Navigation

Limits & Continuity · Topic 02 of 23

Introduction to Limits

limxaf(x)=L\lim_{x\to a}f(x)=L says f(x)f(x) gets as close to LL as demanded once xx is close enough to aa — with xax\neq a, so f(a)f(a) never matters. The limit exists exactly when both one-sided limits exist and agree; a jump, a blow-up to ±\pm\infty or an oscillation means it does not, and the ε\varepsilon-δ\delta definition makes all of this precise.

5 min readCore — on every final7 formulas6 quick checks
01

Key ideas

5 things to remember
  1. 1

    The limit never looks at x=ax=a

    limxaf(x)=L\lim_{x\to a}f(x)=L means f(x)f(x) stays as close to LL as demanded once xx is close enough to aa, with xax\neq a. Whether f(a)f(a) equals LL, differs from it, or is undefined changes nothing.

  2. 2

    Both one-sided limits must agree

    limxa\lim_{x\to a^-} uses only x<ax<a; limxa+\lim_{x\to a^+} only x>ax>a. limxaf(x)=L\lim_{x\to a}f(x)=L exactly when both equal LL. At a domain endpoint (x\sqrt{x} at 00) only the one-sided limit is posed.

  3. 3

    Three ways a limit fails

    A jump (one-sided limits differ), a blow-up (1x2\frac{1}{x^2} or 1x\frac1x at 00: a vertical asymptote), or an oscillation (sin1x\sin\frac1x at 00 sweeps [1,1][-1,1] forever). Writing ==\infty records how it fails; \infty is not a number.

  4. 4

    Circles give limits, dots give values

    Slide a finger along the curve from the left, then from the right, and report the height each branch heads toward — an open circle still counts. A solid dot gives f(a)f(a) only; never jump to it.

  5. 5

    ε\varepsilon-δ\delta: every tolerance must be met

    For every ε>0\varepsilon>0 there must be a δ>0\delta>0 with 0<xa<δf(x)L<ε0<|x-a|<\delta\Rightarrow|f(x)-L|<\varepsilon. δ\delta may depend on ε\varepsilon and aa, never on xx. One δ\delta for one ε\varepsilon proves nothing.

-2-1123412345xyL = 2hole: the limit is 2f(2) = 3.5y = f(x)
f(x)=x24+1f(x)=\frac{x^2}{4}+1 for x2x\neq2, but f(2)=3.5f(2)=3.5. Both branches head to height 22, so limx2f(x)=2\lim_{x\to2}f(x)=2. The solid dot is f(2)f(2) and plays no part in the limit — cover it with your thumb and nothing changes.
12341234xyx = 2left limit 2f(2) = 3 = right limity = x²/2 for x < 2y = 5 − x for x ≥ 2
A jump at x=2x=2: limx2f(x)=2\lim_{x\to2^-}f(x)=2 and limx2+f(x)=3\lim_{x\to2^+}f(x)=3. Both one-sided limits exist but disagree, so limx2f(x)\lim_{x\to2}f(x) does not exist. The solid dot says f(2)=3f(2)=3 — a third fact, independent of either limit.
02

Formulas

What to have memorised
  • Two-sided from one-sided

    limxaf(x)=L    limxaf(x)=L=limxa+f(x)\lim_{x\to a}f(x)=L\iff\lim_{x\to a^-}f(x)=L=\lim_{x\to a^+}f(x)

    Both must exist and agree; either failure kills the two-sided limit. At a domain endpoint only one side is posed.

  • ε\varepsilon-δ\delta definition of limxaf(x)=L\lim_{x\to a}f(x)=L

    ε>0  δ>0:0<xa<δ  f(x)L<ε\forall\,\varepsilon>0\ \ \exists\,\delta>0:\quad 0<|x-a|<\delta\ \Longrightarrow\ |f(x)-L|<\varepsilon

    ε\varepsilon is the opponent's vertical tolerance; δ\delta is your horizontal reply, written in terms of ε\varepsilon only.

  • One-sided limit, precisely

    limxa+f(x)=L    ε>0  δ>0: a<x<a+δ  f(x)L<ε\lim_{x\to a^+}f(x)=L\iff\forall\,\varepsilon>0\ \ \exists\,\delta>0:\ a<x<a+\delta\ \Longrightarrow\ |f(x)-L|<\varepsilon

    Left-hand: use aδ<x<aa-\delta<x<a instead. The superscript names the side xx comes from, not the sign of ff.

  • δ\delta for a linear function

    δ=εmfor f(x)=mx+b, m0\delta=\frac{\varepsilon}{|m|}\quad\text{for } f(x)=mx+b,\ m\neq0

    Because f(x)(ma+b)=mxa|f(x)-(ma+b)|=|m|\,|x-a|. If m=0m=0 the function is constant and any δ>0\delta>0 works.

  • δ\delta with a nuisance factor

    δ=min{1, εK}where xa<1g(x)<K\delta=\min\left\{1,\ \frac{\varepsilon}{K}\right\}\quad\text{where } |x-a|<1\Rightarrow|g(x)|<K

    For f(x)L=xag(x)|f(x)-L|=|x-a|\,|g(x)|: restrict first, bound the other factor, then take the minimum.

  • Infinite limit

    limxaf(x)=    M>0  δ>0: 0<xa<δf(x)>M\lim_{x\to a}f(x)=\infty\iff\forall\,M>0\ \ \exists\,\delta>0:\ 0<|x-a|<\delta\Longrightarrow f(x)>M

    A description of failure — the limit does not exist, because \infty is not a real number.

  • Showing a limit does not exist

    xna,  tna,f(xn)f(tn)c>0  limxaf(x) does not existx_n\to a,\ \ t_n\to a,\quad |f(x_n)-f(t_n)|\ge c>0\ \Longrightarrow\ \lim_{x\to a}f(x)\ \text{does not exist}

    Two families of inputs whose outputs stay apart: for sin1x\sin\frac1x take xn=12πnx_n=\frac{1}{2\pi n}, tn=12πn+π/2t_n=\frac{1}{2\pi n+\pi/2}.

03

Write an $\varepsilon$-$\delta$ proof

The steps, in order
  1. 1

    Scratch work first: simplify f(x)L|f(x)-L| until xa|x-a| appears as a factor — mxa|m|\,|x-a| for a line, xax+a|x-a|\,|x+a| for x2x^2.

  2. 2

    If the other factor still contains xx, agree δ1\delta\le1: then xa<1|x-a|<1 pins xx to an interval and bounds that factor by a constant KK.

  3. 3

    Choose δ\delta: ε/m\varepsilon/|m| for a line, min{1,ε/K}\min\{1,\varepsilon/K\} otherwise. It may involve ε\varepsilon and aa, never xx.

  4. 4

    Write the proof: "Let ε>0\varepsilon>0. Put δ=\delta=\dots. Suppose 0<xa<δ0<|x-a|<\delta. Then f(x)L=<=ε|f(x)-L|=\dots<\dots=\varepsilon."

  5. 5

    Close with "since ε>0\varepsilon>0 was arbitrary". For lim=\lim=\infty, swap in: given M>0M>0, find δ\delta forcing f(x)>Mf(x)>M.

04

Watch out

The mistakes that cost marks
  • ✗ Wrong

    limxaf(x)=f(a)\lim_{x\to a}f(x)=f(a), automatically

    ✓ Right

    The limit never inspects f(a)f(a). That equation is continuity at aa — a property to verify, not part of the definition.

    Why: f(a)f(a) may differ from the limit or not exist at all.

  • ✗ Wrong

    "limx01x2=\lim_{x\to0}\frac{1}{x^2}=\infty, so the limit exists."

    ✓ Right

    It does not exist; "==\infty" only records how it fails.

    Why: \infty is not a real number, and existence needs a real LL.

  • ✗ Wrong

    δ=εx+3\delta=\dfrac{\varepsilon}{|x+3|}

    ✓ Right

    δ\delta may not mention xx. Restrict δ1\delta\le1, bound x+3<7|x+3|<7, then take δ=min{1,ε/7}\delta=\min\{1,\varepsilon/7\}.

    Why: xx is quantified after δ\delta; δ\delta must be fixed before xx is chosen.

  • ✗ Wrong

    "The right-hand limit is 44, so limxaf(x)=4\lim_{x\to a}f(x)=4."

    ✓ Right

    Both one-sided limits must exist and agree — unless aa is a domain endpoint, where only one side is posed.

  • ✗ Wrong

    sinπx\sin\frac{\pi}{x} at x=0.1,0.01,0.001x=0.1,\,0.01,\,0.001 gives 0,0,00,0,0, so the limit is 00

    ✓ Right

    Those samples all sit on zeros of sine; limx0sinπx\lim_{x\to0}\sin\frac{\pi}{x} does not exist.

    Why: A table samples finitely many points; a limit is a claim about all of them.

  • ✗ Wrong

    "For ε=0.06\varepsilon=0.06, δ=0.02\delta=0.02 works. QED."

    ✓ Right

    Give δ\delta as a formula in ε\varepsilon (δ=ε/3\delta=\varepsilon/3) and run the argument for an arbitrary ε>0\varepsilon>0.

    Why: The definition is a promise about every tolerance, not one convenient one.

05

Quick check

Commit to an answer before you reveal one
  1. Q1easy

    Let f(x)={x+1,x<25,x=25x,x>2f(x)=\begin{cases}x+1, & x<2\\ 5, & x=2\\ 5-x, & x>2\end{cases} Find (a) limx2f(x)\lim_{x\to2^-}f(x), (b) limx2+f(x)\lim_{x\to2^+}f(x), (c) limx2f(x)\lim_{x\to2}f(x), (d) f(2)f(2). What is the relationship between (c) and (d)?

    Hint

    Each one-sided limit uses only the formula valid on its side of 22; the value at 22 is a separate question.

    Show answer

    Answer

    (a) 33 (b) 33 (c) 33 (d) f(2)=5f(2)=5. The limit exists but differs from the value: a removable discontinuity.

    Steps

    (a) For x<2x<2 the rule is x+1x+1: limx2f(x)=2+1=3\lim_{x\to2^-}f(x)=2+1=3.

    (b) For x>2x>2 the rule is 5x5-x: limx2+f(x)=52=3\lim_{x\to2^+}f(x)=5-2=3.

    (c) Both one-sided limits exist and equal 33, so limx2f(x)=3\lim_{x\to2}f(x)=3.

    (d) The middle line applies only at x=2x=2: f(2)=5f(2)=5.

    The limit is computed from x2x\neq2 and never sees f(2)f(2), so 353\neq5 is perfectly legal. Redefining f(2)=3f(2)=3 would make ff continuous there — that is why the discontinuity is called removable.

  2. Q2easy

    The graph of hh on [3,4][-3,4] has three pieces: a segment from the solid point (3,1)(-3,-1) up to an open circle at (1,3)(1,3); a single solid dot at (1,0)(1,0); a segment from an open circle at (1,1)(1,1) up to the solid point (4,4)(4,4). Find limx1h(x)\lim_{x\to1^-}h(x), limx1+h(x)\lim_{x\to1^+}h(x), limx1h(x)\lim_{x\to1}h(x), h(1)h(1) and limx4h(x)\lim_{x\to4^-}h(x). Why is limx4h(x)\lim_{x\to4}h(x) not asked?

    Hint

    An open circle still tells you the height a branch is heading toward; only the solid dot gives a value.

    Show answer

    Answer

    33; 11; does not exist (jump); h(1)=0h(1)=0; 44. Nothing is defined for x>4x>4, so only the one-sided limit is meaningful there.

    Steps

    Left piece: through (3,1)(-3,-1) and (1,3)(1,3), slope 11, so h(x)=x+2h(x)=x+2 for 3x<1-3\le x<1. Right piece: through (1,1)(1,1) and (4,4)(4,4), so h(x)=xh(x)=x for 1<x41<x\le4.

    limx1h(x)=1+2=3\lim_{x\to1^-}h(x)=1+2=3 — the circle at (1,3)(1,3) being open is irrelevant. limx1+h(x)=1\lim_{x\to1^+}h(x)=1. The one-sided limits differ, so limx1h(x)\lim_{x\to1}h(x) does not exist. h(1)=0h(1)=0 from the solid dot: left limit, right limit and value are three independent numbers.

    limx4h(x)=4\lim_{x\to4^-}h(x)=4. A two-sided limit needs hh defined on both sides of 44; there is no graph for x>4x>4, so only the left-hand limit is posed.

  3. Q3medium

    Let f(x)=cx+3f(x)=cx+3 for x<1x<1 and f(x)=x2cxf(x)=x^2-cx for x1x\ge1, where cc is a constant. (a) Find every cc for which limx1f(x)\lim_{x\to1}f(x) exists, and give the limit. (b) For that cc, compare f(1)f(1) with the limit. (c) If c=0c=0, find both one-sided limits and describe how the limit fails.

    Hint

    Write each one-sided limit as an expression in cc, then force them to agree.

    Show answer

    Answer

    (a) c=1c=-1 only; the limit is 22. (b) f(1)=2f(1)=2, equal to the limit. (c) Left limit 33, right limit 11: a jump of size 22.

    Steps

    (a) limx1f(x)=c+3\lim_{x\to1^-}f(x)=c+3 and limx1+f(x)=1c\lim_{x\to1^+}f(x)=1-c. The two-sided limit exists exactly when c+3=1cc=1,c+3=1-c\quad\Longrightarrow\quad c=-1, and then both sides equal 22.

    (b) The second rule covers x=1x=1: f(1)=1(1)(1)=2f(1)=1-(-1)(1)=2. Limit equals value, so ff is continuous at 11 when c=1c=-1.

    (c) With c=0c=0: f(x)=3f(x)=3 for x<1x<1 and f(x)=x2f(x)=x^2 for x1x\ge1, so the one-sided limits are 33 and 11. Both exist but disagree — a jump, not a blow-up or an oscillation.

  4. Q4medium

    Let F(x)=1x2F(x)=\dfrac{1}{x-2} and G(x)=1(x2)2G(x)=\dfrac{1}{(x-2)^2}. (a) Find limx2F(x)\lim_{x\to2^-}F(x) and limx2+F(x)\lim_{x\to2^+}F(x); what does this say about limx2F(x)\lim_{x\to2}F(x)? (b) Prove from the definition that limx2G(x)=\lim_{x\to2}G(x)=\infty. (c) Does limx2G(x)\lim_{x\to2}G(x) exist?

    Hint

    For (a) track the sign of x2x-2 on each side. For (b) solve G(x)>MG(x)>M for x2|x-2|.

    Show answer

    Answer

    (a) -\infty and ++\infty; limx2F(x)\lim_{x\to2}F(x) does not exist (vertical asymptote). (b) δ=1/M\delta=1/\sqrt{M} works. (c) No — "==\infty" describes the failure; \infty is not a real number.

    Steps

    (a) As x2x\to2^-, x2x-2 is negative and tiny, so F(x)F(x)\to-\infty; as x2+x\to2^+ it is positive and tiny, so F(x)+F(x)\to+\infty. Neither one-sided limit is a real number, and they run to opposite infinities, so limx2F(x)\lim_{x\to2}F(x) does not exist; x=2x=2 is a vertical asymptote.

    (b) Let M>0M>0 and put δ=1M\delta=\dfrac{1}{\sqrt{M}}. If 0<x2<δ0<|x-2|<\delta then 0<(x2)2<δ2=1MG(x)=1(x2)2>M.0<(x-2)^2<\delta^2=\frac{1}{M}\quad\Longrightarrow\quad G(x)=\frac{1}{(x-2)^2}>M. Since M>0M>0 was arbitrary, limx2G(x)=\lim_{x\to2}G(x)=\infty.

    (c) No. Existence requires a real LL; the symbol \infty records that GG exceeds every bound near 22.

  5. Q5medium

    Let f(x)=3x+2f(x)=3x+2. (a) Find the largest δ>0\delta>0 such that 0<x1<δ0<|x-1|<\delta implies f(x)5<0.06|f(x)-5|<0.06. (b) Repeat for ε=0.003\varepsilon=0.003. (c) Give δ\delta as a formula in an arbitrary ε>0\varepsilon>0 and write the full proof that limx1(3x+2)=5\lim_{x\to1}(3x+2)=5.

    Hint

    Simplify f(x)5|f(x)-5| first; for a linear function it is a constant times x1|x-1|.

    Show answer

    Answer

    (a) δ=0.02\delta=0.02 (b) δ=0.001\delta=0.001 (c) δ=ε/3\delta=\varepsilon/3.

    Steps

    f(x)5=3x3=3x1|f(x)-5|=|3x-3|=3|x-1|.

    (a) 3x1<0.06    x1<0.023|x-1|<0.06\iff|x-1|<0.02, so the largest δ\delta is 0.020.02 (any smaller positive δ\delta also works). (b) x1<0.001|x-1|<0.001: δ=0.001\delta=0.001.

    (c) Let ε>0\varepsilon>0 be given. Put δ=ε/3>0\delta=\varepsilon/3>0. Suppose 0<x1<δ0<|x-1|<\delta. Then (3x+2)5=3x1<3δ=ε.|(3x+2)-5|=3|x-1|<3\delta=\varepsilon. Since ε>0\varepsilon>0 was arbitrary, limx1(3x+2)=5\lim_{x\to1}(3x+2)=5. Check: ε=0.06\varepsilon=0.06 gives δ=0.02\delta=0.02, matching (a).

  6. Q6hard

    Prove from the ε\varepsilon-δ\delta definition that limx2x2=4\lim_{x\to2}x^2=4. Say exactly where the preliminary restriction δ1\delta\le1 is used and why it is needed.

    Hint

    Factor x24x^2-4. You control x2|x-2| directly; the factor x+2|x+2| must be bounded first.

    Show answer

    Answer

    δ=min{1,ε/5}\delta=\min\{1,\varepsilon/5\}. The restriction δ1\delta\le1 is used only to get x+2<5|x+2|<5; without it x+2|x+2| could be arbitrarily large.

    Steps

    Scratch: x24=x2x+2|x^2-4|=|x-2|\,|x+2|. If x2<1|x-2|<1 then 1<x<31<x<3, so x+2<5|x+2|<5 and x24<5x2|x^2-4|<5|x-2|.

    Proof. Let ε>0\varepsilon>0. Put δ=min{1,ε/5}>0\delta=\min\{1,\varepsilon/5\}>0 and suppose 0<x2<δ0<|x-2|<\delta. Since δ1\delta\le1, x+2=(x2)+4x2+4<5|x+2|=|(x-2)+4|\le|x-2|+4<5. Since δε/5\delta\le\varepsilon/5, x2<ε/5|x-2|<\varepsilon/5. Hence x24=x2x+2<ε55=ε.|x^2-4|=|x-2|\,|x+2|<\frac{\varepsilon}{5}\cdot5=\varepsilon. Since ε>0\varepsilon>0 was arbitrary, limx2x2=4\lim_{x\to2}x^2=4.

    The restriction is used only for the bound x+2<5|x+2|<5. Without it xx could be far from 22, making x+2|x+2| huge and the product large even when x2|x-2| is small. Any restriction works: δ12\delta\le\tfrac12 gives x+2<4.5|x+2|<4.5 and δ=min{12,2ε9}\delta=\min\{\tfrac12,\tfrac{2\varepsilon}{9}\}.

06

On the exam

How this topic is marked
  • "Find the limit or explain why it does not exist" is marked on the reason: write both one-sided values, then "they differ, so the limit does not exist" — or name the blow-up or oscillation.

  • An ε\varepsilon-δ\delta proof is marked line by line: "Let ε>0\varepsilon>0", δ\delta as a formula in ε\varepsilon, "suppose 0<xa<δ0<|x-a|<\delta", then a chain of inequalities ending in <ε<\varepsilon. Scratch work alone earns nothing.

  • Never write limxaf(x)=f(a)\lim_{x\to a}f(x)=f(a) without a reason. Here the reason is "both one-sided limits equal f(a)f(a)"; later it will be continuity.

Keep going