My prior 3D NAND post explained how Toshiba’s BiCS cell works, using a silicon nitride charge trap to substitute for a floating gate. This post will look at an alternative technology used by Samsung and Hynix which is illustrated in the first graphic, a diagram Samsung presented at a technical conference. This cell also uses a charge trap.
Let The Memory Guy warn you, if the process in my prior post seemed tricky, this one promises to put that one to shame!
Part of this stems from the use of a different kind of NAND bit cell. You can shrink flash cells smaller if you use a high-k gate dielectric (one with a high dielectric constant “k”) since it provides a very high capacitance (higher coupling) between the charge trap and the control gate. Capacitance decreases with process shrinks since a shrink reduces the plate dimensions of a cell’s capacitors; changing the dielectric from silicon dioxide to a high-k material helps offset this decrease.
High-k dielectrics perform better with a metal control gate than with the polysilicon gate used in conventional flash cells, so some high-k designs deposit a tantalum control gate on top of the high-k dielectric.
Samsung has chosen a cell type called TANOS, which stands for Tantalum-Alumina-Nitride-Oxide-Silicon. Even this is a shortened term, since the “Tantalum” is really tantalum nitride, “Alumina” is short for an aluminum-based high-k gate dielectric (usually Al2O3), “Nitride” is short for silicon nitride, and “Oxide” is an abbreviation for silicon dioxide.
The TANOS structure combines a silicon nitride charge trap with a high-k gate dielectric and a tantalum control gate. According to Samsung this cell provides certain advantages over the BiCS cell, including faster erase speed, wider threshold voltage margins, and better data retention characteristics.
The TCAT (Terabit Cell Array Transistor) structure described in this post was first presented by Samsung at the IEEE’s VLSI Technology Forum (VLSIT) in 2009. Samsung introduced a product based on this technology last August that the company calls V-NAND for “Vertical NAND.”
The series of figures in the second and third graphics illustrate the process steps involved in manufacturing a TCAT string. It’s more extensive than the BiCS series, so I will ask for the reader’s patience.
I will again advise readers to right-click on the graphics and open them in new windows, so that you can “Alt-Tab” back and forth between this blog post and a larger and clearer rendition of the graphic.
Similar to the BiCS design, the process starts by depositing multiple layers onto the surface of a CMOS wafer, (Step a in the graphic) but rather than alternating conductive polysilicon and insulating silicon dioxide layers the TCAT process begins with alternating layers of silicon nitride and silicon dioxide. That may surprise you, since both of these materials are insulators, but we will see that the silicon nitride layer is sacrificial – it serves as a placeholder and will be removed so that the space it consumes can be filled with the gate materials.
Samsung calls this a “gate replacement process.”
As with the BiCS process, a hole is etched through all the deposited layers to the substrate. This is shown as Step b) in the second graphic.
The hole is next lined with polysilicon for the channel, skipping the deposition of all the ONO layers used in the BiCS approach. This is shown as Step c) in the graphic. In Step d) the center of the hole is filled with silicon dioxide, which (I am told) gives better sub-threshold characteristics and threshold voltage distributions than a solid polysilicon channel would. This has been dubbed a “Macaroni Channel.” Although I am not aware of its use in BiCS, the macaroni channel might also appear in that approach once it is introduced.
So how do they make the gate, now that the hole has been plugged? This is the mind-bending part! In Step e) slits are etched – the same as in BiCS (although I didn’t show these in the prior post’s graphic) to separate the columns from each other. All of the processing from here on is performed through these slits. (Note that the layers of oxide and nitride on either side of the drawing are actually attached to adjacent pillars that are not fully shown.)
In Step f) the sacrificial silicon nitride layers are selectively etched away from the sides, leaving a structure resembling a very narrow tower with fins.
The entire tower is then coated with a silicon dioxide tunnel dielectric layer (Step g in the third graphic), followed by a silicon nitride charge trap (Step h) and then the alumina high-k gate dielectric (Step i).
Finally most of the slit and all the remaining gaps in the finned pillar are filled with the tantalum nitride control gate material (Step j) which is subsequently etched out of the center of the slit along with the high-k dielectric on the walls (Step k).
What remains is a vertical NAND string of TANOS transistors: The tantalum nitride plug (gray), an alumina high-k gate dielectric (green), a silicon nitride charge trap (yellow), the silicon dioxide tunnel dielectric (pale blue), and the polysilicon channel (red).
To go back to the layer cake analogy I used in the prior post, we have just built a layer cake, removed all of the frosting, and replaced it with a different and more complex frosting.
The BiCS approach described in the prior post made the transistors from the walls of the hole and worked toward the center, starting with the control gate and ending with the channel. The TCAT design starts with the channel at the center and builds the transistors from the middle out, ending with the control gate. That’s why TCAT’s process is sometimes referred to as “Gate Last” and the BiCS approach is called “Gate First.”
Either way, since the gate surrounds the channel both BiCS and TCAT are called “Gate All Around” structures. Since one of the problems presented by planar scaling is the fact that the coupling capacitance between the control gate and the rest of the transistor diminishes with shrinking processes, a big benefit of the Gate All-Around approach is that it provides a lot of surface area to couple the control gate to the charge trap.
I am told that one strength of the TCAT approach is that it is less constrained in the horizontal plane than is the BiCS approach so the holes may be made smaller, resulting in a smaller die size. This is because of the number of layers deposited into the hole: The hole in BiCS is filled with three concentric tubes to form the oxide-nitride-oxide layers and then is plugged with a polysilicon cylinder. The TCAT hole is filled with a single polysilicon tube and an oxide cylinder. This is expected to allow TCAT to scaled to tighter lithographies than BiCS.
The next post will show how the control gates on all those internal layers are connected to the support logic, a simple job in a planar NAND that becomes a daunting challenge in 3D.
This post is a part of a series called What is 3D NAND? Why do we need it? How do they make it? that was published in weekly segments during the fourth quarter of 2013 on The Memory Guy blog. The different sections are listed below, with a hot link to each section.
- Why Do We Need 3D NAND?
- What Is a 3D NAND?
- Making a Vertical NAND String
- An Alternative Kind of Vertical 3D NAND String
- How Do You Access the Control Gates?
- Benefits of Charge Traps over Floating Gates
- How Do You Erase and Program 3D NAND?
- 3D NAND’s Impact on the Equipment Market
- Who Will Make It and When?
Click on any of the above links to learn more about 3D NAND technology.