HDF5 output structure

The output of a NuRadioMC simulation is saved in the HDF5 file format, as well as (optionally) in .nur files. The data structure of .nur files is explained here. This page outlines the structure of the HDF5 files v3.0. Find the structure of v2.2 here.

Opening the HDF5 file

The HDF5 file can be opened using the h5py module:

import h5py

f = h5py.File("/path/to/hdf5_file", mode='r')
attributes = f.attrs

(...)
f.close()

If you have many HDF5 files, for example because you ran a simulation parallelized over multiple energy bins, NuRadioMC contains a convenience function to correctly merge these files - see here for instructions.

What’s behind the HDF5 files

The hdf5 file is created in NuRadioMC/simulation/simulation.py A list of vertices with different arrival direction (zenith and azimuth) and energy is provided by the event generator. Starting from the vertex, several sub-showers are created along the track. These are not simulated, but the electric field per sub-shower is provided. Sub-showers that happen within a certain time interval arrive at the antenna simultaneous and interfere constructively, therefore, they are summed up.

The event_group_id is the same for all showers that follow the same first interaction. The shower_id is unique for every shower. Shower which interfere constructively are combined into one event and have the same event_id starting from 0.

HDF5 structure

The HDF5 files can be thought of as a structured dictionary:

• The top level attributes, which can be accessed through f.attrs, contain some top-level information about the simulation.

• The individual keys contain some properties (energy, vertex, …) for each stored event or shower.

• Finally, the station_<station_id> key contains slightly more detailed information (triggers, propagation times, amplitudes…) at the level of individual channels for each station. Each station group has its own attributes (f[station_<station_id>].attrs)

HDF5 file attributes

The top-level attributes can be accessed using f.attrs. These contain:

HDF5 attributes

Key	Description
NuRadioMC_EvtGen_version NuRadioMC_EvtGen_version_hash NuRadioMC_version NuRadioMC_version_hash	Versions of the generator/framework as integer as hash
Emin Emax	Define energy range for neutrino energies
phimax phimin	Define azimuth range for incoming neutrino directions
thetamax thetamin	Define zenith range for incoming neutrino directions
flavors	A list of particle flavors that were simulated, using the PDG convention.
n_events	Total number of generated/simulated events(including those that did not trigger)
fiducial_xmax fiducial_xmin fiducial_ymax fiducial_ymin fiducial_zmax fiducial_zmin / fiducial_rmax fiducial_rmin fiducial_zmax fiducial_zmin	Specify the simulated qubic/cylindrical fiducial volume. An event has to produce an interaction within this volume. However, in case of a muon or tau CC interaction the first interaction can occur outside
rmax rmin zmax zmin / xmax xmin ymax ymin zmax zmin	Specify the qubic/cylindrical volume in which neutrino interactions are generated
volume	Volume of the above specified volume
area	Surface area of the above specified volume
start_event_id	event_id of the first event in the file
trigger_names	List of the names of the different triggers simulated
Tnoise	(explicit) noise temperature used in simulation
n_samples	Samples of the to-be generated antenna signals
config	The (yaml-style) config file used for the simulation
deposited
detector	The (json-format) detector description used for the simulation
dt	The time resolution, i.e. the inverse of the sampling rate used for the simulation. This is not necessarily the same as the sampling rate of the simulated channels!

The station-level attributes can be accessed using f[station_<station_id>].attrs. The first two attributes Vrms and bandwidth also exist on the top-level and refer to the corresponding to the first station/channel pair.

HDF5 station attributes

Key	Description
Vrms	RMS of the voltage used as thermal noise floor \(v_{n} = (k_{B} \, R \, T \, \Delta f) ^ {0.5}\). See the relevant section “Noise voltage and power” in this wiki article (last two equations). Determine from Tnoise and bandwidth (see below).
bandwidth	Bandwidth is above equation. Calculated as the integral over the simulated filter response (filt) squared: \(\Delta f = np.trapz(np.abs(filt) ** 2, ff)\).
antenna_positions	Relative position of all simulated antennas (channels)

HDF5 file contents

The HDF5 file contains the following items. Listed are the key and the shape of each HDF5 dataset, where n_events is the number of events stored in the file and n_showers is the number of showers (which may be larger than the number of events), and n_triggers is the number of different triggers simulated. Each “row” correspond to a particle shower which can produce radio emission.

HDF5 items

Key	Shape	Description
event_group_ids	(n_showers)	Specifies the event id to which the corresponding shower belongs (n_events = len(unique(event_group_ids))))
xx yy zz	(n_showers)	Specifying coordinates of interaction vertices
vertex_times	(n_showers)	Time at the interaction vertex. The neutrino interaction (= first interaction) is defined as time 0
azimuths zeniths	(n_showers)	Angle Specifying the neutrino incoming direction (azimuths = 0 points east)
energies	(n_showers)	Energy of the parent particle of a shower. This is typically the energy of the neutrino (for showers produced at the first interaction: all flavor NC, electron CC interactions) or the energy of a muon or tau lepton when those are producing secondary energy losses
shower_energies	(n_showers)	Energy of the shower which is used to determine the radio emission
flavors	(n_showers)	Same as above (the parent of an electromagnetic cascade in an electron CC interaction is the neutrino)
inelasticity	(n_showers)	Inelasticity of the first interaction
interaction_type	(n_showers)	Interaction type producing the shower (for the first interaction that can be “nc” or “cc”)
multiple_triggers	(n_showers, n_triggers)	Information which exact trigger fired each shower. The different triggers are specified in the attributes (f.attrs["triggers"]). The order of f.attrs["triggers"] matches that in multiple_triggers
triggered	(n_showers)	A boolean; True if any trigger fired for this shower, False otherwise
trigger_times	(n_showers, n_triggers)	The trigger times (relative to the first interaction) at which each shower triggered. If there are multiple stations, this will be the earliest trigger time.
n_interaction	(n_showers)	Hierarchical counter for the number of showers per event (also accounts for showers which did not trigger and might not be saved)
shower_ids	(n_showers)	Hierarchical counter for the number of triggered showers
shower_realization_ARZ	(n_showers)	Which realization from the ARZ shower library was used for each shower (only if ARZ was used for signal generation).
shower_type	(n_showers)	Type of the shower (so far we only have “em” and “had”)
weights	(n_showers)	Weight for the probability that the neutrino reached the interaction vertex taking into account the attenuation from the earth (Does not include interaction probability in the volume)

Station data

In addition, the HDF5 file contains a key for each station in the simulation. The station contains more detailed information for each station. Some parameters are per event and some parameters are per shower. See https://doi.org/10.22323/1.395.1231 for a description of how showers relate to events. m_events and m_showers refer to the number of events and showers that triggered the station. NOTE: The simple table structure of hdf5 files can not capture the complex relation between events and showers in all cases. Some fields can be ambiguous (e.g. trigger_times that only lists the last trigger that a shower generated). For more advanced analyses, please use the *.nur files. The event_group_id is the same as in the global dictionary. Therefore you can check for one event with an event_group_id which stations contain the same event_group_id and retrieve the information, which station triggered, with which amplitude, etc. The same approach works for shower_id.

HDF5 station items

Key	Shape	Description
event_group_ids	(m_events)	The event group ids of the triggered events in the selected station
event_group_id_per_shower	(m_showers)	The event group id of every shower that triggered the selected station
event_ids	(m_events)	The event ids of each event that triggered in that station for every event group id. These are unique only within each separate event group, and start from 0.
event_id_per_shower	(m_showers)	The event ids of each event that triggered in that station. This one is for every shower
focusing_factor	(m_showers, n_channels, n_ray_tracing_solutions)
launch_vectors	(m_showers, n_channels, n_ray_tracing_solutions, 3)	3D (Cartesian) coordinates of the launch vector of each ray tracing solution, per shower and channel.
max_amp_shower_and_ray	(m_showers, n_channels, n_ray_tracing_solutions)	Maximum amplitude per shower, channel and ray tracing solution.
maximum_amplitudes	(m_events, n_channels)	Maximum amplitude per event and channel
maximum_amplitudes_envelope	(m_events, n_channels)	Maximum amplitude of the hilbert envelope for each event and channel
multiple_triggers	(m_showers, n_triggers)	A boolean array that specifies if a shower contributed to an event that fulfills a certain trigger. The index of the trigger can be translated to the trigger name via the attribute trigger_names.
multiple_triggers_per_event	(m_events, n_triggers)	A boolean array that specifies if each event fulfilled a certain trigger. The index of the trigger can be translated to the trigger name via the attribute trigger_names.
polarization	(m_showers, n_channels, n_ray_tracing_solutions, 3)	3D coordinates of the polarization vector at the antenna in cartesian coordinates. (The receive vector (which is opposite to the propagation direction) was used to rotate from spherical/on-sky coordinates to cartesian coordinates). The polarization vector does not include any propagation effects that could change the polarization, such as different reflectivities at the surface for the p and s polarization component.
ray_tracing_C0	(m_showers, n_channels, n_ray_tracing_solutions)	One of two parameters specifying the analytic ray tracing solution. Can be used to retrieve the solutions without having to re-run the ray tracer.
ray_tracing_C1	(m_showers, n_channels, n_ray_tracing_solutions)	One of two parameters specifying the analytic ray tracing solution. Can be used to retrieve the solutions without having to re-run the ray tracer.
ray_tracing_reflection	(m_showers, n_channels, n_ray_tracing_solutions)
ray_tracing_reflection_case	(m_showers, n_channels, n_ray_tracing_solutions)
ray_tracing_solution_type	(m_showers, n_channels, n_ray_tracing_solutions)
receive_vectors	(m_showers, n_channels, n_ray_tracing_solutions, 3)	3D (Cartesian) coordinates of the receive vector of each ray tracing solution, per shower and channel.
shower_id	(m_showers)	The Shower ids of showers that triggered the selected station
time_shower_and_ray	(m_showers, n_channels, n_ray_tracing_solutions)
travel_distances	(m_showers, n_channels, n_ray_tracing_solutions)	The distance travelled by each ray tracing solution to a specific channel
travel_times	(m_showers, n_channels, n_ray_tracing_solutions)	The time travelled by each ray tracing solution to a specific channel
triggered	(m_showers)	Whether each shower contributed to an event that satisfied any trigger condition
triggered_per_event	(m_events)	Whether each event fulfilled any trigger condition.
trigger_times	(m_showers, n_triggers)	The trigger times for each shower and trigger. IMPORTANT: A shower can potentially generate multiple events. Then this field is ambiguous, as only a single trigger time per shower can be saved. In that case, the latest trigger time is saved into this field.
trigger_times_per_event	(m_events, n_triggers)	The trigger times per event.