Samsung Electronics has begun applying quantum computing technology to produce chips with a 1-nanometer (nm, one-billionth of a meter) line width, equivalent to one hundred-thousandth the thickness of a human hair. Samsung Electronics
Samsung Electronics (005930.KS), the world’s largest memory chipmaker, is moving beyond its 2-nanometer (nm, one-billionth of a meter) process into the ultra-fine angstrom (Å, one-tenth of a nanometer) range, equivalent to about one hundred-thousandth of the width of a human hair. The company is mobilizing group-wide resources to overcome the physical limits of light at the atomic level.
The AI and quantum computer hybrid simulation technology secured by key affiliate Samsung SDS is a leading example of group-level collaboration aimed at breaking through the technological barriers of Samsung Electronics’ lithography process. Analysts say this reflects the “hyper-gap strategy” of Samsung Electronics Chairman Jay Y. Lee, who seeks to solve the challenges of ultra-fine semiconductor processes with quantum computing, a next-generation technology, and to secure differentiated competitiveness by pooling the group’s research and development (R&D) capabilities.
According to industry sources on Monday, Samsung is pursuing plans to apply quantum computing algorithms to “optical proximity correction (OPC),” a computer simulation technology, to overcome the limits of its most advanced 2nm process and next-generation 1nm (10-angstrom) lithography process. Samsung Electronics’ Semiconductor R&D Center has independently developed this technology for more than a decade, but recently an advanced research unit at Samsung SDS, the group’s system integration (SI) affiliate, joined the OPC research, sharing the workload. Samsung SDS plans to begin a proof of concept (POC) for OPC-related technology in the second half of this year.
OPC is a computer simulation technology that predicts and corrects light distortion in advance so that semiconductor circuit patterns are accurately implemented on the wafer through the lithography process. The Process Development Division of Samsung Electronics’ Semiconductor R&D Center operates dedicated OPC teams for each product and process, carrying out optimization tailored to their respective characteristics. In practice, it runs separate technology development teams for segments such as DRAM, NAND flash, logic (computing chips), and foundry (contract chip manufacturing).
As memory and foundry processes become increasingly fine, the importance of OPC is also growing. The foundry process in particular aims to mass-produce the angstrom-class 1.4nm node beyond the current 2nm within four years. Such next-generation processes implement circuit line widths close to the atomic level, and as line widths shrink, the number of variables that must be calculated through simulation increases exponentially. Whereas only two or three variables, such as light source and lens characteristics, previously needed to be reflected, angstrom-class lithography processes require at least 20 conditions to be calculated simultaneously. As a result, observers say a “computation bottleneck” that sharply increases computing time and costs is inevitable.
To resolve this, attempts to apply quantum computers to OPC continue. Quantum computers are characterized by computing with qubits instead of bits, the computing unit of classical computers. Unlike a bit, which processes only one of 0 and 1, a qubit has the characteristics of “quantum superposition,” representing both states simultaneously, and “quantum entanglement,” in which multiple qubits are connected as one. Using this, the vast number of variables arising in the lithography process can be computed in parallel rather than processed sequentially, resolving the bottleneck in angstrom-class processes.
Samsung is accelerating research on hybrid algorithms that use quantum computers together with graphics processing unit (GPU)-based classical computers. In this approach, the qubits and quantum gates of the quantum computer handle lithography simulation, which calculates numerous variables simultaneously, while the GPU takes charge of the repeated computations that post-process the vast amount of data generated in the process. Computation errors arising in quantum computers, which are still at an early stage, will be supplemented in real time with the latest AI-based error correction technology.
Kim Hyung-jong, a professor in the Division of Intelligent Information Security at Seoul Women’s University and vice president of the Korea Society for Simulation, explained, “Samsung’s research this time is an attempt to preemptively secure simulation technology combining AI and quantum computing at a time when uncertainty in the lithography process is growing as semiconductor line widths approach their extreme limits. Since considerable time is needed until quantum computing is commercialized, at the current stage the significance lies more in preempting core original technology and intellectual property (IP) than in securing commercial technology.”
This research direction is also tied to expanded collaboration with Nvidia. Nvidia is currently recruiting personnel to be stationed in Korea to handle sales of semiconductor process simulation solutions. These personnel will work with domestic semiconductor companies to identify bottlenecks arising in the manufacturing process and propose Nvidia GPU-based software tools to resolve them. Samsung plans to continue a “two-track” strategy that combines the use of external solutions such as Nvidia’s with securing its own algorithms.
An industry official said, “While it is difficult to fully replace the external EDA tools Samsung has used, securing original technology such as simulation algorithms in-house has become essential at this point. This research direction, which pools the capabilities of affiliates, is a strategic move to expand its hyper-gap competitiveness by maximizing Samsung’s unique strength spanning the entire process from design to development and manufacturing.”