superfilling profile monitor
Shao-Yu Chiu, Jia-Min Shieh, Shih-Chieh Chang, Kun-Cheng Lin, Bau-Tong Dai, Chia-Fu Chen, and Ming-Shiann Feng
Citation: Journal of Vacuum Science & Technology B 18, 2835 (2000); doi: 10.1116/1.1322042 View online: http://dx.doi.org/10.1116/1.1322042
View Table of Contents: http://scitation.aip.org/content/avs/journal/jvstb/18/6?ver=pdfcov
Published by the AVS: Science & Technology of Materials, Interfaces, and Processing
Articles you may be interested in
Formation of nearly void-free Cu3Sn intermetallic joints using nanotwinned Cu metallization
Appl. Phys. Lett. 104, 171902 (2014); 10.1063/1.4874608
Self-aligned Cu electroplating of the front surface contact on silicon solar cells
AIP Conf. Proc. 1502, 230 (2012); 10.1063/1.4769147
Characterization of Vacancy Defects in Electroplated Cu Films by Positron Annihilation and its Impact on Stress Migration Reliability
AIP Conf. Proc. 817, 198 (2006); 10.1063/1.2173550
Effects of a new combination of additives in electroplating solution on the properties of Cu films in ULSI applications
J. Vac. Sci. Technol. A 18, 1207 (2000); 10.1116/1.582326
Recrystallization kinetics of electroplated Cu in damascene trenches at room temperature
by the superfilling profile monitor
Shao-Yu ChiuInstitute of Materials Science and Engineering, National Chiao Tung University, Hsinchu 30050, Taiwan Jia-Min Shieha)
National Nano Device Laboratories, 1001-1 Ta-Hseuh Road, Hsinchu 30050, Taiwan Shih-Chieh Chang and Kun-Cheng Lin
Institute of Materials Science and Engineering, National Chiao Tung University, Hsinchu 30050, Taiwan Bau-Tong Dai
National Nano Device Laboratories, 1001-1 Ta-Hseuh Road, Hsinchu 30050, Taiwan Chia-Fu Chen and Ming-Shiann Feng
Institute of Materials Science and Engineering, National Chiao Tung University, Hsinchu 30050, Taiwan
共Received 13 April 2000; accepted 11 September 2000兲
Gap-filling dynamics of several different species of additives for copper electrodeposition was investigated by monitoring the cross section of a partially filled copper profile on the scanning electron microscopy photo. The filling ration⌬y/⌬x between ‘‘bottom-up’’ with ‘‘sidewall shift’’ was found to be proportional to the filling power of additives. The adsorption-diffusion model combined with cathode polarization and cyclic voltammetric stripping measurements was employed to explain the attribution of additives in superfilling phenomena. The superfilling dynamics was achieved under behavior of additives providing selective inhibition gradient within the damascene feature. By means of those analyses, we have optimized the appropriate amount of additives and achieved the superfilling performance for 0.15 m vias with aspect ratio 6 by an acid-copper electrolyte with polyethylene glycol, C1⫺, and 2-mercaptopyridine共2-MP兲. Due to the additive of 2-MP, chelate formed which enhanced adsorption ability on Cu0 surface, and the concentration gradient between side-wall shift and bottom-up in the damascene became high enough to attend superfilling electroplating. © 2000 American Vacuum Society.关S0734-211X共00兲13506-1兴
I. INTRODUCTION
In the future, copper is going to replace aluminum in ultralarge-scale-integrated circuits 共ULSI兲 metallization be-cause of its low resistivity and high electromigration resis-tance. In 1997, IBM published their results on fully inte-grated devices with copper interconnects. Concurrently, a paper discussing six levels of planarized copper intercon-nects was also published by Motorola.1–3Gap-filling require-ment becomes the key issue because the damascene process is necessary for copper metallization. The major reason for the damascene process is that traditional process flow for aluminum metallization cannot be directly adopted for cop-per. Several different deposition techniques could be used to deposit copper films. These include physical vapor deposi-tion, chemical vapor deposideposi-tion, electroless plating, and electroplating.4–7 Among these techniques, electroplating emerges as a viable copper-filling technique for damascene processing due to its low deposition temperature, high throughput, high quality and, excellent via/trench filling capability.8,9
For the three commonly used electroplating baths, i.e., cyanide, pyrophosphate, and acidic baths, the acidic bath is the most suitable approach for copper wiring in ULSI
be-cause of its low cost, ease of control, low toxicity, and high deposition rate.10 Acid copper plating solutions generally contain copper salt as source of copper ions, a support elec-trolyte to provide conductivity of the solution, and the addi-tion agents for modifying kinetics of electroplating process and regulating the properties of deposition films. The addi-tion agents commonly include brighteners and levelers.10 Typically, the overall concentration for addition agents is in the order of parts per million. Nevertheless, both the quality of deposited copper films and gap-filling ability during elec-troplating are controlled by those minute amount of addi-tives.
Copper electroplating for damascene interconnection of submicron vias and trenches by adding leveling agents in the electrolyte to enhance the filling capability has been investigated.3,11–14 Most previous studies of copper electro-plating with similar bath chemistries utilized either unidenti-fied additives or commercial electrolytes. In our study, we researched additives from the patent literatures15 that were typical for electroplating unpatterned substrates and found the recipe of the composition and concentration of electro-lyte for submicron copper metallization application. The fill-ing capability is a key issue for superfillfill-ing in deep submicro and high aspect ratio damascene structure. Accordingly, how to get highly efficient leveling agents is an important work for copper electroplating in ULSI. Several methods for
de-a兲Author to whom correspondence should be addressed; electronic mail:
fining quantitative leveling in electrochemical deposition in the presence of an inhibiting additive have been reported.16–21As early as 1957, Watson and Edwards quan-tified the leveling ability of various additives based on polar-ization curves measured at different bulk leveling agent concentrations.16,17They derived leveling power of an addi-tive for low amplitude features and high electrolyte conduc-tivity, as shown in the following equation:
leveling power⫽Ca i 共/Ca兲i 共/i兲Ca ⫽⫺Ca i
冉
i Ca冊
, 共1兲 where Ca is the concentration of additive in the bulk, i is the current density, and is the overpotential. Further, Watson and Edwards were able to predict roughening or smoothing on copper deposition obtained from an electrolyte containing variable amounts of thiourea. Additionally, in Aroyo’s papers,19–21the definition of the leveling power in an elec-trodeposited coating on the surface of a nonregular profile was defined asLP⫽Ra,⫺Ra,c
d , 共2兲
where Ra,o and Ra,c are the roughness of the substrate and the plated surface, respectively, and d is the actually obtained coating thickness. The LP value that calculated from Eq.共2兲 is independent of the coating thickness. The degree of level-ing with different coatlevel-ing thickness could be compared. As previously mentioned, the leveling effect on the gap-filling ability specified for damascene feature, and not large struc-tures, was more important for very-large-scale-integrated ap-plication. Renewed definition of leveling power for additives in copper electroplating damascene structure is fundamental. We demonstrated a novel approach to monitor the filling profile changes during deposition as shown in Fig. 1. Through calculating the filling ratio⌬y/⌬x of ‘‘bottom-up’’ to ‘‘sidewall-shift’’ from the cross section of a partially filled copper profile on the scanning electron microscopy 共SEM兲 photo, the leveling power of different additives could be graded. This method could be employed to seek a proprietary additive enhancing superfilling performance. In a previous paper,13,14Kelly showed the leveling of 0.2m trenches by an acid copper electrolyte with polyethylene glycol 共PEG兲, Cl⫺, bis共3-sulfopropyl兲 disulfide, and Janus Green B.
Fur-thermore, they concluded that the gap-filling capability strongly depended on the activation overpotential of addi-tives in electrolytes. However, in our study, we achieved the superfilling performance in 0.15m vias with aspect ratio 6 and found that the activation overpotential is not the only factor influencing gap filling. In this work, we employed adsorption-diffusion model combined with cathode polariza-tion and cyclic voltammetric stripping共CVS兲 measurements to explain the attributes of additives in superfilling phenom-ena. By means of those analyses, we have optimized the appropriate amount of additives and achieved the superfilling performance for 0.15m vias with aspect ratio 6 by an acid-copper electrolyte with leveler of 2-mercaptopyridine 共2-MP兲 and several other typical additives. Due to the additive of 2-MP, chelate formed which enhanced adsorption ability on Cu0surface, and the concentration gradient between side-wall with bottom in the damascene became high enough to attend superfilling electroplating. Concluding unitary effects of additives, there are three indispensable requirements for achieving superfilling property; they are low surface tension of electrolyte, adequate activation overpotential, and suffi-cient inhibition gradient. If those conditions were present simultaneously during electroplating, then superfilling could be achieved.
II. EXPERIMENT
Our copper electroplating experiments were carried out in hull cell. The hull cell is a trapezoidal box of nonconducting material with one side at a 37.5° angle.22 The copper anode is laid against the right angle side and the wafer used as cathode is laid against the sloping side. Contact to the elec-trode was implemented outside of the electrolyte with an alligator clip. Agitation air was introduced into the solution from a compressor. In our experiment, the component of electrolyte was listed as Table I. The electrolyte was com-posed of copper sulfate 共30–90 g/l兲, sulfuric acid 共50–200 ml/l兲, chloride ion 共50–100 ppm兲, wetting agent 共80–200 ppm兲, and leveling agent 共10–2000 ppm兲. All electroplating works were proceeded at room temperature. The direct cur-rent 共dc兲 power supply and the sample size utilized in this work was Keithley model 2400 and 1⫻3 cm2. The patterned wafers were composed of a 30-nm-thick ionic metal plasma
共IMP兲–TaN layer as the diffusion barrier and a 200-nm-thick
IMP–Cu film as the seed layer. Results of both dc polariza-tion and CVS measurements were used to analyze the at-tributes of additives in acid copper electrolytes. dc polariza-tion analyses were performed on an EG&G Potentiostat/ Galvanostat Model 273A. The counter electrode was platinum and the working electrode was copper foil with constant surface area 1 cm2. The scanning rate was 2 mV/s
FIG. 1.⌬y/⌬x ratio is taken as the characteristic for comparison of various additives and feature scales to measure the filling capability and judge the effectiveness of the plating process.
TABLEI. Composition of electrolyte.
CuSO4•5H2O 30–90 g/l
H2SO4 100–400 g/l
Cl⫺ 50–100 ppm
PEG 20–2000 ppm
Leveling agents 10–2000 ppm
and the condition time was 30 s. The initial potential was started at⫹250 mV and the final potential was completed at
⫺1.5 V from open circuit potential 共OCP兲. All potentials
were reported relative to a double-junction saturated calomel electrode. A typical CVS technique was applied for evaluat-ing levels of organic additives such as suppressevaluat-ing and brightening agents.23,24 The CVS measurements were oper-ated at room temperature by EG&G PARC Model 616 RDE. The indicator electrode was 1 cm2 platinum electrode mounted in the rotating disc configuration and the Ag/AgCl electrode served as the reference electrode. Typically, oper-ating parameters were 50 mV/s sweep rate and 0–3500 rpm on the working electrode. The scanning range was from
⫺400 to 800 mV. Furthermore, the cross-section profiles of
deposit films were examined by using field emission scan-ning electron microscope.
III. RESULTS AND DISCUSSION
Various organic additives have been typically adopted in acid copper electrolytes to control the current distribution on the macro- and microscale. The inhibitors in electroplating baths with fewer concentrations could yield smoother elec-trodeposits on the microscopic scale. The effect on leveling might be explained through the classical diffusion-adsorption theory.3,16–21 According to the diffusion-adsorption mecha-nism, due to the diffusion distance to the peaks being shorter than the distance to the valleys, more inhibitors arrive on the peaks. Thus, the electrodeposition reaction is more inhibited on the peaks. Obviously, less metal is deposited and the pro-file becomes smoother. In our study, we developed a well-tuned and simple electroplating solution to fill defect-free structures and investigate the optimized recipe of each addi-tive for gap-filling performance. All components in our elec-troplating solution were found from the patent literature or chemical handbooks. Typically, those additives were for electroplating unpatterned or larger-scale-feature substrates. Their mechanisms on a submicron scale within deep vias/ trenches, for our purposes, should be quite different from dynamics on the blanket substrates. In order to distinguish the efficiency of different leveling agents more easily, we adopted the larger dimension patterns 共about 1 m, AR:1兲 and calculated the ⌬y/⌬x ratio of a partially filled trench profile to quantify the leveling power of additives. PEG as wetting agent and four different leveling agents were inves-tigated. They were ammonium peroxydisulfate, thiourea, 4-mercaptopyridine共4-MP兲, and 2-MP. The concentration of all additives was adjusted to the operative concentration for promoting the filling ability. The average ⌬y/⌬x values of each leveling agent in electrolytes containing PEG were 2-MP 共1.83兲⬎thiourea 共1.69兲⬎4-MP 共1.49兲⬎ammonium peroxydisulfate 共1.17兲 as shown in Fig. 2. The results showed that the filling power of 2-MP was higher than other additives. In Fig. 3共c兲, we observed that the superfilling ef-fect was achieved in vias by using 2-MP as the leveler. This resulted because the copper deposition rate on the bottom was higher than the rate on top corners and sidewalls. As seen in Fig. 3共b兲, thiourea as the leveler produced the
con-formal phenomenon with a seam. This resulted because the copper deposition rate on top corners and sidewalls were almost equal to the rate on the bottom. In Fig. 3共a兲, the anticonformal deposited trench profile with a void was found for 4-MP as the leveling agent. In this case, the copper depo-sition rate on top corners and sidewalls were higher than the rate on the bottom. The consistent results also could be seen in Fig. 4 for 0.35 m vias with AR 4. According to the earlier results, the⌬y/⌬x ratio is actually proportional to the filling power of additives.
Although leveling phenomena on anisotropic deposition rates has responded on the transient electroplating contour and resultant of gap-filling within the damascene structure, FIG. 2. ⌬y/⌬x ratio of different additives from the cross-section of a par-tially filled copper profile on the SEM photo.
FIG. 3. 0.8m, AR: 1 cross-section profiles of wafers on SEM photo.共a兲 Anticonformal deposited trench profile within a void was found by using 4-MP as the leveling agent,共b兲 the conformal phenomenon with a seam in a via was produced by using thiourea as the leveler, and共c兲 the superfilling effect was achieved by using 2-MP as the leveler.
the mechanisms of chemical-electrical reaction, interlayer near seed layer, and matter adsorption-diffusion of additives in the sub-0.2 m space, are necessary for purchasing cat-egorical certification. In electrochemical deposition, the
cath-ode polarization measurements provided comparison about inhibition ability between different additives. The higher overpotential is producing by additive, and the higher inhi-bition effect presents on copper deposition, as shown in Fig. 5. Also, it was found that the activation overpotentials of different additives at equal current density 共0.001 A ⫽ 1 A/dm2) in electrolytes containing PEG were thiourea
共⫺0.253 V兲⬎2-MP 共⫺0.186 V兲⬇4-MP共⫺0.188 V兲⬎
per-oxydisulfate 共⫺0.132 V兲. This was because the added thio-urea provided the highest suppression ability on copper deposition. In addition to dc polarization analysis, the CVS method would be employed to make the precise registration of electrochemical reaction changes on the cathode and an-ode, influenced by different organic additives, to be possible.23,24Our investigations with the aid of CVS show a single peak in the anodic area corresponding to copper dis-solution that responded from the deposition of copper in the cathodic area. The equal current density during the cathodic and anodic processes, as well as the existence of a single peak in the cathodic area, suggest that the current density during the cathodic period was used singularly for the reduc-tion of copper ions. In Fig. 6, the effects on the copper strip-ping area of various additives in solution containing PEG at constant electrode rotation rate 2500 rpm were no leveling agent 共6258.72兲⬎2-MP 共1312.78兲⬇4-MP 共1236.07兲
⬎thiourea 共281.42兲. CVS measured on the condition of
ro-tating electrode is advantageous for easily observing the diffusion-adsorption effect of additives on leveling. The lower areas around the peaks in the anodic region for differ-ent additives suggest that the effective concdiffer-entration of cop-per ions involved electrodeposition in the cathodic cycle and was decreased by the leveler. The inhibition effects of level-ing agents in copper electroplatlevel-ing baths with PEG were thiourea⬎2-MP⬇4-MP⬎no leveling agent. The results of CVS analyses were consistent with the results of cathode polarization measurements. But CVS and polarization mea-surements showed some difference from filling ratio⌬y/⌬x. FIG. 4. 0.35m, AR:4 cross-section profiles of wafers on a SEM photo.共a兲
Anticonformal deposited trench profile within a void was found by using 4-MP as the leveling agent,共b兲 the conformal phenomenon with a seam in a via was produced by using thiourea as the leveler, and共c兲 the superfilling effect was achieved by using 2-MP as the leveler.
FIG. 5. dc polarization curves of various additives in the electrolyte. The scanning rate was 2 mV/s and the condition time was 30 s. The initial potential was started at⫹250 mV and the final potential was ended at ⫺1.5 V from OCP.
Further evidences for verifying those disagreements became essential.
As a theoretical consideration, transport of the metal ion
共in this case, the cupric ion, Cu2⫹兲 and of inhibiting additives
is dominated by convection except within a concentration boundary layer that extends several tens of microns from the electrode surface.3From the adsorption-diffusion theory, the rate of electrodeposition is preferentially inhibited on those portions of the electrode surface that are more accessible to these agents by mass transport. Figure 7 shows the CVS results for the electrolyte-containing thiourea as a function of electrode rotation rate, and shows that the copper stripping area decreased共the inhibition effect increased兲 when the ro-tation rate increased from 0 to 3500 rpm. The explanation of this phenomenon is that the enhancement of additives mass transport is higher than the enhancement of cupric ion mass transport. This behavior of quicker transport of additive mass would result in near-equilibrium concentrations of leveler within the whole damascene structure. According to this kind of derivation, we suggested that the added thiourea was un-able to produce superfilling because it did not provide
selec-tive inhibition gradient within the via, though its activation overpotential is high. However, if 2-MP provided adequate overpotential and selective inhibition gradient, then superfill-ing could be obtained. As shown in Figs. 8 and 9, the elec-trolyte containing 2-MP or 4-MP, when the electrode rota-tion rate increased from 0 to 3500 rpm; the inhibirota-tion effect decreased because the copper stripping area increased. For 2-MP or 4-MP, the enhancement of cupric ion mass transport is higher than the enhancement of additives mass transport because their molecular weights are higher than thiourea. This trend was contrary to the case of thiourea. Finally, for superfilling to occur there should be concentration gradient of the leveler within the via.
Although the diffusion model could explain the thiourea reaction on leveling effect within the vias/trenches, the dif-ferent behaviors between 2-MP and 4-MP needed more proof. Diffusion speeds of those two levelers should be ap-FIG. 6. Effects of various additives on copper stripping. Operating
param-eters were 50 mV/s sweep rate and 2500 rpm on the working electrode.
FIG. 7. Effects of thiourea共100 ppm兲 on copper stripping with increasing electrode rotation rate共0–3500 rpm兲. Operating parameters were 50 mV/s sweep rate.
FIG. 8. Effects of 2-MP共12 ppm兲 on copper stripping with increasing elec-trode rotation rate 共0–3500 rpm兲. Operating parameters were 50 mV/s sweep rate.
FIG. 9. Effects of 4-MP共12 ppm兲 on copper stripping with increasing elec-trode rotation rate 共0–3500 rpm兲. Operating parameters were 50 mV/s sweep rate.
proximate because the molecular weights or chemical orga-nizations were similar. It is presumed that the superfilling effect of 2-MP is due to its strong adsorption, both on the copper seed layer and on the growing clusters of the new phase. In addition to the adsorption effect of additives on seed layer to reduce the growth rate of copper crystals, the concentration gradient of additives in vias or trenches also must be present. Following, when the concentration of lev-elers on the sidewall is more than on the bottom in vias or trenches, the stronger inhibitor effect appears on the sidewall more than on the bottom. Then, the growth rate of copper crystals on the bottom is higher than on the sidewall and the superfilling is achieved. When comparing 4-MP with 2-MP with no significant difference between 2-MP and 4-MP on diffusion behavior, we suggested that the inhibition effect depends on the density of monolayer of organic species. Ad-ditionally, the adsorption/desorption ability onto Cu0surface may be the key for their different leveling power. In previous studies,25 Alonso and co-workers concluded that in CVS measurements, if a platinum electrode is pretreated with 1
⫻10⫺3 M solution of 4-MP in water, the copper deposition exhibits behavior similar to that of 2-MP. The overpotential for 4-MP is identical to 2-MP. Since 4-MP is incapable of
forming a chelate, whereas 2-MP is capable, the fact that the same overpotential is observed in both cases suggests that the interaction is dominated by the sulfur atom. In H2SO4
solution共at PH⫽0兲, the nitrogen atom in 4-MP is protonated. Since the nitrogen site is located para to the sulfur, such protonation would共by inductive effects兲 cause a weakening of the sulfur-platinum bond. However, in 2-MP chemical structure, the nitrogen site is located ortho to the sulfur and then forming a chelate on electrode surface. Because the higher adsorption ability of 2-MP, more amount of 2-MP consumed at the top, and less amount to the bottom. Accord-ing to the earlier reasons, the concentration gradient of 2-MP between top and bottom is more than 4-MP because the ad-sorption ability of 2-MP is higher than 4-MP, as seen in Fig. 10. For the other additives, either the overpotential was not sufficient enough to suppress copper deposition at the top corners and sidewalls or the selective inhibition gradient could not be supplied, causing the anticonformal profile. Ad-ditionally, adding PEG as the wetting agent could prevent pits or pores produced in deposited copper by reducing sur-face tension and enhancing the filling capability. In our re-sults, we achieved the super-filling performance of 0.15m vias with aspect ratio 6 by an acid-copper electrolyte with PEG, Cl⫺, and 2-MP as shown in Fig. 11.
IV. CONCLUSIONS
The⌬y/⌬x ratio of a partially filled trench profile is pro-portional to the filling power of additives. Moreover, the results are consistent with the features of filling profiles. The sufficient activation overpotential, lower surface tension of the electrolyte, and selective inhibition gradient within the feature are three indispensable conditions for procuring the superfilling property. Adding PEG in the electrolyte could reduce surface tension and enhance the filling capability. In electrolyte containing PEG, the superfilling phenomenon
FIG. 10. Adsorption/desorption ability onto Cu0surface of 2-MP and 4-MP.
FIG. 11. Superfilling effect was achieved in 0.15m with aspect ratio 6 vias by using 2-MP as the leveling agent.
could be achieved in 0.15 m vias with aspect ratio 6 by using 2MP as the leveling agent, because it provided not only adequate overpotential but also selective inhibition gra-dient. The adsorption-diffusion model combined with cath-ode polarization and CVS measurements was employed to explain the attribution of additives in superfilling phenom-ena. The superfilling dynamics was achieved because the be-havior of additives provided selective inhibition gradient within the damascene feature.
ACKNOWLEDGMENTS
This work is partially sponsored by the National Science Council of the Republic of China under Grant No. NSC89-2721-2317-200. Technical support from the National Nano Device Laboratory and Merck-Kanto Advanced Chemical Ltd. is also acknowledged.
Presented at the ICAMPM Conference, 7–11 Feburary 2000, Santa Clara, CA.
1D. Edelstein et al., Tech. Dig. Int. Electron Devices Meet. 773共1997兲. 2
S. Vendatesan et al., Tech. Dig. Int. Electron Devices Meet. 769共1997兲. 3P. C. Andricacos et al., J. Res. Devel. 42, 567共1998兲.
4S. P. Murarka, Mater. Sci. Eng., R. 19, 87共1997兲. 5S. Gandikota et al., Microelectron. Eng. 50, 547共2000兲. 6
J. P. O’Kelly et al., Microelectron. Eng. 50, 473共2000兲. 7
E. K. Yung and L. T. Romankiw, J. Electrochem. Soc. 136, 206共1989兲. 8R. L. Jackson et al., Solid State Technol. 41, 49共1998兲.
9K. Lowery, VMIC Conference, 1996, p. 507. 10T. Tayler et al., Solid State Technol. 41, 47共1998兲. 11
P. Singer, Semicond. Int. 21, 38共1998兲. 12K. Weiss, Microelectron. Eng. 50, 433共2000兲.
13J. J. Kelly and A. C. West, Electrochem. Solid-State Lett. 2, 561共1999兲. 14J. J. Kelly et al., J. Electrochem. Soc. 146, 2540共1999兲.
15
Watson, U.S. Patent No. 4,376,685共1983兲. 16
J. O. Dukovic et al., J. Electrochem. Soc. 137, 3748共1990兲.
17K. G. Jordan and C. W. Tobias, J. Electrochem. Soc. 138, 1251共1991兲. 18C. H. Nanev et al., Surf. Technol 11, 117共1980兲.
19M. S. Aroyo, Plating Surf. Finishing 82, 53共1995兲. 20
M. Aroyo, Plating Surf. Finishing 85, 69共1998兲. 21M. Aroyo, Plating Surf. Finishing 85, 92共1998兲.
22D. R. Gabe and G. D. Wilcox, Trans. Inst. Metal Finish. 71, 71共1993兲. 23R. Haak et al., Plating Surf. Finishing 68, 52共1981兲.
24R. Haak et al., Plating Surf. Finishing 68, 62共1981兲. 25
