image: Fig. 1：Probabilistic pit based on electronic phase separated manganite nanowire
Credit: ©Science China Press
Storage, computation, and communication are the three pillars of modern information technology, with computation being the central aspect. The von Neumann architecture, based on the Turing machine theoretical model, has dominated the development of computers over the past 80 years due to its computational generality. In 1981, Nobel Prize-winning physicist Richard Feynman posed a question in his famous talk "Simulating Physics with Computers": "Can computers efficiently simulate quantum physical systems?" This question presented a challenge to classical computers, which operate on binary logic (0 and 1) and are not well-suited for capturing the inherent uncertainty of quantum mechanics. Feynman proposed that an ideal way to address this issue would be to build a quantum computer, though this approach faces immense technical challenges. Another method is to build a "probabilistic computer" to handle probabilistic problems. In recent years, with the rapid development of physical materials and devices, probabilistic computing has re-entered the spotlight.
The core unit of probabilistic computing is the probabilistic bit (p-bit), a classical entity that can fluctuate rapidly between 0 and 1 over time. It combines characteristics of both classical and quantum computing, leveraging the probabilistic nature generated by thermal fluctuations, thus incorporating some of the probabilistic principles of quantum computation. Probabilistic computing can address certain classical combinatorial optimization problems, such as prime factorization, the traveling salesman problem, and Bayesian inference, at relatively low cost. We can view probabilistic computing as a bridge between classical and quantum computing, and more closely aligned with the hardware structure of existing classical computers. Under this background, finding physical materials and devices that exhibit high quality, low power consumption, and probabilistic properties is a key factor in the development of probabilistic computing.
[Recently, a team led by Professor Jian Shen and Hangwen Guo from the Institute for Nanoelectronics devices and Quantum Computing at Fudan University fabricated a probabilistic bit device based on manganite nanowires. This device exploits phase separation domains that fluctuate between ferromagnetic metal states (0) and antiferromagnetic insulating states (1), achieving full control of its probabilistic characteristics with nanoampere-level currents. Experiments have demonstrated that this p-bit exhibits extremely high operational stability during repeated computational operations. Simulations have shown that these p-bits play a crucial role in optimizing problems such as Bayesian inference, showcasing exceptional computational potential. This research provides scientific support for the future realization of high-performance probabilistic computers. The results have been published in the 2025 issue of <National Science Review> under the title "Superior probabilistic computing using operationally stable probabilistic-bit constructed by manganite nanowire," with Dr. Weichao Yu, Dr. Hangwen Guo, and Dr. Jian Shen as co-corresponding authors.]
The research team utilized manganite nanowires as probabilistic bit units, enabling spontaneous resistance fluctuations between the ferromagnetic metal (low resistance) state and the antiferromagnetic insulating (high resistance) state. Moreover, the probability of the system being in a state can be precisely and fully controlled through nanoampere-level input currents (Figure 1). Experimental results demonstrated that this p-bit exhibited excellent operational stability under extensive operations, with a standard deviation of less than 1.3% (Figure 2). Through further simulation, the team discovered that this operationally stable p-bit played a critical role in Bayesian inference tasks, where the accuracy of the inference results was significantly higher than that of existing similar probabilistic bits, showcasing superior computational performance potential (Figure 3). In addition, this p-bit can generate high-quality intrinsic true random numbers, indicating potential applications in cryptography. These findings provide a new pathway toward realizing high-performance probabilistic computers.