Ta oxide and Ti oxide may have large defect density and are not suitable for silicon oxide replacement. For future IC generations, protective metal oxides are better candidates. The Pilling-Bedworth ratio is a valuable figure of merit to determine whether a metal oxide can be protective.

Why was silicon chosen over other semiconductor materials by the integrated circuit industry? The single most important reason is its ability to form a robust, high-yielding oxide. Silicon oxide not only guarantees the operation of a MOSFET, but is also the dielectric material in DRAM storage capacitors. As VLSI technology advances towards the nanometer regime, silicon oxide is approaching its practical thickness limit (~2nm). Much effort has been spent to identify alternative insulating materials, particularly high-dielectric constant (k) metal oxides.

Most research on the high-k metal oxides focuses on dielectric constant, leakage current, and other electrical proporties. Very little attention is paid to the defect density: it is considered a fab issue that will be improved when cleaner equipment is used. However, with a dielectric constant of 10-25, the high-k metal oxides should be made no thicker than 12nm. Any pinholes (micropores) in such thin films will be detrimental to IC yield. A first screening test to ensure a high IC yield is that the metal oxide is pinhole-free.

Pilling-Bedworth Ratio
Physical metallurgy relies on a set of guidelines to identify protective oxide coatings for corrosion protection. Such coatings should be pinhole-free, exactly as the IC industry requires for high-k metal oxides. Accordingly, guidelines developed for protective oxide coatings in physical metallurgy can be applied to the protective metal oxides in integrated circuits.

In their 1923 paper 揟he oxidation of metals in high temperature?presented to the Institute of Metals, N. B. Pilling and R. E. Bedworth first correlated the porosity of a metal oxide with the specific density¹. The Pilling-Bedworth ratio, (P-B ratio) R, of a metal oxide is defined as the ratio of the volume of the metal oxide, which is produced by the reaction of metal and oxygen, to the consumed metal volume:

M and D are the molecular weight and density of the metal oxide whose composition is (Metal)a(oxygen)b; m, and d are the atomic weight and density of the metal.

Pilling and Bedworth realized that, when R is less than 1, a metal oxide tends to be porous and non-protective because it cannot cover the whole metal surface. Later researchers found that, for excessively large R, large compressive stresses are likely to exist in metal oxide, leading to buckling and spalling. In addition to R, factors such as the relative coefficients of thermal expansion and the adherence between metal oxide and metal should also be favorable in order to produce a protective oxide.

Using the P-B ratio, Bruce Chalmers, Gordon McKay professor at Harvard University (Cambridge, MA), separated 損rotective?metal oxides from 搉on-protective?metal oxides². The table lists 損rotective?and 搉on-protective?metal oxides and their P-B ratios.

The list can be readily applied to the protective metal oxides used in integrated circuits. The intrinsic protective metal oxides, including the oxides of Be, Cu, Al, Cr, Mn, Fe, Co, Ni, Pd, Pb, and Ce, may be able to replace silicon oxide. On the other hand, a few popular metal oxides, e.g. Ti oxide and Ta oxide, are non-protective, suggesting a possible reason why these oxides have not been successfully used in commercial products after years of research. Besides oxides of elemental metal, the P-B ratio can be applied to oxides of metal alloy, metal nitrides and other metal ceramic systems.

Protective metal oxides can be produced by two classes of methods: growth and deposition. Growth methods include thermal oxidation, plasma oxidation, anodization, and implantation. Deposition methods include direct sputtering, reactive sputtering, and CVD. The IC industry does not have much experience with production of protective metal oxides. Still, manufacturing experience with sub-10 nm silicon oxide suggests that it is better to form the protective metal oxide using a growth method, or to form at least a portion of the protective metal oxide using a growth method and the remaining portion using a depositing method. Improved oxide thickness uniformity is one added advantage of such a composite layer.

Conclusion
The minimum requirement for a metal oxide to be protective is that its P-B ratio larger than 1, preferably smaller than 2. Manufacturing process is also an important factor to ensure a low defect density. Besides being used as the gate material in MOSFET, the protective metal oxides can be used in DRAM storage capacitors and in antifuses for FPGA and 3D-ROM.

About the author
G. Zhang, P.O. Box 9562, Berkeley, CA 94709-0562. Tel: 714-914-8718; E-mail: gzhang@3d-rom.net.

Guobiao Zhang (George) received his M.S. ('92) and Ph. D. ('95) from the University of California at Berkeley. Before that, he studied at the Special Class for the Gifted Young at the University of Science and Technology of China. He did extensive work in IC devices, including antifuse and semiconductor memory.

The protective metal oxide disclosed herein is a preferred embodiment of U.S. Patent 5,838,530 (11/17/98).. It cannot be construed as a limitation to the scope of said patent.