Cauchy definition of definite integral

This note is about Cauchy’s (real-variable) approach to the definite integral as a limit of sums, historically preceding the modern formulation of the Riemann integral. (Not to be confused with the complex-analytic Cauchy integral formula.)

Origins (very brief)

Cauchy sums

Assume f is continuous on [x0,X] and take points

x0<x1<<xn1<X.

Cauchy considers sums like

S=i=0n1f(xi)(xi+1xi),(xn:=X)

and argues that as the subdivision gets finer (the maximum maxi(xi+1xi) tends to 0) the value of S approaches a limit depending only on f,x0,X.

He also notes one can choose sample points inside each subinterval, i.e.

S=i=0n1f(xi+ϑi(xi+1xi))(xi+1xi),ϑi[0,1],

which is essentially the tagged-sum viewpoint later used by Riemann.

Conceptual caveat (modern viewpoint)

Varying the partition does not naturally produce a sequence; it forms a directed family (a net) indexed by refinements. This is one place where the original 19th century discussions can look imprecise by modern standards.

Beyond continuity (Cauchy/Dirichlet/Lebesgue)

Cauchy also sketches extensions beyond continuous functions by splitting the interval into pieces where the function behaves well, and defining the integral by limits near singular/discontinuous points.

Dirichlet extends this idea to certain functions with infinitely many discontinuities but only finitely many accumulation points.

A key limitation (as discussed by Lebesgue) is that the Cauchy–Dirichlet approach essentially requires an antiderivative-type representation on intervals of continuity and restricts the discontinuity set (e.g. to be countable).