To run anything more than examples from our suite, you will need to be able to produce your own initial conditions for SWIFT. We use the same initial conditions format as the popular GADGET-2 code, which uses HDF5 for its type 3 format. Note that we do not support the GADGET-2 types 1 and 2 formats.
One crucial difference is that whilst GADGET-2 can have initial conditions split over many files SWIFT only supports initial conditions in one single file. ICs split over multiple files cannot be read by SWIFT. See the “ICs split over multiple files” section below for possible solutions. In GADGET-2 having multiple files allows multiple ones to be read in parallel and is the only way the code can handle more than 2^31 particles. This limitation is not in place in SWIFT. A single file can contain any number of particles (well… up to 2^64…) and the file is read in parallel by HDF5 when running on more than one compute node.
The original GADGET-2 file format only contains 2 types of particles: gas particles and 5 sorts of collision-less particles that allow users to run with 5 separate particle masses and softenings. In SWIFT, we expand on this by using two of these types for stars and black holes.
As the original documentation for the GADGET-2 initial conditions format is
quite sparse, we lay out here all of the necessary components. If you are
generating your initial conditions from python, we recommend you use the h5py
package. We provide a writing wrapper for this for our initial conditions in
You can find out more about the HDF5 format on their webpages.
Structure of the File
There are several groups that contain ‘auxiliary’ information, such as
Header. Particle data is placed in separate groups depending of the type of
the particles. Some types are currently ignored by SWIFT but are kept in the
file format for compatibility reasons.
HDF5 Group Name
Physical Particle Type
Background Dark Matter
Neutrino Dark Matter
The last column in the table gives the
enum value from
corresponding to a given entry in the files.
Note that the only particles that have hydrodynamical forces calculated
between them are those in
PartType0. The background dark matter
particles are used for zoom-in simulations and can have different masses
(and as a consequence softening length) within the
There are several necessary components (in particular header information) in a
SWIFT initial conditions file. Again, we recommend that you use the
/Header/ group, the following attributes are required:
Dimension, an integer indicating the dimensionality of the ICs (1,2 or 3). Note that this parameter is an addition to the GADGET-2 format and will be ignored by GADGET. SWIFT will use this value to verify that the dimensionality of the code matches the ICs. If this parameter is not provided, it defaults to 3.
BoxSize, a floating point number or N-dimensional (usually 3) array that describes the size of the box. If only one number is provided (as per the GADGET-2 standard) then the box is assumed have the same size along all the axis. In cosmological runs, this is the comoving box-size expressed in the units specified in the
/Unitsgroup (see below). Note that, unlike GADGET, we express all quantities in “h-free” units. So that, for instance, we express the box side-length in
NumPart_Total, a length 6 array of integers that tells the code how many particles of each type are in the initial conditions file. Unlike traditional GADGET-2 files, these can be >2^31.
NumPart_Total_HighWord, a historical length-6 array that tells the code the number of high word particles in the initial conditions there are. If you are unsure, just set this to
[0, 0, 0, 0, 0, 0]. This does have to be present but can be a set of 0s unless you have more than 2^31 particles and want to be fully compliant with GADGET-2. Note that, as SWIFT supports
NumPart_Totalto be >2^31, the use of
NumPart_Total_HighWordis only here for compatibility reasons.
Flag_Entropy_ICs, a historical value that tells the code if you have included entropy or internal energy values in your initial conditions files. Acceptable values are 0 or 1. We recommend using internal energies over entropy in the ICs and hence have this flag set to 0.
You may want to include the following for backwards-compatibility with many GADGET-2 based analysis programs:
MassTable, an array of length 6 which gives the masses of each particle type. SWIFT ignores this and uses the individual particle masses, but some programs will crash if it is not included.
NumPart_ThisFile, a length 6 array of integers describing the number of particles in this file. If you have followed the above advice, this will be exactly the same as the
NumPart_Totalarray. As SWIFT only uses ICs contained in a single file, this is not necessary for SWIFT-only ICs.
NumFilesPerSnapshot, again a historical integer value that tells the code how many files there are per snapshot. You will probably want to set this to 1. If this field is present in a SWIFT IC file and has a value different from 1, the code will return an error message.
Time, time of the start of the simulation in internal units or expressed as a scale-factor for cosmological runs. SWIFT ignores this and reads it from the parameter file, behaviour that matches the GADGET-2 code. Note that SWIFT writes the current time since the Big Bang, not scale-factor, to this variable in snapshots.
Redshift, the redshift at the start of the simulation. SWIFT checks this (if present) against
a_beginin the parameter file at the start of cosmological runs. Note that we explicitly do not compare the
Timevariable due to its ambiguous meaning.
Now for the interesting part! You can include particle data groups for each
individual particle type (e.g.
/PartType0/) that have the following datasets:
Coordinates, an array of shape (N, 3) where N is the number of particles of that type, that are the cartesian co-ordinates of the particles. Co-ordinates must be within the box so, in the case of a cube within [0, L)^3 where L is the side-length of the simulation volume. In the case of cosmological simulations, these are the co-moving positions.
Velocities, an array of shape (N, 3) that is the cartesian velocities of the particles. When running cosmological simulations, these are the peculiar velocities. Note that this is different from GADGET which uses peculiar velocities divided by
sqrt(a)(see below for a fix).
ParticleIDs, an array of length N that are unique identifying numbers for each particle. Note that these have to be unique to a particle, and cannot be the same even between particle types. The IDs must be >= 0. Negative IDs will be rejected by the code. Note, however, that if the parameters to remap the IDs upon startup is switched on (see Initial Conditions), the IDs can be omitted entirely from the ICs.
Masses, an array of length N that gives the masses of the particles.
PartType0 (i.e. particles that interact through hydro-dynamics), you will
need the following auxiliary items:
SmoothingLength, the smoothing lengths of the particles. These will be tidied up a bit, but it is best if you provide accurate numbers. In cosmological runs, these are the co-moving smoothing lengths.
InternalEnergy, an array of length N that gives the internal energies per unit mass of the particles. If the hydro-scheme used in the code is based on another thermodynamical quantity (entropy or total energy, etc.), the conversion will happen inside the code. In cosmological runs, this is the physical internal energy per unit mass. This has the dimension of velocity squared.
Note that for cosmological runs, all quantities have to be expressed in “h-free”
dimensions. This means
Mpc and not
Mpc/h for instance. If the ICs have
been generated for GADGET (where h-full values are expected), the parameter
InitialConditions:cleanup_h_factors can be set to
1 in the
Parameter Files to make SWIFT convert the quantities read in to
h-free quantities. Switching this parameter on will also affect the box size
read from the
/Header/ group (see above).
Similarly, GADGET cosmological ICs have traditionally used velocities expressed
as peculiar velocities divided by
sqrt(a). This can be undone by switching
on the parameter
InitialConditions:cleanup_velocity_factors in the
/Units/ HDF5 group, you cans specify what units your initial conditions are
in. If this group is not present, the code assumes that you are using the same
units for your initial conditions as in your Parameter Files
(i.e. as the internal units system used by the code), but it is best to include
them to be on the safe side. You will need:
Unit length in cgs (U_L)
Unit mass in cgs (U_M)
Unit time in cgs (U_t)
Unit current in cgs (U_I)
Unit temperature in cgs (U_T)
These are all floating point numbers. Note that we specify the time units and not the velocity units.
If the units specified in the initial conditions are different from the internal
units (specified in the parameter file), SWIFT will perform a conversion of all
the quantities when reading in the ICs. This includes a conversion of the box
size read from the
You should have an HDF5 file with the following structure:
Header/ BoxSize=[x, y, z] Flag_Entropy_ICs=0 NumPart_Total=[0, 1, 0, 0, 4, 5] NumPart_Total_HighWord=[0, 0, 0, 0, 0, 0] Units/ Unit current in cgs (U_I)=1.0 Unit length in cgs (U_L)=1.0 Unit mass in cgs (U_M)=1.0 Unit temperature in cgs (U_T)=1.0 Unit time in cgs (U_t)=1.0 PartType0/ Coordinates=[[x, y, z]] Velocities=[[vx, vy, vz]] ParticleIDs=[...] Masses=[...] InternalEnergy=[...] SmoothingLength=[...] PartType1/ Coordinates=[[x, y, z]] Velocities=[[vx, vy, vz]] ParticleIDs=[...] Masses=[...]
ICs split over multiple files
A basic script
tools/combine_ics.py is provided to merge basic GADGET-2
initial conditions split into multiple files into one single valid file. This
script can handle simple HDF5 files (GADGET-2 type 3 ICs) that follow the format
described above but split over multiple files.
The script can also convert ICs using a
MassTable and create the
corresponding particle fields. Note that additional fields present in ICs beyond
the simple GADGET-2 specification will not be merged.
One additional option is to compress the fields in the files using HDF5’s gzip compression. This is very effective for the fields such as masses or particle IDs which are very similar. A checksum filter is also applied in all cases to help with data curation.
We caution that this script is very basic and should only be used with great caution.