CYGNO Optical TPC Detector

Updated 11 January 2026

CYGNO Optical TPC is a gaseous detector that leverages triple-GEM amplification and optical readout to enable detailed 3D imaging of rare particle interactions.
It achieves sub-mm spatial resolution and keV-scale energy thresholds, supporting directional dark matter detection and solar neutrino studies.
Modular prototypes like LIME validate its design, paving the way for scalable, ton-scale observatories for rare-event searches.

The CYGNO Optical Time Projection Chamber (TPC) is a gaseous detector concept developed for high-resolution, low-threshold, three-dimensional imaging of low-energy (O(1 keV)) particle interactions, specifically targeting rare-event searches such as directionally sensitive dark matter (DM) detection and solar neutrino studies. The CYGNO platforms exploit a combination of triple-Gas Electron Multiplier (GEM) amplification and innovative optical readout via scientific CMOS (sCMOS) cameras and fast photomultiplier tubes (PMTs), achieving sub-mm spatial resolution, keV-scale energy thresholds, and robust topological background rejection. The modular approach, demonstrated in prototypes such as LIME (50 L) and projected to scale to multi-m³ arrays, positions CYGNO as a leading effort within the broader CYGNUS collaboration for distributed, ton-scale, directional DM observatories (Amaro et al., 4 Jan 2026, Amaro et al., 2023, Baracchini et al., 2020, Amaro et al., 2022).

1. Detector Architecture and Operating Principle

The CYGNO Optical TPC consists of a gas-filled drift region (typically He:CF₄ 60:40 by volume at 1 atm, 295–300 K), bounded by a cathode and a segmented anode equipped with a triple-GEM stack. Tracks from ionizing particles produce primary ionization electrons, which drift under a uniform electric field (≈1 kV/cm) towards the GEMs. Each GEM is a 50 μm thick foil with 140 μm pitch, providing electron avalanche multiplication and secondary scintillation primarily from excited CF₄ and He fragments. The typical geometry of advanced prototypes such as LIME consists of a 33×33 cm² readout area and 50 cm drift length (active volume ≈50 L) (Amaro et al., 2023, Amaro et al., 4 Jan 2026, Antonietti, 1 Oct 2025).

Optical readout is achieved by imaging the final GEM plane with a high-granularity sCMOS camera (e.g., Hamamatsu ORCA-Fusion, 2304×2304 pixels, ≈150 μm projected pitch) and simultaneously recording time-resolved GEM scintillation with four fast PMTs (single-photon timing resolution ≃2 ns, QE ≃25% at 600 nm). This sensor configuration enables precise two-dimensional (x–y) imaging and timing-based z-coordinate reconstruction, delivering full 3D event information (Amaro et al., 4 Jan 2026, Amaro et al., 2023).

2. Signal Formation: Ionization, Drift, Multiplication, and Scintillation

2.1 Primary Ionization and Attachment

For each energy deposit ΔE, the mean primary electron count is $\bar{N}_e = \Delta E/W_i$ with $W_i = 35\ \mathrm{eV/pair}$ , actualized as a Poisson variate. During drift, electrons may be lost to gas impurities, with spatial dependence $N_e(z) = N_e^{\text{pr}} \exp(-z/\lambda)$ , where λ (attenuation length) is typically 1400 mm (Amaro et al., 4 Jan 2026).

2.2 GEM Charge Amplification

The amplification process in each GEM is parameterized as $G(\Delta V) = 0.034\,\exp(0.021\,\Delta V)$ , where ΔV is the voltage across the GEM. Fluctuations in single-GEM gain are modeled by an exponential distribution (first GEM only for computational efficiency), and extraction efficiency is $\epsilon_{\rm extr}(\Delta V)=0.87\exp(-0.002\,\Delta V)$ . Notably, the third GEM includes a saturation effect parameterized as $G(n_{\rm in}) = A\,g / [1 + \beta n_{\rm in}(g-1)]$ with $A=1.5$ , $\beta=10^{-5}$ , and $g = \exp(\tilde{\alpha} (\Delta V - V_0))$ (Amaro et al., 4 Jan 2026, Amaro et al., 9 May 2025).

2.3 Diffusion During Drift

Primary electrons diffuse both transversely and longitudinally, described as: $\sigma_{xy}(z)=\sqrt{\sigma_T^2\,z + \sigma_{T0}^2},\qquad \sigma_z(z)=\sqrt{\sigma_L^2\,z + \sigma_{L0}^2}$ with optimized parameters $\sigma_T=115\,\mu\mathrm{m}/\sqrt{\mathrm{cm}}$ , $\sigma_L=100\,\mu\mathrm{m}/\sqrt{\mathrm{cm}}$ , $\sigma_{T0}=350\,\mu\mathrm{m}$ , and $\sigma_{L0}=260\,\mu\mathrm{m}$ (Amaro et al., 4 Jan 2026, Amaro et al., 2023).

2.4 Avalanche Scintillation and Optical Transport

Scintillation light is produced in the avalanches, with a yield $\mathrm{LY}=0.07\,\mathrm{ph}/e^-$ . Optical photons enter the imaging system with a collection solid angle $\Omega = 1/[4N(1+1/I)]^2$ (N: f-number, I: magnification), affected by vignetting $\cos^4\phi$ and sensor-specific response. Only a small fraction (O(10⁻⁴)) of photons emitted reach the sensor after all geometric and transmission factors (Amaro et al., 4 Jan 2026, Almeida et al., 1 Dec 2025).

The sCMOS sensor’s pixel signal is $\text{ADC} = C_\text{conv} N_\gamma^\text{pix}$ , typically with $C_\text{conv} = 4$ ADU/photon, and the quantum efficiency $QE(\lambda)$ folded into the overall light-response map. PMT signals record the time structure of the scintillation, enabling z-position extraction by drift timing (Amaro et al., 4 Jan 2026, Antonietti, 1 Oct 2025).

3. Readout, Event Reconstruction, and Data Analysis

Event reconstruction in the CYGNO Optical TPC proceeds by integrating both the high-detailed sCMOS images (x–y) and PMT waveforms (z). Clustering algorithms, notably an intensity-adapted version of DBSCAN (“iDBSCAN”), are used to identify physical event clusters in the sensor output, achieving full detection efficiency at 5.9 keV and providing robust rejection of noise and background clusters (Baracchini et al., 2020, Amaro et al., 4 Jan 2026).

For full 3D reconstruction, various algorithms combine the high-resolution 2D spatial topology from camera frames with depth information from PMT or diffusion-based metrics. A Bayesian network approach has been implemented for 3D event inference based solely on PMT signals, further improving localization when combined with camera data (Amaro et al., 5 Jun 2025). Achievable 3D spatial resolution is sub-mm transversely and mm-scale along z, with energy thresholds of a few keV and energy resolution $\sigma_E/E\leq14\%$ at 5.9 keV (Amaro et al., 2023, Antonietti, 1 Oct 2025, Amaro et al., 4 Jan 2026).

Real-time data selection capabilities have advanced to convolutional autoencoder architectures trained on noise images for effective anomaly detection and rapid region-of-interest extraction. The best autoencoder configuration removes 97.8% of the image area while retaining (93.0 ± 0.2)% of true signal, meeting real-time streaming requirements for high data volumes (Amaro et al., 30 Dec 2025).

4. Detector Performance, Calibration, and Environmental Stability

4.1 Calibrated Performance Metrics

Calibration with external sources (e.g., ^55Fe 5.9 keV X-rays) across a range of energies (3.7–50 keV) demonstrates linearity of response to better than 5%, and energy resolution at 5.9 keV is $\sigma_E/E \approx 14\%$ , with multivariate regression down to <10% (Amaro et al., 4 Jan 2026, Amaro et al., 2023). Drift-dependent diffusion and collection efficiency are uniform over the volume, with spatial resolution dominated by diffusion at long drift: $\sigma_{xy}$ over 50 cm is ≈1.4 mm, and longitudinal $\sigma_z$ from PMT timing is ≈200 μm (Amaro et al., 2023).

4.2 Stability and Sensitivity to Environmental Conditions

Long-term stability of operation at high gain and full drift length is demonstrated. The He:CF₄ 60:40 mixture yields superior light output and electrostatic stability compared to more CF₄-rich mixtures, with typical discharge dead-time <4% even over month-long continuous operation. Pressure and humidity affect light yield at the 1%/mbar and 20–25%/RH% levels, respectively, but online calibration and equalization maintain energy response stability to ±5–10% (Baracchini et al., 2020, Antonietti, 1 Oct 2025, Amaro et al., 27 Oct 2025).

4.3 Sensor Technology Benchmarking

Sensor characterization reveals that the Hamamatsu ORCA-Fusion-BT sCMOS camera, with up to 95% QE at 550 nm and 0.7 e⁻ RMS noise, provides optimal photon yield for the CF₄ emission spectrum, leading to typical SNR~50 for 5.9 keV deposits. Ultra-low noise sensors (ORCA-Quest, 0.27 e⁻ RMS, broader UV sensitivity) open prospects for single-electron sensitivity and background suppression (Almeida et al., 1 Dec 2025).

5. Simulation and Physics Reach

A comprehensive, data-validated simulation framework combines detailed Geant4-based modeling of primary ionization, microscopic electron drift and attachment, GEM gain (including fluctuations and saturation), light transport, and detailed sensor response. Tuned against LIME data, this simulation replicates energy resolution, linearity, and spatial observables across the 3–50 keV energy range to 10–30% accuracy and is used to assess physics sensitivity and optimize operational parameters (Amaro et al., 4 Jan 2026).

Projected sensitivity for a modular 1 m³ CYGNO demonstrator (20 × LIME modules at LNGS, 1 yr exposure) reaches spin-independent WIMP-nucleon cross sections down to $10^{-42}$  cm² for $m_\mathrm{WIMP}\sim1$ –5 GeV/c², exploiting track directionality, head-tail discrimination, and robust background suppression. Directional recoil imaging enables not only sub-neutrino-floor DM detection but also solar neutrino directionality, nuclear recoil spectroscopy, and X-ray polarimetry for astrophysical sources (Amaro et al., 2023, Fiorina et al., 30 Oct 2025, Baracchini et al., 2020).

6. Scalability, Radiopurity, and Future Developments

CYGNO’s modular design—scalable via tiling of 33×33 cm² triple-GEM panels, each with an sCMOS and four PMTs—enables construction of O(1–100 m³) detectors. Radiopurity-driven engineering has validated field cages constructed from low-mass PET/Kapton foils with copper strip grading and nylon supports, achieving high voltage stability, electric field uniformity ( $\Delta E/E < 3$ % across the drift), and minimized radioactive backgrounds. Ongoing R&D focuses on negative-ion drift operation for further diffusion suppression (via SF₆ admixture), real-time clustering, and deployment of advanced readout hardware (Amaro et al., 27 Oct 2025, Marques, 13 Sep 2025).

Large-scale underground deployments will combine multi-m² optical readout areas, extensive shielding, and high-throughput data acquisition, with a roadmap targeting a demonstrator at 0.4 m³ (CYGNO-04) and eventual ∼30 m³ installations. These steps extend the sensitivity of directional rare-event searches into regimes beyond the reach of traditional, non-directional detectors, including statistical subtraction of solar neutrino backgrounds (Antonietti, 1 Oct 2025, Amaro et al., 2023, Marques, 13 Sep 2025).

References:

(Amaro et al., 4 Jan 2026, Amaro et al., 2023, Antonietti, 1 Oct 2025, Amaro et al., 27 Oct 2025, Amaro et al., 5 Jun 2025, Baracchini et al., 2020, Almeida et al., 1 Dec 2025, Amaro et al., 30 Dec 2025, Fiorina et al., 30 Oct 2025, Marques, 13 Sep 2025, Amaro et al., 9 May 2025, Amaro et al., 2022, 2320.16856, Amaro et al., 2023)