Herwig 7.1 Release Note

TL;DR

Herwig 7.1 introduces significant improvements in multi-jet merging, soft interaction modeling, and hadronisation tuning, enhancing the accuracy and versatility of Monte Carlo event simulations in particle physics.

Contribution

This release presents new features like multi-jet merging at LO and NLO, a soft interactions model, and improved top decay handling, advancing the capabilities of Herwig for collider simulations.

Findings

Enhanced multi-jet merging with dipole shower at LO and NLO

New soft interactions and diffraction model

Updated hadronisation parameters and top decay improvements

Abstract

A new release of the Monte Carlo event generator Herwig (version 7.1) is now available. This version introduces a number of improvements, notably: multi-jet merging with the dipole shower at LO and NLO QCD; a new model for soft interactions and diffraction; improvements to mass effects and top decays in the dipole shower, as well as a new tune of the hadronisation parameters.

Full text

1 Introduction

Herwig is a multi purpose particle physics event generator. Its latest version, Herwig 7.0 Bellm:2015jjp , which was based on a major development of the Herwig++Bahr:2008pv ; Bahr:2008tx ; Bahr:2008tf ; Gieseke:2011na ; Arnold:2012fq ; Bellm:2013lba branch, has superseded the Herwig++ 2.x and HERWIG 6.x versions. Building on the technology and experience gained with the higher-order improvements provided by Herwig 7.0, a major follow-up release, Herwig 7.1 is now available and provides multijet merging at next-to-leading order QCD Bellm:2017xxx as one of its main new features. In addition, the new version includes several improvements to the soft components of the simulation, amongst other changes and physics capabilities, which we will highlight in this release note. Please refer to the Herwig++ manual Bahr:2008pv , the Herwig 7.0 Bellm:2015jjp as well as this release note when using the new version of the program. Studies or analyses that rely on a particular feature of the program should also reference the paper(s) where the physics of that feature was first described.

1.1 Availability

The new version, as well as older versions of the Herwig event generator can be downloaded from the website https://herwig.hepforge.org/. We strongly recommend using the bootstrap script provided for the convenient installation of Herwig and all of its dependencies, which can be obtained from the same location. On the website, tutorials and FAQ sections are provided to help with the usage of the program. Further enquiries should be directed to herwig@projects.hepforge.org. Herwig is released under the GNU General Public License (GPL) version 3 and the MCnet guidelines for the distribution and usage of event generator software in an academic setting, see the source code archive or http://www.montecarlonet.org/.

1.2 Prerequisites and Further Details

Herwig 7.1 is built on the same backbone and dependencies as its predecessor Herwig 7.0, and uses the same method of build, installation and run environment. No major changes should hence be required in comparison to a working Herwig 7.0 installation. Some of the changes, though, might require different compiler versions, particularly our switch to the C++ 11 standard. The tutorials at https://herwig.hepforge.org/tutorials/ have been extended and adapted to the new version and serve as the primary reference for physics setups and as a user manual until a comprehensive replacement for the detailed manual Bahr:2008pv is available.

2 Multijet Merging

Based on the Matchbox development Platzer:2011bc which is central to the NLO matching capabilities of Herwig, a multijet merging algorithm detailed in Bellm:2017xxx has been implemented together with the dipole shower algorithm and based on an improved, unitarised merging prescription following the proposal set out in Platzer:2012bs . The algorithm is able to merge cross sections for multiple jet production at the NLO QCD level, and has been tested with a range of standard model processes such as vector boson or Higgs boson plus jets production, top pair production, and pure jet production.

Compared to the simple input file structure of the Matchbox framework, minor additional commands are needed to perform calculations with several jet multiplicities merged to the dipole shower. Input file examples for a range of processes are provided in share/Herwig/Merging. Different from the standard NLO matching input files for use with Matchbox, merging only requires a slightly different process definition. For example,

do MergingFactory:Process p p -> W+ [j j j]

set MergingFactory:NLOProcesses 2

set Merger:MergingScale 10.*GeV

sets up on-shell $W^{+}$ production with up to three jets and including NLO QCD corrections to the inclusive and one-jet process. For the merging scale we recommend some default ranges (LHC at $13\ {\rm TeV}$: $10-30\ {\rm GeV}$, LEP at $91\ {\rm GeV}$: $4-6\ {\rm GeV}$ and for HERA run 2 with $27\ {\rm GeV}$ electrons/positrons on $820\ {\rm GeV}$ protons we have found that a merging scale between $8-15\ {\rm GeV}$ has provided reliable results). For colliders running significantly outside these parameters, and in dependence on acceptance cuts, the value needs to be adjusted, possibly down to small merging scales. This provides stable predictions due to the unitarisation procedure.

Example plots are shown in Fig. 1, highlighting the fact that inclusive quantities do not receive big corrections, while higher jet multiplicities are significantly improved by the procedure. Variations of the factorization and renormalization scales can be obtained as with all other simulation setups.

3 New Soft Model

Our model of soft interactions in the context of multiple partonic interactions (MPI) has been replaced by a new approach. The existing MPI model still forms the basis of the physics simulation by separating hard and soft interactions with the help of the parameter $p_{\perp}^{\rm min,0}$ Bahr:2008dy ; Bahr:2008wk ; Bahr:2009ek . In the context of this framework a number of soft interactions $N_{\rm soft}$ is determined as before. The model for colour reconnections Gieseke:CRmodel stays in place and still has a significant impact on the final state as previously uncorrelated multiple scatters have to acquire some colour correlation.

Instead of the generation of a gluon pair for each soft interaction we now generate a number of rapidity-ordered gluons (and a pair of quarks) as depicted in Fig. 2. The gluons’ rapidities are determined based on multiperipheral kinematics Baker:1976cv , where the longitudinal momenta are slightly smeared, cf. Gieseke:2016fpz . The unordered transverse momenta $p_{\perp}$ of the gluons are limited to the regime of soft transverse momenta, i.e. $p_{\perp}<p_{\perp}^{\rm min,0}$. Further details of the model are described in Gieseke:2016fpz .

In order to complete the model towards softer and more forward interactions we also added a simple model for diffractive scattering which complements the hard MPI model for minimum-bias interactions. The model for diffractive final states heavily uses the cluster hadronization model already used by Herwig. Details of the model and several results have been presented elsewhere Gieseke:2016fpz ; Gieseke:2017yfk ; Gieseke:2017ccw . Here, we highlight two findings: most notably, the unphysical predictions for the distribution of forward rapidity gaps is now replaced by an excellent description of data, Fig. 3, highlighting the expected composition of non-diffractive events at small gap sizes, and diffractive contributions at large gaps. We stress the fact that the old model, which generated a ’bump’ structure in this spectrum due to artificial colour re-connections, was not meant to describe these interactions, so no conclusions from this data comparison could be drawn. Another positive result of the new modelling of soft particle production is the improvement of soft transverse momentum spectra of charged particles, also in minimum-bias interactions, see Fig. 4.

The model for soft interactions has become the new default model. The matrix elements for diffractive scattering are used alongside the hard and soft MPI model by default in the simulation of minimum-bias matrix elements. There are two new parameters for the soft interaction model that determine the number of gluons per soft interaction and its growth with energy. Both parameters have been tuned to minimum-bias data.

4 EvtGen interface

The internal Herwig modelling of hadron decays includes sophisticated modelling with off-shell effects and spin correlations. However, it has proven impossible to provide a good tune to data for the decay of bottom and charm mesons, largely due to the lack of published distributions. Given that EvtGenLange:2001uf has been tuned to non-public data from the B-factory experiments and internally uses similar algorithms to include spin correlations in particle decays, in Herwig 7.1 we include an interface to EvtGen which communicates the spin information between the two programs ensuring that the full correlations are generated. EvtGen is now the default for the decay of bottom and charm mesons. As there is less data for bottom and charm baryons and our modelling of baryonic form-factors is more sophisticated, the decays of heavy baryons continue to be performed by Herwig. This leads to the improvement of a number of distributions, e.g. the momentum distribution of $D^{*}$ mesons Barate:1999bg , Fig. 5, where there is a significant contribution from $D^{*}$ mesons produced in bottom meson decays.

5 Mass Effects in the Dipole Shower

We recall that Herwig 7 contains two shower algorithms, based on angular ordering (which we call QTilde) or dipole showering respectively. The dipole shower has been extended in version 7.1 to include the showering of top quarks in both their production and decay with the option to include the NLO correction to the decay. We show an example of the results for top production in Fig. 6, in comparison with ATLAS dataAad:2015eia . The dipole shower can now perform showering of all Standard Model (SM) processes, including the NLO Powheg-type correction to all SM decays. The NLO correction can be switched on and off by setting,

set DipoleShowerHandler:PowhegDecayEmission Yes/No

and is on by default.

We have also performed a detailed analysis and new derivation of the kinematics used to describe splittings of dipoles involving massive emitters and/or spectators. As part of this we have derived and implemented covariant formulations of the physical momenta of the partons following a splitting in terms of the physical momenta of the partons prior to the splitting for these dipoles. The kinematics for all dipole splittings in the dipole shower and Matchbox now use such a formulation, with an evolution variable which is directly connected to the transverse momentum variable relevant for the collinear or quasi-collinear limits. The effect of these improvements can be clearly seen in our modelling of $B$-Fragmentation in $e^{+}e^{-}$ annihilation at the $Z^{0}$ mass, see Fig. 7, and more details will be covered in a forthcoming publication.

6 Shower Variations and Reweighting

Evaluation of shower uncertainties is an important part of modern Monte Carlo studies. Shower uncertainties are traditionally evaluated by performing a full set of event simulations for each variation of interest.

To reduce the computational cost of evaluating shower uncertainties we have introduced functionality to perform on-the-fly parton shower reweighting in Herwig Bellm:2016voq . In this framework, each event is showered using a central set of parameters. In addition, on a splitting-by-splitting basis, we evolve a weight relative to the central shower for each set of varied parameters. We have currently implemented reweighting to evaluate variations of the factorization and renormalization scales used in the shower however it is a general technique that could be applied to other variations in future developments.

A very efficient sequence of the veto algorithm for the central scale choice can lead to inefficient performance of the algorithm for the variations. We have included a ‘detuning parameter’ which can be used to improve the convergence of the reweighted results at the expense of a less efficient algorithm for the central prediction.

Reweighting is available in both showers. Multiple variations can be included in a single run and each variation requires a unique name, ‘varName’, which is used to identify the weight in the HepMC record. Each variation corresponds to a pair of scale factors, $\xi_{R}$ and $\xi_{F}$, to be applied to the renormalization and factorization scales respectively. Finally each variation can be applied to the showering of the hard process only (Hard), secondary processes only (Secondary) or to both parts (All):

do ShowerHandler:AddVariation VarName xR xF Hard/Secondary/All

set SplittingGenerator:Detuning Factor

do DipoleShowerHandler:AddVariation VarName xiR xiF Hard,Secondary,All

set DipoleShowerHandler:Detuning Factor

On top of using reweighting for the shower variations, the dipole shower offers a number of reweighting and biasing facilities which are e.g. used for the KrkNLO method (see below). These are available through the DipoleSplittingReweight and DipoleEventReweight classes. Very flexible veto functionality is also available for the angular ordered shower through the ShowerVeto and FullShowerVeto classes.

7 KrkNLO

This version of Herwig contains an implementation of the KrkNLO method Jadach:2015mza . This provides NLO QCD corrections to LO matrix elements for specific processes following this paradigm as an alternative to the other matching schemes available. The implementation currently supports the Drell-Yan ($Z/\gamma^{*}$) process, and Higgs production via gluon-fusion (in the large top-mass limit) and is available for the dipole shower Platzer:2009jq ; Platzer:2011bc . For the Drell-Yan process, it is possible to use both the MC and ${\rm MC}_{\rm DY}$ variants of the MC schemeJadach:2016acv . This module was validated against a previous, independent, implementation using the published DY results of Ref. Jadach:2015mza and was also used to simulate the first results for this method in Higgs production Jadach:2016qti . KrkNLO can be enabled by using

read Matchbox/KrkNLO-DipoleShower.in

set KrkNLOEventReweight:Mode H

set KrkNLOEventReweight:PDF MC

set KrkNLOEventReweight:AlphaS_R Q2

set KrkNLOEventReweight:AlphaS_V M2

in combination with an MC-scheme PDF. The MC-scheme PDFs, example input-cards, and other relevant codes are hosted at https://krknlo.hepforge.org/.

8 Other Changes

Besides the major physics improvements highlighted in the previous sections, we have also made a number of smaller changes to the code and build system which we will summarize below. Please refer to the online documentation for a fully detailed description or contact the authors.

8.1 Steering, input files and weights

The steering of the Herwig executable has seen a number of improvements, mainly:

• •

A new run mode has been added to solely perform the merging of integration grids from parallel integration runs,

Herwig mergegrids <run file name>

• •

A high-level run-time interface is now available to steer Herwig within more complex frameworks such as experimental software without the need to execute the binary. This includes all of the read, build, integrate, mergegrids and run steps.

The structure of input files for non-Matchbox-based processes has been adapted to use the snippet input file mechanism and is now in line with steering matched and merged processes. On top of this, a large number of input file switches which have before used On,Off or True,False to indicate their state have been changed to Yes,No.

As far as integration and event generation are concerned, we have made a choice that by default sampling is run in AlmostUnweighted mode, i.e. events carry in general varying weights, most of which are unity. This is to account for the fact that the grid adaption might only have encountered a maximum weight close to the true maximum weight and strict unweighting in this case could skew distributions and cross section estimates. The reference weight to which events are unweighted can also be adjusted to keep weight distributions mostly narrow while reducing fluctuations in tails due to a small frequency of contributing events. Alongside this, the adaption parameters of both the CellGridSampler and MonacoSampler have been revised.

8.2 Minor improvements and bug fixes

A number of minor changes and bug fixes are worth noting, in particular, there have been new options for the physics simulation besides the ones described in the previous text:

• •

Colour reconnection of octet systems into a single cluster are now prevented, improving the description of a number of observables sensitive to these dynamics, as well as some unexpected features which have been observed in preceeding work Gras:2017jty .

• •

For both showers it is now possible to alter the scale choice and ordering properties in $g\to q\bar{q}$ splittings.

• •

New options of shower scale choices are available for NLO matched processes.

• •

The $\alpha_{s}$ running in the dipole shower can now explicitly be switched to use the CMW scheme Catani:1990rr , through both a scaling of its argument as well as by explicitly adding the $\alpha_{s}^{2}K_{g}$ contribution, such that these contributions do not anymore need to be absorbed into a tuned value of $\alpha_{s}$.

• •

Several options have been added for the emission phase space in the dipole shower, which are subject to a more detailed, future study.

• •

Structures in ThePEG have been extended to cover processes which do not exhibit a (tree) diagram-like internal structure, such as instanton- and sphaleron-induced transitions.

Technical issues which have been addressed include:

• •

Matchbox is now able to handle processes which do not contain coloured external legs.

• •

The dipole shower can handle zero-momentum-transfer initial-final colour connections, which have prevented running minimum bias simulation with this shower algorithm before.

• •

Several levels of assumptions (such as Standard Model-like interactions, conservation of lepton flavour number, quark flavour diagonal interactions) can be imposed on the generation of candidate sub-processes to reduce combinatorial complexity for processes with many legs.

8.3 Build and external dependencies

As of version 7.1, Herwig is now enforcing the use of a C++11 compliant compiler, and C++11 syntax and standard library functionality is used widely within the code. The herwig-bootstrap script is able to provide such a compiler along with a full Herwig plus dependencies build. herwig-bootstrap will also enforce the newest versions of external amplitude providers; specifically we now use:

• •

OpenLoops Cascioli:2011va versions $\geq$ 1.3.0 with the Collier library Denner:2016kdg for tensor reduction (should older versions of OpenLoops be required, the input files require the additional option set OpenLoops:UseCollier Off), and

• •

GoSam versions $\geq$ 2.0.4 to pick up the correct normalization for loop induced processes outside of specialized setups.

A number of changes have also been implemented to reduce run-time load for allocating and de-allocating various containers, and to reduce overall memory consumption.

8.4 Licensing

While Herwig 7.0 and older versions of Herwig++ have been licensed under the GNU General Public License GPL, version 2, Herwig 7.1 and future versions are distributed with the GPL version 3. The MCnet guidelines for the distribution and usage of event generator software in an academic setting apply as before, and both the legally binding GPL license and the MCnet guidelines are distributed with the code.

9 Summary and Outlook

We have described a new release, version 7.1, of the Herwig event generator. This new release contains a number of improvements to both perturbative and non-perturbative simulation of collider physics and will form the basis of further improvements to both physics and technical aspects.

Acknowledgments

We are indebted to Leif Lönnblad for his authorship of ThePEG, on which Herwig is built, and his close collaboration, and also to the authors of Rivet and Professor. We are also grateful to Malin Sjödahl for providing the ColorFull library for distribution along with Herwig.

This work was supported in part by the European Union as part of the FP7 and H2020 Marie Skłodowska-Curie Initial Training Networks MCnetITN and MCnetITN3 (PITN-GA-2012-315877 and 722104), the Lancaster-Manchester-Sheffield Consortium for Fundamental Physics under STFC grant ST/L000520/1 and the Institute for Particle Physics Phenomenology under STFC grant ST/G000905/1. AP acknowledges support by the ERC grant ERC-STG-2015-677323. MHS acknowledges support of this project by the Institute for Particle Physics Phenomenology through an IPPP Associateship. SP is grateful to the Erwin-Schrödinger-Institute at Vienna for kind hospitality while part of this work was completed.

CR acknowledges support by the US Department of Energy under grant DE-SC0010102.