Research and Development of a Modular Quantum Computing Software Library Using Quantum-HPC Hybrid Systems

In this sub-project 4, we aim to establish the environment that enables the rapid utilization of cutting-edge quantum computing software developed by QSRH and RIKEN within the quantum-HPC integrated platform environment integrated through this project. Specifically, we will develop libraries targeting four fields: "quantum chemistry," "condensed matter and statistical physics," "machine learning," and "optimization."

The primary objective in the field of quantum chemistry is to analyze molecular structures, understand chemical reaction mechanisms, and theoretically predict molecular properties (such as spectral analysis, optical and magnetic properties, and electrical conductivity) to compare them with experimental data. Improving the accuracy of these predictions is crucial for the efficient design and development of high-performance new materials and pharmaceuticals. Achieving this requires numerically solving quantum mechanical time evolution equations that incorporate the interactions of numerous electrons. Quantum computers, which can directly control quantum states based on the principles of quantum mechanics, provide an ideal environment for handling these equations. From the numerous quantum-classical hybrid algorithms developed by QSRH and RIKEN in the field of quantum chemistry, we will select algorithms that can effectively leverage the capabilities of the platform and proceed with the development of libraries that platform users can utilize.

Figure 1. Illustrative image of extracting molecular properties as bit strings using a quantum computer (from the QSRH website).

In the field of condensed matter and statistical physics, the primary goals include analyzing the crystal structures of solid materials, understanding phase transition phenomena, and, much like in quantum chemistry, comprehending the states formed by a vast number of electrons within a material. This understanding allows for theoretical predictions of the material's physical properties (electrical, thermal, magnetic, optical) and their comparison with experimental data to deepen the knowledge of condensed matter physics. These insights are important for the development of next-generation electronic materials. This field addresses multi-electron problems interacting according to quantum mechanics, often utilizing simple theoretical models to focus on the essence of physics. Therefore, it is recognized as a field where discussions on the quantum advantage may progress earlier than in quantum chemistry. By advancing the library development of quantum-classical hybrid algorithms, which will have been developed and tested for their effectiveness in the quantum-HPC collaborative applications of sub-project 8, we aim to establish an environment where users can easily access cutting-edge technology.

Figure 2. Conceptual diagram of quantum-classical hybrid algorithms.

The main purpose of machine learning is to learn from given data and automate and improve various tasks that humans have traditionally performed, such as pattern recognition and prediction, repetitive tasks, decision-making, personalized service delivery, data analysis and insight extraction, creative generation, etc. Enhancing the support capabilities of machine learning is crucial as it contributes to solving a wide range of societal issues. Quantum machine learning aims to leverage the degrees of freedom of quantum computers to solve problems with complex data structures that are difficult to learn using traditional methods. At QSRH, members who were pioneers in proposing quantum circuit learning have already created and released a quantum machine learning library (scikit-qulacs) and the world's largest quantum circuit learning dataset. This project will provide an environment where some of these functions can be utilized on the platform.

Figure 3. Conceptual diagram of quantum circuit learning (partially modified from the QSRG webpage).

The library developed in this project for optimization aims to provide tools to efficiently solve combinatorial optimization problems. Combinatorial optimization problems play a crucial role in many real-world issues, such as logistics optimization, scheduling, and network design. By leveraging quantum computers, there is a potential to find optimal solutions to these complex optimization problems more quickly than with traditional computing methods. Additionally, the understanding of how quantum state dynamics, which cannot be implemented by leading quantum annealing or similar methods, contribute to approaching optimal solutions is still in its infancy. Through the development of hybrid algorithms that combine classical heuristics, this project aims to not only provide an environment where the computational resources of the platform can be utilized for various optimization problems but also to address these challenges.

Figure 4. Pathfinding problem and the dependence of the number of paths on problem size.