NAND nonvolatile flash memory has been a breakthrough innovation in the storage industry for many years. With the capacity of 2D NAND reaching the limit and the transistor becoming smaller and smaller, the programming time of NAND becomes longer and the number of erasure becomes less. 3D NAND, which can stack memory particles together, has a long history of development As the times require, it can support higher storage capacity in smaller space content, which has great value in the era of massive data storage. < / P > < p > relying on the advanced technology of 3D NAND, the oxide layer is getting thinner and thinner, facing the problem of reliability and stability. How will 3D NAND develop in the future? How to correctly judge the overall efficiency of a 3D NAND? < / P > < p > at the flash memory summit in 2020, joengdong Choe, a senior technical researcher at techinsights, gave a speech on the future of 3D NAND and other emerging memories. Techinsights is a semiconductor product analysis company that includes flash memory. < p > < p > Choe introduced the flash memory roadmap of the world’s leading storage companies from 2014 to 2023, including the 3D NAND technology development routes of Samsung, Jiaxia (formerly Toshiba storage), Intel, micron, SK Hynix and Changjiang storage. < p > < p > after that, Samsung has successively launched 48 layer, 64 layer and 92 layer v-nand, and launched 128 layer products this year. < / P > < p > compared with other companies, domestic companies started 3D NAND later. It was not until the end of 2017 that Changjiang storage launched the first real 32 layer 3D NAND flash memory in China. However, Changjiang storage has developed rapidly. Based on its own xtacking architecture, it has directly leapt from 64 layers to 128 layers. In April this year, it announced the launch of 3D NAND flash memory with 128 layer stack, which has entered the first echelon in terms of the number of flash memory layers. Recently, at the 2020 Beijing Microelectronics International Symposium and IC world academic conference, CEO Yang shining of Changjiang storage Co., Ltd. publicly stated that it will take three years for Changjiang storage to go through the road of international manufacturers for six years, and the current technology is at the world-class level, and the next step is to solve the problem of production capacity. < / P > < p > it is worth mentioning that in the latest Q3 global flash memory report released by China’s flash market a few days ago, the six original flash memory manufacturers of Samsung, Jiaxia, western data, SK Hynix, micron and Intel accounted for 98.4% of the global market share. In the remaining 1.6% of the market, the revenue of Changjiang storage in Q3 is expected to exceed 1%, ranking seventh in the world. < / P > < p > however, Choe points out that it may be misleading for the public to focus on the number of flash memory layers, because the actual number of word lines (active layers with storage units) can vary greatly. For example, other layers can be used as pseudo lines to help alleviate problems caused by higher layers. According to this criterion, Samsung has the best design. However, Samsung does not use multiple layers or stacks, unlike other manufacturers’ current flash memory using “string stack”. < / P > < p > the design of Changjiang storage is special because it has some circuits on the top of flash memory, while CMOS are made in larger process nodes before they are connected to flash memory. Choe thinks the technology has potential, but there are yield problems. < / P > < p > in addition, different companies use different technologies, such as charge trap flash (CTF) and floating gate (FG). < / P > < p > CTF uses silicon nitride to store electrons, rather than the typical doped polysilicon in traditional FG. Specifically, FG stores electrons in the gate, and defects will lead to a short circuit between the gate and the channel, consuming the charge in the gate. That is, every time the data is written, the grid charge will be consumed once. When the grid charge is consumed, the flash memory can no longer store data. The charge of CTF is stored on the insulation layer, the insulator surrounds the channel, and the control gate surrounds the insulator layer. Theoretically speaking, when writing data, the charge is not consumed and the reliability is stronger. < p > < p > Choe pointed out that among current memory chip companies, Intel and micron have been using traditional floating gate technology, while other manufacturers rely on charge capture flash design. Meguiar did not replace the new technology until 176 layers were released recently. Intel’s QLC can maintain better wear performance when using floating gate technology, but this will also affect the durability, reliability, scalability and other performance advantages of its flash memory. In his speech, < / P > < p > Choe mentioned in his speech that the separation gate structure or separation cell structure technology which will be used by armored man in the future is also very interesting. It can directly double the density of memory, and because of the semicircular shape of the separation cell structure, it has a particularly strong floating gate structure, which has stronger durability. < / P > < p > Choe predicts that as the number of platforms or stacks increases (currently up to two), the number of flash memory layers will continue to increase, and the storage capacity of each flash chip will also increase accordingly. Choe believes that this, like other technologies, such as silicon through-hole (TSV), stack packaging (Pop / popop), and migration to 5lc / PLC, will point to more than 500 layers and 3 TB bare chips in the next decade. < / P > < p > in addition, Choe explained in detail that the cost of flash memory is calculated according to how many cents per GB, which means that the architecture of 3D flash memory will become cheaper and cheaper in the future, but the price of 2D flash memory is still expensive, even many times more expensive than 3D flash memory. < / P > < p > when it comes to the advancement of cutting-edge flash technology, Choe thinks that cutting-edge flash memory is always the first to enter mobile and embedded products. For example, 5g mobile phones are the main driving force at present. He also pointed out that there are still some application markets for 2D flat flash memory, which is usually regarded as a low latency SLC as a storage type memory (SCM) alternative for 3D xpion, such as optone or the recently released X100 by micron, although X100 is not common in the consumer market. SpaceX beat blue origin and Northrop Grumman to win a $316 million air force contract