Provided examples¶

Calo-opt¶

This example is based on the existing calo_opt project. In order to run these examples, you need to pull the following container:

With docker:

docker pull jkiesele/minicalosim/latest

With singularity/ apptainer (useful if you lack root access to the machine, do not have docker installed or do not have permissions):
```
apptainer pull docker://jkiesele/minicalosim:latest
```

interface = CaloOptInterface()
interface.container_path = "</path/to/container.sif>"
interface.container_extra_flags = ""  # Extra apptainer settings
interface.verbose = True  # This flag enables / disables Geant4 logs

In essence, the CaloOptInterface creates a subprocess in which Geant4 runs:

def simulate(self, parameter_dict_path: str, sim_output_path: str):
    os.system(
        f"singularity exec {self.container_extra_flags} {self.container_path} python3 \
        examples/calo_opt/simulation.py {parameter_dict_path} {sim_output_path} {self.suppress_output}"
    )

Finally, run the example by calling:

python3 examples/<example.py>

Or add the root directory for AIDO to your $PYTHONPATH and call the examples from the examples folder directly.

If you get Errors of the type Unfulfilled dependencies at Runtime, check if the Simulation was able to run. The first debugging tool is to set interface.verbose=True and check the error logs. For example using a wrong path or not mounting a required filesystem into the container can all lead to a failed Task. The scheduler will then simply inform us that the next Task could not run.

Sampling calorimeter¶

In this example we simulate a sampling calorimeter composed of three layers of absorber material and three of active recording material. More detail can be found in arxiv:2502.02152 Section 6.

Simulation¶

The current examples has some hardcoded parts that set up the geometry of the detector.

Configurable parameters:
- Thickness of the layers
- Material of the layers
- Cost per unit of length
- Probability of sampling one material over another, e.g. the confidence of the model in that material choice.
Currently hardcoded settings in simulation.py but that can be changed in the code of the example:
- Number of layers: expects exactly six layers, three of them absorber, three scintillators.
- Names of the layers: either thickness_absorber_{i}or thickness_scintillator_{i} for $i \in {0, 1, 2}$.
- Lateral granularity: only one cell per layer perpendicularly to the beam axis
- Particles: always uses a 50-50 mix of pions and photons and only shoot one particle per event
- Sampled energy: uniform distribution between 1 and 20 GeV.
- Seed: have to set the seed for Geant4 manually to avoid duplicate events.
Fixed settings that cannot be configured at all
- Shape of the detector: there are only tiles of configurable sizes but no other shapes available
- Orientation: the beam impacts perpendicularly on the detector

For more information about the Simulation software used, consult the dedicated github page by J. Kieseler https://github.com/jkiesele/minicalosim.