KAHIBARO
38.5. Applications
Table of Contents
Accelerator simulation
Phase space is particularly powerful when you need to separate a complex accelerator or linac head from the rest of your simulation while still preserving a realistic particle field. In GATE, you typically use a phase-space actor to record all particles crossing a scoring plane placed at a convenient location, for example at the exit of an accelerator, linac head, or collimation system.
In an accelerator context, a phase-space file stores, for each recorded particle, its position, direction, kinetic energy, particle type, time, and possibly weight. This snapshot represents the full state of the beam at that plane. You can then convert this phase-space into a phase-space source and reuse it in many downstream simulations.
One common workflow in medical applications is to model an accelerator head, such as a medical linac for external beam radiotherapy, only once. You run a detailed simulation that includes target, primary collimator, flattening filter or filter-free configuration, multileaf collimator, and shielding. At the exit of the head, just before the patient or phantom, you place a phase-space actor. After generating a sufficiently large number of primary histories, you obtain a large phase-space file that describes the clinical photon or electron beam emerging from the machine.
This recorded accelerator output can then be used as an input source for many different treatment field configurations, water phantoms, or patient geometries, without re-simulating the entire head each time. You can change geometry, scoring setups, and dose actors downstream while keeping the same accelerator phase space. In this way, phase-space use supports both detailed accelerator design studies and efficient clinical beam characterization in GATE.
Beam modeling
Beam modeling focuses on creating a realistic and reusable description of the radiation field, such as a therapeutic photon or proton beam, or an imaging x ray beam. Phase-space data provide a flexible representation of such beams that is more general than simple analytical descriptions like Gaussian or pencil beams.
With phase-space based beam modeling, you treat the phase-space file as a numerical model of the beam. All relevant beam parameters are implicitly contained: spatial distribution, angular spread, energy spectrum, and even correlations between these quantities. If the underlying accelerator simulation has been validated against measurements, then reusing the resulting phase-space as a source is an effective way to propagate that realism into other studies.
You can construct phase-space based models in several steps. First, record a phase space at a physically meaningful plane. Next, analyze the distribution in energy, position, and direction to verify that it reproduces measured depth-dose curves, profiles, and fluence patterns. If needed, you can apply selection or reweighting to the phase-space file to tune specific beam characteristics, for example by filtering certain energy ranges or restricting the field size in position space.
For imaging applications, such as CT or cone-beam CT, a phase space recorded at the x ray tube exit or after bowtie filters can become a standard beam model used for patient dose calculations or image quality studies. For proton therapy, a phase space recorded at a nozzle exit can model the complex mixed field of primary protons, scattered protons, and secondary particles.
Once you are satisfied with the beam model, you use it as a phase-space source in GATE. Instead of specifying simple source parameters, you instruct GATE to randomly sample particles from the stored phase-space events. Each sampled particle retains its original state, so spatial and angular correlations are preserved automatically. This approach allows you to perform detailed beam modeling without analytically describing each beam component.
A phase-space based beam model is only as accurate as the simulation and validation that created it. Always validate the phase-space beam against measurements such as depth-dose curves and lateral profiles before using it in clinical or research studies.
Performance optimization
Phase-space techniques can dramatically improve computational performance when used carefully. The key idea is to invest computational effort once, to produce a high quality phase-space at an intermediate plane, and then reuse it repeatedly in many downstream simulations.
Without phase-space reuse, each new geometry or dose calculation might require a full re-simulation of the accelerator, collimators, and upstream components. These regions often involve heavy materials and complex geometries that increase computation time per primary event. By recording a phase space just after these components, you effectively cache their effect on the beam, so later runs start from a realistic, already processed particle field.
This approach is particularly useful for parameter studies. For example, suppose you want to evaluate dose in many different phantoms or patient positions for the same accelerator field. You can run one long accelerator simulation to produce a large phase-space file, then launch many shorter simulations that use the same phase-space source but different geometries or scoring configurations. On clusters or HPC systems, each of these downstream simulations can run in parallel with minimal memory overhead, since complex upstream geometry is no longer needed.
Phase-space reuse also helps control statistical uncertainty. If you generate a very large number of particles in the original phase space, you can recycle these events multiple times, or sample them with particle weights to effectively increase the number of usable histories without rerunning the expensive upstream simulation. When doing this, you must be careful to account for correlations if you reuse the exact same events; in many workflows, you randomly subsample or shuffle the phase space to maintain good statistical properties.
Another performance advantage appears when tuning simulation setup. During early development you can work with a small phase-space file that still captures key beam features, which lets you test geometry, actors, and output quickly. Once you are confident in the setup, you can switch to a larger, more statistically rich phase space for production runs.
Finally, phase-space files can minimize I/O and memory when used appropriately. Instead of scoring many different detailed quantities inside the accelerator head for every study, you store only what you need at a single scoring plane. When analyzed carefully and validated, this targeted use of phase space enables you to balance accuracy, flexibility, and computational cost in complex GATE simulations.
Views: 10