pith. machine review for the scientific record. sign in

arxiv: 2605.06735 · v1 · submitted 2026-05-07 · 🧮 math.NA · cs.NA

Recognition: no theorem link

Error estimation for numerical approximations of ODEs via composition techniques. Part II: BDF methods

Authors on Pith no claims yet

Pith reviewed 2026-05-11 01:39 UTC · model grok-4.3

classification 🧮 math.NA cs.NA
keywords BDF methodscomposition techniqueserror estimationODE solverslinear stabilityDahlquist barriercomplex coefficientsvariable step sizes
0
0 comments X

The pith

Composing BDF methods with complex coefficients raises approximation order by one and supplies an embedded error estimate of order p+1.

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

The paper extends a composition technique that uses complex coefficients to the family of implicit Backward Difference Formula (BDF) methods for integrating ordinary differential equations. Standard BDF schemes using p past points reach order p under suitable conditions; the composition produces a new scheme that attains order p+1 while still using only those p points. The real part of the resulting flow advances the solution, and the imaginary part furnishes a built-in error estimator whose leading term is one order higher. Linear stability analysis shows that the new schemes remain stable for orders up to eight, exceeding the classical Dahlquist limit on multistep methods. Numerical experiments confirm both the order increase and improved efficiency relative to ordinary BDF schemes using the same number of steps.

Core claim

The composition of a BDF method with complex coefficients yields a numerical flow whose real component approximates the solution to order p+1 with only p backward points, while its imaginary component estimates the local truncation error to the same order. This construction breaks the Dahlquist barrier and produces stable schemes up to order eight. On non-uniform meshes the characteristic equation for the composed method contains the step-size ratio as a parameter; the presence of a complex root with positive real part supplies an explicit lower bound on admissible ratios (0.4506 for order three, 0.6806 for order four).

What carries the argument

The composed flow formed by applying a complex-coefficient composition operator to the BDF linear multistep formula, with its real part serving as the integrator and imaginary part serving as the error estimator.

If this is right

  • Higher-order integration is obtained without increasing the number of stored solution values or function evaluations per step.
  • Stable integration becomes possible for orders up to eight, whereas classical explicit multistep methods are limited to order two.
  • For variable-step meshes, explicit lower bounds on consecutive step ratios guarantee linear stability for each order.
  • The built-in error estimator can be used for adaptive step-size selection without an auxiliary computation.
  • Overall CPU time decreases for a target accuracy because the same number of past points now delivers one extra order.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same composition operator could be tested on other implicit linear multistep families to check whether the order gain and embedded estimator appear there as well.
  • The stability bounds derived from root locations suggest concrete rules for choosing step ratios when adapting meshes in stiff or long-time problems.
  • Because the imaginary part is available at negligible extra cost, the method may reduce the overhead of separate error-control procedures in existing ODE libraries.

Load-bearing premise

The complex-coefficient composition can be applied directly to implicit BDF schemes while preserving the base order p and without creating new order reductions or instabilities not captured by the linear analysis.

What would settle it

A convergence test on the Dahlquist equation y' = λy (Re λ < 0) in which the observed order of the real part of the composed BDF solution remains exactly p rather than reaching p+1, or in which the imaginary part fails to track the true local error at order p+1.

read the original abstract

Integration of Ordinary Differential Equations (ODEs) using Backward Difference formula (BDF) methods with p backward steps achieves order p accuracy if specific conditions are met. This work extends the composition technique with complex coefficients to the implicit BDF schemes, increasing the approximation order by one without additional backward points. The imaginary part of the composed flow provides an error estimate of order p + 1. Linear stability analysis reveals that the composed schemes break the Dahlquist barrier, achieving stability up to order eight. The computational performance of the composed flow outperforms BDF schemes when using the same number of backward points, allowing for higher accuracy with lower CPU time. For non-uniform meshes, the ratio of consecutive time steps, which influences stability, appears as a parameter in the roots of algebraic equations relative to the composed flow. Having a complex root with a real positive part implies a lower bound to this ratio depending on the order. For example, the bound is 0.4506 for order three and 0.6806 for order four. Numerical tests demonstrate the effectiveness of this technique in improving the accuracy and stability compared to BDF methods.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

0 major / 3 minor

Summary. The manuscript extends composition techniques with complex coefficients to implicit BDF methods for ODEs. It claims that the approach raises the approximation order by one without extra backward points, that the imaginary part of the composed flow supplies an O(h^{p+1}) error estimator, that the resulting schemes are stable up to order 8 (breaking the Dahlquist barrier), and that they outperform standard BDF in accuracy and CPU time for the same number of points. For variable-step meshes the ratio of consecutive steps appears in the characteristic polynomial; the existence of a complex root with positive real part yields explicit lower bounds on that ratio (e.g., 0.4506 for order 3). Numerical experiments are reported to confirm the gains in accuracy and stability.

Significance. If the derivations and stability calculations are correct, the work supplies a practical route to higher-order, self-estimating BDF-type integrators that remain competitive on stiff problems and on non-uniform meshes. The explicit step-ratio bounds and the reported CPU-time advantage are directly usable in existing BDF codes.

minor comments (3)
  1. The linear-stability section should tabulate the explicit complex coefficients and the resulting stability polynomials for orders 3–8 so that the root-locus claims can be reproduced without re-deriving the composition.
  2. In the non-uniform-mesh analysis, the precise algebraic equation whose roots determine the step-ratio bound should be written out (rather than only the numerical bound values) to allow independent verification.
  3. The numerical-experiments section would benefit from a direct comparison of the embedded error estimator against a standard embedded BDF pair of the same effective order, including measured effectivity indices.

Simulated Author's Rebuttal

0 responses · 0 unresolved

We thank the referee for the positive and accurate summary of our manuscript, as well as the recommendation for minor revision. The referee's assessment correctly identifies the key contributions regarding order elevation, error estimation via the imaginary part, stability up to order 8, and step-ratio bounds for variable meshes.

Circularity Check

0 steps flagged

No significant circularity detected

full rationale

The derivation proceeds from the composition of BDF steps with complex coefficients, yielding an order-p+1 error estimator directly from the imaginary part of the resulting flow and stability regions from root-locus analysis of the characteristic polynomial. These steps rely on algebraic construction and root-finding rather than parameter fitting, self-referential definitions, or load-bearing self-citations; the step-ratio bounds for non-uniform meshes are likewise obtained by solving the same algebraic equations for the presence of roots with positive real part. The central claims therefore remain independent of the quantities they predict.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claims rest on standard multistep method theory and the assumption that the composition technique from prior work transfers directly to implicit BDF without loss of properties.

axioms (1)
  • domain assumption BDF methods achieve order p accuracy if specific conditions are met
    Explicitly stated in the abstract as the prerequisite for the order increase.

pith-pipeline@v0.9.0 · 5507 in / 1311 out tokens · 66828 ms · 2026-05-11T01:39:13.281346+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

121 extracted references · 121 canonical work pages

  1. [1]

    doi:https://doi.org/10.48550/arXiv.2409.10548 , month =

    Deeb, Ahmad and Dutykh, Denys , title =. doi:https://doi.org/10.48550/arXiv.2409.10548 , month =

  2. [2]

    Mathematical Methods in the Applied Sciences , doi =

    Laadhari, Aymen and Deeb, Ahmad and Kawi, Badr , title =. Mathematical Methods in the Applied Sciences , doi =

  3. [3]

    Symmetry , volume=

    Laadhari, Aymen and Deeb, Ahmad , title =. Symmetry , volume=. 2023 , doi =

  4. [4]

    Mathematics and Computers in Simulation , volume =

    Proper. Mathematics and Computers in Simulation , volume =. 2023 , issn =. doi:https://doi.org/10.1016/j.matcom.2023.01.008 , author =

  5. [5]

    Advanced Modeling and Simulation in Engineering Sciences , year=

    Razafindralandy, Dina and Salnikov, Vladimir and Hamdouni, Aziz and Deeb, Ahmad , title=. Advanced Modeling and Simulation in Engineering Sciences , year=. doi:https://doi.org/10.1186/s40323-019-0130-2 , abstract=

  6. [6]

    Discrete and Continuous Dynamical Systems - Series S , volume =

    Comparison between. Discrete and Continuous Dynamical Systems - Series S , volume =. 2016 , month=. doi:https://doi.org/10.3934/dcdss.2016003 , author =

  7. [7]

    Performance of

    Ahmad Deeb and Aziz Hamdouni and Dina Razafindralandy , keywords =. Performance of. Applied Mathematics and Computation , volume =. 2022 , doi =

  8. [8]

    ESAIM: Procedings and Surveys , year =

    Borel-. ESAIM: Procedings and Surveys , year =. doi:https://doi.org/10.1051/proc/201445033 , author =

  9. [9]

    11th International Conference on Mathematical Problems in Engineering, Aerospace and Sciences , eventdate =

    Razafindralandy, Dina and Hamdouni, Aziz and Deeb, Ahmad , title =. 11th International Conference on Mathematical Problems in Engineering, Aerospace and Sciences , eventdate =. 2017 , Volume =. doi:https://doi.org/10.1063/1.4972721 , publisher =

  10. [10]

    Soft Computing , volume=

    Ahmad Deeb and Rafik Belarbi , title =. Soft Computing , volume=. 2022 , pages =

  11. [11]

    Mathematics and Computers in Simulation , doi =

    Deeb, Ahmad and Dutykh, Denys , title =. Mathematics and Computers in Simulation , doi =. 2025 , month=

  12. [12]

    A family of embedded. J. Comp. App. Math. , volume =. 1980 , note =. doi:http://dx.doi.org/10.1016/0771-050X(80)90013-3 , author =

  13. [13]

    A 3(2) pair of. App. Math. Lett. , volume =. 1989 , note =

  14. [14]

    2008 , address =

    Numerical Methods for Ordinary Differential Equations , author =. 2008 , address =

  15. [15]

    2009 , series =

    Solving Ordinary Differential Equations I: Nonstiff Problems , author =. 2009 , series =

  16. [16]

    2013 , series =

    Methods of Applied Mathematics for Engineers Scientists , author =. 2013 , series =

  17. [17]

    1996 , address =

    Solving Ordinary Differential Equations II: Stiff and Differential-Algebraic Problems , author =. 1996 , address =

  18. [18]

    A First Course in the Numerical Analysis of Differential Equations , publisher=

    Iserles, Arieh , year=. A First Course in the Numerical Analysis of Differential Equations , publisher=

  19. [19]

    Curtiss, C. F. and Hirschfelder, J. O. , journal =. Integration of Stiff Equations , volume =

  20. [20]

    Ueber die numerische. Math. Ann. , pages =

  21. [21]

    Butcher, John C. , year=. On. Journal of the Australian Mathematical Society , publisher=. doi:10.1017/S1446788700023387 , number=

  22. [22]

    Butcher, John C. , year=. Coefficients for the study of. Journal of Australian Mathematical Society , publisher=

  23. [23]

    , abstract =

    Butcher, John C. , abstract =. A history of. App. Num. Math. , volume =. 1996 , issn =

  24. [24]

    , abstract =

    Verwer, J.G. , abstract =. Explicit. App. Num. Math. , volume =. 1996 , issn =

  25. [25]

    1999 , author =

    Order conditions for numerical integrators obtained by composing simpler integrators , journal =. 1999 , author =. doi:https://doi.org/10.1098/rsta.1999.0365 , publisher =

  26. [26]

    2002 , publisher=

    Geometric Numerical Integration: Structure-Preserving algorithms for Ordinary Differential Equations , author=. 2002 , publisher=

  27. [27]

    Compositions of pseudo-symmetric integrators with complex coefficients for the numerical integration of differential equations , journal =

    Casas, Fernando and Chartier, Philippe and Escorihuela-Tomàs, Alejandro and Zhang, Yong , keywords =. Compositions of pseudo-symmetric integrators with complex coefficients for the numerical integration of differential equations , journal =. 2021 , issn =

  28. [28]

    Adaptive time-stepping and computational stability , journal =

    S\". Adaptive time-stepping and computational stability , journal =. 2006 , issn =

  29. [29]

    2009 , author =

    Splitting methods with complex times for parabolic equations , journal =. 2009 , author =

  30. [30]

    , title =

    LeVeque, Randall J. , title =. SIAM Journal of Numerical Analysis , volume =

  31. [31]

    A novel adaptive. Appl. Math. Lett. , volume =. 2022 , issn =. doi:10.1016/j.aml.2022.107943 , author =

  32. [32]

    , editor=

    Butcher, John C. , editor=. The effective order of. Conference on the Numerical Solution of Differential Equations , year=

  33. [33]

    , title =

    McLachlan, Robert I. , title =. SIAM Journal on Scientific Computing , volume =. 1995 , doi =

  34. [34]

    1990 , issn =

    Construction of higher order symplectic integrators , journal =. 1990 , issn =. doi:10.1016/0375-9601(90)90092-3 , author =

  35. [35]

    Physics Letters A , volume =

    Fractal decomposition of exponential operators with applications to many-body theories and. Physics Letters A , volume =. 1990 , doi =

  36. [36]

    Iserles, Arieh , title =. Numer. Math. , volume =. 1984 , doi =

  37. [37]

    Splitting and composition methods in the numerical integration of differential equations , author =. Bol. Scc. Esp. Mat. Apl. , pages=

  38. [38]

    SIAM Journal on Scientific Computing , volume =

    Blanes, Sergio and Casas, Fernando and Murua, Ander , title =. SIAM Journal on Scientific Computing , volume =. 2006 , doi =

  39. [39]

    2016 , address =

    Blanes, Sergio and Casas, Fernando , title =. 2016 , address =

  40. [40]

    Geometric numerical integration illustrated by the

    Hairer, Ernst and Lubich, Christian and Wanner, Gerhard , year=. Geometric numerical integration illustrated by the. doi:10.1017/S0962492902000144 , journal=

  41. [41]

    , journal=

    Cromer, Alan H. , journal=. Stable solutions using the. 1981 , volume=

  42. [42]

    Simulating

    Leimkuhler, Benedict and Reich, Sebastian , year=. Simulating. doi:10.1017/CBO9780511614118 , publisher=

  43. [43]

    Computational Science and Discovery , year=

    Hybrid symplectic integrators for relativistic particles in electric and magnetic fields , author=. Computational Science and Discovery , year=

  44. [44]

    and Skokos, Ch

    Senyange, B. and Skokos, Ch. , journal=. Computational efficiency of symplectic integration schemes: application to multidimensional disordered. 2018 , volume=

  45. [45]

    2013 , doi =

    New families of symplectic splitting methods for numerical integration in dynamical astronomy , journal =. 2013 , doi =

  46. [46]

    Journal of Computational and Applied Mathematics , volume =

    Novel symplectic integrators for the. Journal of Computational and Applied Mathematics , volume =. 2019 , doi =

  47. [47]

    Butusov, D. N. and Andreev, V. S. and Pesterev, D. O. , booktitle=. Composition semi-implicit methods for chaotic problems simulation , year=

  48. [48]

    International Journal of Non-Linear Mechanics , volume =

    Phase portraits and bifurcations of the non-linear oscillator:. International Journal of Non-Linear Mechanics , volume =. 1980 , issn =. doi:10.1016/0020-7462(80)90031-1 , author =

  49. [49]

    and Cho, Hancheol , title =

    Udwadia, Firdaus E. and Cho, Hancheol , title =. Journal of Applied Mechanics , volume =. 2013 , month =. doi:10.1115/1.4024673 , note =

  50. [50]

    1969 , month =

    Hansen, Eldon , title =. 1969 , month =

  51. [51]

    BIT Numerical Mathematics , volume =

    Bickart, Theodore and Picel, Zdenek , title =. BIT Numerical Mathematics , volume =. 1973 , pages =

  52. [52]

    SIAM Journal on Numerical Analysis , volume =

    Donelson III, John and Hansen, Eldon , title =. SIAM Journal on Numerical Analysis , volume =. 1971 , doi =

  53. [53]

    ACM Transactions on Mathematical Software , volume =

    Tendler, Joel and Bickart, Theodore and Picel, Zdenek , title =. ACM Transactions on Mathematical Software , volume =. 1977 , doi =

  54. [54]

    , title =

    Cash, J.R. , title =. BIT Numerical Mathematics , volume =. 1977 , pages =

  55. [55]

    , title =

    Becker, J. , title =. BIT Numerical Mathematics , volume =. 1998 , pages =

  56. [56]

    Numerische Mathematik , volume =

    Bokanowski, Olivier and Picarelli, Athena and Reisinger, Cristoph , title =. Numerische Mathematik , volume =. 2021 , pages =

  57. [57]

    Advances in Computational Mathematics , volume =

    Wang, Wansheng and Mao, Mengli and Wang, Zheng , title =. Advances in Computational Mathematics , volume =. 2021 , pages =

  58. [58]

    Journal of Mathematical Chemistry , volume =

    Khalsaraei, Mohammad Mehdizadeh and Shokri, Ali and Molayi, Maryam , title =. Journal of Mathematical Chemistry , volume =. 2020 , pages =

  59. [59]

    Journal of Mathematical Chemistry , volume =

    Soomro, Hira and Zainuddin, Nooraini and Daud, Hanita and Sunday, Joshua and Jamaludin, Noraini and Abdullah, Abdullah and Mulono, Apriyanto and Abdul Kadir, Evizal , title =. Journal of Mathematical Chemistry , volume =. 2023 , pages =

  60. [60]

    Mathematica Scandinavica , volume =

    Dahlquist, Germund , title =. Mathematica Scandinavica , volume =. 1956 , pages =

  61. [61]

    Acta Universitatis Apulensis , volume =

    Shokri, Ali and Shokri, Abbasali , title =. Acta Universitatis Apulensis , volume =. 2014 , pages =

  62. [62]

    Journal of ACM , volume =

    Gragg, Williams and Stetter, Hans , title =. Journal of ACM , volume =. 1964 , pages =

  63. [63]

    2001 , doi =

    High order numerical integrators for differential equations using composition and processing of low order methods , journal =. 2001 , doi =

  64. [64]

    2019 , doi =

    Splitting and composition methods with embedded error estimators , journal =. 2019 , doi =

  65. [65]

    2010 , doi =

    Splitting methods with complex coefficients , journal =. 2010 , doi =

  66. [66]

    2005 , doi =

    On the necessity of negative coefficients for operator splitting schemes of order higher than two , journal =. 2005 , doi =

  67. [67]

    1963 , doi =

    A special stability problem for linear multistep methods , journal =. 1963 , doi =

  68. [68]

    2009 , doi =

    General linear methods for ordinary differential equations , journal =. 2009 , doi =

  69. [69]

    Numerical Analysis and Applications , volume =

    Okounghae, R I and Ikhile, M N O , title =. Numerical Analysis and Applications , volume =

  70. [70]

    2004 , doi =

    Linear multistep methods applied to stiff initial value problems--A survey , journal =. 2004 , doi =

  71. [71]

    2014 , doi =

    Two classes of implicit–explicit multistep methods for nonlinear stiff initial-value problems , journal =. 2014 , doi =

  72. [72]

    2013 , doi =

    Implicit--explicit multistep methods for nonlinear parabolic equations , journal =. 2013 , doi =

  73. [73]

    SIAM Journal on Numerical Analysis , volume =

    Akrivis, Georgios and Katsoprinakis, Emmanouil , title =. SIAM Journal on Numerical Analysis , volume =. 2020 , doi =

  74. [74]

    , editor=

    Jay, Laurent O. , editor=. Lobatto Methods , bookTitle=. 2015 , publisher=

  75. [75]

    Journal of Computational and Applied Mathematics , volume =

    Stiff differential equations solved by. Journal of Computational and Applied Mathematics , volume =. 1999 , issn =. doi:https://doi.org/10.1016/S0377-0427(99)00134-X , author =

  76. [76]

    1880 , issn =

    Étude sur les formules d'approximation qui servent à calculer la valeur numérique d'une intégrale définie , journal =. 1880 , issn =

  77. [77]

    Journal of Computational Physics , volume =

    An embedded formula of the. Journal of Computational Physics , volume =. 2017 , issn =. doi:https://doi.org/10.1016/j.jcp.2017.09.046 , author =

  78. [78]

    SIAM Journal on Numerical Analysis , volume =

    Johnson, Claes , title =. SIAM Journal on Numerical Analysis , volume =. 1988 , doi =

  79. [79]

    BIT Numerical Mathematics , volume =

    Axelsson, O , title =. BIT Numerical Mathematics , volume =. 1969 , doi =

  80. [80]

    , title =

    Blanes, Sergio and Casas, Fernando and Ros, J. , title =. SIAM Journal on Scientific Computing , volume =. 1999 , doi =

Showing first 80 references.