Oxide Chemical Mechanical Planarization Research Articles

Single Particle Inductively Coupled Plasma Mass Spectrometry Study of Ceria Nanoparticle Size Distribution from 300mm Wafer Oxide CMP with Microreplicated Pads

This work describes the use of Single Particle Inductively Coupled Plasma Mass Spectrometry (spICP-MS) to measure ceria particle number concentrations and compare changes in size distributions to silicon dioxide wafer removal rates from different chemical mechanical planarization (CMP) processes. Particle number concentrations were measured for the 21 to 559 nm size range at 1nm size resolution. Changes in the ceria particle size distribution after CMP included a decrease in large (>130 nm) particles, an increase in small (<40 nm) particles, an increase in the total number of particles, and a decrease in median particle size. The decrease in median size was as high as 7% and influenced by slurry flow rate, wafer pressure and pad type. A novel microreplicated CMP pad was used which requires no pad conditioning to ensure consistent pad surface features, and the effect of different pad types on removal rate and particle size was isolated. A decrease in the median particle size correlated with higher silicon dioxide removal rates (R2=0.96) for a series of pad types with unique combinations of chemistry and surface features. This new combination of nano particle metrology and control of pad surface features is an improved tool set for modeling advanced CMP processes.

Effect of Applied Pressure on the Ceria Chemical Mechanical Planarization

Ceria abrasive is widely used in the Inter layer dielectric (ILD) Chemical mechanical planarization (CMP) due to its high removal rate and high selectivity ability. Despite the relatively low hardness of the ceria, Ce-O-Si bonding between silicon oxide surface and ceria abrasive enables high removal rate of silicon oxide CMP. Therefore, it is important to investigate the Ce-O-Si bonding for further study of the post CMP cleaning as well as ceria CMP. In this study, the effect of the applied pressure on the ceria CMP was investigated. Dipping and polishing process was conducted to confirm the surface change with applied pressure. The Fourier-transform infrared spectroscopy was measured to compare between dipped wafer and polished wafer. Fig 1 showed FT-IR spectra of each samples. Bare wafer and dipped wafer show similar spectra. Meanwhile, polished wafer shows different spectra in the range of 1000-1200 cm-1 and 800-900cm-1 which were associate with the Si-O-SI and Si-O bonding. The intensity of the both peak decreased due to the Ce-O-Si bonding.Zeta potential is a key factor for adsorption behavior of the ceria CMP. Generally, ceria particle can adsorb on the silicon oxide surface due to electrostatic attraction with opposite zeta potential. however, adsorption phenomenon was different as function of the applied pressure. The applied pressure plays an important role in the adsorption behavior of ceria as well as the zeta potential. This research suggests the importance of the applied pressure for the ceria CMP process.Figure 1. FTIR spectra of the silicon oxide wafer in region of 600-1600 cm-1 Figure 1

Understanding the Role of Slurry Additive Molecular Structure on Modulating the Interfacial Dynamics for Oxide CMP Applications

As feature sizes continue to shrink well beyond the 7-nm node, understanding the delicate balance present in the chemical mechanical planarization (CMP) process is of utmost importance. In order to achieve high through-put and defect-free CMP processes it is critical to develop predictive analytical techniques that directly correlate to macroscopic STI CMP performance metrics (i.e. oxide/nitride removal, defectivity, and dishing/erosion). This work employed a suite of techniques to monitor the CeO2 nanoparticle interfacial redox processes in the presence of structurally diverse rate modulating additives. Specifically, utilizing a UV-Vis spectroscopic technique, the Ce3+/Ce4+ ratio in the presence of different slurry additives was monitored and proved to directly correlate to removal rate performance (i.e. an increase in Ce3+/Ce4+ ratio shows an increase in rate). This finding coupled with the rate of dissolved O2 evacuation and a modified QCM technique determined the mode of interaction/adsorption which validates that the mechanism of oxide removal does not strictly depend on redox capacity, but also depends on the dynamic O2 equilibrium at the CeO2 nanoparticle surface. In this work, the additives selected provide multiple modes of interaction spanning from noncovalent surface adsorbers to redox active functional groups. It was determined that the modulation of oxide removal was directly related to the distribution of interactions (i.e. steric vs redox) and was highly dependent on the slurry additive functionality.

Single Particle Inductively Coupled Plasma Mass Spectrometry Study of Ceria Nanoparticle Size Distribution from Oxide CMP with Microreplicated Pads

This work describes the use of Single Particle Inductively Coupled Plasma Mass Spectrometry (spICP-MS) to measure ceria particle number concentrations and compare changes in size distributions to silicon dioxide wafer removal rates from different chemical mechanical planarization (CMP) processes. Particle number concentrations were measured for the 21 to 559 nm size range at 1 nm size resolution. Changes in the ceria particle size distribution after CMP included a decrease in large (>130 nm) particles, an increase in small (<40 nm) particles, an increase in the total number of particles, and a decrease in median particle size. The decrease in median size was as high as 7% and influenced by flow rate, pressure and pad type. A novel microreplicated CMP pad was used which requires no pad conditioning to ensure consistent pad surface features, and the effect of different pad types on removal rate and particle size was isolated. A decrease in the median particle size correlated with higher silicon dioxide removal rates (R2 = 0.96) for a series of pad types with unique combinations of chemistry and surface features. This new combination of nano particle metrology and control of pad surface features is an innovative tool set for modeling advanced CMP processes.

Evaluation of correlation between chemical modification state of pad and polishing rate in oxide chemical mechanical planarization

In the chemical-mechanical planarization of oxide films, the polishing rate (the rate of material removal) decreases if the polishing pad is not regularly conditioned. Pad brushing, which removed polishing debris but did not otherwise alter the pad surface, was unable to restore the original polishing rate. On the other hand, pad conditioning, in which the pad surface was scraped off, restored the original polishing rate. The morphological state and the chemical modification state of the pad surface were evaluated as factors affecting the polishing rate. FTIR (Fourier-Transform infrared spectroscopy) analysis showed that some of the polyurethane molecules of the pad surface were hydrolyzed. The degree of pad modification at each position in the pad radial direction was quantified based on the FTIR results, and the degree of pad modification was compared with the actual polishing profile within a wafer. The profile based on chemical modification of the pad matched the actual polishing profile. It was shown that not only the morphological factors of the pad but also the chemical modification factors greatly affected the polishing rate.

An Oxide Chemical Mechanical Planarization Model for HKMG Structures

In order to predict the surface topography, a chip-scale oxide chemical mechanical planarization (CMP) model is proposed in poly opening polishing processing of high-k metal gate (HKMG) structures. A geometrical chemical vapor deposition model is first constructed to describe the influence of design pattern effects on the initial surface profiles of dielectric oxides. Based on contact mechanics, the global pressure distribution is then computed to incorporate the long-range features of design structures and polishing pad deformation. The global pressure is further combined with local pattern effects to construct a feature-scale material removal rate model, which can be utilized for surface topography simulation of design chips. The predicted results of the proposed CMP model are consistent with the experimental data collected from the test patterns of HKMG CMP process. Consequently, the oxide CMP model has good potentials of predicting the surface planarity of polishing wafers and optimizing some design rules for HKMG design layouts.

Polystyrene-Core Silica-Shell Composite Abrasives: The Influence of Core Size on Oxide Chemical Mechanical Planarization

Polystyrene (PS)-core silica-shell composite particles with various core sizes were synthesized and applied in chemical mechanical planarization (CMP). The structures of the obtained PS/SiO2 core/shell particles (CSPs) were characterized by field-emission scanning electron microscopy (FESEM), transmission electron microscopy (TEM), and dynamic light scattering (DLS). The effects of the PS core size on oxide CMP behaviors were investigated by atomic force microscopy (AFM) based on the surface profilogram, roughness, and material removal rate (MRR). The roughness of the substrate surface slightly increased with increasing core size, while the MRR obviously increased. The lowest root-mean-square roughness value (0.198 nm) was obtained for the PS/SiO2 CSPs with the smallest cores (216 nm), while the highest MRR (148.9 nm/min) was observed for the CSPs with the largest cores (552 nm). In addition, these polishing results were rationalized according to the indentation-based mechanism, incorporating the total number of active abrasives and the particle indentation depth into the wafer surface.

UV VIS Studies of Ce(III)/Ce(IV) Redox Reactions to Understand Abrasive and Work Surface Interactions during STI CMP

In the current fabrication processes for integrated chips (IC), shallow trench isolation (STI) scheme is used to isolate and insulate transistors. Chemical mechanical planarization (CMP) is a key step in STI integration scheme. High selectivity slurries (HSS) which yield high silicon dioxide to silicon nitride polish rate are often used, and this is achieved by controlling the solid loading and the chemical concentration of additives [1]. Ceria based slurries with suitable additives are known to provide high selectivity [2, 3, 5]. Ceria is known to interact chemically with the silicon dioxide surface during CMP [2,3] and it is proposed that Ce(III) might play a role in the chemical interaction [4], but the exact nature of the species interacting with oxide are not known.Addition of hydrogen peroxide (H2O2) to ceria based slurries is reported to suppress the oxide and nitride removal rates [4]. However, when other abrasives are employed, hydrogen peroxide does not modify the removal rate significantly [4]. Ce (III) and Ce (IV) ions are known to exhibit UV VIS absorption peak at different wavelength regions.In this work a study on the kinetics of Ce (III) and Ce (IV) ions in the presence of H2O2is undertaken using UV-Visible Spectroscopy. Ce (III) chloride and Ce (IV) sulfate are employed for the initial study and the effect of hydrogen peroxide on changing the equilibrium composition is analyzed. The absorption is also monitored as a function of pH and time, to understand the dynamics of the system. Further, ceria abrasives are employed in the study and the nature of the species dissolving into the aqueous medium in presence of and absence of hydrogen peroxide, at different pH values are analyzed. The results show that the equilibrium between Ce (III) and Ce (IV) is shifted by the hydrogen peroxide which may help explain the suppression of oxide and nitride CMP rate by the addition of peroxide to ceria based slurries.

Optimization and Mechanism on Chemical Mechanical Planarization of Hafnium Oxide for RRAM Devices

Chemical mechanical planarization (CMP) of hafnium oxide including chemical and mechanical factors has been investigated in this paper. We demonstrated that the introduction of sodium fluoroborate (NaBF4) into hafnium oxide CMP has efficiently improved the removal rate from 5 nm/min to 95.5 nm/min. According to the static etch rate and effects of down force and platen rotation rate on hafnium oxide CMP, the Preston equation RR = kPV was modified to RR = kPV + Rc which suggests that the hafnium oxide CMP is determined by the combination of chemical and mechanical process. Moreover, we proposed a possible mechanism that the reaction of introduced NaBF4 with hafnium oxide provides a softer surface layer, enabling hafnium oxide to be easily polished and removed. Therefore, the removal rate of hafnium oxide was substantially enhanced. Using this method, the hafnium oxide surface with sub-nanometric roughness (RMS) was achieved, which can provide the related planarization material and technology for hafnium oxide nano-devices.

Pad roughness evolution during break-in and its abrasion due to the pad-wafer contact in oxide CMP

Chemical–mechanical planarization (CMP) is one of the most demanding process steps in interconnect integration, because it is influenced by numerous variables. Among them is the roughness of the polishing pad. Therefore we systematically characterize the break-in process and the pad surface abrasion due to the pad-wafer contact in absence of conditioning. For this we use the pad roughness characterization methodology presented in [1] to extract important roughness parameters like the mean asperities radius of curvature and the asperity density. The subsequent comparison of the pad surface characteristics at different process stages allows the estimation of the break-in time required to achieve a steady state pad texture that enables stable removal rates from wafer to wafer. In addition, by characterizing the abrasion of the pad asperities due to their interaction with the wafer, the contact height between pad and wafer is estimated and the removal rate decay in the absence of conditioning can be modeled.

(Invited) Performance Analysis of a Novel Slurry Injection System for Oxide Chemical Mechanical Planarization

In this study, a novel slurry injection system (SIS) was installed on an Ebara F-REX200® polisher and its performance was analyzed for oxide chemical mechanical planarization process. The slurry injector contacts the pad with a light load and applies slurry in a thin layer at the device trailing edge using single or multiple injection points. Used slurry, rinse water and pad debris are then removed from the pad by the injector leading edge. Blanket 200-mm oxide wafers were polished using Cabot Microelectronics Corporation Semi-Sperse™ 25 slurry at different flow rates (100, 150 and 200 ml/min) on Dow IC1000™ k-groove and XY-perforated pads. The results show that at each flow rate, the oxide removal rate is enhanced sufficiently over the rate from the standard slurry applicator that the flow rate can be reduced by about 35% on the XY-perforated pad and 50% on the k-groove pad without sacrificing removal rate.

Investigation of Source-Based Scratch Formation During Oxide Chemical Mechanical Planarization

The formation of scratches on silicon dioxide surfaces during chemical mechanical planarization in the semiconductor manufacturing process is a significant concern, as it adversely affects yield and reliability. In this study, scratch formation during CMP processing of the oxide surface was examined. The shapes of the resulting scratches were classified into three types: chatter mark type, line type, and rolling type. Chatter mark types were further subdivided into line chatter, broken chatter, and group chatter based on the shape. The effect of three different scratch sources (viz., pad debris, dried particles, and diamond particles) on scratch formation was comprehensively investigated. Chatter-mark-type scratches are predominant in the presence of agglomerated particles and pad debris. Group chatter marks are caused by the addition of pad debris. Unique scratch formation was observed on the wafer with different scratch sources. In particular, multiple-line-type scratches were observed in the presence of diamond particles.

Experimental Evaluation of the Effect of Pad Debris Size on Microscratches during CMP Process

Polishing debris generated by pad surface conditioning has been suspected as a major source of microscratches in the chemical–mechanical planarization (CMP) process. In this study, we investigated the pad debris generated by an in situ conditioning process during oxide CMP as one of the major scratch sources. We evaluated the relationship between the size of pad debris and the shape of microscratches on a wafer to find the cause of scratches. Pad debris was gathered in real time during the polishing process. Then, by transmission electron microscopy we observed a mixed layer of pad material and abrasive particles on the surface of the pad debris and the pad surface, which hardened the pad debris and pad surface. The results reveal a size range of pad debris that led to a minimum scratch count. Pad debris size smaller or larger than the minimum scratch region seems to cause higher scratch count due to the hardened pad surface and pad debris.

Generation of Pad Debris during Oxide CMP Process and Its Role in Scratch Formation

The generation of pad debris during the chemical mechanical planarization (CMP) process was studied. A fresh pad was conditioned with deionized water alone without polishing any wafer in order to characterize the pad debris exclusively. It was found that the pad debris presented in several size ranges with irregular shapes, mostly in agglomerated form. Polishing experiments were performed continuously without conditioning the pad, and the byproducts were gathered several times during the process. No pad debris was found in the gathered slurry during this process, but dispersion of some agglomerated silica particles was observed. The slurry was gathered during CMP processes conducted under both in situ and ex situ pad conditions. The generation of pad debris during polishing was observed with in situ conditioning but not observed with ex situ conditioning. The effect of the presence of pad debris on scratch formation was also studied, and it was found that the occurrence of pad debris seriously increased the number of scratches. An increase in the number of diamonds per unit area of a conditioner could ultimately avoid the generation of pad debris, eventually reducing the scratch formation by sustaining the same removal rate and nonuniformity after the CMP process.

Effects of diamond size of CMP conditioner on wafer removal rates and defects for solid (non-porous) CMP pad with micro-holes

Wafer removal rates and defects were investigated for 200 mm tetraethyl orthosilicate (TEOS) oxide chemical mechanical planarization (CMP) processes using two types of CMP pads: a porous pad and a solid pad with micro-holes. An initial CMP test conducted with fumed silica based-slurry and a conditioner with 180 μm diamond revealed that the wafer removal rates by the solid pad with micro-holes were approximately 10% lower than those by the porous pad, but scratch type defects were reduced. In order to increase the removal rate of a solid pad with micro-holes to the comparable level of a regular porous pad without changing process parameters, it was decided to modify conditioner design by using different diamond size from 70 to 130 μm. It was found that wafer removal rates increased from 2973 to 2587 Å/min and defect counts reduced from 5.3 to 1.7 by decreasing the diamond size from 180 to 70 μm in the case of the solid pad with micro-holes. Various pad surface analysis results, including contact area estimation and microscopic observations, also revealed that a smaller diamond conditioner generated the pad texture with finer and more regular pad asperities.

Effect of Cerium Oxide Particle Sizes in Oxide Chemical Mechanical Planarization

This study explored the effect of different cerium oxide abrasive particle sizes in chemical mechanical planarization of 200 mm blanket plasma-enhanced tetraethylorthosilicate wafers. All polishing experiments were done with a polisher and tribometer capable of measuring shear force and down force in real-time. Coefficient of friction and removal rate were found to correlate well with the slurry median particle size distribution. Removal rate modeling based on particle size was explored to support the interpretation of the experimental results.

Characterization of thermoset and thermoplastic polyurethane pads, and molded and non-optimized machined grooving methods for oxide chemical mechanical planarization applications

This paper systematically studies the effect of pad material, grooving method and grooving pattern on interlayer dielectric chemical mechanical planarization. The tested polishing pads consist of thermoplastic and thermoset polyurethanes synthesized using two different processes. Grooves created using a molding technique are compared with grooves formed by mechanical cutting. The concentric groove design is also compared with the logarithmic positive spiral positive grooving design. Experimental data collected include removal rate, coefficient of friction, shear force variance, pad temperature and dynamic mechanical analyzer measurements. Scanning electron microscope images are used to correlate grooving methods with coefficient of friction and shear force variance measurements. Results show that all of the pads polish wafers in boundary lubrication mode with unique friction coefficient, shear force variance and pad temperature characteristics. Simulations using a two-step removal rate mechanism are performed to estimate the chemical and mechanical rate constants. The analysis indicates that the thermoplastic pad is more mechanically controlled than the thermoset pad and that molded grooving induces a more mechanically controlled process than non-optimized machined grooving.

A Multiscale Bayesian SPRT Approach for Online Process Monitoring

Online monitoring of complex processes, such as semiconductor manufacturing processes, often requires the need to analyze sensor data with multiple characteristics. Some of these characteristics include nonstationary behavior, non-Gaussian distribution, high frequency of data generation, and multiscale (multiple frequencies) noise that mask the true nature of the process. Furthermore, it is necessary to implement process monitoring schemes that take into consideration the cost associated with sampling and incorrect decision making without sacrificing sensitivity, robustness, and ease of implementation. In this paper, a novel multiscale Bayesian sequential probability ratio test (MBSPRT) is developed, which is shown to be efficient in monitoring processes with the above characteristics. The MBSPRT method is also made suitable for online application by developing a moving block data processing strategy, which can match the data processing speed with the rate of data acquisition. The efficacy of the MBSPRT method was tested via detection of the end point occurrence in a chemical-mechanical planarization (CMP) process of semiconductor manufacturing using coefficient of friction (CoF) data. The proposed methodology offers a cost effective alternative to the traditional end point method, which is based on expensive metrology. Test results from both oxide and copper metal CMP are presented which show that MBSPRT is uniquely capable of identifying the start and finish of the end point event.

Experimental dynamics characterization and monitoring of MRR in oxide chemical mechanical planarization (CMP) process

Complex interactions and often dynamic variations in the process parameters and state variables are known to influence material removal rate (MRR) in the chemical mechanical planarization (CMP) of SiO 2 film deposited on a Si wafer. Experimental investigations reported in this paper support the nonlinear stochastic characteristics of the CMP process dynamics. Prior modeling and monitoring efforts have attributed much of the complex patterns of the signals from the CMP process to extraneous noise. Consequently, these models have somewhat limited predictability. Sensor features that quantify the nonlinear stochastic dynamics of the CMP process are found to be effective surrogates to track variations in certain process parameters. A monitoring approach based on combining these nonlinear dynamic features with conventional statistical descriptors of sensor signals as well as process parameter settings is found to improve the tracking of MRR—which is one of the most important performance variables in the CMP process—by over 20% (linear R 2>80%) compared to the use of conventional features in a linear regression setting. The results consistently hold for a variety of process conditions tested using a battery of designed experiments.

Composite Polymer Core–Silica Shell Abrasives: The Effect of the Shape of the Silica Particles on Oxide CMP

We use oxide chemical mechanical planarization (CMP) as a convenient test vehicle for comparing the material removal rate and defectivity induced by different silica abrasives. The effect of particle shape is investigated by comparing fumed and colloidal silica. In particular, the effect of a polymer core on the behavior of a composite abrasive is explored. Overall, fewer and shallower scratches are detected for the composites with a colloidal silica shell as compared with colloidal silica, as well as composites with a fumed silica shell. It is difficult to control the shape and dimension of the fumed silica and fumed silica-based composite agglomerates. Both the length and depth of microscratches increase with particle size and irregularity. Under the same pH conditions, fumed silica and fumed silica-coated abrasives lead to a comparable scratch depth. Fumed silica exhibits agglomerates of greater hardness but higher overall particle stability with respect to its composites, which show lower hardness and lower overall particle stability.