[Retraction]Size measurement and characterization of ceria nanoparticles using asymmetrical flow field-flow fractionation (AsFlFFF)

Abstract

As the size of semiconductors becomes smaller, it is necessary to perform high precision polishing of nanoscale. Ceria, which is generally used as an abrasive, is widely used because of its uniform quality, but its stability is not high because it has a high molecular weight and causes agglomeration and rapid precipitation. Such agglomeration and precipitation causes scratches in the polishing process. Therefore, it is important to accurately analyze the size distribution of ceria particles. In this study, a study was conducted to select dispersants useful for preventing coagulation and sedimentation of ceria. First, a dispersant was synthesized and a ceria slurry was prepared. The defoamer selection experiment was performed in order to remove the air bubbles which may occur in the production of ceria slurry. Dynamic light scattering (DLS) and asymmetrical flow field-flow fractionation (AsFlFFF) were used to determine the size distribution of ceria particles in the slurry. AsFlFFF is a technique for separating nanoparticles based on sequential elution of samples as in chromatography, and is a useful technique for determining the particle size distribution of nanoparticle samples. AsFlFFF was able to confirm the presence of a little quantities of large particles in the vicinity of 300 nm, which DLS can not detect, besides the main distribution in the range of 60-80 nm. AsFlFFF showed better accuracy and precision than DLS for particle size analysis of a little quantities of large particles such as ceria slurry treated in this study.

1. Introduction

Semiconductors are produced on the surface of wafers, and their performance relies on the integration density of the circuitry. The higher the density, the shorter the distance between circuits, which increases the flow rate of currents and improves performance of the product when used in electronic devices. With the recent trend of miniaturization of electronic devices and the consequent miniaturization of semiconductors, the need for high-precision planarization technology has been recognized.^1,2 Among various planarization techniques, the most promising process is chemical mechanical polishing (CMP) that occurs at the area receiving the pressure on the wafer surface being polished and the polishing pad. As the polishing padgets covered with a slurry containing abrasive particles and water moves with respect to the slurry-covered pad, the wafer surface is polished (See Fig. 1).^3,4

 

Fig. 1. Schematic of CMP process.

The slurry is a solution in which small particles are uniformly dispersed. As slurry preparation requires achemical and mechanical element each simultaneously, it gives rise to an industry that necessitates complextechniques. Slurry comprises abrasive particles, dispers ant, and water. For use as abrasive particles, metals exhibiting high levels of durability and hardness are frequently used; these include silica, ceria, and alumina. Among them, ceria seems to be more widely used owing to its high material removal rate. Nonetheless, the large surface area to volume ratio and molecular mass may lead to interparticle cohesion or rapid precipitation in ceria, making itrelatively unstable.⁵ Particle cohesion and precipitation are the two main causes of scratch during polishing.^6-9 It is thus crucial to develop ceria nanoparticles with small particle size, spherical shape, and narrow sizedistribution.

As a general solution to address the problems of particle cohesion and precipitation, dispersants are used.^10-12 A dispersant not only brings about the dispersion of ceria nanoparticles but also helpsmaintain a balance between the three components of slurry (ceria, dispersant, and water), thereby playing an important role in keeping the size distribution consistent and stable. Moreover, the amount of dispers ant added influences the level of slurry dispersion, and in some cases, leads to interparticlecohesion. Thus, it is essential that the proportions of ceria, dispersant, and water are carefully balanced (See Fig. 2).¹³

 

Fig. 2. Three-component system of ceria slurry

For the dispersion of ceria nanoparticles, a dispers ant based on styrene acrylic acid (SAA) or styrenemaleic acid (SMA) was used (see Fig. 3).¹⁴

 

Fig. 3. Chemical structure of dispersant of Styrene acrylate series (a) styrene maleic series (b).

It is generally known that the SMA-based dispers ants are more useful than the SAA-based ones withrespect to the stability of the heat generated during the CMP process.³¹ The foam created during slurry preparation prevents the uniform dispersion of ceriananoparticles, which led us to conduct a series of experiments for stable slurry preparation in this study; we selected the most suitable dispersant forceria slurry preparation by synthesizing dispers ants with different styrene and maleic acid compositions. Foam creation was inhibited by the use of a defoamer during characterization of the prepared slurry and particle size measurement.

The methods commonly used for particle size analysis include scanning electron microscopy (SEM) and dynamic light scattering (DLS). The advantage of using SEM is the direct visualization of particleshape and size. The drawback, however, is the difficulty in acquiring information on particle behavior insolution form due to interparticle cohesion that may occur during drying of the sample to prepare it for SEM. On the other hand, DLS is widely used because it provides the advantage of measuring the sizes of particles dispersed in a solution relatively easily and fast. However, with DLS, it is difficult to obtainaccurate measurements for samples with broad particle size distribution because it measures particlesize using refractive index of the laser based on the Brownian motion of the particles in a confined space.^15,16

Asymmetrical flow field-flow fractionation (AsFlFFF) is a technique useful in isolation of macromolecules, including polymers, as well as nanoparticles, whichensures separation according to the molecular massor particle size.^17-22 In this study, AsFlFFF was used to determine the size distribution of ceria nanoparticles in the slurry to develop a method of particlecharacterization.

 

2. Theory

 

2.1. The theoretical framework for AsFlFFF

AsFlFFF does not have a charged stationary phase in its interior and consists only of an empty channel.²³ It uses a system of separation per particle size based on the flow of the fluid injected to the space between two planes formed upon the insertion of a spacer. Here, a difference in the interaction of samplecomponents is created between the channel flow and the crossflow, which gives rise to the differences in the channel retention time depending on particle size (or molecular mass), and thus, leads the particles to be separated.²⁴

One of the useful features of AsFIFFF is that the sample retention time (t_r) could be used in direct calculation of the hydrodynamic diameter of particles (d_H) using Eq. (1) below.^25-28

\(d_{H}=\frac{2 k T V^{0}}{\pi \eta w^{2} V_{c} t^{0}} t_{r}\)       (1)

Here, k is Boltzmann constant, T is absolutetemperature (K), V⁰ is the volume inside the channel (void volume), η is the viscosity of mobile phase, w is channel thickness, V_c is crossflow strength (i.e.cross flow rate) and t⁰ is the time of passage by a component without retention or a solvent through the channel volume (void time). In Eq. (1), all variablesexcept t_r become the constants once the experimental conditions have been determined so that, by measuring t< sub>r of a given sample using AsFIFFF, the size and sizedistribution of the sample can be directly estimated.^29-31

 

3. Experimentation

 

3.1. Materials

In this study, for ceria, the abrasive particle, ceriumoxide (CeO₂) powder was obtained from the calcination of cerium carbonate (Ce₂(CO₃)₃) powder at 600-1000 oC. For dispersant synthesis, D/I water (Milli QPLUS), SMA-1000 (Styrene maleic anhydride 1000, SARTOMER Co. Ltd., Pennsylvania, USA), SMA-2000 (Styrene maleic anhydride 2000, SARTOMERCo.. Ltd., Pennsylvania, USA), and SMA-3000(Styrene maleic anhydride 3000, SARTOMER Co. Ltd., Pennsylvania, USA) were used, followed by use of ammonia water (25-28 %, Duksan, KOREA) for acidification. Polypropylene glycol triol was used as a defoamer, along with the following : Depol(MW = 280.0), BYK (MW = 510.0), and G-336 (MW= 620.0). As the carrier liquid in AsFlFFF, a solution containing 0.1 % FL-70 (Fisher Chemical, New Jersey, USA) and 0.01 % sodium azide (NaN₃, Sigma-Aldrich, St. Louis, USA) was used.

 

3.2. Measuring devices

For configurational and elemental analyses of Ce₂(CO₃)< sub>3 and CeO₂, the SEM-EDS (JEOL-7800F, JEOL Ltd., Japan) was used.

For the preparation of dispersant and slurry, adigital overhead stirrer (HT-50DX, DAIHAN, Korea) was used, while a heating mantle (WHM12214, DAIHAN, Korea) was used for heating upon stirring.

To measure the molecular mass of the dispers antand the defoamer, size exclusion chromatography (SEC) was used. For SEC, two types of columns: SB-804-HQ and SB-805-HQ (Shodex, USA), were used. For the SEC mobile phase, the HPLC pump (P-6000, FUTECS Co., Ltd, Korea) was operated ata rate of 1 mL/min. For sample detection, the RIdetector (RID-10A, SHIMADZU, Japan) was used.

For the characterization and stability evaluation of the slurry prepared in this study, a pH meter (OHAUSCorp., OHAUS STARTER 2100, USA), a conductivity meter (OHAUS Corp., OHAUS STARTER 3100C, USA), and a viscosity meter (Brookfield, LVDVE230, USA) were used.

To examine the shape and size of ceria, transmissionelectron microscopy (TEM, JEM-F200, JEOL Ltd., Japan) and field emission scanning electron microscopy (FE-SEM, JEOL-7800F, JEOL Ltd., Japan) were used.

To measure particle sizes and size distribution, DLS(ELSZ-2000, Otsuka, Japan) and AsFlFFF were used.

For the AsFlFFF channel, a Wyatt short channel (Wyatt Tech., Europe GmbH, Dernbach, Germany) was used, in addition to a cellulose membrane (Millipore, Bedford, USA) with a 10 kDa cut-offmolecular mass and a 250-µm Mylar spacer, was used.

A solution composed of 0.1 % FL-70 and 0.02 % NaN3 was used as carrier liquid for AsFlFFF. The carrier liquid was introduced using the HPLC pump (P-6000, FUTECS Co., Ltd, Korea). An Optiflow 1000 Liquid Flowmeter (Agilent Technologies, Palo Alto, CA, USA) was used to measure the flow rate.

A UV detector (Spectra UV 150, Thermo Separation Products, USA) was used to detect the samples that were eluted after separation by particle size in AsFlFFF.

The samples were injected using a syringe pump(Legato 110, KD Scientific Inc., Mendon, USA) with a volume of 50 µL and flow rate of 0.2 mL/min. All analyses in this study were repeated three timesto ensure reproducibility.

 

3.3. Dispersant synthesis

For the dispersion of ceria nanoparticles, three types of dispersants were synthesized. Firstly, 42 g of SMA powder was acquired in a 3-neck flask and mixed in 42 g of DI water with stirring.

Using a heating mantel, the temperature of the solution was raised to 80 oC. When the temperaturereached 80 oC, 24 g of ammonia-containing water was added and then stirred for 2 h (See Scheme 1).

Three types of dispersants were prepared by varying the type of SMA powder applied.

Fig. 4. presents the structure of the three types of dispers ants thus (‘SMA 1000’, ‘SMA 2000’, and ‘ SMA 3000’) synthesized in this study.

 

Scheme 1. SMA dispersant synthesis.

 

Fig. 4. The structure of SMA dispersant synthesized in this study, (a) SMA1000, (b) SMA 2000, (c) SMA 3000.

 

3.4. Slurry preparation

For slurry preparation, a vertical centrifugal beadmill (UAM-015, zirconia beads mill, Ultra Apex Mill, Japan) was used. A mixture containing 900 g of ceria powder, 135 g of dispersant, and 1,965 g of distilled water, was applied to UAM-015 for slurry preparation. We used 0.2-mm beads, where the filling factor for the beads was set to 80 %. The flow rate of the pump was 500 mL/min, while the milling was carried out at 1,500 rpm.

 

4. Results and Discussion

 

4.1. Characterization of the SMA dispersant

Table 1 shows the pH and molecular mass of the three types of dispersants synthesized in this study.

The molecular mass was measured at a flow rate of 1 mL/min using size exclusion chromatography. The SMA 1000 dispersant, in comparison to SMA2000 and SMA 3000, displayed a decrease in pH as the number of carboxyl groups increased and stronger hydrogen bonding led to an increasing trend of molecular mass.

 

Table 1. pH and molecular weight of synthesized dispersant

 

4.2. Cerium oxide preparation

For fabrication of the pigment used in slurry preparation, cerium oxide (CeO₂) was used that was obtained from the calcination of cerium carbonate (Ce₂(CO₃)< sub>3) at 600-1000 oC.

Fig. 5 shows the 100 × magnified SEM image of Ce₂(CO₃)< sub>3 and CeO₂. A flower petal-shaped particle of Ce2(CO3)3 can be seen, while CeO₂ particle size was observed to decrease with the progression of calcination.

Fig. 6 presents the SEM-EDS (energy dispersivespectrometry) spectrum of Ce₂(CO₃)₃ and CeO₂; theirrespective EDS analysis results are given in Table 2. The EDS analysis results confirmed a decrease incarbon and oxygen contents and an increase incerium content following calcination.

 

Fig. 5. SEM image of cerium carbonate and cerium oxide (X 100), (a) Cerium carbonate, (b) Cerium oxide.

 

Fig. 6. EDS results of cerium carbonate and cerium oxide, (a) Cerium carbonate, (b) Cerium oxide.

 

Table 2. EDS results of cerium carbonate and cerium oxide.

 

4.3. Dispersant selection

Based on the findings so far, an experiment was conducted to select a dispersant with a good affinity for ceria nanoparticles. The slurry was prepared using an equal amount of dispersant, but with different types of dispersant.

Fig. 7 presents the size distribution of ceriananoparticles in each slurry, containing different types of dispersant, measured using DLS. When SMA 1000 was used as the dispersant, an acceptabledispersion state without interparticle cohesion was achieved. On the contrary, when SMA 3000 was used as the dispersant, interparticle cohesion occurred gradually. The results indicated that the affinity of ceria nanoparticles was stronger for SMA 1000 than for SMA 3000. The results in Fig. 7 verified SMA1000 to be the more stable dispersant for ceria slurry preparation.

 

Fig. 7. DLS size distribution of ceria particles in slurry by dispersant, (a) SMA 1000, (b) SMA 3000.

 

4.4. Dispersant concentration optimization

An important factor in dispersion is the dispers ant concentration. If the concentration is too low, incomplete dispersion may result, and if it is toohigh, slurry cohesion may be caused by bridgingreactions with the dispersant. Hence, optimizing the dispers ant concentration is crucial for acquiring the desired level of dispersion. In this study, an experimenton dispersion stability was conducted to optimize the SMA 1000 dispersant concentration.

Fig. 8 depicts the changes in pH (a), conductivity (b), viscosity (c), and hydrodynamic diameter (d) of ceria nanoparticles measured at 60 oC for 35 days following the addition of varying concentrations of SMA 1000: 4 % (black), 4.5 % (red), and 5 % (green). The hydrodynamic diameter (d_H) was recorded using DLS. Table 3 presents the measured pH, conductivity, viscosity, and hydrodynamic diameter after 35 days as mean ± standard deviation.

Storage stability measured over 35 days at 60 oC showed a subtle, but gradual, increase in pH, conductivity, viscosity, and hydrodynamic diameteras the concentration of SMA 1000 increased from 4 % to 4.5 % and then to 5 %. This change, however, did not exert a significant influence on the slurrystability. As Table 3 shows, the smallest standard deviation for each measured value could be obtained at SMA 1000 concentration of 5 %. Hence, 5 % concentration was selected for SMA1000 in this study.

 

Fig. 8. Stability over time measured at 60 o C for 35 days after addition of SMA 1000 dispersant, (a) pH, (b) conductivity, (c) viscosity, (d) dH from DLS.

 

Table 3. Stability over time according to SMA 1000 concentration

 

4.5. Defoamer selection

The problem of foam formation occurred during slurry preparation (See Fig. 9).

When foam is formed, the dispersion of slurry particles is reduced that prevents uniform polishing. Thus, to eliminate foam formation, an additionalexperiment was conducted for selection of the most suitable defoamer.

For the defoamer, compounds based on polypropyleneglycol triol were used, which exhibited the following molecular mass: 280 kDa for depol, 510 kDa for BYK, and 620 kDa for G-336. Fig. 9 shows animage of slurry preparation after the defoamer had been added. Results showed that the addition of defoamer reduced the foam formation to a negligiblelevel.

 

Fig. 9. Slurry before and after addition of defoamer, (a) Before addition of defoamer, (b) After addition of defoamer.

 

4.6. The impact of defoamer concentration on slurry stability

A defoamer was added during the process of slurry preparation so that foam may not form. The concentration of the added defoamer should be within a range where slurry stability remains intact. An excess amount of defoamer has the potential todestroy slurry stability, and thus, it is added generally inconcentrations below 1 %. In this study, concentrations of the three types of defoamer were set to 0.01 %, 0.025 %, and 0.05 % for evaluating the storagestability at 60 oC for 10 days.

Fig. 10 depicts the changes in hydrodynamic diameter(d_H) for each slurry prepared with an addition of depol, BYK, or G-336 as the defoamer; these were measured using DLS at 60 oC for 10 days. Table 4-6 shows the pH, conductivity, viscosity, and hydrodynamic diameteras mean ± standard deviation measured over 10 days after the addition of each type of defoamer.

For all three types of defoamer, an increase indefoamer concentration was shown to have subtly increased the conductivity and particle size. Based on the results presented in Fig. 10 and Table 4-6, this study selected 0.025 % as the most suitable defoamer concentration.

 

Fig. 10. DLS results for stability according to defoamer concentration. (measurement at 60 o C for 10 days), (a) Depol, (b) BYK, (c) G-336.

 

Table 4. Stability of slurry according to concentration of Depol defoamer

 

Table 5. Stability of slurry according to concentration of BYK defoamer.

 

Table 6. Stability of slurry according to concentration of G-336 defoamer

 

4.7. Slurry preparation

Ceria slurry was prepared with a 5 % concentration of the SMA 1000 dispersant. In addition, 0.025 % defoamer was added to eliminate foam formationduring preparation. In general, the slurry used forpolishing should have small particle size and narrowsize distribution to ensure uniform polishing. Thus, itis necessary to accurately measure particle sizedistribution for the slurry. In this study, the slurry was obtained using a ball mill. The particle size analysis was then carried out using DLS and AsFlFFF.

Fig. 11 depicts the particle size distribution foreach defoamer as measured using DLS for all fourtypes of ceria slurry separately (intensity (a), volume (b), and particle size distribution based on the number (c)). A control slurry with no defoamer was also analyzed for comparison. Table 7 presents the mean particle size in nanometers (nm) as obtained from each size distribution shown in Fig. 11.

The results of DLS presented in Fig. 11 and Table 7 show that all slurry samples exhibited relatively wide particle size distribution with little differences in mean particle size among the samples.

 

Fig. 11. Particle size distribution of ceria slurry by DLS. Size distribution based on (a) intensity, (b) volume, (c) number.

 

Table 7. Particle size distribution of ceria slurry by DLS.

For identical samples, AsFlFFF was used to analyzethe particle size distribution. Upon the As FIFFF analysis, channel flow (V_out) was recorded to be 0.6mL/min with a crossflow (V_c) of 0.5 mL/min. Fig. 12 shows the AsFlFFF fractogram (left) and particlesize distribution (right) for each sample; Table 8 presents the mean particle size obtained from the result in Fig. 12.

Fig. 12 and Table 8 show that all four samplesexhibited the particle size distribution that was in therange of approximately 60-80 nm. Notably, a fewlarge particles that were approximately 300 nm indiameter were found, in addition to the main distribution of 60-80 nm range, in the control and samples to which depol had been added. As can be seen in Fig. 11 and Table 7, DLS did not detect the presence of these large particles. As previously mentioned, incontrast to DLS, which is a non-separation method, AsFIFFF allows separation according to particlesize. As a separation method, AsFIFFF seems to have advantages over DLS in terms of detecting the presence of a few large or small particles that do not belong to the main distribution.

 

Fig. 12. AF4 fractograms of ceria slurry according to defoamer. fractograms (a), size distribution (b). The AF4 channel flow rate was 0.6 mL/min and the cross flow rates was 0.5 mL/min, and the carrier liquid was water containing 0.1 % FL70 and 0.01 % NaN3.

 

Table 8. AF4 particle size analysis of ceria slurry

Following the addition of a defoamer, SEM analysis was carried out to examine the shape of ceriananoparticles in the slurry. Fig. 13 presents the identical FE-SEM images of four sample types.

Fig. 13 shows that the shape of ceria nanoparticles was not significantly altered before and after the addition of defoamer. The BYK and G-336 defoamers, in particular, were shown to have preserved the original shape of ceria nanoparticles without exerting a large influence on dispersion, such as cohesionamong ceria nanoparticles.

The results collectively suggested that as the defoamer was added to eliminate foam formationduring ceria slurry preparation, BYK or G-336 weremore stable than was depol.

 

Fig. 13. SEM image of ceria particles according to defoamer. (X 100,000) (a) Base, (b) Depol, (c) BYK, (d) G-336.

 

5. Conclusions

This study investigated the characterization of ceria nanoparticles. Ceria nanoparticles are used in polishing, where uniform polishing can be achieved only when ceria particles exhibit small size and anarrow size distribution. Even a small number oflarge particles may cause a scratch during polishing. Hence, the determination of accurate particle sizedistribution is critical. The foam created in slurry preparation interferes with the dispersion of ceriananoparticles, therefore, the selection of a suitable defoamer is very important to obtain uniform particlesize distribution.

Conventional particle size analysis generally uses SEM or DLS. While SEM allows the direct visualization of particle shape, the fact that samples have to be dried for measurement makes it difficult to examine the particle state in solution form. On the other hand, DLS, allows the determination of particle size distribution in a solution, but as it is not a method of separation, a small number of largeparticles cannot be selectively detected. Unlike conventional methods, AsFlFFF measures particlesize distribution based on particle separation, whichenables the detection of even a small number of largeparticles that may not be expected with DLS. Notably, AsFlFFF was shown to detect a small number oflarge particles in the main distribution of particles within the range of 60-80 nm, as well as the range above ~300 nm that cannot be detected using DLS. Thus, for particle size analysis in samples containing a small number of large particles, such as the ceriaslurry samples investigated in this study, AsFlFFFprovides higher levels of accuracy and precision thandoes DLS.