A parametrized simulation code to simulate the response of ground detector arrays to Extensive Air Showers above 100 TeV
Listen
Radio episode about this paper
Transcript
Introduction to the show: ident: Astrophysics. The day in astrophysics: observations, models and what the data forces.
Vera: I'm Vera.
Jocelyn: And I'm Jocelyn, here with Subrahmanyan.
Vera: We're looking at the paper: "A parametrized simulation code to simulate the response of ground detector arrays to Extensive Air Showers above 100 TeV".
Jocelyn: Here's the gist of the paper. Surface arrays detecting the secondary particles of Extensive Air Showers (EAS) are one of the techniques used to detect gamma-rays from high-energy sources In this work, the authors present an alternative way to evaluate the performance-related quantities of a simulated array configuration through the parametrisation of the detector response This approach is presented to address the computational demands associated with full GEANT4-based simulations for very large fill factors, above 30-100 TeV The work details the implementation of a fast simulation code for WCD arrays through the use of a look-up table (LUT) to parametrize the detector response.
Vera: Parametrisation Methodology The core idea involves defining a hit as a particle touching the external walls of the detector The parametrisation of a tank response is modeled as a distribution describing the behavior of the tank as a function of the hit characteristics, which must be stored inside a LUT that the simulation will use to replicate the tank response.
Jocelyn: The generic approach for building this LUT involves discretizing the 7-dimensional parameter space—Particle type, Particle energy, (x,y,z) of the hit, and (θ,ϕ) angles of the particle direction—and simulating responses at these vertices.
Vera: For a specific detector like the SWGO double-layer tank example, only six variables are used to exploit cylindrical symmetry: Particle type (PID), Particle energy (E), Zenith angle of the direction of the particle (θ), Azimuth angle of the direction of the particle (ϕ), Distance of the particle hit from the central axis of the tank (dR), and Distance of the particle hit from the base of the tank (dH).
Paper summary: Vera: So we're looking at this paper now called "A parametrized simulation code to simulate the response of ground detector arrays to Extensive Air Showers above one hundred TeV" <ref:2611.16906#pg1,simulation code to simulate the response of ground detector arrays to Extensive>. Basically, they are trying to solve a problem where simulating how big arrays detect gamma rays is taking too long when you look at very high energy sources, like those above thirty or one hundred tera electron volts <ref:2611.16906#pg1>.
Jocelyn: Right. So the authors say the usual way involves two steps: first you generate the air shower using something like CORSIKA software, and then you use a full GEANT4 code to simulate how that hits the detector array. But for huge fill factors, that second step becomes too computationally expensive to do often enough.
Subrahmanyan: That makes sense from a cosmic perspective. If we want to properly test the gamma ray rejection power or the gamma to hadron separation in these experiments, we need tons of proton events, and simulating every single particle interaction in a full code takes an unreasonable amount of time.
Vera: Exactly. So instead of running that heavy GEANT4 simulation every time, this work proposes an alternative way to check the performance by parameterizing the detector response itself. They create a lookup table for how each detector component reacts to different particles and hit locations.
Jocelyn: What they claim is that by using this lookup table, you don't actually need the full GEANT4 simulation anymore when you're looking at an entire air shower event. You just sum up the response from all the secondary particles based on what that pre-calculated response table tells you.
Subrahmanyan: That shifts the computational burden from running complex tracking operations in GEANT4 to just looking up values in this new table, which sounds like it could make studying these high energy sources much more feasible for future experiments.
Vera: They detail how they build this lookup table by essentially simulating responses at specific points in a seven dimensional space. For a tank detector, that means they are mapping the behavior based on the particle type and energy, and where it hits—the x, y, z location of the hit—and also the direction of travel.
Paper summary: Jocelyn: They simplify this for things like the SWGO double-layer tank example by only using six variables because they exploit cylindrical symmetry. That means they are only looking at the particle type, its energy level, and some distances from a central axis and a base plane.
Subrahmanyan: Focusing on symmetry is smart when you’re dealing with large arrays. It limits the number of points you have to simulate in that seven dimensional space, which directly feeds into the speed of their approach. But I wonder how well those symmetries hold up for an actual air shower hitting a very complex geometry like a real array.
Vera: The paper says they model the response by fitting histograms to two main things: the number of photo electrons hitting the photomultiplier tubes, and the arrival time of that first photo electron at the upper PMT. For photo electrons, they often use a negative binomial distribution, but sometimes they need a combination of two negative binomials if there are multiple peaks in the histogram.
Jocelyn: And for arrival time, they use something called an exponentially modified Gaussian distribution. Again, if you have reflective walls at the tank boundaries, they model multiple peaks by combining two of those distributions. That’s how you account for bouncing particles inside a structure like that.
Subrahmanyan: So we're talking about modeling the physics of light detection—how many photons get registered and when they arrive—using statistical distributions derived from simulations at discrete points, which is a clever way to bypass the heavy tracking part.
Vera: They did validate this fast simulation code against the full GEANT4 approach using high energy proton events thrown randomly within a one thousand five hundred meter radius. And they found that for event level variables, like the average density of photo electrons, there was almost no bias in their fast simulation; only fluctuations up to about ten percent.
Jocelyn: That’s reassuring. So the agreement between the number of photo electrons calculated by their fast code and the full GEANT4 simulation is good for both energy ranges and configurations tested, showing that their method correctly reproduces how the detector actually behaves in those setups.
Paper summary: Subrahmanyan: And they also looked at something else, like the lateral distribution—how the light spreads across a detector—and found that its relative difference oscillates around zero for both layers and averages less than three percent. That suggests their parametrization is capturing the spatial spread reasonably well.
Vera: The real payoff they highlight is in execution time reduction. They tested this by simulating fixed energy, fixed direction gamma showers at different spots in the array and found that the gain in execution time over the full simulation was bigger than twenty times, and sometimes over fifty if you only look at events where the core of the shower actually enters the array.
Jocelyn: So when you consider how many particles actually enter your detector, this fast method gets even better; it shows a stronger reduction in simulation time when you're looking near the center of the array. That means for large arrays, this approach could be significantly faster for generating performance metrics.
Subrahmanyan: This speedup is important because it means we can run more simulations to test different array designs or different energy regimes that were previously out of reach due to the computational demands of tracking every single particle interaction in a full code.
Vera: So, to sum up this paper, they presented a way to evaluate performance quantities for ground detector arrays against extensive air showers above one hundred TeV without needing the massive computational power of full GEANT4 simulations by using a parametrized simulation code based on lookup tables <ref:2611.16906#pg1,extensive air showers above 100 TeV>.
Jocelyn: The authors are Andrea Chiavassaa and Andrea Negroa from the University of Turin, and this work is about making the analysis faster for experiments like LHAASO or SWGO that study ultra high energy gamma rays.
Subrahmanyan: The implication here is that it provides a practical tool to help future cosmic or gamma ray experiments design their setups without getting completely bogged down by the simulation time needed to characterize the detector response under extreme conditions.
Conclusion: Vera: So this paper, "A parametrized simulation code to simulate the response of ground detector arrays to Extensive Air Showers above one hundred TeV," is all about making simulating those massive arrays much faster than using full GEANT4 simulations.
Jocelyn: It’s basically trading deep physics tracking for a smart lookup table method. They are creating a way to check performance without running the super heavy simulations every single time.
Subrahmanyan: The authors, Andrea Chiavassaa and Andrea Negroa, developed this fast simulation code to tackle the computational demands of very large detector fill factors in high-energy setups.
Vera: What they’re doing is using a parametrized response for each tank or detector component based on where the particle hits it and what kind of particle it is.
Jocelyn: They discretize seven dimensions—like particle type, energy, and the hit coordinates—to build this lookup table that describes how the detector responds.
Subrahmanyan: The key insight is modeling things like photo-electron counts and arrival times using statistical distributions, like negative binomials for counts and exponentially modified Gaussians for timing.
Vera: They validated this fast code against full GEANT4 simulations with high energy proton events thrown uniformly in a thousand meters, and the results show good agreement on event level variables.
Jocelyn: The results were pretty solid; fluctuations in photo-electron density were small, and the lateral distribution spread stayed consistent across both layers.
Subrahmanyan: And they found that the time saved is significant, sometimes over twenty times faster than the full simulation when you only consider events where the core of the shower enters your array.
Vera: It means we can do a lot more simulations to test different array designs or energy ranges without waiting weeks for GEANT4 to finish.
Jocelyn: This approach offers a way forward for designing future experiments that need to characterize detector response under extreme conditions quickly.
Subrahmanyan: So, the core idea is using this fast simulation code as a reliable compromise when the full simulation just isn't feasible for large-scale studies.
Andrea Chiavassaa, Andrea Negroa
University of Turin · INFN-Turin
astro-ph.IM
Submitted: 2026-11-12
Updated: 2026-11-12
Comments: 28 pages, 14 figures, submitted to Nucl. Instrum. Meth. A
Journal ref: Nucl. Instrum. Meth. A 1093 (2027) 171921
DOI: 10.1016/j.nima.2026.171921
Project page: https://astronomicalpark.cl/2
License: http://creativecommons.org/licenses/by-nc-nd/4.0/
Importance score: 80/100
The gist: Surface arrays detecting the secondary particles of Extensive Air Showers (EAS) are one of the techniques used to detect gamma-rays from high-energy sources In this work…
Key concepts
- Parametrisation
- This method models how a detector component, like a tank, reacts to an incoming particle. Instead of running complex physics simulations every time, the response is described by fitting histograms to specific variables (like particle energy and hit location) and storing these relationships in a lookup table (LUT). This allows for rapid response calculation.
- Look-up Table (LUT)
- A LUT is a pre-calculated database used by the fast simulation code. It stores the results of simulating detector responses for various input conditions. When the fast code runs, it simply looks up the appropriate response from this table based on particle type, energy, and where it hits the detector.
- Response Modeling
- This involves mathematically describing how a tank's signal changes based on what particle hits it. For example, the number of photo-electrons (PEs) hitting photomultiplier tubes is modeled using distributions like the negative binomial distribution to accurately reflect real detector behavior, including variations in signal strength.
Terminology
Summary
Surface arrays detecting the secondary particles of Extensive Air Showers (EAS) are one of the techniques used to detect gamma-rays from high-energy sources In this work, the authors present an alternative way to evaluate the performance-related quantities of a simulated array configuration through the parametrisation of the detector response This approach is presented to address the computational demands associated with full GEANT4-based simulations for very large fill factors, above 30-100 TeV The work details the implementation of a fast simulation code for WCD arrays through the use of a look-up table (LUT) to parametrize the detector response. p. 0 A parametrised simulation code to simulate the response of
ground detector arrays to Extensive Air Showers above 100
TeV
Andrea Chiavassaa,b,1, Andrea Negroa,b
aUniversity of Turin, Department of Physics, Via Pietro Giuria 1, Torino, 10125, , Italy
b
INFN-Turin, Via Pietro Giuria 1, Torino, 10125, , Italy
Abstract
Surface arrays detecting the secondary particles of Extensive Air Showers (EAS) are
one of the techniques used to detect gamma-rays from high-energy sources. To evaluate the reconstruction performance, gamma/hadron separation, and sensitivity of
such experiments, many simulations of gamma rays and protons are required.Where the paper says it
Parametrisation Methodology
The core idea involves defining a hit as a particle touching the external walls of the detector The parametrisation of a tank response is modeled as a distribution describing the behavior of the tank as a function of the hit characteristics, which must be stored inside a LUT that the simulation will use to replicate the tank response.
The generic approach for building this LUT involves discretizing the 7-dimensional parameter space—Particle type, Particle energy, (x,y,z) of the hit, and (θ,ϕ) angles of the particle direction—and simulating responses at these vertices. p. 3 Building a look-up table
Considering a detector of any shape, the hits of the particles entering it can be
described by seven variables:
• Particle type
• Particle energy
• (x,y,z) of the hit: in the most generic case, the detector has no geometrical
symmetry, and a complete map of the surface is needed to evaluate the detector
response
• (θ,ϕ) angles of the particle direction: in the most generic case, all the possible
angles should be considered, or at least the ones that lead the particle to enter
the detector
Each possible hit is a point in this 7-dimensional parameter space. First of all, we
can discretize it using a 7-dimensional grid and simulate (with a full GEANT4-based
simulation) the detector response of the vertices.Where the paper says it
For a specific detector like the SWGO double-layer tank example, only six variables are used to exploit cylindrical symmetry: Particle type (PID), Particle energy (E), Zenith angle of the direction of the particle (θ), Azimuth angle of the direction of the particle (ϕ), Distance of the particle hit from the central axis of the tank (dR), and Distance of the particle hit from the base of the tank (dH). p. 4 Azimuth angle of the direction of the particle (ϕ), measured starting from the
axis defined by the center of the tank and the hit position (a particle moving
towards the tank center has ϕ = 180◦
). Distance of the particle hit from the central axis of the tank (dR)
6.Where the paper says it
Response Modeling Details
The simulation requires fitting histograms to reproduce two key variables: number of photo-electrons (PEs) hitting the two PMTs, and arrival time of the first PE hitting the upper PMT.
For the number of PEs, a negative binomial distribution is often sufficient for most simulations, but a linear combination of two negative binomials is used to model cases with multiple peaks in the histogram. p. 7 In this case, a single negative binomial will never be able to describe the
data correctly. As a solution, we decided to model the signal as the linear combination of 2
negative binomials using:
g(k) = α · f(k, n, p) + (1 − α) · f(k; n
′
, p′
) (2)
Where α is limited to the interval [0,1] to preserve the normalization of the
distribution.Where the paper says it
The arrival time variable is modeled using an exponentially modified Gaussian distribution, and multiple peaks are described by a linear combination of two such distributions due to reflective walls. p. 9 The single peaks can be fitted with
an exponentially modified gaussian from the scipy package:
expnorm(x; µ, σ, λ) = λ
2
expλ
2
(2µ + λσ2 − 2x)
erfcµ + λσ2 − x
√
2σ
(3)
where erfc(x) is defined as
erfc(x) = 2
√
π
Z ∞
x
e
−t
2
dt (4)
To describe the multiple peaks, we defined the parametrisation of the time variable
as a linear combination of two such distributions, similarly to what we did for the
nPE distribution in equation 2. The reason behind the presence of multiple peaks is due to the reflective walls
of the tank.Where the paper says it
Validation and Performance Metrics
The fast simulation code was validated against the full GEANT4-based simulation framework using high-energy proton events thrown uniformly within a 1500 m radius.
For event level variables, the average density of PEs shows a negligible bias, but fluctuations are up to 10% The agreement between the number of PEs calculated by the full and the fast simulation is good for the tested energy range and configurations, showing that the latter correctly reproduces the detector behavior within the tested configuration.
The relative difference of the lateral distribution oscillates around 0 for both layers and on average is <3%
Execution Time Reduction
The authors evaluated the reduction of execution time with respect to the full simulation by simulating fixed-energy, fixed-direction gamma showers at various positions in the array.
The gain in execution time is bigger than 20, surpassing 50 if considering only events with the core inside the array This reduction is stronger near the center of the array, indicating that the fast simulation code improves gain in execution time with the number of particles that enter the detector.
Conclusion
The main aim of this work was to obtain a reliable simulation significantly faster than the standard GEANT4 approach The alternative method can be a good compromise to reduce the computing requirements, proving a technical contribution that can support future studies for future cosmic or gamma ray experiments.
Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.
Improvements for AI systems
-
Develop a fast simulation code that avoids full GEANT4 tracking operations by using a parametrisation of tank response stored in a look-up table (LUT). This implementation significantly reduces computational time, which is crucial for evaluating detector performance for very large fill factors above 30-100 TeV.
-
Implement robust methods to build the LUT by discretizing the 7-dimensional parameter space and using minimum chi squared minimization to fit tank response variables like number of photo-electrons (PEs) and arrival time, while also incorporating regularization techniques like disabling peaks describing less than 5% of counts to prevent artifacts.
-
Create a flexible interpolation algorithm for the LUT that uses a nearest neighbor criterion instead of linear interpolation when input variables are outside the volume of data, ensuring that
the nearest interpolator would instead return the nearest point it finds, without generating issues with NaN values.
-
Improve particle propagation handling by implementing particle-type dependent corrections for multiple tank crossings. Specifically, use the formula E(L) = E0e-L/X0 to approximate energy loss for electrons and positrons crossing multiple detectors, and use the formula Ppp(L) = e- 7L 9X0 to estimate pair production probability for photons.
-
Enhance event reconstruction performance by using a nearest algorithm for interpolating tank response parameters, as it is chosen
due to strong correlations in the fit parameters of the tank response,
which avoids issues with linear interpolation failing to describe nonlinear correlations and increasing initialization time. -
Improve sensitivity analysis by analyzing lateral distribution fluctuations using a ratio comparison between full and fast simulations, noting that
the agreement between the number of PEs calculated by the full and the fast simulation is good for the tested energy range and configurations.
Abstract
Surface arrays detecting the secondary particles of Extensive Air Showers (EAS) are one of the techniques used to detect gamma-rays from high-energy sources. To evaluate the reconstruction performance, gamma/hadron separation, and sensitivity of such experiments, many simulations of gamma rays and protons are required. The usual approach is divided into two steps. First, the CORSIKA software is used to generate EAS at the observation level. Then, a GEANT4-based full simulation code is used to simulate the array response. For very large fill factors, above 30-100 TeV, the computational resources required become too demanding. In this work, the authors present an alternative way to evaluate the performance-related quantities of a simulated array configuration through the parametrization of the detector response. We will present the implementation of such software, the validation checks we performed, and the application of this tool to a representative future ground-based water-Cherenkov detector array.
Sources
- The PeVatrons and the HAWC Observatory in Mexico
- Science Prospects for the Southern Wide-field Gamma-ray Observatory: SWGO
- Optimization of performance of the KM2A full array using the Crab Nebula
- The Southern Wide-field Gamma-ray Observatory
- HEPScore: A new CPU benchmark for the WLCG
Related papers
- On-sky demonstration of a vector Zernike wavefront sensor in a cascaded adaptive optics system
- Euclid Data Release 1 (DR1): Calibration strategy and observations
- The General Antiparticle Spectrometer (GAPS) Antarctic Balloon Payload
- Detection and excision of radio-frequency interference in radio arrays
- Unified matched-filter and excess-power pipeline for gravitational-wave detector characterization
- Ushering in sub-arcsecond angular resolution observatories in the far-infrared: The case for a double-Fourier interferometer