Chapter IV · Limits of functions of a positive integral variable

Limits of sums, products and quotients

§§63–68 45 min

If φ(n) → a and ψ(n) → b then φ + ψ → a + b, φψ → ab and φ/ψ → a/b (when b ≠ 0); with these rules the limit of any rational function of n can be read off.

Why this matters

Proving every limit straight from the definition would be exhausting. §§63–68 proves, once and for all, that limits respect the arithmetic operations. After that, 3n2−12n+52n2+4n+4→32\dfrac{3n^2 - 12n + 5}{2n^2 + 4n + 4} \to \tfrac32 needs no δ\delta at all: it follows from 1/n→01/n \to 0 and a few rules. The proofs are also the first place where the definition is used as a tool rather than checked by hand, and the device they use, splitting the tolerance, appears in almost every later proof in analysis.

The idea

If ϕ(n)\phi(n) is nearly aa and ψ(n)\psi(n) is nearly bb, surely ϕ(n)+ψ(n)\phi(n) + \psi(n) is nearly a+ba + b. The only question is how nearly. The error of the sum is at most the sum of the errors:

∣(ϕ(n)+ψ(n))−(a+b)∣≤∣ϕ(n)−a∣+∣ψ(n)−b∣.|(\phi(n) + \psi(n)) - (a + b)| \le |\phi(n) - a| + |\psi(n) - b|.

So if we want the sum within δ\delta of a+ba + b, it is enough to have each term within 12δ\tfrac12\delta of its own limit. Each of those happens for large nn, say from n1n_1 and from n2n_2 on, so both happen from n0=max⁡(n1,n2)n_0 = \max(n_1, n_2) on.

Explore Split the tolerance

Try this. With “δ/2 each”, move δ\delta and watch n1n_1, n2n_2 and n0=max⁡(n1,n2)n_0 = \max(n_1, n_2). Then switch to “δ each” on the pair with errors of the same sign.

ϕ(n)=1+1n→1\phi(n) = 1 + \frac1n \to 1, ψ(n)=2+1n→2\psi(n) = 2 + \frac1n \to 2
φ(n) with band 1 ± 0.1; n₁ = 11
nφ
ψ(n) with band 2 ± 0.1; n₂ = 11
nψ
φ(n) + ψ(n) with band 3 ± 0.2; n₀ = max(n₁, n₂) = 11
nφ+ψ
  • n ≥ threshold, inside band
  • n ≥ threshold, outside
  • n₁, n₂, n₀
Tolerance for each term
Pair
Every term of the sum from n0=11n_0 = 11 to n=4000n = 4000 lies within δ\delta of a+b=3a + b = 3. This is guaranteed: ∣ϕ+ψ−(a+b)∣≤∣ϕ−a∣+∣ψ−b∣<12δ+12δ|\phi+\psi-(a+b)| \le |\phi - a| + |\psi - b| < \tfrac12\delta + \tfrac12\delta.
What to notice

With half the tolerance each, the sum is always inside its band from n0n_0 on: that is the triangle inequality at work. With the whole δ\delta for each term, two errors of the same sign, each just under δ\delta, add up to nearly 2δ2\delta, and the sum escapes. Halving is not a trick of style; it is what makes the argument correct.

Products need more care, because an error gets multiplied by the other factor. Writing ϕ(n)=a+ϕ1(n)\phi(n) = a + \phi_1(n) and ψ(n)=b+ψ1(n)\psi(n) = b + \psi_1(n), where ϕ1,ψ1→0\phi_1, \psi_1 \to 0 are the errors,

ϕ(n)ψ(n)−ab=a ψ1(n)+b ϕ1(n)+ϕ1(n)ψ1(n).\phi(n)\psi(n) - ab = a\,\psi_1(n) + b\,\phi_1(n) + \phi_1(n)\psi_1(n).

Each of the three terms on the right is small when the errors are small, so the product tends to abab. Quotients come from products once we know 1/ψ(n)→1/b1/\psi(n) \to 1/b, and the key point there is that ψ(n)\psi(n) stays away from 00 when b≠0b \neq 0.

Precise statements

Theorem Limit of a sum Hardy: Theorem I, §63

If ϕ(n)→a\phi(n) \to a and ψ(n)→b\psi(n) \to b, then ϕ(n)+ψ(n)→a+b\phi(n) + \psi(n) \to a + b.

Theorem Limit of a product Hardy: Theorem II, §65

If ϕ(n)→a\phi(n) \to a and ψ(n)→b\psi(n) \to b, then ϕ(n)ψ(n)→ab\phi(n)\psi(n) \to ab. In particular kϕ(n)→kak\phi(n) \to ka for every constant kk.

Theorem Limit of a reciprocal Hardy: Theorem III, §66

If ϕ(n)→a\phi(n) \to a and a≠0a \neq 0, then 1/ϕ(n)→1/a1/\phi(n) \to 1/a.

Since ϕ(n)→a≠0\phi(n) \to a \ne 0, ϕ(n)\phi(n) can be 00 for at most finitely many nn (the proof shows ∣ϕ(n)∣>12∣a∣|\phi(n)| > \tfrac12|a| for large nn), and finitely many terms do not matter, so 1/ϕ(n)1/\phi(n) makes sense where it counts.

Theorem Limit of a quotient Hardy: Theorem IV, §66

If ϕ(n)→a\phi(n) \to a and ψ(n)→b\psi(n) \to b with b≠0b \neq 0, then ϕ(n)/ψ(n)→a/b\phi(n)/\psi(n) \to a/b.

Theorem Rational functions of limits Hardy: Theorem V, §67

Let R=P/QR = P/Q, where PP and QQ are polynomials in several variables. If ϕ(n)→a\phi(n) \to a, ψ(n)→b\psi(n) \to b, χ(n)→c,…\chi(n) \to c, \dots, and Q(a,b,c,… )≠0Q(a, b, c, \dots) \ne 0, then

R{ϕ(n),ψ(n),χ(n),… }→R(a,b,c,… ).R\{\phi(n), \psi(n), \chi(n), \dots\} \to R(a, b, c, \dots).
Theorem Limit of a rational function of nn Hardy: §68

Let S(n)=a0np+a1np−1+⋯+apb0nq+b1nq−1+⋯+bqS(n) = \dfrac{a_0 n^p + a_1 n^{p-1} + \dots + a_p}{b_0 n^q + b_1 n^{q-1} + \dots + b_q} with a0≠0a_0 \ne 0, b0≠0b_0 \ne 0. Then S(n)→0S(n) \to 0 if p<qp < q; S(n)→a0/b0S(n) \to a_0/b_0 if p=qp = q; and if p>qp > q, S(n)→+∞S(n) \to +\infty or −∞-\infty according as a0/b0a_0/b_0 is positive or negative.

Proofs

Proof Theorem I: limit of a sum
Strategy

Given δ\delta, ask each sequence to come within 12δ\tfrac12\delta of its limit, take the later of the two thresholds, and add the two errors with the triangle inequality.

Let δ>0\delta > 0; get two thresholds

Since ϕ(n)→a\phi(n) \to a, applying the definition with 12δ\tfrac12\delta (also a positive number) gives n1n_1 with ∣ϕ(n)−a∣<12δ|\phi(n) - a| < \tfrac12\delta for n≥n1n \ge n_1. Similarly n2n_2 with ∣ψ(n)−b∣<12δ|\psi(n) - b| < \tfrac12\delta for n≥n2n \ge n_2.

Let n0=max⁡(n1,n2)n_0 = \max(n_1, n_2)

For n≥n0n \ge n_0 both inequalities hold.

Add the errors

For n≥n0n \ge n_0, by the triangle inequality, ∣ϕ(n)+ψ(n)−(a+b)∣≤∣ϕ(n)−a∣+∣ψ(n)−b∣<12δ+12δ=δ|\phi(n) + \psi(n) - (a + b)| \le |\phi(n) - a| + |\psi(n) - b| < \tfrac12\delta + \tfrac12\delta = \delta.

Proof Theorem II: limit of a product
Strategy

Split ϕψ−ab\phi\psi - ab into three pieces using the errors ϕ1=ϕ−a\phi_1 = \phi - a, ψ1=ψ−b\psi_1 = \psi - b. Make both errors smaller than a number η\eta chosen so that the three pieces together are below δ\delta.

The identity

ϕψ−ab=(a+ϕ1)(b+ψ1)−ab=aψ1+bϕ1+ϕ1ψ1\phi\psi - ab = (a + \phi_1)(b + \psi_1) - ab = a\psi_1 + b\phi_1 + \phi_1\psi_1, so ∣ϕψ−ab∣≤∣a∣∣ψ1∣+∣b∣∣ϕ1∣+∣ϕ1∣∣ψ1∣|\phi\psi - ab| \le |a||\psi_1| + |b||\phi_1| + |\phi_1||\psi_1|.

Choose η=min⁡{1, δ/(∣a∣+∣b∣+1)}\eta = \min\{1,\ \delta/(|a| + |b| + 1)\}

Then η>0\eta > 0 and η≤1\eta \le 1. As in Theorem I there is an n0n_0 such that ∣ϕ1(n)∣<η|\phi_1(n)| < \eta and ∣ψ1(n)∣<η|\psi_1(n)| < \eta for all n≥n0n \ge n_0.

Estimate

For n≥n0n \ge n_0: ∣ϕψ−ab∣<∣a∣η+∣b∣η+η⋅η≤(∣a∣+∣b∣+1)η≤δ|\phi\psi - ab| < |a|\eta + |b|\eta + \eta\cdot\eta \le (|a| + |b| + 1)\eta \le \delta, using η2≤η\eta^2 \le \eta (as η≤1\eta \le 1). The first inequality is strict because ∣ϕ1∣∣ψ1∣<η2|\phi_1||\psi_1| < \eta^2, so this works even when aa or bb is 00.

The special case

A constant kk is a sequence tending to kk, so kϕ(n)→kak\phi(n) \to ka.

Hardy’s own version takes ∣ϕ1∣<δ/(3∣b∣)|\phi_1| < \delta/(3|b|) and ∣ψ1∣<δ/(3∣a∣)|\psi_1| < \delta/(3|a|), assumes δ<3∣a∣∣b∣\delta < 3|a||b|, and leaves the cases a=0a = 0 or b=0b = 0 to the reader; the choice of η\eta above covers all cases at once.

Proof Theorem III: limit of a reciprocal
Strategy

∣1ϕ−1a∣=∣ϕ−a∣∣a∣∣ϕ∣\left|\dfrac{1}{\phi} - \dfrac1a\right| = \dfrac{|\phi - a|}{|a||\phi|}. The top is small; the danger is a small denominator. So first keep ϕ(n)\phi(n) away from 00, then make the top small.

Keep ϕ(n)\phi(n) away from 00

Apply the definition with δ=12∣a∣>0\delta = \tfrac12|a| > 0: for n≥n1n \ge n_1, ∣ϕ(n)−a∣<12∣a∣|\phi(n) - a| < \tfrac12|a|, so ∣ϕ(n)∣≥∣a∣−∣ϕ(n)−a∣>12∣a∣|\phi(n)| \ge |a| - |\phi(n) - a| > \tfrac12|a|. In particular ϕ(n)≠0\phi(n) \ne 0.

Bound the error

For n≥n1n \ge n_1: ∣1ϕ(n)−1a∣=∣ϕ(n)−a∣∣a∣∣ϕ(n)∣<2∣ϕ(n)−a∣∣a∣2\left|\dfrac{1}{\phi(n)} - \dfrac1a\right| = \dfrac{|\phi(n) - a|}{|a||\phi(n)|} < \dfrac{2|\phi(n) - a|}{|a|^2}.

Make it less than δ\delta

Given δ>0\delta > 0, choose n2n_2 with ∣ϕ(n)−a∣<12∣a∣2δ|\phi(n) - a| < \tfrac12|a|^2\delta for n≥n2n \ge n_2. For n≥max⁡(n1,n2)n \ge \max(n_1, n_2) the error is less than δ\delta.

Theorem IV is now immediate: ϕ/ψ=ϕ⋅(1/ψ)→a⋅(1/b)\phi/\psi = \phi \cdot (1/\psi) \to a \cdot (1/b) by Theorems III and II. Theorem V follows by building RR in stages: each monomial Aϕpψq⋯A\phi^p\psi^q\cdots tends to Aapbq⋯Aa^pb^q\cdots by Theorem II applied finitely many times, so PP tends to P(a,b,… )P(a, b, \dots) by Theorem I applied finitely many times; the same holds for QQ, and Theorem IV finishes the job because Q(a,b,… )≠0Q(a, b, \dots) \ne 0.

The rational function of nn in §68 is the most important application. Try it first.

Explore Only the top degrees matter

Try this. Set p=q=2p = q = 2 and look at the first 2020 terms, then 600600. Then try p=3,q=2p = 3, q = 2 with a0a_0 negative, and p=1,q=3p = 1, q = 3.

S(n)=n2−12n+52n2+4n+4S(n) = \dfrac{n^{2}-12n+5}{2n^{2}+4n+4}
nS(n)
  • S(n)
  • (a₀/b₀) n^(p−q)
  • the limit
Since p=qp = q: S(n)→a0/b0=0.5S(n) \to a_0/b_0 = 0.5. The first terms can point the wrong way (the lower coefficients are chosen to mislead), but for large nn the quotient behaves like the dashed curve (a0/b0) np−q(a_0/b_0)\, n^{p-q}. At n=60n = 60: S(n)=0.38756S(n) = 0.38756.
What to notice

The lower coefficients (such as the −12n-12n on top) can make the early terms point the wrong way, but they lose in the end. Only p−qp - q and the sign of a0/b0a_0/b_0 decide the behaviour, and the terms hug the dashed curve (a0/b0) np−q(a_0/b_0)\,n^{p-q}.

Proof §68: limit of a rational function of nn
Strategy

Divide the top by npn^p and the bottom by nqn^q. What remains is a rational function of 1/n1/n, which tends to a0/b0a_0/b_0 by Theorem V, multiplied by np−qn^{p-q}, whose behaviour we know.

Rewrite

S(n)=np−q T(n)S(n) = n^{p-q}\,T(n), where T(n)=a0+a1/n+⋯+ap/npb0+b1/n+⋯+bq/nqT(n) = \dfrac{a_0 + a_1/n + \dots + a_p/n^p}{b_0 + b_1/n + \dots + b_q/n^q}.

T(n)→a0/b0T(n) \to a_0/b_0

T(n)=R(1/n)T(n) = R(1/n) for the rational function R(x)=(a0+a1x+⋯+apxp)/(b0+b1x+⋯+bqxq)R(x) = (a_0 + a_1x + \dots + a_px^p)/(b_0 + b_1x + \dots + b_qx^q). Since 1/n→01/n \to 0 and the denominator at x=0x = 0 is b0≠0b_0 \ne 0, Theorem V gives T(n)→R(0)=a0/b0T(n) \to R(0) = a_0/b_0.

p≤qp \le q

If p<qp < q, np−q=1/nq−p→0n^{p-q} = 1/n^{q-p} \to 0 (Examples XXIII, 1), so S(n)→0⋅a0/b0=0S(n) \to 0 \cdot a_0/b_0 = 0 by Theorem II. If p=qp = q, np−q=1n^{p-q} = 1 and S(n)=T(n)→a0/b0S(n) = T(n) \to a_0/b_0.

p>qp > q, a0/b0>0a_0/b_0 > 0

For large nn, T(n)>12a0/b0=c>0T(n) > \tfrac12 a_0/b_0 = c > 0 (use the definition with δ=c\delta = c). Then S(n)>c np−q≥c nS(n) > c\,n^{p-q} \ge c\,n, which exceeds any Δ\Delta once n>Δ/cn > \Delta/c. So S(n)→+∞S(n) \to +\infty. If a0/b0<0a_0/b_0 < 0, apply this to −S(n)-S(n).

When the limits are infinite or do not exist

Theorems I–IV assume that both limits exist. §64 works out what can be said about ϕ+ψ\phi + \psi in the other cases, and the pattern is worth knowing. Some combinations are determined; others are indeterminate, meaning the behaviour of the sum depends on the particular sequences.

ϕ\phiψ\psiϕ+ψ\phi + \psi
limit→±∞\to\pm\infty, or oscillates (finitely or infinitely)behaves like ψ\psi
→+∞\to +\infty→+∞\to +\infty, or oscillates finitely→+∞\to +\infty
→+∞\to +\infty→−∞\to -\inftyanything at all
→+∞\to +\inftyoscillates infinitely→+∞\to +\infty or oscillates infinitely
oscillates finitelyoscillates finitelylimit or oscillates finitely
oscillates finitelyoscillates infinitelyoscillates infinitely
oscillates infinitelyoscillates infinitelyanything at all

For example, with ϕ→+∞\phi \to +\infty and ψ→−∞\psi \to -\infty: n+(−n)→0n + (-n) \to 0, n2−n→+∞n^2 - n \to +\infty, n−n2→−∞n - n^2 \to -\infty, (n+(−1)n)−n(n + (-1)^n) - n oscillates finitely, and (n2+(−1)nn)−n2(n^2 + (-1)^n n) - n^2 oscillates infinitely. Products have a similar table (§65), with “tends to 00” as a separate case, since 0×0 \times (anything bounded) →0\to 0.

Worked examples and exercises

Exercise Four rational sequences ●○○ Hardy: Examples XXVI, 1

Find the behaviour of (n−1n+1)2\left(\dfrac{n-1}{n+1}\right)^2, (−1)n(n−1n+1)2(-1)^n\left(\dfrac{n-1}{n+1}\right)^2, n2+1n\dfrac{n^2+1}{n}, (−1)nn2+1n(-1)^n\dfrac{n^2+1}{n}.

Solution

(n−1)/(n+1)(n-1)/(n+1) has p=q=1p = q = 1 and a0/b0=1a_0/b_0 = 1, so it tends to 11, and its square to 11 (Theorem II). With the factor (−1)n(-1)^n, the even terms tend to 11 and the odd ones to −1-1; the sequence is bounded by 11, so it oscillates finitely (as for (−1)n(-1)^n, no ll is within 12\tfrac12 of both 11 and −1-1). (n2+1)/n(n^2+1)/n has p=2>q=1p = 2 > q = 1 and a0/b0=1>0a_0/b_0 = 1 > 0, so it tends to +∞+\infty. With (−1)n(-1)^n, the sizes tend to +∞+\infty and the signs alternate: it oscillates infinitely.

Exercise Which tend to a limit? ●●○ Hardy: Examples XXVI, 2

Which of these tend to a limit?

1cos⁡212nπ+nsin⁡212nπ,1n(cos⁡212nπ+nsin⁡212nπ),ncos⁡212nπ+sin⁡212nπn(cos⁡212nπ+nsin⁡212nπ).\frac{1}{\cos^2\frac12 n\pi + n\sin^2\frac12 n\pi},\qquad \frac{1}{n(\cos^2\frac12 n\pi + n\sin^2\frac12 n\pi)},\qquad \frac{n\cos^2\frac12 n\pi + \sin^2\frac12 n\pi}{n(\cos^2\frac12 n\pi + n\sin^2\frac12 n\pi)}.
Hint
For even nn, cos⁡212nπ=1\cos^2\tfrac12 n\pi = 1 and sin⁡212nπ=0\sin^2\tfrac12 n\pi = 0; for odd nn it is the other way round.
Solution

For even nn the three terms are 11, 1/n1/n and n/n=1n/n = 1. For odd nn they are 1/n1/n, 1/n21/n^2 and 1/n21/n^2. So the first oscillates finitely (even terms 11, odd terms tending to 00), the second tends to 00 (all terms are at most 1/n1/n), and the third oscillates finitely (even terms 11, odd terms tending to 00). Only the second tends to a limit.

Exercise Shifting a rational function ●●○ Hardy: Examples XXVI, 3

With S(n)S(n) the general rational function of §68, show that S(n+1)/S(n)→1S(n+1)/S(n) \to 1.

Hint
Write S(n)=np−qT(n)S(n) = n^{p-q}T(n) with T(n)→a0/b0≠0T(n) \to a_0/b_0 \ne 0.

Solution

S(n+1)S(n)=(n+1n)p−qT(n+1)T(n)\dfrac{S(n+1)}{S(n)} = \left(\dfrac{n+1}{n}\right)^{p-q} \dfrac{T(n+1)}{T(n)} for large nn (when S(n)≠0S(n) \ne 0). Here (n+1)/n=1+1/n→1(n+1)/n = 1 + 1/n \to 1, so its (p−q)(p-q)-th power tends to 11 (Theorems II and III, applied ∣p−q∣|p - q| times). Also T(n+1)→a0/b0T(n+1) \to a_0/b_0 and T(n)→a0/b0≠0T(n) \to a_0/b_0 \ne 0, so their quotient tends to 11 by Theorem IV. By Theorem II the product tends to 11.

Exercise Infinite plus bounded ●○○ Hardy: §64, 2

Prove that if ϕ(n)→+∞\phi(n) \to +\infty and ψ(n)\psi(n) oscillates finitely, then ϕ(n)+ψ(n)→+∞\phi(n) + \psi(n) \to +\infty.

Solution

Finite oscillation gives KK with ∣ψ(n)∣<K|\psi(n)| < K, so ϕ(n)+ψ(n)>ϕ(n)−K\phi(n) + \psi(n) > \phi(n) - K. Given Δ\Delta, choose n0n_0 with ϕ(n)>Δ+K\phi(n) > \Delta + K for n≥n0n \ge n_0 (Definition II with Δ+K\Delta + K). Then ϕ(n)+ψ(n)>Δ\phi(n) + \psi(n) > \Delta for n≥n0n \ge n_0. The same proof works whenever ψ\psi is bounded, in particular when ψ\psi tends to a limit.

Exercise Infinite times finitely oscillating ●●○ Hardy: §65 (i)

Let ϕ(n)=n\phi(n) = n. Find the behaviour of ϕ(n)ψ(n)\phi(n)\psi(n) for ψ(n)=2+(−1)n\psi(n) = 2 + (-1)^n, −2−(−1)n-2 - (-1)^n and (−1)n(-1)^n, and explain why these are the only three possibilities when ϕ→+∞\phi \to +\infty and ψ\psi oscillates finitely.

Solution

n(2+(−1)n)≥n→+∞n(2 + (-1)^n) \ge n \to +\infty; n(−2−(−1)n)≤−n→−∞n(-2 - (-1)^n) \le -n \to -\infty; (−1)nn(-1)^n n oscillates infinitely. In general ϕψ\phi\psi cannot tend to a limit or oscillate finitely: since ψ\psi oscillates finitely it does not tend to 00, so there are ε>0\varepsilon > 0 and infinitely many nn with ∣ψ(n)∣≥ε|\psi(n)| \ge \varepsilon (by the negation theorem of “For large values of n”); at those nn, ∣ϕ(n)ψ(n)∣≥εϕ(n)|\phi(n)\psi(n)| \ge \varepsilon\phi(n), which is unbounded. An unbounded sequence tends to ±∞\pm\infty or oscillates infinitely.

Check yourself

Check yourself Choose one

In the proof of Theorem I, why do we ask for ∣ϕ(n)−a∣<12δ|\phi(n) - a| < \tfrac12\delta rather than <δ< \delta?

Why

The triangle inequality bounds the error of the sum by the sum of the errors. With δ\delta each we would only get an error below 2δ2\delta. Definition I may be applied with any positive number in place of δ\delta, including 12δ\tfrac12\delta; that freedom is what the proof uses.

Check yourself True or false?

If ϕ(n)+ψ(n)\phi(n) + \psi(n) tends to a limit, then ϕ(n)\phi(n) and ψ(n)\psi(n) both tend to limits.

Why

False. ϕ(n)=(−1)n\phi(n) = (-1)^n and ψ(n)=−(−1)n\psi(n) = -(-1)^n both oscillate, but their sum is 00. Theorem I runs in one direction only.

Check yourself Enter a number

What is lim⁡3n2−12n+52n2+4n+4\lim \dfrac{3n^2 - 12n + 5}{2n^2 + 4n + 4}?

Why

Here p=q=2p = q = 2, so the limit is the ratio of the leading coefficients, 3/23/2. Dividing top and bottom by n2n^2 gives (3−12/n+5/n2)/(2+4/n+4/n2)→3/2(3 - 12/n + 5/n^2)/(2 + 4/n + 4/n^2) \to 3/2.

Check yourself Choose one

ϕ(n)→+∞\phi(n) \to +\infty and ψ(n)→−∞\psi(n) \to -\infty. What can you say about ϕ(n)+ψ(n)\phi(n) + \psi(n)?

Why

∞−∞\infty - \infty is indeterminate: n−n→0n - n \to 0, n2−n→+∞n^2 - n \to +\infty, n−n2→−∞n - n^2 \to -\infty, (n+(−1)n)−n(n + (-1)^n) - n oscillates finitely and (n2+(−1)nn)−n2(n^2 + (-1)^n n) - n^2 oscillates infinitely.

Hardy’s language

Summary

Summary
  • Limits respect ++, −-, ×\times and ÷\div (when the limit of the denominator is not 00), and therefore every rational function of convergent sequences.
  • The method: split the tolerance (12δ\tfrac12\delta each), take the larger threshold (max⁡(n1,n2)\max(n_1, n_2)), and combine the errors with the triangle inequality.
  • For reciprocals, first keep the sequence away from 00: ∣ϕ(n)∣>12∣a∣|\phi(n)| > \tfrac12|a| for large nn.
  • A rational function of nn behaves like (a0/b0) np−q(a_0/b_0)\,n^{p-q}: it tends to 00, to a0/b0a_0/b_0, or to ±∞\pm\infty as p<qp < q, p=qp = q or p>qp > q.
  • When a limit is infinite or does not exist, some combinations are determined and others (∞−∞\infty - \infty, two infinite oscillations) can do anything.