Quantum convolutional neural network for predicting nuclear charge radii
First application of quantum convolutional neural network for nuclear charge radii prediction, showing high accuracy.
Key Findings
Methodology
This study employs a hybrid quantum convolutional neural network (QCNN) to predict nuclear charge radii. Based on a classical convolutional neural network (CNN) framework, a small variational quantum convolutional filter is introduced as a quantum feature map to extract local correlations on the nuclear chart. The method encodes and processes local inputs through quantum circuits, generating nonlinear feature maps.
Key Results
- QCNN achieved a validation RMSE of 0.0100 fm, lower than CNN's 0.0124 fm, indicating higher predictive accuracy.
- In representative isotopic chains, QCNN better described odd-even staggering and shell structures, especially in Kr and Ba chains.
- QCNN achieved an RMSE of 0.0147 fm on the extrapolation test set, outperforming CNN's 0.0167 fm.
Significance
This study demonstrates the potential of quantum convolutional neural networks in nuclear physics data analysis, particularly in understanding complex nuclear structures. By introducing quantum feature maps, QCNN can better capture local nuclear chart patterns, which is significant for developing and testing nuclear models.
Technical Contribution
This study introduces a quantum convolutional layer to the classical CNN framework, showcasing the potential of quantum feature maps in nuclear data analysis. Compared to existing methods, QCNN provides a new nonlinear feature representation that better handles complex local nuclear structures.
Novelty
This is the first application of quantum convolutional neural networks for nuclear charge radii prediction. Compared to traditional methods, QCNN achieves more complex feature mapping through quantum circuits, capable of capturing finer structural changes.
Limitations
- The implementation of quantum circuits depends on current quantum computing capabilities, which may be limited by hardware.
- The model performs poorly on very light nuclei due to sparse neighboring nuclear information.
Future Work
Future work could explore richer data encodings, larger quantum filters, and more physically informed input features to verify performance under realistic quantum computing conditions.
AI Executive Summary
Quantum convolutional neural networks (QCNN) have been applied for the first time to predict nuclear charge radii, demonstrating their potential in nuclear physics data analysis. Traditional methods for predicting nuclear charge radii mainly rely on classical convolutional neural networks (CNNs), but these methods have limitations in handling complex local nuclear structures. QCNN, by introducing quantum feature maps, can better capture local nuclear chart patterns, thus improving predictive accuracy.
In experiments, QCNN achieved a validation RMSE of 0.0100 fm, outperforming CNN's 0.0124 fm. Additionally, QCNN showed better performance on the extrapolation test set, indicating its advantage in handling newly measured nuclei. This study shows that quantum feature maps can provide more complex feature representations suitable for describing phenomena like odd-even staggering and shell structures.
Despite QCNN's impressive predictive accuracy, its implementation depends on current quantum computing capabilities, which may be limited by hardware. Future research could explore larger quantum filters and richer data encodings to further enhance the model's performance and applicability.
Deep Analysis
Background
Nuclear charge radii are fundamental observables in nuclear structure studies, reflecting phenomena such as halo structures, shape staggering, and nuclear magic numbers along isotopic chains. Current experimental methods are mainly limited to stable and long-lived unstable nuclei, prompting the development of more accurate predictive methods by combining theoretical models with experimental data.
Core Problem
Existing methods for predicting nuclear charge radii have limited accuracy in handling local variations, especially when neighboring experimental data are sparse. A flexible framework is needed to identify useful correlations in available nuclear data to improve predictive accuracy.
Innovation
This study is the first to apply quantum convolutional neural networks to nuclear charge radii prediction. By introducing a variational quantum convolutional filter, QCNN generates nonlinear feature maps distinct from conventional filters, enhancing predictive accuracy.
Methodology
- �� Encode nuclear chart information using a 5×5 local window, including proton number Z, neutron number N, shell-related features VZ and VN, and odd-even staggering feature Δ.
- �� Replace the final 2×2 convolutional filter in the classical CNN with a four-qubit variational quantum filter.
- �� Minimize mean squared error between predicted and experimental charge radii using the RMSprop optimizer.
Experiments
Experiments used CR2013 and CR2021 datasets, excluding nuclei with Z≤6 and N≤6. The dataset was randomly split into training and validation sets in a 4:1 ratio. The extrapolation test set included 27 nuclei measured after 2021.
Results
QCNN achieved a validation RMSE of 0.0100 fm, outperforming CNN's 0.0124 fm. In representative isotopic chains, QCNN better described odd-even staggering and shell structures. QCNN achieved an RMSE of 0.0147 fm on the extrapolation test set, outperforming CNN's 0.0167 fm.
Applications
QCNN can be used for nuclear physics data analysis, especially when capturing complex local structures is necessary. Its quantum feature maps provide more complex feature representations suitable for describing phenomena like odd-even staggering and shell structures.
Limitations & Outlook
QCNN's implementation depends on current quantum computing capabilities, which may be limited by hardware. The model performs poorly on very light nuclei due to sparse neighboring nuclear information. Future work could explore larger quantum filters and richer data encodings.
Plain Language Accessible to non-experts
Imagine you're in a kitchen cooking. A classical convolutional neural network is like a chef with a fixed recipe, following the same steps every time. A quantum convolutional neural network is like an innovative chef with a magical pot that automatically adjusts cooking methods based on the ingredients, making tastier dishes. In this study, QCNN uses quantum feature maps to better capture complex changes in nuclear charge radii, just like the magical pot adjusts cooking methods based on different ingredients.
ELI14 Explained like you're 14
Imagine you're playing a game with many levels, each with different challenges. A classical convolutional neural network is like an old player who always uses the same strategy to win. A quantum convolutional neural network is like a smart new player with a magical tool that automatically adjusts strategies for each level, making it easier to win. In this study, QCNN uses quantum feature maps to better predict nuclear charge radii, just like the magical tool adjusts strategies for different levels.
Glossary
Quantum Convolutional Neural Network
A novel architecture combining quantum computing and convolutional neural networks, utilizing quantum circuits for feature mapping.
Used for predicting nuclear charge radii, enhancing predictive accuracy through quantum feature mapping.
Variational Quantum Circuit
A trainable quantum circuit that achieves feature mapping by adjusting parameters.
Core component of the quantum convolutional filter, generating nonlinear feature maps.
Root Mean Square Error
A statistical measure of the difference between predicted and actual values; lower values indicate more accurate predictions.
Used to evaluate QCNN and CNN performance in nuclear charge radii prediction.
Odd-Even Staggering
A phenomenon in nuclear structure characterized by systematic differences between odd and even nuclei.
QCNN better describes odd-even staggering in isotopic chains.
Shell Structure
A concept in nuclear physics describing the distribution and energy levels of nucleons within a nucleus.
QCNN better describes shell structures in isotopic chains.
Open Questions Unanswered questions from this research
- 1 How can QCNN be implemented under realistic quantum computing conditions? Are the simulation results still valid on actual hardware?
- 2 How do quantum feature maps perform in other nuclear physics tasks? Can they be extended to more complex nuclear structure predictions?
Applications
Immediate Applications
Nuclear Physics Data Analysis
QCNN can be used to analyze complex nuclear structure data, providing more accurate predictions.
Nuclear Model Development
By providing more accurate charge radii predictions, QCNN can offer important references for developing and testing nuclear models.
Long-term Vision
Quantum Computing in Scientific Research
As quantum computing technology advances, QCNN is expected to be applied in broader scientific research, driving progress in scientific computing.
Abstract
Quantum machine learning has the potential to become a new tool for understanding complex nuclear structures. In this work, we apply a hybrid quantum convolutional neural network (QCNN) to nuclear charge-radius prediction for the first time, aiming to explore the feasibility of quantum machine learning in nuclear-physics data analysis. Based on a classical convolutional neural network (CNN) framework, a small variational quantum convolutional filter is introduced as a quantum feature map to extract local correlations on the nuclear chart. The QCNN shows promising predictive accuracy and training stability, and provides a reliable description of the charge-radius evolution along several representative isotopic chains. These results support further investigation of quantum convolutional architectures for nuclear-structure data analysis.