Discontinuity Calculus Overview

Updated 5 July 2026

Discontinuity Calculus is a framework that treats discontinuities as quantifiable objects using moduli and invariants rather than mere obstacles.
It employs hierarchical classifications and propagation mechanisms to analyze how discontinuities manifest and constrain maps across topology, PDEs, and variational settings.
The approach spans analytic, computable, and numerical methods, enabling precise control over jump sets and singular structures in diverse mathematical applications.

“Discontinuity calculus” does not denote a single standardized formalism. In the arXiv literature, the expression and closely related constructions refer to several programs in which discontinuity is treated as an object of calculation rather than merely as an obstruction: it is quantified by moduli or invariants, organized into hierarchies, propagated through operators, differentiated in variational settings with moving jumps, or encoded by contour deformation on Riemann surfaces. Across topology, computable analysis, the calculus of variations, PDE, complex analysis, and numerical approximation, the common theme is that one studies not only whether continuity fails, but how it fails, what structure that failure has, and how it constrains admissible maps, solutions, or analytic continuations (Adams et al., 10 Nov 2025, Brecht, 2013, Breitkopf et al., 26 Sep 2025, Britto et al., 15 Oct 2025, Jing et al., 8 Jul 2025).

1. Scope and recurrent ideas

A first recurrent pattern is the replacement of binary continuity questions by quantitative or structural data. In “Quantifying discontinuity” (Adams et al., 10 Nov 2025), the classical question “does $X$ embed continuously into $\mathbb R^d$ ?” is replaced by a scale-invariant modulus for injective maps $f:X\to\mathbb R^d$ , defined through the direction map

$\Phi_f\colon \Conf_2(X)\to S^{d-1},\qquad (x,y)\mapsto \frac{f(x)-f(y)}{\|f(x)-f(y)\|},$

and the modulus

$\alpha(f):=\delta(\Phi_f).$

The point is that ordinary diameter-based oscillation is not useful here, because any injective Euclidean map can be rescaled to have arbitrarily small image diameter.

A second pattern is hierarchical classification. In the represented-space setting, jump operators $j$ measure how much non-continuity a realizer is allowed to exhibit; in the Weihrauch setting, one isolates canonical discontinuous problems such as $DIS$ ; and in algebraic-topological settings one introduces invariants such as $\mu(A,d,\Sigma)$ that measure the least unavoidable discontinuity in exact models of an equational theory (Brecht, 2013, Brattka, 2020, Taylor, 2015).

A third pattern is propagation of discontinuity through analytic or variational mechanisms. In free-discontinuity and hyperbolic-PDE problems, the jump set or shock set is part of the unknown, so a calculus must differentiate not only state values but also jump locations. In nonlocal Poisson problems, discontinuities of the source and singularities of the kernel determine where the solution inherits jumps, loses derivatives, or develops cascades of translated derivative discontinuities. In Feynman-integral problems, discontinuity is encoded by contour deformation and monodromy rather than by pointwise jumps (Conti et al., 1 May 2025, Breitkopf et al., 26 Sep 2025, Dang et al., 2 May 2026, Britto et al., 15 Oct 2025).

A common misconception is that discontinuity calculus is always a regularization scheme. Some works do construct analytic approximants to discontinuous data, but others instead prove lower bounds, impossibility theorems, or exact propagation rules for singular structure (Stella et al., 2016, Maor, 2018).

2. Quantified discontinuity in topology and metric-algebraic settings

The quantified-topological strand is centered on explicit moduli. For compact $X$ , if there is no $\mathbb Z/2$ -equivariant map

$\mathbb R^d$ 0

then every injective $\mathbb R^d$ 1 satisfies

$\mathbb R^d$ 2

yielding quantified nonembeddability results of Haefliger–Weber type (Adams et al., 10 Nov 2025). The same framework extends to deleted products and deleted joins, producing quantified versions of van Kampen–Flores and of the topological Tverberg theorem. For prime-power $\mathbb R^d$ 3, every almost $\mathbb R^d$ 4-injective

$\mathbb R^d$ 5

satisfies

$\mathbb R^d$ 6

The paper also relates angular instability to ordinary discontinuity by

$\mathbb R^d$ 7

A different but related metric-algebraic calculus is developed through the local jump

$\mathbb R^d$ 8

its global version $\mathbb R^d$ 9, and the invariant

$f:X\to\mathbb R^d$ 0

which measures the least amount of discontinuity required by exact satisfaction of equations on a metric space (Taylor, 2015). This yields explicit values and bounds. For $f:X\to\mathbb R^d$ 1 with diameter $f:X\to\mathbb R^d$ 2, the paper proves $f:X\to\mathbb R^d$ 3 for several theories, including $f:X\to\mathbb R^d$ 4, commutative idempotent binary operations, and ternary majority laws. It also develops iterated variants $f:X\to\mathbb R^d$ 5, showing that discontinuity can increase when one passes from basic operations to composite terms.

These approaches share a precise feature: they do not ask for an approximation to continuity, but for numerical lower bounds on unavoidable instability. This suggests a shift from qualitative nonembeddability or incompatibility to a genuinely quantitative geometry of discontinuity.

3. Computability-theoretic and represented-space hierarchies

In computable analysis, discontinuity calculus becomes a hierarchy of realizability levels. Jump operators on represented spaces are partial surjections $f:X\to\mathbb R^d$ 6 such that every partial continuous $f:X\to\mathbb R^d$ 7 admits continuous $f:X\to\mathbb R^d$ 8 with $f:X\to\mathbb R^d$ 9; they induce endofunctors $\Phi_f\colon \Conf_2(X)\to S^{d-1},\qquad (x,y)\mapsto \frac{f(x)-f(y)}{\|f(x)-f(y)\|},$0 and define $\Phi_f\colon \Conf_2(X)\to S^{d-1},\qquad (x,y)\mapsto \frac{f(x)-f(y)}{\|f(x)-f(y)\|},$1-realizability (Brecht, 2013). Canonical examples are $\Phi_f\colon \Conf_2(X)\to S^{d-1},\qquad (x,y)\mapsto \frac{f(x)-f(y)}{\|f(x)-f(y)\|},$2 for convergent sequences of guesses, $\Phi_f\colon \Conf_2(X)\to S^{d-1},\qquad (x,y)\mapsto \frac{f(x)-f(y)}{\|f(x)-f(y)\|},$3 for eventually constant guesses, and $\Phi_f\colon \Conf_2(X)\to S^{d-1},\qquad (x,y)\mapsto \frac{f(x)-f(y)}{\|f(x)-f(y)\|},$4 for ordinal-bounded mind changes. The hierarchy satisfies

$\Phi_f\colon \Conf_2(X)\to S^{d-1},\qquad (x,y)\mapsto \frac{f(x)-f(y)}{\|f(x)-f(y)\|},$5

The main correspondence is exact: for admissibly represented countably based $\Phi_f\colon \Conf_2(X)\to S^{d-1},\qquad (x,y)\mapsto \frac{f(x)-f(y)}{\|f(x)-f(y)\|},$6-spaces,

$\Phi_f\colon \Conf_2(X)\to S^{d-1},\qquad (x,y)\mapsto \frac{f(x)-f(y)}{\|f(x)-f(y)\|},$7

and more generally

$\Phi_f\colon \Conf_2(X)\to S^{d-1},\qquad (x,y)\mapsto \frac{f(x)-f(y)}{\|f(x)-f(y)\|},$8

Thus Hertling’s level of discontinuity becomes the minimal ordinal jump needed to realize $\Phi_f\colon \Conf_2(X)\to S^{d-1},\qquad (x,y)\mapsto \frac{f(x)-f(y)}{\|f(x)-f(y)\|},$9.

At the bottom of the continuous Weihrauch lattice, “The Discontinuity Problem” introduces

$\alpha(f):=\delta(\Phi_f).$ 0

and proves

$\alpha(f):=\delta(\Phi_f).$ 1

(Brattka, 2020). The framework is game-theoretic: Player II has a winning strategy iff the problem is continuous, while Player I has a winning strategy iff it is effectively discontinuous. The axiomatic setting matters essentially: under $\alpha(f):=\delta(\Phi_f).$ 2, every problem is either continuous or effectively discontinuous, whereas under $\alpha(f):=\delta(\Phi_f).$ 3 there are discontinuous problems that are not effectively discontinuous.

A further refinement is the study of multidimensional step discontinuities. For a single proper threshold $\alpha(f):=\delta(\Phi_f).$ 4 and a Boolean truth table $\alpha(f):=\delta(\Phi_f).$ 5, the problem $\alpha(f):=\delta(\Phi_f).$ 6 is classified by the alternation invariant $\alpha(f):=\delta(\Phi_f).$ 7, with

$\alpha(f):=\delta(\Phi_f).$ 8

(Hölzl et al., 2024). This yields strict hierarchies such as

$\alpha(f):=\delta(\Phi_f).$ 9

Here discontinuity is not merely present; its combinatorial shape determines Weihrauch degree.

4. Variational and PDE calculi with jump sets, shocks, and weak discontinuity

In the calculus of variations, discontinuity calculus often means a first-order theory for energies or solutions with singular sets. One striking example is the Jacobian determinant functional

$j$ 0

which is shown not to be weakly continuous, hence not weakly lower-semicontinuous, on $j$ 1, even when restricted to conformal diffeomorphisms (Maor, 2018). There exists a sequence of Möbius transformations $j$ 2 with

$j$ 3

Since the weak limit is the constant map $j$ 4, the example gives a geometrically clean counterexample to weak lower-semicontinuity. It also shows that quasiconvexity alone is not sufficient without additional assumptions.

Free-discontinuity models in $j$ 5 extend this logic to energies where the jump set itself is optimized. For

$j$ 6

the superlinear cohesive regime allows $j$ 7 to have infinite measure, because $j$ 8 as $j$ 9 (Conti et al., 1 May 2025). The paper proves lower semicontinuity, relaxation to the quasiconvex and BV-elliptic envelopes, and an Ambrosio–Tortorelli-type phase-field approximation in this setting.

For entropy solutions of the Generalized Riemann Problem, the control is the $DIS$ 0 left state, the $DIS$ 1 right state, and the position $DIS$ 2 of the unique initial jump. Since shocks move with the control, a reference-space transformation fixes shock positions, after which the solution operator is continuously Fréchet differentiable into $DIS$ 3; in physical coordinates, the derivative contains both a smooth part and shock-shift contributions (Breitkopf et al., 26 Sep 2025). The measure-valued first-order variation is written

$DIS$ 4

This is a literal calculus of discontinuities: the derivative includes motion of the jump set.

For the one-dimensional nonlocal Poisson equation, discontinuity analysis is formulated as operator propagation rules. Under $DIS$ 5, jumps in the source are inherited by the solution, singularities of $DIS$ 6 at the origin force blow-up of corresponding derivatives at the source jump, and jumps of $DIS$ 7 or its derivatives at $DIS$ 8 create cascades of derivative jumps at $DIS$ 9 (Dang et al., 2 May 2026). For $\mu(A,d,\Sigma)$ 0, by contrast, source and solution have equivalent piecewise smooth regularity. The same structural analysis leads to a semi-analytic spectral method based on successive smoothing transformations and correction functions $\mu(A,d,\Sigma)$ 1.

5. Complex-analytic discontinuity, monodromy, and generalized differential calculus

In complex analysis and quantum field theory, discontinuity calculus is formulated through continuation operators and contour deformation. One such framework decomposes a function into branch-cut and pole contributions,

$\mu(A,d,\Sigma)$ 2

and introduces continuation kernels and continuation generators $\mu(A,d,\Sigma)$ 3 to cross cuts iteratively (Jing et al., 8 Jul 2025). The formalism reproduces the standard sheet structures: $\mu(A,d,\Sigma)$ 4 Applied to coupled-channel scattering, it yields a $\mu(A,d,\Sigma)$ 5-sheet structure when pseudo-threshold cuts are ignored and a genus formula

$\mu(A,d,\Sigma)$ 6

so the two-channel surface has genus $\mu(A,d,\Sigma)$ 7, the three-channel surface genus $\mu(A,d,\Sigma)$ 8, and higher-channel surfaces require automorphic uniformization.

A closely related geometric calculus is developed in Feynman-parameter space (Britto et al., 15 Oct 2025). There, one distinguishes total discontinuities from monodromies and tracks how the integration contour changes as external kinematics are analytically continued. Sequential discontinuities are constrained by the boundary components retained by the first monodromy contour: if a later singularity requires a boundary that has already been lost, the corresponding iterated discontinuity is forbidden. This translates directly into constraints on symbol words and is effective in dimensional regularization, with higher propagator powers, and for non-uniform transcendental weight.

A different meaning of discontinuous calculus appears in Colombeau’s full algebra, where the scalar ring is the ring $\mu(A,d,\Sigma)$ 9 of full generalized numbers with the sharp ultrametric

$X$ 0

(Cortes et al., 2017). For functions $X$ 1, differentiability at $X$ 2 is defined by the existence of $X$ 3 such that

$X$ 4

The resulting calculus retains sum, product, quotient, and multivariable rules, proves an Embedding Theorem and an Open Mapping Theorem, and develops holomorphic and analytic function theory on generalized points. Here the ambient spaces are totally disconnected, so discontinuity is built into the geometry rather than added as an exception.

Some works use discontinuity calculus in a constructive numerical sense. For piecewise continuous functions with finitely many known discontinuities, a global analytic approximant is built as

$X$ 5

where each $X$ 6 is a hyperbolic-tangent connector centered at a discontinuity and the coefficients $X$ 7 are obtained from a linear system enforcing the interval averages of an auxiliary function $X$ 8 (Stella et al., 2016). The approximants are analytic, free from Gibbs phenomenon, and reported to achieve relative errors of order $X$ 9 in several examples. The same connector method is used for a Filippov-type oscillator with discontinuous forcing and for a thin rectangle approximating a Dirac delta.

Several neighboring frameworks illuminate the breadth of the term while remaining conceptually distinct. Darboux calculus studies partially defined order-preserving maps $\mathbb Z/2$ 0 through lower and upper extensions $\mathbb Z/2$ 1, with the Darboux set

$\mathbb Z/2$ 2

as the locus where a canonical extension exists (Aldi et al., 2016). Continuity, limits, and Riemann integrability then appear as agreement of extremal extensions rather than as primitive $\mathbb Z/2$ 3- $\mathbb Z/2$ 4 notions.

Displacement calculus, by contrast, is a logic of discontinuous syntax in categorial grammar. It extends Lambek calculus by sorted wrapping and intercalation connectives, proves Cut-elimination, the subformula property, and decidability, and analyzes phenomena such as medial extraction, VP ellipsis, gapping, and parentheticals (Morrill et al., 2010). The discontinuity here is syntactic, not analytic.

In causal inference, regression discontinuity designs are analyzed through the calculus of conditional independence. The discontinuity at a treatment threshold becomes identifiable as an average causal effect at the cutoff under precise regime-invariance and continuity assumptions, with separate treatments for sharp and fuzzy designs (Constantinou et al., 2016). This is not usually called discontinuity calculus in the same sense, but it shares the central idea that the jump itself can be manipulated by formal rules rather than treated informally.

Taken together, these literatures show that discontinuity calculus is best understood as a family of rigorous techniques for quantifying, propagating, regularizing, or analytically continuing discontinuities. The unifying idea is methodological rather than terminological: discontinuities are treated as structured mathematical data, subject to operators, invariants, and variational principles, rather than as defects to be ignored or smoothed away.