Doped Ta₂O₅ and mixed HfO₂–Ta₂O₅ films for dynamic memories applications at the nanoscale

Abstract

The doping of Ta₂O₅ films with a proper element or its mixing with another high-k dielectric as a breakthrough to extend the potential of Ta₂O₅ toward meeting the criteria for future technological nodes is discussed. Essential issues in the engineering of storage capacitor parameters for dynamic memories based on Ti-doped Ta₂O₅, Hf-doped Ta₂O₅ and mixed HfO₂–Ta₂O₅ layers are presented. The benefits and the disadvantages of these modified Ta₂O₅ stacks are discussed.

Introduction

As scaling becomes less effective in boosting the performance of technologies for nanoscale nodes, high-k materials have attracted a great deal of interest to meet the formidable challenges for replacing the SiO₂-based insulators. The demise of SiO₂ scaling have been predicted since about 1980, considering the high leakage current as a result of direct tunneling. Thus dielectrics with permittivity higher than SiO₂ have been proposed as possible solution. The studied at that time high-k layers demonstrated severe crystallization induced current and poor reliability. So the research on high-k films was phased out as Si-oxinitride was successfully implemented. In the mid-1990s interest in high-k alternatives was renewed and focused on TiO₂ and Ta₂O₅ which were inherited from DRAM capacitor dielectric research. Investigations quickly converted on another high-k insulators, mainly metal oxides and their silicates. The requirements for the alternatives include high-k value, acceptable band gap, good chemical stability in contact with Si and preferentially being amorphous even after post-deposition temperature processing (to avoid increased surface roughness and leakage current due to the crystallization effects). The ultimate high-k choice depends on both the technological node and the specific device applications (MOSFETs or DRAMs). Due to the intensive studies of high-k dielectrics in the last decade, the knowledge of their material and electrical properties has reached some stage of maturity. As a result a consensus about the most suitable dielectrics for the various microelectronics applications exists. After years of development HfO₂ was selected as the strongest high-k candidate for nanoscale MOSFETs. DRAMs are a driving force for high-k dielectrics integration and the first application of these materials has been marked with fabrication of memories with Ta₂O₅ [1]. High storage ability and a low leakage current are the essential requirements for a material to be used as a dielectric in DRAM capacitors. Given these requirements, Ta₂O₅ has been recognized as the leading candidate for nanoscale DRAMs, (a stored charge several times higher than that of other candidates). It should be emphasized that the scaling of equivalent oxide thickness (EOT) for memory applications is more aggressive and accordingly a specific problem appears for nanoscale dynamic memories (so called minimum cell capacitance value, which should not be less than ∼25 fF/cell, in order to guarantee (stable memory cell operation). Besides its advantages and merits Ta₂O₅ has also drawback typical of the most high-k dielectrics [2], [3] (a trap-rich material with unstable interface with Si resulting in an inevitable lower-k SiO₂-like interface layer at the Si substrate, which layer preserves a good quality of Si interface but reduces the effective k value and compromises the EOT. Due to the unstable interface with Si, Ta₂O₅ tends to phase separate into Si-oxide(s), Ta-oxide(s) and Ta-silicate(s). Since the memory capacitors require mainly a low leakage current and a high capacitance to ensure high charge storage, the interface quality is not that critical as in the case of MOSFETs. Yet, some control of the interface is necessary in order to limit interface reaction and to keep the total capacitance high. In the search for a solution of the problems with the bulk traps and the interface layer, different approaches for Ta₂O₅ have been used: optimization of the technology for its fabrication, annealing treatments, deposition on nitrided Si, or their appropriate combinations. The two key breakthroughs that have enabled the progress in high-k are the understanding of their electrical behavior which is different from that of single SiO₂ layer, and realizing the necessity of scaling of high-k layer thickness for future device generations. The thinning of high-k layer, however, is only a temporary measure that works for a limited range of EOT. The further reduction of physical thickness of high-k films can result to the same problems as SiO₂ scaling. Although encouraging results with high-k dielectrics have been achieved, further scaling of these materials for EOT ≪ 1 nm faces severe challenges. The right solution at the moment is optimization of both the high-k composition and the interfacial layer, considering also the effect of the metal gate. A path to control high-k composition is the doping of certain high-k dielectric with a proper element or mixing of various high-k films. Accordingly, the doped high-k materials and the composite films of two or more high-k films were recognized as a next step in the efforts of extending the potential of pure high-k dielectrics [1]. The concept of doping is the idea that the dopants act as network modifiers, compensate the oxygen vacancies in metal oxides, and by this way reduce the leakage current and trap density [4]. It was supposed that by mixing of high-k dielectrics it would be possible to engineer the electrical properties of these materials – combining the favorable properties of each of them while suppressing their individual disadvantages. The results in general confirm the expectation for improved properties of chemically modified high-k dielectrics compared to the pure ones. Over the past years the focus on Ta₂O₅ was also shifted on the doped Ta₂O₅ with different elements (usually metal agents) and on mixed films with Ta₂O₅ as a main component [5], [6], [7], [8], [9], [10], [11], [12], [13]. Since the structure of Ta₂O₅ does not change significantly with the addition of another oxide it is a good base oxide for doping process [14]. The results on doped Ta₂O₅ reveal indeed an improvement of the stack characteristics [5], [6], [7], [8], [9], [10], [11], [12], [13], [14], [15], [16], [17], [18], [19], [20], [21], [22], [23], but only when the amount of dopant(s), the method of its incorporation and the fabrication conditions of Ta₂O₅ are simultaneously optimized. The end effect of the presence of a third element into Ta₂O₅ matrix depends on the dopant amount (some times the enhancement of the permittivity is a non-linear function of the dopant concentration [8], [20]), and the method of its incorporation [4], [5], [6], [10], [11], [12], [15], [19]. Therefore a knowledge on the impact of specific dopant on the electrical behavior of Ta₂O₅ stack is necessary.

A comparative analysis of the effect of Ti and Hf as dopants in Ta₂O₅ in conjunction with the influence of Si nitridation on the stack characteristics is presented in this work, (some of the data are previously published). The relevance of mixing of two favorite high-k dielectrics for devices applications (Ta₂O₅ and HfO₂) is also a subject of the study. A special emphasis is put on the role of dopant on the mechanism(s) of conductivity and on the stack stability under electrical stress. The specific purpose of the work is to take a more general look at our investigations and experience on doped and mixed with another high-k Ta₂O₅, and to demonstrate the wide diversity of phenomena which influence and control the electrical behavior of modified Ta₂O₅ films. This motivation is based on the simple fact that the multicomponent nature of such kind of films defines the great complexity of phenomena typically observed for high-k dielectrics.