SiW Dataset: Precision Electromagnetic Calorimetry

• SiW Dataset is a collection of precise, high-granularity detector data from silicon-tungsten ECAL prototypes designed for accurate electromagnetic shower reconstruction.
• The dataset supports particle flow algorithms and shower separation studies by offering detailed energy deposition maps across tens of thousands to millions of segmented channels.
• Validated through beam tests at DESY and CERN, the data ensures high efficiency and reliability for advanced event reconstruction in high-energy physics experiments.

The SiW dataset refers to the collection of high-granularity, precision-calibrated detector data generated by the silicon–tungsten electromagnetic calorimeter (SiW ECAL) prototypes developed by the CALICE collaboration for linear collider detectors, as exemplified in large-scale R&D for the International Large Detector (ILD) at the International Linear Collider (ILC). SiW datasets serve as a foundation for benchmark studies in particle flow algorithms, shower separation, reconstruction efficiency, electronics integration, and the development of technological solutions for next-generation calorimetry. The dataset encapsulates detailed energy deposition information from tens of thousands to, ultimately, hundreds of millions of 5–5.5 mm² silicon channels arranged in highly segmented, multi-layered volumes, supporting advanced event reconstruction and analysis required for future high-energy physics experiments.

1. Design Principles and Detector Architecture

SiW datasets originate from detector prototypes whose architecture specifically fulfills high-granularity calorimetry requirements for linear $e^+e^-$ colliders. The principal design considerations include:

• Granularity: Silicon sensors segmented into pads of $5.5\times5.5\,\mathrm{mm}^2$ (Pöschl, 2012), with pad sizes as small as $5\times5\,\mathrm{mm}^2$ in technological prototypes (Pöschl, 2022).
• Layering and Depth: The electromagnetic calorimeter stacks 15–30 layers of silicon interleaved with tungsten absorbers for a total thickness of $\sim24\,X_0$ (radiation lengths), with individual layers varying between 3.4 mm and 4.4 mm depending on position (Pöschl, 2012).
• Materials: Tungsten absorbers (radiation length $X_0 = 3.5\,\mathrm{mm}$, Molière radius $R_\mathrm{M}=9\,\mathrm{mm}$, interaction length $\lambda_\mathrm{I}=96\,\mathrm{mm}$) are employed for containment and spatial compactness.
• Mechanical Structure: Sensitive units and front-end electronics are integrated in a tungsten–carbon composite alveolar housing (Pöschl, 2012), with advanced PCBs (including chip-on-board, COB, solutions as thin as 1.2 mm) supporting compactness and minimal dead space (Irles, 2020, Irles, 2020).

The SiW dataset structure is defined by this granularity and segmentation, enabling complete spatial imaging of electromagnetic showers for the reconstruction of complex events in particle physics.

2. Electronics, Readout, and Data Characteristics

The SiW ECAL features custom ASICs—principally the SKIROC family—directly bonded to ultra-thin PCBs within the calorimeter layers (Pöschl, 2012, Pöschl, 2022). Key aspects include:

• Signal Processing: Each SKIROC ASIC handles 64 channels, providing analog preamplification, digitization (via a 12-bit Wilkinson ADC), and zero-suppression to optimize data throughput (Pöschl, 2022).
• Trigger and Thresholds: The typical trigger threshold corresponds to a signal-to-noise ratio of $\mathrm{SNR}\sim8$, facilitating auto-triggering capabilities at half-a-MIP energy (Pöschl, 2012, Pöschl, 2022).
• Power Pulsing: A novel scheme cycles electronics in synchronization with the ILC bunch train ($\sim1\,\mathrm{ms}$ active, $\sim200\,\mathrm{ms}$ idle) to constrain heat dissipation and enable integration inside the solenoid coil (Pöschl, 2012).
• Data Structure: For a prototype with 15,360 readout channels (15 layers $\times$ 1,024 per layer), SiW datasets record event-wise channel ADC values, trigger patterns, and timing information, post-processed to yield calibrated MIP distributions and energy maps (Pöschl, 2022).
• PCB and ASU Technologies: The COB solution integrates ASICs via wire bonding to 1.2 mm thick, multi-layer PCBs (compared to the 3–3.5 mm of BGA-packaged boards), reducing dead volume and facilitating up to $10^8$ channels in a $\leq20\,\mathrm{cm}$ global ECAL thickness (Irles, 2020, Irles, 2020).

This electronic integration produces datasets with minimal noise and high spatial fidelity, allowing robust analysis of shower profiles, separation, and energy deposition patterns.

3. Calibration, Testing Regimes, and Data Quality

The calibration of the SiW dataset is performed via dedicated beam tests (notably at DESY and CERN), utilizing MIP signals from electron beams for calibration constants and efficiency mapping (Pöschl, 2022).

• Calibration Workflow: Test beams at energies 1–6 GeV (DESY) and 10–150 GeV (CERN) are used to calibrate individual cells, yielding $>230,400$ calibration constants for a 15,360-channel prototype (Pöschl, 2022).
• Efficiency: Efficiency maps constructed per layer show cell efficiencies $>90\%$ in well-functioning regions, with inhomogeneities—primarily due to routing or sensor delamination—identified for further investigation.
• Quality Metrics: MIP spectra, peak finding, layout displays, and event overlays are employed to assess channel uniformity and spatial separation capability (Irles, 2020). Added decoupling capacitances can mitigate retrigger noise, ensuring that signal fidelity is competitive with BGA-based systems (Irles, 2020).

The dataset thus enables precise event reconstruction in both test-beam and operational environments, forming the basis for reliable benchmarking of calorimeter performance.

4. Particle Flow Algorithms, Shower Separation, and Event Reconstruction

SiW datasets are critical for the development and optimization of Particle Flow Algorithms (PFAs), which reconstruct every individual particle within a jet using combined tracking and calorimetric information (Shpak, 2018).

• Shower Separation Studies: Overlapping EM–EM or EM–hadronic showers are emulated through event mixing of single-shower data, with reconstructions benchmarked across cell sizes (1×1 cm², 5×5 mm², and 2.5×2.5 mm²) and algorithm variants (Pandora, Garlic, Arbor) (Shpak, 2018).
• Performance Metrics: Efficiency curves $P(\Delta)$ quantify the probability of correctly reconstructing objects as a function of inter-shower separation $\Delta$, demonstrating strong agreement between test-beam and simulation for canonical configurations (Shpak, 2018).
• Algorithmic Trade-offs: Pandora excels in $\pi^+$–$\gamma$ separation with standard granularity, Garlic performs best in $\gamma$–$\gamma$ separation, while Arbor benefits from finer segmentation but requires algorithmic tuning to suppress over-splitting (Shpak, 2018).
• Calibration Formula: Layer-wise energy calibration is described by

$$E = \sum_{j=0}^9 E_j^{(\mathrm{ECAL})} + 2 \sum_{j=10}^{19} E_j^{(\mathrm{ECAL})} + 3 \sum_{j=20}^{29} E_j^{(\mathrm{ECAL})} + \sum_{k=0}^{37} E_k^{(\mathrm{AHCAL})}$$

where layer weights reflect the increasing tungsten thickness (Shpak, 2018).

Ultimately, SiW datasets allow for refined assessment of jet energy resolution targets, confusion contributions ($>100$ GeV), and high-fidelity event imaging needed for W/Z boson and $\tau$-decay analyses.

5. Technological Developments and Scaling for Future Experiments

Significant R&D within CALICE has advanced the SiW dataset infrastructure toward mass production and deployment in full-scale detectors.

• Production and Integration: Industrial-scale fabrication of $\sim3000\,\mathrm{m}^2$ active silicon and $10^8$ cells is being prepared, necessitating ultra-thin, planar PCBs and robust wirebonding methodologies (Pöschl, 2012, Irles, 2020).
• Component Innovations: COB solutions achieve $\leq1.2\,\mathrm{mm}$ PCB thickness and $\leq0.5\,\mathrm{mm}$ planarity deviation, vital for integrating many layers inside $\leq20\,\mathrm{cm}$ detector envelopes (Irles, 2020).
• Mechanical Reliability: Direct chip attachment, advanced gluing robots, and improved sensor guard-ring designs (segmented or none) minimize cross-talk and mechanical stresses, improving long-term reliability (Pöschl, 2022).
• Electronics Scaling: Ultra-compact front-end systems (SL-Board, CORE-module) demonstrate capacity for reading out $>10,000$ channels per board, supporting full detector scaling (Irles, 2020, Pöschl, 2022).

These developments ensure that SiW datasets will be available at the channel density, spatial compactness, and electronic integration necessary for future collider experiments (ILC, FCCee, CEPC).

6. Applications and Impact on High-Energy Physics

SiW datasets underpin several critical domains in experimental particle physics:

• Precision Jet Reconstruction: Achieving $\sigma_E / E \simeq 30\%/\sqrt{E}$ and imaging calorimetry supports PFA-driven jet reconstruction for multi-boson final states (Pöschl, 2012).
• Detector Benchmarking: Beam test data informs algorithm tuning, sensor development, and system integration for detector commissioning (Pöschl, 2022).
• Physics Studies: The ability to resolve closely overlapping EM–EM or EM–hadronic showers enables sophisticated analyses of $\pi^0$/$\tau$ decay modes, W/Z separation, and background rejection (Shpak, 2018).
• Transferability: Technological solutions developed for the SiW dataset (thin PCB, COB, advanced front-end) inform both large-scale collider applications and smaller precision experiments (e.g., LUXE at DESY) (Pöschl, 2022).

These capabilities underscore the foundational role of SiW datasets for particle flow calorimetry and future high-energy physics detector technologies.

7. Future Directions

Advancements in SiW dataset acquisition and associated detector technology continue with:

• Engineering Prototypes: Scaling to modules with local power pulsing, autonomous bias distribution, and further material budget reduction (Pöschl, 2022).
• Timing Resolution: Achieving $\sim30\,\mathrm{ps}$ per centimeter granularity for following shower development temporally (imposing causality constraints for clustering) (Pöschl, 2022).
• Wider Applicability: Adaptation for use in new facilities with higher collision rates and cooling demands, such as Higgs factories or next-generation circular colliders (Pöschl, 2022).
• Algorithmic Integration: Ongoing optimization of PFAs and software compensation informed by SiW dataset characteristics and simulation frameworks (DD4HEP/GEANT4) (Pöschl, 2022).

A plausible implication is that as SiW datasets scale in scope and fidelity, synergistic advances in electronics, sensor integration, and calibration will further enhance the precision and versatility of calorimeter-based measurements in particle physics.