Accumulation and the integral

In brief

A car's speed is not constant over a long trip, it climbs on the highway, drops through town, stops at lights. You cannot find the total distance traveled by multiplying one speed by the whole trip's duration, because there is no single speed to multiply. Yet the total distance is a perfectly real, perfectly well-defined number, the odometer reads it at the end. The integral is the tool built to recover exactly this kind of total from a rate that will not sit still: it works by slicing the whole trip into pieces short enough that the speed barely changes within each piece, adding up the small distance covered in every piece, and then asking what that sum approaches as the pieces are sliced ever thinner. The same method recovers total mass from a density that varies through a material, or total energy from a power output that varies over time, whenever a fixed rate can no longer be multiplied by a fixed span to get the right answer.

The full treatment

First look: distance from a speed that will not sit still

Suppose a car's speed, in meters per second, is given by v of t, equal to 2t, so the car is accelerating steadily, its speed doubling the time in seconds. How far has it traveled between t equals 0 and t equals 4 seconds? If the speed were constant, distance would just be speed times time, but here the speed itself is changing throughout the interval. Split the four seconds into four one-second pieces. On each piece, approximate the speed as constant, using its value at the start of that piece: from 0 to 1 second, speed is roughly v of 0, equal to 0, contributing 0 meters; from 1 to 2 seconds, speed is roughly v of 1, equal to 2, contributing 2 meters; from 2 to 3, speed is roughly v of 2, equal to 4, contributing 4 meters; from 3 to 4, speed is roughly v of 3, equal to 6, contributing 6 meters. Adding these gives 0 plus 2 plus 4 plus 6, equal to 12 meters, a first estimate. Splitting into eight half-second pieces instead, using the same method, gives a different, closer estimate, since the speed changes less within each shorter piece. As the pieces get thinner and thinner, the estimate settles down toward a single number, and that limiting number is the exact distance traveled, 16 meters in this case, which can be checked against the exact answer once the fundamental link between the derivative and the integral is established in the next stage of this sequence.

Building the idea: slices, sums, and the shrinking-width limit

Generalize the process used above. Take an interval from a to b over which some rate or density, call it f of x, is varying. Divide the interval into n equal pieces, each of width delta-x, equal to (b minus a) divided by n. In each piece, pick a representative point, say the left endpoint of that piece, and evaluate f there; multiplying that value by the piece's width delta-x gives an estimate of that piece's contribution to the total, treating the rate as if it were constant across the tiny piece. Adding up all n of these small contributions gives a sum, called a Riemann sum, which approximates the true total. As n grows larger, so that delta-x shrinks toward zero, this Riemann sum, under the controlled-approach conditions established earlier, settles toward a single limiting value, provided f behaves reasonably, such as being continuous, on the interval. That limiting value is the definite integral of f from a to b, and it is the exact total accumulated quantity, not an approximation, precisely because the approximation error introduced by treating each thin slice as constant shrinks to zero as the slices themselves shrink to zero width.

What the integral recovers, and what it assumes

The integral answers a specific question: given a rate or a density that is known at every point of an interval, what is the exact total it produces over that whole interval. It assumes the rate or density is known as a function of the variable being integrated over, and that this function is well behaved enough, at minimum bounded and with only a limited number of breaks, for the Riemann sum limit to actually settle down to a single value rather than wander or blow up. If the underlying rate itself is not known or cannot be written as a function, the integral cannot be computed, though it may still exist in principle, and this is an important practical boundary: real measured data, sampled only at a finite number of points, gives only an approximation to the true integral, exactly the Riemann sum itself, rather than the limiting exact value.

Formal model: the definite integral defined precisely

State the definition with every symbol named. Let f be a function defined and bounded on the interval from a to b. Partition the interval into n pieces of width delta-x-i, each not necessarily equal in general, though equal widths are the simplest case, and in each piece choose a sample point x-i-star. Form the Riemann sum: the sum, over all n pieces, of f evaluated at x-i-star, times delta-x-i. The definite integral of f from a to b, written as the integral sign with a at the bottom and b at the top, of f of x, dx, is defined as the limit of this Riemann sum as the largest piece width shrinks toward zero, provided that limit exists and does not depend on how the pieces or sample points were chosen. Geometrically, when f is positive, each term f of x-i-star times delta-x-i is the area of a thin rectangle of height f of x-i-star and width delta-x-i, so the Riemann sum is the total area of a stack of thin rectangles approximating the region under the curve of f, and the integral itself is the exact area under that curve, the limit as the rectangles become infinitesimally thin.

A worked accumulation: mass from a varying density

Consider a metal rod of length 3 meters whose density varies along its length, given by the function d of x, equal to 2 plus x, in kilograms per meter, where x measures distance from one end. To find the rod's total mass, slice it into thin pieces of width delta-x; each piece's mass is approximately its density at that point times its small width, d of x-i-star times delta-x. Summing over all pieces and taking the limit as the pieces shrink gives the exact total mass as the definite integral of d of x, from 0 to 3. Approximating first with three one-meter pieces, using left endpoints, gives d of 0 times 1, plus d of 1 times 1, plus d of 2 times 1, equal to 2 plus 3 plus 4, equal to 9 kilograms, a rough estimate; refining the slices further would sharpen this toward the exact value, exactly as the falling-speed and accelerating-car examples sharpened toward their exact distances.

Lineage

The idea of finding an area or a total by summing many thin pieces goes back to the ancient Greek method of exhaustion, used by Eudoxus and famously by Archimedes to compute the area of a circle and the volume of a sphere by inscribing and circumscribing polygons and shrinking their difference toward zero, over two thousand years before calculus proper existed. Isaac Newton, in the 1671 Method of Fluxions, treated the recovery of a total quantity from its instantaneous rate as one half of his calculus, paired directly with the derivative as its counterpart. The definite integral did not receive a fully rigorous definition as a limit of sums, independent of any appeal to infinitesimals, until Augustin-Louis Cauchy's work in the 1820s, later refined and generalized by Bernhard Riemann in the mid-nineteenth century into the Riemann sum formulation used in this entry and standard in calculus courses today.

The strongest case for it

The integral succeeds precisely where simple multiplication of a rate by a time span fails: any time a quantity accumulates from a rate or a density that genuinely varies, whether distance from a changing speed, mass from a varying density, work from a varying force, or total revenue from a varying sales rate, the integral gives the exact total, not merely an estimate, and it does so by a single unified method, slice, approximate, sum, and shrink, that works identically regardless of what the varying quantity represents physically. Because the Riemann sum construction is built directly on the controlled-approach limit established earlier in this sequence, the integral inherits the same rigor: for well-behaved functions, the limit is guaranteed to exist and to be independent of exactly how the slices were chosen, which is what makes the definite integral a reliable, checkable number rather than an approximation that depends on arbitrary choices.

The strongest case against it

The Riemann sum limit defining the integral requires the underlying function to be reasonably well behaved, at minimum bounded with only a limited number of discontinuities, and functions that oscillate too wildly or blow up within the interval can defeat the construction entirely, requiring more advanced generalizations of the integral not covered here. A second real limitation is practical rather than theoretical: real measurements are almost always sampled at a finite number of points, so any integral computed from actual data is itself only a Riemann sum, an approximation whose accuracy depends entirely on how finely the data was sampled, and treating a coarse sample's sum as if it were the true limiting integral is a common source of error. A third common misconception is assuming the integral always represents a literal geometric area; when the function being integrated takes negative values, the integral instead computes a signed accumulation, area above the axis counted positively and area below counted negatively, canceling rather than adding, which can produce a total smaller than either region alone, a fact easy to miss if area is assumed to always be a positive quantity.

Where it stands now

The Riemann sum construction of the definite integral has been the standard, rigorously settled definition since the mid-nineteenth century, with no live dispute about its correctness for continuous or piecewise continuous functions, the overwhelming majority of cases arising in ordinary applications. More general notions of integration, built to handle stranger functions that defeat the Riemann construction, were developed later and remain useful in advanced settings, but they extend rather than replace the accumulation idea taught here: they agree with the Riemann integral on every function reasonable enough for the Riemann sum to already work.

Test yourself

A factory's power consumption varies over an eight hour shift according to the function p of t, equal to 10 plus 3t, measured in kilowatts, where t is hours since the shift began. Using the slice, approximate, sum method directly, divide the eight hour shift into four two-hour pieces, estimate the energy used in each piece by treating power as constant at its value at the start of the piece, and sum these to get a first estimate of the total energy consumed, in kilowatt hours, over the shift. Then explain, using the logic of shrinking slice width, why splitting the shift into eight one-hour pieces instead would produce a more accurate estimate, and state what specific total quantity, beyond energy, mass, or distance, you could compute with this same method if you were instead given a factory's water usage rate varying over the same eight hours.