Abstract:
The numerical integration of stiff ordinary differential equations is frequently hindered by the classical Dahlquist barrier and severe absolute stability constraints, forcing a reliance on prohibitively small step sizes or computationally expensive extended formulas. To resolve these algorithmic bottlenecks, this study formulates a robust class of A-stable, two-step multi-derivative hybrid block methods for direct integration. Derived via continuous polynomial power series interpolation and dynamic intra-step collocation, these discrete formulas incorporate analytical higher-order derivatives up to the fourth order to guarantee unconditional A-stability across the open left-half complex plane. Theoretical analysis establishes that the schemes are zero-stable, consistent, and globally convergent, uniformly achieving an algebraic order of eight (\( p = 8 \)). The primary novelty lies in allocating a dynamically shifted single off-grid collocation point within a fourth-derivative architecture, successfully bypassing traditional stability limitations while maintaining a rigidly low step number (\( k = 2 \)). Extensive numerical simulations on severe stiff systems, rapid transient modes, and chemical reaction dynamics demonstrate that the proposed integrators suppress stiffness-induced error accumulation, yielding exceptionally tighter precision bounds and significant computational speedups over existing state-of-the-art block techniques.





Open Access