Light Constructions - Nanostructured dielectrics good candidates for next-generation computer chips

IBM researchers have reported a new technique for producing highly insulating materials that are compatible with existing processes in the semiconductor industry. The procedure involves chemically creating air-filled nanopores inside an organosilicate matrix. Because air is such a good insulator, its presence inside the other material greatly enhances its dielectric properties.

Researchers recently demonstrated that, with just 20 percent of layer volume made up of these pores, they can produce a thin film with a very low dielectric constant of 2.2. Although the technology has not yet been fully developed, it already meets basic criteria of thermal stability and fabrication simplicity.

Materials with low dielectric constants (low-k) are important for the next generation of computer chips, because of the problems of multilevel wiring. As feature sizes become smaller and wires are pushed closer together, the problem of capacitive coupling or crosstalk increases. Although moving from aluminum-based films used today to copper will help, increasing the electrical barrier between wires is also likely to be necessary.¹ To do this requires good insulators.

Unfortunately, although there are many materials that can act as dielectrics, few of these have values of k in the ultra-low range, and those that do tend to be difficult to integrate with conventional technology. To be successful, a material has to meet a number of tough constraints. Its chemistry has to be compatible with that of the rest of the chip; it has to stick well; and it must be thermally stable up to relatively high temperatures, easy and cheap to produce, and environmentally stable.

The IBM team, which is comprised of researchers from both the Almaden (San Jose, CA) and T.J. Watson (Yorktown Heights, NY) research centers, were initially interested in trying to improve the dielectric constant of organic compounds (plastics).^2,3 These materials have the advantage of being very easy to deposit -- they can be spin-coated -- and are otherwise easy to work with. They also have dielectric constants that are significantly lower than silicon dioxide. However, in their normal polymeric form they have a number of problems -- most notably, the long-chain molecules form a layer that is both optically and dielectrically anisotropic. Researchers say the fact that the in-plane dielectric constant could be as much as 0.8 higher than that of the out-of-plane constant would be an unacceptable constraint on chip designers.

To overcome this problem, and to further decrease the dielectric constant of the materials, researchers tried to find a way to chemically create nanopores inside the material layers. The imposed structure would stop the plastics from forming even, aligned layers, and so prevent the undesired anisotropy. Making unconnected nanopores (so the layer won't act like a sponge) of the right size is complex chemically, but the basic idea is simple. The matrix material, which has high thermal stability, is combined with another plastic that starts to break down at a much lower temperature. When this temperature is reached, the polymers start to split apart, eventually producing a gas of monomers that are small enough to escape the material. These are replaced by air.

This approach was successful, but only for porosity levels below about 18 percent. At higher levels, the bubbles were found to be interconnected. Researchers speculated that this might be because the sacrificial pore material broke down and expanded into a gas more quickly than the individual monomers could escape: this would cause the pores to blow up like a balloon and improve the chances that they could meet and merge. To combat this, they tried to strengthen the matrix through crosslinking, but this had almost no effect.

More recently, the IBM team has been working with organosilicates-in particular, methylsilsesquioxane (MSSQ). With a dielectric constant of 2.85, MSSQ is another material that starts off much lower than silicon dioxide (which is in the 3.9 to 4.2 range) and meets the basic constraints for integration. However the mechanism for pore production is subtly different than in the organic case.

Figure 1. A schematic of the chain extension and vitrification of low-molecular-weight MSSQ materials. The material starts to harden around 150 deg. C and continues to do so up to 400 deg. C.

The MSSQ is a resin that vitrifies upon heating (Figure 1). First it goes through a liquid-like stage and then, at around 150 deg. C, it starts to become hard. Within this matrix are porogens, specially designed molecules that, like those described above, are designed to break down and escape from the deposited layer (Figure 3). These start to disintigrate at around 300 deg. C. The sequence in which the stiffening of the matrix and the breakdown of the porogen occur is critical: if the trapped gas escapes before the matrix has fully hardened, then the pores could collapse. The technique successfully produced the non-interconnecting nanopores that researchers were looking for (Figure 3). The resulting structures not only have dielectric constants in the ultra-low range, but can also withstand temperatures higher than 400 deg. C.

Figure 2. A generic porogen, here with six arms, showing the basic features of a pore-generating macromolecule. As the porogen breaks down within the MSSQ matrix, it forms small molecules, which then diffuse through the matrix and escape as gases leaving holes.

Figure 3. Transmission electron microscope photograph of the final product: porous MSSQ. This sample was prepared from a hybrid containing 10 percent of a four-arm star porogen. This material has a low dielectric constant and is highly thermally and environmentally stable.

References:

1. Robert D. Miller, In Search of Low-k Dielectrics, Science 286, pp. 421-423, 15 October 1999.

2. J.L. Hedrick, K.R. Carter, J.W. Labadie, R.E. Miller, W. Volksen, C.J. Hawker, D.Y. Yoon, T.P. Russell, J.E. McGrath, R.M Briber, Nanoporous Polyimides, Advances in Polymer Science 141, pp. 1- 43, 1999.

3. James L. Hedrick, Robert D. Miller, Craig J. Hawker, Kenneth R. Carter, Willi Volksen, Do Y. Yoon, and Mikael Trollsås, Templating Nanoporosity in Thin-Film Dielectric Insulators, Advanced Materials 10 (13), pp. 1049-1053, 1998.

4. R.D. Miller, W. Volksen, J.L. Hedrick, C.J. Hawker, J.F. Remenar, P. Furuta, C.V. Nguyen, D. Yoon, M. Toney, D.P. Price, J. Hay, Sacrificial macromolecular porogens: A route to porous organosilicates for on-chip insulator applications, Reference to come.