Slurry Design for Chemical Mechanical Polishing

Chemical Mechanical Polishing (CMP) process is widely used in the microelectronics industry for planarization of metal and dielectric layers to achieve multi layer metallization. For an effective polishing, it is necessary to minimize the surface defects while attaining a good planarity with optimal material removal rate. These requirements can be met by controlling the chemical and mechanical interactions during the polishing process, or in other words, by engineering the slurry chemistry, particulate properties and stability. This paper reviews the impact of chemical, inter-particle and pad-particle-substrate interactions on CMP performance. It is shown that for consistently high performing slurries, stability of abrasive particles must be achieved under the dynamic processing conditions by providing sufficient pad-particle-wafer interactions.


Introduction
Chemical mechanical polishing (CMP) is a surface smoothing and material removal process, made possible by combination of chemical and mechanical interactions.The wafer surface (on which microelectronic devices are to be built) is planarized by a slurry consisting of submicron sized abrasive particles and chemical reagents.The ultimate goal of CMP is to achieve an optimal material removal rate creating an atomically smooth surface with minimal number of defects, while maintaining global planarity.The chemical effect in CMP is provided by the addition of pH regulators, oxidizers or stabilizers depending on the process.The mechanical action on the other hand, is mostly provided by the submicron sized abrasive particles contained in the slurry, as they f low between the polishing pad and the wafer surface.
As the rapid advances in the microelectronics industry demand a continuous decrease in the sizes of the microelectronic devices, minimal material removal with atomically f lat and clean surface finish has to be achieved during manufacturing (1).This requires close monitoring of the CMP process that can be attained by controlling not only the operational variables (such as the applied head pressure and the relative pad velocity) but also the slurry particulate properties and chemistries.This paper focuses on the development of consistently high performing CMP slurries by selecting effective (i) chemistries, (ii) particulate properties and (iii) methods of stabilization.
tures are repeatedly oxidized by the chemicals and abraded by the slurry particles until the planarization is reached.Consequently, slurry oxidizers provide topographic selectivity in metal CMP.In silica polishing on the other hand, chemical effect is provided by increasing the slurry pH so that silica dissolution rate is very high (4) forming a soft hydrated surface layer due to the weakening of silica bonds (5)(6)(7)9).In both cases, material properties of the chemically modified layer are expected to be different than the film to be polished, which gives control over the CMP process.The absence of chemically modified layer was found to result in no material removal in tungsten polishing (10).Similarly, in silica CMP, material removal rate was observed to decrease with the decreasing slurry pH, where the chemical activity (dissolution of silica) is reduced (11).Polishing mainly by chemical means leaves an isotropically etched surface with no planarity, whereas polishing by only mechanical means a rough surface is obtained.As a result, the formation of a chemically altered film is necessary to achieve an optimal CMP performance and its properties must be tailored by the design of effective slurry chemistries.

Effect of Slurr y Particulate Properties
Abrasive CMP slurry particles provide the mechanical action during the polishing process.The chemically modified surface layer of the wafer is abraded continuously with the submicron size slurry abrasives resulting in material removal.To achieve an optimal polishing performance with minimal deformations and good planarity, it is necessary to optimize the rates of chemical modification and mechanical abrasions.The intensity of the mechanical abrasion also varies with the slurry particle size and concentration as these factors determine the load applied per particle.Furthermore, the frequency of abrasion depends on the number of slurry abrasives in contact with the wafer surface.Therefore, slurry particle size and concentration as well as the particle size distribution are very important factors in determining the polishing performance and should be studied carefully to understand the CMP process.A small variation in the slurry particle size distribution by oversize particle contamination or due to slurry destabilization may result in major changes in the particle-substrate interactions.Consequently, the material removal rate response may vary resulting in poor process control and the number of surface deformations may increase giving rise to defective microprocessors.

Slurr y Particle Size and Concentration
Effects of slurry particle size and concentration on CMP material removal rate response have been investigated by a number of researchers (5,(12)(13)(14)(15)(16)(17).The reported results however, are often contradictory in relating the material removal rate to the particle size.This discrepancy can be explained based on the type of the CMP process as well as the selected ranges of the particle size and concentration.
In a study on tungsten CMP (10), material removal rate was observed to increase with the increasing number of particle contacts, achieved by increasing the solids content of the slurry and decreasing the abrasive particle size.In silica CMP however, a maximum removal rate was reached for any given particle size at a particular solids concentration (11).These findings indicate that different polishing mechanisms may become predominant as a function of the type of CMP application as well as the slurry particulate properties for a given polishing system.
Two material removal mechanisms have been introduced to clarify the observed variations in the polishing rates, which can be explained on the basis of silica CMP explicitly (10).Figure 1 illustrates the material removal rate obtained with 0.5µm particle size silica slurry as a function of the slurry solids concentration.At low solids loading of the slurries (0.2-5%wt.), material removal rate increases with the increasing solids concentration.The surface micrographs of the wafers obtained with the Atomic Force Microscopy (AFM) illustrates scratches on the polished wafer surface at these concentrations indicating that the abrasive particles were indenting and 'sliding' across the wafer surface resulting in mechanical removal.This type of material removal is defined as the indentation mecha- nism (10).A maximum was observed in the material removal rate at 5wt% solids concentration after which a significant decrease was detected.This transition occurs due to the change in the particle motion as the load per particle decreases with the increasing number of particles in contact with the wafer.More particles tend to start rolling rather than sliding across the wafer surface as the solids loading is increased from 5 to 15wt% (18).Consequently, total indent volume and the material removal rate decrease with the increasing solids concentration.AFM surface micrograph at 15wt% shows pitting type of deformation on the wafer surface rather than scratches, indicating that abrasives are in rolling motion.Hence, at these higher solids concentrations, mechanical indentation reduces significantly and material removal occurs predominantly due to chemical interactions.This type of material removal mechanism is defined as the contact area mechanism (10).In support of the proposed mechanisms, in-situ friction force measurements conducted on the same system have also shown similar trend in the total friction force as a function of the slurry solids loading for 0.5µm size slurry (Figure 1).In situ friction force measurements in CMP have been shown to correlate with the material removal rate response previously (19)(20).As the indentation of the particles become negligible due to the decreasing load per particle with the increasing number concentration, particles start rolling resulting in decreased friction force.To reduce the number of surface deformations in CMP, it is necessary to operate in the contact area regime that requires the use of small size particles at high solids concentrations.

Slurr y Particle Size Distribution
In order to achieve an optimal polishing performance with minimal surface deformations, it is necessary to use monosized particles for the CMP slurries (5).In practical applications, however, there may be a few oversize particles in the slurries in the form of larger size particles (hard agglomerates) due to insufficient filtration, or the agglomerates of the primary slurry particles (soft agglomerates) due to poor stability.Presence of agglomerates in the CMP slurries result in unequal distribution of the applied head load on the abrasives, which may lead to surface deformations (21).
Hard Agglomerates: Although the presence of hard agglomerates was suspected to result in major surface deformations (5,21) their impact on polishing performance was only recently quantified in a systematic study (22).Polishing tests conducted in the presence of hard agglomerates at the established detection limits verified significant degradation in the polishing performance.Surface analyses of the silica wafers polished with spiked slurries showed increased surface roughness, and more surface deformations relative to the baseline polishing as illustrated by the AFM images in Figures 2a and b.In addition, significant variations were detected in the material removal rate response in the presence of hard agglomerates indicating that they have to be removed from the slurries not only to protect the surface quality but also to achieve consistent material removal rate.
Soft Agglomerates: To remove coarser particles, filtration of CMP slurries is commonly practiced.Nevertheless, even after filtering the slurries, the defect counts on the polished surfaces are often observed to be higher than expected (26).It has been suspected that some of the defects may be created by the agglomerate formation during the CMP operations.A study conducted on silica-silica system by substituting a fraction of baseline slurry with dry aggregated, polymer f locculated and salt coagulated particulates have shown that even the agglomerates, which breakdown under the applied load can result in major surface deformations (Figure 2c) (27).These observa- tions indicate that CMP slurries must remain stable during polishing to obtain optimal polishing performance.
There are several size analysis techniques available for detecting the larger size particles in the CMP slurries (23)(24)(25).Among these, the number counting techniques are the most promising in detecting the large size particles at the lowest concentrations (24).However, as the slurries are diluted for counting, some of the techniques may not be effective in detecting the soft agglomerates formed during slurry preparation.Therefore, combination of sizing techniques that can analyze both the dilute and concentrated (25) slurries must be used for effective detection of coarser size particulates.

Stabilization of CMP Slurries
In CMP processes, polishing slurries have to be stabilized under extreme environments of pH, ionic strength, pressure and temperature, in the presence of reactive additives.Most of the commonly used stabilization techniques, such as, electrostatic stabilization, inorganic or polymeric dispersion may not perform adequately under these severe environments.An alternative is to use surfactant structures at the solid-liquid interface for the stabilization of particulate systems (28)(29)(30)(31)(32)(33).
Self-assembled structures of C 12 TAB (dodecyl trimethyl ammonium bromide), a cationic surfactant, have been shown to provide stability to silica suspensions at high ionic strengths and extreme pH by introducing a strong repulsive force barrier (33)(34).This concept was utilized to stabilize the silica CMP slurry in the presence of 0.6M salt at pH 10.5.C 12 TAB surfactant was used at 32mM concentration, which is two times the critical micelle concentration (CMC) in the absence of electrolyte (35).The baseline polishing slurry contained 12wt%, 0.2µm monodispersed silica particles at pH 10.5.Addition of 0.6 M salt coagulated the particles by screening the surface charge (36).At 8mM C 12 TAB, a jump was reported in the repulsive force barrier due to the beginning of the formation of self-assembled surfactant aggregates (34) which led to increased slurry stability.At 32mM C 12 TAB concentration, where a sufficient number of the surfactant aggregates form at the surface, the slurry became completely stable due to strong repulsive force barrier between particles.
Slurries stabilized with C 12 TAB were used for polishing silica wafers.The surface quality response was observed to be optimal with minimal surface rough-ness and deformation.However, material removal rate was only 70 Å/min.This observation highlighted the importance of particle-substrate interactions along with particle-particle interactions for optimal CMP operation.
Two reasons were suggested for the negligible material removal in the presence 32mM C 12 TAB (35).First, it is known that the presence of surfactants can result in lubrication between the abrasive and the surface to be polished (37).This may decrease the frictional forces thus reducing material removal in the presence of surfactants.Alternatively, the high repulsive force barrier induced by the C 12 TAB selfaggregated structures may be preventing the particlesurface engagement and, therefore, resulting in a very low material removal rate.
In order to isolate the effect of lubrication and the repulsive force barrier on material removal response, it was necessary to calculate the force applied on a single abrasive particle during polishing.Initially, this value was estimated to be 100-1000nN for a 0.2µm size particle based on the assumption of hexagonal close packed surface coverage of particles at 2.7҂10 Ҁ3 normalized pad area contact at 7psi head load.Furthermore, the exact value was calculated to be 750Ȁ150nN for a 0.2µm size particle by determining the pad-substrate contact area at the applied head pressure and the particle concentration at the area of contact of the pad (18), which agreed with the modeling values in literature (38).Comparing the load applied per particle to the repulsive force barrier of a C 12 TAB surfactant aggregate (6nN), it was clear that the repulsion introduced by the surfactant structures was overcome under the applied head load.Hence, it was more likely that the lubrication effects introduced by the surfactants controlled the material removal by varying the frictional forces between the abrasive particles and the wafer.Similar studies using C 10 TAB and C 8 TAB with added salt showed nonlinear behavior of lateral forces with respect to normal force (Figure 3).Though material removal rate values were low compared to the baseline, this approach opened a new venue of tailoring the friction forces encountered in CMP.Detailed studies of in-situ friction force response and single particle-substrate friction simulations with AFM for the surfactant mediated slurries have proved that the strength of surfactant adhesion and the chemistry of the slurries (pH, ionic strength and surfactant chain length) could be tailored to provide sufficient frictional force to achieve optimal polishing performance (35).
Palla has showed that in tungsten polishing the addition of a mixture of ionic (SDS) and nonionic (Tween 80) surfactants over CMC concentrations could stabilize alumina particles in the presence of high concentrations of potassium ferricyanide by creating sufficiently high repulsive force barriers (39).The mechanism of stabilization in this system was explained by enhanced adsorption of nonionic surfactant using the strongly adsorbing ionic surfactant as a binding agent.The stabilizing ability of the surfactant system was also found to increase with increasing hydrophobicity of the nonionic surfactant and increasing hydrophobicity of the ionic surfactant.The effect of surfactant concentration on dispersion ability revealed an optimum concentration range for a given surfactant.When tungsten polishing was conducted using the slurries stabilized by the mixed surfactant system, 30% less material removal rate was obtained compared to the baseline slurry.However, a much better surface quality was obtained.Accordingly, surfactants can be used to modify the particle-particle and particle-substrate interactions during CMP to optimize the process performance based on the selected performance criteria.

Conclusions
In CMP, slurry design has significant effect on the process performance and should be carefully monitored for optimal results.Slurry chemistry has to be tailored so that the slurry is free of any hard or soft agglomerates (even in low concentration), which not only would deteriorate the surface quality but would also lead to inconsistent removal rates.The use of surfactants to stabilize the slurry in extreme environments can take advantage of the vast literature available for the solution behavior of surfactants.

Fig. 1
Fig.1Material removal rate and friction force of silica as a function of solids loading of 0.5µm silica abrasives.

Fig. 3
Fig. 3 AFM friction force measurements on silica wafer with 7µm size particle attached to the tip.(a) Solutions containing C 12 TAB, C 10 TAB and C 8 TAB surfactants at 32, 68 and 140mM concentrations without NaCl at pH 10.5 in the presence of 0.6 M salt in the solution.Reprinted from Basim et.al.(35) with permission from Elsevier.