The value a function approaches as its input gets arbitrarily close to a point.
IntuitionWhat happens right next to a point
Suppose we cannot evaluate a formula right at x=x0 — for instance the difference quotient hf(x0+h)−f(x0) at h=0, which gives 00. Even though the point itself is off-limits, we can walk arbitrarily close to x0 from both sides and watch what value f(x) settles toward. That target value is the limit limx→x0f(x)=L, and it does not care whether f(x0) even exists.
Curve with a fixed point x0 and a second point x0+h; as h shrinks toward 0, the secant line rotates into the tangent line.
Secant slope hf(x0+h)−f(x0): slide h→0 and watch the secant line settle onto the tangent — a limit at h=0 where the formula itself is undefined.
SchoolFormal definition and fundamental limits
Definition: Limit of a function at a point (ε–δ definition)
Let f be defined on an open interval around x0, possibly excluding x0 itself. We write limx→x0f(x)=L if for every tolerance ε>0 there is a radius δ>0 such that ∀ε>0,∃δ>0:0<∣x−x0∣<δ⇒∣f(x)−L∣<ε. Notice the strict inequality 0<∣x−x0∣: x is never allowed to equal x0 in the test.
∀ε>0,∃δ>0:0<∣x−x0∣<δ⇒∣f(x)−L∣<ε
Here ε is the vertical tolerance around L on the output axis, and δ (delta) is the horizontal window around x0 on the input axis. Two foundational limits — proved by geometry and by monotone bounds — unlock all derivatives of trigonometric, exponential, and logarithmic functions: limx→0xsinx=1 (with x in radians) and limx→∞(1+x1)x=e.
A two-sided limit exists if and only if both one-sided limits exist and are equal: limx→x0f(x)=L⟺limx→x0−f(x)=limx→x0+f(x)=L.
Why is it true?
Walking toward a doorway from the left hallway and from the right hallway only lands you in the same room if both hallways actually meet at the same door.
Proof
(⇒) Suppose limx→x0f(x)=L. Given ε>0, the ε–δ definition gives δ>0 such that 0<∣x−x0∣<δ⇒∣f(x)−L∣<ε.
In particular, if x0<x<x0+δ (the right-hand condition) or x0−δ<x<x0 (the left-hand condition), we automatically have 0<∣x−x0∣<δ, so ∣f(x)−L∣<ε. Thus both one-sided limits equal L.
(⇐) Conversely, suppose both one-sided limits equal L. For a given ε>0, the left-hand limit gives δ1>0 working on (x0−δ1,x0), and the right-hand limit gives δ2>0 working on (x0,x0+δ2).
Set δ=min(δ1,δ2)>0. Then any x with 0<∣x−x0∣<δ lies either in (x0−δ1,x0) or in (x0,x0+δ2), and in both cases ∣f(x)−L∣<ε. Hence limx→x0f(x)=L. ■
On a unit circle, a very small arc of length x, its vertical chord leg sinx, and the outer tangent segment tanx become visually indistinguishable as the angle closes toward 0.
Proof
Take 0<x<2π on the unit circle with center O, point A=(1,0), point B=(cosx,sinx) on the circle, and point T=(1,tanx) where the ray OB meets the vertical tangent at A.
The triangle △OAB sits strictly inside the circular sector OAB, which sits strictly inside the right triangle △OAT. Comparing their areas gives 21sinx<21x<21tanx.
Since sinx>0 on (0,2π), multiply through by sinx2: 1<sinxx<cosx1. Taking reciprocals reverses the inequalities: cosx<xsinx<1.
As x→0+, cosx→1, so the Squeeze Theorem forces limx→0+xsinx=1. Finally, g(x)=xsinx is an even function (g(−x)=−xsin(−x)=−x−sinx=g(x)), so limx→0−xsinx=1 as well, giving limx→0xsinx=1. ■
UndergraduateReal-World Applications and Worked Examples
Function limits turn average rates into instantaneous rates in physics (velocity, acceleration, electric current), justify the paraxial small-angle approximation sinθ≈θ in geometric optics (camera lenses, telescopes) and pendulum mechanics, and turn discrete compounding into continuous compounding via limx→∞(1+x1)x=e.
Example
A falling stone has position s(t)=5t2 (in meters) at time t (in seconds). Using a limit of average velocities over [2,2+h], find its instantaneous velocity at t=2 s.
Solution
On the time interval [2,2+h] with h=0, the average velocity is vavg(h)=hs(2+h)−s(2)=h5(2+h)2−5⋅22.
Expand the numerator: 5(4+4h+h2)−20=20h+5h2=h(20+5h).
Since h=0 in the limit, cancel h: vavg(h)=20+5h.
Now take the limit as h→0: v(2)=limh→0(20+5h)=20 m/s. Notice how the 00 indeterminate form disappears once we simplify before taking the limit.
Example
In paraxial optics, lens designers replace sinθ in Snell's law n1sinθ1=n2sinθ2 with θ (in radians) thanks to limx→0xsinx=1. Using the squeeze bound cosx<xsinx<1, show that for an incident ray at θ=0.1 rad (≈5.7∘), the relative error sinθθ−sinθ is less than 1%.
Solution
From cosx<xsinx<1 for 0<θ<2π, taking reciprocals gives 1<sinθθ<cosθ1.