GRANDProto300 (GP300) Radio Array Prototype

• GRANDProto300 is a prototype radio array designed to detect extensive air showers from cosmic rays, validating key self-triggered detection techniques.
• The array deploys optimized radio detection units with integrated low-noise amplifiers, advanced digital processing, and hierarchical data acquisition for precise event reconstruction.
• GP300 provides high-statistics cosmic-ray measurements in the transition energy range, paving the way for future GRAND stages such as GRAND10k through improved trigger and calibration methods.

GRANDProto300, usually abbreviated GP300, is the main experimental effort of the prototype phase of the Giant Radio Array for Neutrino Detection (GRAND). It is a large, autonomous radio array being deployed in China to validate the detection principle and technology of self-triggered radio detection of extensive air showers, especially highly inclined showers, while simultaneously delivering a cosmic-ray science program in the transition region between Galactic and extragalactic origin. In its current phase, GP300 has operated with 65 deployed detection units and is intended to evolve toward a full array over about $200\,\mathrm{km}^2$, providing the empirical basis for GRAND10k and later GRAND stages (Martineau-Huynh, 9 Jul 2025).

1. Position within the GRAND program

GP300 occupies a specific place in the staged development of GRAND. GRAND is conceived as a next-generation observatory for ultra-high-energy particles, with ultra-high-energy neutrinos as the principal long-term target, but the prototype phase is centered on demonstrating that autonomous radio detection of air showers is technically robust and reconstructively precise under realistic field conditions. Within that prototype phase, GP300 is the largest and most representative instrument; the companion arrays GRAND@Auger and GRAND@Nançay serve, respectively, cross-calibration and hardware-test roles (Guelfand, 3 Jan 2025).

Its immediate scientific domain is not neutrino discovery. Current GP300 analyses explicitly state that the prototype is limited to detecting inclined extensive air showers from cosmic rays, and that neutrinos cannot be observed because of the restricted detector size. This is an important distinction, because GP300 is often conflated with the ultimate GRAND neutrino observatory. In practice, GP300 is a pathfinder for the detection regime required by Earth-skimming $\nu_\tau$ searches: near-horizontal geometries, sparse kilometer-scale arrays, radio-only triggering, and long-term autonomous operation (Guelfand et al., 6 Jul 2025).

Historically, GP300 was formulated as the first science-capable GRAND array, designed both to validate instrumentation and to open a high-statistics cosmic-ray program in the $10^{16.5}$–$10^{18}\,\mathrm{eV}$ or $10^{16.5}$–$10^{18.5}\,\mathrm{eV}$ range, depending on the study and analysis threshold adopted. Earlier design papers emphasized nested or hybrid spacings, while current status papers emphasize end-to-end validation: autonomous triggering, multi-station coincidence, timing, data transfer, and reconstruction of arrival direction, energy, and nature of the primary (Kotera, 2021).

2. Site, deployment history, and array geometry

GP300 is deployed at XiaoDuShan in Gansu Province, China, on a plateau in the Gobi Desert at $$(40.99^\circ\,\mathrm{N},\,93.94^\circ\,\mathrm{E})$$ and $1200\,\mathrm{m}$ above sea level. The site is described as electromagnetically excellent, with a large, relatively flat area suitable for a sparse radio array. A strong daily modulation of the stationary noise is observed, consistent with the transit of the Galactic Plane through the antennas’ field of view; this is interpreted as evidence for low terrestrial radio-frequency interference and a sky-noise-dominated bandpass (Martineau-Huynh, 9 Jul 2025).

Deployment has proceeded in stages. Earlier commissioning used smaller configurations, including 13-unit and 45-unit phases. The current status paper reports 65 deployed detection units, with 65 units running since the end of 2024 and enough active stations to detect a few tens of cosmic rays per day in the energy range $10^{17}$–$10^{18}\,\mathrm{eV}$. Commissioning has concentrated on mitigation of self-generated noise and optimization of trigger parameters, and the array response is reported as now homogeneous (Ma et al., 8 Jul 2025).

The full array is described somewhat differently in different GP300 studies. The current status overview states that GP300 will ultimately consist of 289 detection units covering a total area of $\nu_\tau$0, but does not specify the station spacing or detailed grid topology for that final configuration. By contrast, dedicated exposure and hybrid-performance papers model approximately 299-detector or 300-station layouts, typically with a denser infill at $\nu_\tau$1 spacing embedded in a sparser $\nu_\tau$2 grid. This suggests that the precise deployment geometry remains analysis-dependent while the global scale—about 300 stations over about $\nu_\tau$3—is stable across the literature (Kato et al., 9 Jul 2025).

The site choice has a dual rationale. For the present prototype, radio quietness and operational practicality are primary. For GRAND more broadly, the surrounding topography is also relevant to the Earth-skimming $\nu_\tau$4 concept, in which mountains can act as interaction targets and nearby terrain can help project radio footprints of the resulting nearly horizontal showers (Guelfand, 3 Jan 2025).

3. Detection units, front-end chain, and data acquisition

Each GP300 detection unit uses the Horizon Antenna, a three-polarization design with independent East–West, North–South, and vertical radiators. The low-noise amplifiers are integrated into the antenna head mounted at $\nu_\tau$5 height, which is intended to reduce attenuation for near-ground-propagating signals. The antenna sensitivity is reported as optimal in the $\nu_\tau$6–$\nu_\tau$7 range, while the operational analog filtering in the deployed front-end is in the $\nu_\tau$8–$\nu_\tau$9 band (Martineau-Huynh, 9 Jul 2025).

Signals are routed to a front-end board at the base of the pole, where they undergo analog filtering and additional amplification matched to the digitizer dynamic range. The digitizer is a 14-bit, $10^{16.5}$0 ADC, and first-level online processing is implemented on a Xilinx system-on-chip combining FPGA logic and CPUs. On-board digital processing includes filtering, notably notch and FIR filters, before local transient triggering. Timestamping is provided by a Trimble GPS module integrated into the front-end board (Martineau-Huynh, 9 Jul 2025).

The current DAQ architecture is hierarchical. After local first-level triggers, data are transmitted by Wi-Fi to a central DAQ. When multiple units—typically three to five in the status description—register coincident signals within a causal time window, the central DAQ issues a second-level trigger. Upon that trigger, $10^{16.5}$1-long waveforms for all three channels are recorded for each triggered antenna, and raw data are transferred to the IN2P3 computing center for offline processing (Martineau-Huynh, 9 Jul 2025).

Operational constraints are treated as first-class design inputs. Each unit is powered by a battery recharged by a $10^{16.5}$2 solar panel. Wireless transfer has been tested up to about $10^{16.5}$3 in point-to-point mode; under a typical multiplexed configuration of about 20 detection units per Wi-Fi receiver, the measured throughput is about $10^{16.5}$4 per unit, supporting a maximum second-level trigger rate of $10^{16.5}$5 for an event size of $10^{16.5}$6 (Martineau-Huynh, 9 Jul 2025).

Commissioning has also included dedicated board-level validation. An automated test system for the GP300 DAQ board reported an AD9694 14-bit $10^{16.5}$7 ADC, a configurable FPGA IIR notch filter with at least $10^{16.5}$8 attenuation demonstrated at 71 and 137 MHz, and an inter-board GPS timing jitter of $10^{16.5}$9 in laboratory tests. These are development-stage hardware figures rather than array-level physics performance, but they document the engineering maturation needed for mass production (Chen et al., 2023).

4. Detection principle, trigger logic, and reconstruction formalism

GP300 targets coherent radio emission from extensive air showers in the tens-of-MHz regime. In the GRAND status paper, the dominant mechanism in the relevant band is the geomagnetic effect, arising from the deflection of shower electrons and positrons in the Earth’s magnetic field. Pulses are short, of order $10^{18}\,\mathrm{eV}$0, and the radio technique is considered robust for showers above about $10^{18}\,\mathrm{eV}$1 (Martineau-Huynh, 9 Jul 2025).

The same geometric regime is central to the broader GRAND neutrino program. Earth-skimming $10^{18}\,\mathrm{eV}$2 interactions in rock can produce emerging $10^{18}\,\mathrm{eV}$3 leptons that decay promptly in the atmosphere, generating nearly horizontal air showers. GP300 does not have the aperture to detect that neutrino flux directly, but it is explicitly used to validate the radio-only trigger, background rejection, and reconstruction machinery required for such events (Guelfand et al., 6 Jul 2025).

In deployed operation, first-level triggers are produced locally after digital filtering, with exact pulse criteria not specified in the status overview. Companion candidate-search analyses on the early 46-DU layout defined Coincidence Data as events with at least four detection units triggered within a $10^{18}\,\mathrm{eV}$4 window, after which a conservative offline pipeline applied direction reconstruction, time-space clustering rejection, polarization cuts, footprint cuts, and additional quality cuts (Lavoisier et al., 9 Jul 2025).

Reconstruction in current GRAND analyses combines timing and amplitude models tailored to very inclined events. A point-source-like wavefront description is used after an initial plane-wave seed. In one formulation, the seed obeys

$10^{18}\,\mathrm{eV}$5

after which a spherical or point-source curvature fit reconstructs an emission point $10^{18}\,\mathrm{eV}$6. Amplitude information is then modeled with an Angular Distribution Function (ADF) in the $10^{18}\,\mathrm{eV}$7–$10^{18}\,\mathrm{eV}$8 band: $10^{18}\,\mathrm{eV}$9 Here $10^{16.5}$0 is a global amplitude scale, $10^{16.5}$1, $10^{16.5}$2 is the geomagnetic angle, $10^{16.5}$3 is measured from the $10^{16.5}$4 direction, $10^{16.5}$5 is the Cherenkov angle, and $10^{16.5}$6 controls the width of the ring-like enhancement. This phenomenological representation is used directly on voltage traces in current GP300 candidate analyses (Guelfand et al., 6 Jul 2025).

The status overview explicitly notes that many closed-form analytic expressions commonly requested in radio-air-shower work are not specified there: no explicit geomagnetic-angle scaling law, no GP300-specific lateral distribution formula, no analytic energy estimator, and no exposure formula. Those elements instead appear in companion contributions, including the ADF-based reconstruction work, exposure calculations, and GNN-based reconstruction developments implemented in GRANDlib (Martineau-Huynh, 9 Jul 2025).

5. Commissioning performance and first cosmic-ray candidates

Commissioning has established several instrumental benchmarks. Relative timing resolution better than $10^{16.5}$7 has been reported. Beacon-based tests achieve source-position reconstruction better than $10^{16.5}$8, and for a distant static ground source the azimuthal angle reconstruction is around $10^{16.5}$9. A major commissioning effort reduced self-generated noise to nominal levels, after which strong daily Galactic modulation in the $10^{18.5}\,\mathrm{eV}$0–$10^{18.5}\,\mathrm{eV}$1 band confirmed sky-noise dominance and excellent electromagnetic conditions (Martineau-Huynh, 9 Jul 2025).

Trigger and throughput performance have also been quantified. Trigger parameters were optimized to reach a measured trigger efficiency of about $10^{18.5}\,\mathrm{eV}$2 for event rates up to $10^{18.5}\,\mathrm{eV}$3, evaluated with a beacon generator emitting prompt pulses similar to extensive-air-shower signals. Narrow-band aeronautical lines between 119 and 136 MHz are present but mitigated by digital filtering (Martineau-Huynh, 9 Jul 2025).

The first cosmic-ray searches have been deliberately conservative. A published GP300 search pipeline on stable data from the first 46 operational antennas used six successive stages: coincidence search, planar-wave-front direction reconstruction, exclusion of clustered noise in time and space, a geomagnetic-polarization cut based on the median $10^{18.5}\,\mathrm{eV}$4-ratio, a footprint-multiplicity cut, and additional quality cuts including visual inspection. Applied to 533,466 coincidence events, this pipeline yielded 41 cosmic-ray candidates; after stricter reconstruction-quality criteria, 26 “solid” candidates remained (Lavoisier et al., 9 Jul 2025).

Those candidates are consistent with the expected phenomenology of inclined radio showers. The search paper reports compact elliptical footprints with 5–10 triggered units, short pulses in all three channels, low $10^{18.5}\,\mathrm{eV}$5-ratio values, and timing gradients consistent with a single incoming wave. The same work notes that a dominant transformer source accounts for about 70% of all raw events and that the clustering cut alone removes 79% of the coincidence sample, which underscores that background rejection remains a central operational problem rather than a solved preprocessing detail (Lavoisier et al., 9 Jul 2025).

Independent reconstruction methods are now used in parallel. The status paper states that arrival direction and energy have been reconstructed using three approaches: ADF fits on voltages, LDF fits on electric-field energy fluence, and a graph neural network. In separate GNN studies tailored to GP300-like conditions, physics-informed graph learning using peak-time and peak-amplitude features achieved angular resolution of $10^{18.5}\,\mathrm{eV}$6 in the abstract and about $10^{18.5}\,\mathrm{eV}$7 primary-energy resolution, with uncertainty estimates that remain slightly under-calibrated under simulation-to-data domain shift. These figures are reconstruction-development results, not yet a statement about the deployed full array’s end-to-end measured physics performance (Ferrière et al., 10 Jul 2025).

6. Exposure, science case, and transition to GRAND10k

GP300’s scientific core is high-statistics radio detection of cosmic rays in the transition region. A dedicated exposure study for the full array computes trigger efficiency, exposure, and expected event rates for zenith angles above $10^{18.5}\,\mathrm{eV}$8. In that calculation, the array-level trigger efficiency reaches 50%, 80%, and 90% at $10^{18.5}\,\mathrm{eV}$9, $$(40.99^\circ\,\mathrm{N},\,93.94^\circ\,\mathrm{E})$$0, and $$(40.99^\circ\,\mathrm{N},\,93.94^\circ\,\mathrm{E})$$1, respectively. The one-day exposure is about $$(40.99^\circ\,\mathrm{N},\,93.94^\circ\,\mathrm{E})$$2 near $$(40.99^\circ\,\mathrm{N},\,93.94^\circ\,\mathrm{E})$$3, increasing to about $$(40.99^\circ\,\mathrm{N},\,93.94^\circ\,\mathrm{E})$$4 at $$(40.99^\circ\,\mathrm{N},\,93.94^\circ\,\mathrm{E})$$5 and about $$(40.99^\circ\,\mathrm{N},\,93.94^\circ\,\mathrm{E})$$6 at $$(40.99^\circ\,\mathrm{N},\,93.94^\circ\,\mathrm{E})$$7 (Kato et al., 9 Jul 2025).

That study uses a simulation-based exposure estimator,

$$(40.99^\circ\,\mathrm{N},\,93.94^\circ\,\mathrm{E})$$8

with $$(40.99^\circ\,\mathrm{N},\,93.94^\circ\,\mathrm{E})$$9, $1200\,\mathrm{m}$0, and weighted simulated events thrown over $1200\,\mathrm{m}$1. Under those assumptions, GP300 is expected to observe about 130 cosmic-ray events per day for $1200\,\mathrm{m}$2 and $1200\,\mathrm{m}$3, corresponding to about $1200\,\mathrm{m}$4 events in one year of continuous operation (Kato et al., 9 Jul 2025).

The resulting measurement program is broader than simple event counting. The same exposure paper projects that GP300 will be able to measure the cosmic-ray energy spectrum from $1200\,\mathrm{m}$5 to $1200\,\mathrm{m}$6 in one year, with statistical precision about five times better than a previous mono-fluorescence measurement by TALE. For composition studies, it adopts LOFAR-like reconstruction performance and $1200\,\mathrm{m}$7 per event, yielding projected statistical uncertainties on the mean $1200\,\mathrm{m}$8 of about $1200\,\mathrm{m}$9, $10^{17}$0, and $10^{17}$1 at $10^{17}$2, $10^{17}$3, and $10^{17}$4, respectively. The same paper also stresses that systematic uncertainties, rather than counting statistics, are expected to dominate interpretation of chemical composition (Kato et al., 9 Jul 2025).

GP300’s science case is not restricted to cosmic-ray transition physics. Prospects papers include ultra-high-energy gamma-ray searches, possible operation with a ground particle array for electromagnetic–muonic complementarity, and radio-transient studies including fast radio bursts. One GP300 prospects paper estimates about one FRB per month in a beam-forming mode, while commissioning reports have already described detections of Galactic emission and solar radio bursts used for calibration and environmental characterization (Chiche, 2024).

The forward link to GRAND10k is explicit. GP300 is meant to validate the self-triggered radio approach, the power and communications model, and the reconstruction performance that will inform the next GRAND phase, defined as two arrays of about 10,000 detection units each, to be deployed from 2030 onward, most likely in China and Argentina. Near-term developments include advanced trigger methods such as NUTRIG and machine-learning denoising, possible movement toward distributed triggers, and hybrid concepts such as HERON. A plausible implication is that GP300’s most durable legacy may be less a single physics result than a calibrated operating envelope—trigger purity, timing, RFI tolerance, and data-handling architecture—on which the practical realization of GRAND10k depends (Martineau-Huynh, 9 Jul 2025).