DEFINING EXPONENTIAL FUNCTIONS VIA LIMITS

STEVE FAN

Abstract. In this short note we prove a few classical properties of the exponential function
ex from the simple definition  x n
ex := lim 1 + .
n→∞ n
This note is a result of the author’s curiosity in exploring and possibly teaching ex from a
different point of view.

1. The Exponential Function ex

In standard calculus textbooks, the exponential function ex is defined by
∞
X xn
ex := (1)
n=0
n!
x
for all x ∈ R. The continuity of e then follows from the uniform convergence of the power
series that defines it on closed and bounded intervals. Similarly, one can show that ex is
differentiable everywhere. An alternative way of defining ex is by considering the following
initial value problem
d
f (x) = f (x),
dx
f (0) = 1.
One defines ex to be the unique solution f (x) to this problem. One of the advantages of this
approach is that one has the differentiability of ex for free. However, there is a third method
when it comes to defining ex . It is well known that the constant e is commonly defined by
 n
1
e := lim 1 + .
n→∞ n
It is hence natural to define ex by
x n

ex := lim 1+
. (2)
n→∞ n
This definition was introduced by L. Euler [2] who actually derived the power series expansion
(1) from it (in a somewhat unrigorous way). Using Bernoulli’s inequality that (1 + x)n >
1 + nx holds for all positive integers n ≥ 2 and all nonzero x > −1, it is not hard to show
that for every x ∈ R \ {0}, the sequence {an (x)}∞n=1 defined by
 x n
an (x) := 1 +
n
is strictly increasing for n > max(0, −x). It is also not hard to prove that the sequence
{an (x)}∞n=1 is bounded for every fixed x ∈ R. Indeed, let us fix x ∈ R and denote by m the

1
2 STEVE FAN

least positive integer greater than or equal to |x|. Then for sufficiently large n, we have
 mn
 m n  m mn 1
|an (x)| ≤ 1 + ≤ 1+ = 1+ ,
n mn n
where we have used the monotonicity of {an (x)}∞n=1 . Since
 n n   n n
1 X n 1 X 1 X 1 1
1+ =2+ k
≤ 2 + < 2 + = 3 − ,
n k=2
k n k=2
k! k=2
k(k − 1) n
it follows that  m
1
|an (x)| ≤ 3 −
n
for all sufficiently large n. This proves that {an (x)}∞ n=1 is bounded for every fixed x ∈ R.
n
Here the treatment for (1 + 1/n) is classical. However, the author discovered a slightly
different way to prove that the sequence {an (x)}∞ n=1 is bounded for every fixed x ∈ R. The
starting point is another simple inequality due to Bernoulli:
1
(1 + x)n ≤ ,
1 − nx
where n is a positive integer and x ∈ (−1, 1/n). This inequality can be proved easily by
induction. As a consequence of this inequality, we have
 2mn  2mn
 m n 2m 1
|an (x)| ≤ 1 + ≤ 1+ = 1+ ≤ 4m
n 2mn 2n
for all sufficiently large n. This again proves that {an (x)}∞
n=1 is bounded for every fixed x ∈ R.
Now it follows from the monotone convergence theorem that {an (x)}∞ n=1 is convergent for
every x ∈ R. This justifies Euler’s definition (2). Moreover, the monotonicity of {an (x)}∞ n=1
implies that for every x ∈ R \ {0}, the inequality
 x n
ex > 1 +
n
holds for all n > max(0, −x). In particular, this yields ex > max(0, 1 + x) for all x ∈ R \ {0},
by Bernoulli’s inequality. It is also clear from (2) that ex is increasing on R.

2. The Continuity and Differentiability of ex

Now we derive from (2) the fact that ex is continuous everywhere. Fixing x0 ∈ R and
 > 0, we have
x x0
h x n  x0  n i
e − e = lim 1 + − 1+
n→∞ n n
n−1 
1 X x n−1−k  x 0 k
= (x − x0 ) lim 1+ 1+ .
n→∞ n n n
k=0

If |x − x0 | < , then
 n n−1  n
x0 −  1 X x n−1−k  x0  k x0 + 
1+ ≤ 1+ 1+ ≤ 1+
n n k=0 n n n
 DEFINING EXPONENTIAL FUNCTIONS VIA LIMITS 3

for sufficiently large n. It follows that

ex0 − |x − x0 | ≤ |ex − ex0 | ≤ ex0 + |x − x0 | (3)
for all x ∈ (x0 − , x0 + ). This proves that ex is continuous at x0 . Since x0 ∈ R is arbitrary,
we conclude that ex is continuous on R.
Now that we have established the continuity of ex , it follows immediately from (3) and
the monotonicity of ex that ex is differentiable with
d x
(e ) = ex > 0
dx
for all x ∈ R. Thus e is strictly increasing and ex ∈ C ∞ (R). From this the power series
x

expansion (1) of ex follows naturally from Taylor’s theorem. If f (x) is a differentiable function
satisfying f 0 (x) = f (x), then
f 0 (x) − f (x)
 
d f (x)
= =0
dx ex ex
for all x ∈ R. This implies f (x) = Cex for all x ∈ R, where C ∈ R is a constant. Hence ex
is the unique solution to the initial value problem
d
f (x) = f (x),
dx
f (0) = 1.
We have thus shown that the definition (2) implies both the power series definition and the
differential equation definition of ex . One advantage of this approach is that ex defined by
(2) provides an explicit and elementary solution to the initial value problem in consideration,
from which the uniqueness follows naturally as we saw above.

3. The Addition Law for ex

The addition law for ex states that ex+y = ex ey for all x, y ∈ R. It is not immediately
clear how the addition law follows from (1), but it can be easily verified once we know ex is
differentiable with derivative ex . Indeed, we have by the chain rule that
d c−x x d c−x x d
(e e ) = (e )e + ec−x (ex ) = −ec−x ex + ec−x ex = 0
dx dx dx
for all x ∈ R, where c ∈ R is a constant. Thus ec−x ex is constant. Since its value at x = 0
is ec , we find that ec−x ex = ec for all x, c ∈ R. Taking c = x + y yields the addition law.
Without doubt, this classical argument [1, §3.1] is neat and elegant.
On the other hand, the author found a way to derive the addition law without using the
differentiability of ex . Note first that
n
x2

x −x
 x n  x n
e e = lim 1 + 1− = lim 1 − 2 .
n→∞ n n n→∞ n
By Bernoulli’s inequality we have
n
x2 x2

1− ≤ 1− 2 ≤1
n n
4 STEVE FAN

for sufficiently large n. It follows that

n
x2

lim 1 − 2 = 1.
n→∞ n
Hence ex e−x = 1. More generally, suppose that x, y ∈ R are arbitrary. Then
 n
x y x + y xy
e e = lim 1 + + 2 .
n→∞ n n
For any  > 0 we have
xy 
2
<
n n
for all sufficiently large n. It follows that
 n  n  n
x+y− x + y xy x+y+
1+ < 1+ + 2 < 1+ .
n n n n
Thus we have
ex+y− ≤ ex ey ≤ ex+y+ .
Since ex is continuous and  > 0 is arbitrary, we have ex ey = ex+y .

4. The Natural Logarithm

The natural logarithm, denoted by log x or ln x, is defined to be the inverse function of ex ,
namely, elog x = x. It is strictly increasing on its domain (0, +∞). We now show that log x
is continuous everywhere. Fix x0 > 0 and 0 <  < x0 . As in Section 2, we have
n−1  n−1−k  k
1X log x log x0
x − x0 = (log x − log x0 ) lim 1+ 1+ .
n→∞ n n n
k=0

for all x > 0. If |x − x0 | < , then

 n n−1  n−1−k  k  n
log(x0 − ) 1X log x log x0 log(x0 + )
1+ ≤ 1+ 1+ ≤ 1+
n n k=0 n n n
for sufficiently large n. Hence
| log x − log x0 |(x0 − ) ≤ |x − x0 | ≤ | log x − log x0 |(x0 + ).
This shows that log x is continuous at x0 . Hence log x is continuous on (0, +∞). Moreover,
we have
1 log x − log x0 1
≤ ≤
x0 +  x − x0 x0 − 
for all x > 0 with x 6= x0 and |x − x0 | < . This implies that log x is differentiable at x0 with
d 1
(log x) = .
dx x=x0 x0
Hence log x is differentiable on (0, +∞) with
d 1
(log x) = .
dx x
 DEFINING EXPONENTIAL FUNCTIONS VIA LIMITS 5

Now one can define ax := ex log a , where x ∈ R and a > 0. Then it is easy to see that
ax+y = ax ay and
x
(ax )y = ey log a = exy log a = axy
for all x, y ∈ R and a > 0.

5. Extension to the Complex Exponential Function ez

Both the power series expansion (1) and the differential equation approach [1, §3.1] can
be extended to define the complex exponential function ez . It is thus tempting to generalize
(2) as well by defining
 z n
ez := lim 1 + (4)
n→∞ n
for z = x + iy ∈ C. Once the existence of the limit on the right-hand side is established for
every z ∈ C, one can show as in Section 2 that ez defined in this way is continuous on C, but
some extra effort is needed to show that ez is holomorphic everywhere. The author found a
way to prove that the sequence {an (z)}∞ n=1 defined by
 z n
an (z) := 1 +
n
converges uniformly on every bounded subset of C. Let r > 0 be a constant and let E ⊆
{z ∈ C : |z| ≤ r}. For any positive integers n > m ≥ 2 we have
m      n   k
X n 1 m 1 k
X n |z|
|an (z) − am (z)| ≤ k
− k
|z| + k
.
k=2
k n k m k=m+1
k n

It is clear that uniformly for all z ∈ E, we have

X n 1    n   k   k X rk
m 1 k
X n |z| X n r
k
− k
|z| + < < →0
√4
k n k m k=m+1
k nk √
4
k nk √4
k!
m<k≤m m<k≤n k> m
√
as m → ∞. For 2 ≤ k ≤ m, it follows by Bernoulli’s inequality that
4

k−1 k
k2
     Y  
n 1 m 1 l k 1
k! k
− k
< 1 − 1 − < 1 − 1 − < ≤√
k n k m l=0
m m m m
for sufficiently large m. Hence uniformly for all z ∈ E, we have
∞
X n 1 1 X rk
  
m 1 k
− |z| < √ →0
√
4
k nk k mk m k=2 k!
2≤k≤ m

as m → ∞. We have thus shown that an (z) − am (z) → 0 uniformly for all z ∈ E as m → ∞.

By Cauchy’s uniform convergence test, we conclude that {an (z)}∞
n=1 is uniformly convergent
∞
on E. This proves that {an (z)}n=1 is uniformly convergent on every bounded subset of C.
By a theorem of Weierstrass [1, Theorem 1, §5], we know that an (z) converges to an entire
function, which we denote by ez , with derivative
d z d  z n  z n−1
(e ) = lim 1+ = lim 1 + = ez .
dz n→∞ dz n n→∞ n
6 STEVE FAN

Having proved this, we obtain immediately the addition law ea+b = ea eb for all a, b ∈ C. It
is also clear that this approach allows one to define the exponential function in an arbitrary
unital Banach algebra.
In the argument above, we used the fact that the series
∞
X rn
n=0
n!

is convergent for every r ≥ 0. This is almost trivial when 0 ≤ r ≤ 1. For r > 1, the
convergence of this series follows from the fact that n! > r2n for all sufficiently large n. It
is worth noting that in comparison to the power series definition of ez , the definition (4)
ix
does not give clear clues√to Euler’s√ formula xe = cos x + i sin x, though it does yield at
z̄ z
z
once e = e and |e | = e e = ez z̄ z+z̄ = e for any z = x + iy ∈ C. Thus eiy : R → S 1
defines a homomorphism from R to the unit circle S 1 , where both R and S 1 are considered
as topological groups.
Now we give a proof of Euler’s formula eix = cos x + i sin x . Let us write

eix = f (x) + ig(x),

where
n x2k
X  
k
f (x) = lim (−1) ,
n→∞ 2k n2k
0≤k≤n/2
  2k+1
X
k n x
g(x) = lim (−1) .
n→∞ 2k + 1 n2k+1
0≤k≤(n−1)/2

Since
d ix
(e ) = ieix = −g(x) + if (x),
dx
we have f 0 (x) = −g(x) and g 0 (x) = f (x). Therefore, f (x) and g(x) are solutions to the
following initial value problem

f 0 (x) = −g(x), g 0 (x) = f (x),

f (0) = 1, g(0) = 0.

It follows that f (x) = cos x and g(x) = sin x. Euler’s formula makes it reasonable to define,
for all z ∈ C, the complex trigonometric functions

eiz + e−iz
  2k
X
k n z
cos z := = lim (−1) ,
2 n→∞ 2k n2k
0≤k≤n/2
iz −iz
  2k+1
e −e X
k n z
sin z := = lim (−1) .
2i n→∞ 2k + 1 n2k+1
0≤k≤(n−1)/2
 DEFINING EXPONENTIAL FUNCTIONS VIA LIMITS 7

6. Final Comments
There are other alternative ways of defining ex . For instance, it is well known that the set
of all positive continuous solutions to the Cauchy functional equation f (x + y) = f (x)f (y)
is {ax : a > 0}. Thus we may define ex to be the unique positive continuous solution to this
Cauchy equation with its value at 1 given by e. On the other hand, we may first define
log x to be the the unique nonzero real-valued continuous solution to the Cauchy equation
f (xy) = f (x) + f (y) on (0, +∞) with its value at e given by 1, and then define ex to be its
inverse function. Of course, if one is willing
R x to resort to the theory of integration, then log x
can be defined by the definite integral 1 1/t dt, though this is not as elementary as the ones
suggested above.

References
[1] L. V. Ahlfors, Complex Analysis: An Introduction to the Theory of Analytic Functions of One Complex
Variable, 3rd ed., International Series in Pure and Applied Mathematics, McGraw-Hill, 1979.
[2] L. Euler and J. D. Blanton (transl.), Introduction to Analysis of the Infinite, Book I, Springer-Verlag
New York, 1988.

Department of Mathematics, Dartmouth College, Hanover, NH 03755, USA

Email address: [email protected]