Researchers Chinedu Nwaigwe and Sanda Micula have introduced the first verifiable convergent second-order numerical method with complete theoretical analysis for one-dimensional viscoelastic fluid flow models. Viscoelastic fluids, particularly third-grade non-Newtonian substances, are widely used across diverse industrial sectors, including commercial paint manufacturing, food processing, and drilling operations. Despite their extensive industrial utility, the partial differential equations describing third-grade fluid motion exhibit severe mathematical complexity due to high-order mixed spatio-temporal derivatives, nonlinear advective transport, variable transport properties, and flow resistance within porous channels. Because these equations resist exact analytical solutions, computational fluid dynamics practitioners require stable numerical algorithms that preserve underlying physics while guaranteeing verifiable convergence. The newly developed algorithm addresses this challenge by introducing a stable finite difference framework that achieves second-order spatial accuracy and provable stability without omitting any physical mechanisms.
Prior computational research examining viscoelastic third-grade flows routinely compromised the physical fidelity of governing models to circumvent mathematical intractability. Previous investigators frequently dropped critical physical mechanisms, including convective transport, unsteady time-dependent terms, and variable transport properties such as state-dependent fluid viscosity. For example, earlier investigations into plate-driven third-grade flows developed hybrid finite difference techniques that omitted variable viscosity and convection, leaving the resulting formulations unverified against theoretical error bounds. Other studies exploring species diffusion over porous plates discarded unsteadiness and variable viscosity altogether, producing algorithms whose mathematical convergence and truncation errors were never established. Similarly, investigations analyzing radiative surfaces, magnetized pipe flows, and horizontal porous channels omitted variable suction velocity, unsteady dynamics, or convective acceleration. Across this prior literature, existing numerical solvers remained limited to first-order accuracy and lacked rigorous convergence proofs, meaning calculated solutions could not be trusted for industrial applications.
The physical problem investigated by the authors models the unsteady flow of a third-grade viscoelastic fluid through a rectangular channel filled with porous material. In this governing initial-boundary value formulation, fluid velocity depends on spatial position across a unit channel and temporal progression, subject to variable suction velocity and variable dynamic viscosity. The porous environment is characterized by a positive Darcy number, while constant positive parameters govern third-grade non-Newtonian viscoelastic behavior alongside applied external body forces. Mathematically, the governing equation equates temporal acceleration and convective transport to the spatial divergence of viscous shear stresses, a third-order mixed spatio-temporal viscoelastic derivative, a nonlinear second-degree derivative scaling with squared velocity gradients, linear porous drag inversely proportional to the Darcy number, and external forcing. Prescribed boundary conditions fix fluid velocity at both channel boundaries, while a known function defines the initial state across the spatial domain.
To establish a stable discretization framework, the researchers restructured the governing equation by regrouping key differential operators. Specifically, they combined the first-order temporal derivative of velocity with the third-order mixed spatio-temporal viscoelastic term, expressing them collectively as the time derivative of a second-order spatial differential operator. This analytical rearrangement isolates the temporal evolution of the viscoelastic stress state while transforming remaining terms into an explicit balance of convective transport, variable diffusion, nonlinear gradient interactions, Darcy porous damping, and external source terms. Spatial and temporal domains are systematically discretized by selecting positive integers defining spatial mesh spacing and discrete time step size up to a designated terminal simulation time. Spatial grid points span the unit interval uniformly, while temporal steps mark discrete evaluation intervals, setting the foundation for a discrete representation that avoids artificial numerical oscillations during simulations.
A central innovation in the proposed algorithm lies in the coupled numerical discretization of convective and variable diffusive transport terms. Convective transport is governed by an effective convective velocity defined as the negative of suction velocity. To prevent numerical instability without introducing excessive artificial diffusion, the authors decomposed the convective velocity into non-negative and non-positive components using absolute value formulations. These split velocity components are scaled by variable fluid viscosity and incorporated into localized mesh parameters that account for local cell Reynolds numbers. The combined convection-diffusion operator is then discretized across neighboring spatial nodes using weighted forward and backward differences scaled by arithmetic mean viscosities evaluated at spatial half-steps. This formulation dynamically adapts to local flow direction and viscosity gradients, ensuring robust upwind characteristics where advection dominates while smoothly transitioning to central diffusion approximations in regions dominated by viscous dissipation.
The remaining differential terms within the rearranged governing equation are discretized using a carefully balanced combination of implicit and explicit algebraic approximations. The complex nonlinear second-degree term, which multiplies the second spatial derivative by the squared spatial gradient of velocity, is approximated at the advanced time level using a central difference second derivative evaluated implicitly, multiplied by a central difference spatial gradient computed explicitly from the preceding time level. This semi-implicit decoupling linearizes quadratic gradient interactions without degrading overall spatial accuracy. Meanwhile, the Darcy porous medium resistance term is evaluated implicitly at the forward time level using fluid viscosity calculated from the known velocity field of the current time step. External source terms are evaluated directly at the forward time level, while the combined unsteady viscoelastic operator is approximated using a backward temporal difference quotient applied to second-order central spatial differences.
Assembling these discrete approximations yields a linear algebraic system at each forward time level possessing an elegant tridiagonal structure. The discrete formulation takes a canonical linear form where the unknown solution at the current spatial index and its adjacent neighboring nodes are equated to a known right-hand source vector. The tridiagonal coefficients, designated as lower, central, and upper matrix entries, are constructed directly from spatial grid dimensions, time step size, viscoelastic constants, Darcy porous parameters, and directional convective-diffusive weights. The diagonal central coefficient incorporates contributions from both neighboring connection weights alongside an intrinsic positive unit term and Darcy damping. Known boundary conditions enforce prescribed solution values at channel boundaries, fully closing the algebraic system. Because the resulting coefficient matrix is strictly tridiagonal and structurally diagonally dominant, the linear system can be solved with exceptional computational efficiency during each advancing time step.
To establish the formal accuracy of their proposed numerical algorithm, the authors conducted a rigorous consistency analysis based on multidimensional Taylor series expansions. By substituting the exact analytical solution into the discrete finite difference scheme, they defined the local truncation error as the difference between discrete operational components and exact differential terms. Expanding each discrete spatial difference around individual grid points revealed that higher-order expansion terms cancel systematically due to symmetric stencil placements. The coupled convection-diffusion approximation was shown to reproduce true analytical derivatives with residual errors scaling strictly with the square of spatial grid spacing. Similarly, temporal differencing yielded first-order temporal truncation residuals. Combining these expansions demonstrated that the global local truncation error is second-order accurate in space and first-order accurate in time. Consequently, the scheme satisfies formal mathematical consistency, confirming that the discrete equation converges toward the continuous physical system as mesh intervals vanish.
Beyond theoretical consistency, practical computational algorithms must guarantee that computed solutions remain physically bounded and free from spurious numerical oscillations. The authors proved that their discrete canonical formulation inherently satisfies a strict positivity condition across all spatial grid points and time levels. By analyzing the structural definitions of the tridiagonal matrix entries, they verified that off-diagonal coefficients remain strictly non-negative while the diagonal central entry remains strictly positive. Furthermore, the difference between the diagonal coefficient and the sum of off-diagonal entries is strictly positive, governed by the porous Darcy resistance parameter. Leveraging this structural positivity, the authors proved a two-sided boundedness theorem establishing that the computed numerical solution at any forward time step is bounded above and below by the minimum and maximum scaled source terms from the preceding time level. This analytical guarantee prevents the emergence of unphysical solution blowup during prolonged simulation runs.
The foundational theoretical milestone of the paper is the complete mathematical proof of convergence establishing that discrete numerical approximations converge to the exact continuous solution. The convergence theorem applies to any viscoelastic flow model whose physical coefficients remain smooth with continuous first derivatives. To bridge the gap between theory and computation, the authors formulated an intermediate lemma demonstrating that matrix coefficients evaluated using numerical approximations deviate from exact counterparts by an error proportional to the maximum solution discrepancy. Substituting exact solutions into the discrete system and subtracting the numerical formulation produced an explicit error recurrence relation. By applying the previously established positivity bounds and diagonal dominance properties, the authors derived an upper bound for the maximum absolute error at the forward time level, demonstrating that global computational error contracts toward zero as spatial spacing, time step size, and initial discrete errors vanish.
To corroborate their theoretical deductions, the authors implemented the complete numerical algorithm within an in-house C++ computational software package utilizing the Eigen linear algebra library for matrix operations. They conducted rigorous verification benchmarks across multiple test cases equipped with known exact analytical solutions under varying parameter settings. In the primary benchmark problem featuring exponential state-dependent viscosity, simulations were executed across progressively refined spatial grids ranging from coarse meshes to highly refined partitions, setting time steps proportional to the square of grid spacing. The observed experimental order of convergence matched the theoretical value of two, confirming second-order spatial accuracy. Surface plots demonstrated flawless numerical stability over extended simulation intervals without spurious wiggles or computational damping. Additionally, contour plots comparing numerical approximations directly with exact analytical fields revealed complete visual and quantitative agreement, validating the algorithm as a reliable tool for fluid dynamics simulations.
Subject of Research: Second-order numerical scheme and theoretical convergence analysis for one-dimensional viscoelastic fluid flow models
Article Title: Second order algorithm with convergence analysis for viscoelastic flow model
Article References: Nwaigwe, C., & Micula, S. (2025). Second order algorithm with convergence analysis for viscoelastic flow model. Journal of Mathematics in Industry, 15(1), Article 11. https://doi.org/10.1186/s13362-025-00176-x
Image Credits: AI Generated
DOI: 10.1186/s13362-025-00176-x
Keywords: viscoelastic fluids, third-grade fluid, numerical analysis, convergence analysis, second-order method, fluid dynamics, finite difference scheme, variable viscosity, suction velocity, Darcy number, positivity preserving, truncation error, Taylor series, convection diffusion, porous material, Eigen C++ library, experimental order of convergence, initial boundary value problem
Cite Scienmag News
Reid Dalton. (November 17, 2026). Second-Order Numerical Method Proves Verifiable Convergence for One-Dimensional Viscoelastic Flow Models. Scienmag. https://scienmag.com/second-order-numerical-method-proves-verifiable-convergence-for-one-dimensional-viscoelastic-flow-models/
Reid Dalton. "Second-Order Numerical Method Proves Verifiable Convergence for One-Dimensional Viscoelastic Flow Models." Scienmag, 17 November 2026, https://scienmag.com/second-order-numerical-method-proves-verifiable-convergence-for-one-dimensional-viscoelastic-flow-models/. Accessed 17 November 2026.
Reid Dalton. "Second-Order Numerical Method Proves Verifiable Convergence for One-Dimensional Viscoelastic Flow Models." Scienmag. November 17, 2026. https://scienmag.com/second-order-numerical-method-proves-verifiable-convergence-for-one-dimensional-viscoelastic-flow-models/








