The Fundamental Theorem, Part 1
Watch rain collect in a gauge during a storm. The total caught so far climbs fast when the
rain is heavy, slows when it eases, and holds steady when it stops — the rate the
total rises is simply how hard it is raining at that instant. That everyday link
between a running total and its current rate is exactly the theorem on this page, and it is
the hinge the whole of calculus swings on.
Calculus has two halves, and for most of its history they were built as separate
subjects. One half measures slope — the
derivative,
born from tangent lines and instantaneous speed. The other measures area — the
definite integral,
born from chopping regions into slivers and adding them up. Tangents and areas: two problems that
look about as related as a speedometer and a warehouse inventory.
The Fundamental Theorem of Calculus throws open the connecting door. Part 1 says:
accumulating area, then differentiating, returns the integrand you started with.
Differentiation undoes integration. The two halves are one subject running in opposite
directions — and that single sentence is what turned a pile of clever tricks into the machine
called calculus.
Here is the object the theorem is about. Fix a continuous f and define
the accumulation function — the area so far, swept from a base point
a up to a moving right edge x:
F(x) = \int_{a}^{x} f(t)\,dt.
Notice the cast of characters: x is the input — where the right edge
of the shading currently sits — and t is a private, internal
dummy variable that runs across the shaded region and is used up by the integral. (We
write it as t precisely so it isn't confused with the limit
x.) Slide x right and
F grows; slide it back and F shrinks.
F is a perfectly ordinary function of x —
it just happens to be defined by an area.
The claim — the shock — is that this area function has a derivative, and that derivative is
exactly the function under the integral sign:
F'(x) = f(x).
Watch the theorem emerge: build F by hand
Before any proof, let's catch the theorem in the act. For simple integrands we can
compute the accumulated area with nothing but school geometry — and then differentiate the
result and see what comes out.
Example 1 — a constant. Take f(t) = 3 and base
point a = 0. The region from 0 to
x under a flat line of height 3 is a
rectangle: width x, height 3. So
F(x) = \int_0^x 3\,dt = 3x, \qquad\text{and}\qquad F'(x) = 3 = f(x).\;\checkmark
The derivative of the area function is the integrand. Once might be a coincidence.
Example 2 — a linear integrand. Take f(t) = t. The
region from 0 to x under the line
y = t is a right triangle with base x and
height x:
F(x) = \int_0^x t\,dt = \tfrac12 x^2, \qquad\text{and}\qquad F'(x) = x = f(x).\;\checkmark
Twice is a pattern. Note what happened to the degree: accumulating pushed
t up to \tfrac12 x^2, and differentiating
pulled it straight back down. The two operations are running the same staircase in opposite
directions.
Example 3 — where the theorem starts paying rent. Now take
f(t) = \cos t. There is no rectangle-and-triangle trick for
F(x) = \int_0^x \cos t\,dt,
and if you only knew Riemann sums, computing F'(x) would look
hopeless: a limit of a limit. But the theorem answers instantly, without computing the area
at all:
\frac{d}{dx}\int_0^x \cos t\,dt = \cos x.
No sums, no geometry — just read the integrand off and evaluate it at the upper limit. That is
the power move: FTC Part 1 turns "differentiate this area" from a research project into
a lookup.
-
The variable of integration is a dummy. In
F(x) = \int_0^x f(t)\,dt, the letter t
is pure bookkeeping — \int_0^x f(t)\,dt,
\int_0^x f(u)\,du and \int_0^x f(s)\,ds
are the same number. F depends only on
x. Never differentiate "with respect to t"
— by the time the integral is done, t no longer exists. (Writing
\int_0^x f(x)\,dx, with x doing double
duty, is the notational crime that causes this confusion — don't.)
-
The derivative is the integrand at the upper limit.
\frac{d}{dx}\int_a^x f(t)\,dt = f(x) — evaluated at
x, the moving edge, not at a and not
"at t". The base point a is frozen; it
shifts F by a constant and contributes nothing to the
derivative. \frac{d}{dx}\int_{7}^{x} f(t)\,dt is still exactly
f(x).
-
A moving limit that isn't plain x needs the chain
rule. If the upper limit is g(x), then
\frac{d}{dx}\int_a^{g(x)} f(t)\,dt = f(g(x))\cdot g'(x). Writing
f(g(x)) and forgetting the g'(x) is the
classic exam slip — see the "moving the goalposts" section below.
Why it's true: the sliver argument
The honest heart of the theorem fits in one picture. Ask how much new area
F gains when the right edge nudges from x
to x + h. The gain, F(x+h) - F(x), is a
skinny sliver of width h standing under the curve. For tiny
h the curve barely moves across the sliver, so the sliver is almost a
rectangle of height f(x):
F(x+h) - F(x) \;\approx\; f(x)\cdot h \qquad\Longrightarrow\qquad \frac{F(x+h)-F(x)}{h} \;\approx\; f(x).
Let h \to 0 and the "approximately" hardens into "equals":
F'(x) = f(x). Everything below is just this picture made airtight —
the word "almost a rectangle" replaced by a squeeze.
The proof — line by line
We compute F'(x) straight from the
limit definition
of the derivative. The whole argument is a squeeze.
Step 1 — form the difference quotient. By definition,
F'(x) = \lim_{h\to 0}\frac{F(x+h) - F(x)}{h}.
Step 2 — turn the numerator into one integral. Using additivity of the
definite integral, the area up to x+h minus the area up to
x is just the area of the thin sliver
[x, x+h]:
F(x+h) - F(x) = \int_a^{x+h} f(t)\,dt - \int_a^{x} f(t)\,dt = \int_{x}^{x+h} f(t)\,dt.
Step 3 — write the difference quotient as an average. Divide that sliver by its
width:
\frac{F(x+h) - F(x)}{h} = \frac{1}{h}\int_{x}^{x+h} f(t)\,dt.
The right side is the average value of f over the little
interval [x, x+h].
Step 4 — bound that sliver. Let m_h and
M_h be the smallest and largest values of f
on [x, x+h] (they exist because f is
continuous on a closed interval). The
bounding inequality
m_h\, h \le \int_x^{x+h} f \le M_h\, h, divided by
h > 0, gives
m_h \;\le\; \frac{1}{h}\int_{x}^{x+h} f(t)\,dt \;\le\; M_h.
Step 5 — squeeze as h \to 0. As the sliver shrinks
to the single point x, continuity of f
forces both the minimum and the maximum on it to converge to f(x):
\lim_{h\to 0} m_h = f(x) \qquad\text{and}\qquad \lim_{h\to 0} M_h = f(x).
Step 6 — apply the squeeze theorem. The difference quotient is trapped between
two quantities that both tend to f(x), so it has no choice but to do
the same:
\lim_{h\to 0}\frac{F(x+h) - F(x)}{h} = f(x).
Step 7 — read off the conclusion. The left side is
F'(x). Hence
\boxed{\;F'(x) = f(x).\;}
Differentiation undoes accumulation. (For h < 0 the same argument
runs with the inequalities and the average reversed, giving the identical limit, so the
two-sided derivative exists.)
Think of F(x) as a tank filling with water: at time
x, F is the total volume collected so
far. The instantaneous rate the tank fills is the height of the inflow at this
instant — which is exactly f(x). So F'(x),
the rate of change of accumulated area, equals the current height of the curve. Where
f is tall, area piles up fast and F rises
steeply; where f = 0, the area momentarily stalls and
F is flat; where f < 0, area is being
removed and F falls.
Step 5 is the only place we used continuity, and the proof leans its whole weight there. If
f had a jump at x, the slivers
approaching from the left and the right would average to different values, so
m_h and M_h would not both close on a
single number — the squeeze collapses, and F'(x) need not equal
f(x) there. (The accumulation function F
is still continuous — area can't jump — but it develops a corner, so it isn't differentiable
at the jump.) Continuity of the integrand is exactly what guarantees a clean derivative.
Let f be continuous on [a, b], and define
the accumulation function
F(x) = \int_{a}^{x} f(t)\,dt \qquad (a \le x \le b).
Then F is differentiable on (a, b) and
F'(x) = f(x) \qquad\text{for every } x \in (a, b).
In words: the accumulation function is an antiderivative of its integrand — every continuous
function has an antiderivative, namely its own area-so-far function.
The area filling, the slope matching
On top is f(t) = \tfrac12\,t with the area from
a = 0 up to the moving edge x shaded.
Below is the accumulation function F(x) = \int_0^{x}\tfrac12 t\,dt = \tfrac14 x^2,
with a dot riding along it at the current x. Slide
x: the shaded area on top equals the height of the dot
below, and the slope of F at the dot equals the height of
f on top — that is F'(x) = f(x), made
visible.
Two things worth hunting for as you drag. First, near x = 0 the
integrand is short, so area trickles in and the lower curve is nearly flat; out at
x = 4 the integrand is tall, area floods in, and
F climbs steeply — the lower graph's steepness is a live
readout of the upper graph's height. Second, watch the two printed numbers: the tangent's
slope always agrees with f(x), at every position, to the last decimal.
The theorem isn't an end-of-chapter fact; it's happening continuously, everywhere along the curve.
Moving the goalposts: a chain-rule upper limit
The theorem hands you the derivative when the upper limit is plain x.
But limits of integration love to be sneakier — say
G(x) = \int_0^{x^2} \cos t\,dt. This is a composition in disguise:
it is the accumulation function F(u) = \int_0^{u} \cos t\,dt with
u = x^2 plugged into it, so
G(x) = F(x^2). Differentiate with the chain rule and use
F'(u) = \cos u from FTC Part 1:
G'(x) = F'(x^2)\cdot \frac{d}{dx}\!\left(x^2\right) = \cos(x^2)\cdot 2x.
The recipe for any differentiable upper limit g(x): evaluate the
integrand at the limit, then multiply by the limit's derivative —
\frac{d}{dx}\int_a^{g(x)} f(t)\,dt \;=\; f\!\big(g(x)\big)\cdot g'(x).
And if x sits at the bottom? Flip the limits (which flips the
sign of the integral) and apply the theorem:
\frac{d}{dx}\int_x^{b} f(t)\,dt = -f(x). A limit at both ends splits
into two pieces through any convenient midpoint — one term per moving endpoint, each with its own
sign and chain factor.
Nobody woke up one morning and proved FTC. The pieces assembled across a generation. Isaac
Barrow — Newton's teacher at Cambridge, and the man who gave up his professorship so the young
Newton could have it — proved a geometric theorem in the 1660s that is essentially
FTC Part 1: the tangent to an area-curve is governed by the original curve's height.
But he stated it in the heavy costume of classical geometry, and seems not to have grasped that
he was holding a master key.
Newton and
Leibniz, working independently (and later
feuding bitterly about it), each saw what Barrow hadn't: that "areas undo tangents" is not a
curiosity about one curve but an engine. Instead of inventing a bespoke trick for every
area problem — the state of the art since Archimedes — you could solve them all at once by
running differentiation backwards. That reversal is why the subject is called calculus,
"a method of calculating": the theorem turned two thousand years of one-off ingenuity into an
algorithm. It is the reason this page's theorem gets the adjective fundamental.
See it explained