Introductory Analysis 1–Fall 2007

The number e

The best definition of this very important number is probably

∞
X 1
e= .
n=0
n1

But, because we do not officially know anything about series yet, we have to be
a bit more explicit. We will define a sequence {sn }∞
n=0 by

1 1 1
s0 = 1, s1 = 1 + 1 = 2, s2 = 1 + 1 + = 2.5, . . . , sn = 1 + 1 + + · · · + , . . .
2! 2! n!
briefly,
n
X 1
sn = .
k!
k=0
We now have the rather simple lemma:
Lemma 1 The sequence {sn } is strictly increasing and bounded above, hence
converges.
Proof. We have
   
1 1 1 1
sn+1 − sn = 1 + ··· + + − 1 + ··· + = > 0;
n! (n + 1)! n! (n + 1)!
it follows that the sequence is strictly increasing. To show that it is bounded we
will use a rather crude estimate; we use the fact that k! > 2k for k = 4, 5, . . ..
Crude because, for example, 10! = 3, 628, 800 while 210 = 1024; the difference
in size gets to be quite notable. The proof of this estimate is a trivial exercise
in mathematical induction. From it we get, if n ≥ 4:
   
1 1 1 1 1 1 1 1
sn = 1 + 1 + + + + ··· + ≤1+1+ + + + ··· + n
2 6 4! n! 2 6 24 2
1
1 1 1 1 − 2n−3 1 1 1 1
= 1+1+ + + 4· ≤1+1+ + + 4 ·
2 6 2 1 − 12 2 6 2 1− 1
2
67
= ≈ 2.7917.
24
We used here the maybe well known formula for sum of a geometric progression:
If r 6= 1, then
1 − rm+1
1 + r + r2 + · · · + rm = .
1−r
With r = 1/2, m = n − 4, one gets
1
1 1 − 2n−3
 
1 1 1 1 1
+ · · · + = 1 + + · · · + = · .
24 2n 24 2 2n−4 24 1 − 12
The sequence is bounded. In fact,
67
(1) sn ≤ <3 for n = 0, 1, 2, . . .
24

Because of the lemma we can officially define the number e by

(2) e = lim sn .
n→∞
Since {sn } is strictly increasing and s1 = 2, it is clear that e > 2. In view of (1)
we also see that e ≤ 67/24 thus e < 3. That is,

(3) 2 < e < 3.

We will now prove what I think is a rather non-obvious fact, namely that we
also have
 n
1
(4) e = lim 1 + .
n→∞ n

The proof is adapted (copied?, plagiarized?) from Rudin’s Principles of Mathe-

matical Analysis and is presented here mostly to show how one uses the concepts
of lim sup and lim inf. Let us set
 n
1
tn = 1 + , n = 0, 1, 2, . . . .
n

We now need to remember the binomial formula, another formula that isn’t too
hard to prove by induction. Namely,
n  
n
X n n(n − 1) 2 n−2
(a + b) = an−k bk = bn + nabn−1 + a b + · · · + an .
k 2
k=0

We apply it to tn , so we use it with a = 1, b = 1/n. The general term in the

binomial expansion is then
 
n 1 n! n(n − 1) · · · (n − k + 1)(n − k)! n(n − 1) · · · (n − k + 1)
= = =
k nk k!(n − k)!nk k!(n − k)!nk k!nk
n(n − 1) · · · (n − k + 1) 1 n−1 n−2 n−k+1 1
= · = · ··· · .
nk k! n n n k!
The factors preceding 1/k! are all less than 1, so is their product and it follows
that this general term is ≤ 1/k!. That is,
n   n
X n 1 X 1
(5) tn = ≤ = sn , n = 0, 1, 2, . . .
k nk k!
k=0 k=0

Actually, a careful reader would notice that our expression for the k-th term
assumes implicitly that k ≥ 2; however one sees directly that for k = 0, 1 he
terms in the expansion of tn are equal to those in the expression for sn (they are
both equal to 1), so (5) follows without a problem. In fact, we proved tn < sn
if n ≥ 2.

Here is where limsup and liminf come into the picture. We don’t know
whether {tn } converges yet, so we can’t talk of its limit. But its limsup and
liminf always exist. We apply te following result; the proof is left as an exercise:
Lemma 2 Let {an }, {bn }, {cn } be sequences of real numbers. Assume {an }
converges; limn→∞ an = a.
1. If bn ≤ an for all n, then lim supn→∞ bn ≤ a.

2. If cn ≥ an for all n, then lim inf n→∞ cn ≥ a.

2
 Applying the lemma with an = sn and bn = tn , we get lim supn→∞ tn ≤
limn→∞ sn ; that is

(6) lim sup tn ≤ e.

n→∞

Let us return to the binomial expansion

 of tn . We’ll use the expression we got
n 1
above for the k-th term; that is for nk
. We repeat the expression here
k
in a slightly different way, namely
      
n 1 1 2 k−1 1
= 1− 1− ··· 1 − · .
k n k n n n k!
noticing this time that this expression is only valid for k ≥ 2. Let us fix for a
while m ∈ N, m ≥ 2 and et n ≥ m. Then
n   m  
X n 1 X n 1
tn = ≥
k nk k nk
k=0 k=0
         
1 1 1 2 1 1 2 m−1 1
= 1+1+ 1− + 1− 1− + ··· + 1 − 1− ··· 1 − · ..
n 2! n n 3! n n n m!
Here is the tricky thing. We keep m fixed and let n → ∞. The expression on
the right hand side above; that is
    
1 2 m−1 1
1+1+ 1− 1− ··· 1 − · ,
n n n m!
is a sum of m + 1 terms and it should be clear that, because we are dealing
with a finite sum of terms!, that
      
1 2 m−1 1 1 1
lim 1 + 1 + 1 − 1− ··· 1 − · = 1+1+ +· · ·+ = sm .
n→∞ n n n m! 2! m!
By Lemma 2, applied now with
    
1 2 m−1 1
an = 1 + 1 + 1 − 1− ··· 1 − · , cn = tn ,
n n n m!
we get

(7) lim inf tn ≥ sm ,

n→∞

and since m ∈ N, m ≥ 2 was arbitrary, (7) holds for all m ∈ N, m ≥ 2.

But now we use the fact that if sm ≤ τ for all m and {sm } converges, then
limm→∞ sm ≤ τ . Applying this with τ = lim inf n→∞ tn and considering that
limm→∞ sm = e, we proved

(8) lim inf tn ≥ e.

n→∞

Together with (6), we proved

(9) lim sup tn ≤ e ≤ lim inf tn .

n→∞ n→∞

Since ne always has that lim inf n→∞ tn ≤ lim supn→∞ tn , we actually proved
that
lim inf tn = e = lim sup tn
n→∞ n→∞

hence that limn→∞ tn = e.