• Skip to main content
  • LOG IN
  • REGISTER
Coventor_New_LogoCoventor_New_LogoCoventor_New_LogoCoventor_New_Logo
  • COMPANY
    • ABOUT
    • CAREERS
    • PRESS RELEASE
    • PRESS COVERAGE
    • EVENTS
  • PRODUCTS
    • SEMulator3D®
      Semiconductor Process Modeling
    • CoventorMP®
      MEMS Design Automation
      • CoventorWare®
      • MEMS+®
  • SOLUTIONS
    • SEMICONDUCTOR SOLUTIONS
    • MEMS SOLUTIONS
  • RESOURCES
    • CASE STUDIES
    • BLOG
    • VIDEOS
  • CONTACT
  • SUPPORT
Contact Us
✕
  • Home
  • Coventor Blog
  • Challenges in 3D NAND Flash Processing
Sneak Peak: New Capabilities for Micro Scanning and Projection Mirrors
May 6, 2014
What will the next 30 years of MEMS bring?
June 26, 2014

Challenges in 3D NAND Flash Processing

Published by Coventor at May 23, 2014
Categories
  • Coventor Blog
Tags
  • 3D NAND
  • Technology Reviews
3D NAND flash memory array, based on TCAT [1], with 16 cells per string, top gate-select layer and bottom source-select layer.

Figure 1. 3D NAND flash memory array, based on TCAT [1], with 16 cells per string, top gate-select layer and bottom source-select layer.

With 2D planar NAND flash hitting scaling issues at sub-20nm technology nodes, 3D NAND flash has become all the rage. Instead of restricting memory cells to a single plane and scaling the devices horizontally, memory cells can also be stacked vertically, allowing high cell density while side-stepping scaling issues (for now). Major NAND flash manufacturers have each developed their own designs and technology for 3D NAND flash, and with the addition of vertical cell stacking, new issues in 3D process integration arise.

For instance, in Samsung’s Terabit Cell Array Transistor (TCAT) technology [1], a memory cell array is formed of NAND flash strings with vertically-oriented channels and word lines arranged in planes. Of particular interest is the gate integration scheme: TCAT uses charge-trapping (SONOS/TANOS) with metal replacement gates, the combination which is expected to result in faster erase speed, wider threshold voltage margins, etc. The cell gates are created using a sacrificial nitride layer combined with a damascene process: the entire stack of SiO2/SiN layers is etched (“word line cut”) after staircase formation, then nitride is removed through wet etching with hot phosphoric acid, leaving behind gaps separated by the oxide. These gaps are then filled with dielectric and gate metal to create gate-all-around structures.

3D NAND flash memory array, based on TCAT [1], with 16 cells per string, top gate-select layer and bottom source-select layer.
Figure 1. 3D NAND flash memory array, based on TCAT [1], with 16 cells per string, top gate-select layer and bottom source-select layer.


TCAT replacement gate processing was investigated using SEMulator3D, with the process integration based on publicly available sources [1-2]. In the first attempt at a 3D model (Figure 2a), the sacrificial SiN removal resulted in unsupported SiO2 beams, not unlike cantilever structures that result from release etching in MEMS processing — but not commonly found in memory or logic devices. This initial model looked untenable and did not make much sense, particularly if more cells (24-32 total) are to be added.

After stumbling on some Samsung patents [3] and evaluating other process flows, the model was revised to include a capping dielectric (oxide) to cover the staircase during word line cut and before SiN removal (Figure 2b). While the cap holds the oxide layers together and avoids free-standing oxide beams, the word line cut required some adjustment: both the oxide/nitride stack and the capping dielectric should be etched fully, so that an optimal etch process requires control of the etch selectivity of oxide to nitride, and vice versa.

(a) First iteration of the TCAT model after sacrificial SiN removal results in unsupported oxide beams. (b) Second iteration of the model now has top dielectric to hold the oxide layers together during SiN removal.
Figure 2. (a) First iteration of the TCAT model after sacrificial SiN removal results in unsupported oxide beams. (b) Second iteration of the model now has top dielectric to hold the oxide layers together during SiN removal.

Gate stack processing is but one of the many process integration issues in 3D NAND flash processing. Other challenges with vertically-stacked memory cells include film deposition uniformity and high-aspect ratio etching. And with flash memory design being taken to its extremes, it is an exciting time to be a memory process integrator!

References

[1] J. Jang et al, “Vertical cell array using TCAT(Terabit Cell Array Transistor) technology for ultra high density NAND flash memory,” VLSIT 2009.
[2] W. Cho et al, “Highly reliable vertical NAND technology with biconcave shaped storage layer and leakage controllable offset structure,” VLSIT 2010.
[3] B. Jang et al, “Methods of manufacturing three dimensional semiconductor memory devices using sub-plates” US Patent 8,697,498,.

Share
Coventor
Coventor

Related posts

Figure 1 displays a single cell of a conventional DRAM that consists of 2 Word Lines (WLs), a Bit Line (BL) and 2 Storage Node Contacts (SNC) in Figure 1(a). There are 3 images in the figure. The Saddle Fin is produced during the WL etch step (prior to WL metal deposition) and is located below the cell wordline (Figure 1(b), right center inside a yellow dotted circle). The Saddle Fin structure can be seen in detail by making a vertical cut in the wordline direction (Fig.1(b), right). During device simulation, the Saddle Fin performance can be measured by virtually cropping a transistor and adding ports at the Gate, Source and Drain after an SNC Process (Fig.1(c), showing the gate, source and drain).
May 30, 2023

Improving DRAM Device Performance Through Saddle Fin Process Optimization


Read more - Improving DRAM Device Performance Through Saddle Fin Process Optimization
Figure 6 (left to right): Different profiles using pattern dependence for the antenna and sharp head shapes. a) Antenna shape with POR flow (b) Antenna profile with a gate CD of 26nm (c) Sharp head profile with a gate CD of 28nm (d) Sharp head profile with an etch.

Figure 6 (left to right): Different profiles using pattern dependence for the antenna and sharp head shapes. a) Antenna shape with POR flow (b) Antenna profile with a gate CD of 26nm (c) Sharp head profile with a gate CD of 28nm (d) Sharp head profile with an etch.

April 13, 2023

The Impact of Metal Gate Recess Profile on Transistor Resistance and Capacitance


Read more - The Impact of Metal Gate Recess Profile on Transistor Resistance and Capacitance
Figure 1a (left) displays the process of performing Physical Vapor Deposition (PVD), including Cu bombardment and filling of voids. Figure 1b (right) displays the process of performing Ion Beam Etch (IBE), including ion beam bombardment, mask shadowing and etch regions.

Fig 1a Physical Vapor Deposition (PVD); Fig 1b Ion Beam Etch (IBE)

March 22, 2023

A Deposition and Etch Technique to Lower Resistance of Semiconductor Metal Lines


Read more - A Deposition and Etch Technique to Lower Resistance of Semiconductor Metal Lines
Left to right: SEMulator3D virtual structures of NON, Low K and Airgap spacers for a DRAM cell, with highlighted SiO2, Polysilicon, Silicon, Si3N4, TIN and W layers

Fig. 1: (a) NON, (b) Low k and (c) Airgap spacer

February 28, 2023

A Comparative Evaluation of DRAM bit-line spacer integration schemes


Read more - A Comparative Evaluation of DRAM bit-line spacer integration schemes

Comments are closed.

Product Information

  • Product Offerings
  • Technical Support & Training
  • Licensing
  • System Requirements

Resources

  • Blog
  • Case Studies
  • Videos
  • 2018 MEMS Design Contest

Company

  • About
  • Press
  • Partners & Programs
  • Contact
© Copyright Coventor Inc., A Lam Research Company, All Rights Reserved
Privacy Policy • Terms of Use
Contact Us
  • LOG IN
  • REGISTER