Since its initial development in the 1970s, DRAM(Dynamic Random Access Memory) has played an increasingly important role in the electronics industry, and its significance has become even more pronounced in the era of generative Artificial Intelligence...
Since its initial development in the 1970s, DRAM(Dynamic Random Access Memory) has played an increasingly important role in the electronics industry, and its significance has become even more pronounced in the era of generative Artificial Intelligence (AI). DRAM is a memory device composed of one transistor and one capacitor in a cell. During operation, the potential difference (ΔV) generated as the charge stored in the capacitor is transferred to the bit line is amplified by a sense amplifier circuit, enabling the distinction between logic states “0” and “1”. A larger potential difference reduces errors during data sensing; therefore, memory cells must be designed to achieve as high a capacitance as possible, since capacitance is directly proportional to this potential difference.
Capacitance is proportional to the capacitor area and the dielectric constant of the dielectric layer, and inversely proportional to the thickness of the dielectric. However, continuous scaling in advanced technology nodes inevitably reduces the capacitor area. In addition, further reduction of dielectric thickness is limited due to leakage current caused by tunneling effects. Consequently, the need for high-k dielectric materials with high dielectric constants has become increasingly critical for achieving large capacitance.
Among binary dielectric materials, TiO2 (Titanium dioxide) is known to exhibit the highest dielectric constant and can achieve values of 90–170 when deposited in the rutile phase under well-controlled thin film growth conditions. However, the rutile phase is difficult to form under typical DRAM process environments, which involve high vacuum and temperatures below 1000 °C. Moreover, its relatively small bandgap of approximately 3.1 eV is disadvantageous in terms of capacitor leakage current. By appropriately engineering the bottom electrode prior to TiO2 deposition, it is possible to simultaneously address both challenges: the difficulty of rutile phase formation under DRAM-compatible conditions and the small bandgap.
In ultrathin films, not only Gibbs free energy but also surface energy can determine phase formation. If the bottom electrode possesses a crystal structure similar to that of rutile TiO2, the formation of the rutile phase may be promoted during TiO2 deposition in order to reduce surface energy. Meanwhile, leakage current in capacitors is primarily governed by tunneling, which depends on the dielectric thickness and bandgap, and by Schottky emission, which is determined by the work function of the electrode. Recently, RuO2 (ruthenium dioxide), which has a rutile crystal structure and a high work function of approximately 5.0 eV, has been extensively studied as a bottom electrode for rutile TiO2. However, ruthenium is expensive and exhibits sublimation at high temperatures, leading to thermal instability and limiting its applicability as a DRAM capacitor electrode.
As an alternative to ruthenium, MoO2 (molybdenum dioxide), which is relatively inexpensive, has a high work function, and possesses a monoclinic crystal structure similar to rutile, can be employed as a seed layer for rutile TiO2. Unlike MoO3 (molybdenum trioxide), which is an insulator, MoO2 contains Mo–O bonds of different lengths within its crystal structure, which cause delocalized electrons and metallic-like electrical conductivity. Therefore, MoO2 can serve as an effective bottom electrode for rutile TiO2.
In this study, MoO2 thin films deposited by atomic layer deposition (ALD) were employed as bottom electrodes for rutile TiO2 to overcome the high aspect ratio challenges of DRAM capacitors. MoO2 thin films were deposited on molybdenum film using a molybdenum precursor and water (H2O), successfully inducing the growth of rutile TiO2 films on top. To minimize leakage current, the deposition and annealing conditions of MoO2 were optimized, resulting in low resistivity and low surface roughness. The excellent properties of the ALD-deposited MoO2 films and the subsequently formed rutile TiO2 films were confirmed as a function of the underlying substrate. This study demonstrates a DRAM-compatible process using low-cost molybdenum-based thin films, highlighting its potential for practical application in future DRAM manufacturing.