🔭 astrophysics

Event types in H.E.S.S.: a combined analysis for different telescope types and energy ranges

This paper introduces a novel event-type-based analysis for the H.E.S.S. experiment that successfully combines monoscopic and stereoscopic data across different telescope types and energy ranges, achieving optimal sensitivity with 25–45% improvements over standard configurations and validated by Crab Nebula observations.

Original authors: Rodrigo Guedes Lang, Tim Unbehaun, Lars Mohrmann, Simon Steinmassl, Jim Hinton, Stefan Funk

Published 2026-04-01

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Rodrigo Guedes Lang, Tim Unbehaun, Lars Mohrmann, Simon Steinmassl, Jim Hinton, Stefan Funk

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you are trying to listen to a symphony orchestra, but your ears are split into two very different types:

The "Big Ears" (CT5): A massive, sensitive ear that can hear the quietest whispers (low-energy gamma rays) from far away, but sometimes gets confused when the music gets too loud and complex.
The "Small Ears" (CT1–CT4): Four smaller, sharper ears that work best when they listen together (stereoscopically). They are great at figuring out exactly where a loud sound is coming from, but they can't hear the quiet whispers at all.

For years, astronomers using the H.E.S.S. telescope array in Namibia had to choose: "Do I use the Big Ears to hear the quiet stuff, or the Small Ears to get the best details on the loud stuff?" They couldn't easily combine them because the data looked so different that the computer software didn't know how to mix the recipes.

This paper introduces a brilliant new way to listen to the whole orchestra at once. Here is the breakdown of their solution:

1. The Problem: The "One-Size-Fits-All" Trap

Previously, the analysis software was like a chef who only knew two recipes:

Recipe A (Mono): Use the Big Ear alone. Good for low energy, but blurry and prone to mistakes at high energy.
Recipe B (Stereo): Use the Small Ears together. Great for high energy, but completely deaf to low energy.

If a sound was in the middle, the chef had to guess which recipe to use, often missing the best details or the quietest notes.

2. The Solution: The "Event Type" Menu

The authors realized that instead of forcing every sound into one of two recipes, they should sort the sounds into three specific categories (Event Types) based on how the "sound" (the image of the gamma ray) looks when it hits the camera.

Think of it like sorting mail into three different bins based on the size and shape of the envelope:

Type M (The "Monoscopic" Whisper): These are the faint, low-energy signals caught only by the Big Ear. They are like a whisper in a library. The team upgraded the "Big Ear" with new software (neural networks) to make sure it doesn't get confused about where the whisper is coming from.
Type B (The "Bridge" Chorus): These are medium-energy signals caught by the Big Ear and at least two Small Ears. They are the "bridge" between the quiet and the loud. This is the tricky middle ground that used to be ignored or handled poorly.
Type A (The "Stereoscopic" Roar): These are the loud, high-energy signals caught by the Big Ear and all the Small Ears. They are like a thunderclap. Because so many telescopes see them, the location is pinpointed with extreme accuracy.

3. The Magic Trick: The "Smart Mixer"

The real genius of this paper is the Joint Likelihood Fit.

Imagine you have three different audio tracks recorded by different microphones. In the past, you had to pick one track to listen to. Now, the authors built a "Smart Mixer."

It takes the Type M track for the quiet parts.
It takes the Type B track for the middle parts.
It takes the Type A track for the loud parts.
Crucially: It knows that some "Type B" sounds are a bit fuzzy, so it gives them less weight in the final mix, while the super-sharp "Type A" sounds get more weight.

This allows them to create one single, perfect song that covers the entire range from the quietest whisper to the loudest roar without any gaps.

4. The Results: Hearing More, Better

By using this new method, the astronomers found:

Lower Threshold: They can now hear sounds that are 25–45% quieter than before. It's like suddenly being able to hear a pin drop from across the room.
Better Clarity: The "blur" in the middle energy range is gone. The map of the sky is sharper.
Time Travel: The improvement is so good that it's equivalent to doubling the observation time. If they wanted to get this much data with the old method, they would have needed to stare at the sky for twice as long.

5. The "Weather" Check (Systematic Uncertainties)

The paper also highlights a tricky problem: The Weather.
Just like fog or a dirty window can make a camera see things differently, changes in the atmosphere or the telescope mirrors can trick the sorting system. If the air is hazy, a "Type B" sound might look like a "Type M" sound, and the mix gets ruined.

The authors showed that they must apply a "Run-by-Run Correction" (like adjusting the white balance on a camera for every single photo taken) to ensure the final song sounds right, regardless of the weather on the night of the observation.

The Bottom Line

This paper is a masterclass in adaptability. Instead of forcing the universe to fit into two rigid boxes, the astronomers built a flexible system that sorts the data into three smart categories, treats each category with the specific tools it needs, and then blends them together.

The result? A telescope that can finally hear the entire symphony of the high-energy universe, from the faintest notes to the loudest crashes, all in one go. This approach is also a blueprint for the next generation of telescopes (CTAO), which will rely on this exact "mix-and-match" strategy.

1. Problem Statement

The High Energy Stereoscopic System (H.E.S.S.) employs a hybrid array consisting of four 12-meter telescopes (CT1–CT4) and one large 28-meter central telescope (CT5). Historically, H.E.S.S. analysis has been split into two distinct configurations:

Mono Mode: Uses only CT5 in monoscopic trigger mode. It achieves a low energy threshold ( $\sim$ 30–50 GeV) but suffers from poor energy/angular resolution and reduced effective area at energies above $\sim$ 200 GeV due to the difficulty of reconstructing single-camera images.
Stereo Mode: Requires at least two telescopes (typically CT1–CT4). It offers excellent performance above $\sim$ 150 GeV but cannot access the low-energy regime where CT5 operates alone.

The Core Challenge: Combining these two modes into a single, unified analysis has been difficult because events detected by CT5 alone, events detected by CT5 + small telescopes, and events detected by small telescopes alone have vastly different reconstruction characteristics and systematic uncertainties. Previous attempts to simply merge Mono and Stereo analyses or select the "best" reconstruction for overlapping events failed to provide sensitivity improvements across the full energy range.

2. Methodology

The authors propose a novel Event Type analysis framework that unifies the entire energy range by classifying events based on pre-reconstruction Hillas parameters (image size, number of pixels, and local distance) rather than post-reconstruction quantities. This approach is inspired by the Fermi-LAT analysis but adapted for the specific multiplicity and hardware differences of H.E.S.S.

A. Definition of Event Types

Events are categorized into three mutually exclusive types based on preselection cuts on Hillas parameters:

Type A (High-Quality Stereo): Requires CT5 and at least two images from CT1–CT4 with high signal size ( $\ge$ 200 p.e. for CT1–4, $\ge$ 50 p.e. for CT5). Dominates at high energies ( $E \gtrsim 500$ GeV).
Type B (Lower-Quality Stereo): Requires stereoscopic observation (CT5 + $\ge$ 2 CT1–4 or CT1–4 only) but with lower signal sizes. Dominates the intermediate range ( $\sim$ 150 GeV – 500 GeV).
Type M (Monoscopic): Events detected only by CT5 (no small telescopes triggered). Dominates the low-energy range ( $E \lesssim 150$ GeV).

B. Technical Improvements

The analysis incorporates several advanced reconstruction and separation techniques:

Monoscopic Enhancements (Type M):
- Time-based Cleaning: Replaces standard tail-cut cleaning for CT5, utilizing pixel timing correlations to retain more signal in faint images.
- Neural Networks (NN): Used to resolve the "flip" ambiguity (head vs. tail of the shower image) in monoscopic reconstruction, significantly improving angular resolution.
- Size-Dependent Cuts: Gamma-hadron separation cuts are optimized based on image size rather than a single global cut.
Stereoscopic Enhancements (Type A & B):
- CT5 Integration: Updated Monte Carlo simulations and fine-tuned thresholds allow CT5 data to be combined with CT1–4 without introducing systematic biases.
- Variable Separation: Mean-scaled variables are separated for CT1–4 and CT5 to handle different trigger systematics.
- Energy-Dependent Cuts: BDT (Boosted Decision Tree) cuts are optimized per energy bin to maximize the $q$ -factor (gamma efficiency / $\sqrt{\text{background}}$ ).

C. Statistical Combination

Instead of a simple sum of effective areas, the analysis uses a Joint Likelihood Fit. The contributions of each event type to the final statistical significance (Test Statistic, TS) are weighted. High-quality events (Type A) contribute more to the TS than low-quality events (Type B), allowing the fit to effectively "neglect" poorly reconstructed events that would otherwise degrade the sensitivity.

3. Key Contributions

Unified Analysis Framework: The first H.E.S.S. analysis capable of simultaneously exploiting monoscopic and stereoscopic data across the full energy spectrum ( $\sim$ 30 GeV to 100 TeV) in a single statistical framework.
Event Type Classification: A robust definition of event types based on pre-reconstruction Hillas parameters, ensuring non-overlapping categories that respect the physical limitations of the different telescope configurations.
Systematic Correction Scheme: Implementation of a run-by-run correction for observation conditions (atmospheric transparency and mirror degradation) to prevent biases in the event type fractions, which is critical for this method.
Performance Benchmarking: Comprehensive comparison against standard H.E.S.S. configurations (Mono, Stereo) and intermediate "improved" configurations (Mono++, Stereo++).

4. Results

Effective Area: The "Event types" configuration provides continuous coverage of the energy range, whereas Mono and Stereo have gaps or reduced efficiency at the boundaries.
Sensitivity Improvements:
- Overall: Sensitivity improvements of 25% to 45% are achieved across most of the energy range compared to standard configurations.
- Low Energy: The energy threshold is reduced from $\sim$ 230 GeV (standard Stereo) to $\sim$ 130 GeV (Event types), and further down to $\sim$ 50 GeV for the monoscopic component.
- Intermediate Energy: Significant gains ( $\sim$ 25–45%) are observed between 300 GeV and 5 TeV, attributed to the ability of the joint fit to down-weight low-quality Type B events while retaining high-quality Type A events.
Resolution:
- Angular: Improved by $\sim$ 57% at $\sim$ 100 GeV due to the NN-based flip resolution.
- Energy: Improved by breaking the degeneracy between energy and impact parameter using time-sensitive variables.
Validation (Crab Nebula):
- Analysis of real Crab Nebula data confirms the method's validity.
- The reconstructed spectrum matches the established H.E.S.S. (2024) model within $\pm$ 10% residuals.
- The method eliminates the artificial spectral curvature seen in Mono (due to pixel saturation at high energies) and the low-energy deviations in Stereo.
- Crucial Finding: Without run-by-run corrections, the transition between event types (e.g., Type M to Type B) introduces significant spectral biases. With corrections, the spectrum is stable.

5. Significance

Scientific Impact: This analysis allows H.E.S.S. to observe fainter sources and extend the reach of existing datasets to lower energies, effectively doubling the observation time equivalent for intermediate energies.
Future Relevance: The methodology is directly applicable to the Cherenkov Telescope Array Observatory (CTAO), which will rely heavily on a mix of Large, Medium, and Small Size Telescopes. The "Event Type" approach provides a blueprint for handling heterogeneous telescope arrays in next-generation gamma-ray astronomy.
Robustness: The work demonstrates that combining disparate telescope types is feasible if systematic uncertainties (atmospheric conditions, hardware calibration) are rigorously corrected on a run-by-run basis.

In conclusion, this paper presents a paradigm shift in IACT analysis, moving from disjointed mono/stereo modes to a unified, event-type-driven approach that maximizes the scientific return of the H.E.S.S. array and sets a precedent for future multi-telescope observatories.