Electrocoagulation for removal of silica nano-particles from chemical-mechanical-planarization wastewater

(1)

Electrocoagulation for removal of silica nano-particles from

chemical–mechanical-planarization wastewater

Walter Den

a,∗

, Chihpin Huang

b,1

a_{Department of Environmental Science, Tunghai University, Taichung-Kan Road, Sec. 3, #181, P.O. Box 818, 407 Taichung, Taiwan, Republic of China} b_{Institute of Environmental Engineering, National Chiao Tung University, Po-Ai St., #75300 Hsinchu, Taiwan, Republic of China}

Received 16 April 2004; accepted 26 November 2004 Available online 30 December 2004

Abstract

Continuous-flow electrocoagulation process with vertical flow-channels was investigated as a method to treat synthetic chemical–mechanical-planarization (CMP) wastewater containing highly charged ultrafine silica particles (ζ = −55 mV, mean Rp= 45 nm at pH 9.5). The parallel-plate, monopolar electrochemical cells resembled a series of closed electrical circuits such that the electrical field strength was highly dependent of the current density and aqueous conductivity, but independent of the inter-electrode gap. The residual tur-bidity of the CMP wastewater decreased with the increases in either hydraulic retention time or applied current density, and removal efficiency as high as 95% was achieved for wastewater with both low (70 NTU) and high (400 NTU) initial turbidities. The charge loading linearly correlated with turbidity removal efficiency up to a level of 8 F m−3, presenting an appropriate design parameter. Further analysis indicated that turbidity removal was limited by the quantity of liberated ferrous ions at lower range of current density, but seemingly reached a critical level of current density beyond which the process performance gradually deteriorated. Comparisons between the effective particle retention time and the estimated electrophoretic migration time revealed that the electrocoagulation process was predominantly controlled by the rate of particle aggregation occurring near the anodic surfaces. Furthermore, this process generates lesser amount of dry sludge as compared to chemical coagulation with polyaluminum chloride, and does not require pH adjustment prior to treatment.

Keywords: Electrocoagulation; Electrochemical; Parallel plate; Monopolar; Silica colloids; Chemical–mechanical-planarization; CMP wastewater treatment

1. Introduction

Chemical–mechanical-planarization (CMP) is one of the fastest growing processes in semiconductor manufacturing, and has become an integral part of the state-of-the-art fab-rication line for the multi-level design of integrated circuit (IC). The process is primarily used for “polishing” the de-vice side of a semiconductor wafer through the mechanical downforce of slurry abrasive in combination with the chemi-cal oxidation of wafer surface. Due to its capability to achieve

∗_{Corresponding author. Tel.: +886 4 23590121x3050;} fax: +886 23594276.

E-mail addresses: wden@mail.thu.edu.tw (W. Den),

cphuang@mail.nctu.edu.tw (C. Huang). 1 _{Tel.: +886 3 5726463; fax: +886 5725958.}

uniform surface topography with high throughput, the use of CMP technology has expanded from the traditional applica-tion in metal interconnects (i.e., metal CMP) to more com-plicated applications in dielectric barrier layer planarization (i.e., oxide CMP). With the continuing pursuit of functionally stronger devices with smaller feature sizes (sub 0.25␮m), the device fabrication will undoubtedly rely more and more on

the CMP processes[1]. As a consequence, the quantity of

CMP wastewater generated is expected to increase propor-tionally with the growing needs of the CMP processes[2]. Furthermore, as the International Technology Roadmap for Semiconductors[3]specifies, the objective of reducing total water consumption by roughly five-fold for fabrication pro-cesses must be met by 2005, with more than 80% water recy-cling capability for the new plants. This industry-wide effort of water conservation and recycling translates to higher solid

(2)

82 W. Den, C. Huang / Colloids and Surfaces A: Physicochem. Eng. Aspects 254 (2005) 81–89

Nomenclature

Ap effective area of anode (m2)

E electrical field strength (V m−1)

Eeq equilibrium potential difference between

elec-trodes (V)

F Faraday constant (96,500 C)

i applied current (A)

j current density (A m−2)

lp inter-electrode distance (m)

N number of anodic plates

q particle charge (C)

Rp mean particle radius (m)

T0, T initial and variable wastewater turbidity (NTU)

u electrophoretic migration velocity (m s−1)

Vr working volume of wastewater in reactor (m3)

Greek letter

ε dielectric constant

η over-potential (V)

κ aqueous conductivity (mho m−1)

µ dynamic viscosity of water (kg m−1s−1)

τ hydraulic retention time (min)

ζ zeta potential (V)

concentration in CMP wastewater, which must be properly treated to meet the local discharge regulations.

CMP effluents typically contain suspended solids origi-nated from slurry abrasive particles of SiO2, Al2O3, or CeO2,

depending on the nature of the CMP application. Other con-taminants, including insoluble metal oxides and nitrides and soluble chemicals, also exist in the wastewater in a much lesser quantity. Golden et al.[4,5]have described the chem-istry of CMP wastewater in detail. In principal, CMP wastew-ater contains very dilute slurry abrasives that are narrowly ranged between 50 and 200 nm, and possess highly nega-tive surface charges that repel adjacent particles when they are immersed in base solutions. These wastewater charac-teristics render conventional coagulation–flocculation tech-nology less ideal to remove such nano-scale particles from the CMP wastewater. A common problem experienced by many system operators is the need to increase the coagu-lant dosage, leading to bulky sludge volume. Additionally, the poor settling characteristics of the flocs often prolong the settling time to achieve the necessary separation efficiency. To overcome these problems, membrane separation process has been adapted as a single unit or as a post separator for co-agulation/flocculation process[6–9]. The use of membrane filtration has demonstrated success in the removal of sus-pended solids, and owns the added value of re-concentrate the valuable slurry for possible reuse. However, maintain-ing the necessary degree of permeate flux is still an issue of concern.

As an alternative treatment method, electrochemical pro-cesses have been proposed and investigated to treat wastewa-ter containing dispersed ultra-fine particles without chemical additions[10–14]. Belongia et al.[10]first demonstrated that, when passing an electrical current through CMP suspensions, two distinct solid separation phenomena, namely electrode-cantation and electrocoagulation, occurred. The occurrence of these phenomena strongly depends on the type of slurry and the conductivity of the suspension. For electrocoagula-tion, the sacrificial anodic electrodes of iron and aluminum are commonly used to continuously supply metallic ions as the source of coagulants. These electrochemically generated metallic ions can hydrolyze near the anode to form a series of metallic hydroxides capable of destabilizing dispersed parti-cles. The simultaneous electrophoretic migration of the neg-atively charged particles (e.g., silica particles) toward the an-ode forces chemical coagulation between particles and metal-lic hydroxides in the vicinity of the anode, forming flocs that either settles or re-deposits onto the anode (Fig. 1). Matteson

et al.[13]developed a mathematical model based on

elec-trophoresis and charge neutralization to simulate the kinetics of electrocoagulation process for ultra-fine kaolin particles. They found that a second-order reaction kinetics could be used to generically describe the particle aggregation and re-moval processes. Den et al.[14]have also demonstrated ef-ficient particle separation by electrocoagulation of dispersed silica particles in a continuous-flow mode, although the roles of particle transport and reaction were not clarified. There-fore, the primary objective of the present work was to investi-gate the effects of major parameters, including transport and electrochemical parameters, in the design and operation of the continuous-flow electrocoagulation reactor for treatment of oxide CMP wastewater consisting primarily of nano-size silica.

2. Materials and methods

2.1. CMP wastewater preparation

Silica suspensions were prepared by diluting a commer-cially available, KOH-based CMP slurry containing 25 wt.% fume silica (Semi-Sperse®25 slurry, Cabot Microelectron-ics Corp., USA). The original slurry was diluted to 0.05% and 0.4% with deionized water, yielding suspension turbid-ity of approximately 70 and 400 NTU, respectively. These slurry dilution levels correspond to particle number concen-trations of 4.55× 1017and 3.65× 1018m−3for a nominal sil-ica density of 2.1 g cm−3and particle size of 0.1␮m.Table 1

presents the physical and chemical properties of several oxide CMP wastewater from different chip manufacturing facil-ities, as well as those of the synthetic wastewater used in this study. The oxide CMP wastewater primarily consists of dilute silica slurry with turbidity ranging between 100 and 350 NTU, pH between 8 and 9.5, conductivity between

(3)

Fig. 1. (a) Experimental apparatus of continuous-flow electrocoagulation system. The magnified portion (b) schematically depicts the electrocoagulation mechanisms for the removal of colloidal particles, with the image (c) of the anode surface covered with gelatinous film.

−80 mV. In this study, the silica CMP suspension was

main-tained in a weak alkaline condition (pH∼9.5) by 0.01 M KOH for the purpose of maintaining particle dispersion and soft-ening the surface of silica particle by hydrolysis of Si O Si bonding:

Si O Si + H2O ↔ Si OH (1)

Si OH + OH ↔ Si O−+OH : H (2)

The silica particles submerged in base solutions tend

to adsorb OH− on particle surface and form Si O− on

the surface extension. Therefore, increasing the OH−

con-centration leads to greater degree of negative surface potential.

The particle size and distribution were measured by a particle analyzer (Honeywell UPA-150, Microtrac Inc., Montgomeryville, PA) using dynamic light scattering with

an effective range between 0.001 and 6␮m. The

elec-trophoretic mobility of the particles was measured by a zeta potential analyzer (ZetaPlus, Brookhaven Instruments Corp., Holtsville, NY) using Doppler shift analysis with laser light scattering. All samples were measured in duplicate to ensure data reproducibility, and an additional measurement was performed if necessary.

Table 1

Wastewater properties of the actual and synthetic oxide CMP effluent Sample source Mean particle

size (nm)

Discharge

ζ-potential (mV) Discharge pH Turbidity(NTU)

Total organic carbon (mg L−1) Discharge conduc-tivity (␮S cm−1) Source 1 165 −63.4 8–9 200–350 35–50 100 Source 2 170 −47.6 8.7–9.4 100–140 1.5–3.0 104–138 Source 3 >100a _−78.23 _8.57 ₁₃₅ _15.16 _127.2 Synthetic wastewater (this study) 88 −55 9.3–9.7 70–400 0.5–10 55–150

Source 1: actual oxide CMP wastewater from a DRAM fabrication (Hsinchu, Taiwan). Source 2: actual oxide CMP wastewater from a DRAM fabrication (Hsinchu, Taiwan)[7]. Source 3: wafer fabrication from Southern Taiwan[9].

(4)

84 W. Den, C. Huang / Colloids and Surfaces A: Physicochem. Eng. Aspects 254 (2005) 81–89 2.2. Monopolar electrocoagulation experiments

An 8-L continuous-flow reactor channelized by the

elec-trode plates (20 cm× 14 cm) was designed and fabricated

with clear acrylic to perform the experiments. As shown in

Fig. 1a, alternate electrodes were connected to the opposite terminal power supply to form a monopolar configuration, giving a number of reactor cells electrically in parallel with one another. The anodes (iron plates) were completely sub-merged, and the cathodes (stainless steel plates) were par-tially submerged in the suspension. The channel width can be varied by adding or removing an equal number of electrode plates. The outlet port was located rather close to the bottom to prevent floating flocs from exiting the reactor. Electrical current was provided by a manually controllable dc power supply (Model GPS, Goodwill Instruments, Taiwan) oper-ating in constant-current mode (range: 0–3 A). The voltage response (range: 0–200 V) was monitored and recorded by a data-acquisition system.

At the commencement of each experiment, the reactor was filled with CMP suspension such that the liquid level was approximately 1 cm above the anodic plates and below the top of the cathodic plates. The synthetic wastewater was continuously pumped from a 100-L reservoir into the reactor using a microprocessor-controlled peristaltic pump (Master-flex L/S, Cole Parmer, IL) at various flow rates between 80 and 380 mL min−1, corresponding to mean hydraulic reten-tion times between 20 and 100 min. Hence, an overall charge loading can be calculated using the following expression: charge loading (CF)= iτ

FVr (3)

where i is the applied current (ampere),τ the mean hydraulic retention time (s), F the Faraday constant (96,500 C), and Vr

the working volume (m3) of the wastewater in the tank. Dur-ing operation, a ball valve installed at the effluent end allowed control of steady liquid level in the reactor. Samples from the influent and effluent ports were taken at a 5-min interval and measured for turbidity, conductivity, and pH by a Hach ratioTM/XR turbidity meter (Hach Co., Loveland, CO). The organic content in the suspensions was measured by a non-dispersive infrared TOC analyzer (Shimadzu TOC-5000A, Japan) after samples were filtered with 0.45-␮m membrane,

acidified, and N2 purged to remove inorganic carbon. At

the end of each experimental run, the total residual iron concentration was analyzed by inductively-coupled plasma-atomic emission spectroscopy (ICP-AES, Jobin-Yzon JY24, France).

2.3. Chemical coagulation experiments

Jar tests (PB-700 Jar tester, Phipps & Bird Inc., USA) were also performed on the synthetic CMP wastewater as a comparative basis between electro and chemical coagu-lation processes. Polyaluminum chloride (PACl) was cho-sen as the chemical coagulant. The stock PACl solution of

0.1 wt.% was freshly prepared by dissolving the reagent-grade PACl (Showa Chemicals Inc., Japan) in the deionized water. The alkalinity was adjusted to 100 mg L−1(as CaCO3)

by adding NaHCO3. The mixing conditions were provided

with 1-min rapid mixing (200 rpm, G = 350 s−1), followed by 20-min slow mixing (30 rpm, G = 25 s−1). Sedimentation was allowed to occur for 40 min, and the residual turbidity of the supernatant was measured. These coagulation and mix-ing conditions were similar to those used to successfully treat clay suspensions as described elsewhere[15]. All coagulant dosages reported in this paper were in the unit of mg L−1as Al.

3. Results and discussion

3.1. Characterization of synthetic CMP wastewater Fig. 2a shows the surface potential (i.e., ζ-potential) of the diluted silica slurry and the corresponding particle size as a function of solution pH. The isoelectric point (IEP) for the suspension was 2.3, resulting in a mean particle size of approximately 130 nm. Under the weak alkaline condi-tion (pH between 9 and 10) into which the wastewater dis-charged from the oxide CMP process typically falls, the dispersed particle formed a logarithmically normal distri-bution with a geometric mean particle size of 90 nm and a standard deviation of 1.42. Furthermore, the turbidity of the silica suspension exhibited a linear relationship with the slurry dilution levels for slurry concentration less than 1 wt.%, as shown in Fig. 2b. The typical range of turbid-ity of CMP wastewater was 50–500 NTU, which is equiv-alent to a dilution level of 50–200 times from the original slurry. The conductivity of the solution also increased lin-early with the increasing slurry concentration (or turbidity) due to the presence of oxidizers or additives (e.g., KOH, NH4OH) in the raw slurry. The range of these values is

con-formed to those reported for the actual oxide CMP wastew-ater (seeTable 1), and thus justified the appropriateness of using the synthetic CMP wastewater as the model wastewa-ter. The solution conductivity was not altered by any external ionic addition throughout the study of the electrocoagulation process.

3.2. Reactor analysis

The monopolar arrangement of anodes and cathodes con-nected in parallel to the dc power supply created multiple cells, each could be viewed as a closed electrical circuit hav-ing identical electrical potential, provided that the wastewater conductivity was sufficiently low to prevent “leaks” from the circuit. Theoretically, The required voltage for each electro-chemical cell should include the equilibrium potential differ-ence between the anode and cathode (Eeq), the over-potentials

(5)

Fig. 2. (a) Variation ofζ-potential and mean particle size with wastewater pH and (b) aqueous turbidity and conductivity corresponding to the slurry dilutions (concentration) for preparation of the synthetic CMP wastewater.

kinetics, as well as the ohmic potential drop of the solution:

Up= Eeq+ |η| + j _lp κ (4) where lp(m) is the inter-electrode distance,κ (mho m−1) the

solution conductivity, and j (A m−2) represents the current density uniformly distributed over the active surface of each anode in the monopolar configuration.

Fig. 3shows the measured cell potential as a function of the applied current density for wastewater with a range of

aqueous conductivity between 60 and 250␮S cm−1, along

with the theoretical curves calculated from the Ohmic po-tential drop. It can be seen that the Ohmic popo-tential curves substantially underestimated the actual cell voltages, espe-cially at lowκ and large j, indicating that over-potentials were

significant under the operating conditions. In principal, the small differences between the Ohmic and measured voltage at very low j were mainly due to Eeq, and the widening gaps at

higher j were mainly attributed to the j-driven over-potential as depicted by the well-known Tafel equation. Furthermore, in the electrocoagulation process with iron anodes, the pre-cipitates of iron oxides or hydroxides were primarily gelati-nous and tended to collect on the anodic surface (Fig. 1c), producing a gelatinous surface film that could also increase the anodic over-potential by passivation, particularly at large

j that intensifies the surface passivation. Mass transport

re-sistance due to concentration polarization near the anodes could also contribute to the over-potential since the system was operated with primarily laminar-flow regimes with little mixing.

(6)

Fig. 3. Dependence of electrolysis voltage on the wastewater con-ductivity and applied current density (pH 9.5, temperature = 26–28◦C, flowrate = 130 mL min−1). The symbols with dash-lines indicate the mea-sured cell potential, whereas the solid lines represent the Ohmic potential drops.

3.3. Effect of charge loading for electrocoagulation

In the electrocoagulation process with iron as the anodes, the following cell reactions are expected:

Anode : Fe 2e−→ Fe2+ (5)

Cathode : 2H2O + 2e−→ H2(g)+ 2OH− (6)

Overall cell : Fe + 2H2O → Fe2++ H2(g)+ 2OH− (7)

Theoretically, each Faraday of charge passing through a cell circuit produces 27.9 g of Fe2+in the cell, and thus the re-actor charge loading defined in Eq.(3)should be proportional to the quantity of ferrous ions available for electrocoagula-tion.Fig. 4presents a consolidated plot showing the effect of charge loading on the steady-state turbidity removal effi-ciency. It can be observed that the resulting turbidity removal efficiencies clustered into a single characteristic profile,

al-Fig. 4. Dependence of turbidity removal efficiencies on charge loading for wastewaters with low turbidity ((): T0= 70 NTU) and high turbidity (():

T0= 400 NTU).

though the removal efficiency for the wastewater with low initial turbidity (70 NTU) was somewhat higher than those with higher initial turbidity (400 NTU) under similar charge loadings. The discrepancy, however, could be attributed both to the difference in the initial turbidity as well as to the dif-ference in the wastewater conductivity. As mentioned ear-lier, higher suspension concentration (i.e., higher turbidity) resulted in higher aqueous conductivity, which in turn led to lower electrical field strength under the same applied current. This sequence, therefore, may potentially lower the turbidity removal efficiency. More importantly, the profile exhibited in

Fig. 4indicates that the turbidity removal efficiency increased sharply with charge loadings up to a level of approximately 8 F m−3, then leveled off with larger charge loadings. Appar-ently, under conditions of low charge loading, the released ferrous ions were not sufficient to destabilize the silica col-loids in the wastewater, limiting the removal efficiency of turbidity. Under larger values of charge loading (>8 F m−3), the slowing of turbidity removal efficiency was likely con-trolled by the rate of particle aggregation, which is known to be a much slower process as compared to the colloidal desta-bilization process. From this perspective, it would also be possible to experience reduced turbidity removal efficiency under larger charge loadings, because excessive dosage of Fe2+may cause restabilization of colloids by fully surround-ing the silica and turnsurround-ing them into positively charged parti-cles.

3.4. Effects of applied current density

Since charge loading is a surrogate parameter that com-bines contributions from the applied current and the hydraulic retention time, it would also be of interest to examine their effects on the turbidity removal as individual parameters. For this purpose,Fig. 5presents the turbidity removal efficiency as a function of applied current density for four different hy-draulic retention times. Each value of the residual turbidity was taken after the electrocoagulation process had reached

Fig. 5. Effect of applied current density on the steady-state turbidity removal efficiency at different hydraulic retention times (T0= 70 NTU, lp= 5 cm).

(7)

an apparent steady-state condition, normally after 60 min of operation. As can be seen from the figure, the residual tur-bidity in the effluent invariably decreased with an increase in either applied current density or retention time due to the greater quantity of iron dissolution for coagulation, as well as the stronger electrical field that accelerated particle mi-gration toward the anodes. In addition, the turbidity removal was not satisfactory (<70%) at the applied current density less than 8 A m−2, unless long retention times (e.g.,τ = 100 min) were allowed. More interestingly, the removal efficiencies increased almost linearly with the applied current density at the retention times of 20 and 30 min. When the retention time was prolonged to 60 and 100 min, however, the corresponding removal efficiencies tended to level off, or even decreased, at larger applied current densities. Incidentally, the quantity of residual iron concentration in the effluents also demon-strated a strong dependency on the current density, as shown inFig. 6. At the retention time of 30 min, the residual iron concentration fluctuated between 1 and 2 mg L−1, indicating that the liberated ions were effectively consumed. This result, coupled with the linear increase of the turbidity removal effi-ciency with current density, suggests that the process was still limited by the dosage of ferrous ion available for coagulation. For the retention times of 60 and 100 min, the residual iron concentrations were substantially less than those for the re-tention time of 30 min, which was consistent with the results of better removal efficiency under the same current densities for longer retention times. However, the residual iron con-centration sharply increased (>7 mg L−1) when the current density was raised to 10.9 A m−2, then further to 14.5 A m−2. Also, the fact that the profiles of residual iron concentrations were very similar for the retention times of 60 and 100 min indicates that the effect of flowrate on turbidity removal was less significant than the current density when sufficiently long retention time was allowed. The drastic increase in the resid-ual iron concentration was also visresid-ually observed by noting the intensified yellowish color in the reactor as the current density was increased. As mentioned previously, applying

Fig. 6. The effect of current density on the residual iron concentration in the effluents for wastewater treated at various hydraulic retention times (T0= 70 NTU, lp= 5 cm).

higher current density would have caused restabilization of colloids, eventually leading to reduced turbidity removal ef-ficiency as well as very high residual iron concentration.

3.5. Effect of hydraulic retention time

The effect of hydraulic retention time on the colloidal re-moval in parallel-plate electrodes are two-fold, namely the time required for the colloidal migration due to electrophore-sis toward the anodic surface, and the time required for the actual coagulation process to occur. Liberation of ferrous ions is considered instantaneous and should not be a limiting factor as compared to colloidal migration and coagulation. For the purpose of identifying the predominant effect of the flowrate, the characteristic time of particle migration was estimated and compared to that of the effective particle retention time in the active region of the electrodes. Conceptually, colloidal destabilization and aggregation mainly occur near the anodic surface because the mass transport of the liberated ions was driven by diffusion only[16]. This is also supported by the growth of gelatinous film on the anodes as well as the accu-mulating sediments at the foot of each anode. Consequently, the characteristic time for colloidal migration is defined as the time required for particles locating farthest from the anodes to reach the anodic surface. By performing a force balance (i.e., electrophoretic force and drag force) in the direction normal to the bulk fluid flow, a steady-state electrophoretic migration velocity can be obtained:

u = qE

6πµRp (8)

where E is the electrical field strength (V m−1),µ the dynamic viscosity of water, Rp the mean particle radius, and q the

particle charge (C) which can be calculated from a modified Debye–H¨uckel expression[17]:

q = 4πεRp

|ζ| (9)

In the last equation, ζ is the measured zeta potential

(−55 mV at pH 9.5), and ε is the dielectric constant of water (6.95× 10−19C2J−1m−1 at 25◦C). Accordingly, applica-tion of current densities from 3.6 to 14.5 A m−2would yield characteristic migration times between 2 and 4 s with an inter-electrode distance of 50 mm. The range of these character-istic times was more than two orders of magnitude smaller than the effective particle retention times used in this study, implying that the migration velocity under the experimental conditions was not a limiting factor to the particle removal efficiency. In contrast, the rate of actual coagulation appeared to dictate the system performance, and that the hydraulic re-tention time must be sufficiently long to allow coagulation

process to complete. Lai and Lin [12]previously reported

that, in their batch study for electrocoagulation of CMP sus-pension using cast iron as the electrodes (E = 375 V m−1), approximately 30 min was needed for particle aggregation

(8)

Fig. 7. Effect of inter-electrode distance (solid symbols: 5 cm, hollow sym-bols: 2 cm) on the turbidity removal (T0= 400 NTU,τ = 60 min).

to take effect and 80 min to complete. Matteson et al.[13]

also estimated the coagulation time required for 50% parti-cle reduction of kaolin suspensions (medium Rp= 1200 nm, ζ = −33 mV at pH 10.8) ranged between 50 and 100 min for

applied potentials between 15 and 50 V with steel mesh as the electrodes. These ranges of coagulation time appear to concur with our experimental results, and help explain the enhanced turbidity removal efficiencies at hydraulic retention times of 60 and 100 min.

3.6. Effects of inter-electrode distance

In a parallel-plate monopolar reactor, the electrical field strength in each cell channel can be controlled by varying the applied current, but is theoretically independent of the inter-electrode distance under the same current. In order to verify whether this presumption holds true for the channel-ized cell reactor used in this study, the reactor was arranged such that the inter-electrode distance was positioned at ei-ther 20 mm or 50 mm apart, while the number of electrodes

remained identical. Fig. 7 shows the normalized turbidity

removal profile (initial turbidity of 400 NTU) using the dif-ferent inter-electrode distances under two current densities

(10.9 and 14.5 A m−2). The results reveal improved turbidity removal at greater current density, which was attributed to the liberation of ferrous ions as mentioned earlier. However, for operation under both applied current densities, no discern-able differences between the removal profiles corresponding to the two different inter-electrode distances were observed. These results demonstrate that the continuous-flow, channel-ized reactor indeed operates very similarly to a monopolar reactor with fully divided cells.

3.7. Process comparison between chemical and electrocoagulation

Since chemical coagulation is still considered as the pri-mary method for treatment of CMP wastewater for majority of the chip-making facilities, it is of particular interest to compare the performance of the electrocoagulation vis-`a-vis chemical coagulation. As shown inTable 2, the quality of the treated water by chemical coagulation and electrocoagula-tion with respect to residual turbidity was comparable, both reaching below 5 NTU when correctly treated. The electro-coagulation process, however, does not require preliminary adjustment of pH, which in itself stages a significant advan-tage because pH adjustment for a large quantity of wastew-ater is never cost-effective. Moreover, it was observed that electrocoagulation process was capable of removing organic compounds from the CMP wastewater. In this study, the re-moval of total organic carbon (TOC) was typically more than 60% at the charge loadings of 8 F m−3for a wide range of ini-tial wastewater TOC (5–30 mg L−1). It has been documented that, in electrocoagulation, indirect oxidation via electron transfer of metal ions plays a much more important role than direct oxidation of adsorbed molecules on the anodic surface

[12,18]. Nevertheless, the nature of low conductivity of the CMP wastewater may have inhibited more efficient removal of the organic components.

Table 2also shows the jar-test results using PACl as the chemical coagulant for wastewater with different initial tur-bidities. From the data presented for the wastewater with low turbidity (74 NTU), it was determined that the pH must be maintained in the range between 4 and 7 to achieve particle

Table 2

Comparison of electrocoagulation with chemical coagulation processes

Process Initial turbidity (NTU) PACl dosage (mg L−1as Al) pH ζ-Potential (mV) Mean particle size (nm) Residual turbiditya (NTU)

Sludge dry solid (wt.% kg m−3water) Chemical coagulationa ₇₄_.2 ₃_.0 ₂ _+11.2 ₉₀ ₇₀ ₀_.66 4 +2.5 2350 4.9 6 +5.8 3650 2.8 8 −10.1 200 18 10 −43.2 150 53 200 10 6.0 Not measured 1 1.96 400 25 6.0 Not measured 3 4.43

Electrocoagulationb ₂₅₀ _– _9.5 _{Not measured} ₃ _1.09

a _{Residual turbidity measured after 40 min of settling.}

(9)

instability and growth in the particle size, and consequently effective turbidity removal. This result is consistent with the knowledge that the hydrolysis products such as Al(OH)2+,

Al(OH)2+, and Al(OH)3 generally form from PACl under

a narrow pH range between 5 and 6 [19]. The jar-test

re-sults also indicate that the minimum PACl dosage required for effective turbidity removal progressively increased with the initial wastewater turbidity, from as little as 3 mg L−1(as Al) for turbidity of 70 NTU up to 25 mg L−1for turbidity of 400 NTU at pH of 6. As a consequence, the dry solid weight of the sludge also increased accordingly.

To make a preliminary assessment with respect to sludge production by electrocoagulation, the system was continu-ously operated for more than 13 h for a wastewater turbidity of 250 NTU. The sludge was then collected by gathering the bottom sediment and by scraping as much deposited flocs off the anode surfaces as possible. As shown inTable 2, the dry solid weight of the resulting sludge from the electrocoag-ulation process (1.1 kg m−3) was significantly less than that from PACl coagulation (>2 kg m−3) for treatment of wastew-ater with similar initial turbidities. The low sludge volume (18 L m−3) generated by iron electrocoagulation process has also been documented in other studies[12]. Although these results could only serve as preliminary assessment by dis-counting important sludge properties such as water content, they nevertheless provide a strong evidence that the elec-trocoagulation process could conceivably reduce substantial sludge volumes, especially in view of the operating diffi-culties for chemical coagulation that often lead to coagulant overdosing.

4. Conclusions

The performance of a continuous-flow, channelized parallel-plate electrocoagulation system for treating synthetic oxide CMP wastewater was investigated. The dilute silica slurry was representative to the actual oxide CMP wastewa-ter, possessing a discharge pH between 9 and 10,ζ-potential

more than −50 mV, and aqueous conductivity of

approxi-mately 250␮S cm−1 at 0.4% dilution from original silica

CMP slurry, equivalent to an initial turbidity of 400 NTU. The experimental results revealed that the turbidity removal efficiency could be enhanced by increasing either the applied current or the hydraulic retention time, and that the removal efficiency was linearly correlated with the charge loading up to a level of 8 F m−3. Under short hydraulic retention time (τ ≤ 30 min), the turbidity removal efficiency was controlled by the current density, indicating that the process was lim-ited by the dosage of the liberated ferrous ions. For longer retention times, however, a threshold level of current density

appeared to exist, and operations beyond this level would lead to process deterioration. Calculation of the characteristic mi-gration time of particles due to electrophoresis indicated that the electrocoagulation process was not limited by the rate of particle migration, but by the rate of actual particle aggre-gation and coagulation that occur near the anodic surfaces. Preliminary appraisal between electro and chemical coagu-lation demonstrated that the former process produced lesser amount of dry sludge, and did not require pre-adjustment of wastewater pH.

Acknowledgements

This work is funded by the National Science Council (NSC) of the Republic of China through project grant NSC90-2211-317-001. The authors also wish to acknowledge the National Nano Device Laboratories (NDL) for their general supply of CMP slurries and assistance in particle size mea-surements.

References

[1] M.S. Kim, S.W. Woo, J.G. Park, Jpn. J. Appl. Phys. 41 (2002) 6342.

[2] B. Maag, D. Boning, B. Voelker, Semicond. Int. 23 (2000) 101. [3] SEMATECH International, International Technology Roapmap for

Semiconductors, SEMATECH, Austin, TX, 2001.

[4] J.H. Golden, R. Small, L. Pagan, C. Chang, S. Ragavan, Semicond. Int. 23 (2000) 85.

[5] J.H. Golden, J.E. Carrubba, Semicond. Fabtech. 13 (2001) 123. [6] S. Browne, V. Krygier, J. O’ Sullivan, E.L. Sandstrom, MICRO 17

(1999) 77.

[7] W. Jiang, C. Huang, H. Wu, Proceedings of the International Wa-ter Association Asia Environmental Technology Conference, Suntech City, Singapore, 30 October–1 November, 2001.

[8] R. Woodling, Semicond. Fabtech. 14 (2001) 69.

[9] G.C.C. Yang, T.-Y. Yang, S.-H. Tsai, Water Res. 37 (2003) 784. [10] B.M. Belongia, P.D. Haworth, J.C. Baygents, S. Raghavan, J.

Elec-trochem. Soc. 146 (1999) 4124.

[11] C. Tsouris, D.W. DePaoli, J.T. Shor, Z.-C. Hu, T.-Y. Ying, Colloids Surf. A 177 (2001) 223.

[12] C.L. Lai, S.H. Lin, Chem. Eng. J. 95 (2003) 205.

[13] M.J. Matteson, R.L. Dobson, R.W. Glenn, N.S. Kukunoor, W.H. Waits, E.J. Clayfield, Colloids Surf. A 104 (1995) 101.

[14] W. Den, C. Huang, S.-S. Chiou, Proceedings of the Ninth Interna-tional Semiconductor Environmental Health and Safety Conference (ISESH), San Diego, California, 9–13 June, 2002.

[15] C. Kan, C. Huang, J.R. Pan, Colloids Surf. A 203 (2002) 1. [16] G. Chen, X. Chen, P.L. Yue, Chem. Eng. Sci. 57 (2002) 2449. [17] P.C. Hiemenz, R. Rajagopalan, Principles of Colloid and Surface

Chemistry, third ed., Marcel Dekker, New York, 1997.

[18] J. Dziewinski, S. Marczak, E. Nuttall, G. Purdy, W. Smith, J. Taylor, C. Zhou, Waste Manage. 18 (1998) 257.

[19] R.D. Latterman, in: R.D. Latterman, A. Amirtharajah, C.R. O’Melia (Eds.), Water Quality and Treatment, McGraw-Hill, New York, 1999.