1 Definition and basic ideas
Bounded variation is a way to quantify how much a real-valued function changes across an interval. Instead of measuring only the size of the values, it measures the cumulative size of all upward and downward movements. This makes it a useful notion for functions that may not be smooth but still have controlled irregularity.
A function with bounded variation need not be monotone or differentiable everywhere. However, it remains sufficiently structured to support decomposition theorems, measure-theoretic interpretations, and strong differentiation results. In one variable, the concept is built from the variation of the function over partitions of an interval.
1.1 Variation on an interval
For a function defined on an interval, the variation over a finite partition is obtained by adding the absolute differences between successive function values. Refining the partition may increase this sum, so the overall variation is defined by taking the supremum over all partitions.
This construction captures the full oscillatory behavior of the function. A function that repeatedly rises and falls can still have finite variation if the total size of those changes remains bounded.
1.2 Total variation
The total variation of a function on an interval is the supremum of all partition sums of absolute increments. It is a nonnegative quantity, and it vanishes exactly when the function is constant on the interval.
Total variation provides a norm-like measure of irregularity. It is additive over adjacent subintervals in an appropriate sense and serves as the basic numerical invariant for the theory.
1.3 Functions of bounded variation
A function is said to be of bounded variation, often abbreviated BV in one dimension, if its total variation on the interval is finite. Such functions may have jumps and other discontinuities, but they cannot oscillate without bound.
Every monotone function on a closed interval has bounded variation. More generally, functions of bounded variation form a broad class that includes many piecewise smooth and piecewise monotone functions.
1.4 Equivalent formulations
Several equivalent characterizations describe bounded variation. One common formulation requires that the function be expressible as a difference of two increasing functions. Another uses the total variation as a finite supremum over partitions.
These formulations connect bounded variation to order structure, measure theory, and decomposition methods. They also make the class robust under algebraic operations and limits.
2 Fundamental properties
Functions of bounded variation enjoy a number of structural properties that distinguish them from arbitrary bounded functions. They are stable under many natural operations, and their discontinuities are tightly constrained.
The theory is especially effective because bounded variation sits between monotonicity and absolute continuity. This position allows the class to retain enough regularity for analysis while remaining broad enough to include useful examples with jumps and corners.
2.1 Relation to monotone functions
Every monotone increasing or decreasing function on a closed interval has bounded variation. In fact, monotone functions form the simplest nontrivial examples of the theory.
Bounded variation generalizes monotonicity by permitting both upward and downward motion, as long as the total amount of motion is finite. This makes the class suitable for functions that are not ordered but still have controlled behavior.
2.2 Algebraic properties
The class of bounded variation functions is closed under addition, scalar multiplication, and subtraction. If two functions have bounded variation, then so do their linear combinations.
Products of bounded variation functions can also have bounded variation under suitable boundedness assumptions. Compositions with well-behaved functions often preserve bounded variation as well, making the class flexible in analysis.
2.3 Continuity and discontinuities
Bounded variation imposes strong restrictions on how a function can fail to be continuous. Although discontinuities are allowed, they cannot accumulate in an arbitrary way, and their total effect remains limited.
A function of bounded variation has at most countably many discontinuities of jump type. Moreover, the set of discontinuities is small in a measure-theoretic sense.
2.3.1 Jump discontinuities
A jump discontinuity occurs when the left and right limits exist but differ. For functions of bounded variation, such jumps may occur, but their total magnitude is controlled by the variation.
This makes jump behavior especially important in applications involving piecewise smooth models, signal processing, and weak formulations of differential equations.
2.3.2 Continuity almost everywhere
A bounded variation function is continuous almost everywhere with respect to Lebesgue measure. Although it may have exceptional points of discontinuity, these form a set of measure zero.
This property is one reason bounded variation is so useful in integration and differentiation theory. It ensures that the function behaves regularly on most points of the interval.
2.4 Bounded variation and absolute continuity
Absolute continuity is stronger than bounded variation. Every absolutely continuous function has bounded variation, but not every bounded variation function is absolutely continuous.
The difference lies in how variation is distributed. Absolutely continuous functions spread their change in a way that is directly controlled by integrable derivatives, whereas bounded variation functions may also contain singular or jump components.
3 Examples and nonexamples
Examples clarify the scope of bounded variation by showing both common and subtle cases. They also illustrate that finite variation is not the same as smoothness or even continuity.
Nonexamples are equally important because they show how a function may be bounded yet still fail to have bounded variation due to excessive oscillation.
3.1 Simple piecewise monotone functions
Piecewise monotone functions on a closed interval typically have bounded variation. Each monotone segment contributes a finite amount of variation, and the total remains finite if there are only finitely many pieces.
Functions of this type include many elementary graphs with corners, flat sections, and finitely many jumps.
3.2 Functions with finite total variation
Many functions used in analysis have finite total variation even when they are not differentiable everywhere. Examples include cumulative distribution functions, functions with finitely many jump discontinuities, and certain trigonometric or exponential combinations on compact intervals.
These examples show that bounded variation is a natural regularity condition in settings where smoothness is too restrictive.
3.3 Functions of infinite variation
A function may be bounded in value but still have infinite total variation. This occurs when the function oscillates too rapidly or too often.
Classical examples include suitably chosen oscillatory functions on intervals, especially those whose oscillations do not diminish fast enough. Such functions lie outside the BV class despite being otherwise well behaved in range.
4 Decomposition theorems
One of the most powerful features of bounded variation is that such functions can be decomposed into simpler components. These decompositions reveal the internal structure of variation and separate increasing behavior from more general motion.
This viewpoint is central to both classical analysis and modern measure-theoretic treatments.
4.1 Jordan decomposition
The Jordan decomposition theorem states that a function of bounded variation can be written as the difference of two increasing functions. This is one of the foundational results of the theory.
The theorem shows that bounded variation is, in a sense, precisely the class generated by monotone building blocks. It provides a bridge between oscillatory functions and ordered functions.
4.2 Positive and negative variation
The total variation of a function can be split into positive variation and negative variation, corresponding to the accumulated upward and downward movement. These two quantities are both increasing in the interval variable.
Their difference reconstructs the original function up to an additive constant, while their sum gives the total variation. This decomposition is often used in proofs and in the construction of associated measures.
4.3 Decomposition into absolutely continuous, singular, and jump parts
Functions of bounded variation admit a refined decomposition into three parts: an absolutely continuous part, a singular continuous part, and a jump part. Each component captures a different mode of variation.
The absolutely continuous part is governed by an integrable derivative. The jump part accounts for discontinuities, while the singular continuous part changes continuously but concentrates on a set of measure zero.
5 Integration and measure theory
Bounded variation is deeply connected to integration theory because such functions naturally define integrators in the sense of Stieltjes. They also correspond to measures through their variation and distributional derivatives.
These links make the class central in the passage from classical calculus to modern analysis.
5.1 Stieltjes integrals
A function of bounded variation can serve as an integrator in a Riemann–Stieltjes or Lebesgue–Stieltjes integral. The finiteness of variation ensures that the integral is well controlled under broad hypotheses.
This is especially important when integrating against functions with jumps or singular behavior. The Stieltjes framework extends ordinary integration by allowing the integrator itself to carry significant structure.
5.2 Connection with measures
A bounded variation function determines a measure through its increments. This measure-theoretic interpretation is one of the most influential aspects of the subject.
It allows variation to be studied using the tools of signed measures, Radon measures, and distribution theory.
5.2.1 Signed measures associated with BV functions
From a function of bounded variation, one can construct a signed measure whose values on intervals match the function’s increments. This measure encodes the same information as the function up to an additive constant.
The correspondence is especially transparent in one dimension, where intervals and endpoint values provide a natural way to record variation.
5.2.2 Distributional derivatives
The derivative of a BV function can be interpreted in the distributional sense as a measure. This extends the classical derivative and accommodates jumps and singular parts.
In this framework, the derivative of the function is not necessarily a function in the usual sense, but a finite signed measure or vector measure depending on the setting.
5.3 Functions of bounded variation as measures
The relationship between BV functions and measures goes both ways: bounded variation can be viewed as generating measure-valued derivatives, while measures can be used to reconstruct BV functions by integration.
This duality is one reason the class is so important in modern analysis. It provides a unified language for discontinuities, weak derivatives, and cumulative effects.
6 Differentiation results
Although bounded variation does not imply classical differentiability everywhere, it yields strong differentiability properties almost everywhere. These results are among the most striking features of the theory.
They show that finite variation prevents pathological oscillation from persisting on large sets.
6.1 Almost everywhere differentiability
A function of bounded variation on an interval is differentiable almost everywhere. This theorem places BV functions among the important classes of functions for which classical derivatives exist at most points.
The result is compatible with the presence of jumps and singular parts, since these exceptional behaviors occur on negligible sets.
6.2 The variation measure and derivative
The variation of a BV function gives rise to a measure that controls the derivative in a broad sense. Where the function is classically differentiable, the density of this measure agrees with the ordinary derivative.
This measure-theoretic derivative captures the full variation, including parts not visible to pointwise differentiation.
6.3 BV functions and weak derivatives
In weak form, a function of bounded variation has a derivative represented by a finite measure rather than by a classical function alone. This makes BV functions natural objects in the study of weak derivatives and generalized solutions.
The weak derivative viewpoint is particularly useful in contexts where solutions may develop edges, interfaces, or discontinuities.
7 Generalizations
The notion of bounded variation extends well beyond one-dimensional intervals. In higher dimensions, it becomes a powerful framework for describing sets and functions with controlled irregularity.
These generalizations play a major role in modern analysis, geometric measure theory, and the theory of partial differential equations.
7.1 Bounded variation on higher-dimensional domains
On multidimensional domains, bounded variation is defined using vector-valued distributional derivatives or equivalent geometric conditions. The total variation of the derivative measure replaces the one-dimensional partition-based definition.
This setting captures functions whose gradients may fail to exist classically but remain finite in a generalized sense.
7.2 Functions of bounded variation in several variables
In several variables, BV functions are those whose distributional derivatives are finite measures. They can exhibit jump sets and singular behavior along hypersurfaces, yet still retain enough structure for analysis.
The higher-dimensional theory is closely tied to interfaces, edges, and discontinuity surfaces.
7.2.1 Sets of finite perimeter
A set of finite perimeter is one whose characteristic function has bounded variation. Such sets provide a geometric version of the BV concept.
Their boundaries may be irregular, but their total boundary size is controlled in a measure-theoretic sense.
7.2.2 Caccioppoli sets
Caccioppoli sets are another name for sets of finite perimeter. They are central objects in geometric measure theory and the calculus of variations.
These sets generalize smooth domains by allowing rough boundaries while preserving a usable notion of surface area.
7.3 BV spaces in analysis
The collection of BV functions on a domain forms a function space equipped with a natural variation-based seminorm or norm. This space is important because it is large enough for weak limits of many analytic processes.
BV spaces often serve as the correct setting when derivatives may concentrate or when minimizing sequences develop discontinuities.
8 Applications
Bounded variation appears in many areas of analysis because it balances flexibility with control. The class is well suited to problems involving weak convergence, discontinuous behavior, and generalized derivatives.
Its applications range from classical integration to modern variational methods.
8.1 Real analysis
In real analysis, bounded variation provides a framework for studying pointwise properties, monotone approximations, and convergence of functions. It also supports rigorous treatment of functions with jumps and singular behavior.
The theory is often used in the study of integral representations and fine regularity properties.
8.2 Functional analysis
In functional analysis, BV spaces offer examples of Banach spaces with rich compactness and lower-semicontinuity properties. They are useful in understanding weak convergence and the behavior of function sequences under variation bounds.
These properties make BV spaces important in approximation theory and in the study of linear and nonlinear operators.
8.3 Partial differential equations
In partial differential equations, BV functions arise naturally when solutions develop discontinuities or sharp interfaces. They are especially relevant for conservation laws, free-boundary problems, and equations with nonsmooth solutions.
The BV framework allows one to formulate and analyze weak solutions whose derivatives are measures rather than classical functions.
8.4 Calculus of variations
The calculus of variations often uses BV spaces to treat minimization problems where classical smoothness fails. Energy functionals may admit minimizers with jump discontinuities or singular gradients, and BV provides the correct setting for their analysis.
This has made bounded variation a standard tool in modern variational methods, particularly for problems involving image segmentation, fracture models, and phase transitions.