Source-linked AI summary
A proposal for a standard interface between Monte Carlo tools and one-loop programs
T. Binoth, F. Boudjema, G. Dissertori, A. Lazopoulos, A. Denner, S. Dittmaier, R. Frederix, N. Greiner, S. Hoche, W. Giele, P. Skands, J. Winter, T. Gleisberg, J. Archibald, G. Heinrich, F. Krauss, D. Maitre, M. Huber, J. Huston, N. Kauer, F. Maltoni, C. Oleari, G. Passarino, R. Pittau, S. Pozzorini, T. Reiter, S. Schumann, G. Zanderighi
TL;DR
Existing tree-level Monte Carlo tools provide realistic collider simulations but leave absolute-rate predictions sensitive to missing higher-order corrections. The paper proposes a standard interface linking these tools with one-loop programs, enabling their combination for next-to-leading-order predictions and establishing an initial implementation framework.
Problem
Tree-level Monte Carlo predictions of absolute rates suffer from uncertainties from missing higher-order terms, motivating integration of one-loop corrections.
Method
The paper proposes a standardized interface for exchanging one-loop information between one-loop programs and Monte Carlo tools that provide tree amplitudes and phase-space integration.
Results
The paper presents the first proposal for a standardized interface between leading-order Monte Carlo tools and one-loop programs.
Takeaways & Limitations
The proposed interface is intended to accelerate next-to-leading-order results while allowing experimentalists to use familiar Monte Carlo tools.
Abstract
from arXiv · showhide
Many highly developed Monte Carlo tools for the evaluation of cross sections based on tree matrix elements exist and are used by experimental collaborations in high energy physics. As the evaluation of one-loop matrix elements has recently been undergoing enormous progress, the combination of one-loop matrix elements with existing Monte Carlo tools is on the horizon. This would lead to phenomenological predictions at the next-to-leading order level. This note summarises the discussion of the next-to-leading order multi-leg (NLM) working group on this issue which has been taking place during the workshop on Physics at TeV colliders at Les Houches, France, in June 2009. The result is a proposal for a standard interface between Monte Carlo tools and one-loop matrix element programs.
F. Boudjema ∗
This section lists contributors and their institutional affiliations across France, Switzerland, Germany, the United Kingdom, the United States, and Belgium.
- The listed affiliations include institutions in France, Germany, Switzerland, and the United Kingdom.
- Contributors include J. Archibald, G. Heinrich, F. Krauss, D. Maître, G. Dissertori, and A. Lazopoulos.
- Additional affiliations are given for Michigan State University, FNAL, SLAC, and CP3 at Université Catholique de Louvain.
G. Passarino
This section lists affiliations in Italy and Spain.
- The listed affiliations include Università di Torino and Università di Granada.
- The affiliations are associated with INFN, the Department of Theoretical Physics, CAFPE, and the Department of Theoretical Physics and the Cosmos.
S. Pozzorini
This section lists CERN/PH in Geneva, Switzerland.
- The listed affiliation is CERN/PH in Geneva, Switzerland.
- The address is given as CH–1211 Geneva 23, Switzerland.
- The section contains an institutional affiliation rather than research findings.
T. Reiter
This section lists Nikhef in Amsterdam, the Netherlands.
- The listed affiliation is Nikhef, located at Science Park 105 in Amsterdam.
- The address is given as 1098 XG Amsterdam, The Netherlands.
- The section contains an institutional affiliation rather than research findings.
S. Schumann
The section identifies the Heidelberg affiliation.
- The affiliation is Institut für Theoretische Physik, Universität Heidelberg, D-69120, Heidelberg, Germany.
G. Zanderighi
The section identifies an Oxford affiliation and dedicates the paper to Thomas Binoth, while naming its subject keywords.
- The Rudolf Peierls Centre for Theoretical Physics is located at 1 Keble Road, OX13PN, Oxford, UK.
- The paper honors Thomas Binoth and names the proposal the Binoth Les Houches Accord.
- The keywords are Monte Carlo tools, one-loop computations, and the Les Houches Accord.
1. Introduction
Existing tree-level Monte Carlo tools are mature, while advances in one-loop calculations make their combination for NLO predictions feasible and valuable. The paper reports a working-group proposal for a standardized interface to support this integration.
- Existing Monte Carlo event generators use tree matrix elements, sophisticated generation and phase-space methods, parton showers, and hadronisation models to simulate realistic events.
- Including higher-order corrections is desirable because renormalisation-scale sensitivity remains, while electroweak Sudakov logarithms can produce corrections of about ten percent at high energies.
- One-loop matrix-element evaluation has matured through unitarity-based and improved Feynman-diagram methods, removing a prior computational bottleneck.
- Several groups can now make NLO predictions for processes with up to four final-state particles, bringing broad LHC-relevant coverage within reach.
- Modularity makes combining one-loop information with existing tree-level Monte Carlo tools feasible and can accelerate NLO production by separating loop evaluation from tree-like computation and phase-space integration.
- The NLM working group’s Les Houches discussion produced a proposal for a standardized interface between Monte Carlo tools and one-loop programs.
2. Modular structure of one-loop computations
NLO computations decompose into tree-level, virtual, real-emission, collinear, and subtraction components whose divergences are organized to yield finite contributions. This modularity motivates an interface in which providers supply renormalized one-loop information while retaining flexibility in infrared treatment.
- The modular NLO structure combines Born, real, virtual, and collinear contributions for each partonic subprocess, with real emission adding one final-state particle.
- Existing tree-amplitude generators can efficiently evaluate Born, real-emission, and collinear terms, whereas the virtual term integrates the Born–one-loop interference over the m-particle phase space.
- The one-loop amplitude contains UV and IR divergences; UV renormalization removes UV poles, while infrared regularization determines the finite remainder and remaining IR poles.
- The proposed interface assumes providers perform UV renormalization internally and treat infrared singularities uniformly, while allowing flexible regularization and subtraction choices.
- Subtraction methods add local real-emission counterterms and integrated counterparts so infrared divergences cancel separately in the m- and m+1-particle contributions.
- Figure 1 separates the one-loop module from IR subtraction modules, with integrated subtraction terms compensating loop divergences and horizontal-line contributions becoming finite after summation.
- For loop-induced processes absent at tree level, the modular structure contains only a one-loop module with the squared one-loop amplitude instead of an interference term.
3. A computational model for an interface
The proposal defines a two-phase MC/OLP interaction model that separates setup from phase-space evaluation while preserving flexibility in one-loop implementations and schemes.
- Interface goals: The interface transfers information between one-loop programs and Monte Carlo tools without constraining how either program internally operates.The common platform standardizes exchanged information while allowing different implementation models, including on-the-fly amplitude generation and hard-coded libraries.
- Two-phase interaction: The computational model has an initialization phase for process availability, parameters, and options, followed by a run-time phase for phase-space-dependent one-loop contributions.Initialization establishes what can be computed; run-time queries return contributions for cross-section evaluation.
- Run-time information: Run-time output should include the finite interference term and the squared leading-order amplitude, with the interference optionally split for more efficient Monte Carlo sampling.The squared LO amplitude also supports consistency checks and flexibility in renormalisation and factorisation schemes.
- Communication mechanism: Initialization and run-time communication are preferably implemented through files and function calls, respectively, with model parameters passed in Les Houches Accord format.Table 1 summarizes the basic input/output information exchanged by the one-loop program.
- Schemes and outputs: The interface permits explicit labels for regularisation schemes, provider approximations, and the minimum colour- and helicity-summed interference output.Approximations must be transparently defined and flagged during initialization, while additional return values remain provider-documented options.
4. Proposal for the implementation of the MC/OLP interface
The implementation proposal defines compatible communication between MC and OLP programs through initialization files, contract files, and run-time function calls, with explicit settings for processes, schemes, approximations, and outputs.
- Implementation model: The MC/OLP connection standardizes compatible input/output values while allowing linking, bridge programs, or transmission protocols for different languages and environments.The proposal deliberately leaves the underlying connection mechanism outside its scope.
- Initialization files: Initialization creates an order file describing requested subprocesses and settings, while the OLP returns a contract file specifying how the programs connect at run time.Subprocesses receive integer labels used later to identify contributions.
- Settings and conventions: The proposal requires explicit treatment of colour/helicity information, correction types, infrared regularisation, electroweak settings, model parameters, and provider-defined approximations.The OLP is assumed to provide ultraviolet-renormalised output, while infrared-subtracted results may be optional and must identify their subtraction scheme.
- Infrared treatment: The interface discourages passing only infrared-subtracted information because doing so would couple the MC, OLP, and subtraction provider and reduce modularity.The default infrared-subtraction setting is None, although schemes such as DipoleSubtraction, FKSSubtraction, and AntennaSubtraction may be available.
- Subprocess decomposition: The OLP may subdivide requested subprocesses into separately labelled contributions so multi-channel integration can exploit differences in their Monte Carlo weights.A single unsplit subprocess may also be returned, depending on the SubdivideSubprocess setting.
- Order-file contents: The order file records subprocesses using particle data group codes and can specify coupling powers, correction type, colour/helicity treatment, regularisation, model files, and operational options.The listed process format distinguishes decays from scattering processes and supports flags such as AlphasPower and AlphaPower.
5. Extending the interface to include EW OLP
The proposed electroweak extension requires additional communication between the MC and OLP to coordinate resonance, regularisation, coupling, and renormalisation choices. The interface preserves flexibility while ensuring the MC receives the information needed for consistent infrared cancellation and real-correction calculations.
- Electroweak interfacing requires more MC–OLP communication because renormalisation and unstable-particle treatment are more complicated than in QCD.The exchanged input parameters must be interpreted consistently with the chosen computational scheme.
- 5.1 Treatment of resonances: The interface supports ComplexMassScheme, FixedWidthScheme, PoleScheme, or an OLP-defined resonance treatment, with unsupported requests returned in the contract file.Alternative schemes may require additional information such as the gauge choice or included diagrams.
- 5.2 Infrared regularisation: Mass-regulated and dimensionally regulated infrared singularities are communicated through IRregularisation and related flags, including the regulator masses when required.The framework also allows the OLP to provide an infrared-subtracted result when regulator-mass dependence is otherwise less transparent.
- 5.3 Electroweak renormalisation scheme: The default electroweak renormalisation choice is the α(0) scheme, while alphaMZ, alphaGF, alphaRUN, alphaMSbar, and OLPdefined are also supported.The OLP can return effective couplings such as αOLP and αIR for use by the MC during runtime.
- 5.4 The run-time phase: The MC must obtain OLP-consistent Born information and infrared-module parameters, especially when masses, sin^2 θW, or couplings differ between programs.A coupling-stripped MC mode and runtime transfer of coupling parameters help maintain consistency and enable Born-level checks.
6. Summary and outlook
The paper presents a first proposal for a standard interface between leading-order Monte Carlo tools and one-loop programs, while identifying future refinements and extensions.
- The initiative aims to improve coordination between Monte Carlo and loop-calculator communities and support more efficient resource use beyond leading order.The authors invite feedback and hope the approach will help describe a wide range of Tevatron and LHC data.
- The group presents the first proposal for a standardised interface between leading-order Monte Carlo tools and one-loop programs.Several program authors were committed to implementing it, and initial experiences were described as promising.
- The interface is provisional: details such as input/output formats may change as standardisation continues.
- A natural extension would add information needed to combine the proposed next-to-leading-order fixed-order setup with a parton shower.This concerns infrared subtraction and is treated as orthogonal to passing one-loop information.