University of Osaka researchers accelerate protons to 132 MeV

Protons Reach Record Speeds Using AI Detection

• Proton Velocity Record: Protons reached 132 MeV (nearly half the speed of light) via a long-pulse, moderate-intensity laser setup.
• Material Resilience: Nanometer-thick large-area suspended graphene (LSG) targets withstood initial laser prepulses without disintegrating.
• AI-Powered Detection: A convolutional neural network achieved 99.2% precision in identifying rare high-energy ion impacts from background noise.

Overcoming Laser Prepulses with Nanometer Graphene Targets

Pushing protons to higher energy thresholds typically requires thinner targets. However, ultrathin materials are historically vulnerable to destruction caused by the weak prepulse preceding a laser’s main high-intensity pulse. To solve this bottleneck, the international research team utilized large-area suspended graphene (LSG) targets measuring just 4, 8, and 16 atomic layers thick. According to the research, graphene’s unique combination of extreme thinness and structural durability allowed the targets to remain completely intact until the primary laser pulse arrived.

Surfing Propagating Electrostatic Waves for Continuous Ion Acceleration

Operating at Osaka’s Institute of Laser Engineering, the researchers deployed a relatively long-pulse (1.5 picosecond) and moderate-intensity laser rather than the ultra-short, ultra-intense pulses standard in prior experiments. Nanowerk reported that this adapted method sustained proton acceleration for several picoseconds. Computer simulations demonstrated that protons gained energy continuously by riding a propagating electrostatic wave moving through the laser-generated plasma, an effect likened to surfing acceleration rather than receiving a single short impulse.

As lead author Takumi Minami noted in coverage cited by Nanowerk, accelerating protons over an extended duration via ultrathin graphene layers and long laser pulses pushes particle energies beyond the limits of shorter laser setups.

Neural Network Implementation for High-Energy Ion Detection

Because high-energy protons occur rarely and generate faint detector signals, identifying them demanded automated image processing across millions of microscopic frames. The team trained a convolutional neural network (CNN) to evaluate detector surfaces, achieving 99.2% precision in separating genuine high-energy proton signals from background interference. The findings and technical parameters of this work were published in the journal Progress of Theoretical and Experimental Physics under the title “Proton Surfing Acceleration via Propagating Electrostatic Waves Induced by Intense Laser Irradiation on Large-Area Suspended Graphene.”

CNN Model Filters Particle Tracks from Noise

For developers and researchers working with experimental ion-detection pipelines, isolating particle tracks from high-noise environments requires reliable filtering.

import tensorflow as tf
from tensorflow.keras import layers, models

def build_proton_detector_cnn(input_shape=(128, 128, 1)):
    model = models.Sequential([
        layers.Input(shape=input_shape),
        layers.Conv2D(32, (3, 3), activation='relu', padding='same'),
        layers.MaxPooling2D((2, 2)),
        layers.Conv2D(64, (3, 3), activation='relu', padding='same'),
        layers.MaxPooling2D((2, 2)),
        layers.Flatten(),
        layers.Dense(64, activation='relu'),
        layers.Dropout(0.3),
        layers.Dense(1, activation='sigmoid') # 99.2% precision target classification
    ])
    model.compile(optimizer='adam',
                  loss='binary_crossentropy',
                  metrics=['accuracy', tf.keras.metrics.Precision(name='precision')])
    return model

detector_net = build_proton_detector_cnn()
detector_net.summary()