A thrust to trust minimum thrust

Technical Summary: A Thrust to Trust Minimum Thrust

Problem Statement
The paper addresses the determination of the minimum kinematically allowed value of the thrust variable, $T_N$, for a system of $N$ particles in momentum space, subject to momentum conservation ($\sum p_i = 0$). Thrust is a standard event shape variable in high-energy physics, defined as the maximum projection of particle momenta onto an axis $\mathbf{n}$:
$$T = \max_{\mathbf{n}} \frac{\sum_i |\mathbf{p}_i \cdot \mathbf{n}|}{\sum_i |\mathbf{p}_i|}$$
The objective is to find $T_N^{\min}$, or equivalently, the maximum value of the variable $\tau_N = 1 - T_N$. While analytical solutions exist for $N=3$ (equilateral triangle, $\tau_3^{\max} = 1/3$) and $N=4$ (regular tetrahedron, $\tau_4^{\max} = 1 - 1/\sqrt{3}$), no analytical solutions were known for $N \geq 5$. Previous numerical attempts (e.g., Ref. [3]) using stochastic algorithms provided estimates but lacked guarantees of finding the global maximum due to the non-convex nature of the optimization landscape.

Methodology
The authors employ a hybrid approach combining exact algorithmic evaluation of thrust with stochastic global optimization and statistical inference.

1. Thrust Calculation: To evaluate $T_N$ for a given configuration, the paper utilizes an exact algorithm (detailed in Appendix A) that replaces the continuous maximization over the axis $\mathbf{n}$ with a discrete search over sign patterns $s_i \in \{-1, 1\}$. The maximum thrust is given by $\max_S |\sum s_i \mathbf{p}_i|$. For small $N$, a brute-force search over $2^{N-1}$ patterns is efficient; for larger $N$, an iterative algorithm (restarting with random axes) is used to mitigate local minima.
2. Global Optimization: To find the configuration of momenta $\{\mathbf{p}_i\}$ that minimizes $T_N$, the authors treat the problem as a minimization over $3(N-2)$ degrees of freedom (after fixing scale, direction, and momentum conservation). They utilize stochastic evolutionary algorithms, specifically Differential Evolution (DE) and Covariance Matrix Adaptation Evolution Strategy (CMA-ES), implemented via SciPy and Pymoo.
3. Statistical Analysis: Recognizing that stochastic optimizers may not reach the absolute global maximum, the authors apply Extreme Value Theory (EVT) for cases where convergence is ambiguous. They model the distribution of observed maxima using the Weibull distribution (a type of Generalized Extreme Value distribution with a finite endpoint). Parameters are estimated using Maximum Likelihood Estimation (MLE) and Bayesian analysis (using PyMC) to derive credible intervals and upper bounds for the true $\tau_N^{\max}$.

Key Contributions and Results

• Exact Result for N=5: The paper presents, for the first time, an exact analytical result for the minimum thrust of 5 particles in three dimensions. The configuration resembles a specific polyhedral shape (described as having a plane symmetry and resembling an "Imperial Star Destroyer"). The exact value is derived as $\tau_5^{\max} = 1 - T_5^{\min} \approx 0.4539948658$, where $T_5^{\min}$ is determined by the largest real root of a degree-10 polynomial.
• Numerical Determinations for N=6 to N=20: The authors provide numerical values for $\tau_N^{\max}$ up to $N=20$.
  • For $N=6, 7, 8$, the results are presented as highly probable true values based on frequent convergence to a unique maximum.
  • For $N=9$ through $N=12$, where multiple occurrences of the maximum are rare, statistical analyses (MLE and Bayesian) provide estimated endpoints and credible intervals.
  • For $N > 12$, only credible intervals are provided.
• Comparison with Previous Work: The results generally supersede previous estimates from Ref. [3], yielding slightly larger values for $\tau_N^{\max}$ (indicating slightly lower minimum thrust). The differences range from approximately 0.14% at $N=6$ to nearly 1% at $N=14$.
• Dimensional Extensions: The paper includes exact and numerical results for two-dimensional configurations (planar) and selected cases in $d$ dimensions. For $d$ dimensions with $N=d+1$ particles, a conjecture is offered regarding the regular $d$-simplex configuration, yielding specific analytical formulas for even and odd $d$.

Significance and Claims
The paper claims to provide the most accurate estimates to date for the minimum thrust of $N$-particle configurations. A primary contribution is the first exact solution for $N=5$, which corrects and refines previous numerical estimates. The authors note that the configurations yielding minimum thrust are often less isotropic than naive expectations (e.g., regular polyhedrons) might suggest.

The significance of these results is framed within the context of theoretical high-energy physics, where event shapes like thrust are used for theory-experiment comparisons. The authors modestly state that while their numerical results improve upon previous estimates, the small magnitude of the differences (sub-1%) suggests they may not have major consequences at the phenomenological level, though they provide a more rigorous baseline for theoretical calculations. The work also serves as a validation of the statistical methods (EVT) used to bound the results when global optimization is computationally intractable.