The fundamental link between change and accumulation

In brief

A car's speedometer and its odometer seem to be reporting two different things: one shows how fast the car is going right now, the other shows how far it has gone in total. Yet anyone who has driven knows the two are locked together. Add up the speed carefully enough over every instant and you get the distance; watch how the distance is changing at any moment and you get back the speed. Two operations that look unrelated, one that reports a rate and one that reports a running total, turn out to undo each other exactly. That fact is the fundamental theorem of calculus, and it is not a convenient coincidence. It is the reason integrals, which are naturally infinite sums of vanishingly small pieces, can instead be computed by a finite bit of arithmetic: evaluate one function at two points and subtract.

The full treatment

First look: the tank, the tap, and the gauge

Picture a tank fed by a tap whose flow rate wobbles over the course of an hour: fast at first, slowing, then surging again. Call the flow rate at time t the function f(t), measured in liters per minute. A gauge on the tank reports the accumulated volume V(t) at any moment. Two questions are obviously connected. If you know f at every instant, you should be able to work out V at any later time by tallying up all the water that has flowed in. And if you watch the gauge V(t) closely enough, its rate of rise at any instant should just be whatever the tap is doing right then, namely f(t). The volume and the flow rate are two views of one physical process. The puzzle is showing precisely how the view built from adding (the volume) relates to the view built from an instantaneous snapshot (the rate).

Two operations invented for different problems

Historically, these two questions were solved by entirely different methods. Finding a rate of change at an instant is what the derivative does: it is the limit of an average rate over a shrinking interval, a way of extracting a slope at a single point (see rate of change). Finding a running total is what the integral does: it is the limit of a sum of thin slices, each slice's contribution equal to a rate multiplied by a sliver of time (see accumulation and the integral). Nothing in the definition of either operation mentions the other. The derivative is built from division and a limit; the integral is built from multiplication and a sum. That they are related at all needs an argument.

The accumulation function and why its own rate is the original rate

Fix a starting time a and define an accumulation function A(x), equal to the integral of f from a up to x: the total amount that has flowed in by time x. Now ask for the derivative of A, using the derivative's own definition: look at how much A changes between x and x plus a small step h, and divide by h.

A(x+h) minus A(x) is exactly the water that flows in during the short interval from x to x+h. If f is continuous, then over that short interval f does not move far from its value f(x). So the strip of water added is trapped between h times the smallest value f takes on that little interval and h times the largest value it takes there. Divide through by h, and both of those bounds squeeze toward f(x) as h shrinks toward zero, because a continuous function's values on a shrinking interval crowd around its value at the point. So the limit of (A(x+h) minus A(x)) divided by h is exactly f(x). In symbols, A'(x) = f(x). The derivative of the accumulation function is the original rate. This is half of the fundamental theorem, and it is the more surprising half: accumulating and then differentiating gets you back exactly where you started, with no residue.

Undoing the derivative: recovering net change from a rate

The other half runs in reverse. Suppose F is any function whose derivative equals f, what is called an antiderivative of f. The accumulation function A built above is one such antiderivative, since A'(x) = f(x). Here is the key fact that finishes the argument: if two functions have the same derivative everywhere on an interval, they differ by a constant. This is because a function whose rate of change is zero everywhere cannot be rising or falling anywhere, so it must be flat. Since A and F have the same derivative f, their difference has derivative zero, so F(x) = A(x) + C for some constant C. Evaluate at x = a: A(a) is zero by definition (no time has passed to accumulate anything), so F(a) = C. Evaluate at x = b: F(b) = A(b) + C = A(b) + F(a). Rearranging, A(b) = F(b) minus F(a). But A(b) is exactly the integral of f from a to b. So the integral of f from a to b equals F(b) minus F(a), for any antiderivative F you happen to find. An infinite sum of slivers collapses into two evaluations and a subtraction.

Why this was ever a surprise

For roughly two thousand years, area problems (how much space a curved region encloses) and tangent problems (how steeply a curve rises at a point) were treated as unrelated branches of geometry, solved by separate ad hoc tricks for each new curve. The fundamental theorem says they were never separate: accumulation and instantaneous rate are inverse operations on the same underlying process, in the same sense that undoing a knot is the inverse of tying it.

Lineage

Area finding by exhaustive slicing goes back to Eudoxus and Archimedes, who computed areas and volumes for specific shapes by squeezing them between simpler figures, centuries before algebraic notation existed. Tangent line problems were sharpened in the 1600s by Fermat and Descartes. Isaac Barrow, Newton's teacher at Cambridge, came close to stating the inverse relationship geometrically around 1670. Newton and Leibniz, working independently in the 1660s through 1680s, turned that hint into a systematic method, giving calculus its calculating power. The arguments of that era relied on infinitesimals that were not fully rigorous by later standards. Augustin-Louis Cauchy, in his 1821 lectures, gave the integral a rigorous definition as a limit of sums and proved the connection to the derivative with the care modern mathematics expects.

The strongest case for it

The theorem's power is that it replaces an infinite process (summing ever-thinner slices) with a finite one (evaluate an antiderivative twice and subtract). This is why quantities that are naturally accumulations, distance from speed, work from force, charge from current, probability from a density, can be computed from a formula instead of estimated by brute-force slicing. It holds for the broad class of continuous functions used throughout science and engineering, and versions of it extend to functions with a finite number of jump discontinuities.

The strongest case against it

The theorem needs f to be integrable on the interval in question; for continuous functions this is automatic, but wild enough functions (unbounded, or discontinuous at infinitely many points in certain ways) require more delicate versions or fail outright. A separate and common misconception is thinking the theorem guarantees every antiderivative can be written down using familiar functions. It does not: the accumulation function of e to the power of negative x squared exists and is perfectly well defined by this theorem, but it has no expression in terms of ordinary algebraic, trigonometric, or exponential functions. The theorem guarantees the inverse relationship, not that the answer will look tidy.

Where it stands now

This is settled, foundational mathematics, part of the broad consensus underlying all of calculus. What has changed since Cauchy is not the theorem's truth but the reach of the word integral itself: Riemann and later Lebesgue built more general definitions of the integral precisely to extend the theorem's reach to rougher functions, and in each generalization the core inverse relationship between accumulating and differentiating survives intact.

Test yourself

A reactant's concentration in a beaker is falling at a rate described only by a rough graph: fast at first, nearly flat in the middle, then dropping quickly near the end of a ten-minute run. You are not given a formula for the rate, only the graph. Explain how you would estimate the total amount of reactant consumed over the ten minutes using the ideas in this entry, and then explain what it would mean, physically, to differentiate your accumulated-consumption estimate and check that you recover the original rate graph. Finally, identify one feature the rate graph could have, a sudden vertical jump, for instance, that would force you to justify your method more carefully before trusting it.