Evaluating spintronics-compatible implementations of Ising machines

Andrea Grimaldi,Luciano Mazza,Eleonora Raimondo,Pietro Tullo,Davi Rodrigues,Kerem Y. Camsari,Vincenza Crupi,Mario Carpentieri,Vito Puliafito,Giovanni Finocchio

DOI: https://doi.org/10.1103/PhysRevApplied.20.024005

2023-04-09

Abstract:The commercial and industrial demand for the solution of hard combinatorial optimization problems push forward the development of efficient solvers. One of them is the Ising machine which can solve combinatorial problems mapped to Ising Hamiltonians. In particular, spintronic hardware implementations of Ising machines can be very efficient in terms of area and performance, and are relatively low-cost considering the potential to create hybrid CMOS-spintronic technology. Here, we perform a comparison of coherent and probabilistic paradigms of Ising machines on several hard Max-Cut instances, analyzing their scalability and performance at software level. We show that probabilistic Ising machines outperform coherent Ising machines in terms of the number of iterations required to achieve the problem s solution. Nevertheless, high frequency spintronic oscillators with sub-nanosecond synchronization times could be very promising as ultrafast Ising machines. In addition, considering that a coherent Ising machine acts better for Max-Cut problems because of the absence of the linear term in the Ising Hamiltonian, we introduce a procedure to encode Max-3SAT to Max-Cut. We foresee potential synergic interplays between the two paradigms.

Other Condensed Matter

What problem does this paper attempt to address?

The problem that this paper attempts to solve is: how to effectively use spintronics technology to implement Ising machines (IMs) in order to solve complex combinatorial optimization problems (COPs). Specifically, the paper compares the performance and scalability of two Ising machine implementation methods - coherent Ising machines (cIMs) and probabilistic Ising machines (pIMs) - in solving the maximum - cut problem (Max - Cut). The paper evaluates the advantages and disadvantages of these two methods through software simulation and explores their potential in hardware implementation, especially the application prospects of hybrid spintronics technology combined with CMOS technology.

### Main problems of the paper:
1. **Solving complex combinatorial optimization problems**: The paper focuses on how to efficiently solve combinatorial optimization problems belonging to the NP class, especially the maximum - cut problem (Max - Cut).
2. **Application of spintronics technology**: Research on the application of spintronics technology in implementing Ising machines, especially how to use magnetic tunnel junctions (MTJs) to implement the binary states of Ising machines.
3. **Comparing the performance of different Ising machines**: Through software simulation, compare the performance of coherent Ising machines (cIMs) and probabilistic Ising machines (pIMs) in solving the Max - Cut problem, and analyze their respective advantages and limitations.
4. **Exploring the potential of hardware implementation**: Discuss how to improve the performance of Ising machines through hardware implementation (such as using high - frequency spintronic oscillators), especially the potential at high - frequency operation.

### Specific research contents:
- **Theoretical background**: Introduce the Ising model and its application in combinatorial optimization problems, as well as the basic principles of spintronics technology.
- **Model implementation**: Describe in detail how to use three - terminal magnetic tunnel junctions (MTJs) to implement the binary states of Ising machines, including the specific designs of coherent Ising machines and probabilistic Ising machines.
- **Performance comparison**: Through software simulation, compare the performance of the two Ising machines in solving the Max - Cut problem, including indicators such as the number of required iterations and the quality of solutions.
- **Hardware potential**: Explore the potential of high - frequency spintronic oscillators in implementing efficient Ising machines, especially their advantages at high - frequency operation.

### Main conclusions:
- **Software performance**: Probabilistic Ising machines (pIMs) perform better than coherent Ising machines (cIMs) in software implementation, especially in terms of the number of required iterations.
- **Hardware potential**: Although pIMs perform better in software implementation, cIMs have the advantage of high operating frequency in hardware implementation and may perform well in practical applications.
- **Future prospects**: Propose the potential synergy between the two Ising machines in hardware implementation, and suggest further research on how to combine the advantages of both to implement more efficient Ising machines.

Through these studies, the paper provides a theoretical basis and technical route for the development of efficient and low - power Ising machines, and is expected to make breakthroughs in solving complex combinatorial optimization problems.