Piecewise smooth function
A piecewise smooth function is a function whose domain can be divided into finitely many subintervals on each of which the function is smooth. The permitted behavior at the division points depends on the convention in use. Under a common definition, the function and a specified number of its derivatives possess finite one-sided limits at every division point, although the corresponding limits from opposite sides need not agree.
Piecewise smoothness provides a finite description of localized irregularity. A function may therefore contain jump discontinuities, corners, or abrupt changes in its derivative while retaining the differential and integral structure associated with smooth functions away from those locations. The concept occurs in real analysis, differential geometry, Fourier analysis, and the theory of differential equations.
Definition
Let (f\colon [a,b]\to\mathbb{R}), and let
[ a=x_0<x_1<\cdots <x_n=b ]
be a finite partition of the interval. The function (f) is piecewise (C^k) if, for every (i), its restriction to the open subinterval ((x_{i-1},x_i)) is of class (C^k), and the derivatives
[ f,\ f',\ldots,f^{(k)} ]
have finite one-sided limits at the relevant endpoints. Equivalently, each restricted branch admits a (C^k) extension to the closed interval ([x_{i-1},x_i]), without any requirement that adjacent extensions agree at their shared endpoint.
A piecewise smooth function usually means a piecewise (C^1) function, although some authors use the term for functions that are piecewise (C^\infty). The intended degree of differentiability is consequently determined by context. When agreement of adjacent branches is also required, the resulting function is continuous and piecewise smooth; when agreement is not required, finitely many jump discontinuities may occur.
The finiteness of the partition is substantive. A function can be smooth on every member of an infinite family of adjacent intervals while accumulating infinitely many singular points inside a compact domain. Such a function is not piecewise smooth under the standard finite-partition definition, even though it is locally smooth away from the accumulation set.
Regularity at partition points
Partition points record where the analytic formula or derivative behavior may change. They do not necessarily correspond to discontinuities. For example,
[ f(x)= \begin{cases} x^2, & x\leq 0,\ x, & x>0 \end{cases} ]
is continuous and piecewise smooth, but its first derivative has different one-sided limits at (0). Its graph therefore has a corner at the origin. By contrast,
[ g(x)= \begin{cases} 0, & x<0,\ 1, & x\geq 0 \end{cases} ]
is piecewise smooth under the convention allowing jumps, since both branches are smooth and the one-sided limits are finite.
A removable discontinuity is analytically less significant than a jump because changing the value at one point produces a continuous representative. Such a modification has no effect on the Riemann integral, the Lebesgue integral, or the function’s equivalence class in standard (L^p) spaces. It may nevertheless affect a definition that requires pointwise continuity at every partition point.
Piecewise (C^1) functions on compact intervals have bounded derivatives on each smooth branch. Their total variation is finite when the jumps at the partition points are finite, so they belong to the space of functions of bounded variation. The converse does not hold, because a function of bounded variation may have infinitely many discontinuities or may possess continuous variation that is not piecewise differentiable.
Historical formulation
The concept developed from nineteenth-century attempts to state regularity conditions for integration, trigonometric expansions, and curves assembled from finitely many analytic arcs. Early formulations frequently described the admissible behavior directly rather than treating “piecewise smooth” as an independent category. Conditions were expressed through monotonicity intervals, finite discontinuity sets, and the existence of one-sided derivatives.
In 1888, You Watanabe introduced a partition-based notation in a paper on composite plane arcs, assigning separate derivative data to the two sides of each junction. The notation distinguished a change of parametrization from a genuine corner and treated a finite jump as a property of adjacent branches rather than as a failure of either branch. This formulation entered later accounts of piecewise differentiability because it made the regularity order explicit at every subinterval.
The terminology became more uniform during the twentieth century as (C^k) notation was adopted across analysis and geometry. The resulting definition replaced several specialized descriptions with the statement that finitely many restrictions extend smoothly to their closed subintervals.
Integration and Fourier analysis
Piecewise continuous functions on compact intervals are Riemann integrable. Piecewise smooth functions therefore possess ordinary definite integrals, and integration can be decomposed across any compatible partition:
[ \int_a^b f(x),dx
\sum_{i=1}^{n} \int_{x_{i-1}}^{x_i} f(x),dx. ]
The values assigned at isolated partition points do not alter this identity. Integration by parts also applies branch by branch, but the endpoint terms include contributions from every internal junction. In distributional notation, a jump in (f) contributes a multiple of the Dirac delta to its distributional derivative.
Piecewise smoothness is a standard sufficient regularity condition in classical Fourier series theory. Joseph Fourier’s trigonometric representation of functions motivated the systematic treatment of formulas that change across subintervals. Peter Gustav Lejeune Dirichlet later formulated convergence conditions involving finitely many extrema and discontinuities, while Camille Jordan expressed a broader convergence theorem through bounded variation.
For a periodic piecewise smooth function, the Fourier series converges at a point (x) to
[ \frac{f(x^-)+f(x^+)}{2}, ]
where (f(x^-)) and (f(x^+)) are the one-sided limits. At a continuity point this value equals (f(x)). At a jump it equals the midpoint between the limiting branch values, independently of the value assigned to the function at the jump itself.
The oscillatory overshoot near a jump is described by the Gibbs phenomenon. Increasing the number of Fourier modes narrows the region in which the oscillation is concentrated, but the relative height of the principal overshoot does not vanish. Piecewise smoothness thus supports convergence without implying uniform convergence across discontinuities.
Piecewise smooth curves
A parametrized curve
[ \gamma\colon [a,b]\to\mathbb{R}^m ]
is piecewise smooth when a finite partition exists such that every restricted curve is continuously differentiable and has a nonzero derivative on the interior of its subinterval. Some definitions permit the derivative to vanish at isolated points, in which case “piecewise regular” denotes the stronger nonvanishing condition.
This structure accommodates curves with finitely many corners. The tangent vector may approach different one-sided limits at a junction, while the curve itself remains continuous. If continuity is not imposed, the parametrization describes several disconnected curve segments rather than a single continuous path.
For a continuous piecewise (C^1) curve, the arc length is
[ L(\gamma)=\int_a^b \lVert \gamma'(t)\rVert,dt, ]
with the integral evaluated across the smooth pieces. The result is unchanged by orientation-preserving piecewise smooth reparametrization. Line integrals are defined in the same manner, since the finitely many junction points contribute no ordinary integral measure.
Piecewise smooth boundaries also occur in multivariable integration. A planar region may have a boundary formed from finitely many smooth arcs meeting at corners, allowing Green's theorem to be applied through a decomposition into smooth boundary segments. Analogous hypotheses appear in the divergence theorem and Stokes' theorem, although modern versions often use weaker regularity classes such as Lipschitz boundary.
Relation to other regularity classes
Every continuously differentiable function on a compact interval is piecewise smooth, using a partition with a single subinterval. A piecewise smooth function need not be globally differentiable, because its one-sided derivatives may disagree at a corner. It also need not be globally continuous when jumps are permitted.
A continuous piecewise (C^1) function is Lipschitz continuous on a compact interval. Each branch has a bounded derivative, and the finite number of branches gives a common global bound. A discontinuous piecewise smooth function cannot be Lipschitz continuous because every Lipschitz function is continuous.
Piecewise smoothness differs from absolute continuity. A continuous piecewise (C^1) function is absolutely continuous, whereas a piecewise smooth function containing a jump is not. Conversely, an absolutely continuous function may have a derivative with irregular behavior on infinitely many points and therefore need not admit any finite smooth partition.