Delta Shock Waves in Hyperbolic Conservation Laws

• Delta shock waves are singular discontinuities in hyperbolic conservation laws marked by Dirac delta concentrations on moving interfaces.
• They arise when classical self-similar wave patterns collapse, forcing mass, momentum, or energy to concentrate on lower-dimensional sets.
• Generalized Rankine–Hugoniot conditions and overcompressive criteria rigorously govern their dynamics and admissibility in various model systems.

Searching arXiv for recent and foundational papers on delta shock waves to ground the article. A delta shock wave is a measure-valued discontinuity in a hyperbolic conservation law for which one or more conserved variables develop a Dirac delta singularity supported on a moving interface. In the cited literature, delta shocks occur in pressureless Euler systems, Eulerian droplet models with drag, generalized zero-pressure systems with linear damping, Temple-class systems, relativistic Chaplygin flows, and two-dimensional zero-pressure gas dynamics with internal energy. They arise precisely when classical self-similar wave patterns fail or when singular limits force concentration of mass, momentum, or energy onto a lower-dimensional set (Keita et al., 2017, Pandey et al., 23 Jul 2025, Shao, 2017).

1. Distributional structure and measure-valued representation

In one space dimension, the standard pressureless Euler representation writes the density and momentum as a piecewise regular part plus a Dirac mass supported on a curve $x=X(t)$: $$\rho(x,t)=\rho^{-}(x,t)\,H\big(X(t)-x\big)+\rho^{+}(x,t)\,H\big(x-X(t)\big)+w(t)\,\delta\big(x-X(t)\big),$$

$$m(x,t):=(\rho u)(x,t)=m^{-}(x,t)\,H\big(X(t)-x\big)+m^{+}(x,t)\,H\big(x-X(t)\big)+w(t)\,V(t)\,\delta\big(x-X(t)\big),$$

with $V(t)=X'(t)$ and $w(t)\ge 0$ the concentrated mass (Gao et al., 2022). In the zero-pressure Riemann problem this concentrated measure replaces the ill-posed classical discontinuity when characteristics overlap, and the over-compressive condition $u_L \ge V(t) \ge u_R$ enforces sticky dynamics (Gao et al., 2022).

The same structural idea persists in other systems, but the singular component need not be attached to the same variable. In Korchinski-type and embedded constructions, the state takes the form

$U(x,t)=\overline U(x,t)+w(t)\delta(x-s(t))\eta,$

and, in the systems studied there, the Dirac mass appears only in the second component of $U=(u,v)^T$ (Castañeda, 2019). In the Eulerian droplet model, the droplet volume fraction $\alpha$ carries the singular concentration while $u$ remains piecewise regular away from the shock path (Keita et al., 2017). In the relativistic full Euler equations for a Chaplygin gas, the singularity is stronger: both the rest-mass density $$\rho(x,t)=\rho^{-}(x,t)\,H\big(X(t)-x\big)+\rho^{+}(x,t)\,H\big(x-X(t)\big)+w(t)\,\delta\big(x-X(t)\big),$$0 and the proper energy density $$\rho(x,t)=\rho^{-}(x,t)\,H\big(X(t)-x\big)+\rho^{+}(x,t)\,H\big(x-X(t)\big)+w(t)\,\delta\big(x-X(t)\big),$$1 simultaneously contain Dirac masses on the same shock curve (Shao, 2017). In two dimensions, the support becomes a moving curve $$\rho(x,t)=\rho^{-}(x,t)\,H\big(X(t)-x\big)+\rho^{+}(x,t)\,H\big(x-X(t)\big)+w(t)\,\delta\big(x-X(t)\big),$$2, and both density $$\rho(x,t)=\rho^{-}(x,t)\,H\big(X(t)-x\big)+\rho^{+}(x,t)\,H\big(x-X(t)\big)+w(t)\,\delta\big(x-X(t)\big),$$3 and internal energy density $$\rho(x,t)=\rho^{-}(x,t)\,H\big(X(t)-x\big)+\rho^{+}(x,t)\,H\big(x-X(t)\big)+w(t)\,\delta\big(x-X(t)\big),$$4 may concentrate on $$\rho(x,t)=\rho^{-}(x,t)\,H\big(X(t)-x\big)+\rho^{+}(x,t)\,H\big(x-X(t)\big)+w(t)\,\delta\big(x-X(t)\big),$$5 with strengths $$\rho(x,t)=\rho^{-}(x,t)\,H\big(X(t)-x\big)+\rho^{+}(x,t)\,H\big(x-X(t)\big)+w(t)\,\delta\big(x-X(t)\big),$$6 and $$\rho(x,t)=\rho^{-}(x,t)\,H\big(X(t)-x\big)+\rho^{+}(x,t)\,H\big(x-X(t)\big)+w(t)\,\delta\big(x-X(t)\big),$$7 (Pandey et al., 23 Jul 2025).

This distributional formulation is not a secondary technicality; it is the definition of the wave. A delta shock is not a classical jump with large amplitude, but a solution in which conservation laws are satisfied only after the singular measure supported on the moving interface is explicitly included (Yang et al., 2014).

2. Formation mechanisms

The cited works exhibit several distinct formation mechanisms. The most direct is characteristic overlap in pressureless or linearly degenerate systems. For the one-dimensional pressureless Euler equations,

$$\rho(x,t)=\rho^{-}(x,t)\,H\big(X(t)-x\big)+\rho^{+}(x,t)\,H\big(x-X(t)\big)+w(t)\,\delta\big(x-X(t)\big),$$8

the Riemann solution is classical only when $$\rho(x,t)=\rho^{-}(x,t)\,H\big(X(t)-x\big)+\rho^{+}(x,t)\,H\big(x-X(t)\big)+w(t)\,\delta\big(x-X(t)\big),$$9, in which case two contact discontinuities bound a vacuum region; when $$m(x,t):=(\rho u)(x,t)=m^{-}(x,t)\,H\big(X(t)-x\big)+m^{+}(x,t)\,H\big(x-X(t)\big)+w(t)\,V(t)\,\delta\big(x-X(t)\big),$$0, the classical Rankine–Hugoniot jump fails and a delta shock forms instead (Shao, 2015, Gao et al., 2022).

A closely related mechanism appears in the Eulerian droplet model

$$m(x,t):=(\rho u)(x,t)=m^{-}(x,t)\,H\big(X(t)-x\big)+m^{+}(x,t)\,H\big(x-X(t)\big)+w(t)\,V(t)\,\delta\big(x-X(t)\big),$$1

whose velocity subsystem reduces, for smooth solutions with $$m(x,t):=(\rho u)(x,t)=m^{-}(x,t)\,H\big(X(t)-x\big)+m^{+}(x,t)\,H\big(x-X(t)\big)+w(t)\,V(t)\,\delta\big(x-X(t)\big),$$2, to the inviscid Burgers equation with source term

$$m(x,t):=(\rho u)(x,t)=m^{-}(x,t)\,H\big(X(t)-x\big)+m^{+}(x,t)\,H\big(x-X(t)\big)+w(t)\,V(t)\,\delta\big(x-X(t)\big),$$3

There the loss of regularity criterion is explicit: $$m(x,t):=(\rho u)(x,t)=m^{-}(x,t)\,H\big(X(t)-x\big)+m^{+}(x,t)\,H\big(x-X(t)\big)+w(t)\,V(t)\,\delta\big(x-X(t)\big),$$4 with blow-up time

$$m(x,t):=(\rho u)(x,t)=m^{-}(x,t)\,H\big(X(t)-x\big)+m^{+}(x,t)\,H\big(x-X(t)\big)+w(t)\,V(t)\,\delta\big(x-X(t)\big),$$5

The same criterion carries to the full droplet model and its subsystem because the velocity satisfies the Burgers equation with source along characteristics (Keita et al., 2017). The paper also states that increasing $$m(x,t):=(\rho u)(x,t)=m^{-}(x,t)\,H\big(X(t)-x\big)+m^{+}(x,t)\,H\big(x-X(t)\big)+w(t)\,V(t)\,\delta\big(x-X(t)\big),$$6 delays or prevents characteristic crossing, which reflects the stabilizing role of drag (Keita et al., 2017).

Singular-limit mechanisms are equally important. In the isentropic magnetogasdynamics equations for a Chaplygin gas, the Riemann solution with two shocks converges, as both the pressure and the magnetic field vanish, to a delta shock of the transport equations; by contrast, the two-rarefaction configuration converges to two contact discontinuities with vacuum between them (Shao, 2015). An analogous conclusion is established by flux approximation for zero-pressure gas dynamics: as the perturbation parameters vanish, parameterized delta shocks converge to the delta-shock solutions of the pressureless limit, while constant-density intermediate states converge to vacuum (Yang et al., 2014).

The embedded constructions show that concentration need not appear only as an isolated overcompressive discontinuity between two constant states. In the modified systems studied in “Embedded delta shocks,” a delta shock can sit inside a fan containing one or two centered rarefactions, so that the singularity is embedded in a multi-wave pattern rather than separating two constants directly (Castañeda, 2019).

3. Generalized Rankine–Hugoniot relations and admissibility

The motion of a delta shock is governed by generalized Rankine–Hugoniot relations. For the one-dimensional pressureless Euler system with a single moving interface $$m(x,t):=(\rho u)(x,t)=m^{-}(x,t)\,H\big(X(t)-x\big)+m^{+}(x,t)\,H\big(x-X(t)\big)+w(t)\,V(t)\,\delta\big(x-X(t)\big),$$7, the ODE system is

$$m(x,t):=(\rho u)(x,t)=m^{-}(x,t)\,H\big(X(t)-x\big)+m^{+}(x,t)\,H\big(x-X(t)\big)+w(t)\,V(t)\,\delta\big(x-X(t)\big),$$8

$$m(x,t):=(\rho u)(x,t)=m^{-}(x,t)\,H\big(X(t)-x\big)+m^{+}(x,t)\,H\big(x-X(t)\big)+w(t)\,V(t)\,\delta\big(x-X(t)\big),$$9

$V(t)=X'(t)$0

For constant left and right states, these equations determine the trajectory, the interfacial speed, and the accumulated mass (Gao et al., 2022). Self-similar constant-speed solutions satisfy the quadratic constraint

$V(t)=X'(t)$1

whose entropy-admissible root is

$V(t)=X'(t)$2

This is the standard square-root-weighted delta-shock speed for the pressureless Riemann problem (Gao et al., 2022).

Source terms modify the jump law. For the Eulerian droplet model, the generalized Rankine–Hugoniot system becomes

$V(t)=X'(t)$3

$V(t)=X'(t)$4

The extra term $V(t)=X'(t)$5 is the contribution of the drag source to the concentrated momentum balance (Keita et al., 2017).

In two dimensions the same principle holds along a moving curve $V(t)=X'(t)$6, but the balance laws are written as ODEs for the geometric motion and the concentrated strengths: $V(t)=X'(t)$7 together with evolution equations for $V(t)=X'(t)$8, $V(t)=X'(t)$9, $w(t)\ge 0$0, and $w(t)\ge 0$1 (Pandey et al., 23 Jul 2025). For the relativistic Chaplygin gas, the generalized relations couple the shock speed to the delta weights in both $w(t)\ge 0$2 and $w(t)\ge 0$3, not only one state variable (Shao, 2017).

Admissibility is consistently overcompressive. In pressureless Euler this is the sticky inequality $w(t)\ge 0$4 (Gao et al., 2022). In the droplet model it is the Lax-type condition $w(t)\ge 0$5 (Keita et al., 2017). In two dimensions it becomes

$w(t)\ge 0$6

which requires characteristics to enter the discontinuity from both sides (Pandey et al., 23 Jul 2025). In the Temple-class $w(t)\ge 0$7 system the admissibility condition is

$w(t)\ge 0$8

and in the relativistic Chaplygin setting it becomes

$w(t)\ge 0$9

These conditions distinguish delta shocks from classical Lax shocks and are the core selection principle used across the cited literature (cruz et al., 2015, Shao, 2017).

4. One-dimensional model classes and Riemann structures

The one-dimensional literature shows that delta shocks are not confined to a single PDE class. The following model families all admit explicit Riemann constructions.

Model class	Concentrated variable(s)	Distinctive feature
Pressureless Euler / transport equations	$u_L \ge V(t) \ge u_R$0, $u_L \ge V(t) \ge u_R$1	Vacuum for $u_L \ge V(t) \ge u_R$2; delta shock for $u_L \ge V(t) \ge u_R$3
Eulerian droplet model with drag	$u_L \ge V(t) \ge u_R$4	Shock states and speeds relax exponentially to $u_L \ge V(t) \ge u_R$5
Generalized zero-pressure system with linear damping	$u_L \ge V(t) \ge u_R$6	Shock path and weight saturate as $u_L \ge V(t) \ge u_R$7
$u_L \ge V(t) \ge u_R$8 Temple-class system	$u_L \ge V(t) \ge u_R$9	All characteristic fields are linearly degenerate
Relativistic full Euler for Chaplygin gas	$U(x,t)=\overline U(x,t)+w(t)\delta(x-s(t))\eta,$0, $U(x,t)=\overline U(x,t)+w(t)\delta(x-s(t))\eta,$1	Both densities carry Dirac masses

For the pressureless Riemann problem, the canonical dichotomy is exact. If $U(x,t)=\overline U(x,t)+w(t)\delta(x-s(t))\eta,$2, the solution is

$U(x,t)=\overline U(x,t)+w(t)\delta(x-s(t))\eta,$3

while if $U(x,t)=\overline U(x,t)+w(t)\delta(x-s(t))\eta,$4, the admissible delta shock moves with

$U(x,t)=\overline U(x,t)+w(t)\delta(x-s(t))\eta,$5

in the transport formulation (Shao, 2015). The free-piston formulation gives the same dynamics and interprets the delta shock as a rigid body that sticks all impinging particles (Gao et al., 2022).

For the Eulerian droplet model, the Riemann solver is constructive: first solve the Burgers equation with source, then the $U(x,t)=\overline U(x,t)+w(t)\delta(x-s(t))\eta,$6-subsystem, then the full model. In the shock case $U(x,t)=\overline U(x,t)+w(t)\delta(x-s(t))\eta,$7, the left and right states evolve as

$U(x,t)=\overline U(x,t)+w(t)\delta(x-s(t))\eta,$8

and the Burgers shock speed is

$U(x,t)=\overline U(x,t)+w(t)\delta(x-s(t))\eta,$9

In the rarefaction case $U=(u,v)^T$0, the solution contains an intermediate vacuum with $U=(u,v)^T$1 in the fan region (Keita et al., 2017).

For the generalized one-dimensional zero-pressure system with linear damping,

$U=(u,v)^T$2

the delta-shock solution for $U=(u,v)^T$3 has

$U=(u,v)^T$4

with admissibility $U=(u,v)^T$5. The damping causes the shock to slow and the accumulated mass to saturate to a finite limit (cruz et al., 2020).

Other one-dimensional systems modify the same pattern. In the $U=(u,v)^T$6 Temple-class Suliciu relaxation system, all characteristic fields are linearly degenerate, and the Riemann problem admits a unique admissible delta shock when $U=(u,v)^T$7 and $U=(u,v)^T$8, with explicit formulas for $U=(u,v)^T$9, $\alpha$0, and the shock value of the third component (cruz et al., 2015). In the relativistic full Euler equations for a Chaplygin gas, the classical regime consists of three contact discontinuities, but when

$\alpha$1

the three-contact construction collapses and a single delta shock forms, now carrying Dirac masses in both $\alpha$2 and $\alpha$3 (Shao, 2017).

5. Multidimensional geometry

The two-dimensional theory adds geometric structure absent in one space dimension. For the zero-pressure Euler system with internal energy, the state variables are $\alpha$4, $\alpha$5, $\alpha$6, and $\alpha$7, and in any direction $\alpha$8 all four eigenvalues coincide,

$\alpha$9

The characteristic fields are linearly degenerate, and in self-similar variables $u$0 the pseudo-characteristics are straight lines. Delta shocks arise when these families overlap globally and must be resolved by measure-valued solutions with concentration on a moving curve (Pandey et al., 23 Jul 2025).

For three-state Riemann data posed in three quadrants, the cited work derives nine topologically distinct solution patterns. These include configurations with three delta shocks, mixed patterns with contacts and delta shocks, and cases in which vacuum wedges are enclosed by delta-shock boundaries (Pandey et al., 23 Jul 2025). Some patterns are described as Mach-reflection-like: an incident delta shock interacts at a triple point, new delta shocks bifurcate, and a triangular vacuum domain is enclosed by the resulting fronts (Pandey et al., 23 Jul 2025).

The two-dimensional formulas retain the square-root weighting familiar from one-dimensional pressureless dynamics. When two constant states are joined by a straight delta shock, the shock velocity is

$u$1

and the concentrated mass and internal-energy strengths are explicit functions of the jumps and of the transport direction (Pandey et al., 23 Jul 2025). This suggests that the weighted-average speed is not restricted to the one-dimensional Riemann problem, but persists in higher-dimensional self-similar geometries when the shock front remains straight.

A common misconception is that vacuum and delta shocks are mutually exclusive alternatives. The two-dimensional three-state problem shows otherwise: vacuum regions may be bounded by delta shocks and contact discontinuities, and some global patterns require both concentration and vacuum simultaneously (Pandey et al., 23 Jul 2025).

6. Physical interpretations, singular limits, and computation

The physical interpretation most directly emphasized in the cited literature is sticky aggregation. In pressureless Euler, particles collide and stick because there is no pressure to separate them; in the free-piston formulation, the delta shock is exactly a moving piston that accumulates all incident particles and obeys the same ODE as the generalized Rankine–Hugoniot system (Gao et al., 2022). In the Eulerian droplet model, delta shocks represent accumulation of droplets: faster trailing droplets catch slower leading droplets and concentrate droplet volume fraction $u$2 on a moving front (Keita et al., 2017).

Several analytical constructions justify delta shocks as limits of regularized problems. The generalized damped zero-pressure system is treated by a time-dependent viscous approximation and a vanishing-viscosity argument in similarity variables; the limit produces a single concentration point $u$3 with explicit delta strength (cruz et al., 2020). The flux-approximation framework shows that delta shocks and vacuum states are stable as flux perturbations vanish (Yang et al., 2014). The vanishing pressure and magnetic-field limit for Chaplygin magnetogasdynamics shows how a two-shock classical solution collapses into a delta shock, with the intermediate density blowing up while the width of the intermediate layer shrinks to zero (Shao, 2015).

The numerical literature represented here is designed to capture both concentration and vacuum without spurious oscillations. For the Eulerian droplet model, numerical illustrations use the Transport–Collapse method, following Bouchut, with typical parameters $u$4, $u$5, $u$6, and $u$7; the computed concentrated peak in $u$8 and the discontinuity in $u$9 agree with the theoretical shock trajectory, shock speed, and weight growth (Keita et al., 2017). For the two-dimensional three-state Riemann problem, the analytical constructions are validated by a second-order semidiscrete central-upwind scheme on a uniform $$\rho(x,t)=\rho^{-}(x,t)\,H\big(X(t)-x\big)+\rho^{+}(x,t)\,H\big(x-X(t)\big)+w(t)\,\delta\big(x-X(t)\big),$$00 grid on $$\rho(x,t)=\rho^{-}(x,t)\,H\big(X(t)-x\big)+\rho^{+}(x,t)\,H\big(x-X(t)\big)+w(t)\,\delta\big(x-X(t)\big),$$01, with final time $$\rho(x,t)=\rho^{-}(x,t)\,H\big(X(t)-x\big)+\rho^{+}(x,t)\,H\big(x-X(t)\big)+w(t)\,\delta\big(x-X(t)\big),$$02 and $$\rho(x,t)=\rho^{-}(x,t)\,H\big(X(t)-x\big)+\rho^{+}(x,t)\,H\big(x-X(t)\big)+w(t)\,\delta\big(x-X(t)\big),$$03; the simulations reproduce delta shocks, triple-point geometry, and vacuum wedges with excellent agreement (Pandey et al., 23 Jul 2025).

The present literature also delineates unresolved directions. The embedded constructions provide explicit self-similar solutions but do not prove uniqueness or stability for those patterns (Castañeda, 2019). The vanishing-viscosity analysis for linear damping is confined to Riemann data and odd $$\rho(x,t)=\rho^{-}(x,t)\,H\big(X(t)-x\big)+\rho^{+}(x,t)\,H\big(x-X(t)\big)+w(t)\,\delta\big(x-X(t)\big),$$04, and does not address general well-posedness beyond the Riemann class (cruz et al., 2020). For the $$\rho(x,t)=\rho^{-}(x,t)\,H\big(X(t)-x\big)+\rho^{+}(x,t)\,H\big(x-X(t)\big)+w(t)\,\delta\big(x-X(t)\big),$$05 Temple-class system, the authors identify regions where neither the classical solution nor the delta-shock construction applies, leaving the appropriate generalized solution as an open problem (cruz et al., 2015).