Response | Horizon: Manufacturing requirements for new memory

Applied | Vision Bring you the latest blog post by Mr. Gill Lee, General Manager of Storage Technology, Semiconductor Products Division, Applied Materials , to help you pay attention to the development trend of emerging memory technologies.

From the official website of CES ASIA 2019

At the just concluded CES Asia 2019, technologies including 5G , AI , and VR have become the new trends in the field of consumer electronics in Asia this year. The advancement of AI and the full penetration of 5G have brought many challenges to big data and data storage. Let's read the following blog post to explore the manufacturing requirements for new memory in the future.

Manufacturing requirements for new memory

By Gill Lee

General Manager of Storage Technologies, Semiconductor Products Division, Applied Materials

After decades of research and development, three new types of memory, magnetic RAM (MRAM), phase change memory (PCRAM) and resistive RAM (ReRAM), are about to be commercialized. This is an exciting time for the semiconductor and computing industries. All three emerging memories will be built with new materials and require breakthroughs in process technology and manufacturing.

In reality, neither existing nor emerging memories have all of the above characteristics. Chip and computer designers will continue to dig deeper and use various memory technologies to achieve their goals. Let's take a look at the characteristics of these three emerging memories and predict their application directions in the industry.

MRAM has many advantages, including fast random read and write speeds, non-volatility, and low power consumption (magnetic technology). In addition, MRAM has great potential for cost reduction because only three additional masks are needed to embed the memory array into the back-end interconnect layer of the chip.

Although MRAM is slightly slower than SRAM, it is sufficient as working memory for many embedded computing applications and is also sufficient to meet the needs of third-level cache.

In fact, several leading logic/foundry companies have announced that their system-on-chip designs using embedded MRAM are in early production. Specifically, as the demand for artificial intelligence (AI) computing in Internet of Things (IoT) devices increases, MRAM has become the memory of choice for these devices. In time, MRAM may prove capable of meeting the high temperature requirements of the automotive sector in the embedded market.

From a technology and manufacturing perspective, the biggest challenge facing MRAM is how to accurately deposit numerous thin film stacks to form a magnetic tunnel junction (MTJ), which is the basic programming element used to represent digital data. Many metal and insulating layers need to be deposited using physical vapor deposition (PVD) in a dust-free, high vacuum environment below one atmosphere. Each layer requires precise control and measurement. Most importantly, the core magnesium oxide (MgO) film layer needs to be accurately deposited in a form close to a crystalline arrangement, supplemented by strict control, because even a single atomic height variation can greatly affect performance and reliability.

PCRAM is based on "phase change" materials, using heat as the programming mechanism to transform the highly amorphous material arrangement into a crystalline arrangement.

PCRAM supports random access, is lower cost than DRAM, is 3D scalable, and is non-volatile. As a result, industry-leading memory companies are evaluating PCRAM for applications such as non-volatile DIMMs to replace some DRAM-based DIMMs and high-end solid-state drives.

PCRAM requires precise deposition of multiple material layers to form key structures. Although PCRAM layers are not as thin as MRAM, their materials are extremely susceptible to impurities, so a PVD process technology that can handle multiple materials and prevent particles and impurities from entering is required. After the PCRAM stack is formed, plasma etching technology is used to form each memory cell, and packaging is used to protect the exposed phase change material.

Scalability is the key driver of PCRAM cost planning. Using 2D scaling, critical dimensions can be significantly reduced, with half pitch as low as 20nm, and 15-16nm designs are coming soon. 3D scaling of PCRAM has a broader prospect: its initial design uses a two-layer stack, but the technology blueprint shows that 4-layer or even 8-layer stacking is possible in the future.

ReRAM technology comes in many forms. Some ReRAMs are embedded in metal filaments within an ion bridge, while others create oxygen vacancies in the matrix. The bits of information are stored in a resistive material (usually a metal oxide), which is programmed by applying an electric current to the resistive material and read by sensing different resistance levels. Another feature of ReRAM is that a variety of materials can be used to implement the technology.

To date, ReRAM has shown significant limitations in terms of endurance, and more work is needed to identify its failure mechanisms in order to improve materials and manufacturing techniques to ensure field reliability. Assuming these challenges can be addressed, ReRAM has the potential to further increase the density and reduce the cost of storage applications.

Today’s mainstream memories, such as DRAM, flash memory, and SRAM, have taken decades to mature and are still evolving. However, these memories are becoming increasingly difficult to scale, especially in terms of performance, power, and cost. At the same time, the application areas of computing technology continue to expand, and many people envision a future world where tens of billions of low-cost computers will be embedded in various industrial and consumer products, forming the Internet of Things, and leading to an explosion in the amount of data that needs to be stored in public and private cloud data centers. Emerging memories such as MRAM, PCRAM, and ReRAM are expected to achieve higher performance, lower power consumption, and lower cost, and are the preferred supplements, and in some cases, even alternatives to today’s mainstream technologies.

Emerging memory technologies will use new materials and structures formed by precise thin film deposition, measurement, etching/single and packaging techniques under clean conditions. When manufacturing these next-generation memories, changes in even a few atoms can have a big impact. Applied Materials is committed to providing customers with new materials engineering solutions, and we plan to share more information in the coming weeks and months, so stay tuned.

What are your thoughts and insights on the future of memory and storage?

In the future of 5G, which new type of memory will shine?

Welcome to express your views in the comment section and contribute to the future of technology.

Ying| Horizon: Outlook for the electronics industry in the next 25 years

2019-04-19

Applied Materials and IBM to collaborate on AI chip technology

2019-02-22

Ying| Horizon: New materials drive progress in the era of artificial intelligence (Part 2)

2018-12-07