arxiv: 2604.18466 · v1 · submitted 2026-04-20 · ⚛️ physics.acc-ph

Recognition: unknown

pyTRAIN -- a modern TRAIN implementation

Michi Hostettler , Xavier Buffat , Tobias Persson , Tatiana Pieloni , Jorg Wenninger

Authors on Pith no claims yet

Pith reviewed 2026-05-10 02:58 UTC · model grok-4.3

classification ⚛️ physics.acc-ph

keywords pyTRAINTRAIN codebeam-beam long-rangePACMAN effectsclosed orbitsLHCtunes and chromaticitiesaccelerator simulation

0 comments

The pith

A Python re-implementation of the 1995 TRAIN code reproduces its calculations of beam-beam effects while adding support for arbitrary particles and interaction points.

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

The paper presents pyTRAIN as a modern Python version of the TRAIN code originally developed to compute self-consistent closed orbits, tunes, and chromaticities for bunch trains under beam-beam long-range and PACMAN effects at facilities such as LEP and the LHC. It uses numpy and scipy for the core numerics, with an optional connection to MAD-X, to improve usability and allow extensions beyond the original Fortran implementation. The new code is shown to match the classic TRAIN outputs and real LHC operating data while introducing the ability to handle any particle type, any number or arrangement of beam-beam points, and extrapolation of orbit distortions to arbitrary locations.

Core claim

The pyTRAIN code reproduces results from the classic TRAIN code and matches observations from LHC physics operation while adding support for arbitrary particle types, an arbitrary number of beam-beam interaction points, and extrapolation of beam-beam induced closed-orbit effects.

What carries the argument

The self-consistent numerical solution for closed orbits, tunes, and chromaticities under beam-beam long-range and PACMAN perturbations, now performed in Python with numpy and scipy.

If this is right

Beam-beam studies can now treat arbitrary particle species instead of being limited to protons.
An arbitrary number and distribution of beam-beam interaction points can be modeled without code changes.
Beam-beam induced closed-orbit effects can be evaluated at any chosen point in the accelerator lattice.
The Python structure improves maintainability and allows direct comparison against current LHC data.
The optional MAD-X link enables seamless use within existing accelerator design workflows.

Where Pith is reading between the lines

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

The added flexibility in particle type and interaction-point layout makes it straightforward to explore beam-beam effects in proposed future colliders with non-standard parameters.
Extrapolation of orbit effects to arbitrary locations could reduce the need for repeated full-lattice runs when assessing local beam quality.
The open Python implementation invites combination with other modern tools for integrated beam-dynamics studies.
Validation against LHC observations suggests the code can serve as a reliable reference for operational tuning of bunch trains.

Load-bearing premise

The numerical implementation in Python using numpy and scipy accurately captures the original TRAIN physics models for self-consistent closed orbits, tunes, and chromaticities under BBLR and PACMAN effects without introducing discrepancies or numerical instabilities.

What would settle it

A direct numerical comparison that reveals a mismatch in computed tunes, chromaticities, or closed-orbit shifts between pyTRAIN and the original TRAIN code for the same LHC bunch-train configuration.

Figures

Figures reproduced from arXiv: 2604.18466 by Jorg Wenninger, Michi Hostettler, Tatiana Pieloni, Tobias Persson, Xavier Buffat.

**Figure 1.** Figure 1: Closed orbits of beam 1 at IP5 in a typical 2023 LHC physics scenario, predicted by the classic TRAIN and pyTRAIN codes. The results for the other beams, and in other points of the machine, are similar. The differences are at the level of 10−9 m. 3 [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗

**Figure 2.** Figure 2: Parasitic separation of the colliding bunch pairs at LHC IP5, introduced by long-range beam-beam effects. The right panels show a magnification of one bunch train structure. LHC proton physics fill 10066 (August 2024), at β ∗ = 1.2 m with 1.6 · 1011 protons per bunch. Luminous Centroid Positions at the Collision Points During collisions, the LHC experiments reconstruct the primary vertices of each recorded… view at source ↗

**Figure 3.** Figure 3: Transverse luminous region position (centroid) of the bunch pairs colliding at LHC IP1, as predicted by pyTRAIN and measured by the ATLAS experiment. The right panels show a magnification of one bunch train structure. LHC calibration transfer fill 9635 (May 2024), β ∗ = 30 cm with 1.3 · 1011 protons per bunch. The preliminary reconstructed luminous region data is provided courtesy of the ATLAS collaboratio… view at source ↗

**Figure 4.** Figure 4: shows bunch offsets measured by the wire scanners in comparison to the pyTRAIN orbit prediction at the wire scanner location. It is worth noting that several wire scans show almost no closed orbit distortion; these measurements were taken when only the first 12 bunches, but no longer bunch trains, were circulating in one of the two beams of the machine. 0 20 40 60 80 100 120 140 −50 0 50 100 25ns Bunch Slo… view at source ↗

read the original abstract

The TRAIN code, developed in 1995 as a post-processor for second-order transport maps from MAD, has been used extensively at the LEP and the LHC to study self-consistent closed orbits, tunes and chromaticities of bunch trains under the presence of beam-beam long-range (BBLR) and PACMAN effects.. This paper presents a modern re-implementation of the TRAIN concept in Python using well-known numeric libraries (numpy, scipy) and an optional link to MAD-X via cpymad. This greatly improves the usability, maintainability and extensibility of the code. New functionality includes the support for arbitrary particle types, an arbitrary number and distribution of beam-beam interaction points, and the extrapolation of the beam-beam induced closed-orbit effects to arbitrary points in the machine. The code is benchmarked against the classic TRAIN code, and simulation results are compared to observations from LHC physics operation.

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.

Desk Editor's Note private letter to a colleague

A practical Python port of the 1995 TRAIN code with added flexibility but thin quantitative validation.

read the letter

This is a Python reimplementation of the 1995 TRAIN code for calculating self-consistent closed orbits, tunes and chromaticities under beam-beam long-range and PACMAN effects in bunch trains. The core idea is the same as the original post-processor for MAD maps, but now built on numpy and scipy with an optional cpymad link to MAD-X. New features include arbitrary particle types, any number and distribution of beam-beam points, and extrapolation of orbit effects to other locations in the ring. These changes make the tool easier to use and extend for current work. The authors benchmark it against the classic TRAIN and compare results to LHC observations, which is the right approach for a reimplementation. The main limitation is that the validation stays at a high level. The abstract mentions benchmarking and data comparisons, but gives no numbers on orbit shifts, tune values, chromaticities, or maximum differences between versions. Without tables or reported error metrics, it is hard to judge how closely the new solver matches the old one on the iterative closed-orbit and tune calculations. The stress-test concern about possible small divergences in matrix handling, interpolation, or PACMAN indexing is fair based on what is shown. This kind of detail matters for users who rely on precise reproduction for LHC studies. The paper is aimed at accelerator physicists who already run beam-beam simulations and need a maintainable, open version they can modify. It does not introduce new physics models or methods. I would bring it to a reading group only if the group focuses on beam dynamics tools. I would not cite the paper in my own work. It does not need a serious referee process; the contribution is incremental and the evidence for exact numerical fidelity is not strongly presented.

Referee Report

2 major / 2 minor

Summary. The manuscript presents pyTRAIN, a Python reimplementation of the 1995 TRAIN code (originally a post-processor for MAD second-order transport maps) used to compute self-consistent closed orbits, tunes, and chromaticities for bunch trains under beam-beam long-range (BBLR) and PACMAN effects. It emphasizes improved usability via numpy/scipy, optional cpymad/MAD-X linkage, and new capabilities including arbitrary particle species, arbitrary numbers and distributions of beam-beam points, and extrapolation of beam-beam orbit effects to any machine location. The work claims successful benchmarking against the legacy TRAIN code and qualitative agreement with LHC operational observations.

Significance. If the numerical reproduction holds, pyTRAIN supplies a maintainable, extensible, and open tool that lowers the barrier to studying beam-beam effects in modern accelerators. The added support for arbitrary particles and flexible interaction-point configurations, together with orbit extrapolation, directly addresses limitations of the original Fortran implementation and could facilitate studies for HL-LHC or future colliders. The use of standard scientific Python libraries and MAD-X interoperability is a clear practical advance.

major comments (2)

[Benchmarking and validation] The benchmarking claim against the original TRAIN code (mentioned in the abstract and presumably in the results section) is not supported by quantitative metrics. No side-by-side tables, maximum absolute or relative differences, or reported values for closed-orbit shifts, Qx/Qy, or chromaticities on identical LHC configurations are provided, leaving open the possibility of silent discrepancies arising from matrix inversion, interpolation, PACMAN bunch indexing, or convergence tolerances in the numpy/scipy iterative solver.
[Comparison to LHC data] The comparison to LHC physics-operation data is described only qualitatively. Without reported error bars, specific observables (e.g., measured vs. simulated orbit offsets at particular IPs), or details of the test cases, it is impossible to judge how well the reimplementation captures the combined BBLR+PACMAN physics for the claimed arbitrary particle types and beam-beam point counts.

minor comments (2)

[Numerical methods] The manuscript would benefit from an explicit statement of the convergence criteria and iteration limits used in the closed-orbit solver, as these choices can affect reproducibility when porting between languages.
[Figures] Figure captions and axis labels should be expanded to indicate which quantities are plotted (e.g., horizontal/vertical orbit shifts, tune shifts) and whether the data are from pyTRAIN, original TRAIN, or measurements.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback and positive evaluation of pyTRAIN's potential utility. We address the major comments point by point below, proposing specific revisions to improve the quantitative support for our claims.

read point-by-point responses

Referee: [Benchmarking and validation] The benchmarking claim against the original TRAIN code (mentioned in the abstract and presumably in the results section) is not supported by quantitative metrics. No side-by-side tables, maximum absolute or relative differences, or reported values for closed-orbit shifts, Qx/Qy, or chromaticities on identical LHC configurations are provided, leaving open the possibility of silent discrepancies arising from matrix inversion, interpolation, PACMAN bunch indexing, or convergence tolerances in the numpy/scipy iterative solver.

Authors: We agree that the current presentation lacks explicit quantitative metrics, which weakens the validation claim. In the revised manuscript we will add a dedicated table (or set of tables) comparing pyTRAIN and legacy TRAIN results for identical LHC bunch-train configurations. The table will report closed-orbit shifts at selected IPs, horizontal and vertical tunes, and chromaticities, together with maximum absolute and relative differences. We will also document the convergence tolerances and solver settings used in both codes to address possible sources of discrepancy. revision: yes
Referee: [Comparison to LHC data] The comparison to LHC physics-operation data is described only qualitatively. Without reported error bars, specific observables (e.g., measured vs. simulated orbit offsets at particular IPs), or details of the test cases, it is impossible to judge how well the reimplementation captures the combined BBLR+PACMAN physics for the claimed arbitrary particle types and beam-beam point counts.

Authors: We acknowledge that the LHC comparison section is currently qualitative only. We will revise it to include concrete test cases with explicit machine parameters (fill number, bunch pattern, beam species), tabulated measured versus simulated orbit offsets at specific IPs, and any available experimental uncertainties. Where quantitative data exist we will report them; where only qualitative agreement can be shown we will state the limitation clearly. revision: yes

Circularity Check

0 steps flagged

No circularity: pure reimplementation with external benchmarking

full rationale

The paper is a software port of the 1995 TRAIN code into Python (numpy/scipy + optional MAD-X link), adding features like arbitrary particle types and beam-beam points while claiming reproduction of the original results plus LHC data matches. No derivation chain exists; the work consists of implementing known physics models and validating them by direct comparison to the classic TRAIN output and operational observations. No equations reduce to self-referential definitions, fitted inputs renamed as predictions, or load-bearing self-citations. The reproduction claim is externally falsifiable via the provided benchmarks rather than being forced by construction.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

This is a code reimplementation paper; no new free parameters, axioms, or invented entities are introduced beyond those inherited from the original TRAIN physics models.

pith-pipeline@v0.9.0 · 5452 in / 1099 out tokens · 38365 ms · 2026-05-10T02:58:38.911524+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

24 extracted references

[1]

Truly Self-Consistent Treatment of the Side Effects with Bunch Trains

E. Keil, “Truly Self-Consistent Treatment of the Side Effects with Bunch Trains”, CERN, 1995

1995
[2]

Features and implications of different LHC crossing schemes

W. Herr, “Features and implications of different LHC crossing schemes”, LHC Project Report 628, CERN, 2003

2003
[3]

The MAD-X project,https://madx.web.cern.ch/
[4]

Xsuite: An Integrated Beam Physics Simulation Framework

G. Iadarolaet al., “Xsuite: An Integrated Beam Physics Simulation Framework”,JACoW HB2023, TUA2I1, 2024

2024
[5]

Non-linear Dynamics in Accelerators

W. Herr, E. Forest, “Non-linear Dynamics in Accelerators”,European Physical Journal C, 2017

2017
[6]

Coherent Beam–Beam Effects

X. Buffat, “Coherent Beam–Beam Effects”,Proceedings of CAS-CERN Accelerator School: In- tensity Limitations in Particle Beams, CERN-2017-006-SP, 2017

2017
[7]

Closed expression for the electrical field of a two-dimensional Gaussian charge

M. Bassetti, G. Erskine, “Closed expression for the electrical field of a two-dimensional Gaussian charge”, CERN ISR-TH/80-06, 1980

1980
[8]

Self-consistent Orbits for Beam-beam Interactions in the LHC

H. Grote, “Self-consistent Orbits for Beam-beam Interactions in the LHC”, CERN Report, 2003

2003
[9]

A Study of Beam-Beam Effects in Hadron Colliders with a Large Number of Bunches

T. Pieloni, “A Study of Beam-Beam Effects in Hadron Colliders with a Large Number of Bunches”, PhD Thesis, EPFL, 2008

2008
[10]

Luminosity Control and Beam Orbit Stability with Beta Star Leveling at LHC and HL-LHC

A. Gorzawski, “Luminosity Control and Beam Orbit Stability with Beta Star Leveling at LHC and HL-LHC”, PhD Thesis, EPFL, 2016

2016
[11]

LHC Luminosity Performance

M. Hostettler, “LHC Luminosity Performance”, PhD Thesis, University of Berne, 2018

2018
[12]

Studies of PACMAN Effects in the HL-LHC

A. Ribes-Metidieri, X. Buffat, “Studies of PACMAN Effects in the HL-LHC”, 2021

2021
[13]

Hostettler, “pyTRAIN”,https://gitlab.cern.ch/mihostet/pytrain/

M. Hostettler, “pyTRAIN”,https://gitlab.cern.ch/mihostet/pytrain/
[14]

Array Programming with NumPy

H. Harriset al., “Array Programming with NumPy”,Nature585, 2020

2020
[15]

SciPy 1.0: Fundamental Algorithms for Scientific Computing in Python

P. Virtanenet al., “SciPy 1.0: Fundamental Algorithms for Scientific Computing in Python”, Nature Methods17, 2020

2020
[16]

Cython: The Best of Both Worlds

S. Behnelet al., “Cython: The Best of Both Worlds”,Computing in Science & Engineering, 2011, https://cython.org/

2011
[17]

The LHC Design Report

O. Bruniget al.(editors), “The LHC Design Report”, CERN, 2004

2004
[18]

Luminosity Scans for Beam Diagnostics

M. Hostettleret al., “Luminosity Scans for Beam Diagnostics”,Phys. Rev. Accel. Beams21, 2018

2018
[19]

Calibration of the Effective Beam Height in the ISR

S. van der Meer, “Calibration of the Effective Beam Height in the ISR”, CERN Report ISR-PO- 68-31, 1968

1968
[20]

Emittance Scans for CMS Luminosity Calibration

O. Karchebanet al., “Emittance Scans for CMS Luminosity Calibration”, inThe LHCC Poster Session, CERN, 2018

2018
[21]

Reconstruction of Primary Vertices at the ATLAS Experiment in Run 1 Proton–Proton Collisions at the LHC

ATLAS Collaboration, “Reconstruction of Primary Vertices at the ATLAS Experiment in Run 1 Proton–Proton Collisions at the LHC”,JINST9, 2014

2014
[22]

Development of ATLAS Primary Vertex Reconstruction for LHC Run 3

ATLAS Collaboration, “Development of ATLAS Primary Vertex Reconstruction for LHC Run 3”, CERN, 2023

2023
[23]

Moenig, W

K. Moenig, W. Kozanecki, C. Bernius, private communication, 2024

2024
[24]

Review of Beam Wire Scanners for LHC Runs 1 and 2

F. Roncarolo et al., “Review of Beam Wire Scanners for LHC Runs 1 and 2”, presented at Beam Size Review, CERN, 2019. 7

2019