Support Recombination and Scintillation-Quenching Models for LArTPC-based Experiments

[whokion]

Liquid Argon Time Projection Chambers (LArTPCs) experiments including DUNE, ICARUS, SBND, etc require scintillation light production models that incorporate electric-field–dependent (and temperature-dependent?) recombination, not only dE/dx quenching. The number of scintillation optical photons per step in LAr depends on both the intrinsic scintillation yield ($S \sim (dE/dx) / W_{ph}$ where $W_{ph} \approx 19.5 eV$) and a recombination factor ${R( \vec{\mathcal{E}}, dE/dx)}$ determined by the local drift field ($\vec{\mathcal{E}}$). These effects are essential for reproducing the experimentally observed anti-correlation between ionization charge and scintillation light.

Celeritas currently implements scintillation photon generation through ScintillationOffload where the mean number of optcal photons is calculated by a used provided material property (i.e., scintillation yield), but does not include:

1. Recombination-based scintillation suppression (field-dependent), and
2. Birks-like empirical saturation applied to liquid argon or other scintillators.

Because these models, particularly "Recombination parameteization" are used by LArTPC simulations (DUNE, ICARUS, MicroBooNE, SBND) at Fermilab, adding support for field-dependent recombination (and optional Birks quenching) would require improve to integrate optical physics of Celeritas into LArSoft or to support LArTPC simulations (related issues, #1548, #1344, #1345).

Motivation

LArTPC experiments require models that incorporate:

• Electric-field–dependent recombination, using a parameterized field map, ${\vec{\mathcal{E}}(x,y,z)}$
• Anti-correlated ionization (${Ar^{+}}$) and scintillation (${Ar^{*}}$) signals
• Suppression of scintillation light yield as a function of both dE/dx and ${\vec{\mathcal{E}}}$
• Optional Birks-law–type saturation for more general scintillator materials.

These effects are currently handled through ad-hoc, parameterized optical-photon models in LArTPC simulations. Native support in Celeritas would enable full, offloading GPU-accelerated optical photon generation and propagation within LArSoft.

Support common LArTPC recombination models

The request is to introduce built-in support for the standard recombination models used in LArTPC detectors, parameterized by
${dE/dx}$ and the local drift field ${\vec{\mathcal{E}}}$:

1. Birks-like recombination model: Used in ICARUS and earlier DUNE simulations
   $R(E, dE/dx) = \frac{A_B}{1 + k_B * \frac{dE/dx} {|\mathcal{E}|}}$​

• $k_B$ : Birks constant
• $A_B$: normalization parameter
• Incorporates suppression from both ionization density and electric field

2. Modified Box model: Used by DUNE, SBND, and MicroBooNE:
   $R(E, dE/dx) = \frac{\ln(\alpha + \beta * \frac{dE/dx} {|\mathcal{E}|}) } { \beta * \frac{dE/dx} {|\mathcal{E}|}}$

• $\alpha, \beta$: tuned parameters
• Produces more accurate quenching across the full dE/dx spectrum
• Reproduces measured anti-correlation between light and charge

3. User-supplied / configurable models
   Permit experiments to load parameters, field maps, or functional forms so Celeritas can support a broad range of detector configurations.

Requirements (based on discussion with Eric Snider/LArSoft coordinator)

1. Decouple optical photon generation from Ionization

• Separate the optical photon generation and propagation from the ionization (drift-electron) process.
• Maintain the correlation between these channels by dividing the total energy deposited in each simulation step into optical and ionization components.

2. Preserve the existing ionization workflow

• Keep the current ionization (charge-drift) workflow intact.
• Fork a parallel branch that uses Celeritas to offload optical-photon generation and propagation.

3. Generate optical photon hit data

• Ensure that the optical simulation produces hit data with OpDetBacktrackerRecord object for downstream digitization compatibility.

4. Use a two-stage processing workflow (preferred)

• First, perform particle tracking to generate SimEnergyDepositSD data for all steps inside the TPC volume.
• After the event completes, run the optical-photon simulation and charge-drift simulation based on the stored SimEnergyDeposit data.
• (If necessary, support a more efficient integrated workflow as an alternative.)

Work Plan (to be discussed)

1. Extend ScintillationOffload for Configurable Recombination and Yield Models

• Extend ScintillationOffload to support recombination models (Modified Box model, Elliptic Box recombination, and Birks), See larsim/IonizationScintillation/ISCalcCorrelated::CalcIonAndScint(...)
• Make all model parameters to be configurable, See lardataalg/DetectorInfo/DetectorPropertiesData
• Consider whether separate distributions for ionization electrons are required.

2. Provide an interface to access electric field values

• Implement a method or interface that retrieves the electric field (from map?) at a given simulation step

3. Define APIs / Data-Flow for Integration with LArSoft

• Add the necessary interfaces to pass data back to LArSoft and support decoupled workflows, i.e., APIs for transferring ionization-electron information to the charge-drift chain.
• Document the workflow for decoupling photon propagation (Celeritas) from electron drift (LArSoft), including sata structures produced at each stage, required data conversions (?), event-level synchronization and sequenc ing requirements

[sethrj]

@whokion Is this the physics implemented by NEST, which is used by LZ (#1345 (comment))?

[whokion]

Yes, it should also be related to issue #1345 (which I forgot to include). This issue focuses on adding more direct native API support in Celeritas for estimating the number of optical photons by the parameterized recommbination models, without involving models/interfaces of generating and transporting drifts electrons (LArSoft reconstruction module, NEST-like, Garfield, etc) which may be another potential extension, though it may fall outside the current scope of our development.