Pith. sign in

REVIEW 3 major objections 5 minor 10 references

A thrust to trust minimum thrust

T0 review · 3 major / 5 minor · reviewed 2026-08-03 · deepseek-v4-flash

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

Pith's one-line read For five particles, the minimum possible thrust now has an exact value, derived from a degree-10 polynomial.

desk verdict A careful, honest numerical study of minimum thrust with a genuinely new exact-looking N=5 candidate — but global optimality is asserted, not proven, so treat the headline result as a strong conjecture. read the letter →

arxiv 2512.04244 v2 pith:JF5FEUWK submitted 2025-12-03 hep-ph

Matteo Cacciari This is my paper
classification hep-ph
keywords thrusteventshapesminimumN=5exactresultglobaloptimizationextremevaluetheoryWeibulldistributionparticlemomenta
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

Thrust measures how collimated a set of particle momenta is; this paper asks how small thrust can be for a fixed number N of particles, equivalently how large τ=1−T can get. For N=5 in three dimensions it obtains, for the first time, an exact value, expressed through the largest root of a degree-10 polynomial. For N up to 20 it gives numerical maxima, and for the hardest cases it uses extreme-value statistics to put credible bounds on the true maxima. The results improve on earlier numerical estimates and show the maximizing configurations are generally less symmetric than one might expect.

What carries the argument

The paper's main computational device is the exact sign-pattern algorithm for thrust: the continuous maximization over thrust directions is replaced by a finite search over 2^(N−1) − 1 sign patterns, so each thrust evaluation is exact and fast. Around this sits a global-optimization layer using stochastic evolutionary optimizers with local polishing to perform the outer minimization over momenta, and an extreme-value-theory layer that models the distribution of observed maxima with a Weibull distribution to estimate the unobserved true endpoint. For N=5 the exact result is obtained by reverse-engineering the numerical configuration and proving it via the sign-pattern algorithm.

What would settle it

For N=5, evaluate the polynomial in Eq. (19) to high precision and confirm the largest root reproduces Eq. (20); any mismatch invalidates the exact result. For N=9, run an independent, much longer stochastic global optimization search and see whether any configuration yields τ larger than the quoted maximum 0.4834291842117; if so, the claim that this is likely the true maximum is falsified, though the statistical bounds may already cover it.

Watch

Extended reading notes

Core claim

The central discovery is that for five particles in three dimensions the minimum thrust is exactly T_min = 2 / (1 + sqrt(1 + R_+)), where R_+ is the largest real root of an explicit degree-10 polynomial, giving τ_max = 0.45399486580028350678876698759830... . This is the first exact result beyond N=4. For N=6, 7, and 8, the stochastic optimizers repeatedly returned the same maximum, so the paper treats these as very likely true values. For N=9 to 20, where repeated hits become rare, it fits a Weibull extreme-value distribution to the observed maxima and reports endpoint estimates and credible intervals, explicitly noting the i.i.d. assumption may not be fully met because deep local minima can

Load-bearing premise

The statistical claims for N≥9 rest on the assumption that the maxima collected from stochastic runs are independent, identically distributed samples from a Weibull distribution with a finite endpoint; the paper explicitly concedes this may not hold because deep local minima can skew the distribution, and this violation is not quantified.

Editorial extensions

If this is right

  • The exact N=5 value can be used as a benchmark for future event-shape calculations and Monte Carlo event generators.
  • The new numerical maxima for N up to 20 supersede earlier estimates, so any physical analysis relying on minimum-thrust configurations should use these values.
  • For N≥13 the paper reports credible intervals rather than definitive values, giving a principled upper range for the true maxima when the optimizer cannot settle them.
  • The d-dimensional formula for the regular simplex provides a benchmark for N=d+1 in any dimension.
  • The observation that minimal-thrust configurations are less isotropic than expected could affect understanding of event-shape tails at colliders.

Reading between the lines

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

  • If the extreme-value-theory assumption is violated by deep local minima, the credible intervals may be too tight; the paper's own tests against known maxima suggest the method errs conservatively, but only for the cases tested.
  • The exact N=5 result might inspire closed forms for N=6–8 if their configurations are recognized as having algebraic descriptions; the 2D results already show composite constructions from lower-N configurations.
  • The sign-pattern algorithm's cost is exponential in N, so for larger N the paper's iterative method is faster but not exact; future work could combine branch-and-bound or mixed-integer methods to recover exactness for moderate N.
  • The reported values can be used to test the reliability of generic global optimizers on non-convex event-shape problems.
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

3 major / 5 minor

Summary. The paper determines the minimum of the thrust event shape (equivalently, the maximum of τ=1−T) for finite N-particle momentum configurations with zero total momentum. In three dimensions, it reports an exact algebraic value for N=5, τ_max^5 = 1 − T_min^5 = 0.45399486580028350678876698759830..., expressed in Appendix B through the largest root of a degree-10 polynomial, and quotes improved numerical estimates for N=6–20 obtained with stochastic global optimizers (DE, CMA-ES), sometimes supplemented by Extreme Value Theory (EVT) MLE and Bayesian endpoint analyses. Analytic results are also given for planar (2D) configurations up to N=12 (with N=8 left as an exercise) and for d+1 particles in d dimensions assuming regular-simplex optimality. The main claims are: (i) exact N=5 result in 3D; (ii) numerical values for N=6–8 that are very likely exact; (iii) credible bounds for N=9–20; (iv) 2D exact results.

Significance. If the N=5 exact value is indeed the global minimum, it is a welcome first exact result beyond N=4 for a classic event-shape optimization, a result that would sharpen the earlier numerical work of Aglietti et al. and provide a nontrivial benchmark for optimization algorithms. The paper’s numerical estimates for N≥6 supersede those of ref. [3], and the EVT cross-checks in Appendix C are a praiseworthy attempt to quantify uncertainty. The 2D analytic results and the d-dimensional simplex formula are useful ancillary contributions. However, the headline N=5 exact result is presented without a proof of global optimality; as written, the exact expression is for a candidate configuration whose global optimality rests on numerical evidence. This gap is load-bearing for the central claim, though plausibly fixable within the manuscript’s scope.

major comments (3)
  1. [Appendix B / Eq. (20)] The central claim—an exact minimum thrust for N=5—is not proven. Appendix B provides an exact value T_min^5 for a specific configuration whose shape was inferred from numerical optimization (‘Hints taken from the configuration obtained via the numerical optimisation’), but the text gives no theorem, inequality, or stationarity analysis showing that this configuration is the global minimizer over all momentum-conserving N=5 configurations. Since Eq. (2) is non-convex, reproducible convergence of stochastic minimizers does not establish global optimality. The abstract and Section 1 call this an ‘exact result’ and ‘determine for the first time an exact result’; as written this is an exact algebraic value for a candidate configuration, i.e., a strong conjecture supported by numerics. A proof, or at minimum a decisive change of claim (e.g., ‘exact candidate value, conjectured to be the true m
  2. [Section 3 / Table 1, rows N=9–20] The EVT-based endpoint estimates and credible bounds for N≥9 assume that the observed maxima from stochastic runs are i.i.d. samples from a Weibull-type distribution with finite endpoint. Footnote 4 concedes these conditions ‘may not be fully satisfied’ because deep local minima can skew the distribution, yet the manuscript proceeds to quote MLE endpoints and 95% bounds as though this caveat did not affect them. Appendix C tests the procedure on known cases and shows that the true maximum falls within the 95% or 99.7% bound, which is reassuring; however, the test cases (N=3,4 in 3D; N=13,15 in 2D) are not demonstrated to share the same failure modes as the hard 3D cases N=9–12. The impact of the violated i.i.d./Weibull assumption is not quantified. I would like to see either a justification that the deep-minimum skew is negligible for the quoted rows, or a propagation of this uncertainty
  3. [Section 3, Eq. (10) and Table 1 footnotes] The Bayesian prior in Eq. (10) sets the endpoint prior x_max as a half-Normal with mean equal to the observed maximum τ_max. This makes the observed maximum a lower bound for the endpoint by construction, so the posterior upper bound is necessarily ≥ τ_max. This is logically sound as a prior, but it means the credible intervals are not independent evidence for the location of the true maximum; they are conditional on the observed maximum being close to the truth. The text in footnote 6 says this prior ‘appears entirely justified’, but the circularity should be acknowledged more prominently when interpreting Table 1’s 95% upper bounds as evidence that the true value is close to the observed value. The paper does note in footnote 7 that for N=9–12 the statistical analyses point to a larger true maximum, so the caveat is partly present; I recommend making it explicit in the main text near T
minor comments (5)
  1. [Appendix D / Table 3] The N=8 2D row has no analytic value, with the calculation ‘left as an exercise for the reader.’ A paper reporting exact results would be stronger if this were supplied or deferred to future work; as is, the table has a gap that is unusual.
  2. [Appendix E / Eq. (22)] Eq. (22) states the d-dimensional regular-simplex result under an explicit assumption (‘Assuming that … the configuration … is the regular d-simplex’). The wording in the main text later calls it a result ‘one finds’, which slightly overstates the status. Please mark it as conjectural/conditional.
  3. [General] Several typos exist: ‘satisfied’ spelled ‘satistfied’ in footnote 4; ‘valuable’ for ‘value’ in Fig. 2 caption; ‘the maximum’, likely meant ‘the maximum value’ in the last paragraph of §3. The prose would benefit from proofreading.
  4. [Section 2.2] The hyperparameter choices for DE and CMA-ES (popsize, tolerance, mutation, restarts) are stated, but there is no convergence study demonstrating that the final results are stable with respect to these choices. Given the load-bearing nature of the numerical results, a small stability test (e.g., varying popsize or restarts) would strengthen Table 1.
  5. [References] Ref. [14] is a Star Wars fan-site reference; while the author’s comparison is amusing, a journal would likely prefer a more formal reference or none at all.

Circularity Check

1 steps flagged · score 6.0 of 10

Headline N=5 'exact' result is reverse-engineered from the numerical optimum; rest of paper has independent content.

  1. fitted input called prediction [Appendix B, Eq. (20); footnote 1]
    "An exact expression for the coordinates of the configuration with five particles giving the minimum thrust, and therefore for the value of minimum thrust itself, can be found using the algorithm given in Appendix A. Hints taken from the configuration obtained via the numerical optimisation can be used to guide the derivation. ... in perfect agreement with the one found numerically and given in Table 1."

    The N=5 exact configuration was obtained with 'hints taken from the configuration obtained via the numerical optimisation,' i.e. the candidate was read off from the same stochastic minimization that yields the Table 1 value. The Appendix A algorithm is exact only for a given set of momenta; it does not prove that this set is the global thrust minimizer (the objective in Eq. (2) is non-convex, Sec. 2). Eq. (20) is therefore an algebraic expression constructed to match the numerical optimum, and its agreement with Table 1 is by construction rather than an independent confirmation. Labeling this a first-time exact 'minimum thrust' result makes the headline claim partly circular, since the input to the derivation is the numerical value/output it is said to validate.

full rationale

The sole circular element is the N=5 'exact' result. Appendix B states that the exact expression can be found using the exact thrust algorithm after taking 'hints' from the numerically optimized configuration, and then notes that Eq. (20) is in perfect agreement with the value found numerically. Since the candidate configuration was itself obtained from the same stochastic optimization whose result it is used to confirm, the exact value is a post-hoc algebraic parametrization of the numerical optimum rather than an independent derivation of global minimality. No global-optimality proof is supplied; the thrust objective is non-convex (Sec. 2), so the algorithm in Appendix A only evaluates thrust for a given configuration. This is a circular confirmation loop for the headline claim. The rest of the paper has substantial independent content: exact checks against known N=3,4 and 2D values, EVT tests against exactly known maxima (Table 2), and new numerical estimates for N=6-20. The Bayesian prior (Eq. 10) is transparently data-informed and is tested in Appendix C, so it is not a circularity. No load-bearing self-citations appear. Overall, the central N=5 claim is partially circular, but the paper is not wholly reducible to its inputs.

Assumptions & free parameters 2 free parameters · 4 assumptions · 0 invented entities

No new physical entities are introduced. The free parameters are computational/statistical hyperparameters that affect the reported numerical and Bayesian results. The axioms are the EVT modeling assumption (explicitly flagged), the regular-simplex assumption for dD, the subjective Bayesian priors, and the implicit trust in stochastic optimizer convergence.

free parameters (2)
  • Prior width factor for endpoint (10^-1 in Eq. 10) = 0.1
    The width of the half-Normal prior on the endpoint x_max is set to (τ̄-τ_max)^2 / σ_τ × 10^{-1}, chosen by hand to mimic a width of 0.1 on the Weibull shape ξ. It affects the Bayesian credible intervals for N≥9.
  • DE/CMA-ES hyperparameters (popsize, tolerance, mutation, restarts, etc.) = popsize 100/200, tolerance 1e-14 to 1e-6, restarts 6, etc.
    Chosen by hand after manual tests; the paper makes no claim of optimality. These affect convergence and the observed maxima distribution, hence the statistical results.
assumptions (4)
  • domain assumption The distribution of observed maxima follows a generalized extreme value (Weibull) distribution with finite endpoint.
    Invoked in §3; footnote 4 admits the i.i.d. condition may not be fully satisfied due to deep local minima, but this is ignored.
  • ad hoc to paper For N=d+1 particles in d dimensions, the regular d-simplex gives the minimum thrust.
    Stated as an assumption in Appendix E to derive Eq. (22); only verified for d=2,3.
  • domain assumption The Bayesian priors in Eqs. (8)-(10) are reasonable representations of prior knowledge.
    Subjective Bayesian analysis; the paper checks stability, but the prior on the endpoint is informed by the observed maximum, which is a data-dependent choice.
  • domain assumption The numerical optimizers (DE, CMA-ES) sample the configuration space well enough that repeated convergence to the same maximum indicates global optimality.
    Used throughout; for N=5, all 100 runs converge to the same value, but no mathematical guarantee is provided.

how reviews work

0 comments
Cite this review

Pith. "Pith review of A thrust to trust minimum thrust." pith.science (2026). https://pith.science/paper/JF5FEUWK

@misc{pith2026251204244,
  author       = {Pith},
  title        = {Pith review of: A thrust to trust minimum thrust},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/JF5FEUWK}},
  note         = {Machine review of arXiv:2512.04244}
}
read the original abstract

We determine the minimum value of thrust for a number of N-particle configurations. For N=5 in three dimensions an exact result is found for the first time. For larger N we obtain numerical results through optimisation. When a definite value cannot be reliably identified, the results are analysed in the context of Extreme Value Theory, using a Maximum Likelihood Estimate and a Bayesian analysis. Results are given for three spatial dimensions, two dimensions, and selected cases in d dimensions.

Figures

Figures reproduced from arXiv: 2512.04244 by the authors.

Figure 1
Figure 1. Observed and posterior predictive distribution for the maxima (top row), and prior and posterior [PITH_FULL_IMAGE:figures/full_fig_p006_1.png] view at source ↗
Figure 2
Figure 2. Configuration of minimum thrust for five particles. The second and third plot show a front [PITH_FULL_IMAGE:figures/full_fig_p008_2.png] view at source ↗

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

10 extracted references · 4 linked inside Pith

  1. [3]

    U. G. Aglietti, G. Ferrera, W. L. Ju and J. Miao, Phys. Rev. Lett.134(2025) no.25, 251904 doi:10.1103/dv7n-qvyp [arXiv:2502.01570 [hep-ph]]

    arXiv 2025

  2. [1]

    Brandt, C

    S. Brandt, C. Peyrou, R. Sosnowski and A. Wroblewski, Phys. Lett.12(1964), 57-61 doi:10.1016/0031- 9163(64)91176-X

    doi:10.1016/0031- 1964

  3. [2]

    Farhi, Phys

    E. Farhi, Phys. Rev. Lett.39(1977), 1587-1588 doi:10.1103/PhysRevLett.39.1587

    doi:10.1103/physrevlett.39.1587 1977

  4. [4]

    P. F. Monni, T. Gehrmann and G. Luisoni, JHEP08(2011), 010 doi:10.1007/JHEP08(2011)010 [arXiv:1105.4560 [hep-ph]]

    arXiv 2011

  5. [5]

    Weinzierl, JHEP06(2009), 041 doi:10.1088/1126-6708/2009/06/041 [arXiv:0904.1077 [hep-ph]]

    S. Weinzierl, JHEP06(2009), 041 doi:10.1088/1126-6708/2009/06/041 [arXiv:0904.1077 [hep-ph]]

    arXiv 2009

  6. [6]

    Banfi, G

    A. Banfi, G. Marchesini, Y. L. Dokshitzer and G. Zanderighi, JHEP07(2000), 002 doi:10.1088/1126- 6708/2000/07/002 [arXiv:hep-ph/0004027 [hep-ph]]. 11

    arXiv 2000

  7. [8]

    Storn, and K

    R. Storn, and K. Price, Differential Evolution - a Simple and Efficient Heuristic for Global Optimisa- tion over Continuous Spaces, Journal of Global Optimization, 1997, 11, 341 - 359

    1997

  8. [9]

    Nature Methods, 17(3), 261-272

    Pauli Virtanenet al., Fundamental Algorithms for Scientific Computing in Python. Nature Methods, 17(3), 261-272. DOI: 10.1038/s41592-019-0686-2. https://scipy.org

    doi:10.1038/s41592-019-0686-2

Show all 10 references
  1. [10]

    Nikolaus Hansen and Andreas Ostermeier, Completely derandomized self- adaptation in evolution strategies. Evol. Comput., 9(2):159–195, June 2001. URL: http://dx.doi.org/10.1162/106365601750190398, doi:10.1162/106365601750190398

    2001 doi

  2. [14]

    https://www.starwars.com/databank/imperial-star-destroyer 12

Pith tools

Reviewed August 3, 2026 · model on record in the stance chip above.