Our increasingly connected and ever “smarter” world generates increasing amounts of data, putting pressure on manufacturers who face new technical challenges in delivering the increasing capacity required for processing and storage. The ALD Tungsten process is helping 3D NAND manufacturers overcome the technical challenges of producing memory chips with higher storage capacity.
3D NAND is a type of non-volatile memory, which means it holds onto data when power is turned off. 3D NAND is found in phones, tablets, PCs and memory sticks, as well as in cars and enterprise servers.
As consumer needs become more sophisticated, greater storage capacity is needed. For example, in addition to taking photos on a phone, we expect to shoot high resolution video and store it on the device. Instead of downloading one movie to a tablet, we download a box set so we can binge-watch our favorite shows.
Devices that understand natural speech, like Siri and Alexa, and those that perform facial recognition, require large amounts of memory both on the devices and in servers. Cars have multiple cameras and hundreds of sensors, requiring large amounts of memory and processing power to deal with the data generated. This will only increase as we move towards driverless vehicles.
As AI (artificial intelligence) and machine learning become more embedded in manufacturing, medical and other industries, there’s even more pressure on memory capacity
The 3D NAND challenge is depositing and etching ever taller, deeper, structures. A channel hole etched for 90+ layer 3D NAND with an aspect ratio of 40 to 1 (left side). In comparison, the Burj Khalifa, the tallest structure in the world, has an aspect ratio of 9 to 1 (right side).
A 3D NAND chip consists of multiple layers, like floors in a building. Each layer has an array of storage cells, each of which generally hold up to three bits of data. To increase capacity on a memory device, manufacturers need to add more layers. All the cells in each layer are electrically connected using tungsten, which must be applied evenly throughout the device – the more layers and cells you have, the more difficult this becomes.
In device interconnect with only a few layers, copper is used and deposited by electroplating from a liquid. With 3D NAND storage stacks, the horizontal layering and large number of layers make this impossible, so tungsten is used as it can be applied as a gas, better able to reach down through all the layers in the device.
One way of doing this is by using CVD (chemical vapor deposition), which deposits the tungsten and removes any by-product, also as gas, in a continuous process lasting many tens of seconds.
Key challenge for wordline tungsten fill is to ensure void-free uniform fill of every layer as layer count continues to increase.
The next-generation solution is to use ALD (atomic layer deposition). ALD supplies different required gases in alternating cycles, controlling specific reactant molecules precisely within each cycle, unlike CVD which supplies all the molecules together continuously.
ALD is like painting a surface layer by layer to obtain an even coating. It’s highly precise, producing smooth, void-free layers that conform to the deposition surface. This means that semiconductor device companies can build flash memory structures with more layers, slimmer wordlines, smaller cells and greater capacity.
Since the inception of this solution, advanced memory chipmakers have come to rely on the technology, enabling them to quickly move to next capacity devices.