Skip to content

Latest commit

 

History

History
65 lines (43 loc) · 7.97 KB

SOURCES.md

File metadata and controls

65 lines (43 loc) · 7.97 KB

Mathematical sources and acknowledgments

Course contents · Method explanations

Course presentation

The grouped table of contents, chapter-oriented Markdown reading flow, and links between explanation and code are inspired by CSC207-UofT/207-course-notes, inspected for this documentation edition. AstroSimBench's physics prose, examples, and captions are written for this project. No CSC207 Java lesson text or repository assets are included, and no affiliation or endorsement is implied.

Published algorithms and local implementations

Local method Mathematical source Relationship to this project
Euler / RK4 / Leapfrog Standard numerical integration formulas retained from the existing project Educational implementation, not a new algorithm claim
RK45 Dormand & Prince (1980) Published 5(4) tableau; local stage evaluation and controller integration
Hermite4 Makino & Aarseth (1992) Basic shared-step PEC equations; no Ahmad–Cohen neighbor or block-step scheme
BS Bulirsch & Stoer (1966) Modified midpoint and polynomial extrapolation in an educational implementation
GL4 Hairer: symplectic integrators Standard two-stage Gauss tableau, local finite-difference Newton solve
WH2 Wisdom & Holman (1991) Jacobi Hamiltonian splitting with the project's spatial Kepler subflow
WH4 Yoshida (1990) Published symmetric composition coefficients applied to WH2
Universal Kepler propagation Universal-variable / Stumpff f,g equations; background in Shepperd (1985) Bracketed Newton and Stumpff evaluation; not a reproduction of Shepperd's continued-fraction or state-transition implementation
IAS15 Rein & Spiegel, IAS15 paper External production reference through REBOUND

Algorithm formulas and theory remain attributable to their published sources. Implementing a published algorithm is not a claim to invent it. The preserved code's provenance states that third-party solver source was not copied into this package.

External software and concepts

REBOUND remains an optional package dependency needed for the reference experiments; its source is not vendored here. The existing core test suite calls SciPy DOP853 as an additional validation where available. Saved settings record the actual reference version used (5.1.1 in the original runs, 5.2.2 in Part 4); the supported dependency range is declared in pyproject.toml.

Licensing scope

Version 1.0.0 licenses the original code under MIT and the original teaching prose, project-generated figures and result collections under CC BY 4.0, where those rights apply. Dependencies retain their own terms, including REBOUND's GPL-3.0-only license. See Licensing and attribution for scope and combined-distribution considerations. Citation of an algorithm is separate from licensing its implementation.

Part 4 processor and external controls

Part 4B and 4C

  • Wisdom (2015), Resolving the Pericenter: motivation for testing the local resolution of perturbed eccentric motion. Its empirical step counts are not universal acceptance conditions.
  • Rein et al. (2019), MERCURIUS, and the official adapter settings: external hybrid candidate, with no local changeover implementation.
  • Lu, Hernandez & Rein (2024), TRACE, and the official settings: external almost time-reversible encounter candidate. No paper or REBOUND source code is vendored.

Part 4E: secular dynamics and reference limits

The bundled Part 4 references and external candidates were run with REBOUND 5.2.2. Verification explains the independent BS and output-sampling checks, including the limits of full-trajectory agreement in the 120-degree case. New compatible environments record their own versions and empirical reference floors.

Part 4 hybrid distance context

The official MERCURIUS settings describe multiple critical-distance criteria and a pair changeover based on the larger critical distance. The official TRACE settings describe a modified Hill scale using heliocentric distance, together with a separate pericenter condition. Our new Hill-normalized columns are explicitly geometric context, not solver telemetry or a count of internal switches.

Resolution-scale comparison in 4B

Wisdom (2015), Figures 1–4 and the conclusion, supplies a reported empirical scale, not data points for our state-error plot. The lesson converts the period convention explicitly. Kaib, Quinn & Brasser (2011), AJ 141, 3, preprint Section 1, discusses corrector error scaling and timestep/partition changes. No figures from either paper are reproduced.