Effects of underlying films on the chemical-mechanical polishing for shallow trench isolation technology

We report on the use of mixed abrasive slurries ͑MAS͒ containing alumina and ceria abrasives for chemical mechanical planarization ͑CMP͒ of silicon dioxide and silicon nitride films for shallow trench isolation applications, extending an earlier investigation of alumina/silica MAS for the CMP of copper and tantalum films. These slurries show a polish rate selectivity between oxide and nitride films that is as high as 65 and show an excellent surface quality even without additives. Analysis of dried slurry particles using transmission electron microscopy indicates formation of a sheath of smaller ceria particles around larger alumina particles. Possible explanations and supportive arguments for the improved performance of MAS during CMP are presented based on the particle-particle and particle-film interactions.Chemical mechanical planarization ͑CMP͒ is widely accepted as an effective technique for meeting current ͑2002͒ and near future planarization requirements of the semiconductor industry. 1 Damascene and dual-damascene processing, coupled with CMP, have successfully implemented copper ͑Cu͒ interconnects in integrated circuit ͑IC͒ fabrication, 2 leading to greatly reduced resistancecapacitance ͑RC͒ delays and superior chip performance. Another important application of CMP occurs in the use of shallow trench isolation ͑STI͒ for device isolation. STI offers better dimensional control ͑trench and width͒ and greater packing density 3,4 relative to the earlier isolation achieved by local oxidation of silicon ͑LOCOS͒. Better planarity is achieved due to the elimination of local encroachment of field oxide. 5 In STI, isolation trenches are plasma etched in the silicon substrate using a thin nitride mask layer, which simultaneously acts as a stop and a capping layer. The trenches are overfilled with chemical vapor deposited ͑CVD͒ oxide. 4 As in dualdamascene patterning, the overburden oxide is removed using CMP while removing as little of the nitride as possible. Finally, the nitride is etched using hot phosphoric acid and active devices are fabricated in the exposed silicon areas. It is critical to minimize microscratching, particulate adhesion, and overpolishing during STI CMP. 4 Introduction of a thin nitride film combined with a ceria-based, highly selective CMP process for overburden oxide removal can eliminate the need for overpolish. 6,7 These highly selective ceria-based slurries, however, can generate microscratches and hence performance must be improved further for an effective application to STI CMP.Thus, an important area of current research in CMP continues to be the development of slurries that can eliminate or minimize the problems associated with the conventional CMP process. 8,9 Typically, a slurry consists of two phases, namely, liquid and solid phases. Liquid phase consists primarily of deionized ͑DI͒ water with several additives like oxidizers, complexing agents, inhibiting agents, and surfactants. Solid phase is comprised of abrasives, which are usually inorganic oxides, e.g., alumina, silica, cer...

mentioning

confidence: 99%

Chemical Mechanical Polishing of Dielectric Films Using Mixed Abrasive Slurries

Jindal¹,

Hegde²,

Babu

2003

“…Such a phenomenon is frequently encountered in the CMP process, and can now be attributed to the stress (curvature) effect. In addition, the so-called substrate effect, 16,17 in which the removal rate was found to increase or decrease depending on the substrate material beneath the layer being polished, may actually arise, at least partially, from the curvature effect described in the present study.…”

Section: Stress-induced Pressure Distribution and Polish Nonuni-mentioning

confidence: 62%

“…From Eq. 5, 10, and 15, the pressure can be expressed as a function of [16] Equation 16 will be singular at ϭ 0, (x ϭ Ϯa) unless we choose an a such that [17] [18]…”

mentioning

confidence: 99%

Effects of Film Stress on the Chemical Mechanical Polishing Process

Tseng

Wang

Chin

1999

A theoretical model is constructed to simulate the pressure distribution arising from wafer curvature (film stress) during chemical mechanical polishing (CMP), based on theories of elastic contact stress. Results from oxide CMP experiments suggest that the wafer curvature results in a nonuniform polish rate distribution across the wafer, in agreement with the simulation based on the model. This stress-dependent polish nonuniformity is attributed to the nonuniform pressure distribution across the wafer, induced by the wafer radius of curvature (film stress). Also, it was found that the magnitude of oxide film stress itself has little effect on removal rate. Oxides with tensile stress tend to have a weakened bond structure and enhanced chemical reactivity, both of which result in slightly higher removal rates. The reverse is true for oxides with compressive stress. Deviations from the model prediction may result from the stress induced by slurry flow, local variations in wafer shape and form, and pad surface properties.

“…Fixed abrasive polishing has special performance characteristics, making it well suited for the special needs of STI CMP. 4,5 For example, during fixed abrasive STI CMP, the polish rate drops significantly once planarity is achieved unlike in conventional STI CMP, [6][7][8] and a slight overpolishing ensures that any residual oxide on the wafer is removed. This helps not only in achieving total planarity but also in reducing the oxide overfill needed, lowering the oxide deposition costs and increasing the throughput.…”

mentioning

confidence: 99%

Study of Pattern Density Effects in CMP Using Fixed Abrasive Pads

Gorantla¹,

Venigalla

Economikos

et al. 2003

Pattern density effects during chemical-mechanical planarization ͑CMP͒ for shallow trench isolation applications using fixed abrasive pads were investigated in this work. Wafers patterned with various test masks, having a mix of pattern densities, were polished, and their polishing characteristics are correlated to their respective pattern density distributions in the dies. Single patterned density wafers were also polished in order to understand the affect of individual pattern densities on the polish rates as well as ''pad activation.'' It was observed that the low pattern density features in a die affected the polishing characteristics of the high pattern density features present on the same die in a wafer. Compared to the high pattern density structures, it was observed that the low pattern density features activated the pad more effectively.