I3Particle

Purpose and properties

The I3Particle class is a very generic class that we use to describe both simulated particles (primaries, secondaries) and reconstructed muon tracks (through-going, starting, stopping, contained) and cascades. The class has nine floating point data members that describe the kinematical properties of a given particle:

  • position (3 degrees of freedom)

  • time

  • direction (2 d.o.f.)

  • energy

  • length

  • speed

The actual meaning of those properties depends on the type, shape, location and fit status of the particle, which are specified with integer data members (enum types, actually) and described in more detail below.

Simulated particles are usually collected in an I3MCTree object which is stored in the Q frame with the name "I3MCTree". This tree can contain any number of particles, from just a few to millions.

Reconstructed particles are usually stored directly in the I3Frame. Some reconstructions produce some extra (algorithm specific) information that is stored in one or more separate I3FrameObject objects. Usually this extra information specifies diagnostics of the reconstruction process and/or reconstructed properties of the particle. Often the name (frame key) of the additional object is the name of the particle object with “Params” or “FitParams” appended.

Data members

Floating point data members

Position and time

The time data member of an I3Particle object is given with respect to the event time window. The position data member (stored as an I3Position object) then gives the location where the particle was at that time. What the particle was doing at that position and what happened before/after that time depends on the type, shape and location information.

Note that the time is not necessarily within the event window and the position can also be far out of the detector array.

The coordinate system for the position should always be the local coordinate system of the detector you are working with Wiki: Coordinate_system.

Direction

The direction of an I3Particle object is stored in an object of type I3Direction. The direction can be represented in several ways, most commonly we use astronomy-oriented zenith and azimuth angles, which specify where the particle is coming from. The zenith angle is a polar angle, equal to zero for particles coming in vertically from above. The reconstruction-oriented theta and phi angle point in the opposite direction, so they indicate where the particle is going to. This propagation direction can also be represented as a normalized Cartesian 3-vector, using the IceCube coordinate system.

Length

If the length has a positive value then this usually means what you think it means: the (simulated or reconstructed) particle traveled over the specified length into the specified direction (see below) from a location specified by the position data member.

A zero length usually indicates a shower that short enough to be considered pointlike.

Negative length can be used for track-like stopping particles (simulated or reconstructed).

For reconstructed through-going tracks (shape InfiniteTracks) the length is set to NAN.

Energy

For simulated particles the energy represents the energy that the particle had at the given position (primary at interaction, or secondary at creation).

For reconstructed the meaning of the energy value depends on the algorithm. It may be an energy proxy, e.g. a number that correlates with the mean energy loss (averaged dE/dX, number of emitted photons per meter) along the part of the track that intersects the array. I may also be an estimate of the muon energy at some point (the point where it entered the array, or where it was closest to the center of the detector or the COG of the pulses). It might even be an estimate of the neutrino energy, assuming the event was caused by a neutrino energy. For reconstructed showers the energy usually represents the electron equivalent energy, i.e. the energy that an electron should have to cause (a shower that causes) the observed amount of light in the array.

In order to understand the issues surrounding the reconstructed energy it is highly recommended to read the energy reconstruction paper.

Speed

For most use cases, the speed should be zero or c=299792458 m/s.

However, some reconstruction methods actually try to reconstruct the speed of the particle, for instance the “line fit” provides a speed. For a vanilla muon track this should not be too different from c, for shower-like events it is probably more compatible with zero. There are also more elaborate reconstruction methods that are used in for analyses searching for events triggered by magnetic monopoles, which may (or may not) travel at speeds less than c.

IsCascade

A Cascade interaction in the detector would look like a blob instead of a track. An example of an interaction that produces a blob is a neutrino neutral current Deep Inelastic Scattering in the ice. This will produce a shower of hadrons (many of those are Cascades) in the final state of the interaction.

An I3Particle can be set as a cascade via ParticleShape or ParticleType in the constructor of I3Particle class. A description of ParticleShape and ParticleType is given below of this documentation.

  • The ParticleShape options to set an I3Particle as a Cascade are: Cascade or CascadeSegment.

  • Regarding to the type: EPlus, EMinus, Brems, DeltaE, PairProd, NuclInt, Hadrons, PiPlus or PiMinus. All these type of particles are cascades.

  • Furthermore, a ParticleShape set as Primary and with a type of ParticleType as PPlus, PMinus, IsNucleus or Gamma will be a Cascade.

Below an example of how to set the type and other properties of an I3Particle in python:

from icecube import dataclasses

particle= dataclasses.I3Particle()
particle.pos= dataclasses.I3Position(2.5e6,4.8e6,-1.7e6)
particle.dir= dataclasses.I3Direction(0.8,1.1)
particle.type= dataclasses.I3Particle.PiPlus
particle.energy= 8.6e2

If the user makes: print particle all the information of the particle will be displayed. Furthermore, the user will obtain a True value by executing particle.is_cascade. This value is True since a PiPlus is defined as a Cascade.

The user can also set the shape of an I3Particle through particle.shape= dataclasses.I3Particle.Cascade.

Integer data members

Type

In simulation we know exactly which type a particle has: a neutrino, a muon, a calcium nucleus, etcetera. The Particle Data Group made a list of all known particles and gave them a number. For particles in that list we use their PDG number (note: in the past we used a different convention, but if you always use an enum to specify a type and not a literal integer then you should be able to stay happily oblivious of that detail). There are some “particle” types that we use in our simulation that do not exist in the PDG list. Frequently used types include PPlus (proton), Fe56Nucleus (iron), MuPlus and MuMinus (muons), and NuE and NuEBar (electron neutrinos). Some of these are particle types that have not yet been observed in experiments, such as monopoles and staus. Others have to do with shortcuts in our simulation, such as Hadrons for a hadronic showers and Pair, Brems and NuclInt for the various kinds of stochastic interactions. And finally there are particle type values that do not describe an actual particle but a calibration device, such as a laser or a flasher in calibration-related simulations.

Our current reconstruction algorithms do not identify any type, so the type data member is usually left “UnSet”. Even for infinite tracks the type is left as ‘unknown’. If the fit is good then it might be a muon (MuMinus), but it could also be an anti-muon (MuPlus) or a bundle of several muons (we don’t have a type for that).

Shape

For reconstructions it is more meaningful to assign a “shape” than a “type”. The shape data member is also used for simulated particles, but not very consistently. The shape can take the following values:

  • Null: Not set, not defined, not relevant.

  • Primary: Could be used for MC primaries, but e.g. in CORSIKA this needs to be fixed.

  • TopShower: Cosmic ray air shower reconstructed from IceTop data.

  • Cascade: A shower in the ice, reconstructed with in-ice data

  • InfiniteTrack: A through-going track (usually a muon).

  • StartingTrack: A starting track (supposedly due to a CC numu/nutau interaction).

  • StoppingTrack: A stopping track (supposedly due to ranged out muon or tau).

  • ContainedTrack: A track starting and stopping within the in-ice array.

  • MCTrack: simulated tracks [1]

  • Dark: tracks that do not emit any light [2]

Location

For fast selection of particles in an I3MCTree it is useful to have a rough indicator of “where the particle is”, because the position alone leaves that sometimes ambiguous. Values are Anywhere, IceTop, InIce and InActiveVolume. If KM3NET gets built on multiple sites, we could consider to add enum values to specify those sites.

Also this data member is not used very consistently. For instance, simulated neutrinos may have their interaction vertex somewhere in the atmosphere above Arizona and still be classified as InIce.

NB: This will be deprecated in future releases and no longer used.

Fit status

The fit status is an integer type data member. The possible values are defined by the FitStatus enum type:

  • NotSet: particle object is not the result of a fit (e.g. if it is a MC “true” particle)

  • OK: according to the fit algorithm the fit was successful

  • InsufficientHits: there was not enough data to do a fit

  • FailedToConverge: indicates that a likelihood maximization fit failed to find a maximum

  • MissingSeed: the fit algorithm requires a seed which was not available for the current event

  • InsufficientQuality: the fit succeeded but according to the algorithm it is not good enough to use

  • GeneralFailure: something else went wrong in the fit

You can in principle often use the integer values for these status values, but that makes your code harder to read. If you use the actual enum constants (e.g. p.fit_status == dataclasses.I3Particle.OK in python) then you and your colleages do not need to memorize the meanings of those numbers.

Major and Minor ID

In icetray, the “ID” of a particle object is a pair of (integer) numbers that serve as its unique identifier. No two particles, even two particles in different data sets, can have the same identifier. This is important because simulation is not done in one single process, but instead in stages, where particle generation and propagation can happen on different machines. So if we simply assigned ‘0’ to the first particle created in a generator process, its ID would clash with the first particle created by the propagator, both of which will be added to the same I3MCTree.

  • Major ID - This is a hash of the time (of I3Particle object creation), process ID, and hostname. So two particles created at the same time on the same machine, but in different processes will have different major IDs.

  • Minor ID - This is meant to provide the bone-headed (and commonly used) method of particle ID creation. The first particle created is ‘0’ and the next ‘1’, and so on.

This numerical particle ID should not be confused with term “PID” that is commonly used in other HEP experiments to denote the (process of determining the) type of a detected particle.

For instance, in simulation data this ID can be used to store maps that give for each simulated particle a list of PMT pulses (I3MCPulse) that were caused by photoelectrons due to Cherenkov photons emitted that particle. These maps only store the particle ID numbers, which can then be used (by an inquisitive student) to find the corresponding particle back in the I3MCTree, if desired.

In simulated data, a primary and its associated secondaries can have the same major ID but different minor ID. This is used mainly for storage in the I3MCTree.

This is likely not widely (if at all) used in experimental data.

NB : Copying particles - When I3Particles are copied they preserve their particle IDs. This is intentional. I3Particles are copied when added to STL containers. It would defeat the purpose of having a particle ID if it changed each time the particle was added to a container. If you want to create a new particle with all the same properties of another particle, the method you’re looking for is I3Particle::Clone.

Use case overview

(Note: this table does not contain all use cases.)

Use case

Shape

Location

Length

Meaning of position

CR primary

Primary

Anywhere

NAN

interaction vertex (cosmic ray gets destroyed) [3]

Neutrino primary

Null

Anywhere?

>0

interaction vertex (neutrino gets destroyed) [4]

Secondary muon

Null

InIce

>0

interaction vertex (muon gets created)

Secondary (other)

Null

InIce or IceTop

0

shower location (secondary gets created)

Reconstructed cascade

Cascade

InIce

0

shower location (possibly: neutrino interaction vertex)

Reconstructed stopping track

Stopping

InIce

NAN

end point (possibly: muon/tau decay vertex)

Reconstructed starting track

Starting

InIce

NAN

starting point (possibly: neutrino interaction vertex)

Reconstructed contained track

Contained

InIce

>0

starting point (possibly: neutrino interaction vertex)

Reconstructed through-going muon

InfiniteTrack

InIce

NAN

random point on the track [5]

Conventions and recommendations

  • Like most things in icetray, if you want to specify the I3Particle data members in specific units, you should use I3Units, e.g. (in python):

    p=dataclasses.I3Particle() p.dir.zenith = 42.0 * I3Units.degree print(“The zenith angle is %.1f degrees” % (p.dir.zenith/I3Units.zenith))

  • Most data members (all, except the major and minor ID) have a default initialization value that indicates that it is not yet set. For floating point data members this is NAN (Not a Number, TODO: add link), for the enum data members it is unknown (type), Null (shape), NotSet (fit status) and Anywhere (location).

  • Never use the explicit numerical values of the enum types. First, the enum constants have descriptive names, so they are much more informative. Secondly, it may occasionally happen that the definition of the enum changes (e.g. the type enum changed from the AMANDA/rdmc convention to the PDG convention). The icetray versioning system guarantees that old data is converted correctly when you read it in with new software but it will not update the literal integer constants in your code.

History

The current documentation attempts to provide a correct and more or less complete description of the current state (winter 2015) of I3Particle. In older data (many years older) you may discover that e.g. some conventions were different or some data members or enum values were not yet in use, but the basics have been relatively stable.

The first version of the I3Particle class was added to icetray in 2005. It was at least partly inspired by the mtrack struct in the rdmc library that was the foundation of the early AMANDA data processing and analysis software. In the early stages of icetray development (2004) we first tried to give literally every kind of simulated particle and reconstructed phenomenon (track, shower, double bang, monopole, elephantino) its own special dedicated class, but things that were common (like position or direction) would always have to represented in the same way. The result was very template-heavy and impressive but hard to work with. That’s why we went back to the single simplistic 8-dimensional (x, y, z, zenith, azimuth, length, energy) AMANDA solution (well, adding a 9th: speed), with the type and shape given by enum data members.

Some elements of I3Particle were added only a few years ago. The versioning system of icetray classes helps to make this backwards compatible, in the sense that if you read older data with newer software, the software deals with the version mismatches and you should not even notice that there was a version mismatch. Still, if you ever need to process older data and you run into strange problems, it may be useful to know which items are new and why they were added/changed:

  • Particle ID: this used to be a single number, but in 20XY this was replaced by a “major” and “minor” ID, to guarantee unique IDs between all particles created in mass production.

  • Particle types: we have been using a set of enum values that were invented in the AMANDA days. This reinvented wheel was eliminated by changing to an enum system mostly based on the PDG (Particle Data Group) list of elementary particles and nuclei.

  • The LocationType is relatively new and will be going away soon. This was meant to make it easier for analysts to pull the muon, for example, out of the I3MCTree. It’s proved to cause more problems than it solves and a more sophisticated method is in the works.