Real-time coefficient of friction (COF) analysis is used to determine the extent of normal and shear forces during post-CMP PVA brush scrubbing and identify the tribology of the process. Fluid pH and applied pressure have pronounced effects on process tribology and the magnitude of COF. In cases where brush rotation is the only kinematic attribute of the system, low pH results in ‘mixed lubrication’ where increasing brush velocity causes a dramatic reduction in COF in accordance with classical tribological arguments. At lower pressures the tribological mechanism shifts to ‘hydrodynamic lubrication’ with significantly lower values of COF. Regardless of pressure, increasing the pH to 7 and 10.7 causes the tribology to remain in ‘hydrodynamic lubrication’ and results in even lower values of COF. Trends are explained qualitatively by considering the effect of pH on the solubility and gellation characteristics of silica in the silica-water system. Stribeck curves fail to shed light on the tribology of the system when complex ranges of tool kinematics (i.e. wafer rotation and brush oscillation superimposed on brush rotation) are employed. The results are counter-intuitive since they indicate nearly an order of magnitude increase in COF with increasing brush velocity. Nevertheless, results demonstrate the utility of having complex ranges of motion in PVA brush scrubbing such that very low or very high values of COF can be realized by simply increasing the rotational velocity of the brush while maintaining a constant wafer rotation and brush oscillation.

A combined experimental and theoretical approach has been devised to understand the mechanical behavior of CMP pad. The pad response is examined at different length scale utilizing a nano-indentation with a conical tip and a flat-punch. The experimentally observed trends showed the competition between the local contact field of the abrasive particle and the global porous pad cell deformation. From the experiments, the local particle-level contact, cell bending, long range pad surface asperity contact as well as the bulk response of the pad can be integrated into mechanism-based models for better understanding the applied force partitioning and to predict the characteristics of the material removal rates.

It is well known and accepted that the viscosity of CMP-slurries has an effect on polishing results. Even though the literature on rheology recognizes that viscosity is not always constant and the slurry can show non-Newtonian behavior or even dilatant effects, all calculations have been performed with constant viscosity.

However, the “real” viscosity of a CMP-slurry during polishing can change significantly with shear rate.

The typical equipment for viscosity measurement is based on a rotating cylinder or a plate. But even with a plate system it is only possible to reach a shear rate range up to 50,000 1/sec. A calculation of the shear rate between the wafer and the polishing pad is based on a relative velocity of 1 m/sec and a distance between the wafer and the pad of 20 μm; this correlates to a shear rate of 50,000 1/sec. If parts of the polishing pad come closer to the wafer or especially closer to the edge of structures on the wafer (for example 1 μm), the shear rate will increase locally to 1,000,000 1/sec.

When the shear rate is high enough, viscosity depends mostly on hydrodynamic factors like viscosity of continuos phase, solids content, particle size, particle size distribution and shape of the particles.

The shape of fumed metal oxides is controlled during the synthesis in the flame process. But the slurry-making process is also responsible for particle size distribution, shape of the particle and the high shear rate viscosity of the CMP-slurry.

The high shear rate viscosity of different silica slurries in dependence from BET-surface area, used milling energy, concentration and preparation direction was measured in this investigation.

Ceria containing slurries are increasingly used in the chemical mechanical polishing of CVD silicon oxide films to obtain STI structures. Unlike silica or alumina, ceria has redox characteristics. Because of this characteristic, removal of ceria particles from planarized surfaces may be possible using chemical reagents that can participate in redox reactions. One such reagent is hydroxylamine, which is already being used in copper CMP. The objective of the work reported in this paper was to characterize the reaction between ceria and hydroxylamine, especially with respect to dissolution of ceria particles. A kinetic study of the dissolution of ceria in hydroxylamine solutions maintained at various pH values has been performed. The extent and kinetics of dissolution of ceria has been determined by ICPMS. Removal of ceria particles from oxide surfaces using hydroxylamine-based chemistries has been investigated.

The drive for improved performance of microelectronic devices has led to the prevalence of copper metallization and the aggressive development of low-permittivity (low-κ) dielectric materials for use as interlayer dielectrics in BEOL interconnect structures. Progressive scaling of metal line widths coupled with the need to incorporate ultra low-κ (ULK) dielectrics, with κ<2.2, presents numerous challenges for integration and reliability. Perhaps the most significant challenge for the 90nm technology node and beyond is successful planarization of Cuinterconnect structures by chemical mechanical polishing (CMP). The present paper will discuss the general integration challenges and key structural reliability issues for chemical mechanical polishing of Cu-interconnects incorporating ULK dielectric materials.

Copper metallization in sub-0.18 μm semiconductor devices is achieved by combining the dual damascence techniques followed by chemical mechanical planarization (CMP). Tantalum and its nitride have been identified as the diffusion barrier layer for copper metallization. However, the wide differences in properties between copper and tantalum layers result in selectivity problems during CMP process. The aim of this work is to obtain a better understanding on the slurry selectivity for copper and tantalum and to develop slurries with best selectivity performance. In this work, the effect of several chemical parameters (abrasive type, oxidizer type, concentration, pH etc.) was studied through static and dynamic tests using advanced electrochemical techniques and surface analysis techniques. The surface layers of the statically etched copper and tantalum discs were investigated using X-ray photoelectron spectroscopy (XPS) and surface planarity was studied using atomic force microscopy (AFM). Polishing rates results show that alumina-based slurry polished copper very well whereas tantalum removal rate was low. However, for the silica-based slurry the tantalum shows much higher removal rate than copper and better surface planarity was obtained.

The polishing pad is one of the prime components in a typical Chemical Mechanical Polishing (CMP) process. The structure and transport properties of a polishing pad are critical in determining the particle and chemical utilization in a conventional CMP process. Our earlier paper investigated the particle retention and transport on two different polishing pads, IC-1400 and Suba-500, during copper polishing. In this paper, the results of chemical retention and transport of IC-1400 and Suba-500 pads during copper polishing are presented. The polish rate results from slurry-step-flow experiments with H2O2-glycine based slurries, where the concentrations of chemicals in the slurry are altered in steps during polishing, are correlated to the chemical retention and transport characteristics of these pads. It is found that IC-1400 has a higher chemical transport capability than Suba-500 pad, which is shown to affect the polish rates of copper.

The residence time distribution of slurry in the pad-wafer interface was experimentally determined and used to calculate the slurry utilization efficiency (η) of the CMP process. Slurry utilization efficiency represents the percentage of slurry that actually participates in the polish by entering the region bounded between the wafer and the pad. Results show that η ranges from 2 to 22 percent, depending on operating conditions such as applied wafer pressure, relative pad wafer velocity, slurry flow rate and pad surface texture (i.e. type of pad grooving).

A new approach using Fixed Abrasive (FA) pads has been undertaken to overcome the problem of non-uniform thick film Silicon-on-Insulator (SOI) wafers after CMP polishing. The theoretical models indicating the advantages of the 2-body system of the fixed abrasive configuration vs. the conventional 3-body system of slurry based polishing have been convincingly demonstrated in practise upon experiments in a wide range of parameters. As a result it is possible to maintain or improve the flatness of wafers after back grinding, while simultaneously removing the sub-surface damage. A surface quality of prime wafers can be reached on the device layer. Capacitive thickness measurement scans and atomic force microscopy (AFM) monitoring confirm the results. A detailed comparison with conventional processing has been carried out to clarify the advantages on bulk silicon wafers. Decoration etching is used to analyse the wafer surface quality in terms of oxide induced stacking faults (OISF). As a result an alternative processing method is proposed for manufacturing thick film SOI substrates with improved uniformity.

Chronoamperometry was used to investigate the reaction/passivation kinetics and thickness of the chemically modified surface layer on the copper during chemical mechanical polishing (CMP). The result showed that the reaction/passivation kinetics and the thickness of the chemically modified surface layer are strongly dependent on the chemistry of CMP slurry in the chemical aspect of CMP and play critical keys in the selection of the chemistry and its concentration. BTA and H2O2 enhanced the passivation kinetics, resulting in thinner layer on the copper surface. In addition, the reaction kinetics increased as pH decreased.

A previously presented model of CMP is extended to include the role of inhibitors. In CMP, a chemical reaction forms a surface film which is removed mechanically by abrasives. When inhibitor molecules bond to the surface film, the mechanical abrasion rate is reduced. The general model will be discussed, and then applied to W-CMP explaining differences in the reduction of polishing rates for different inhibitors.

A novel method for forming interdigitated electrodes for GHz Surface acoustic wave (SAW) devices on diamond using a damascene-like polishing technique is described. Low aspect ratio Al electrodes are recessed into the diamond substrates to create a near planar surface. This allows the deposition of higher quality oriented ZnO. CVD diamond samples were ion-beam etched with groove widths in the range 1-5μm to a depth of 80-120nm. Al (120nm thick) was sputter deposited on to the etched diamond and was polished using either (i) a solution of 0.05μm silica on a neoprene polishing cloth or (ii) using a ∼0.12mm thick PVDF (polyvinylidene difluoride) filter pad with colloidal solutions of silica. Using AFM scans and optical microscopy it was observed that the combination of method (i) with the small sample size resulted in significant hollowing-out of the Al electrodes and complete removal of Al from over 90% of the pad-interconnect area. The alternative technique in (ii) demonstrated far less Al removal and better uniformity in the electrode patterns.

Soak of polyurethane-based CMP pads in tungsten slurry and de-ionized water and its effect on retention of thermal and mechanical properties of the pads was studied using Dynamic Mechanical Analysis (DMA), Thermal Mechanical Analysis (TMA), Thermal Gravimetric Analysis (TGA), and Modulated Differential Scanning Calorimetry (MDSC). Simultaneous cross-linking and plastisizing due to soak were established using DMA and MDSC analysis. The stable operating temperature range and its dependence on soak time were determined using TMA analysis. Substantial difference in diffusion behavior of the “soft” and “hard” pads was discovered: diffusion into the hard pads followed Fickian law [1], while diffusion into the multi-layer soft pads was dominated by the fast filling of the highly porous pad surface with liquid.

During a traditional CMP process, which involves application of polishing pads and slurry, the pad properties can be substantially and irreversibly changed as the result of slurry/rinse water absorption.

The retention of the pad properties after exposure was monitored using such thermal and mechanical techniques, as Thermal Mechanical Analysis (TMA), Dynamical Mechanical Analysis (DMA), Modulated Differential Scanning Calorimetry (MDSC), Thermal Gravimetric Analysis (TGA).

The development of a direct polish process for STI CMP on 200mm wafers using highselectivity slurry (HSS) has been achieved for production of 0.13μm technology microelectronic devices. The new process has improved on-wafer performance compared to standard STI CMP processes. The step height range across the wafer was decreased by 84%, planarity Cpk values (silicon nitride thickness and step-height uniformity) were increased by >25%, leakage current statistics were superior, and the cost of ownership was lowered by 78%. Cross-sectional SEMs both after direct polish CMP and after removal of the silicon nitride show improved planarity.

In this work, we present an integrated prediction of thickness variations in electroplating and chemical mechanical polishing (CMP) processes across an entire chip for random layouts. We achieve chip-scale prediction by first calibrating both electroplating and CMP models with experimental data using the same test mask. Using the calibrated plating model in conjunction with a discretized and binned layout extraction for a random chip layout, a prediction of plated copper topography is then performed. Finally, using the plated thickness prediction as the initial condition, the CMP model predicts the dishing and erosion across the chip. Layout geometry extraction for each discretized region of a chip as well as layout parameter manipulation and model output integration are all critical elements, in addition to the model development itself, enabling the integrated chip-scale prediction of final copper interconnect thickness variation.

This study explores aspects of the fluid dynamics of CMP processes. The residence time distribution of slurry under the wafer is experimentally determined and used to calculate the Dispersion Number (Δ) of the fluid in the wafer-pad region based on a dispersion model for non-ideal reactors. Furthermore, lubrication theory is used to explain flow behaviors at various operating conditions. Results indicate that at low wafer pressure and high relative pad-wafer velocity, the slurry exhibits nearly ideal plug flow behavior. As pressure increases and velocity decreases, flow begins to deviate from ideality and the slurry becomes increasingly more mixed beneath the wafer. These phenomena are confirmed to be the result of variable slurry film thicknesses between the pad and the wafer, as measured by changes in the coefficient of friction (COF) in the pad-wafer interface.

A new cleaning technique using gas dissolved water has been found to be effective in protecting against corrosion in aluminum metallization, which is also useful in post-cleaning of Cu CMP as already reported. Corrosion can be a significant concern in aluminum as well as in copper metallization. In Al CMP process, corrosion often occurs in contact with DIWbecause of the big potential difference between Al and the barrier metal. In this study, it has been confirmed that gas dissolved water is able to change the potential of various metal films, and that following Al CMP, post-cleaning using gas dissolved water instead of DIW can successfully protect against corrosion.

Furthermore this technique is also effective in the post-cleaning process of Al wiring formed by RIE, which contains a slight amount of Cu. Because segregation of copper at the side-wall of wiring often occurs during RIE and the following processes, Al-Cu wiring is easily corroded during the DIW rinsing step in wet cleaning, which may cause killer defects. Gas dissolved water can remarkably decrease the potential difference between Al-Cu alloy and segregated Cu, as well as Al-Cu alloy and the Ti/TiN used as a barrier metal. Moreover, because it seems that a reaction occurs between cathode water and Ti/TiN, where electrons may be supplied there to corrosive sites, corrosion can be prevented.

Lower mechanical strength, reduced cohesive strength and lack of compatibility with other interconnect materials, are the major challenges involved in chemical mechanical polishing (CMP) of Cu metallization with ultra low-k materials as interlayer dielectrics. In this study we have investigated the polishing behavior of patterned Cu samples with underneath different low-k materials using two different slurries and a wide range of machine parameters. CMP micro tribometer was used to polish the samples with different rotations of platen (50 to 250 RPM) and down forces (1-6 PSI). Friction co-efficient and wear behavior were also investigated at different conditions. Optical and scanning electron microscopy was used to investigate the polished surface. It was observed that the two different Cu slurries used for polishing have marked effects on the polishing of Cu-low-k stack with respect to wear and delamination.

Nanometer scale wear caused by a single asperity of silicon nitride was examined by measuring the wear caused by atomic force microscope tips translated against sodium trisilicate glass, soda-lime glass, or fused silica in aqueous solutions. As a function of contact force, FN, and scan duration, t, the wear to both tip and substrate scales approximately as (FNt)0.5. Substrate wear was independent of temperature from 5°C to 60°C, whereas tip wear showed a temperature dependence on soda-lime glass corresponding to an activation energy of 60 kJ/mol on soda-lime glass.

With an aim to understanding the fundamental mechanisms underlying chemical mechanical planarization (CMP) of copper, we simulate the nanoscale polishing of a copper surface with molecular dynamics utilizing the embedded atom method. Mechanical abrasion produces rough planarized surfaces with a large chip in front of the abrasive particle, and dislocations in the bulk of the crystal. The addition of chemical dissolution leads to very smooth planarized copper surfaces and considerably smaller frictional forces that prevent the formation of bulk dislocations. This is a first step towards understanding the interplay between mechanistic material abrasion and chemical dissolution in chemical mechanical planarization of copper interconnects.