Mechanisms and Selectivity during Atomic Layer Deposition of Germanium Chalcogenides for Storage Class Memory Applications

Memory and data storage technologies can be classified in a hierarchical way, with the CPU-internal DRAM memory on top, offering extremely fast access to a very limited number of data but at a high cost/byte. At the bottom there are Hard Disk Drives, more and more replaced by Solid State Drives using NAND flash technology, for cheap and reliable storage of tons of data but quite a bit slower. Between these two extremes, a new class is emerging, i.e. Storage Class Memory (SCM), offering performance, endurance, capacity and cost in between these of DRAM and NAND flash. Unlike DRAM, SCM is persistent in nature and retains data written to it across power cycles. Compared to NAND flash, SCM is orders of magnitude faster for read/write operations. It delivers these benefits over NAND flash at a much lower cost/GB as compared to DRAM. SCM is considered an important building-block for high-speed data transfers, next-generation in-memory/near-memory computing and scale-out clusters for computation and storage. SCM cells contain a Resistive RAM (RRAM) device as memory element, and a selector device that allows to target one cell only for read/write in a so-called crossbar array. After many years of R&D, germanium chalcogenides emerged as strong candidate materials for both devices. Most important however is that there is no widely accepted deposition technology available yet, meeting all material specifications and fit to grow thin layers of these materials in advanced 3D architectures with excellent step coverage and uniformity of thickness and composition. This poses a formidable barrier to introducing this technology in High Volume Manufacturing. The most promising candidate to break down this barrier is Atomic Layer Deposition (ALD), offering low processing temperature, excellent 3D step coverage and flexible precursor choice. Today, initial results have been reported on ALD processes for GeSe (for selector devices) and Ge2Sb2Te5 (a phase-change material for RRAM). The general aim of this PhD project is to generate fundamental insights into the nucleation and growth mechanisms during ALD of germanium chalcogenide materials. First, it is clear that we will need more complex ternary and even quaternary compounds for the stabilization of these materials and for boosting the memory cell performance. We will therefore investigate the incorporation of extra elements during ALD by using new precursor reactions. We will investigate the impact of additional precursor reactions on the growth mechanisms, to design deposition processes for complex materials with well controlled composition and stability. A second new aspect is the selectivity of germanium chalcogenide ALD processes. The concept is to grow germanium chalcogenide materials from the bottom up, only where needed in the advanced 3D structures, as this could greatly simplify the patterning process. We will therefore investigate the surface dependence of the germanium chalcogenide ALD processes, surface passivation and activation approaches, and study the growth mechanism in nanoscale structures to design selective ALD processes. The application of these understandings and the performance benefit will be tested in memory devices in close collaboration with the memory team at imec. We will leverage imec’s 300mm state of the art R&D line and advanced characterization techniques to deliver industrially relevant research. Important to note is that this work will run in close collaboration with the world’s top-notch chemical suppliers as well as industrial tool manufacturers.