To meet the need for higher transistor speed while keeping power consumption under control, the semiconductor industry is working to introduce high-k gate dielectrics in leading-edge transistor-manufacturing processes.
To meet the need for higher transistor speed while keeping power consumption under control, the semiconductor industry is working to introduce high-k gate dielectrics in leading-edge transistor-manufacturing processes. But the attempts are hampered by problems with high-k device performance (in particular, low mobility and poor control of threshold-voltage values), which were not fully anticipated at the start of high-k development activities years ago.
Are these issues, which may potentially delay introduction of high-k dielectrics, of an intrinsic, material nature, or are they mostly related to material integration within the fabrication process? The answer lies in understanding the fundamental factors controlling the values of dielectric constants.
Unlike conventional silicon dioxide dielectric, where electronic polarization plays the major role, the most important contribution to the value of the dielectric constant in high-k materials comes from the dipole momentum generated by appropriate ion displacements: The larger the amplitude of the displacements, the greater the k values. To occur, larger displacements must provide an additional advantage for the atomic system; that is, they must generate energy gain.
It has become clear from extensive studies of a wide class of transition metal compounds that the driving force for large atomic displacements is a very specific feature of the d-electrons present in all high-k materials; that is, their highly anisotropic spatial distribution. Because of this anisotropy, the appropriate dipole momentum generating displacements of the metal ion may increase the overlap between its d-electrons and the electrons of the surrounding oxygen atoms; this additional overlap lowers the total energy of the system.
The atomic-system instability associated with the ionic displacements, caused by their strong coupling with the low-symmetry electrons, generally is known as the Jahn-Teller effect (formulated by Edward Teller, later known as the "father" of the H-bomb).
The presence of the d-electrons affects most of the critical properties of high-k dielectrics. For instance, all high-k materials exhibit low band offsets, which define the energy barrier for the electron injection from the silicon substrate into the dielectric. This band offset strongly depends on the dielectric bandgap, which is reduced (with respect to the SiO2 dielectric) by a poor overlap of certain oxygen electronic states with metal d-electrons, because the latter do not point directly to the oxygen atoms.
Since these d-electrons are highly delocalized and are shared among the surrounding oxygen atoms, the atomic structure of high-k dielectrics is much more rigid than that of silicon dioxides. Such rigidity leads to a high density of structural defects, which may be a source of fixed charges and electron traps affecting mobility, threshold voltage and reliability.
For the same reason-delocalized d-electrons-the transition metal compounds tend to form a long-range order that promotes crystallization of high-k dielectrics. This is a very undesirable material property for the transistor application, since it introduces nonuniformity by randomly modulating the applied electric field in the transistor channel, as well as by generating grains boundaries prone to the contaminant ion diffusion. In particular, oxygen diffusion along these boundaries greatly contributes to the parasitic growth of the interfacial layer between the high-k dielectric and silicon substrate.
The d-electrons that deliver high "k" values are also responsible for material limitations. Many current issues with high-k dielectrics are related to intrinsic, fundamental material properties rather than to the integration of those materials into the transistor process flow. Most of these fundamental challenges could and should be uncovered and analyzed in advance, before the industry launches a costly frontal assault on the high-k problem. That is the lesson that should be learned for future attempts at radical innovations.
Gennadi Bersuker is a fellow and Howard Huff is a senior fellow at International Sematech.