Quantum Random Access Memory (QRAM) Researchers Announce Milestone in Scalable Superconducting Quantum Computer Architecture

Related Vendors

FAULHABER Drive Systems

SIMTOS (Seoul International Machine Tool Show)

Chinese researchers say they have achieved a breakthrough in testing a coherent routing system for QRAM memory. This could enable deeper circuits for quantum computers and thus prevent errors that occur during scaling.

A research team from the University of Science and Technology of China, the Chinese quantum computing startup Origin Quantum, and the Institute of Artificial Intelligence at the Hefei Comprehensive National Science Center has demonstrated a coherent routing system for quantum memory on a superconducting quantum processor. The accompanying study was open access in the journal Physical Review X. The work was based on “Wukong,” a quantum processor developed by Origin Quantum with 72 superconducting transmon qubits, ten of which were used for the experiments.

The focus is on so-called quantum random access memory, or QRAM for short. This type of memory is designed to retrieve data based on quantum addresses and is considered a potential building block for applications such as quantum search and quantum machine learning. In the bucket-brigade architecture under investigation, quantum routers are responsible for routing information through a binary network based on the state of an address. However, as the scale increases, conventional implementations quickly become complex: the number of required gates and the circuit depth increase, causing errors and decoherence to have a greater impact.

Auxiliary States Shorten the Circuit

The researchers are therefore relying on a method known as the Transition Composite Gate (TCG). In this approach, the higher energy levels of the superconducting transmons are temporarily used as auxiliary states. A controlled SWAP operation required for the router can thus be constructed using three two-qubit transition gates. According to the study, this significantly reduces the circuit depth of a router compared to a conventional Clifford decomposition: in the variant examined, with error detection, it drops from 30 to 12 levels.

At the same time, the third energy level of the transmons serves for error detection. For this purpose, the quantum address is encoded in the non-adjacent states |0⟩ and |2⟩. If, due to gate errors, decoherence, or thermal effects, it enters the intermediate state |1⟩, this process can be detected during measurement. The affected runs are then discarded. According to the authors, this post-selection method requires no additional auxiliary qubits.

In tests of three quantum routers evaluated individually, less than two percent of the population remained at the input after routing, which corresponds to a transmission efficiency of approximately 98 percent. In the random-access tests developed by the researchers, the three routers with error detection achieved fidelities of 95.1, 93.5, and 95.7 percent. On average, this results in the value of 94.8 percent reported in the study.

Two Router Levels as a Scalability Test

The team then cascaded three routers to form a two-level network. To achieve this, the modules were arranged in a triangular structure on the chip; in addition, routers on a single level can be partially controlled in parallel to limit the circuit depth. According to the study, the network achieved a transmission efficiency of 93 percent. In the random-access test, the average fidelity with error detection was 82.4 percent.

However, the results also highlight the limitations of the experiment. The authors cite decoherence and leakage into undesirable states as the most significant sources of error. In the two-stage network, crosstalk between components is likely an additional factor. As a result, the benefit of error detection decreases as the system size increases, to a lesser extent than would be expected based on individual routers. Furthermore, what was demonstrated was a two-stage routing system—not yet a complete, large-scale QRAM memory.

The work thus primarily provides experimental evidence that coherent and reversible quantum routers with comparatively low circuit complexity can be cascaded on superconducting hardware. For larger QRAM systems, the researchers see a need for further development in chip layout, coherence times, pulse control, crosstalk suppression, and quantum connections between more distant components. Only such advances will enable the transition from the routing network demonstrated here to more comprehensive memory access.

Quantum Computers

World's First Functional Prototype of a Diamond-Spin Quantum Computer