Adherence to the prescript of Moore’s law continues to drive materials development for new and lower dielectric constant materials for use as back-end-of-line (BEOL) interlayer dielectric in advanced logic IC’s. As is the case for the current generation of low-K materials (<3.0), these ultra-low K materials (<2.2) will need to meet the variety of integration and reliability requirements for successful product development. Excluding the incorporation of fluorine to lower the material polarity, further reductions of dielectric constant can only be achieved by reduced density. Based upon the industry’s experience with the current class of full density dielectrics, process integration may be challenging for ultra-low K materials. This anticipated difficulty derives from the profound differences in material properties, e.g. mechanical integrity, as one lowers the material density, which in turn confounds existing manufacturing processes that have evolved over 35 years based on silicon dioxide. Minimizing these material and processing differences by extending leveraged learning from previous technology nodes is essential for timely and cost-efficient development cycles. As a result, material selection of a full density low-K is somewhat influenced by the ability of that material to be extended into future generations. Understanding how the material properties will change as its density is lowered is vital to this selection process. In this paper, we present a summary of models for calculating effective properties as a function of density and apply these to current low-K materials with emphasis on mechanical integrity. We will also review experimental methods for measuring the mechanical integrity of ultra-low K materials and compare the results to the various models described herein.

The integration of organic low k interconnection dielectric materials into wafer fab manufacturing processes requires changes to the equipment set and the process sequence. The capital investment required for conversion and the impact of the new equipment and materials on the finished wafer cost are a concern to wafer manufacturers. We have modeled the costs of converting 180 nm logic/microprocessor fabs from Al/W/SiO/sub 2/ and Cu/SiO/sub 2/ dual damascene interconnect processes to Al/W/Low-k and Cu/Low-k processes. Equipment was chosen to minimize the capital outlay required for conversion. The output of the models includes equipment capital requirements: changes to staffing and facilities costs, and the cost of finished wafers. Our results indicate that conversion to low k processing should require only modest capital expenditures and effects the finished wafer costs by less that 5%. For example, one case study shows 10% of a fab's capacity could be converted from Al/W/SiO/sub 2/ interconnect technology to Al/W/Low-k by the addition of a wafer apply track. In this case wafers with the Al/W/Low-k interconnection technology are produced in a new fab at a lower cost than those with Al/W/SiO/sub 2/ interconnection technology.

Currently, the IC industry is researching the integration of a variety of materials to meet the low dielectric constant requirement for improved back-end of line (BEOL) interconnect performance. One critical dimension for successful ntegration of these new materials is maintaining mechanical integrity through multilayer processes. This includes both cohesive and adhesive fracture resistance. The latter adds additional complexity in that adhesive toughness is a function of the adherend materials and the processes used to join them. Hence, many good dielectric materials may be rematurely eliminated from further research not because of inherently poor adhesion but because of the necessity to optimize processing strategies. In this paper, we use the modified Edge Liftoff Test (m-ELT) to quantify the mechanical adhesion of multilayer blanket coatings. A specific example is used to demonstrate the utility of combining the m-ELT with surface analysis to optimize the reliability of low-K dielectric resins for use in ULSI applications. The system studied consists of a Cyclotene ™ 5021(BCB) low-K material integrated with CVD aluminum for single level, damascene structures. The effects of liner layer metallurgy and surface plasma treatments are measured. Surface analysis is done on the failed parts to understand the location of the failure. In this way recommendations for process optimization can be made.

Polymer films of DVS-BCB (CYCLOTENETM 5021) exhibit a combination of material and processing properties which make them an attractive low k interlayer dielectric (ILD) material for integration into IC manufacturing processes. Key DVS-BCB film properties include isotropic dielectric constant (2.65 @ 1 MHz); equilibrium moisture absorption (< 0.2% @ 25°C and 81% RH); high purity (< 10 ppb individual metals concentrations); Tg > 350°C; low alpha-particle emission rates. The DVS-BCB films are easily produced using simple resist spin track equipment and are subsequently cured without generation of corrosive or volatile by-products under anaerobic conditions (less than 100 ppm oxygen). Films are completely cured for typical processing conditions at 250°C for 1 h or as rapidly as 30 s at 325°C. DVS-BCB has been demonstrated to fill 0.20 μm gaps at 5:1 aspect ratios. Fully cured films exhibit greater than 95% degree-of-planarization for isolated feature widths of 20 μm and less; 80% for widths between 20 μm and 100 μm. Currently, the acceptance of DS-BBC is challenged by today's thermal requirements of greater than 400°C for CD-W via/plug and post device/contact anneals. The development of moderate temperature metal deposition processes and reduced final device anneal temperatures would enable implementation of DS-BBC as an LID material with the realization of its low dielectric constant, excellent gap-fill and other attractive features.

A novel polymer dielectric derived from tris-perfluorocyclobutene(PFCB) monomer has been investigated as a thin film dielectric. The dielectric constant of this material has been measured at 2.35. PFCB is applied from a hydrocarbon solvated solution using conventional spin coating equipment onto the substrate and produces highly uniform films after curing at 300 °C. This paper describes the processing and properties of films derived from PFCB, and examines the interactions between the polymer and metals at the interfaces. In particular, the interfaces formed by the deposition of Cr or Co onto PFCB are unchanged after 30 minutes at 390 °C. The interface formed by the application of the polymer onto Ta is similarly unaffected by high temperatures. The interfaces formed by the deposition of Ta or Ti onto the polymer surface did not remain intact, with substantial quantities of the metals permeating the polymer after high temperature exposures. This behavior is discussed relative to the thermo-physical properties of the metal fluorides.

Three low dielectric constant organic polymers are being investigated for possible use in a conventional, subtractive etch, multi-level metal process with PVD Al plugs. Material properties, physical properties, planarization ability and etch chemistries are compared, as well as the possibility of using these materials in a low temperature PVD A1 plug process.

The results of a series of novel synchrontron-based in situ x-ray scattering experiments of monofilament fiber drawing from lyotropic solutions of poly(cis-benzoxazole) (PBO) and poly(trans-benzothiazole) (PBZT) are reported. The purpose of the study is to determine orientation and microstructure development in the draw zone as a function of shear rate in the capillary die, and spin draw ratio (SDR). The transition of the extrudate from opaque to the transparent is complete at about a SDR = 3 and f of 0.9. The filament orientation parameter (f) was found to depend strongly on spin draw ratio, but not shear rate. The orientation was found to increase down the extrudate toward completion of the draw down as one proceeds further from the die face up to an extrudate length of 3.8 cm. Coherence lengths on the order of 19 nm (axial), and 4.5 nm (lateral) have been observed. These ''microdomain'' sizes are consistent with the ''crystallite'' sizes typically observed in coagulated fiber. The occurrence of these microdomains in the draw zone as a precursor to the microfibrillar structure is believed to be the origin of low filament compressive strength. (C) 1994 John Wiley and Sons, Inc.