Zirconium (Zr) and hafnium (Hf) silicates are promising high-k dielectrics developed largely to overcome the interface stability issues suffered by many high-k binary metal oxides [52]−[54]. Ternary phase diagrams of the Zr-Si-O system reveal that the binary metal oxide ZrO2 as well as the compound silicate ZrSiO4 should be thermodynamically stable in direct contact with Si. Since Zr and Hf are isoelectronic elements, it is expected that HfO2 and HfSiO4 should also be stable on Si.
In practice, interface reactions, much like those described in the previous section for other metal oxides, have been observed for nearly all ZrO2 and HfO2 films deposited directly on Si [50], [55], [56]. High resolution transmission electron microscope (HRTEM) measurements by Campbell et al. revealed that interfacial layers of 0.9 nm and 1.2 nm thickness were formed when depositing ZrO2 and HfO2 directly on Si, respectively [50].
Medium energy ion spectroscopy (MEIS) and x-ray photoelectron spectroscopy (XPS) analysis showed that the interfacial layer was SiO2-like, not a silicate. In contrast, Lee et al.
found that ZrO2 and HfO2 films deposited directly on Si using a magnetron sputtering technique gave rise to a silicate interfacial layer [55]. Copel et al. also found that ZrO2 deposition by atomic layer chemical vapor deposition (ALCVD) on HF-last-Si led to discontinuous nucleation with islands of ZrO2 interspersed along the Si interface [56]. Atomic force microscopy also revealed a large RMS roughness of 0.57 nm along the interface.
The variability in the interfacial layer following ZrO2 and HfO2 deposition leads to the conclusion that the growth of such layers is difficult to control. Instead, recent efforts toward integrating ZrO2 and HfO2 films have focused on first growing a high quality, well controlled, ultrathin SiO2 layer prior to ZrO2 deposition. Perkins et al. deposited ALCVD ZrO2 films on chemically grown oxides and obtained EOT < 1.4 nm [57]. Having an underlying SiO2 layer
also leads to the desirable electrical properties of a Si/SiO2 interface which helps to maintain high channel carrier mobility. Of course, the minimum EOT is still limited by the extent of the interfacial SiO2 layer as described previously.
Another potential problem with ZrO2 and HfO2 is that they have been observed to crystallize at relatively low temperatures. Polycrystalline films may exhibit high leakage paths along grain boundaries which act as trapping centers. Thus, in general, amorphous dielectrics which resist recrystallization up to relatively high temperature are desirable for gate dielectric applications.
To overcome the challenges of pure ZrO2 and HfO2 films, Wilk and Wallace proposed the use of Zr and Hf silicates as promising high-k gate dielectrics [42], [52]−[54]. By alloying two different oxides, such as ZrO2 and SiO2 in the case of Zr silicate, (ZrO2)x(SiO2)1-x, they believed that it may be possible to retain the desirable properties of both oxides while eliminating the undesirable properties of each. By explicitly incorporating SiO2 during deposition of ZrO2 precursors, the driving force for reaction between the dielectric and the Si substrate is reduced, so that the interface is more likely to behave like the desirable Si/SiO2
interface. This allows better control of the Si interface properties. It is generally believed that silicate deposition results in a single, uniform high-k layer in direct contact with Si.
Combining a poly-crystalline film like ZrO2 with an amorphous one like SiO2 also leads to an amorphous silicate phase.
At the same time, by incorporating some amount of ZrO2 into the SiO2 film, the enhanced polarizability of Zr-O bonds relative to Si-O bonds leads on average to a higher dielectric constant for material. While many different silicate systems are possible, column IVB elements such as Zr and Hf are expected to substitute well for Si atoms, thus reducing the possibility of forming dangling (i.e. unsaturated) bonds at the Si interface. The notion of
doping SiO2 with Zr or Hf also leads to a natural scaling scenario for the silicate system. By progressively increasing the Zr or Hf concentration, the k-value can be steadily increased, up to a limit, to meet the gate capacitance requirements for future technology generations.
The increased control over interface properties comes at the expense of a lower dielectric constant relative to pure ZrO2 or HfO2. Depending on the metal concentration, the dielectric constant is believed to scale between 4, corresponding to pure SiO2, and approximately 15 to 20 for stoichiometric ZrSiO4 or HfSiO4, since pure ZrO2 and HfO2 are thought to have dielectric constants close to 25 and 35, respectively [54]. Si-rich compositions are preferred to maintain thermal stability with the Si substrate. Preliminary theoretical work by Jun et al.
suggests that both ZrSiO4 and HfSiO4 have dielectric constants near 12, toward the lower end of the above range.
The lower dielectric constant of the silicates relative to the pure metal oxides does not necessarily imply a smaller gate capacitance. To illustrate this point, Figure 2.7 shows two possible gate stack structures. Figure 2.7(a) depicts a pure metal oxide (e.g. ZrO2 with k = 25) which requires a 0.5 nm interfacial SiO2 layer with k = 4. Figure 2.7(b) shows a silicate layer (e.g. Zr silicate with k = 16) which forms a single, uniform high-k layer in direct contact with Si. Using Equations (2.9) and (2.11), it is straightforward to show that both gate stack structures achieve the same equivalent gate capacitance, corresponding to EOT values of 1.0 nm. Surprisingly, the lower-k silicate layer can be physically thicker than the higher-k ZrO2
layer, so that less tunneling is expected for the silicate stack. This is due to the fact that the silicate is believed to form a single, uniform high-k layer, thus avoiding the formation of a low-k interfacial region.