Continuity at a Point
You have known what continuity means since you were about six years old. Put a pen on paper
and trace a curve: if you can draw the whole thing without lifting the pen,
the curve is continuous. Lift the pen — to hop over a gap, to jump to a different height —
and it isn't. Rollercoaster tracks and the temperature through an afternoon are
pen-never-lifts curves; a staircase, or a stock price at the moment bad news lands, is not.
Calculus keeps the childhood picture but sharpens it into something you can actually
check. Why bother? Because the pen picture misleads for exactly the functions where
it matters. A graph can look unbroken on a calculator screen and hide a missing point one
pixel wide. A function can be given by a formula too wild to draw at all — try sketching
\sin(1/x) near zero and your pen gives up long before the
mathematics does. And "I drew it in one stroke" is not an argument you can write in a proof.
We need a test that works on formulas, not drawings — and it turns out to be a
checklist of exactly three conditions, all built from the
limits you
already own.
A function f is continuous at x = c when all three of the following hold:
- f(c) exists — the function has an actual value at c;
- \displaystyle\lim_{x \to c} f(x) exists — the curve is heading somewhere definite, and both sides agree on where;
- they are equal: \displaystyle\lim_{x \to c} f(x) = f(c).
Read the three conditions as a tiny story. Condition 1 says the point is there.
Condition 2 says the curve knows where it's going. Condition 3 is the punchline:
the curve is going exactly where the point is. Where the function is heading and
where it actually lands are the same place — no surprise on arrival. That third line is also
why direct
substitution works: "just plug in c" computes
f(c), and it gives the limit precisely when the two are equal —
which is the definition of continuity. Substitution isn't a lucky trick; it is continuity in
disguise.
The three-part test, live
Here is the checklist in action. All four functions below use the same left half —
the line y = x approaching c = 2 — but
each treats the point c differently. Flip between the cases and,
before reading the verdict, run the checklist yourself: is there a dot at
c? Do the two sides of the curve aim at the same height? Does the
dot sit at that height?
Notice how each broken case fails at a different line of the checklist. The hole has
a curve that aims perfectly at height 2 — condition 2 passes — but
there's no dot to arrive at: condition 1 fails. The jump has a dot, but the two halves of the
curve can't agree on a destination, so the limit doesn't exist: condition 2 fails. The
infinite case fails even harder — no value and no limit. One picture, three distinct
diseases; the checklist is the diagnosis.
Worked example 1: a polynomial passes with full marks
Is f(x) = x^2 + 1 continuous at x = 2?
Run the checklist in order — each line is a genuine, separate check:
-
Condition 1 — does f(2) exist? Substitute:
f(2) = 2^2 + 1 = 5. A perfectly good number. ✓
-
Condition 2 — does \lim_{x \to 2} f(x) exist?
By the limit laws, \lim_{x \to 2}(x^2 + 1) = \left(\lim_{x \to 2} x\right)^2 + 1
= 2^2 + 1 = 5. Both sides of 2 agree on
5. ✓
-
Condition 3 — are they equal?
\lim_{x \to 2} f(x) = 5 = f(2). ✓
Verdict: continuous at 2 — the pen never lifts.
And nothing about the point 2 was special: the same three lines go
through for any polynomial at any point, because the limit laws push a limit
through sums, products and powers of x. That is the real content of
the slogan "polynomials are continuous everywhere". The interesting functions, then,
are the ones built with seams — piecewise definitions, fractions whose denominators can vanish
— and that's where the checklist earns its keep.
Three ways to break
When a function is not continuous at c, we call
c a discontinuity — and the type of
discontinuity is simply a record of which condition failed, building directly on
when a limit doesn't
exist. Learn to name them on sight:
-
Removable (a hole) — the limit exists, but f(c)
is missing, or sits at the wrong height (a "relocated dot"). Everything about the curve is
fine except one single point. As the name promises, you could repair it by
redefining that one point — worked example 3 below does exactly that.
-
Jump — the two one-sided limits both exist but disagree, so the
two-sided limit doesn't exist (condition 2 fails). The graph steps from one height to
another like a staircase. No single value of f(c) could ever
patch it: wherever you put the dot, at least one side of the curve refuses to meet it.
-
Infinite — the function blows up near c (a
vertical asymptote), so the limit doesn't exist in the most dramatic way possible. Postage
due: conditions 1 and 2 usually both fail.
The chart below shows the crudest of the three — a jump at
x = 0. Trace it with an imaginary pen: coming in from the left you
arrive at height 2; the curve resumes at height
0.5. The pen must lift, and the size of the lift —
1.5 here — is exactly the gap between the one-sided limits. (Flip
back to the live figure above to compare it against the hole and the blow-up.)
Worked example 2: diagnosing three patients
Three piecewise functions, each unwell at one point. For each, run the checklist and name the
disease.
Patient (a):
h(x) = \begin{cases} x^2 & x \neq 1 \\ 5 & x = 1 \end{cases}
Check at c = 1. Condition 1: h(1) = 5 —
defined. ✓ Condition 2: for every x near
1 (but not equal to it) the rule is x^2,
so \lim_{x \to 1} h(x) = 1 — exists. ✓ Condition 3:
1 \neq 5. ✗ The curve aims at height 1
but the dot has been relocated up to 5. Diagnosis:
removable discontinuity — two conditions pass, only the match fails.
Patient (b):
g(x) = \begin{cases} x + 1 & x < 2 \\ x + 3 & x \ge 2 \end{cases}
Check at c = 2. Condition 1: g(2) = 5 —
defined. ✓ Condition 2: from the left, \lim_{x \to 2^-} (x+1) = 3;
from the right, \lim_{x \to 2^+} (x+3) = 5. The sides disagree
(3 \neq 5), so the limit does not exist. ✗ Stop — once condition 2
fails, condition 3 is meaningless (there is no limit to compare against). Diagnosis:
jump discontinuity, of size 2.
Patient (c):
k(x) = \frac{1}{(x-4)^2}
Check at c = 4. Condition 1: k(4) = \tfrac{1}{0}
— undefined. ✗ Condition 2: as x \to 4 from either side,
(x-4)^2 is a tiny positive number, so k(x) \to +\infty
— the limit does not exist either. ✗ Diagnosis: infinite discontinuity. Note
the severity ordering: patient (a) failed one condition, (b) effectively one, and (c) failed
everything. That ordering matters for the next question — which patients can be cured?
Worked example 3: repairing a removable discontinuity
Here is a function you will meet constantly in calculus — a fraction whose top and bottom both
vanish at the same point:
f(x) = \frac{x^2 - 9}{x - 3}.
Step 1 — find the sore point. At x = 3 the
denominator is 0, so f(3) is undefined:
condition 1 fails, and f is not continuous at
3. But how badly is it broken?
Step 2 — take the limit anyway. For every x \neq 3
we may factor and cancel:
\frac{x^2 - 9}{x - 3} = \frac{(x-3)(x+3)}{x-3} = x + 3,
so \lim_{x \to 3} f(x) = 3 + 3 = 6. The limit exists! Away from the
sore point, f is the straight line
y = x + 3 — a perfectly healthy function with one point punched
out. That is the signature of a removable discontinuity: condition 2 passes, only the value is
missing.
Step 3 — prescribe the cure. The curve is aiming at height
6; all that's missing is a dot there. So define one. Build
the repaired function
\tilde f(x) = \begin{cases} \dfrac{x^2-9}{x-3} & x \neq 3 \\[2pt] 6 & x = 3 \end{cases}
and re-run the checklist at 3: \tilde f(3) = 6
exists ✓, the limit is still 6 ✓ (limits never care about the value
at the point, only near it), and 6 = 6 ✓. Healed. This is
why "removable" is exactly the right word: the whole defect was one point's worth of missing
(or misplaced) data, and redefining that single point removes it. The recipe is
always the same — compute L = \lim_{x \to c} f(x), then
declare f(c) := L. Note what it cannot fix: a
jump has no single L to aim for, and a blow-up has none at all.
Only the mildest disease is curable.
-
All three conditions are genuinely needed — including the third. The
sneakiest failure is patient (a) above: f(c) exists and
the limit exists, yet they simply disagree — a curve aiming at one height with its
dot relocated to another. Two out of three is not continuity. If you only ever check "is it
defined?" and "does the limit exist?", the relocated dot sails straight past you.
-
"Defined at c" is not the same as "continuous at
c". The jump function of patient (b) is perfectly well
defined at 2 — g(2) = 5, no missing
value anywhere — and still discontinuous there. Having a value is one condition out of
three, not the whole test.
-
Continuity is a pointwise property. The definition is "continuous
at c" — one point at a time. A function is continuous
where the checklist passes and discontinuous where it fails, and both can be true of the
same function at different points: f(x) = 1/x is continuous at
every point except 0. So beware the phrase "a continuous
function" used as if it were a species. It is shorthand for "continuous at every point of
its domain" — and an exam answer that says where a function is and isn't continuous
will always beat one that issues a blanket verdict.
Watch Sal define continuity
Sal Khan walks through the same three-part definition with the epsilon-free pictures used
here — a good second pass, especially on telling the hole and the relocated dot apart.
Here is a claim about your own life that mathematics can prove: at some instant in your
childhood, your height was exactly one metre. Not roughly — exactly, to infinitely
many decimal places. How can anyone know that without a time machine and a very fussy ruler?
Because height is continuous in time — you grow, but you never teleport from
99 cm to 101 cm without passing through every height in between. Once a function is continuous
on an interval, it inherits a superpower called the Intermediate Value Theorem:
a continuous function that starts below a value and ends above it must hit that value
somewhere along the way. You were 50 cm at birth and are more than a metre now; one metre lies
in between; therefore some instant caught you at precisely one metre. The pen picture makes it
obvious — a pen travelling from below a line to above it must cross the line — but the theorem
only holds because all three checklist conditions hold at every point. Give the
function a single jump and the guarantee dies: a staircase can step over any value it likes.
The same superpower pulls off a party trick with the weather: at every moment there are two
points on exactly opposite sides of the Earth with precisely the same temperature
(compare each equator point with its opposite; the difference starts positive, ends negative,
so somewhere it is exactly zero). And it is why engineers trust root-finding: a continuous
function that is negative here and positive there is guaranteed to solve
f(x) = 0 in between. Continuity is the fine print on that
guarantee — which is why mathematicians fuss so much about checking it.
The checklist, one last time
Faced with "is f continuous at c?", work
in order and stop at the first failure — each later condition only means anything if the
earlier ones passed:
- 1. Value: compute f(c). Undefined? Not continuous. (If the limit exists anyway, the hole is removable — you know the cure.)
- 2. Limit: compute \lim_{x \to c^-} f(x) and \lim_{x \to c^+} f(x). Disagree, or infinite? Not continuous — a jump or a blow-up, and no redefinition can save it.
- 3. Match: equal or not? Equal means continuous; unequal means a relocated dot — removable again.
Everything on this page compresses into that triage — and it feeds directly into what comes
next: continuity at a point is the entry requirement for
having a
derivative there, where a fourth, stricter condition (no corners!) joins the
list.