Mathematics
Riemann vs. Lebesgue Integration: What Changes and Why
Quick fact
The Dirichlet function—which is 1 on rational numbers and 0 on irrationals—is not Riemann integrable on any interval, yet its Lebesgue integral is exactly 0 because the rationals have Lebesgue measure zero.
Why this is interesting
You know the area under a curve. But what if the curve is so messy that the area is undefined? The Riemann integral fails, and the Lebesgue integral rescues us—but how?
Read the full explanation
Understanding Riemann vs. Lebesgue Integration: What Changes and Why
Imagine you want to measure the area under a curve. The Riemann integral splits the x-axis (the domain) into small slices and sums up the areas of rectangles whose heights are given by the function. That works well for functions that don't wiggle too wildly, like polynomials or sines. But if the function jumps everywhere—like the Dirichlet function which is 1 on rational numbers and 0 on irrationals—the Riemann sum becomes inconsistent: if you pick sample points at rationals you get one value, at irrationals another. The upper and lower sums never agree, so no area exists. Now flip your perspective. Instead of slicing the x-axis, slice the y-axis (the range). For a given height, ask: on what set of x-values does the function stay close to that height? Then measure the 'size' (the Lebesgue measure) of that set. Multiply that size by the height, and sum over all heights. This is the Lebesgue integral. It ignores the arrangement of the function's values; it only cares about the sizes of the sets where the function takes those values. For the Dirichlet function, for height 0 you get the set of irrationals (which has measure the length of the interval), and for height 1 you get the rationals (which have measure zero). So the Lebesgue integral is the size of the irrationals times 0 plus the size of the rationals times 1, which is 0. This simple example shows how Lebesgue integration can assign an area where the Riemann integral fails.
A deeper explanation
The core difference lies in the partitioning strategy. The Riemann integral partitions the domain into intervals, then approximates the function as constant on each piece. This works only if the function is 'well-behaved'—meaning the set of discontinuities is small (in a precise sense, it must have measure zero). The Lebesgue integral, on the other hand, partitions the range, then measures the preimage of each slice. This requires that the preimages are measurable, a condition that holds for a much wider class of functions (the measurable functions). The Lebesgue integral is not just a different way to compute areas; it is a more powerful tool because it allows us to integrate functions that are limits of sequences of integrable functions, and it gives us powerful convergence theorems. For instance, the Dominated Convergence Theorem states that if a sequence of functions converges pointwise and is dominated by an integrable function, then the limit's integral is the limit of the integrals. This is not true for the Riemann integral, where such exchanges can fail even for seemingly tame functions. So, the Lebesgue integral is a generalization: every Riemann integrable function is also Lebesgue integrable, and the two integrals agree when both exist. But the Lebesgue integral also handles many more functions, including those that are unbounded or defined on sets without a natural interval structure. This makes it the standard integral in real analysis, probability theory, and functional analysis, because it aligns with measure theory and allows for elegant mathematical statements and proofs.