IN A NUTSHELL: We are currently developing a new technology for manufacturing ultra-low dielectric constant materials (ULκD, ĸ< 2.0) for leading-edge logic devices for the 22 nm technology node and beyond. It is widely recognized that materials presently used will not remain effective with further miniaturization of integrated circuits either because they do not have sufficiently low ĸ values, or because their morphological features will not provide acceptable mechanical properties for process integration. Many also agree that industry cannot continue to simply engineer around materials shortcomings indefinitely. Our approach is based on a unique concept of bottom-up synthesis of honeycomb-like nano-structured films from nanoscopic core-shell-type PAMAMOS dendrimers in which a porogen component is pre-built into the nano-sized cells and can be decomposed in a controlled manner. This provides geometrically precisely organized closed pores with diameters that can be selected at will by choosing the right dendrimer generation from the range of 1-5 nm with precision of ±1 nm while separated by 1-3 nm thin organo-inorganic hybrid walls.

THE PROBLEM: While copper has emerged as the preferred conductor material, the future of integrated circuits (ICs) will greatly depend on new and more efficient insulators, usually referred to as interlayer dielectrics (ILDs). For example, it is expected that in the next 5-10 years an average IC will contain a billion or more transistors, and over 10 kilometers of “wiring”. Hence, in order to minimize electrical “cross-talk” between so closely packed wires (the number in the designation of the technology node corresponds to the separation distance between the neighboring conductors expressed in nm) and ensure optimum chip performance, it will be necessary to employ dramatically improved ILDs since the presently available ones will just not be good enough and the industry cannot continue to engineer around materials shortcomings indefinitely. In fact, it can be shown that in order to advance beyond the 22 nm technology node it will be necessary to develop completely new insulators with dielectric constants (κ, a measure of the insulating power of a material) significantly lower than 2 (the lower the κ number, the better the insulator).

Road Map for the Development of Low Dielectric Constant Materials

However, in spite of some major advances in the technology of low κ dielectrics in recent years, good ultra-low dielectric constant insulators have not yet been developed. Furthermore, while it is hard to generalize about low ĸ technology today, since every material producer and every chip maker has its own integration schemes and processes and many details are proprietary, an overview of available data clearly shows that porosity is the key for reaching into the required κ<2 zone. This is because the dielectric constant decreases with the host matrix density, as ĸ of air is equal to 1.0006 and represents one of the lowest dielectric constants known.  

THE STATE OF TECHNOLOGY: Development of nanoporous dielectrics has advanced in two main directions: the templating and the copolymerization approach. The templating approach consists of three steps: (a) physical dispersion of thermally degradable material (porogen) into the selected thermosetting dielectric matrix, (b) heating the matrix to crosslink and phase separate the porogen into nanoscopic domains, and (c) second heating to decompose the porogen and enable its degradation products to diffuse out of the resulting porous material under conditions that prevent pore collapse. Typical examples of this approach include Dow Chemical’s porous SiLK and IBM’s DendriGlass, and competing products by JSR (Tokyo), Asahi Chemical, Shipley Co., Schymacher and Honeywell, etc. Major advantages of this method are applicability of a wide range of thermoset matrices (from inorganic to organic) and porogens (ranging from polymers to solvents), as well as utilization of the relatively simple spin-on deposition techniques. On the other hand, its disadvantages include the design and structure of the porogen used, uniformity and size distribution of the dispersed porogen domains and regularity of porogen phase separation throughout the ILD film at the nano level. In contrast, the copolymerization method involves the covalent attachment of a thermally decomposable organic component (often polymeric) to a high glass temperature organic matrix which results in the formation of a copolymer structure in which microphase separation is induced to create nanoscopic domains by self-assembly. An example of this approach is the surfactant-directed polymerization and crosslinking of silicate oligomers developed by Schumacher and Sandia National Laboratories. However, this method also encounters problems with precise control of pore sizes as well as pore size and pore disposition distribution which depend on the phase-separation of the constitutive nano domains. Finally, while both approaches have their advantages and disadvantages, the major problem remains the same: the volume fraction of the pores in the resulting porous material must be high enough to achieve the desired dielectric constant reduction, but at the same time also low enough to ensure sufficient mechanical consistency of the deposited ILD films.

OUR NOVELTY: Today, it has become clear that a successful nanoporous dielectric must combine the following major requirements: (a) the volume fraction of pores must be high enough to achieve the desired dielectric constant reduction (usually about 20-40 vol.%), but at the same time low enough to ensure sufficient mechanical consistency of the deposited ILD films, (b) the pores must be closed and spherical, with very uniform size and topological distribution throughout the entire volume of the ILD film, and (c) the degradation products that evaporate through the surrounding dielectric matrix during the thermal treatment must not leave any non-degradable residue in the resulting pores. To accomplish this, we are developing a unique new technology for preparation of ULκDs based on our geometrically precise, core-shell PAMAMOS dendrimers which can be crosslinked into 3D honeycomb-like films in which thermally degradable (at temperatures between 150 and 200°C), spheroidal, nano-scaled PAMAM domains act as porogen species which are perfectly distributed throughout the matrix while separated by much more thermally stable (up to 300-350°C), 1-3 nm thin organosilicon (OS) walls. By the mere nature of dendrimer technology, the sizes of both of these domains are controllable at will, with a precision of ±1 nm, by selecting the generation of the dendrimer used for the film preparation. Thermal treatment of such films results in the formation of pores with a potential for control of their sizes and topological distribution which leads to the ability to dial-in the desired ĸ values. Since both PAMAM decomposition and OS crsslinking chemistry occur at relatively low temperatures between 150 and 250°C, this process also offers significant energy savings in the computer chip manufacturing industry with respect to temperatures (often > 400°C) that are currently used.