• Start
  • Research print
  • Materials for Micro- and Nanoelectronics print
  • Former Projects print
  • DRAM print
  • Introduction


Dynamic Random Access Memory (DRAM) Cells with integrated high-k dielectrics


There are a number of solid-state memory technologies that are currently being mass-produced or are being aggressively pursued in the research laboratories [1]. The vision in the field is mass production of the so-called “universal memory cell”, integrated in the conventional Si process technology. The “universal memory cell” is a concept which unifies the advantages of the various solid-state memory approaches nowadays manufactured, namely non-volatile data retention, high scalability, quick read and write times, long endurance and low power characteristics. Flash [2] as well as resistive, magnetoresistive and ferroelectric random access memory cells (RRAM, FeRAM [3] and MRAM [4]) are among the most promising approaches nowadays discussed in the literature. However, in particular the storage density of these technologies is far behind the integration levels of the well-established memory technologies in semiconductor devices, such as non-volatile magnetic hard disc drives and volatile semiconductor memories (static and dynamic random access memory cells (SRAM and DRAM)). Therefore, until any competitive universal memory cell technology will appear on the market, the thirst for storage space will force technology drivers to fully exploit the potential of these classical memory approaches.


Research in our group is focussed on the development of dielectric layers for DRAM cells. DRAM is a 1 transistor – 1 capacitor cell technology. Scaling DRAM cells requires reducing the dielectric layer cell in the transistor as well as in the capacitor. It turned out in the past that, due to dielectric breakdown issues, thickness reduction of the dielectric layer in the capacitor is particularly challenging. The scaling rule requires to reduce the thickness of the dielectric proportionally to L2 (L: feature size of the technology node), but this is not possible anymore since the late eighties (4-Mbit generation). Instead, 3D capacitors architectures were integrated, either in form of trenches in the bulk Si wafer (Qimonda) or stacks on top of the wafer (Samsung). Figure 1 shows as a vertical cross-section of the fully integrated 512Mb deep-trench DRAM using the 58 nm technology of Qimonda (left), and a schematic representation of a deep-trench DRAM cell (right).

Fig. 1: Vertical cross-section of the fully integrated 512Mb deep trench (DT) DRAM using the 58 nm technology of Qimonda (left), schematic representation of a deep-trench DRAM cell (right).  

For future Gbit DRAM generations manufacturers will have to integrate advanced high-k dielectric materials, fulfilling the following, very challenging requirements [5-7]:

          1) Equivalent oxide thickness (EOT) < 1 nm

          2) low leakage current (10-7 – 10-8 A / cm2 at 1 V)

          3) excellent time dependent dielectric breakthrough (TDDB) characteristics

          4) good step coverage for 3-D cells

          5) good microprocessing capability and thermal stability

          6) small polarization reversal at reversing biases


To fulfill the above requirements Praseodymium- and Hafnium-based high-k dielectric layers are currently being developed at IHP. These research activities are performed in the framework of the MEGA EPOS project supported by Federal Ministry of Education and Research (BMBF).

The building and the infrastructure of the IHP were funded by the European Regional Development Fund of the European Union, funds of the Federal Government and also funds of the Federal State of Brandenburg.