Limits, Continuity & the Big Theorems

Study Sheet

Limits, Continuity & the Big Theorems

IVT, MVT, Taylor with remainder, and limits in disguise

The Theorems That Do the Work

Tip
Limits that are derivatives, integrals, or exponentials in disguise

n(21/n1)ln2n(2^{1/n} - 1) \to \ln 2 (a derivative); 1nf(kn)01f\tfrac1n\sum f(\tfrac kn) \to \int_0^1 f (an integral); (1+an)nea(1 + \tfrac an)^n \to e^a (an exponential); (n!)1/nn1e\tfrac{(n!)^{1/n}}{n} \to \tfrac1e (a log of a Riemann sum).

The first reflex for any strange limit: rewrite it as a difference quotient, a Riemann sum, or exp(log)\exp(\text{log}). The second: Taylor expand — sinxxx316\tfrac{\sin x - x}{x^3} \to -\tfrac16 because sinx=xx36+\sin x = x - \tfrac{x^3}{6} + \cdots, and lim(cosx)1/x2=e1/2\lim(\cos x)^{1/x^2} = e^{-1/2} because lncosxx22\ln\cos x \approx -\tfrac{x^2}{2}. L'Hôpital is the tool of last resort: it is slow and it hides the structure.

Tip
IVT, MVT, Rolle, and Taylor with remainder

Continuous ff on [a,b][a,b] takes every value between f(a)f(a) and f(b)f(b). Differentiable ff has f(b)f(a)=f(c)(ba)f(b) - f(a) = f'(c)(b - a) for some c(a,b)c \in (a,b). And f(x)=k<nf(k)(a)k!(xa)k+f(n)(ξ)n!(xa)nf(x) = \sum_{k<n}\tfrac{f^{(k)}(a)}{k!}(x-a)^k + \tfrac{f^{(n)}(\xi)}{n!}(x-a)^n.

Reminder — The Mean Value Theorem:f(c)=f(b)f(a)bafor some c(a,b)f'(c)=\frac{f(b)-f(a)}{b-a} \quad\text{for some } c \in (a,b)
Reminder — The Intermediate Value Theorem:f continuous on [a,b], f(a)<k<f(b)  f(c)=k for some c(a,b)f \text{ continuous on } [a,b],\ f(a)<k<f(b) \ \Longrightarrow\ f(c)=k \text{ for some } c\in(a,b)

Existence of roots is IVT; uniqueness is monotonicity (a derivative that keeps one sign) — x=cosxx = \cos x has exactly one solution, ex=3xe^x = 3x exactly two. MVT converts a statement about VALUES into one about a DERIVATIVE at some unknown point; Rolle is the case f(a)=f(b)f(a) = f(b), which proves that between two roots of ff lies a root of ff' (so a polynomial with nn real roots has n1n - 1 real critical points). Taylor with the Lagrange remainder gives INEQUALITIES: cosx1x22\cos x \ge 1 - \tfrac{x^2}{2}, ex1+xe^x \ge 1 + x, ln(1+x)x\ln(1+x) \le x — each a one-line remainder-sign argument.

Side note
Continuity versus uniform continuity, and why graders care

On a closed bounded interval, continuous functions are bounded, attain their extrema, and are uniformly continuous — three facts a Putnam solution may cite by name. On open or infinite intervals none of them is automatic, and the counterexamples (1x\tfrac1x on (0,1)(0,1), x2x^2 on R\mathbb R) are exactly the ones that appear in the hard problems.

Try it
Try it: count the solutions

How many real solutions does x33x+1=0x^3 - 3x + 1 = 0 have?

f=3x23f' = 3x^2 - 3 vanishes at ±1\pm1; f(1)=3>0f(-1) = 3 > 0 and f(1)=1<0f(1) = -1 < 0 with ff \to \mp\infty at \mp\infty: three sign changes, hence three real roots (and a cubic has at most three). They are 2cos2πk92\cos\tfrac{2\pi k}{9}-type numbers — not needed for the count.

Proofs & Why It Matters

Tip
Proof: the mean value theorem from Rolle

Rolle: if ff is continuous on [a,b][a,b], differentiable on (a,b)(a,b), and f(a)=f(b)f(a) = f(b), then f(c)=0f'(c) = 0 for some cc.

Rolle: ff attains a max and min on [a,b][a,b] (extreme value theorem). If both occur at endpoints then ff is constant and any cc works; otherwise an interior extremum has f(c)=0f'(c) = 0 (Fermat: the difference quotients from the two sides have opposite signs). MVT: apply Rolle to g(x)=f(x)f(b)f(a)ba(xa)g(x) = f(x) - \tfrac{f(b) - f(a)}{b - a}(x - a), which has g(a)=g(b)=f(a)g(a) = g(b) = f(a); g(c)=0g'(c) = 0 gives f(c)=f(b)f(a)baf'(c) = \tfrac{f(b)-f(a)}{b-a}. \blacksquare Significance: the MVT is how every inequality between a function and its derivative is proved, and how Taylor's remainder is derived (apply it to the right auxiliary function).

Tip
Proof: (n!)1/nn1e\frac{(n!)^{1/n}}{n} \to \frac1e

A limit that looks like it needs Stirling and does not.

ln(n!)1/nn=1nk=1nlnklnn=1nk=1nlnkn\ln\tfrac{(n!)^{1/n}}{n} = \tfrac1n\sum_{k=1}^n\ln k - \ln n = \tfrac1n\sum_{k=1}^n\ln\tfrac kn, a Riemann sum for 01lnxdx\int_0^1\ln x\,dx. The integrand is unbounded at 00, so justify: the sum is a right-endpoint sum of a monotone function, hence squeezed between 01lnxdx=1\int_0^1\ln x\,dx = -1 and 1/n1lnxdx+1nln1n1\int_{1/n}^1\ln x\,dx + \tfrac1n\ln\tfrac1n \to -1. Exponentiate: the limit is e1e^{-1}. \blacksquare Significance: "log, Riemann sum, exponentiate" handles every limit of a geometric-mean shape — and handling the endpoint singularity honestly is what earns the 1010.

Going Deeper: Worked Problems

Example
Worked: an MVT inequality

Prove ln(1+x)x1+x\ln(1 + x) \ge \dfrac{x}{1 + x} for x>0x > 0, and find limnnln ⁣(1+1n)\displaystyle\lim_{n\to\infty}n\ln\!\left(1 + \tfrac1n\right).

Step 1 — MVT on ln\ln over [1,1+x][1, 1 + x]: ln(1+x)ln1=1cx\ln(1 + x) - \ln 1 = \tfrac1c\cdot x for some c(1,1+x)c \in (1, 1 + x), and 1c>11+x\tfrac1c > \tfrac{1}{1+x}, giving ln(1+x)>x1+x\ln(1+x) > \tfrac{x}{1+x}.

Step 2 — the same MVT also gives ln(1+x)<x\ln(1+x) < x (since 1c<1\tfrac1c < 1).

Step 3 — squeeze with x=1nx = \tfrac1n: 1/n1+1/n<ln(1+1n)<1n\tfrac{1/n}{1 + 1/n} < \ln(1 + \tfrac1n) < \tfrac1n, multiply by nn: nn+1<nln(1+1n)<1\tfrac{n}{n+1} < n\ln(1 + \tfrac1n) < 1, so the limit is 11.

Step 4 — exponentiate to recover (1+1n)ne(1 + \tfrac1n)^n \to e, now PROVED rather than assumed.

Example
Worked: Taylor's remainder bounds an error

Show cosx(1x22)x424\left|\cos x - \left(1 - \tfrac{x^2}{2}\right)\right| \le \dfrac{x^4}{24} for all real xx.

Step 1 — Taylor at 00 to order 33 with Lagrange remainder: cosx=1x22+0x3+cos(4)(ξ)4!x4\cos x = 1 - \tfrac{x^2}{2} + 0\cdot x^3 + \tfrac{\cos^{(4)}(\xi)}{4!}x^4 for some ξ\xi between 00 and xx.

Step 2 — cos(4)=cos\cos^{(4)} = \cos, and cosξ1|\cos\xi| \le 1, so the remainder is at most x424\tfrac{x^4}{24} in absolute value.

Step 3 — in fact cosξ0\cos\xi \ge 0 is not guaranteed, but the BOUND is; and since the next omitted term is +x424cosξ+\tfrac{x^4}{24}\cos\xi, we also get cosx1x22\cos x \ge 1 - \tfrac{x^2}{2} whenever cosξ0\cos\xi \ge 0 — e.g. for xπ2|x| \le \tfrac\pi2. The remainder term is a precise, citable error bar.