The Event Record
The Event
class for event records basically is a vector of
Particle
s, so that it can expand to fit the event size.
The index operator is overloaded, so that event[i]
corresponds to the i
'th particle of an Event
object event
. Thus event[i].id()
returns the
identity of the i
'th particle. References to the first,
i
'th and last particle are obtained with
event.front()
, event.at(i)
and
event.back()
, respectively.
The event size can be found with size()
, i.e. valid
particles are stored in 0 <= i < event.size()
.
Line 0 is used to represent the event as a whole, with its total
four-momentum and invariant mass, but does not form part of the
event history. Lines 1 and 2 contains the two incoming beams, and
only from here on history tracing works as could be expected. That
way unassigned mother and daughter indices can be put 0 without
ambiguity. Depending on the task at hand, a loop may therefore start
at an index 1 without any loss. Specifically, for translation to other
event record formats such as HepMC [Dob01], where the first
index is 1, the Pythia entry 0 definitely ought to be skipped in order
to minimize the danger of errors.
New particles can be added to the end of the current event record
with append(Particle)
, or
append(id, status, mother1, mother2, daughter1, daughter2,
col, acol, p, m, scale)
where p
is the four-momentum vector, and everything except
id
defaults to 0. The append
method returns
the index of the new particle position.
The existing particle at index iCopy
can be copied to the end
with copy(iCopy, newStatus = 0)
. By default, i.e. with
newStatus = 0
, everything is copied precisely as it is,
which means that history information has to be modified further by hand
to make sense. With a positive newStatus
, the new copy is set
up to be the daughter of the old, with status code newStatus
,
and the status code of iCopy
is negated. With a negative
newStatus
, the new copy is instead set up to be the mother
of iCopy
.
The event record can be emptied for the next event by a
clear()
. The last n
entries can be removed by
popBack(n)
, where n = 1
by default.
The =
operator is overloaded to allow a copying of
an event record, and +=
to append an event to an
existing one. In the latter case mother, daughter and colour tags are
shifted to make a consistent record. The zeroth particle of the
appended event is not copied, but the zeroth particle of the combined
event is updated to the full energy-momentum content.
A listing of the whole event is obtained with list()
.
The basic id, status, mother, daughter, colour, four-momentum
and mass data are always given, but the method can also be called with
a few optional arguments for further information:
method
list(showScaleAndVertex, showMothersAndDaughters, os)
where
argument
showScaleAndVertex (default = false
) : optionally give a
second line for each particle, with the production scale (in GeV), the
production vertex (in mm or mm/c) and the invariant lifetime
(also in mm/c).
argument
showMothersAndDaughters (default = false
) :
gives a list of all daughters and mothers of a particle, as defined by
the motherList(i)
and daughterList(i)
methods
described below. It is mainly intended for debug purposes.
argument
os (default = cout
) : a reference to the ostream
object to which the event listing will be directed.
Each particle in the event record has a pointer to the corresponding
particle species in the particle database, used to find some particle
properties. This pointer is automatically set whenever the particle
identity is set by one of the normal methods. Of course its value is
specific to the memory location of the current run, and so it has no
sense to save it if events are written to file. Should you use some
persistency scheme that bypasses the normal methods when the event is
read back in, you can use restorePtrs()
afterwards to set
these pointers appropriately.
The user would normally be concerned with the Event
object
that is a public member event
of the Pythia
class.
Thus, having declared a Pythia
object pythia
,
pythia.event[i].id()
would be used to return the identity
of the i
'th particle, and pythia.event.size()
to
give the size of the event record.
A Pythia
object contains a second event record for the
hard process alone, called process
, used as input for the
generation of the complete event. Thus one may e.g. call either
pythia.process.list()
or pythia.event.list()
.
To distinguish those two rapidly at visual inspection, the
"Pythia Event Listing"
header is printed out differently,
in one case adding "(hard process)"
and in the other
"(complete event)"
. When +=
is used to
append an event, the modified event is printed with
"(combination of several events)"
as a reminder.
One data member in an Event
object is used to keep track of the
largest col()
or acol()
tag set so far, so that
new ones do not clash. The lastcolTag()
method returns the
last tag assigned, i.e. largest value in the current event, and
nextColTag()
ups it by one before returing the value. The
latter method thus is used when a new colour tag is needed.
mode
Event:startColTag
(default = 100
; minimum = 0
; maximum = 1000
)
This sets the initial value used, so that the first one assigned is
startColTag+1
, etc. The Les Houches accord [Boo01]
suggests this number to be 500, but 100 works equally well.
The scale()
methods can be used to set or get the scale
(in GeV) of the event as a whole. Further methods for event properties
may be added in the future.
A few methods exist to rotate and boost events. These derive from the
Vec4 methods, and affect both the
momentum and the vertex (position) components of all particles.
method
rot(theta, phi)
rotate by this polar and azimuthal angle (expressed in radians).
method
bst(betaX, betaY, betaZ)
boost by this vector. Optionally you may provide the gamma
value as a fourth argument, if you deem this may help avoid roundoff
errors for big boosts. You may alternatively supply a Vec4
four-vector, in which case the boost vector beta = p/E.
method
rotbst(M)
rotate and boost by the combined action encoded in the
RotBstMatrix
M
.
There are also a few methods with an individual particle index
i
as input, but requiring some search operations in
the event record, and therefore not possible to define inside the
Particle
class:
method
motherList(i)
returns a vector<int>
containing a list of all the
mothers of a particle. This list is empty for entries 0, 1 and 2,
i.e. the "system" in line 0 is not counted as part of the history.
Normally the list contains one or two mothers, but it can also be more,
e.g. in string fragmentation the whole fragmenting system is counted
as mothers to the primary hadrons. Mothers are listed in ascending order.
method
daughterList(i)
returns a vector<int>
containing a list of all the
daughters of a particle. This list is empty for a particle that did
not decay (or, if the evolution is stopped early enough, a parton
that did not branch), while otherwise it can contain a list of
varying length, from one to many. Many partons may have the same
daughterList
, e.g. in the hard process and fragmentation
steps. For the two incoming beam particles, all shower initiators and
beam remnants are counted as daughters, with the one in slot 0 being
the one leading up to the hardest interaction. The "system" in line 0
does not have any daughters, i.e. is not counted as part of the history.
method
iTopCopy(i), iBotCopy(i)
are used to trace carbon copies of the particle at index i
up
to its top mother or down to its bottom daughter. If there are no such
carbon copies, i
itself will be returned.
method
iTopCopyId(i), iBotCopyId(i)
also trace top mother and bottom daughter, but do not require carbon
copies, only that one can find an unbroken chain, of mothers or daughters,
with the same flavour id
code. When it encounters ambiguities,
say a g -> g g branching or a u u -> u u hard scattering,
it will stop the tracing and return the current position. It can be confused
by nontrivial flavour changes, e.g. a hard process u d -> d u
by W^+- exchange will give the wrong answer. These methods
therefore are of limited use for common particles, in particular for the
gluon, but should work well for "rare" particles.
method
sisterList(i)
returns a vector<int>
containing a list of all the
sisters of a particle, i.e. all the daughters of the first mother,
except the particle itself.
method
sisterListTopBot(i)
returns a vector<int>
containing a list of all the
sisters of a particle, tracking up and back down through carbon copies
if required. That is, the particle is first traced up with
iTopCopy()
before its mother is found, and then all
the particles in the daughterList()
of this mother are
traced down with iBotCopy()
, omitting the original
particle itself. Any non-final particles are removed from the list.
Should this make the list empty the search criterion is widened so that
all final daughters are allowed, not only carbon-copy ones. A second
argument false
inhibits the second step, and increases
the risk that an empty list is returned. A typical example of this
is for ISR cascades, e.g. e -> e gamma where the photon
may not have any obvious sister in the final state if the bottom copy
of the photon is an electron that annihilates and thus is not part of
the final state.
method
isAncestor(i, iAncestor)
traces the particle i
upwards through mother, grandmother,
and so on, until either iAncestor
is found or the top of
the record is reached. Normally one unique mother is required,
as is the case e.g. in decay chains or in parton showers, so that
e.g. the tracing through a hard scattering would not work. For
hadronization, first-rank hadrons are identified with the respective
string endpoint quark, which may be useful e.g. for b physics,
while higher-rank hadrons give false
. Currently also
ministrings that collapsed to one single hadron and junction topologies
give false
.
The Junction Class
The event record also contains a vector of junctions, which often
is empty or else contains only a very few per event. Methods are
available to add further junctions or query the current junction list.
This is only for the expert user, however, and is not discussed
further here, but only the main points.
A junction stores the properites associated with a baryon number that
is fully resolved, i.e. where three different colour indices are
involved. There are two main applications,
- baryon beams, where at least two valence quarks are kicked out,
and so the motion of the baryon number is notrivial;
- baryon-number violating processes, e.g. in SUSY with broken
R-parity.
Information on junctions is set, partly in the process generation,
partly in the beam remnants machinery, and used by the fragmentation
routines, but the normal user does not have to know the details.
For each junction, information is stored on the kind of junction, and
on the three (anti)colour indices that are involved in the junction.
The possibilities foreseen are:
kind = 1
: incoming colourless particle to three
outgoing colours (e.g. baryon beam remnant or
neutralino -> q q q);
kind = 2
: incoming colourless particle to three
outgoing anticolours;
kind = 3
: one incoming anticolor (stored first)
and two outgoing colours (e.g. antisquark decaying to quark);
kind = 4
: one incoming color (stored first) and two
outgoing anticolours;
kind = 5
: incoming colour octet to three colours,
where the incoming colour passes through unchanged and so need not
be bokkept here, while the incoming anticolor (stored first) and the
two outgoing colours are (e.g. gluino decay to three quarks);
kind = 6
: incoming colour octet to three anticolours,
where the incoming anticolour passes through unchanged and so need not
be bookkept here, while the incoming color (stored first) and the two
outgoing colours are.
The odd (even) kind
codes corresponds to a +1 (-1) change in
baryon number across the junction.
Warning: Currently only kind = 1, 2
are
implemented.
The kind and colour information in the list of junctions can be set
or read with methods of the Event
class, but are not of
common interest and so not described here.
A listing of current junctions can be obtained with the
listJunctions()
method.
Subsystems
The event record also contains a few vectors where parton indices can be
stored, classified by subsystem. Such information is needed to interleave
multiple interactions, initial-state showers, final-state showers
and beam remnants. It could also be used in other places. It is intended
to be accessed only by experts, such as implementors of
new showering models.
A listing of current subsystems can be obtained with the
listSystems()
method.