2.6 Event Generators
59
In order to obtain clusters with mesonic quantum numbers, non-perturbative gluon
splittings g → qq must be enforced at the scale t min . The decay of clusters to hadrons
is modelled by treating clusters as excited hadrons, where all decay channels allowed
by flavour constraints and kinematics need to be included. Most clusters undergo a
sequential chain of two-body decays. The limited cluster mass spectrum leads to
limited transverse momenta of the produced hadrons and a natural suppression of
heavy flavours. Heavier clusters with masses above 3–4 GeV appear in events with
little parton showering and are typically forced to fission into lighter clusters before
the decay into hadrons is started. Parameters can be introduced to steer the momentum distribution of light quarks from gluon splittings, the flavour distribution, and
the gluon splitting to light diquark/anti-diquark pairs for baryon production. The two
implementations of cluster hadronisation in use in Herwig and Sherpa are based
on [396] and [397], respectively.
A number of unstable hadrons are produced in the hadronisation phase of event
generation. The decay of these hadrons to particles stable on the time scale of collider
experiments is simulated using experimental data wherever possible (most notable
the decay tables from the Particle Data Group (PDG) [398]), and theoretically well
motivated choices in cases where no data is available. The level of sophistication
for the simulation of decays is quite high nowadays, including matrix elements for
certain decay modes and spin correlations. The simulation of particle decays is closely
related to the hadronisation process, such that the free parameters of both models
need to be adjusted simultaneously when comparisons to data are made.
In the context of jet substructure, hadronisation and particle decays lead to a
dispersion of the energy flow and angular distances within jets. Ultimately, this results
in a broadening of distributions in substructure observables, comparable to the effect
of finite resolutions and efficiencies from particle detectors, albeit hadronisation
typically results in smaller modifications. The difference between the effects from
hadronisation and multiple parton interactions is due to the characteristic radiation
pattern. While hadronisation is related to the dynamics of the partonic final state
and thus leads to a broadening of distributions, multiple parton interactions result in
uncorrelated radiation with respect to the hard interaction, leading to a dilution of jet
substructure observables.
2.6.7 Tuning
It is apparent from the discussion above that event generators are complicated algorithms with many free parameters. These parameters are usually adjusted such that
a variety of measurements are well described (referred to as tuning). While final
state showers are constrained by data from e
+ e
− collisions (mostly by using measurements from the LEP and SLD collaborations), the parameters affecting initial
state showers can only be adjusted through data from hadronic collisions (ep, p p,
pp). Additional complications arise because the parameters of the parton shower
are intricately connected with parameters from the modelling of multi-parton inter-
59
In order to obtain clusters with mesonic quantum numbers, non-perturbative gluon
splittings g → qq must be enforced at the scale t min . The decay of clusters to hadrons
is modelled by treating clusters as excited hadrons, where all decay channels allowed
by flavour constraints and kinematics need to be included. Most clusters undergo a
sequential chain of two-body decays. The limited cluster mass spectrum leads to
limited transverse momenta of the produced hadrons and a natural suppression of
heavy flavours. Heavier clusters with masses above 3–4 GeV appear in events with
little parton showering and are typically forced to fission into lighter clusters before
the decay into hadrons is started. Parameters can be introduced to steer the momentum distribution of light quarks from gluon splittings, the flavour distribution, and
the gluon splitting to light diquark/anti-diquark pairs for baryon production. The two
implementations of cluster hadronisation in use in Herwig and Sherpa are based
on [396] and [397], respectively.
A number of unstable hadrons are produced in the hadronisation phase of event
generation. The decay of these hadrons to particles stable on the time scale of collider
experiments is simulated using experimental data wherever possible (most notable
the decay tables from the Particle Data Group (PDG) [398]), and theoretically well
motivated choices in cases where no data is available. The level of sophistication
for the simulation of decays is quite high nowadays, including matrix elements for
certain decay modes and spin correlations. The simulation of particle decays is closely
related to the hadronisation process, such that the free parameters of both models
need to be adjusted simultaneously when comparisons to data are made.
In the context of jet substructure, hadronisation and particle decays lead to a
dispersion of the energy flow and angular distances within jets. Ultimately, this results
in a broadening of distributions in substructure observables, comparable to the effect
of finite resolutions and efficiencies from particle detectors, albeit hadronisation
typically results in smaller modifications. The difference between the effects from
hadronisation and multiple parton interactions is due to the characteristic radiation
pattern. While hadronisation is related to the dynamics of the partonic final state
and thus leads to a broadening of distributions, multiple parton interactions result in
uncorrelated radiation with respect to the hard interaction, leading to a dilution of jet
substructure observables.
2.6.7 Tuning
It is apparent from the discussion above that event generators are complicated algorithms with many free parameters. These parameters are usually adjusted such that
a variety of measurements are well described (referred to as tuning). While final
state showers are constrained by data from e
+ e
− collisions (mostly by using measurements from the LEP and SLD collaborations), the parameters affecting initial
state showers can only be adjusted through data from hadronic collisions (ep, p p,
pp). Additional complications arise because the parameters of the parton shower
are intricately connected with parameters from the modelling of multi-parton inter-