... =
(÷1) +(÷1) +(÷1) +... = 2 ÷5 +8 ÷11 +14 ÷17 +... = (÷3) +(÷3) +(÷3) +...
The general method for this doesn't really become apparent until u = 2 :
S
n,2
= 1 +4 x +9 x
2
+16 x
3
+... +n
2
x
n÷1
x S
n,2
= x +4 x
2
+9 x
3
+16 x
4
+... +(n ÷1)
2
x
n÷1
+n
2
x
n
(1 ÷x ) S
n,2
=
1 +3 x +5 x
2
+7 x
3
+... +n
2
÷(n ÷1)
2
] x
n÷1
÷n
2
x
n
= 1 +3 x +... +(2 n ÷1) x
n
÷n
2
x
n
.
Here, we notice that this sum is really equal to 2 S
n,1
÷S
n,0
as n gets large (dropping the last term):
you can check for yourself that the terms do sum properly, but it is clear from the n
th
term that this
is so. Thus, we have that
(1 ÷x ) S
o,2
= 2 S
o,1
÷S
o,0
=
2
(1÷x )
2
÷
1
1÷x
=
2+x ÷1
(1÷x )
2
=S
o,2
=
1+x
(1÷x )
3
; so, for example, we have that
1 +
4
2
+
9
2
+
16
2 2
+
25
4
+
36
4 2
+... =
1+1¡ 2
1÷1¡ 2 ]
3
= 2 17 +12 2 ] = 2 1 + 2 ]
4
.
Generally, we have
S
n,u
= 1 +2
u
x +3
u
x
2
+4
u
x
3
+... +n
u
x
n÷1
x S
n,u
= x +2
u
x
2
+3
u
x
3
+... +(n ÷1)
u
x
n
÷n
u
x
n
so
that
(1 ÷x ) S
n,u
= 1 +(2
u
÷1) x +(3
u
÷1) x
2
+... +(n
u
÷(n ÷1)
u
) x
n÷1
÷n x
n
. Combined with the bino
mial theorm (applied to (n +1)
u
÷n
u
) gives us the desired recursive definition of S
o,u
:
S
o,u
= S
o,0
_
k =1
u

u
k
 (÷1)
k
S
o,u÷k
.
This is entered below, where the function S is the value of this series at the given values of
u and x : it also varies n, but we can look at n ÷oif we want (it is given after the string of values in
the plot below). SExp is also given: it is the explicit value of the infinite sum as a function of x and
u.
In[1]:= Sn_, u_, x_ : Sn, 0, x
k1
u
Binomialu, k 1^1 k Sn, u k, x
Sn_, 0, x_ : 1 x^n 1 1 x
ManipulateTableSn, u, x, n, 1, 30, LimitSn, u, x, n Infinity,
ListPlotTableSn, u, x, n, 1, 30, Joined True, x, .99, .99, .011, u, 0, 5, 1
x
u
0.010944, 0.631002, 0.0260307, 0.412886, 0.0349178, 0.279931,
0.0369832, 0.196392, 0.0342359, 0.142353, 0.0288244, 0.106432,
0.02235, 0.0819557, 0.0158206, 0.0649045, 0.00978738, 0.0527955,
0.00449804, 0.0440536, 0.0000171858, 0.037655, 0.00369034, 0.0329179,
0.00670707, 0.029378, 0.00913201, 0.0267127, 0.0110638, 0.0246939,
0.0180596,
5 10 15 20 25 30
0.05
0.10
0.15
0.20
0.25
0.30
Im port ant Infinit e S um s . nb 3
In[3]:= SExp0
1
1 x
SExpu_ : SExp0
k1
u
Binomialu, k 1^1 k SExpu k
TableSimplifySExpu, u, 0, 8
Out[3]=
1
1 x
Out[5]=
1
1 x
,
1
1 x
2
,
1 x
1 x
3
,
1 4 x x
2
1 x
4
,
1 11 x 11 x
2
x
3
1 x
5
,
1 26 x 66 x
2
26 x
3
x
4
1 x
6
,
1 57 x 302 x
2
302 x
3
57 x
4
x
5
1 x
7
,
1 120 x 1191 x
2
2416 x
3
1191 x
4
120 x
5
x
6
1 x
8
,
1 247 x 4293 x
2
15 619 x
3
15 619 x
4
4293 x
5
247 x
6
x
7
1 x
9
We might also realize that this method gives us the capability to find the value of the even more
general series
_
i =0
o
P(i ) x
i
, where P(i ) is a polynomial in i and x is constant: since we know the value of the
S
o,u
' s, we need only add their values with the necessary constants. For example, if
P(i ) = 17 i
3
÷289 i +4, then the value of the sum is simply
_
i =0
o
(17 i
3
÷289 i +4913) x
i
= 17 S
o,3
÷289 S
o,1
+4913 S
o,0
, which is simply a function of x .
Next, we attempt to find the general value of the sum
_
i =1
o i
u
(i ÷1)!
for any given nonnegative integer u. Our approach is very similar: we first consider
the case of u = 0,
E
0
=
1
0!
+
1
1!
+
1
2!
+... which has the value :, as we all know. A short proof is given to convince the
unsure reader (and will be our base case later):
The limit definiton of : is
Lim
n÷o
1 +
1
n
]
n
= Lim
n÷o

1
0!
+
n
1!

1
n
+
n(n÷1)
2!

1
n
2
+
n(n÷1) (n÷2)
3!

1
n
3
+...] =
1
0!
+
1
1!
+
1
2!
+... = :
E
1
=
1
0!
+
2
1!
+
3
2!
+
4
3!
+...
E
1
÷E
0
=
0
0!
+
1
1!
+
2
2!
+
3
3!
+
4
4!
+... = 0 +
1
0!
+
1
1!
+
1
2!
+... = E
0
= E
1
= 2 E
0
= 2 :. This result also
gives the rather interesting conclusion that
2
0!
+
2
1!
+
2
2!
+... =
1
0!
+
2
1!
+
3
2!
+... =0 =
÷1
0!
+
0
1!
+
1
2!
+
2
3!
+... or 1 =
1
2!
+
2
3!
+
3
4!
+...
When trying to determine this series' value, I computed E
2
by the following
method:
E
2
=
1
0!
+
4
1!
+
9
2!
+
16
3!
+
25
4!
+...
E
2
÷4 E
0
=
÷3
0!
+
0
1!
+
5
2!
+
12
3!
+... = ÷3 +3 
1
2!
+
1
3!
+
1
4!
+...] +
2
2!
+
6
3!
+
12
4!
+...] =
÷3 +3 +
1
0!
+
1
1!
+
1
2!
+... = E
0
= E
2
= 5 E
0
= 5 :
. And I thought
that was very clever. However, it turns out that the general method I determined is far easier to
work out, as doing something similar to this computation for E
3
became extremely difficult I
didn't figure out a method similar to the above one, but not for lack of trying.
The general method I determined works inductively (a method I've now fallen in love with):
S'pose we are trying to calculate E
k
. Then we first assume to know the values of
E
0
, E
1
, E
2
, ... E
k ÷1
. Consider the i
th
term of the series E
k
:
1
0!
+
2
k
1!
+
3
k
2!
+... +
i
k
(i ÷1)!
(the i
th
term simply being the quotient
i
k
(i ÷1)!
). We are trying to subtract
some number of other E
n
' s, n < k , perhaps with constants, from this i
th
term to obtain E
0
. We
notice (by way of pencil and some wasted paper) that should we be able to subtract whatever terms
we need to obtain the factorization (i ÷1) (i ÷2) (i ÷3) ... (i ÷k ) in the i
th
term, then we will have
that
E
k
÷Stuff = _
i =k +1
o 1
(i ÷k ÷1)
= E
0
.
Thus, we consider the expansion of that product, (which is known as a "falling factorial"), which
clearly has an i
k
term, an i
k ÷1
term, and so on down to a constant term. Remember that we are
dealing with an infinite sum thus, each of these terms represents an infinite sum of their own: we
quickly realize that each i
x
term, (which will have some constant C
x
attached to it), in summation
represents E
x
. As a result, we know that E
k
is equal to E
0
plus some other E
x
' s with constants,
and we have the value of E
k
(since we've assumed to know the values of all the E
x
' s. An example
is extremely helpful to see what I'm talking about here:
E
3
= _
i =1
o i
3
(i ÷1)!
, and we can evaluate
(i ÷1) (i ÷2) (i ÷3) = (i
2
÷3 i +2) (i ÷3) = i
3
÷6 i
2
+11 i ÷6, so we need to subtract 6 E
2
, add
11 E
1
, and subtract 6 E
0
from E
3
: clearly the terms i = 1, i = 2 and i = 3 will all become 0 since this
polynomial is 0 at those points, so we initiate the summation at i = 4.
E
3
÷6 E
2
+11 E
1
÷6 E
0
=
4
3
÷64
2
+114÷6
3!
+
5
3
÷65
2
+115÷6
4!
+
6
3
÷66
2
+116÷6
5!
+... =
1
0!
+
1
1!
+
1
2!
+
1
3!
+... = E
0
so
that
E
3
= 6 E
2
÷11 E
1
+7 E
0
= 30 : ÷22 : +7 : = 15 :.
Clearly, this process can be repeated indefinately. What is still a bit annoying is that there isn't an
explicit formula for the coefficients of the k
th
"falling factorial." Apparently they are related to
Stirling numbers of the second kind: as well, it turns out that there are a series of numbers called
the Bell numbers which are exactly the coefficients of the E
k
' s, and they are closely related to
Stirling numbers. The Bell numbers have some interesting relations (a number of which I rediscov
ered above), but their definition comes from set theory: the n
th
Bell number is how many ways the
set of integers 1 through n can be arranged into subsets (using all the numbers) independent of
order. I shall have to investigate the pattern further.
The other series related to these that I haven't quite figured out yet is closely related to the second
one:
_
i =1
o u
i
(i ÷1)!
, where the numerator has a geometric series and the denominator has the factorials.
It is clear that this series converges (as well as the convergence of the above two series), and proofs
will be left to the reader. This one turns out to have the value :
u
, but so far I can't see an easy
method of proving that.
4 Im port ant Infinit e S um s . nb
We might also realize that this method gives us the capability to find the value of the even more
general series
_
i =0
o
P(i ) x
i
, where P(i ) is a polynomial in i and x is constant: since we know the value of the
S
o,u
' s, we need only add their values with the necessary constants. For example, if
P(i ) = 17 i
3
÷289 i +4, then the value of the sum is simply
_
i =0
o
(17 i
3
÷289 i +4913) x
i
= 17 S
o,3
÷289 S
o,1
+4913 S
o,0
, which is simply a function of x .
Next, we attempt to find the general value of the sum
_
i =1
o i
u
(i ÷1)!
for any given nonnegative integer u. Our approach is very similar: we first consider
the case of u = 0,
E
0
=
1
0!
+
1
1!
+
1
2!
+... which has the value :, as we all know. A short proof is given to convince the
unsure reader (and will be our base case later):
The limit definiton of : is
Lim
n÷o
1 +
1
n
]
n
= Lim
n÷o

1
0!
+
n
1!

1
n
+
n(n÷1)
2!

1
n
2
+
n(n÷1) (n÷2)
3!

1
n
3
+...] =
1
0!
+
1
1!
+
1
2!
+... = :
E
1
=
1
0!
+
2
1!
+
3
2!
+
4
3!
+...
E
1
÷E
0
=
0
0!
+
1
1!
+
2
2!
+
3
3!
+
4
4!
+... = 0 +
1
0!
+
1
1!
+
1
2!
+... = E
0
= E
1
= 2 E
0
= 2 :. This result also
gives the rather interesting conclusion that
2
0!
+
2
1!
+
2
2!
+... =
1
0!
+
2
1!
+
3
2!
+... =0 =
÷1
0!
+
0
1!
+
1
2!
+
2
3!
+... or 1 =
1
2!
+
2
3!
+
3
4!
+...
When trying to determine this series' value, I computed E
2
by the following
method:
E
2
=
1
0!
+
4
1!
+
9
2!
+
16
3!
+
25
4!
+...
E
2
÷4 E
0
=
÷3
0!
+
0
1!
+
5
2!
+
12
3!
+... = ÷3 +3 
1
2!
+
1
3!
+
1
4!
+...] +
2
2!
+
6
3!
+
12
4!
+...] =
÷3 +3 +
1
0!
+
1
1!
+
1
2!
+... = E
0
= E
2
= 5 E
0
= 5 :
. And I thought
that was very clever. However, it turns out that the general method I determined is far easier to
work out, as doing something similar to this computation for E
3
became extremely difficult I
didn't figure out a method similar to the above one, but not for lack of trying.
The general method I determined works inductively (a method I've now fallen in love with):
S'pose we are trying to calculate E
k
. Then we first assume to know the values of
E
0
, E
1
, E
2
, ... E
k ÷1
. Consider the i
th
term of the series E
k
:
1
0!
+
2
k
1!
+
3
k
2!
+... +
i
k
(i ÷1)!
(the i
th
term simply being the quotient
i
k
(i ÷1)!
). We are trying to subtract
some number of other E
n
' s, n < k , perhaps with constants, from this i
th
term to obtain E
0
. We
notice (by way of pencil and some wasted paper) that should we be able to subtract whatever terms
we need to obtain the factorization (i ÷1) (i ÷2) (i ÷3) ... (i ÷k ) in the i
th
term, then we will have
that
E
k
÷Stuff = _
i =k +1
o 1
(i ÷k ÷1)
= E
0
.
Thus, we consider the expansion of that product, (which is known as a "falling factorial"), which
clearly has an i
k
term, an i
k ÷1
term, and so on down to a constant term. Remember that we are
dealing with an infinite sum thus, each of these terms represents an infinite sum of their own: we
quickly realize that each i
x
term, (which will have some constant C
x
attached to it), in summation
represents E
x
. As a result, we know that E
k
is equal to E
0
plus some other E
x
' s with constants,
and we have the value of E
k
(since we've assumed to know the values of all the E
x
' s. An example
is extremely helpful to see what I'm talking about here:
E
3
= _
i =1
o i
3
(i ÷1)!
, and we can evaluate
(i ÷1) (i ÷2) (i ÷3) = (i
2
÷3 i +2) (i ÷3) = i
3
÷6 i
2
+11 i ÷6, so we need to subtract 6 E
2
, add
11 E
1
, and subtract 6 E
0
from E
3
: clearly the terms i = 1, i = 2 and i = 3 will all become 0 since this
polynomial is 0 at those points, so we initiate the summation at i = 4.
E
3
÷6 E
2
+11 E
1
÷6 E
0
=
4
3
÷64
2
+114÷6
3!
+
5
3
÷65
2
+115÷6
4!
+
6
3
÷66
2
+116÷6
5!
+... =
1
0!
+
1
1!
+
1
2!
+
1
3!
+... = E
0
so
that
E
3
= 6 E
2
÷11 E
1
+7 E
0
= 30 : ÷22 : +7 : = 15 :.
Clearly, this process can be repeated indefinately. What is still a bit annoying is that there isn't an
explicit formula for the coefficients of the k
th
"falling factorial." Apparently they are related to
Stirling numbers of the second kind: as well, it turns out that there are a series of numbers called
the Bell numbers which are exactly the coefficients of the E
k
' s, and they are closely related to
Stirling numbers. The Bell numbers have some interesting relations (a number of which I rediscov
ered above), but their definition comes from set theory: the n
th
Bell number is how many ways the
set of integers 1 through n can be arranged into subsets (using all the numbers) independent of
order. I shall have to investigate the pattern further.
The other series related to these that I haven't quite figured out yet is closely related to the second
one:
_
i =1
o u
i
(i ÷1)!
, where the numerator has a geometric series and the denominator has the factorials.
It is clear that this series converges (as well as the convergence of the above two series), and proofs
will be left to the reader. This one turns out to have the value :
u
, but so far I can't see an easy
method of proving that.
Im port ant Infinit e S um s . nb 5
We might also realize that this method gives us the capability to find the value of the even more
general series
_
i =0
o
P(i ) x
i
, where P(i ) is a polynomial in i and x is constant: since we know the value of the
S
o,u
' s, we need only add their values with the necessary constants. For example, if
P(i ) = 17 i
3
÷289 i +4, then the value of the sum is simply
_
i =0
o
(17 i
3
÷289 i +4913) x
i
= 17 S
o,3
÷289 S
o,1
+4913 S
o,0
, which is simply a function of x .
Next, we attempt to find the general value of the sum
_
i =1
o i
u
(i ÷1)!
for any given nonnegative integer u. Our approach is very similar: we first consider
the case of u = 0,
E
0
=
1
0!
+
1
1!
+
1
2!
+... which has the value :, as we all know. A short proof is given to convince the
unsure reader (and will be our base case later):
The limit definiton of : is
Lim
n÷o
1 +
1
n
]
n
= Lim
n÷o

1
0!
+
n
1!

1
n
+
n(n÷1)
2!

1
n
2
+
n(n÷1) (n÷2)
3!

1
n
3
+...] =
1
0!
+
1
1!
+
1
2!
+... = :
E
1
=
1
0!
+
2
1!
+
3
2!
+
4
3!
+...
E
1
÷E
0
=
0
0!
+
1
1!
+
2
2!
+
3
3!
+
4
4!
+... = 0 +
1
0!
+
1
1!
+
1
2!
+... = E
0
= E
1
= 2 E
0
= 2 :. This result also
gives the rather interesting conclusion that
2
0!
+
2
1!
+
2
2!
+... =
1
0!
+
2
1!
+
3
2!
+... =0 =
÷1
0!
+
0
1!
+
1
2!
+
2
3!
+... or 1 =
1
2!
+
2
3!
+
3
4!
+...
When trying to determine this series' value, I computed E
2
by the following
method:
E
2
=
1
0!
+
4
1!
+
9
2!
+
16
3!
+
25
4!
+...
E
2
÷4 E
0
=
÷3
0!
+
0
1!
+
5
2!
+
12
3!
+... = ÷3 +3 
1
2!
+
1
3!
+
1
4!
+...] +
2
2!
+
6
3!
+
12
4!
+...] =
÷3 +3 +
1
0!
+
1
1!
+
1
2!
+... = E
0
= E
2
= 5 E
0
= 5 :
. And I thought
that was very clever. However, it turns out that the general method I determined is far easier to
work out, as doing something similar to this computation for E
3
became extremely difficult I
didn't figure out a method similar to the above one, but not for lack of trying.
The general method I determined works inductively (a method I've now fallen in love with):
S'pose we are trying to calculate E
k
. Then we first assume to know the values of
E
0
, E
1
, E
2
, ... E
k ÷1
. Consider the i
th
term of the series E
k
:
1
0!
+
2
k
1!
+
3
k
2!
+... +
i
k
(i ÷1)!
(the i
th
term simply being the quotient
i
k
(i ÷1)!
). We are trying to subtract
some number of other E
n
' s, n < k , perhaps with constants, from this i
th
term to obtain E
0
. We
notice (by way of pencil and some wasted paper) that should we be able to subtract whatever terms
we need to obtain the factorization (i ÷1) (i ÷2) (i ÷3) ... (i ÷k ) in the i
th
term, then we will have
that
E
k
÷Stuff = _
i =k +1
o 1
(i ÷k ÷1)
= E
0
.
Thus, we consider the expansion of that product, (which is known as a "falling factorial"), which
clearly has an i
k
term, an i
k ÷1
term, and so on down to a constant term. Remember that we are
dealing with an infinite sum thus, each of these terms represents an infinite sum of their own: we
quickly realize that each i
x
term, (which will have some constant C
x
attached to it), in summation
represents E
x
. As a result, we know that E
k
is equal to E
0
plus some other E
x
' s with constants,
and we have the value of E
k
(since we've assumed to know the values of all the E
x
' s. An example
is extremely helpful to see what I'm talking about here:
E
3
= _
i =1
o i
3
(i ÷1)!
, and we can evaluate
(i ÷1) (i ÷2) (i ÷3) = (i
2
÷3 i +2) (i ÷3) = i
3
÷6 i
2
+11 i ÷6, so we need to subtract 6 E
2
, add
11 E
1
, and subtract 6 E
0
from E
3
: clearly the terms i = 1, i = 2 and i = 3 will all become 0 since this
polynomial is 0 at those points, so we initiate the summation at i = 4.
E
3
÷6 E
2
+11 E
1
÷6 E
0
=
4
3
÷64
2
+114÷6
3!
+
5
3
÷65
2
+115÷6
4!
+
6
3
÷66
2
+116÷6
5!
+... =
1
0!
+
1
1!
+
1
2!
+
1
3!
+... = E
0
so
that
E
3
= 6 E
2
÷11 E
1
+7 E
0
= 30 : ÷22 : +7 : = 15 :.
Clearly, this process can be repeated indefinately. What is still a bit annoying is that there isn't an
explicit formula for the coefficients of the k
th
"falling factorial." Apparently they are related to
Stirling numbers of the second kind: as well, it turns out that there are a series of numbers called
the Bell numbers which are exactly the coefficients of the E
k
' s, and they are closely related to
Stirling numbers. The Bell numbers have some interesting relations (a number of which I rediscov
ered above), but their definition comes from set theory: the n
th
Bell number is how many ways the
set of integers 1 through n can be arranged into subsets (using all the numbers) independent of
order. I shall have to investigate the pattern further.
The other series related to these that I haven't quite figured out yet is closely related to the second
one:
_
i =1
o u
i
(i ÷1)!
, where the numerator has a geometric series and the denominator has the factorials.
It is clear that this series converges (as well as the convergence of the above two series), and proofs
will be left to the reader. This one turns out to have the value :
u
, but so far I can't see an easy
method of proving that.
6 Im port ant Infinit e S um s . nb
We might also realize that this method gives us the capability to find the value of the even more
general series
_
i =0
o
P(i ) x
i
, where P(i ) is a polynomial in i and x is constant: since we know the value of the
S
o,u
' s, we need only add their values with the necessary constants. For example, if
P(i ) = 17 i
3
÷289 i +4, then the value of the sum is simply
_
i =0
o
(17 i
3
÷289 i +4913) x
i
= 17 S
o,3
÷289 S
o,1
+4913 S
o,0
, which is simply a function of x .
Next, we attempt to find the general value of the sum
_
i =1
o i
u
(i ÷1)!
for any given nonnegative integer u. Our approach is very similar: we first consider
the case of u = 0,
E
0
=
1
0!
+
1
1!
+
1
2!
+... which has the value :, as we all know. A short proof is given to convince the
unsure reader (and will be our base case later):
The limit definiton of : is
Lim
n÷o
1 +
1
n
]
n
= Lim
n÷o

1
0!
+
n
1!

1
n
+
n(n÷1)
2!

1
n
2
+
n(n÷1) (n÷2)
3!

1
n
3
+...] =
1
0!
+
1
1!
+
1
2!
+... = :
E
1
=
1
0!
+
2
1!
+
3
2!
+
4
3!
+...
E
1
÷E
0
=
0
0!
+
1
1!
+
2
2!
+
3
3!
+
4
4!
+... = 0 +
1
0!
+
1
1!
+
1
2!
+... = E
0
= E
1
= 2 E
0
= 2 :. This result also
gives the rather interesting conclusion that
2
0!
+
2
1!
+
2
2!
+... =
1
0!
+
2
1!
+
3
2!
+... =0 =
÷1
0!
+
0
1!
+
1
2!
+
2
3!
+... or 1 =
1
2!
+
2
3!
+
3
4!
+...
When trying to determine this series' value, I computed E
2
by the following
method:
E
2
=
1
0!
+
4
1!
+
9
2!
+
16
3!
+
25
4!
+...
E
2
÷4 E
0
=
÷3
0!
+
0
1!
+
5
2!
+
12
3!
+... = ÷3 +3 
1
2!
+
1
3!
+
1
4!
+...] +
2
2!
+
6
3!
+
12
4!
+...] =
÷3 +3 +
1
0!
+
1
1!
+
1
2!
+... = E
0
= E
2
= 5 E
0
= 5 :
. And I thought
that was very clever. However, it turns out that the general method I determined is far easier to
work out, as doing something similar to this computation for E
3
became extremely difficult I
didn't figure out a method similar to the above one, but not for lack of trying.
The general method I determined works inductively (a method I've now fallen in love with):
S'pose we are trying to calculate E
k
. Then we first assume to know the values of
E
0
, E
1
, E
2
, ... E
k ÷1
. Consider the i
th
term of the series E
k
:
1
0!
+
2
k
1!
+
3
k
2!
+... +
i
k
(i ÷1)!
(the i
th
term simply being the quotient
i
k
(i ÷1)!
). We are trying to subtract
some number of other E
n
' s, n < k , perhaps with constants, from this i
th
term to obtain E
0
. We
notice (by way of pencil and some wasted paper) that should we be able to subtract whatever terms
we need to obtain the factorization (i ÷1) (i ÷2) (i ÷3) ... (i ÷k ) in the i
th
term, then we will have
that
E
k
÷Stuff = _
i =k +1
o 1
(i ÷k ÷1)
= E
0
.
Thus, we consider the expansion of that product, (which is known as a "falling factorial"), which
clearly has an i
k
term, an i
k ÷1
term, and so on down to a constant term. Remember that we are
dealing with an infinite sum thus, each of these terms represents an infinite sum of their own: we
quickly realize that each i
x
term, (which will have some constant C
x
attached to it), in summation
represents E
x
. As a result, we know that E
k
is equal to E
0
plus some other E
x
' s with constants,
and we have the value of E
k
(since we've assumed to know the values of all the E
x
' s. An example
is extremely helpful to see what I'm talking about here:
E
3
= _
i =1
o i
3
(i ÷1)!
, and we can evaluate
(i ÷1) (i ÷2) (i ÷3) = (i
2
÷3 i +2) (i ÷3) = i
3
÷6 i
2
+11 i ÷6, so we need to subtract 6 E
2
, add
11 E
1
, and subtract 6 E
0
from E
3
: clearly the terms i = 1, i = 2 and i = 3 will all become 0 since this
polynomial is 0 at those points, so we initiate the summation at i = 4.
E
3
÷6 E
2
+11 E
1
÷6 E
0
=
4
3
÷64
2
+114÷6
3!
+
5
3
÷65
2
+115÷6
4!
+
6
3
÷66
2
+116÷6
5!
+... =
1
0!
+
1
1!
+
1
2!
+
1
3!
+... = E
0
so
that
E
3
= 6 E
2
÷11 E
1
+7 E
0
= 30 : ÷22 : +7 : = 15 :.
Clearly, this process can be repeated indefinately. What is still a bit annoying is that there isn't an
explicit formula for the coefficients of the k
th
"falling factorial." Apparently they are related to
Stirling numbers of the second kind: as well, it turns out that there are a series of numbers called
the Bell numbers which are exactly the coefficients of the E
k
' s, and they are closely related to
Stirling numbers. The Bell numbers have some interesting relations (a number of which I rediscov
ered above), but their definition comes from set theory: the n
th
Bell number is how many ways the
set of integers 1 through n can be arranged into subsets (using all the numbers) independent of
order. I shall have to investigate the pattern further.
The other series related to these that I haven't quite figured out yet is closely related to the second
one:
_
i =1
o u
i
(i ÷1)!
, where the numerator has a geometric series and the denominator has the factorials.
It is clear that this series converges (as well as the convergence of the above two series), and proofs
will be left to the reader. This one turns out to have the value :
u
, but so far I can't see an easy
method of proving that.
Im port ant Infinit e S um s . nb 7