Nonlinear Compressible Potential Equation

• Nonlinear compressible potential equations are quasilinear PDEs that model inviscid, irrotational, compressible flows using a velocity potential, with nonlinearity driven by the equation of state.
• The equations exhibit varying analytical structures—elliptic in subsonic flows and hyperbolic in supersonic conditions—necessitating distinct solution techniques.
• Applications span aerodynamic design, porous media flow, and theoretical gas dynamics, with advanced solvers like multi-stage PINNs enhancing accuracy and error control.

The nonlinear compressible potential equation is a class of quasilinear partial differential equations (PDEs) modeling inviscid, irrotational, compressible flows where nonlinearity arises from the equation of state and the structure of the velocity potential. These equations constitute the cornerstone for the analysis and computation of gas dynamics phenomena in subsonic, transonic, and supersonic regimes, including applications to flow in porous media, aeroacoustics, and high-speed aerodynamics. Several distinct, but mathematically related, formulations appear in the literature depending on the physical context—namely, the steady or unsteady full potential equation for compressible flow, doubly nonlinear potential equations for Forchheimer flows in porous media, and perturbation potential equations for subsonic aerodynamics.

1. Mathematical Formulation and Origin

Nonlinear compressible potential equations arise from the isentropic or barotropic Euler equations under the assumption of irrotational flow. For a fluid with density $\rho$ and pressure $p(\rho)$, the velocity field %%%%2%%%% is written in terms of a potential: $v = \nabla \phi$ (Elling, 2014). The governing mass conservation and momentum equations reduce to a system for $\phi$ via

$$\partial_t \left[\pi^{-1}\left(-\phi_t - \tfrac12 |\nabla \phi|^2\right)\right] + \nabla \cdot \left(\pi^{-1}\left(-\phi_t - \tfrac12 |\nabla \phi|^2\right) \nabla \phi\right) = 0$$

with $\pi'(\rho) = p'(\rho)/\rho$. For steady flow, the compressible potential equation takes the form (Kalinowski, 2016):

$$\sum_{i=1}^n (c^2 - \phi_{x_i}^2) \phi_{x_ix_i} - 2 \sum_{i<j} \phi_{x_i} \phi_{x_j} \phi_{x_ix_j} = 0$$

where $c^2 = p'(\rho)$ is the local speed of sound and the structure of nonlinearity depends on the Mach number.

In the context of flows through porous media obeying generalized Forchheimer laws, the nonlinear compressible potential equation further involves both the nonlinear dependence of flux on gradients and the compressible equation of state. Introducing a pseudo-pressure variable $u$ such that $\nabla u = \rho \nabla p$, the canonical doubly nonlinear parabolic equation is (Celik et al., 2016):

$$\partial_t\big(u^\alpha\big) = \nabla \cdot \big[ K(|\nabla u + Z(u)|)(\nabla u + Z(u)) \big] + f(x,t,u)$$

where $\alpha \in (0,1]$ captures the gas law nonlinearity, $K(\cdot)$ encodes nonlinear permeability, and $Z(u)$ represents gravity or shift effects.

2. Analytical Structure and Classification

The analytical structure of the nonlinear compressible potential equations is governed by their nonlinearity in both the principal part and, in the Forchheimer case, in the time derivative. The steadiness, the value of the Mach number, and the specific equation of state lead to differences in ellipticity, hyperbolicity, and solution regularity. In the subsonic regime ($|v| < c$), the equations are typically elliptic or parabolic, and admit stable, unique solutions; in the supersonic regime ($|v| > c$), the system becomes non-elliptic and more akin to hyperbolic PDEs (Kalinowski, 2016).

For the steady three-dimensional compressible potential flow, the second-order quasilinear PDE admits a geometric classification of solutions via Riemann invariants and the characteristic cone. The analysis by explicit “algebraization” methods leads to a parametrization of all rank-1 integral elements, corresponding to simple wave solutions. Eight families (denoted $F_1, ..., F_4$ and primed extensions) exhaust all simple-wave types and enable the explicit integration of multi-parametric solutions even in the supersonic regime (Kalinowski, 2016).

For the Forchheimer model, the doubly nonlinear diffusion term, the nonlinear source, and the polynomial growth in boundary and lower-order terms complicate the analysis. Nevertheless, existence, uniqueness, a priori $L^p$ and $L^\infty$ estimates, and regularity can be achieved by combining advanced techniques (trace theorems, Sobolev inequalities, and Moser iteration) (Celik et al., 2016).

3. Boundary Value Problems and Physical Conditions

The physical relevance of nonlinear compressible potential equations derives from their boundary value problem formulations. Typical boundary conditions include:

• Far-field asymptotics: Ensuring correct decay for $\phi$ and its derivatives at infinity. For aerodynamic simulations, this is crucial to suppress truncation artifacts (Qian et al., 1 Jan 2026).
• Solid boundary (body) conditions: These may be encoded by a flow tangency (Neumann-type) constraint, e.g., $\mathbf{V}\cdot \mathbf{n}=0$ on solid surfaces, expressed for a parametrized boundary such as an ellipse (Qian et al., 1 Jan 2026).
• Nonlinear Robin-type boundary fluxes: Especially in porous-media contexts, flux across boundaries is modeled as $$(K(|\nabla u + Z(u)|)(\nabla u + Z(u)))\cdot n = B(x,t,u)$$ with $B(x,t,u)$ of polynomial growth (Celik et al., 2016).

In the full potential flow, initial conditions are also crucial for unsteady problems, as they strongly affect finite-speed propagation and regularity (Elling, 2014).

4. Numerical Methods and Machine-Learning-Based Solvers

Classical numerical methods rely on either finite difference/volume discretizations or iterative solvers for nonlinear PDEs. However, recent advancements have focused on Physics-Informed Neural Networks (PINNs) and, in particular, multi-stage PINN (MS-PINN) architectures for resolving the nonlinear compressible potential equation in truly unbounded domains (Qian et al., 1 Jan 2026).

A summary of the methodology:

• Equation form: The fully nonlinear perturbation potential equation is posed for steady, irrotational, compressible flow, with locally varying coefficients and cross-derivative terms.
• Domain treatment: A coordinate mapping $q = (1 + r^2)^{-\alpha/2}$, $\beta = x^2/(x^2 + y^2)$ compactifies the infinite physical domain.
• Network architecture: A deep feedforward network encodes the potential, with asymptotic constraints (hard-decay embedding) and symmetry built in.
• Multi-stage residual minimization: The MS-PINN iteratively fits residuals of increasing frequency, achieving exponential error decay and machine-precision accuracy.
• Loss function: A weighted sum of mean-squared errors on boundary conditions (including at infinity), body-surface conditions, and the PDE residual guides training.

This approach not only achieves much higher global accuracy compared to truncated-domain or linearized solvers but also enables quantification of linearization and truncation errors throughout the domain (Qian et al., 1 Jan 2026).

5. Entropy Structure, Uniqueness, and Weak Solutions

For the compressible potential flow equation, the system can be cast as a conservation law system for $(\rho, v)$, restricted to curl-free velocity fields. The total energy $E(\rho,v)$ provides a strictly convex entropy only for $|v| < c$, i.e., in the subsonic regime (Elling, 2014). The relative entropy framework leads to crucial qualitative properties:

• Finite speed of propagation: Perturbations to smooth entropy solutions propagate only finitely fast, in strong contrast to certain ill-posed Euler flows with non-unique weak solutions.
• Conjectured uniqueness: Owing to the vorticity-free constraint, it is conjectured—supported by the absence of convex-integration counterexamples for this system—that compressible potential flow admits at most one admissible (entropy) weak solution for given irrotational initial data, unlike the non-unique multidimensional Euler system (Elling, 2014).

A plausible implication is that the potential flow model provides a more robust mathematical foundation for weak solutions in compressible gas dynamics than the general Euler equations.

6. Exact and Special Solutions

In certain cases, notably supersonic stationary flows, the nonlinear compressible potential equation admits explicit families of solutions in terms of simple Riemann invariants. For the quasilinear PDE in three dimensions, solutions of the form $u(x) = f(R)$ with $R = \lambda \cdot x$ can be constructed, where $\lambda$ is a simple element solving a characteristic cone equation in phase space. Classes of solutions (labeled $F_i$) are parametrized by free functions and constants, yielding a multi-parametric catalogue of nonlinear stationary waves, their envelopes, and associated shock-formation loci (Kalinowski, 2016).

The methodology enables the construction of nontrivial standing-wave solutions, multi-dimensional analogs, and a gauge/Bäcklund transformation structure linking related solutions within a class. The approach generalizes to unsteady flows, yielding closed-form families for the time-dependent nonlinear potential equation as well.

7. Applications and Modeling Implications

The nonlinear compressible potential equation models a broad range of physical phenomena:

• Aerodynamics: Accurate computation of subsonic and transonic flow over complex geometries, as in aircraft design, benefits from full nonlinear treatment rather than simplified (linearized) equations. Quantitative estimates show that linearization can induce $O(1)$-level relative errors at moderate Mach numbers, particularly near solid bodies (Qian et al., 1 Jan 2026).
• Porous Media Flows: Realistic modeling of gas flows in fractured rock or high-Reynolds-number porous domains necessitates Forchheimer extensions, captured by doubly nonlinear pseudo-pressure equations. These models account for gas-law nonlinearity and allow treatment of gravity, complex boundary conditions, and strongly nonlinear fluxes (Celik et al., 2016).
• Theoretical Gas Dynamics: The equation provides a testbed for analysis of nonlinear wave phenomena, finite-speed propagation, uniqueness, and formation of singularities/gradient catastrophes (Kalinowski, 2016).

Rigorous a priori estimates for solutions and gradients underpin further studies of qualitative behavior (e.g., asymptotic states, free-boundary evolution) and are foundational for the design of robust numerical schemes (Celik et al., 2016). The conjectured uniqueness of entropy solutions highlights the desirability of potential-flow-based models for well-posedness in multidimensional gas dynamics (Elling, 2014).