Key issues in CMP today include reduction of surface defectivity and enhancement of planarization efficiency. More specifically, the polished surface should be free of defects such as scratches, pits, corrosion spots, trench copper loss, and residue particles. For copper/low k CMP, one of the most promising strategies to accomplishing these goals is an Abrasive-Free Process (AFP). By eliminating abrasive particles from the process, either free or fixed to the pad, it has been anticipated and realized that defects such as severe scratching, particle contamination and slurry instability via particle aggregation or settling will be significantly reduced. In addition, with proper formulation, an abrasive free process can also yield an excellent over polishing window and desired step function of pressure for material removal rate. Coupled with a supramolecular design, some of the characteristic advantages seen in abrasive containing system such as step height reduction efficiency can be realized without the side effects often introduced from solid particles. In this presentation, some designing principles for an abrasive free system will be first presented. The potential advantages of a supramolecular design for copper CMP will be illustrated. The CMP performance on a set of testing blanket and patterned wafers will be discussed.

Effect of ceria particle-size distribution and pressure interactions in CMP of dielectric materials and the subsequent surface generation mechanisms is investigated in detail. The removal rate is observed to correlate primarily with the slurry mean particle-size distribution (D50) and reach early rate saturation with decreasing particle size. Slurries with tighter particlesize distribution exhibit a logarithmic relationship with pressure, while a linear relationship is observed for wider distribution slurries. In contrast to the removal rate, surface roughness and degree of microscratches depend primarily on the tail distribution (D99) and increase with increasing particle size. The addition of a selective component to the slurry increases the rate differential between the slurries.

This study presents the effect of pad properties, such as elastic modulus and surface roughness, on planarity in a CMP process. A systematic method to measure planarization length, which represents the die-scale planarity in a quantitative manner, has been proposed. It has been shown that the planarization length is highly dependent on the bulk modulus of the pad. The effect of elastic modulus and roughness of the pad on dishing amount, which represents the feature-scale planarity, has been shown. Dishing amount is determined by the elastic modulus of the superficial layer of the pad, which is typically tens of microns thick, rather than by the bulk elastic modulus of the pad. A double layer pad model has been proposed based on the observed results, which can explain that the dishing amount is reduced by increasing elastic modulus of the pad superficial layer, or by decreasing the surface roughness of the pad.

Two first-step copper damascene slurries and one commercial second-step slurry are characterized in terms of their intrinsic properties and CMP performance. A prototype first-step slurry with high static etch rate (∼150 nm/min) yielded higher dishing in the copper lines (∼200 nm in 100 μm lines) compared to a commercial first-step slurry with negligible static etch rate. In both the cases, dishing in copper lines is observed to be a strong function of line width and radial position on the wafer. High static etch rate of the prototype slurry is believed to be responsible for the high dishing. Non-selective second-step polishing removes the liner layer while maintaining planarity.

Conventional CMP for Cu/Ultra-low k (k<2.4) integration faces significant technical challenges [1-2]. The majority of ULK materials are made porous to reduce the dielectric constant, while trading off on the mechanical strength [3-6]. With diminished hardness, elasticity and adhesion, the CMP process has to be “kinder and gentler”: lower down force, lower relative velocity, softer pad, and slurry with lower abrasive content [1,7]. In a word, the mechanical portion of the planarization process would be greatly reduced. To maintain the same performance, one has to rely on the chemical reactions to make Cu/ULK CMP a viable process.

Spectral analysis of real-time friction data obtained during ILD CMP is used to elucidate the tribology of the process in terms of stick-slip phenomena. Fourier transform analysis is employed to quantify the total amount of mechanical interaction in the pad-slurry-wafer interface as a function of various IC-1000 pad surface textures, PL-4217 fumed silica concentrations, relative pad-wafer velocities and applied wafer pressures. A new parameter termed the ‘Interfacial Interaction Index’ (γ) is defined and determined empirically by integrating the amplitude of the force spectra over a wide range of frequencies. Values of γ extracted from individual force spectra are in qualitative agreement with the tribological information obtained in previous studies using Stribeck curve analysis. This new method is remarkable since it has the potential to eliminate having to perform a multitude of experiments needed for constructing and interpreting Stribeck curves. For a given tribological mechanism, analysis of the spectra for various types of pad textures indicates significant differences between the K-Grooved pad and other types of pads. A qualitative model relating the observed spectra to pad storage modulus is presented as a potential explanation for the above observation.

Copper has been observed to passivate in CMP slurries containing glycine, when hydrogen peroxide is used as an oxidant, even under acidic conditions where no solid oxidized phases appear on the potential-pH diagram. This passivation behavior is highly desirable for effective CMP. In contrast, passivation is not seen in slurries of similar pH and complexing agent concentration, where the potential is increased electrochemically. In order to model the effects of chemistry on CMP rates, we are endeavoring to better understand the mechanisms responsible for passivation in slurries containing hydrogen peroxide. Here we report tests that characterize the development and degree of passivation seen in slurries with a reasonably wide range of compositions.

This paper proposes a three-dimensional wafer process model for nanotopography. This model allows us to predict the effect and behavior of nanotopography on the wafer processes especially for the wafer Mechanical/Chemical Lapping (MLP/CLP), wafer Single/Double Side Polishing (SSP/DSP) and device Chemical Mechanical Polishing (CMP).

Two complexing agents, glycine and citric acid, in hydrogen peroxide based slurries for planarizing copper have been compared. Copper dissolution and polish rates and in situ electrochemical experimental results at various slurry pH values and hydroxyl radical concentrations at pH=8.4 are presented. It was observed that the pH of the slurry has a strong influence on copper dissolution and polish rates. While high copper removal rates were observed with citric acid-peroxide solutions at low pH values, glycineperoxide system yielded high Cu removal rates at alkaline pH values. Copper dissolution rates in both the systems at pH 4 and 8 were consistent with the electrochemical measurements. The concentration of hydroxyl radicals generated in citric acid-peroxide system was less than that of those generated in glycine peroxide system at pH=8.4 indicating low copper removal rates at alkaline conditions in the former system.

The process of copper chemical-mechanical planarization (CMP) can be considered as an erosion corrosion process. Such a process can be efficiently studied by in situ and ex situ electrochemical techniques, such as potentiodynamic scan and electrochemical impedance spectroscopy (EIS). Using a copper disk as the working electrode in an electrochemical cell, slurries with different oxidizer concentrations have been investigated with the aforementioned techniques. Corresponding dissolution tests were also studied and compared. It is shown that changing the oxidizer concentration leads to the formation of surface films with different structure and composition on the copper surface during CMP process. The nature of these films controls the rate of copper corrosion. These results could be used to explain the change of copper removal rate in different oxidizer concentration, as well as to understand the copper CMP mechanism.

Chemical Mechanical Polishing is a key technology for Cu damascene wiring process in integrated circuit (IC) manufacturing. It is important to understand the effects of mechanical and tribological properties of the coatings that form a part of Cu damascene structure on the CMP process in order to successfully evaluate and implement these materials. In this paper, we present tribological properties of different interlayer coatings (SiLKTM, Ta and Cu) during their CMP process using different Cu and barrier selective slurries. A micro CMP tester was used to study the fundamental aspects of the CMP process. The coefficient of friction (COF) and acoustic emission (AE) signals were acquired in situ to monitor the process and to see the difference in polishing behavior of these coatings. The issues of end point detection and slurry selectivity are discussed in detail.

Business/foundry driven requirements have necessitated the creation of modified ILD CMP processes for terminal metal-die planarization for BiCMOS SOC/MEMS applications. Therefore, despite much CMP process evolution, this paper ‘steps-back’ to test CMP assumptions first introduced within ILD process norms on these new applications at 10X typical process topographies. Evaluation of planarization targets, density effects, oxide budgets, as well as associated integration, throughput and metrology considerations within this expanded regime are discussed. ANOVA on inter- and intra-die (WIWNU and WIDNU) variation are used to quantify results throughout the work.

Organic passivating agents have been measured by a galvanostatic method during abrasive free Chemical Mechanical Planarization (CMP) of copper coated silicon wafers. CMP process steps include (a) removal of the passivation layer to expose the metal surface, (b) wet chemical oxidation of the metal surface and (c) mechanical removal of the oxidation products by means of the pad with a colloidal polishing aid such as silica or alumina1. CMP of necessity seeks to establish a unique equilibrium between little or no metal removal due to corrosion effects and rapid metal removal resulting from a mechanically aided chemical oxidation or polishing. Chemical oxidizing agents can corrode the metal surface during polishing and non-polishing steps alike so protective passivation agents are necessary to inhibit or block surface corrosion. Historically, benzotriazole (BTAH) has been the copper protective agent of choice because it performed well at concentrations of 50 to 500ppm, however BTAH can scratch copper. Self-assembly of monolayers was reported by Sung et al2 demonstrating a tendency toward organizational energy of molecular surface protection. Cho et al3 have reported formation of a flat, well ordered BTAH based monolayer on copper supporting a three dimensional BTAH stack above it. Metrikos-Hukovic et al4 reported evidence for a monolayer of a CuBTA compound and additional adsorption/crystallization layers of 7nm thickness5 were measured. Discussions in this work refer to the combined adsorption/chelation layers on copper as passivation providing measurements that distinguish between them. It will be shown that addition of BTAH forms a thin layer of CuBTA covered by a second layer creating a protective film with relatively high electrical resistance. The purpose of this electrochemical effort is to measure the effect of BTAH during a non-Prestonian, slurry free CMP process, to affect a direct measure of bi-layer passivation and to identify other organic compounds that may also provide protection of the copper surface against corrosion.

Nano-sized cerium oxide particles in the size range of 5 to 500nm have been synthesized for use in chemical mechanical planarization (CMP) applications. The CeO2 particles were prepared using cerium nitrate salts with bases such as ammonium water, ethylamine, or other alkylamine /polyalkylamine compounds. Other additives such as urea can be added to affect crystallization and size growth of the final particles. The particle sizes of the resulting CeO2 particles depend on the initial concentrations of cerium salts, additive concentration, and solution pH. Reaction duration seems to have little effect on CeO2 particle sizes. Temperature effects on size were moderate. These CeO2 particles showed excellent surface quality and desirable polishing rate during CMP tests.

This study explores the effect of pH on the chemical mechanical polishing (CMP) characteristics of copper in H2O2 and KIO3 based slurries under various dynamic and static conditions. High purity copper disc was used to study the dissolution and oxidation kinetics at various pH (2 to 10) with 5% H2O2 or 0.1M KIO3. Electrochemical techniques were used to investigate the dissolution/passivation behavior of Cu. The affected surface layers of the statically etched Cu-disc were investigated using X-ray photoelectron spectroscopy (XPS), and scanning electron microscopy (SEM). In 5% H2O2, the Cu removal rate decreases with an increase in pH and reaches minimum at pH 6, and then increases under alkaline conditions. XPS results indicate that the surface oxide formed at various pH values was responsible for this CMP trend. However, with 0.1M KIO3, the CMP removal rates were found to be lower at pH 2. The maximum was observed at pH 4, then the removal rate decreased with the increase of pH. The lower value of removal rate at pH2 was due to the fast interaction between Cu and KIO3 and the precipitation of CuI on the pad, which makes the pad glassy, resulting in lowered removal rates. This was confirmed by XPS measurements. The decreased CMP removal rates when the pH is higher than 4 might be due to the weaker oxidation power of KIO3 with the increase of pH.

Real-time coefficient of friction (COF) analysis is used to determine the extent of normal and shear forces during CMP and identify the lubrication regimes associated with the process. Pads with different surface textures and slurries with varying abrasive concentrations are used to polish ILD films over a wide range of operating parameters. Results show that by varying abrasive concentration and pad surface texture, one can cause the process tribology to change from ‘boundary lubrication’ to ‘partial lubrication’, to ‘hydrodynamic lubrication’. A two-phase model relating average coefficient of friction and Preston's constant is presented. At abrasive concentrations up to 9 percent, material removal is proportional to the extent of contact between the abrasives and the wafer. At abrasive concentrations between 9 to 25 percent, removal rate is directly influenced by average COF. A new parameter termed the ‘tribological mechanism indicator’ is defined and extracted from the data, which coupled with the information on COF and ILD removal rate, results in a series of ‘universal’ correlations. A qualitative model based on pad storage modulus is used to explain the trends.

We conducted fundamental investigation of interfacial interactions between copper and polyurethane surfaces. Using in situ surface analysis techniques we were able to evaluate effects of water molecules on both materials surfaces during rubbing. Evidence of transferring elements between copper and urethane surfaces due to friction was found. Further investigation on pad wear indicated that transfer wear was taking a major role during copper CMP and mixed wear modes dominated pad wear.

This paper presents temperature and friction force data at the pad-slurry-wafer interface during real time CMP polishing with in situ pad conditioning. Experiments are performed on a 1:2 scale laboratory tabletop rotary polisher with variable pad speed and wafer down force control. Dual emission laser induced fluorescence (DELIF) techniques are used to optically measure the temperature directly beneath the wafer during polishing using a two camera imaging system. An infrared camera and a thermocouple are alternately used to measure bow wave temperatures. Optically transparent BK-7 glass wafers with either concave (wafer edges sloping toward the pad) or convex (wafer edges sloping away from the pad) curvature were used. When concave wafers are polished, the bow wave temperatures are 3°C to 5°C higher than the corresponding value for convex wafers. Similarly, slurry temperatures under the concave wafers are 5°C to 6°C higher than the value for convex wafers (±0.5°C). The friction force per unit area is typically 2 kPa to 3 kPa higher for concave wafers. Temperatures beneath the wafer are as high as 12°C above the ambient temperature for a concave wafer at a high applied wafer pressure (41.4 kPa) or linear velocity (0.93 m/sec). Bow wave temperatures reach as high as 9°C above ambient at a linear velocity of 0.93 m/sec. The lowest temperatures, within 1°C of ambient at the bow wave and 5°C above ambient under the wafer, were found with convex wafers at low applied wafer pressures (20.7 kPa). Linear velocity has little effect on the slurry temperature while polishing convex wafers. Increasing slurry abrasive concentration causes an increase in temperature, despite a decrease in friction force. A correlation, with an R-squared value greater than 0.96, exists between the bow wave temperature and the temperature beneath the wafer. This correlation holds at constant linear velocities across wafer shapes, applied wafer pressures, and slurry concentrations.

Our group has proposed several chip-scale CMP models, with key assumptions including the notion of planarization length in the pattern density model [1], and step height dependent polishing rate in the density step height model [2]. In the effective density model, planarization length is the characteristic length of an elliptic weighting function based on the long-range pad deformation and pressure distribution during CMP. This semi-physical model is often adequate and usually gives a fitting error of a few hundred angstroms. As ever-shrinking device size pushes for tighter control of post CMP uniformity, however, we need a chip-scale CMP model with better accuracy.

In this work, we re-examine the physical basis for averaging weighting functions and step height dependence, particularly in the context of contact mechanics based model formulations. The comparison of the two models confirms that the analytical density and step height models can be viewed as approximations to the contact wear model. The study also suggests a new dependence of contact height on line space and pattern density.

The present investigation was focused on understanding of the oxidation, dissolution and modification of Cu surface in slurries at various pH using hydrogen peroxide as oxidizer, glycine as complexing agent and 3-amino-triazol (ATA) as inhibitor during Cu-CMP. The electrochemical process involved in the oxidative dissolution of copper was investigated by potentiodynamic polarization studies. Surface modification of copper was investigated using Xray photoelectron spectroscopy to understand the interaction of Cu-H2O2-glycine-ATA during CMP. In the absence of glycine and ATA, the copper removal rate is found to be high in a slurry with 5% H2O2 at pH 2, then it decreases with increasing pH and reaches the minimum at pH 6, it continuously increases at alkaline condition. In the presence of 0.01M glycine, the removal rate of copper decreases in acidic slurries while increases significantly in alkaline slurries. With the further addition of ATA, the copper removal rate was reduced. However, better surface planarity was obtained. The present investigation enhanced understanding of the mechanism of Cu CMP in the presence of oxidizer, complexing agent and inhibitor for formulation of a highly effective CMP-slurry.