The Event Record

A Pythia instance contains two members of the Event class. The one called process provides a brief summary of the main steps of the hard process, while the one called event contains the full history. The user would normally interact mainly with the second one, so we will exemplify primarily with that one.

The Event class to first approximation is a vector of Particles, so that it can expand to fit the current event size. The index operator is overloaded, so that e.g. event[i] corresponds to the i'th particle of the object event. Thus event[i].id() returns the identity of the i'th particle, and so on. Therefore the methods of the Particle class are at least as essential as those of the Event class itself.

As used inside PYTHIA, some conventions are imposed on the structure of the event record. Entry 0 of the vector<Particle> 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 index 1 rather than 0 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 indexing errors.

In the following we will list the methods available. Only a few of them have a function to fill in normal user code.

Basic output methods

Some methods are available to read out information on the current event record:

Particle& Event::operator[](int i)  
const Particle& Event::operator[](int i)  
returns a (const) reference to the i'th particle in the event record, which can be used to get (or set) all the properties of this particle.

Particle& Event::front()  
Particle& Event::at(int i)  
Particle& Event::back()  
returns a reference to the zeroth, i'th or last particle in the event record, as an alternative to the methods above.

int Event::size()  
The event size, i.e. the size of the vector<Particle>. Thus valid particles, to be accessed by the above indexing operator, are stored in the range 0 <= i < size(). See comment above about the (ir)relevance of entry 0.

void Event::list()  
void Event::list(ostream& os)  
void Event::list(bool showScaleAndVertex, bool showMothersAndDaughters = false)  
void Event::list(bool showScaleAndVertex, bool showMothersAndDaughters, ostream& os)  
Provide a listing of the whole event, i.e. of the vector<Particle>. The methods with fewer arguments call the final one with the respective default values, and are non-inlined so they can be used in a debugger. The basic identity code, status, mother, daughter, colour, four-momentum and mass data are always given, but the methods can also be called with a few optional arguments for further information:
argument showScaleAndVertex (default = false) : optionally give a second line for each particle, with the production scale (in GeV), the particle polarization (dimensionless), 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 has two mother and two daughter indices. These may be used to encode zero, one, two or more mothers/daughters, depending on the combination of values and status code, according to well-defined rules. The two methods below can do this job easier for you.

vector<int> Event::motherList(int i)  
returns a vector of all the mother indices of the particle at index i. 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. Many particles may have the same motherList. Mothers are listed in ascending order.

vector<int> Event::daughterList(int i)  
returns a vector of all the daughter indices of the particle at index i. 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. 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. Many partons may have the same daughterList. Daughters are listed in ascending order.

int Event::statusHepMC(int i)  
returns the status code according to the HepMC conventions agreed in February 2009. This convention does not preserve the full information provided by the internal PYTHIA status code, as obtained by Particle::status(), but comes reasonably close. The allowed output values are:

Further output methods

The above methods are the main ones that a normal user would make frequent use of. There are some further methods that also could come in handy, in the exploration of the history of an event, but where the outcome is not always obvious if one is not familiar with the detailed structure of an event record.

int Event::iTopCopy(int i)  
int Event::iBotCopy(int 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. A carbon copy is when the "same" particle appears several times in the event record, but with changed momentum owing to recoil effects.

int Event::iTopCopyId(int i)  
int Event::iBotCopyId(int 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.

vector<int> Event::sisterList(int i)  
returns a vector of all the sister indices of the particle at index i, i.e. all the daughters of the first mother, except the particle itself.

vector<int> Event::sisterListTopBot(int i, bool widenSearch = true)  
returns a vector of all the sister indices of the particle at index i, 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. That is, starting from the top mother, all its daughters are found, except for the traced-up one, and then all their daughters, and so on. This can produce quite extensive and rather useless output, notably when tracing parton showers. 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.

bool Event::isAncestor(int i, int 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.

One data member in an Event object is used to keep track of the largest col() or acol() colour tag set so far, so that new ones do not clash.

mode  Event:startColTag   (default = 100; minimum = 0; maximum = 1000)
This sets the initial colour tag 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.

void Event::initColTag(int colTag = 0)  
forces the current colour tag value to be the larger of the input colTag and the above Event:startColTag values.

int Event::lastColTag()  
returns the current maximum colour tag.

int Event::nextColTag()  
increases the current maximum colour tag by one and returns this new value. This method is used whenever a new colour tag is needed.

Many event properties are accessible via the Info class, see here. Since they are used directly in the event generation, a few are stored directly in the Event class, however.

void Event::scale( double scaleIn)  
double Event::scale()  
set or get the scale (in GeV) of the hardest process in the event. Matches the function of the scale variable in the Les Houches Accord.

void Event::scaleSecond( double scaleSecondIn)  
double Event::scaleSecond()  
set or get the scale (in GeV) of a second hard process in the event, in those cases where such a one has been requested.

Constructors and modifications of the event record

Although you would not normally need to create your own Event instance, there may be times where that could be convenient. The typical example would be if you want to create a new event record that is the sum of a few different ones, e.g. if you want to simulate pileup events. There may also be cases where you want to add one or a few particles to an existing event record.

Event::Event(int capacity = 100)  
creates an empty event record, but with a reserved size capacity for the Particle vector.

Event& Event::operator=(const Event& oldEvent)  
copies the input event record.

Event& Event::operator+=(const Event& addEvent)  
appends an event to an existing one. For the appended particles 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.

void Event::init(string headerIn = "", ParticleData* particleDataPtrIn = 0, int startColTagIn = 100)  
initializes colour, the pointer to the particle database, and the header specification used for the event listing. We remind that a Pythia object contains two event records process and 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.

void Event::clear()  
empties event record. Specifically the Particle vector size is reset to zero.

void Event::reset()  
empties the event record, as clear() above, but then fills the zero entry of the Particle vector with the pseudoparticle used to represent the event as a whole. At this point the pseudoparticle is not assigned any momentum or mass.

void Event::popBack(int n = 1)  
removes the last n particle entries; must be a positive number. History (and other) information of remaning entries is untouched, and so may be internally inconsistent.

void Event::remove(int iFirst, int iLast)  
removes particles in the range between indices iFirst and iLast, including the endpoints. History (and other) information of remaning entries is untouched, and so may be internally inconsistent.

bool Event::undoDecay(int i)  
removes the decay chain of the particle i and thus restores it to its undecayed state. It is only intended for "normal" particle decay chains, and will return false in other cases, notably if the particle is coloured. The procedure would not work if non-local momentum shifts have been performed, such as with a Bose-Einstein shift procedure (or for a dipole shower recoiler). The history information of the remaining particles, many of which may have new indices, is updated to be internally consistent.

int Event::append(Particle entryIn)  
appends a particle to the bottom of the event record and returns the index of this position.

int Event::append(int id, int status, int mother1, int mother2, int daughter1, int daughter2, int col, int acol, double px, double py, double pz, double e, double m = 0., double scale = 0., double pol = 9.)  
appends a particle to the bottom of the event record and returns the index of this position; see here for the meaning of the various particle properties.

int Event::append(int id, int status, int mother1, int mother2, int daughter1, int daughter2, int col, int acol, Vec4 p, double m = 0., double scale = 0., double pol = 9.)  
appends a particle to the bottom of the event record and returns the index of this position, as above but with four-momentum as a Vec4.

int Event::append(int id, int status, int col, int acol, double px, double py, double pz, double e, double m = 0., double scale = 0., double pol = 9.)  
int Event::append(int id, int status, int col, int acol, Vec4 p, double m = 0., double scale = 0., double pol = 9.)  
appends a particle to the bottom of the event record and returns the index of this position, as above but with vanishing (i.e. zero) mother and daughter indices.

int Event::setEvtPtr(int iSet = -1)  
send in the this pointer of the current Event itself to the particle iSet, by default the most recently appended particle. Also generates a pointer to the ParticleDataEntry object of the identity code of the particle.

int Event::copy(int iCopy, int newStatus = 0)  
copies the existing particle in entry iCopy to the bottom of the event record and returns the index of this position. 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, while the status code of iCopy is negated. With a negative newStatus, the new copy is instead set up to be the mother of iCopy. An attempt to copy an out-of-range entry will return -1.

void Event::restorePtrs()  
each particle in the event record has a pointer to the event itself and another to the particle species it belongs to. The latter pointer is automatically set/changed whenever the particle identity is set/changed by one of the normal methods. Of course the pointer values are specific to the memory locations of the current run, and so it has no sense to save them 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.

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.

void Event::rot(double theta, double phi)  
rotate all particles in the event by this polar and azimuthal angle (expressed in radians).

void Event::bst(double betaX, double betaY, double betaZ)  
void Event::bst(double betaX, double betaY, double betaZ, double gamma)  
void Event::bst(const Vec4& vec)  
boost all particles in the event by this three-vector. Optionally you may provide the gamma value as a fourth argument, which may help avoid roundoff errors for big boosts. You may alternatively supply a Vec4 four-vector, in which case the boost vector becomes beta = p/E.

void Event::rotbst(const RotBstMatrix& M)  
rotate and boost by the combined action encoded in the RotBstMatrix M.

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 properties associated with a baryon number that is fully resolved, i.e. where three different colour indices are involved. There are two main applications,

  1. baryon beams, where at least two valence quarks are kicked out, and so the motion of the baryon number is nontrivial;
  2. 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:

The odd (even) kind codes corresponds to a +1 (-1) change in baryon number across the junction.

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.


Separate from the event record as such, but closely tied to it is the PartonSystems class, which mainly stores the parton indices of incoming and outgoing partons, classified by collision subsystem. Such information is needed to interleave multiparton interactions, initial-state showers and final-state showers, and append 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.