GRANDProto300: Autonomous Radio Detection Prototype

• GRANDProto300 is a mid-scale prototype experiment that autonomously detects extensive air showers over a 200 km² area using advanced radio antennas.
• It pioneers detection technologies such as FPGA-based signal processing and machine learning algorithms to achieve high-precision cosmic-ray energy measurements.
• The facility serves as a technology pathfinder within the GRAND framework, refining methods to study the Galactic–extragalactic transition in cosmic particles.

GRANDProto300 is a mid-scale prototype experiment for the Giant Radio Array for Neutrino Detection (GRAND), a next-generation observatory designed to detect ultra-high-energy cosmic particles—specifically neutrinos, cosmic rays, and gamma rays—via the radio emission generated by extensive air showers (EAS) in the atmosphere. GRANDProto300 is both a technology pathfinder and a stand-alone scientific facility, targeting high-statistics, high-precision measurement of cosmic-ray-induced air showers in the critical energy regime where the transition from Galactic to extragalactic sources is expected. The experiment is currently under phased deployment in the Gobi Desert of western China, with the aim to achieve autonomous radio detection and robust event reconstruction of very inclined air showers over a 200 km² area with an array of 300 autonomous radio antennas.

1. Role within the GRAND Framework and Science Objectives

GRANDProto300 is the first large-scale field prototype in the multi-stage GRAND roadmap. Its objectives are:

• To validate the autonomous radio detection of very inclined extensive air showers (EAS), crucial for scaling to the much larger GRAND10k subarrays and ultimately the full GRAND observatory (consisting of 20 subarrays of 10,000 antennas each, totaling 200,000 antennas over 200,000 km²) (Martineau-Huynh, 2019, Kotera, 2021, Martineau-Huynh, 9 Jul 2025).
• To pioneer and benchmark key detection technologies (antenna hardware, front-end electronics, FPGA-based digital filtering and triggering, DAQ wireless communication, and on-the-fly signal processing) tailored for detection in the 50–200 MHz frequency band (Chiche, 3 Sep 2024, Collaboration et al., 25 Sep 2025).
• To provide high-statistics measurements of cosmic-ray-induced EAS in the range $E \in [10^{16.5}, 10^{18}]~{\rm eV}$, enabling detailed studies of the energy spectrum, mean chemical composition ($\langle \ln A \rangle$), arrival anisotropies, and physics of the Galactic–extragalactic transition (Zhang et al., 19 Jun 2025, Guelfand et al., 6 Jul 2025).
• To serve as a versatile experimental platform for developing, optimizing, and validating both the triggering and reconstruction algorithms (including advanced machine learning approaches) that will underpin the full GRAND detection paradigm (Mitra, 2023, Ferrière et al., 10 Jul 2025).

2. Array Architecture, Detection Principle, and Hardware Implementation

The GRANDProto300 array is sited at XiaoDuShan near Dunhuang (Gansu, China), an exceptionally radio-quiet plateau specifically chosen after extensive RFI surveys (Decoene, 2019, Chiche, 3 Sep 2024). The array design and hardware realize the following principles:

Antenna and Mechanical Design

• Each detection unit (DU) features a HorizonAntenna, a three-arm structure (NS, EW dipoles and vertical monopole) optimized for the 50–200 MHz band and positioned 3.5 m above ground to maximize sensitivity to near-horizontal air showers (Neto, 2023, Collaboration et al., 25 Sep 2025).
• Mechanical robustness addresses harsh Gobi Desert conditions (temperature extremes, high winds, sand); each DU includes a triangle-shaped base for power infrastructure (solar panel, battery, charge controller) and weather-sealed enclosures for electronics.

Front-End Electronics and Digital Signal Processing

• Analog signals from the antennas are amplified (LNA, ~20 dB), filtered (30–200 MHz elliptical bandpass), and digitized by a 14-bit, 500 MSPS ADC on the Front-End Board (FEB). The analog chain includes a tunable variable-gain amplifier (Collaboration et al., 25 Sep 2025).
• The digitized stream is continuously analyzed by an FPGA-based SoC, which executes real-time baseline subtraction, digital IIR notch filtering for narrowband RFI rejection (suppression >40dB at targeted frequencies), and a flexible self-trigger algorithm (Chen et al., 2023).
• The trigger scheme requires at least $N_{\rm trig}=5$ antennas out of 9 in a local group to exceed a programmable voltage threshold (typically 30–75 μV, i.e., several $\sigma$ above stationary noise), ensuring selection of impulsive, spatially-correlated EAS signals.

Data Acquisition and Wireless Communication

• When triggered, local circular memory buffers store digitized waveforms; time-stamping is provided by GPS-disciplined clocks with per-DU timing accuracies below 12 ns (Gaussian width) (Chen et al., 2023, Ma et al., 8 Jul 2025).
• Data are transmitted over a 5 GHz WiFi mesh network (bandwidth $\sim$150 Mbps) to the central station for secondary trigger evaluation (coincidence and topology validation across units), archiving, and subsequent offline analysis (Collaboration et al., 25 Sep 2025).

3. Signal Processing, Triggering, and Event Selection

Autonomous radio-detection in GRANDProto300 relies primarily on distinguishing nanosecond-scale, polarized broadband pulses from the immense anthropogenic and natural RFI background:

• Template-Matching Algorithms: Offline analyses use matched filters based on realistic simulated pulse templates. The detection threshold on the cross-correlation maximum is tuned to balance purity (signal/background discrimination, typically ~85%) and efficiency (reaching 80% for $E \gtrsim 2$ EeV) (Mitra, 2023).
• Multi-Level Trigger: T0 (instantaneous amplitude threshold), T1 (pulse-shape and frequency features), and T2 (spatial–temporal coincidence across neighboring units) progressively suppress background by >99% while retaining genuine EAS candidates (Decoene, 2019).
• Polarization Cuts: Event selection exploits the geomagnetic effect, favoring signals perpendicular to the local magnetic field: $b$-ratio $= \vec{e}_V \cdot \vec{e}_B < 0.25$ is typically employed to reject RFI (Lavoisier et al., 9 Jul 2025).
• Clustering and Spatial Consistency: Spatiotemporal clustering algorithms reject locally and temporally clustered events (e.g., manmade or atmospheric sources) that are inconsistent with EAS footprints (Lavoisier et al., 9 Jul 2025).
• Coincidence Time Windows and Footprint Size: Only events triggering at least 5 spatially distributed DUs within a window of a few microseconds are retained, ensuring compatibility with extensive air shower topologies (Chiche, 3 Sep 2024).
• Hybrid Option: The array can be complemented by a particle detector array for muon/electromagnetic separation—essential for cosmic gamma-ray and ultra-inclined EAS identification (Martineau-Huynh, 2019).

4. Event Reconstruction and Performance Metrics

Shower reconstruction in GRANDProto300 leverages unique modeling and novel computational approaches:

• Wavefront Reconstruction: Arrival directions are first reconstructed using a plane-wave fit to antenna times, then refined by modeling the radio wavefront as emerging from a point-like emission region ($X_e$), with a spherical curvature adjusted for atmospheric refraction (Guelfand et al., 6 Jul 2025).
• Angular Distribution Function (ADF) Fitting: The lateral and angular amplitude profiles across antennas are fit with a phenomenological ADF model:

$$f_i^{\rm ADF}(A, \theta, \phi, \delta\omega; X_e) = \frac{A}{l} \left[1+ G_A \frac{\cos\eta}{\sin\alpha}\right] \cdot \frac{1}{1 + 4 \frac{\left(\frac{\tan\omega}{\tan\omega_c} - 1\right)^2}{\delta\omega^2}}$$

(see details in (Guelfand et al., 6 Jul 2025)). This yields directional resolutions of $\leq0.1^\circ$ and links the amplitude normalization $A$ to the electromagnetic shower energy.

• Calibration and Energy Estimation: The energy estimator applies:

$$E^{\ast}_{\rm em} = \frac{A}{\sin\alpha\,f(\rho, \sin\alpha)}$$

where $\alpha$ is the geomagnetic angle, $f(\rho, \sin\alpha)$ is a correction for air density and coherence (Guelfand et al., 6 Jul 2025).

• Advanced Reconstruction (ML/GNN Methods): Graph neural networks (GNNs), informed by planar wavefront fits and physical priors, achieve energy resolutions of ~15% and angular uncertainty of ~0.14°, with built-in aleatoric and epistemic uncertainty estimates (Ferrière et al., 10 Jul 2025). Model calibration and domain adaptation strategies have been tested on real versus simulated data.
• Exposure and Trigger Efficiency: At $10^{17.5}$ eV, the trigger efficiency is about 50%, growing to 90% at $10^{18.3}$ eV. The daily exposure is 50 km²·day·sr at $10^{17.5}$ eV. The anticipated event rate (for $E>10^{17}$ eV and $\theta>65^\circ$) is ~130 events/day (Kato et al., 9 Jul 2025).
• Statistical Power: One year of operation yields $\sim$5×10⁴ EAS events, enabling energy spectrum and $X_{\rm max}$ (depth of shower maximum) measurements with fivefold improved statistical precision over previous radio-based and fluorescence techniques (Kato et al., 9 Jul 2025).

5. Science Case and Implications

GRANDProto300 addresses a broad spectrum of high-energy astrophysics and cosmic particle physics:

• Galactic–to–Extragalactic Transition: The target energy window ($10^{16.5}$–$10^{18}$ eV) spans the expected shift in cosmic-ray origin. Accurate spectra, mean mass, and anisotropy searches are enabled by high-statistics, unbiased radio detection (Zhang et al., 19 Jun 2025, Guelfand et al., 6 Jul 2025).
• Proton and Composition Studies: $X_{\rm max}$ can be reconstructed with a precision of 20 g/cm², enabling discrimination of light (proton) versus heavy primaries. A 90% purity proton spectrum can be extracted for tests of acceleration and propagation models (Zhang et al., 19 Jun 2025).
• Spectrum and Anisotropy: Resolutions allow fine mapping of the spectrum's knee, ankle, and potential large-scale anisotropy. The dipole amplitude sensitivity after five years is $A=5\times10^{-3}$ (for $E\lesssim10^{17.1}$ eV) (Zhang et al., 19 Jun 2025).
• Gamma-Ray and UHE Photon Searches: The hybrid approach with muon detectors enables gamma–hadron discrimination for constraints on UHE gamma-ray flux.
• Radio Astronomy and Transients: The field of view (~steradian scale) and continuous duty cycle facilitate sensitive searches for astrophysical radio transients (fast radio bursts, solar bursts, EoR 21-cm line—sensitivity at the mK level with a sub-array) (Decoene, 2019, Chiche, 3 Sep 2024, Ma et al., 8 Jul 2025).

6. Status, Achievements, and Operational Experience

• Current Deployment: As of mid-2025, 65/300 antennas are installed at XiaoDuShan with ongoing scaling. Early campaigns (GP13/65) demonstrated per-DU timing precision (stddev ≈5 ns), position reconstruction better than 15 m, and stable communication under harsh environmental conditions (Ma et al., 8 Jul 2025).
• Commissioning Results: Daily and sidereal variations in the galactic background have been mapped, solar bursts with fine time structure detected, and ≳40 cosmic-ray candidate events reconstructed with multiple independent algorithms (Ma et al., 8 Jul 2025, Lavoisier et al., 9 Jul 2025, Collaboration et al., 25 Sep 2025).
• Algorithm Validation: Coincidence analyses, polarization, and footprint cuts have efficiently filtered anthropogenic and atmospheric backgrounds; robust candidate selection pipelines have been established (Lavoisier et al., 9 Jul 2025).
• DAQ and Firmware Maturation: Mass-production prototypes of the DAQ board have been validated up to required ADC dynamic range (ENOB ≈10.7) and GPS-timing accuracy (Chen et al., 2023).
• Scaling Prospects: Full deployment to 300 antennas is expected by ∼2026, followed by scaling to GRAND10k in the 2030s (Chiche, 3 Sep 2024).

7. Limitations, Challenges, and Future Directions

• Systematic Uncertainties: While statistical uncertainties on $X_{\rm max}$ and energy are projected to be an order of magnitude improved over previous radio or fluorescence arrays, systematic errors in calibration (e.g., antenna model, timing, RFI) remain the limiting factor for composition studies (Kato et al., 9 Jul 2025, Zhang et al., 19 Jun 2025). Dedicated calibration campaigns and simulation–data cross-checks are ongoing.
• RFI Environment: Despite careful site selection, time-varying and airplane/communication radiofrequency backgrounds necessitate evolving signal-processing thresholds and multi-algorithm event validation (Collaboration et al., 25 Sep 2025).
• Algorithmic Extensions: Next stages will further integrate deep learning, graph-based, and physically informed reconstruction pipelines, aiming for both improved accuracy and explainable uncertainty intervals (Ferrière et al., 10 Jul 2025).
• Neutrino Readiness: Although GRANDProto300 is not expected to directly detect UHE neutrinos because of its limited area, its results directly validate and inform the technologies, algorithms, and operational approaches essential for the larger GRAND arrays ($\sim$20 subarrays, each with 10,000 antennas) that will target point-source and cosmogenic neutrino fluxes at sensitivities of $\sim 10^{-10}~{\rm GeV\,cm^{-2}\,s^{-1}\,sr^{-1}}$ above $5\times 10^{17}$ eV (Kotera, 2021).

Summary Table: Key GRANDProto300 Parameters

Parameter	Value/Feature	Context
Number of Antennas	300	Full deployment (2026)
Coverage Area	200 km²	Sparse array, Gobi Desert
Frequency Band	50–200 MHz	Radio air-shower pulse detection
ADC Resolution / Sampling Rate	14-bit / 500 MSPS	Per DU (Front-End Board)
Angular Resolution (reconstruction)	≤0.1°	Simulations, validated in events
Energy Resolution	~15%	ML/ADF-based, confirmed in sim/exp
$X_{\rm max}$ Precision	20 g/cm²	Full radio array, high statistics
Trigger Efficiency ($10^{17.5}$, $10^{18.3}$ eV)	50%, 90%	Array simulation
Exposure (@$10^{17.5}$ eV)	50 km²·day·sr	Per day, per energy bin
Daily CR Event Rate ($E>10^{17}$ eV, $\theta>65^\circ$)	~130	Full array, predicted
Main Science Targets	CR spectrum/composition, anisotropy, gamma and radio transients	Transition region physics, testbed

References to Primary Literature

All findings and technical descriptions are traceable to the cited arXiv preprints: (Martineau-Huynh, 2019, Decoene, 2019, Kotera, 2021, Mitra, 2023, Neto, 2023, Chen et al., 2023, Duan et al., 2023, Chiche, 3 Sep 2024, Zhang et al., 19 Jun 2025, Guelfand et al., 6 Jul 2025, Ma et al., 8 Jul 2025, Kato et al., 9 Jul 2025, Lavoisier et al., 9 Jul 2025, Martineau-Huynh, 9 Jul 2025, Ferrière et al., 10 Jul 2025, Collaboration et al., 25 Sep 2025).

---

GRANDProto300 thus represents a pivotal, rigorously validated experimental platform driving forward the realization of large-scale, autonomous, and high-precision radio arrays for astroparticle physics. Its ongoing technical and scientific results will define both the detection methodology and discovery potential of GRAND in the paper of the most energetic particles in the universe.