New Materials in Semiconductor Fabrication: An Evolutionary Process

An evolution has been taking place in CMOS fabrication processes, which received an injection of pace when Intel (Santa Clara, Calif.) and IBM (Yorktown Heights, N.Y.) both announced the use of hafnium in the high-k dielectric gate material. It is becoming clear that novel materials are going to play an important role in sustaining Moore's Law, along with the traditional scaling. The voracious appetite for rapidly storing and accessing information has prompted the need for materials with higher and higher k values, ranging from the oxides of aluminum to rare earth metals and those of every possible element in between. The challenges are enormous, as the novel molecules to make it all happen show some very unusual properties that make their transport onto the wafer more difficult and their breakdown chemistries much more complex.

New materials integration requires careful risk management. (Source: SAFC Hitech)

The move to 45 nm

The age of new materials has arrived in the CMOS fabrication process with the announcement of 45 nm microprocessors in January of 2007. This announcement introduced hafnium-based high-k dielectric, along with novel metal gate electrode materials, to improve the transistor performance of processors. With this transition, the silicon industry took a major step toward adapting elements from the periodic table that did not belong to the core list of elements, such as silicon, oxygen, nitrogen and carbon. It is now generally recognized that scaling alone will not be sufficient in achieving ever-increasing performance of the silicon device. The component of innovation will involve newer materials from the periodic table of elements, and play a major role in driving performance.

Of course, this change did not occur overnight. The replacement of silicon dioxide and silicon oxynitride as the transistor gate insulator material was being investigated even before 2002. The 1999 International Technology Roadmap for Semiconductors (ITRS) highlighted the introduction of new materials to the IC manufacturing process. It was becoming very clear that at a thickness below 2 nm, the gate oxide would suffer high leakage currents and a material with a higher dielectric constant would be necessary to allow a thicker gate dielectric while reducing the leakage. Wallace and Wilk¹ highlighted the manufacturing issues facing the selection of the high-k material to offer permittivity, thermodynamic stability, gate electrode compatibility, and interfacial stability, amongst others. The periodic table of elements was further assessed by Scholm and Haeni² to propose a thermodynamic approach to selecting alternative gate dielectric oxides and nitrides; materials including compounds of hafnium, zirconium and rare earth elements.

The exploration of new materials is not limited to high-k gate dielectrics and metal gates. Potential applications include capacitor dielectrics, interconnect layers, metal contacts and plugs. A golden opportunity has evolved for the chemists and technologists who like to play at the interface of chemistry and material science. The periodic table³ (Fig. 1) illustrates the range of elements that are being explored at various stages of development via various collaborations between academia, OEMs, IDMs and material suppliers. The importance of these collaborations cannot be overemphasized, as significant investment is required by each partner.

1. Many elements of the periodic table are being considered to meet the electrical, mechanical and physical needs of nanoelectronic devices.

The introduction of new materials requires a series of new molecular precursors with the right properties. Both CVD and ALD have been recognized as viable techniques to deposit new materials on the chip. As features shrink, ALD becomes more favorable because of its ability to produce layers with atomic-level thickness, uniformity over a wide range of planar structures, conformality over non-planar structures, and overall manufacturing process control. ALD is expected to play a vital role as 3-D transistors begin to emerge where uniformity and conformal thicknesses will be required, and would draw on the availability of new molecules. For a molecular precursor to be suitable for ALD:

• It should be thermally stable at the deposition temperature
• It should have some volatility, preferably >50 mTorr at the delivery temperature
• It should possess the right chemistry to evenly form oxides, nitrides, carbides or reduction to the metal itself
• It should be economically acceptable within the process requirements

The acceptance of ALD in semiconductor fabs is at a different level with the DRAM industry at the helm. Most capacitor materials like Al₂O₃, HfO₂ and ZrO₂ are using ALD as the deposition technology of choice. For gate dielectrics, ALD is beginning to be adopted, with implementation in flash technology following closely behind.

Most of the material suppliers are designing novel molecules to either introduce a new material on the chip or replace an existing class of precursors with limitations. One such case includes the use of hafnium oxide, hafnium silicate or HfSiON materials. The conventional molecules used for this application in ALD are hafnium chloride, tetrakis(dimethylamido)hafnium (TDMAH) or tetrakis(ethylmethylamido)hafnium (TEMAH). While all of these molecules have worked in the process, the amides are limited because of their thermal instability at the high temperature of vapor transport or deposition temperature, which is typically <300°C. Chemists and materials engineers studied^4–6 novel non-amide precursors and have shown ideal self-limiting deposition behavior at temperatures as high as 450°C with negligible incorporation of impurities within the layers (Figs. 2 and 3).

2. Novel precursors of hafnium and zirconium have the potential to replace TDMAH and TEMAH, amides currently used to deposit hafnium oxides by ALD.

Researchers are already extending their reach into the periodic table by exploring oxides with even higher-k materials, such as zirconium oxides, silicates, rare earth-doped oxides of hafnium and zirconium, and rare earth oxides themselves. The objective of this exercise is not only to find an ideal ALD precursor, but also to understand issues such as interface stability and the integrity of the gate oxide after subsequent processing. Recent deposition of lanthanum zirconate with tris(isopropylcyclopentadienyl) lanthanum and ZrD-04 (Fig. 2) by liquid injection ALD and CVD highlights some of the issues described here.⁷ It is predicted that some of these rare earth metal oxides are going to be adopted in the 32 nm node. Rare earth oxide precursors suitable for CVD/ALD belong to the following general categories: sterically hindered alkoxides, β-diketonates, β-ketoiminates, substituted cyclopentadienyls, tris silylamides, amidinates and pyrazolates.

3. The HfD-02 and HfD-04 precursors led to the lowest levels of impurities in the deposited hafnium oxide films.

As these precursor systems become complex, their volatility and thermal stability become a concern. The evolution of precursors from gases (e.g., silane) to liquids (e.g., TEOS) to less volatile solids (e.g., rare earth molecules) is not without problems. Already the industry has begun to warm up to the concept of direct liquid injection (DLI) and solution vaporization techniques. Of course, this mandates the development of precursors with high purity and low residues.

Research is continuing at all levels of the supply chain to find a suitable metal gate material as well. Intel and IBM have announced the use of new metal gate materials at the 45 nm node. This trend is going to continue in the 32 nm node, with the possibility of next-generation metal gate electrode material such as ruthenium, platinum and transition metal nitrides and carbides. Whatever the materials are, they should be able to withstand temperatures as high as 1000°C caused by a likely annealing process. ALD precursors of cobalt and nickel and their silicides have started to attract a great deal of attention as well.

Volatile metal precursors of ruthenium, platinum, cobalt, nickel and other noble metals consist of ligands of the types cyclopentadienyls, carobonyls and amidinates. If the desired properties are not achieved, functional groups are added or modified. Figure 4 gives an example of precursor development with a view to produce a liquid precursor with high vapor pressure. Newly developed ruthenium precursors with the highest volatility and an ideal thermal stability are shown in Figure 5. The presence of an alkyl group bonded directly to the ruthenium metal center has been shown to assist in the ALD process.

4. The ideal ruthenium precursor can be delivered as a liquid with high vapor pressure.

Material suppliers cannot "go it alone" and are seeking early engagement with OEMs and IDMs in their development cycle. Various collaborations are developed with academic institutions to understand the basic fundamental properties of new materials.

5. Some of the most promising ruthenium precursors have an alkyl group bonded directly to the ruthenium metal center, which assists in the ALD process.8

Finally, the complexity of these molecules, along with the concern of their availability and expense of raw materials, may make these molecules appear less economically viable in comparison to traditional precursors. The value could be in the form of smaller feature size and added performance, leading to efficient utilization of the precursors. In all instances, cost-of-ownership models will enable developers and users to assess the value offered by these materials. No doubt, the company with experience at all levels of the materials supply chain (from R&D through high-volume manufacturing), application knowledge base, and innovative delivery solutions will prove to be the enabling partner in this technology space.