Controls which particles are clustered by the jet algorithm.
1 : This option explicitly disallows top quarks, leptons and photons. All other particle types are passed to the jet algorithm.
2 : No extra particles are disallowed.
Alpgen-specific parameters
JetMatching:jetMatch | (default = 1 ; minimum = 1 ; maximum = 2 ) |
Criteria for matching a clustered jet to a parton.
1 : This option can be used with both the CellJet
and SlowJet
algorithms. The delta R between each parton and jet is taken, and the minimal value compared against coneMatchLight * coneRadius for light jets or coneMatchHeavy * coneRadiusHeavy for heavy jets. Note that by default coneRadiusHeavy = coneRadius, see below. If below this value, the parton and jet are considered to match. With CellJet
, the delta R measure is in (eta, phi), while with SlowJet
it is in (y, phi).
2 : This option can only be used with the SlowJet
algorithm. The hard partons are inserted into the parton level event as "ghost" particles, but at the correct (y, phi) position. If this particle is then clustered into a jet, it is considered a match.
The coneMatchLight parameter used when
JetMatching:jetMatch = 1
.
The coneRadiusHeavy parameter used when
JetMatching:jetMatch = 1
. When assigned a negative value,
the value of JetMatching:coneRadius
is used.
The coneMatchHeavy parameter used when
JetMatching:jetMatch = 1
.
Madgraph-specific parameters
JetMatching:doShowerKt On
Off
(default = off
)
This switch changes the merging prescription to the shower-kT matching scheme
outlined in [Alw08]. This scheme differs from "classical" MLM jet
matching with respect to when the matching veto is checked. The shower-kT
scheme considers already immediately after the first shower emission if an
event should be discarded. A detailed comparison of the "classical" MLM
prescription (e.g. Madgraph-style matching with
JetMatching:doShowerKt = off
) and the shower-kT scheme are
in preparation.
kT scale for merging shower products into jets.
JetMatching:nQmatch | (default = 5 ; minimum = 3 ; maximum = 6 ) |
Controls the treatment of heavy quarks.
5 : All quarks (except top) are treated as light quarks for matching.
4 : Bottom quarks are treated separately. Currently, they are unmatched.
The clFact parameter determines how jet-to parton matching
is done. A match is defined as a squared cluster scale that equals:
|clFact| * qCut for inclusive mode,
|clFact| * max(qCut,min(pT(parton)))
for exclusive mode, clFact ≥ 0, or
|clFact| * min(kT(parton)) for exclusive mode,
clFact < 0.
JetMatching:doVeto On
Off
(default = on
)
If turned off, then no jet matching veto will be applied internally in Pythia.
Instead, it is assumed that the (expert) user enforces all necessary vetoes
externally by some other means. Do not change the default value unless
you are an expert in MLM jet matching and want to use your own code
to perform the necessary vetoes.
A preliminary implementation of the FxFx prescription for combining multiple
NLO calculations [Fre12] is available. We would like to stress that
this implementation is still undergoing validation. FxFx merging with aMC@NLO
shares most parameters with the leading-order (MadGraph-style) MLM
prescriptions and can be activated by using the three additional settings
below.
JetMatching:doFxFx On
Off
(default = off
)
If turned on, then FxFx merging with aMC@NLO inputs is performed. Note that
this requires event samples that are specifically generated for this task.
mode
JetMatching:nPartonsNow
(default = -1
)
The number of partons in Born-like events for the current input LHEF. If
the current sample e.g. contains pp → e+e- + 2 partons
Born-like configurations, and pp → e+e- + 3 partons
Real-emission-type events, then JetMatching:nPartonsNow = 2
applies.
The cut applied to regulate multi-jet matrix elements. Note that this cut can
differ from the matching scale.
Alpgen-style parton-jet matching and merging
This section describes the Alpgen-style MLM merging algorithm for PYTHIA8.
The most common reference to the algorithm is [Man02]. In many
respects, however, the implementation provided in the ALPGEN package should
be considered the official description of the MLM merging procedure.
Although designed primarily to work with events generated with ALPGEN, it
can in principle also be used with events from a different source. This
should not be done without thought, however, and it is up to the user to
understand the details of the algorithm and the implications of using a
different hard process generator.
First, either the CellJet
or SlowJet
jet
algorithm is chosen. Both of these algorithms have an R and
an etaMax parameter. In addition, CellJet
has an
eTmin and SlowJet
has a pTmin parameter.
These are the primary three parameters of the merging procedure, and in
practice are set dependent on the cuts applied to the matrix element (ME)
generation. We stress that the merging procedure is not tied to the
geometry of a specific physical detector, but only to the match between
the original partons and the resulting jets, using standard jet algorithms
in the phase space region where partons have been generated.
ME samples with different jet multiplicities are run through the event
generator, and the generation interrupted after parton showers have been
applied, but before resonance decays and beam remnants have been
processed. Note in particular that top quarks will not yet be decayed,
which may lead to slight differences with the PYTHIA 6 interface included
with the ALPGEN package. In what follows, the hardness measure of
jets/partons is taken to be eT when CellJet
is used and pT when SlowJet
is used. The hard
system (ignoring all MPI systems) is then analysed:
-
The particles in the original matrix element process are sorted into
light partons, heavy partons and other particles. For backwards
compatibility, a light parton is defined as the set (d, u, s, c,
b, g) with zero mass. A heavy parton is defined as the set
(c, b, t) with non-zero mass.
-
All particles not originating from the heavy partons or other
particles are passed to the jet algorithm and clustered.
-
Clustered jets are matched to the light partons in the original ME
process. There are two different methods which can be used:
-
Method 1: The following is done for each parton, in order
of decreasing hardness. The delta R between the parton
and all jets is calculated and the smallest value taken. If
this is less than the jet R parameter, possibly
multiplied by a constant, the jet and parton are considered to
match, and the jet is removed from further consideration.
Note that for
CellJet
the delta R measure
is in (eta, phi), while for SlowJet
, it is
in (y, phi).
-
Method 2: This method is only possible when using the
SlowJet
algorithm. Before the clustering is performed,
extremely soft "ghost" particles are added to the event at the
(y, phi) coordinates of the original matrix element
partons. If such a particle is clustered into a jet, the parton
and jet are considered to match. The idea of "ghost" particles
was originally introduced by FastJet as a way to measure jet
areas [Cac06] and should not affect clustering with an
infrared-safe jet algorithm.
-
If there is a light ME parton remaining which has not been matched
to a jet, then the event is vetoed. If all ME partons have been
matched to a jet, but there are still some extra jets remaining,
then two options are possible:
-
Exclusive mode: the event is vetoed. This is typically used when
there are ME samples with higher jet multiplicities, which would
fill in the extra jets.
-
Inclusive mode: the event is retained if the extra jets are softer
than the softest matched jet. This is typically used when
there is no ME sample with higher jet multiplicity, so the parton
shower should be allowed to give extra jets.
-
All particles originating from the heavy partons are passed to the
jet algorithm and clustered.
-
The clustered jets are again matched to the original partons, but
there is no requirement for a match to be present; all matched jets
are immediately discarded. The matching procedure is much the same
as for light partons, but with two differences when delta R
matching is used. First, a different R parameter than that
used by the jet algorithm may optionally be given. Second, all jets
that are within the given radius of the parton are matched, not
just the one with the smallest delta R measure. If there
are still extra jets remaining then in exclusive mode the event is
immediately vetoed, while in inclusive mode the event is retained if
the extra jets are softer than the softest light matched jet.
Some different options are provided, specified further above in the
parameters section. These are set so that, by default, the algorithm
closely follows the official MLM interface provided in the ALPGEN package.
All vetoing of events is done through the usual
";?>User Hooks machinery, and is
therefore already taken into account in the cross section. In the output
from ";?>Pythia::stat()
,
the difference between the "Selected" and "Accepted" columns gives the
number of events that have not survived the vetoing procedure. It is
still the responsibility of the user to add together the results from
runs with different jet multiplicities. In the simplest case, when
ALPGEN input is used and the hard process parameters are used to guide
the merging procedure, it is enough to set the
JetMatching:nJetMax
parameter.
Madgraph-style parton-jet Merging and Matching
This section describes the Madgraph-style parton-jet matching algorithm
for PYTHIA8.
First, the kT jet algorithm is applied using the PYTHIA8
SlowJet
implementation. The useStandardR = false
is used, ie. the (delta R)^2 separation is defined as
2 (cosh(delta y) - cos(delta phi)) rather than the more common
(delta y)^2 + delta phi)^2. The R, etaMax,
and a pTmin parameters are specified. By default, R = 1
and pTmin = qCut . It is not recommended to change these.
These should match the algorithm parameters used in the Madgraph
Matrix Element (ME) generation.
ME samples with different jet multiplicities are run through the event
generator, and the generation is interrupted after parton showers have
been applied, but before resonance decays and beam remnants have been
processed. In what follows, the hardness measure of jets/partons is taken
to be kT relative to qCut.
The hard system (ignoring all MPI systems) is analyzed:
-
The hard partons in the original matrix element process, provided by
the LHEF, are sorted into light partons, heavy partons and other
particles. A heavy parton is defined by the
JetMatching:nQmatch
or by the maxjetflavor
value in the LHEF. nQmatch refers to the absolute value of
the quark PDG identity code.
-
All partons arising from the parton shower are sorted based on their
motherhood. A showered parton arising from a heavy parton or "other"
parton classified in the previous step is not passed to the jet
algorithm. All other partons are clustered into light jets.
- It is checked whether there are "too few" or "too many" light jets.
If the number of light jets is less than the number of light partons
defined by nQmatch, the event is vetoed. If the number is
larger, the event is vetoed only in exclusive mode (defined below).
- In exclusive mode, the number of jets matches the number of light
partons. In inclusive mode, the jets are re-clustered until the number
of jets equals the number of light partons. Next, each light hard
parton is clustered, one at a time, with the jets until a match is found.
A match is defined as a squared cluster scale that equals:
- |clFact| * qCut for inclusive mode,
- |clFact| * max(qCut,min(pT(parton)))
for exclusive mode, clFact ≥ 0, or
- |clFact| * min(kT(parton)) for exclusive
mode, clFact < 0.
If no match is found, the event is vetoed. When a parton
matches a jet, the jet is removed from the collection, and
the process continues. The process terminates when all partons
are matched to a jet, or a parton is unmatched.
-
All particles originating from the heavy partons are not used.
In exclusive mode, it is expected that ME samples with higher parton
multiplicity are available to fill the phase space above qCut.
The inclusive mode is when there are no such samples, and the parton
shower is used to fill the phase space.
Some different options are provided, specified further above. These
are set so that, by default, the algorithm closely follows the
FORTRAN interface ME2Pythia
provided in the Madgraph
package.
All vetoing of events is done through the usual
";?>User Hooks machinery, and is
therefore already taken into account in the cross section. In the output
from ";?>Pythia::stat()
,
the difference between the "Selected" and "Accepted" columns gives the
number of events that have not survived the vetoing procedure. It is
still the responsibility of the user to add together the results from
runs with different jet multiplicities. In the simplest case, when
the hard process parameters are used to guide the merging procedure,
events will be matched in the exclusive mode.
Madgraph-style jet matching with no internal vetoes (assuming an external
veto implementation)
This section describes the facilities that allow expert users to use their own
veto code to perform a Madgraph-style jet matching. This can e.g. be useful
to assess jet matching uncertainties without having to process the same input
events multiple times.
As a first step, any vetoes in the Pythia Jet Matching need to be
disabled by using JetMatching:doVeto = off
. In this mode,
Pythia only stores all the information that is necessary to check (and apply)
the shower-kT or kT-MLM vetoes externally by hand. This information can be
accessed by calling the functions
Event JetMatchingMadgraph::getWorkEventJet()
Return the event after parton showering, without resonance decay products and
particles that should not be included in the jet matching, as necessary to
implement the vetoes in the kT-MLM scheme.
Event JetMatchingMadgraph::getProcessSubset()
Return the event record containing the hard process, without resonance decay
products and particles that should not be included in the jet matching, as
necessary to implement the vetoes in the shower-kT and kT-MLM schemes. In the
former, this event is needed to find the lowest pT in the ME inputs. In
the latter, the event record is used to count the number of hard-process
partons, minimal hard process pT, and to perform the matching of hard-process
particles to shower jets.
bool JetMatchingMadgraph::getExclusive()
Return flag to identify if exclusive or inclusive vetoes should be applied
to this event.
double JetMatchingMadgraph::getPTfirst()
Return the transverse momentum (w.r.t. the beam) of the first parton
shower emission, as needed for the shower-kT scheme.
vector <double> JetMatchingMadgraph::getDJR()
Return a vector of jet clustering scales produced by running the jet algorithm
used for jet matching on the event record without resonance decay products and
particles that should not be included in the matching. In this vector,
clustering scales for combining few jets appear before scales from combining
many jets. This function is useful for the
kT-MLM scheme, or to have quick access to this information for histogramming
and sanity checks.
We do not currently supply example code for this very advanced functionality.
Interested expert users should feel free to contact the Pythia authors for
further explanations.
A note on combining UserHooks
As have been noted above, the matching is implemented using classes
derived from the ";?>UserHooks
class, thereby gaining access to the event generation process at the
relevant locations. For native ALPGEN files, which do not adhere to
the Les Houches standards, it is also necessary to intervene with
a UserHooks
-derived AlpgenHooks
to handle
the extraction and setting of relevant extra information.
One must then combine multiple UserHooks
classes,
such that the functionality of both is present. A prerequisite
is that the different UserHooks
classes should be
declared with virtual inheritance, e.g.
class JetMatching : virtual public UserHooks
Without this option, when combining two UserHooks
-derived
classes, two copies of the base UserHooks
class would be
created, leading to ambiguities.
The two first classes in CombineMatchingInput.h
combine
ALPGEN input with the two different matching schemes, e.g. for the first
class JetMatchingAlpgenInputAlpgen : public AlpgenHooks,
public JetMatchingAlpgen {
public:
// Constructor and destructor.
JetMatchingAlpgenInputAlpgen(Pythia& pythia) : AlpgenHooks(pythia),
JetMatchingAlpgen() { }
~JetMatchingAlpgenInputAlpgen() {}
// Initialisation.
virtual bool initAfterBeams() {
if (!AlpgenHooks::initAfterBeams()) return false;
if (!JetMatchingAlpgen::initAfterBeams()) return false;
return true;
}
// Process level vetos.
virtual bool canVetoProcessLevel() {
return JetMatchingAlpgen::canVetoProcessLevel();
}
....
};
This class inherits from both AlpgenHooks
and
JetMatchingAlpgen
. Any functions which are present
in both classes should be overridden with a function that calls
the different parent methods in the desired order. In the
above example, the only shared methods are the constructor and
initAfterBeams()
.
"?>