Pith. sign in

REVIEW 5 major objections 6 minor 34 references

An adaptive symplectic integrator for gravitational dynamics

T0 review · 5 major / 6 minor · reviewed 2026-08-06 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read The paper claims that SQQ-PTQ, an adaptive symplectic integrator built from SQQ with a time transformation, Chebyshev interpolation, and Broyden's quasi-Newton solver, improves efficiency and applicability while keeping energy bounded…

desk verdict A modest but useful engineering extension of the SQQ symplectic integrator with adaptive time stepping; the numerical evidence is decent, but symplecticity is asserted rather than proven and the paper skips comparisons against the standard adaptive symplectic methods. read the letter →

arxiv 2507.14881 v1 pith:XWCDIEMI submitted 2025-07-20 math.NA astro-ph.GAcs.NA

Keqi Ye , Zizhe Cai , Mingji Wang , Kun Yang , Xiaodong Liu This is my paper · ORCID
classification math.NAastro-ph.GAcs.NA MSC 65P1070F1065H10
keywords symplecticintegratoradaptivetimesteptransformationquasi-NewtonmethodBroyden'sgeneratingfunctiongravitationalN-bodyChebyshevinterpolation
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper sets out to show that a fixed-step symplectic integrator can be made adaptive without giving up its long-term energy behavior. It introduces SQQ-PTQ, a variant of the generating-function integrator SQQ, which adds a time transformation for adaptive step sizes, Chebyshev interpolation to avoid the Runge phenomenon, a projection method to speed up interpolation evaluations, and Broyden's quasi-Newton method to solve the resulting implicit equations. The reported experiments—Kepler orbits with eccentricity up to 0.99, a three-body configuration with close encounters, and the outer Solar System—show bounded energy error over thousands of periods, with CPU times substantially below Newton-based and non-adaptive alternatives. A sympathetic reader would take the contribution to be a practical adaptive symplectic integrator for gravitational N-body problems.

What carries the argument

The machinery is the generating-function action S(q(ta), q(tb)): stationarity of the discretized action yields the nonlinear system whose solution defines the map from one endpoint to the other. SQQ is a fixed-step symplectic integrator built from this least-action generating function with generalized coordinates as independent variables at both ends. SQQ-PTQ approximates the action with Chebyshev interpolation and Gaussian quadrature, accelerates the interpolation evaluations with a projection method that precomputes basis functions on a reference interval, makes the step size adaptive through the time transformation dt/dτ = σ(p,q) with the Darboux-Sundman Hamiltonian K = σ(H - H0), and solves the implicit equations with Broyden's update for the inverse Jacobian.

What would settle it

Compare SQQ-PTQ's orbital phase, such as perihelion time or longitude, against a high-precision reference over 10,000 Kepler periods, and rerun the three-body test with the nonlinear tolerance loosened to 1e-8 and tightened to 1e-14; if phase error grows linearly while energy error stays bounded, or if the close-encounter timing shifts without an energy-error change, the energy diagnostic is masking a loss of symplecticity.

Watch

Extended reading notes

Core claim

The central claim is that the time-transformed Hamiltonian can be integrated symplectically at low cost if the implicit equations are solved by Broyden's method rather than Newton's method. Because Broyden's method reuses an approximate inverse Jacobian from the previous step, it avoids the complicated analytic Jacobian of the transformed system; the paper reports that SQQ-PTQ achieves essentially the same maximum energy error as SQQ-PTN while using one-third to one-sixth of the CPU time. Over 10,000 Jupiter periods the energy error stays bounded near 3.5e-8, which the authors read as confirmation that the map remains symplectic in practice.

Load-bearing premise

The load-bearing premise is that the quasi-Newton iteration converges, at every time step, to the exact solution of the discrete map's nonlinear equations within the 1e-12 tolerance; if that solve is not accurate enough, the map is no longer truly symplectic even when the energy error looks bounded.

Editorial extensions

If this is right

  • High-eccentricity Kepler orbits with e = 0.9 and e = 0.99 can be followed for 500 periods with bounded energy error, a regime where fixed-step SQQ becomes inefficient.
  • A three-body periodic configuration with close encounters down to separation approximately 0.014 is integrated for 500 periods without failure, with energy error oscillating stably.
  • In the outer Solar System, SQQ-PTQ keeps maximum energy error near 3.5e-8 over 10,000 Jupiter periods and uses about 58% of SQQ-PN's CPU time.
  • Quasi-Newton solving gives nearly identical energy-error curves to Newton solving while using less than one-third of the Newton-based CPU time in the three-body test and about one-sixth in the outer Solar System test.
  • The projection method suppresses the Runge phenomenon at high interpolation orders, keeping energy error bounded where plain SQQ drifts.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • A sharper test of the symplectic claim would track phase error, such as perihelion passage time, over 10,000 orbits, since bounded energy error does not by itself rule out secular phase drift.
  • Because Broyden's method is warm-started from the previous step, very stiff encounters not covered by these examples might require periodic Jacobian refresh; the paper does not explore that regime.
  • The projection method's speed advantage grows with interpolation order, so the same construction may make higher-order SQQ variants or higher-dimensional systems practical.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

5 major / 6 minor

Summary. The paper proposes an adaptive symplectic integrator, SQQ-PTQ, built on the fixed-step generating-function integrator SQQ. It combines Chebyshev interpolation to avoid the Runge phenomenon, a projection method to accelerate interpolation-function evaluation, a time transformation to implement adaptive time steps, and Broyden's quasi-Newton method to solve the resulting nonlinear equations without explicit Jacobians. The method is tested on the Kepler problem at high eccentricity, a three-body close-encounter problem, and an outer Solar System model. The reported experiments show bounded energy error over long integrations and CPU-time reductions relative to SQQ variants and ODE45.

Significance. If the central claim is established, SQQ-PTQ would be a practically useful adaptive symplectic integrator for gravitational N-body dynamics, combining the long-term energy behavior of symplectic methods with adaptive step sizing and reduced Jacobian cost. The paper contains several strengths: the method parameters (a, b, m, n, Delta-tau, epsilon) are user-chosen rather than fitted to the reported errors; the comparisons against SQQ variants and ODE45 provide independent baselines; and the three experiments cover close encounters and long integrations, where adaptive behavior is essential. However, the theoretical basis for the 'symplectic' claim is not demonstrated, and the numerical evidence consists almost entirely of energy-error plots, which are necessary but not sufficient for symplecticity.

major comments (5)
  1. [§3.3, Algorithm 1] The stopping criterion in Algorithm 1 is ek+1 = ||x_{k+1}-x_k||/||x_k|| < epsilon, i.e., a relative update size, not the residual norm ||F(x_k)||. For Broyden's method a small update does not imply a small residual when the approximate inverse Jacobian is ill-conditioned, so the paper does not establish that the converged iterate actually satisfies Eq. (8). Since the symplecticity of the discrete map depends on the nonlinear equations being solved accurately, this is a load-bearing gap. The paper should either prove that the Broyden iteration drives the residual to a prescribed tolerance, or report residual histories, iteration counts, and failure diagnostics for the three experiments.
  2. [§2, §3.2, and §5] The central claim that SQQ-PTQ is a symplectic integrator is asserted but never proved. The exact generating-function map with a fixed step in transformed time is symplectic only when Eq. (8) is solved exactly; the paper uses an approximate quasi-Newton solve with a loose coupling between the update size and the residual. No theorem, symplecticity-condition check (e.g., J^T Omega J = Omega), or back-error analysis is provided. The conclusion's statement that bounded energy 'confirms its characteristics as a symplectic algorithm' is not logically sufficient, because non-symplectic integrators with small steps can show bounded energy over hundreds of periods while phase errors accumulate.
  3. [§4.1, §4.2, and §4.3] All numerical validations report only energy error; no position or phase error is measured. The visual claim that 'the phase-space structure remains nearly intact' in Section 4.1 and Section 4.2 is not a quantitative substitute for a phase-error diagnostic. The paper should compare the computed trajectory against a high-accuracy reference solution (or check the periodicity of the three-body orbit) and report configurational error, since bounded energy can coexist with secular phase drift.
  4. [§4.2 and §4.3, Tables 4-6] The efficiency comparisons are based on CPU time at a single, fixed step size and fixed solver tolerance, rather than on error-vs-cost curves. The statement that SQQ-PTQ 'improves computational efficiency' is not fully supported unless it is shown to reach a given accuracy in position and energy with less cost than the alternatives. In particular, the ODE45 comparison in Table 4 uses a relative tolerance of 1e-10 while the outer Solar System comparisons use 1e-8, and the tables report only energy error; a fair accuracy-controlled comparison would strengthen the efficiency claim.
  5. [§3.3, last paragraph] The paper acknowledges that the quasi-Newton method can converge more slowly than Newton and suggests that smaller step sizes mitigate this, but it does not provide any convergence analysis, iteration counts, or evidence that the maximum iteration count kmax is never reached in the reported runs. Because the adaptive time step may produce arbitrarily small steps during close encounters, a detailed account of the solver's behavior in those regimes is needed to rule out silent solver failures.
minor comments (6)
  1. [§1, References] The reference 'Vogelaere 1900' appears to be a typo; the relevant work by de Vogelaere is from 1956. Please correct the citation and bibliography entry.
  2. [§4.2, first paragraph] The text contains the typo 'The The three-body problem.' Please fix.
  3. [§4.3, Table 5 discussion] The statement that SQQ-PTQ requires 'only 50% of the computational time' relative to SQQ-PN is imprecise: from Table 5 the ratio is 1.1e1 / 2.6e1 = 0.42. Please either say 'about 40%' or correct the numbers.
  4. [§3.3, Algorithm 1] The relative error definition uses ||x_k|| in the denominator, which can be problematic when components of x_k approach zero; consider a combined absolute/relative tolerance or a safeguard for near-zero norms.
  5. [§4.1] For the high-eccentricity Kepler experiments the parameters are listed as m = n = 3, but the figure captions and text do not specify the number of Gaussian quadrature points g; please state g and all related implementation parameters for reproducibility.
  6. [Figures 4 and 8] The phase-space-trajectory figures would be more informative if the axes were labeled with the specific variables plotted (e.g., q1 vs p1) and if a reference orbit from a high-accuracy solver were overlaid for comparison.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the derivation is self-contained and the numerical demonstrations are independent checks, not fitted predictions.

full rationale

The paper's derivation chain is self-contained and does not reduce any claimed result to its own inputs. The method parameters (a = 1e-6, b = 1e2, m, n, and delta-tau) are user-chosen constants, not fitted to the reported energy errors. The symplectic structure is inherited from the generating-function construction of SQQ, with Chebyshev interpolation and the time transformation introduced as algorithmic modifications, and Broyden's method used only as a nonlinear solver. Validation is performed against SQQ variants and the MATLAB built-in ODE45 integrator, which are independent comparison baselines rather than outputs of the method. No self-citation is load-bearing: the foundational SQQ integrator is cited to Gao et al. (2012), whose authorship does not overlap with the present paper, and the other cited works supply standard techniques (time transformation, Broyden's method, step-size control functions) rather than the paper's central claims. The strongest possible concern is that Algorithm 1 stops on the relative update norm rather than the residual, so the bounded energy error reported in the experiments does not by itself prove exact symplecticity; however, this is a correctness or verification gap, not a circular reduction of a conclusion to an input. Accordingly, the circularity score is 0.

Assumptions & free parameters 2 free parameters · 5 assumptions · 0 invented entities

The method introduces no new physical entity or fitted model constant. The parameters a and b in the step-size control function are hand-chosen hyperparameters. The core assumption is inherited: the discrete action (Chebyshev interpolation plus Gaussian quadrature) and the quasi-Newton solve define a sufficiently accurate symplectic map. No code or data is shipped to independently verify the numerical results.

free parameters (2)
  • a (lower bound in step-size control function) = 1e-6
    Chosen by hand in Eq. (20) to prevent excessively small time steps during close encounters; not fitted to data.
  • b (upper bound in step-size control function) = 1e2
    Chosen by hand in Eq. (20) to cap the time step; not fitted to data.
assumptions (5)
  • standard math Hamilton's canonical equations and the principle of least action underpin the generating-function construction (Eqs. 1-4).
    Used without proof in Section 2; standard mathematical background.
  • domain assumption Darboux-Sundman transformation K = sigma (H - H0) with fixed-step integration in tau yields a symplectic map that conserves the original Hamiltonian when K is conserved (Eq. 17).
    Invoked with citations to Nacozy 1977 and Hairer; not re-derived for the bounded sigma used in Eq. 20.
  • domain assumption Chebyshev interpolation plus Gaussian quadrature approximates the action S with negligible error for the chosen m, n, and time steps.
    Assumed in Section 3.1 and 3.2; convergence in m and n is not analyzed.
  • ad hoc to paper Broyden's method converges to the solution of Eq. 8 within tolerance epsilon = 1e-12, making the implicit solve exact enough to preserve symplecticity.
    Stated in Section 3.3 with no convergence guarantee for these nonlinear systems beyond observed iterations.
  • domain assumption The bounded step-size function in Eq. 20 is positive, smooth, and a valid time reparametrization for all trajectories considered.
    a and b are chosen as 1e-6 and 1e2; no proof that sigma avoids degenerate cases in close encounters.

how reviews work

0 comments
Cite this review

Pith. "Pith review of An adaptive symplectic integrator for gravitational dynamics." pith.science (2026). https://pith.science/paper/XWCDIEMI

@misc{pith2026250714881,
  author       = {Pith},
  title        = {Pith review of: An adaptive symplectic integrator for gravitational dynamics},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/XWCDIEMI}},
  note         = {Machine review of arXiv:2507.14881}
}
read the original abstract

This paper presents an adaptive symplectic integrator, SQQ-PTQ, developed on the basis of the fixed-step symplectic integrator SQQ. To mitigate the Runge phenomenon, SQQ-PTQ employs Chebyshev interpolation for approximating the action, enhancing both the precision and stability of the interpolation. In addition, to reduce the computational cost of evaluating interpolation functions, SQQ-PTQ introduces a projection method that improves the efficiency of these computations. A key feature of SQQ-PTQ is its use of the time transformation to implement an adaptive time step. To address the challenge of computing complicated Jacobian matrices attributed to the time transformation, SQQ-PTQ adopts a quasi-Newton method based on Broyden's method. This strategy accelerates the solution of nonlinear equations, thereby improving the overall computational performance. The effectiveness and robustness of SQQ-PTQ are demonstrated via three numerical experiments. In particular, SQQ-PTQ demonstrates adaptability in handling close-encounter problems. Moreover, during long-term integrations, SQQ-PTQ maintains the energy conservation, further confirming its advantages as a symplectic algorithm.

Figures

Figures reproduced from arXiv: 2507.14881 by the authors.

Figure 1
Figure 1. Illustration of the length coordinates for the time intervals [−1, 1] and [ta, tb]. Although the horizontal coordinates of ξˆ and ξ are not equal and their length coordinates λ1 and λ2 are equal, the values of the cor￾responding interpolation functions are equal (Eq. 12). where L = tb − ta. As illustrated in [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Evolution of the energy error for SQQ-P and SQQ. Here, e = 0.5, and m and n denote the number of interpolation points for the general￾ized displacement and generalized momentum, respectively [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Evolution of the energy error for SQQ-PTQ for the Kepler prob￾lem with high eccentricities. the latter requiring about 2 to 3 times of the CPU time at e = 0.9 and e = 0.99, respectively [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (6 more)
Figure 4
Figure 4. Figure 4: Phase space trajectories over 500 orbital periods for the Kepler problem with different eccentricities. periodic solutions for the three-body problem. In this section, we use one specific periodic configuration from their study to test SQQ-PTQ’s ability to automaticall…
Figure 5
Figure 5. Figure 5: Trajectories of particles in the three-body problem, with dashed boxes a and b indicating the locations of two close encounters. Box a corresponds to a close encounter occurring at approximately 0.4T, and box b corresponds to the second close encounter occurring at ap￾…
Figure 6
Figure 6. Figure 6: Evolution of the velocity magnitudes of the three particles over one period in the three-body problem. 0 500 1000 1500 2000 2500 3000 Time 10!16 10!15 10!14 10!13 10!12 10!11 10!10 10!9 10!8 10!7 10!6 Energy error [PITH_FULL_IMAGE:figures/full_fig_p007_6.png]
Figure 7
Figure 7. Figure 7: Evolution of the energy error for the integration of the three￾body problem by SQQ-PTQ. of the masses and orbital parameters of the planets, we refer to Hairer et al. (2006). We perform simulations over 100 Jupiter orbits using three integrators: SQQ-PN, SQQ-PTN, and S…
Figure 9
Figure 9. Figure 9: Evolutions of the energy error for the integration of the outer Solar System by SQQ-PN, SQQ-PTN, and SQQ-PTQ, with a total inte￾gration time of 100 Jupiter orbital periods. 0 0.5 1 1.5 2 2.5 3 3.5 4 Time (days) #10 7 10!12 10!11 10!10 10!9 10!8 10!7 Energy error SQQ-PN…
Figure 10
Figure 10. Figure 10: Evolutions of the energy error for the integration of the outer Solar System by SQQ-PN and SQQ-PTQ, with a total integration time of 10000 Jupiter orbital periods [PITH_FULL_IMAGE:figures/full_fig_p008_10.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

34 extracted references · 34 canonical work pages

  1. [1]

    & Iserles, A

    Blanes, S. & Iserles, A. 2012, Celest. Mech. Dyn. Astron., 114, 297

    work page 2012

  2. [2]

    Broyden, C. G. 1965, Math. Comput., 19, 577

    work page 1965

  3. [3]

    Broyden, C. G. 1970, Comput. J., 12, 393

    work page 1970

  4. [4]

    & Piggott, M

    Budd, C. & Piggott, M. 2003, Handbook of Numerical Analysis (North-Holland Publishing Co Amsterdam) Bulín, R. & Hajžman, M. 2021, Nonlinear Dyn., 103, 2475

    work page 2003

  5. [5]

    & Song, L

    Cai, Q. & Song, L. 2012, in 2012 IEEE International Conference on Information Science and Technology, IEEE, 745–748

    work page 2012

  6. [6]

    Channell, P. J. & Scovel, C. 1990, Nonlinearity, 3, 231

    work page 1990

  7. [7]

    Chen, J.-B., Guo, H.-Y ., & Wu, K. 2003, J. Math. Phys., 44, 1688

    work page 2003

  8. [8]

    2022, Appl

    Chen, Y ., Li, Y ., Huang, Y ., Li, M., & Liu, Y . 2022, Appl. Math. Model., 106, 742

    work page 2022

Show all 34 references
  1. [9]

    & Read, J

    Dehnen, W. & Read, J. I. 2011, Eur. Phys. J. Plus, 126, 1 Emel’yanenko, V . V . 2007, Celest. Mech. Dyn. Astron., 98, 191

    2011

  2. [10]

    1985, Proceedings of the 1984 Beijing Symposium on Differential Ge- ometry and Differential Equations: Computation of Partial Differential Equa- tions (Science Press)

    Feng, K. 1985, Proceedings of the 1984 Beijing Symposium on Differential Ge- ometry and Differential Equations: Computation of Partial Differential Equa- tions (Science Press)

    1985

  3. [11]

    2012, Int

    Gao, Q., Tan, S., Zhang, H., & Zhong, W. 2012, Int. J. Numer. Methods Eng., 89, 438

    2012

  4. [12]

    1997, Appl

    Hairer, E. 1997, Appl. Numer. Math., 25, 219

    1997

  5. [13]

    2006, Geometric Numerical Integration, 2nd edn

    Hairer, E., Wanner, G., & Lubich, C. 2006, Geometric Numerical Integration, 2nd edn. (Springer Berlin, Heidelberg)

    2006

  6. [14]

    & Leimkuhler, B

    Huang, W. & Leimkuhler, B. 1997, SIAM J. Sci. Comput., 18, 239

    1997

  7. [15]

    Lanczos, C. 1938, J. Math. Phys., 17, 123

    1938

  8. [16]

    2021, Nonlinear Dyn., 106, 2919

    Li, F., Peng, H., Yang, H., & Kan, Z. 2021, Nonlinear Dyn., 106, 2919

    2021

  9. [17]

    2022, New Astron., 96, 101850

    Liao, S., Li, X., & Yang, Y . 2022, New Astron., 96, 101850

    2022

  10. [18]

    & Murua, A

    Makazaga, J. & Murua, A. 2009, Numer. Math., 113, 631

    2009

  11. [19]

    & Pesch, H

    Miesbach, S. & Pesch, H. J. 1992, Numer. Math., 61, 501

    1992

  12. [20]

    & Tanikawa, K

    Mikkola, S. & Tanikawa, K. 1999, MNRAS, 310, 745

    1999

  13. [21]

    1977, Celest

    Nacozy, P. 1977, Celest. Mech., 16, 309

    1977

  14. [22]

    & Wright, S

    Nocedal, J. & Wright, S. J. 2006, Numerical optimization (Springer)

    2006

  15. [23]

    2019, IEEE Trans

    Peng, H., Li, F., Liu, J., & Ju, Z. 2019, IEEE Trans. Ind. Electron., 67, 3819

    2019

  16. [24]

    2023, Appl

    Peng, H., Shi, B., Song, J., & Wang, X. 2023, Appl. Math. Model., 114, 205

    2023

  17. [25]

    2020, Multibody Syst

    Peng, H., Song, N., & Kan, Z. 2020, Multibody Syst. Dyn., 49, 119

    2020

  18. [26]

    & Tremaine, S

    Preto, M. & Tremaine, S. 1999, Astron. J., 118, 2532

    1999

  19. [27]

    Ruth, R. D. 1983, IEEE Trans. Nucl. Sci., 30, 2669

    1983

  20. [28]

    Schnabel, R. B. & Frank, P. D. 1984, SIAM J. Numer. Anal., 21, 815

    1984

  21. [29]

    & Morrison, W

    Sherman, J. & Morrison, W. J. 1950, Ann. Math. Stat., 21, 124

    1950

  22. [30]

    2020, Nonlinear Dyn., 102, 1375

    Song, N., Peng, H., Kan, Z., & Chen, B. 2020, Nonlinear Dyn., 102, 1375

    2020

  23. [31]

    2005, MNRAS, 364, 1105 V ogelaere, R

    Springel, V . 2005, MNRAS, 364, 1105 V ogelaere, R. D. 1900, Methods of Integration which Preserve the Contact Trans- formation Property of the Hamilton Equations, Tech. rep., University of Notre Dame

    2005

  24. [32]

    & Nitadori, K

    Wang, L. & Nitadori, K. 2020, MNRAS, 497, 4384

    2020

  25. [33]

    2020, MNRAS, 493, 3398

    Wang, L., Nitadori, K., & Makino, J. 2020, MNRAS, 493, 3398

    2020

  26. [34]

    2021, AJ, 907, 66 Article number, page 8 of 8

    Wang, Y ., Sun, W., Liu, F., & Wu, X. 2021, AJ, 907, 66 Article number, page 8 of 8

    2021

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.