(1+1/n)n is a tug of war. The base 1+1/n tends to 1, which pulls towards 1n=1; the exponent n tends to ∞, which pulls towards +∞. Neither the algebra of limits nor the result on xn settles it, because the base changes with n. §73 settles it with the monotone theorem: the sequence increases and is bounded, so it converges, and its limit is one of the two most important constants in mathematics, e. Chapter IX takes this limit as the starting point of the exponential function.
The idea
Compound interest is the classic picture. A pound at 100% interest for a year becomes 2. Paid half-yearly, it becomes (1+21)2=2.25; monthly, (1+121)12≈2.613; daily, about 2.7146. More frequent compounding always helps, but by less and less. Does the amount grow without bound, or approach a ceiling?
Expanding by the binomial theorem shows what is going on. With n fixed,
The term with index p is p!1(1−n1)(1−n2)⋯(1−np−1).
ExploreThe expansion, term by term
Try this.Move n from 1 up to about 10, one step at a time, with “mark the terms for n+1” on. Then switch on the bound 1/2p−1 and push n to 60.
(1+31)3=∑p=03p!1(1−n1)⋯(1−np−1)
pterm
term p for this n
cap 1/p!
term p for n + 1
Sum of all 4 terms: (1+1/3)3=2.37037. For n+1=4: 2.441406, larger, because every term grew and a new one appeared. The caps give ∑p=031/p!=2.666667, which is below 1+1+21+41+⋯<3. The limit is e=2.718282; at this n the gap is 0.3479.
What to notice
Every bar grows when n increases (each bracket 1−k/n gets closer to 1), and a new bar appears; so the total increases. But no bar ever passes its cap 1/p!, and the caps are themselves below 1/2p−1, which add up to less than 2. So the total stays below 1+2=3.
Precise statements
TheoremThe number eHardy: §73
The sequence ϕ(n)=(1+1/n)n is strictly increasing and satisfies 2≤ϕ(n)<3 for every n. Hence it tends to a limit
e=n→∞lim(1+n1)n,2<e≤3.
(We shall see in the exercises that in fact e<2.72; the true value is e=2.718281828…)
Proofs
Proof(1+1/n)n increases and is less than 3
Strategy
Write the binomial expansion so that each term is visibly an increasing function of n; then bound each term by 1/p!, and 1/p! by 1/2p−1, and sum a geometric progression.
Rewrite the binomial coefficients
For 1≤p≤n: (pn)np1=p!npn(n−1)⋯(n−p+1)=p!1(1−n1)⋯(1−np−1), since each factor (n−k)/n equals 1−k/n.
Each term increases with n
For fixed p and k<p, the factor 1−k/n is positive and increases with n (strictly, when k≥1). So each term with p≥2 is strictly larger for n+1 than for n; the terms p=0,1 are always 1.
And there are more terms
The expansion for n+1 has one more term than that for n, and it is positive. So ϕ(n+1)>ϕ(n): the sequence is strictly increasing. In particular ϕ(n)≥ϕ(1)=2.
Bound by 1/p!, then by 1/2p−1
Each bracket is at most 1, so the p-th term is at most 1/p!. And p!=1⋅2⋅3⋯p≥2p−1 (each factor after the first is at least 2). Hence
By Cor. 2 of §69, an increasing sequence bounded by 3 tends to a limit e≤3. Since the sequence is strictly increasing, e>ϕ(n) for every n (Hardy’s remark in §69), in particular e>ϕ(1)=2.
∎
Notice what the proof did not do. It never found n0(δ), and it never identified e with anything else. The monotone theorem guarantees the limit exists; e is then defined as that limit. How fast the sequence approaches e is a separate question, and the answer is: slowly.
ExploreHow slowly (1+1/n)n approaches e
Try this.Find the smallest n0 for δ=0.1 and then for δ=0.01 (show more terms when you need them, and use “Zoom in on the band”).
ϕ(n)=(1+n1)n , candidate limit l=2.718
nφ(n)
band l ± δ
l
n₀ (drag)
n ≥ n₀, good
n ≥ n₀, bad
12 terms with n ≥ n₀ fall outside the band: n = 1, 2, 3, 4, 5, ….
What to notice
You need n0=13 for δ=0.1 and n0=135 for δ=0.01: ten times the accuracy costs about ten times as many terms, since e−(1+1/n)n is roughly e/(2n). The series 1+1+1/2!+1/3!+…, whose terms are the caps in the first explorer, reaches e far faster.
Worked examples and exercises
ExerciseA better upper bound●●○Hardy: §73
Improve the bound 3: show that (1+1/n)n<2.72 for every n, so that e≤2.72.
Hint
Keep the exact values 1/2! and 1/3!, and bound the terms from p=4 on by a geometric series with ratio 51.
Solution
For p≥4, p!=24⋅5⋅6⋯p≥24⋅5p−4, so 1/p!≤241(1/5)p−4. Therefore
By Bernoulli, (1+n2−11)n≥1+n2−1n>1+n2n=1+n1=nn+1. So the ratio exceeds 1 and ψ(n−1)>ψ(n).
Bounds.ψ decreases to e, so ψ(n)>e strictly (it is strictly decreasing, so it is never equal to its limit); and ϕ(n)<e as shown above. For n=10 this traps e between 2.5937 and 2.8531.
Check yourself
Check yourselfChoose one
A student argues: ”1+1/n→1 and 1n=1, so (1+1/n)n→1.” What is wrong?
Why
The rules of Limits of sums, products and quotients cover a fixed number of operations. Here the number of factors grows with n, so the rule for products does not apply. The opposite argument (“base greater than 1, so it tends to +∞”) fails for the same reason: the theorem on xn is about a fixed x.
Check yourselfTrue or false?
(1+1/n)n<e for every positive integer n.
Why
True. The sequence is strictly increasing, so each term is less than the next, and the limit of an increasing sequence is at least every term (Cor. 3); hence each term is strictly below e.
Check yourselfEnter a number
What is (1+1/n)n for n=2?
Why
(1+21)2=(23)2=49=2.25. Compare ϕ(1)=2 and ϕ(3)=(34)3=2.370…: increasing, as the theorem says.
Hardy’s language
Summary
Summary
In the binomial expansion of (1+1/n)n, the p-th term is p!1(1−n1)⋯(1−np−1), which grows with n; there are also more terms. So the sequence increases strictly.
Each term is at most 1/p!≤1/2p−1, so the sum is less than 3.
By the monotone theorem the sequence has a limit, which defines e, with 2<e≤3 (in fact e≈2.71828).
"1∞" is an indeterminate form: a base tending to 1 raised to a growing power can do anything.