A hyperbolic Kepler equation solver with four Newton iterations and a computer-assisted uniform error bound
Year: 2026
Type: article
Abstract: The hyperbolic Kepler equation $$e\sinh (H) - H = M$$ e sinh ( H ) - H = M , relating the mean anomaly M to the hyperbolic anomaly H at eccentricity $$e > 1$$ e > 1 , is solved at every epoch along planetary flybys and the passages of interstellar objects. Toward the parabolic limit $$e = 1$$ e = 1 at small M , the factor multiplying the squared error at each Newton step grows without bound. I present a solver that works in the variable $$x = \sinh H$$ x = sinh H : one comparison selects a closed-form starting value, four Newton iterations follow with no stopping test, and one inverse hyperbolic sine returns H . In exact arithmetic the relative error in H is at most $$C_4 = 9.02\times 10^{-20}$$ C 4 = 9.02 × 10 - 20 for all $$e > 1$$ e > 1 and $$M \ne 0$$ M ≠ 0 , below the double-precision machine epsilon $$2^{-52}$$ 2 - 52 . The proof reduces to finitely many numerical inequalities, each verified in interval arithmetic. On $$102\,600$$ 102 600 test points over $$e - 1 \in [10^{-14}, 10^{8}]$$ e - 1 ∈ [ 10 - 14 , 10 8 ] and $$M \in [10^{-48}, 10^{12}]$$ M ∈ [ 10 - 48 , 10 12 ] , the largest relative error in H
Author: Hitaishi Chillara
Institution: University of North Texas
Language: English
Subfield: Computational Theory and Mathematics
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
Estimated APC paid: $3,190
OpenAlex ID: W7204821633
