Influence of Molecular Weight of Polyethylene ... - University of Idaho [PDF]

Journal of The Electrochemical Society, 152 共11兲 C769-C775 共2005兲

C769

0013-4651/2005/152共11兲/C769/7/$7.00 © The Electrochemical Society, Inc.

Influence of Molecular Weight of Polyethylene Glycol on Microvia Filling by Copper Electroplating Wei-Ping Dow,*,z Ming-Yao Yen, Wen-Bing Lin, and Shih-Wei Ho Department of Chemical Engineering, National Yunlin University of Science and Technology, Touliu, Yunlin 640, Taiwan The influence of the molecular weight 共Mw兲 of polyethylene glycol 共PEG兲 on the microvia filling by copper electroplating was demonstrated and examined by cross-sectional images using an optical microscope. The electrochemical behavior of PEG of different molecule weights in the copper electroplating was characterized by galvanostatic measurement. In the presence of excess Cl−, the surface coverage of PEG of various Mw adsorbed on the copper surface was characterized by observing the size and distribution of CuCl precipitates using a scanning electron microscope. As PEG Mw was increased, the best filling performance of plating formula was obtained when the PEG Mw ranged from 6000 to 8000 g/mol. Only large PEG amounts whose Mw exceeds 2000 g/mol can effectively polarize the cathode, in turn inducing the catalytic effect of bis共3-sulfopropyl兲 disulfide on copper deposition, resulting in a synergistic interaction between the suppressor and accelerator on the microvia filling. © 2005 The Electrochemical Society. 关DOI: 10.1149/1.2052019兴 All rights reserved. Manuscript submitted April 26, 2005; revised manuscript received July 1, 2005. Available electronically October 4, 2005.

Copper electroplating has been widely employed for fabricating the wires of printed circuit boards 共PCBs兲. The interconnection between two conducting layers in PCB is normally connected by means of metallization of microvia or through hole. In a conventional method, sidewall metallization of the microvia and through hole by copper electroplating was a crucial technique in the fabrication of multilayer PCBs. However, multilayer PCBs have not met the demands of state-of-the-art electronic products, of which the features are light, thin, short, and small. Instead, high-density interconnection 共HDI兲 designs with microvias formed by laser ablation have dominated the current trend of PCB fabrication. In this new process technology, the microvias, and even the micro through holes, must be fully filled by copper electroplating.1-3 In other words, the rate of copper electrodeposition at the bottom of the microvia must exceed that at the via opening and at the board surface. This phenomenon is called bottom-up filling or superfilling.2,4,5 In micro through hole filling, the rate of copper electrodeposition must be maximal at the center of the hole wall, leading to a sidewall center-up filling.3 The filling mechanism associated with the micro through hole differs from that associated with the microvia because its geometric feature and, therefore, the motion of the fluid, are different. This special filling behavior in copper electrodeposition was first applied to fabricate semiconductor devices to meet the requirements of dual-Damascene process.6 The plating electrolyte must contain at least two specific organic additives. One is called the suppressor, which is composed of polyethylene glycol 共PEG兲 and chloride ions and inhibits the copper deposition;7-14 the other is called the accelerator, which commonly bears a thiol 共i.e., 3-mercapto-1propanesulfonate, MPS兲 or disulfide 关i.e., bis共3-sulfopropyl兲 disulfide, SPS兴 functional group and enhances the rate of copper deposition.2,15-19 According to the superfilling mechanism proposed in previous works,4,5 a synergy between the suppressor and accelerator has to be established during electroplating. The suppressor functions mainly on the board surface, whereas the accelerator functions mainly at the via bottom; even the accelerators can accumulate on the copper/electrolyte interface at the via bottom.2,4,5 It has been shown that competitive adsorption between the suppressor and accelerator occurs on the copper surface.5,17,20 However, the domination of the competitive adsorption is associated with certain physical factors, such as geometric position around the features and forced convection of the plating solution.18,19,21-24 Several studies have reported that the inhibition effect of PEG on copper electrodeposition is a function of its molecular weight 共Mw兲. PEG can markedly exert its inhibition effect on copper electrodepo-

* Electrochemical Society Active Member. z

E-mail: [email protected]

sition when its Mw is beyond a certain value and Cl− ions are simultaneously present in the electrolyte.7,8 However, the influence of PEG Mw on the filling performance of microvia metallization remains unclear. A number of investigations have reported that the inhibition effect of PEG on copper electrodeposition is caused by a synergistic interaction among PEG, Cu+, and Cl−.7,10,13,14,25 These studies have showed that a complex film composed of PEG, Cu+, and Cl− adsorb on the copper surface and in turn inhibits the copper deposition. Moreover, previous works have also manifested that the inhibition effect of the complex film 共i.e., PEG–Cu+–Cl−兲 is related to the coordination of the ether group of PEG.10,14 Therefore, PEG Mw is an interesting topic in relation to the filling of microvia. The chemically competitive adsorption between the suppressor and accelerator has been demonstrated to depend on the strength of forced convection.21-24 The convection-dependent adsorption of these additives can be characterized by galvanostatic measurements 共GMs兲 of copper deposition conducted at two different rotation speeds of a copper rotating disk electrode 共Cu-RDE兲. If the overpotential of the Cu-RDE obtained at a slow rotation speed is smaller than that obtained at a fast one, then this plating solution certainly exhibits good filling performance.22,23 This work also characterized the competitive adsorption between various PEG and an accelerator using the aforementioned GM method and further correlated the potential difference 共⌬␩兲 between the two polarization curves, which were respectively measured at slow and fast rotation speed, with the filling performance. Experimental results indicate that the best filling performance of various PEG occurred when the PEG Mw ranged between 6000 and 8000 g/mol and the variation of the ⌬␩ with the PEG Mw was consistent with the variation of filling performance at various PEG. Therefore, the ⌬␩ obtained from the given GM seems to be a useful criterion that tells someone whether the plating solution is workable for microvia filling or not. Experimental The electroplating conditions, procedures, and pretreatment methods have been described in detail elsewhere.2,21 PCB fragments with many microvias formed by CO2 laser ablation were used herein as test samples for the filling plating. The dimensions of the PCB fragment were 6 ⫻ 15 cm. The diameters of the microvias were 65 and 105 ␮m. The depth of the microvias was 55 ␮m. The sidewall of the microvia was first metallized by electroless copper plating and subsequently deposited by copper electroplating in order to increase the thickness of the copper seed layer to be 3–4 ␮m before filling plating was conducted. The test PCB was plated at a current density of 18 A/ft2 共ASF, ⬃19.4 mA/cm2兲 for 70 min. Although the current density was based on the superficial area of the patterned PCB fragment, the examination of via cross sections using an optical micro-

C770

Journal of The Electrochemical Society, 152 共11兲 C769-C775 共2005兲

Figure 1. Filling performance definition of plating formula in microvia filling, given by 共H2 /H1兲 ⫻ 100%.

scope 共OM兲 共Olympus BX51兲 showed that the thickness of the electrodeposited film exterior to the microvia was within 10% of the nominal thickness. In each plating test, the PCB fragment was predipped in the plating solution and simultaneously shaken for 10 min in order to wet the microvias thoroughly and remove the air bubbles that were blocked in the microvias. Two phosphorus-containing copper slices were used as anodes and directly placed in the plating bath with a working volume of 1500 mL. Constant agitation was performed by continuous flows of air bubbles during electroplating to ensure good mass transfer. The plating bath was illustrated in detail in a previous work.21 The composition of the base electrolyte used for all plating tests was 0.88 M CuSO4·5H2O 共Riedel-de Haën, ACS兲 and 0.54 M H2SO4 共Merck, 96%, Ultrapure兲. The additives, 0.3 ppm SPS 共Raschig GmbH, Germany兲, 200 ppm PEG with various molecular weights 共200–20,000 g/mol兲 共Fluka兲, and 60 ppm Cl− ions 共NaCl, Fisher, Certified ACS兲, were added concomitantly to the base electrolyte. In the following results and discussion, the various PEG amounts are expressed as PEG-number, where the number is the molecular weight in g/mol. The temperature of the plating solution was maintained at 28°C. The filling performance of various plating formulas defined by 共H2 /H1兲⫻100% is illustrated in Fig. 1. The filling performance was evaluated from the cross-sectional OM pictures. The cross-sectional samples prepared for OM examinations were pretreated with polishing and etching in order to carefully show the microstructures of the copper deposits. The etching solution formulated for cross-sectional analysis was composed of H2SO4, NaCl, and Na2CrO4. All of the electrochemical analyses of additives were performed in a glass vessel that contained 100 mL of electrolyte solution using a PGSTAT30 共Autolab兲 potentiostat with three-electrode cell. The temperature of the electrolyte solution was maintained at 28°C during the electrochemical analysis. A platinum rotating disk electrode 共Pt-RDE兲 with a diameter of 8 mm was employed as a working electrode 共WE兲. The counter electrode 共CE兲 was a small copper bar placed in a small cylindrical cell containing a base electrolyte 共i.e., 0.88 M CuSO4·5H2O + 0.54 M H2SO4兲. The end-side of the cylindrical cell was sealed by a porous material in order to prevent a direct contact of additives with the CE during analysis. A saturated mercurous sulfate electrode 共SSE兲 served as the reference electrode 共RE兲. Before each electrochemical analysis, a thin copper layer with a thickness of 1.34 ␮m was predeposited onto the Pt-RDE in a predeposition bath, which only contained 0.06 M CuSO4·5H2O and 0.9 M H2SO4. Galvanostatic measurements 共GMs兲 of additive injection were performed using the Cu-RDE at a current density of 18 A/ft2. The rotation speed of Cu-RDE was kept at 400 rpm during the GM. At the beginning of the GMs, the base electrolyte contained only 0.88 M CuSO4·5H2O and 0.54 M H2SO4. After the GM was performed for ca. 200 s, 60 ppm Cl− was injected into the base electrolyte. Following the Cl− injection, 200 ppm PEG was subse-

quently injected into the Cl−-containing electrolyte at 400 s. After ca. 800 s, 0.3 ppm SPS was subsequently injected into the electrolyte. The filling performance of various plating formulas was characterized by means of a measurement of potential difference 共⌬␩兲 between two polarization curves obtained from GMs using Cu-RDE, which was individually operated at 100 and 1000 rpm. The compositions and concentrations of the additives used in the GMs were the same as those used for the filling plating. Millipore Direct-Q DI water 共18.2 M⍀ cm兲 was used to make all solutions that were used in the electrochemical analysis. To characterize the coverage of the adsorbed PEG, several PCB fragments were individually dipped in several electrolytes that were composed of 0.88 M CuSO4·5H2O, 0.54 M H2SO4, 150 ppm Cl−, and 200 ppm PEG with various Mws. The dipping time was 15 min. The coverage of the adsorbed PEG was characterized according to the crystal size and distribution of CuCl precipitates formed on the copper surface.21 The surface morphologies of the PCB fragments and the distribution of CuCl precipitates following the dipping treatments were examined using a scanning electron microscope 共SEM兲 共Hitachi S-3500N兲. Results and Discussion Filling capabilities of various PEG.— Figure 2 shows the typical plating results of microvias obtained from several plating solutions which contained 200 ppm PEG with various Mw, 60 ppm Cl−, and 0.3 ppm SPS. When the PEG Mw was smaller than 600 g/mol, the copper electrodeposition in the microvia was conformal, indicating that the deposition rate along the via profile was isotropic. According to the filling mechanism proposed in the previous works,4,5,26 two possible reasons may explain the results. The acceleration of copper deposition by the accelerator 共i.e., SPS + Cl−兲 at the via bottom was poor under the plating condition. Alternatively, the inhibition of copper deposition by the suppressor 共i.e., PEG + Cl−兲 at the board surface was inadequate. Actually, these two additives have to cooperatively exert their effects on the copper electrodeposition. Apparently, the plating formula lost the synergy between the suppressor and the accelerator when the PEG Mw was smaller than 600 g/mol. The loss of the synergistic effect led to a mat and rough appearance of the copper deposits, as shown in Fig. 2c and f, indicating that the locally vertical and lateral growth rates of the copper deposits on the plated surface were uneven. When the PEG Mw was larger than 600 g/mol, bottom-up filling, even superfilling, occurred. Moreover, the appearance of the copper deposits became smooth and bright, indicating that the synergistic effect between the suppressor and the accelerator was working well under this plating condition. The mass concentration of PEG was constant 共i.e., 200 ppm兲 in all cases, so the corresponding molar concentration was gradually low with increasing the PEG Mw. However, the inhibiting effect of the suppressor was increasingly strong with the PEG Mw 共see Fig. 3兲. A number of studies7,8,11-14 have shown that PEG–Cl− is adsorbed onto a copper surface in the form of monolayer. This adsorptive behavior implies that the projective coverage of the PEG–Cl− complex crucially affects the inhibition of copper deposition. Kelly and West11,12 reported that the adsorbed monolayer of PEG–Cl− complex is composed of numerous spherically packed PEG molecules, which leave ordered and abundant defects 共i.e., bare copper surfaces兲 available for accelerator adsorption. Hence, an optimal ratio of the concentration of suppressor to that of accelerator should exist, at which the synergistic effect is the strongest. Figure 2 shows that the appearance of copper deposits was smooth and bright when the PEG Mw exceeded 600 g/mol. These results mean that the copper growth rates in the vertical and lateral directions are approximately equal within a local area. In contrast, the mat and rough copper deposits caused by the addition of small PEG are attributed to the lower projection coverage of PEG. In other words, the small PEG results in much more coverage defects than

Journal of The Electrochemical Society, 152 共11兲 C769-C775 共2005兲

C771

Figure 2. Typical OM pictures after electroplating. 共c兲, 共f兲, 共i兲, 共l兲, 共o兲, and 共r兲 are the top views of the board surfaces. The others are the cross sections of various microvias. PEG molecule weights used in the plating solutions are 共a兲–共c兲 200, 共d兲–共f兲 400, 共g兲–共i兲 600, 共j兲–共l兲 4000, 共m兲–共o兲 8000, and 共p兲–共r兲 20,000 g/mol. All OM magnifications are 200⫻.

the large PEG. Surface-enhanced Raman spectroscopy14 demonstrated that the adsorption of PEG on the copper surface is associated with the presence of Cu+ and Cl−, where Cu+ serves as an intermedium to link PEG and Cl−, and Cl− acts as a dynamic anchor

to immobilize the PEG–Cu+–Cl− complex onto the as-deposited copper. It was demonstrated that the acceleration by thiol or disulfide molecules on copper deposition also needs the participation of Cl−.22 Therefore, if the PEG–Cu+–Cl− complex is not large enough,

C772

Journal of The Electrochemical Society, 152 共11兲 C769-C775 共2005兲

Figure 3. Galvanostatic measurements of additives injection. Current density of Cu-RDE is kept at 18 A/ft2. Rotating speed of Cu-RDE is 400 rpm. PEG molecule weights are 共a兲 200, 共b兲 600, 共c兲 2000, 共d兲 6000, 共e兲 8000, and 共f兲 20,000 g/mol.

the dynamic anchors 共i.e., Cl− ions兲 which underlie the PEG have many chances to interact with the contiguous thiol or disulfide molecules, as shown in Fig. 4a, turning them into practical accelerators.22 The binding strength between the anchors and the PEG is proportional to the number of ether groups of PEG, because the ether groups function as ligands to coordinate the Cu+ ions which are linked to the Cl− ions.14 Hence, a small PEG molecule corresponds to few anchors, weak PEG adsorption, and weak inhibition. Figure 4 presents the adsorption mechanism and is supported by the following results. Figure 5 shows the relationship between the filling performance and the PEG Mw. The filling performance is the best when PEG Mw is between 6000 and 8000 g/mol. When the PEG Mw was larger than 8000 g/mol, the drop extent in the filling performance with the PEG Mw in the large via was much more obvious as compared to that in the small via. This difference is related to the PEG Mw and the via size, suggesting that the fluid motion relatively acts easily in the large via and in turn favors the inhibition effect of the large PEG on the copper deposition at the bottom of the large via due to a strongly convection-dependent adsorption of the large PEG. The convection-dependent adsorption of the suppressor is confirmed by the electrochemical analyses shown in Fig. 8 and 9. The relevant discussion is addressed in the last section. The influence of Cl− transport on filling performance has been discussed in a previous work.23

Figure 4. Illustrations of additive adsorption on copper surface before passage of cathodic current. R represents 共CH2兲3–SO−3 and n is equal to 1, 2, 3,… .

Figure 5. Relationships between filling performances and PEG molecule weights. Via diameters are 共䊊兲 65 and 共䊐兲 105 ␮m. Via depth is 55 ␮m. Definition of filling performance is illustrated in Fig. 1.

The best filling performance occurring at the PEG Mw ranging between 6000 and 8000 g/mol suggests that a proper molecule size 共i.e., projective coverage兲 of the suppressor is critical for an appropriate synergy with the accelerator. When the defects of the suppressor coverage were more, its inhibition effect at the board surface was weak. Then the local enhancement of copper deposition due to the accumulation of the accelerator at the via bottom proposed in the previous works4,5,26 did not function, as shown in Fig. 2a-f. Therefore, the locally catalytic effect of the accelerator on the copper deposition at the via bottom must be established on a premise that an effective inhibition must simultaneously operate at the board surface. The selective exertion of the suppressor at the board surface has been demonstrated to be related to forced convection.22,23 However, the following GM will show that if the defects of the suppressor coverage are more, the assistance of forced convection in the PEG adsorption is limited. Consequently, conformal deposition occurred when the small PEG was used in the plating solution. Galvanostatic measurements.— GMs were carried out at constant current density of 18 A/ft2 to characterize the adsorption behavior of PEG with various Mws. Figure 3 shows that the addition of 60 ppm Cl− into the base electrolyte resulted in rate enhancement of copper deposition due to the effect of inner sphere electron transfer of chloride bridge.22,27 The potential oscillation following Cl− injection was ascribed to the transient formation and consumption of CuCl precipitates.28,29 The potential of the Cu-RDE almost reverted to the level that was in the absence of additives after 200 ppm PEG-200 was injected into the electrolyte. This result implies that the catalytic effect of the chloride bridge on copper deposition disappeared due to the formation of PEG–Cu+–Cl− complexes. However, the small PEG could not effectively block the copper deposition, so the Cu–RDE exhibited an additive-free potential. When 200 ppm PEG-600 was injected into the Cl−-containing electrolyte, copper deposition was significantly inhibited. The inhibiting effect of the suppressor increased with the PEG Mw. However, when the PEG Mw was larger than 6000 g/mol, a further increase in the inhibition was insignificant. Surface morphologies of copper surface.— If the copper surface is exposed to an electrolyte that contains Cu2+, PEG, and Cl− at high concentration, CuCl precipitates are formed at which copper surfaces are not covered by the PEG–Cu+–Cl− complex.21 This method was also used herein to verify the above statements, as illustrated in Fig. 4. The crystal growth and distribution of CuCl precipitates are compared using SEM images to further confirm the projective cov-

Journal of The Electrochemical Society, 152 共11兲 C769-C775 共2005兲

C773

Figure 7. Galvanostatic measurements of additives injection. Current density of Cu-RDE is kept at 18 A/ft2. Rotating speed of Cu-RDE is 400 rpm. PEG molecule weight is 8000 g/mol.

Fig. 6a, because the growth rate of CuCl was sluggish in this case, where little of copper surface that was not covered by the PEG–Cu+–Cl− complex was available.21

Figure 6. SEM images of PCB fragments after the fragments were dipped in various electrolytes for 15 min. Electrolyte is composed of 0.88 M CuSO4·5H2O, 0.54 M H2SO4, 150 ppm Cl−, and 200 ppm PEG. PEG molecule weights used are 共a兲 200, 共b兲 8000, and 共c兲 20,000 g/mol. SEM specimen is inclined at 30°.

erage of the PEG–Cu+–Cl− complex. The crystal size and particle density of the CuCl precipitate can be used to judge the extent of the suppressor coverage.21 Figure 6 shows three surface morphologies of PCB fragments that had undergone immersion in an electrolyte that contained 0.88 M CuSO4·5H2O, 0.54 M H2SO4, 150 ppm Cl−, and 200 ppm PEG. Apparently, the crystal number of the CuCl precipitate is a function of the PEG Mw and is not associated with the molar concentration of PEG. Namely, a larger PEG Mw corresponds to a lower crystal number. The details of the mechanism of CuCl formation have been reported in a previous work.21 The copper surface that is covered by the PEG–Cu+–Cl− complex is passivated against corrosion by Cl− ions. The corrosion that in turn leads to the formation of CuCl precipitates occurs at the coverage defects of PEG. This evidence is in agreement with the surface morphologies shown in Fig. 6, in which the corrosive extent of the bare copper surface shown in Fig. 6a was the most serious. Notably, the bare copper surface shown in Fig. 6a swarmed with small pinholes, indicating that there were much more defects among the PEG–Cu+–Cl− complexes as illustrated in Fig. 4a. These results support the adsorption mechanism of the projective coverage, proposed in Fig. 4. The defects among the adsorbed PEG, in which are the attack points of Cl− or SPS, are inversely proportional to the PEG Mw, as shown in Fig. 4. Hence, the mean crystal size of CuCl precipitate, shown in Fig. 6b and c, is smaller than that shown in

Potential-dependent competitive adsorption.— Figure 3 also reveals another interesting phenomenon that the competitive adsorption between the suppressor and accelerator on the copper surface depends on the cathodic potential rather than the PEG Mw. Curves 共a兲 and 共b兲 plotted in Fig. 3 show that the depolarization of the Cu-RDE was insignificant after 0.3 ppm SPS was injected into the electrolytes which contained 60 ppm Cl− and 200 ppm PEG-200 or PEG-600. In contrast, significant but slow depolarization on the CuRDE occurred after 0.3 ppm SPS was injected into the electrolytes containing 60 ppm Cl− and 200 ppm PEG-2000–20,000. To further confirm that this slow displacement of the adsorbed suppressor by the accelerator is not associated with the PEG Mw but with the overpotential of the Cu-RDE, a similar galvanostat analysis of Cl− ions with various concentrations and 200 ppm PEG-8000 were injected in sequence into the base electrolyte and shown in Fig. 7. The depolarization extent of the Cu-RDE increased with the injected Cl− concentrations. The polarization extent of the Cu-RDE caused by the subsequent injection of 200 ppm PEG-8000 varied with the Cl− concentration. The strongest polarization effect appeared at 30 ppm Cl−, which is in agreement with the results obtained from the potential sweep and step methods.9,11,12,30 The polarization curve 共a兲 reverted approximately to the level that was free of additive, similar to that in case 共a兲, shown in Fig. 3. This suggests that these Cl− ions were almost entirely subject to the PEG. However, the Cl− concentration was too low to form a compact complex film. The compact and stable complex film was not formed until the Cl− concentration exceeded 20 ppm. When the potential of the CuRDE was higher than −0.55 V vs SSE, the injection of 0.3 ppm SPS could markedly depolarize the Cu-RDE. The most apparent depolarization of the Cu-RDE occurred in the case indicated by curve 共d兲 plotted in Fig. 7. These results show that the catalytic effect of SPS on copper deposition is not only dependent on the cathodic potential26,31,32 but also on the Cl− concentration. The required overpotential for the SPS activation is induced by the interaction between the large PEG and the sufficient Cl− ions under the galvanstatic plating condition. Hebert33 proposed that when the Cl− coverage is equal to the critical coverage of PEG, a monolayer of PEG–Cu+–Cl− complex can be formed on the copper surface, which results in the strongest inhibition effect in the absence of SPS. This may explain the result

C774

Journal of The Electrochemical Society, 152 共11兲 C769-C775 共2005兲

of curve 共d兲 plotted in Fig. 7. After the SPS injection, SPS can interact with the adsorbed Cl− to become a practical catalyst for Cu2+ reduction.22 At the moment, if the Cl− coverage approximates to the critical coverage of PEG, a marked depolarization of Cu-RDE can be observed due to the interaction of SPS with the limited Cl− ions. If the Cl− coverage is higher than the critical coverage of PEG, as shown by curve 共e兲 in Fig. 7, the depolarization of Cu-RDE caused by the SPS injection is alleviatived as compared with curve 共d兲, because more available Cl− can still interact with the PEG. A similar phenomenon showing that depolarization extent of copper electrode caused by the addition of SPS is potentialdependent was also observed in previous works.26,31,32 Moffat et al.26,32 attributed these phenomena to the floatability and segregation of SPS on the as-deposited copper surface. In fact, the Cl− ion adsorption on the copper surface is also potential-dependent,14,34,35 and the SPS adsorption on the copper surface is similar to the adsorption of thiol self-assembly monolayer 共SAM兲 on gold and copper surfaces, which is also potential-dependent.36-40 Furthermore, the catalytic effect of SPS on copper deposition must be implemented under two premises: those Cl− ions must be present in the electrolyte and SPS must be reduced to MPS in advance.22 Accordingly, the potential-dependent catalysis of SPS can be reasonably interpreted as involving the reductive electrodesorption of MPS from the copper surface and its subsequent floating on the as-deposited copper atoms such that it can interact with the surface Cl− ions to form an electron-transfer net, which is the practical catalyst of copper deposition.22 Otherwise, the adsorbed MPS bonded to the copper surface was confirmed to be like SAM, which inhibits copper deposition.22 Figure 7 also reveals that there is a time delay in SPS activation after the SPS injection. Apparently, the delayed time constant 共i.e., ␶1, ␶2, and ␶3兲 depended on the Cl− concentration, which the lower the Cl− concentration, the larger the delayed time constant. Beyond 20 ppm Cl−, no time delay in SPS activation was observed. These results mean that Cl− is a promoter for SPS activation, which is consistent with a previous work.22 The behavior of the time delay in SPS activation also reveals that SPS can grab the adsorbed Cl− ions which have beforehand interacted with the PEG. Therefore, the depolarization extent in the case of 30 ppm Cl− caused by the SPS injection is more significant than that in the case of 60 ppm Cl−, whereas their polarization effects are reversed in the absence of SPS.

Filling performance characterization of plating formula.— Previous works22,23 have shown that if the plating formula can result in superfilling of microvia, a significant potential difference 共⌬␩兲 between the two polarization curves measured with Cu-RDE can be obtained in the GMs due to the convection-dependent adsorption of these additives. The slow rotation speed of the Cu-RDE gives rise to a relative depolarization curve, whereas the fast one leads to a relative polarization curve. Here, this approach is also employed to characterize the filling performance of various plating formulas, of which the PEG Mw is various. Figure 8 shows the results of the GMs. It is evident that ⌬␩ varied with the PEG Mw. To further clarify the relationships among the filling performance, the PEG Mw, and ⌬␩, the ⌬␩ values were plotted vs the PEG Mw as shown in Fig. 9. Evidently, when the PEG Mw is larger than 600 g/mol, a significant ⌬␩ can be obtained, which is consistent with the plating results shown in Fig. 2 and 5. Namely, bottom-up filling behavior is observed as the PEG Mw is larger than 600 g/mol. However, the ⌬␩ values could only be maintained well as the PEG Mw ranged from 6000 to 8000 g/mol. Beyond 8000 g/mol, small ⌬␩ value was obtained, which also coincides with the trend in filling performance with the PEG Mw shown in Fig. 5. Figure 8 also shows that PEG-20000 displays the strongest convection-dependent adsorption, because its overpotential is the highest one when the rotation speed of the Cu-RDE is kept at 1000 rpm. This can explain the significant drop of filling perfor-

Figure 8. Galvanostatic measurements of various plating formulas at two different forced convections. Cl− ion concentration is 60 ppm and SPS concentration is 0.3 ppm. PEG concentration is 200 ppm. PEG molecule weights are 共a and b兲 200, 共c and d兲 600, 共e and f兲 2000, 共g and h兲 8000, and 共i and j兲 20,000 g/mol. Solid lines are measured at 100 rpm and dotted lines are measured at 1000 rpm. Potential difference 共⌬␩兲 of each formula varying with plating time, which is evaluated by subtracting the potential of the dotted line from that of the solid line, is shown in Fig. 9.

mance in the large via when the PEG Mw is larger than 8000 g/mol, because the fluid motion is more active at the bottom of the large via compared to the small one. These results suggest that this electrochemically analytic measurement is a promising method for characterizing and predicting the filling performance of a plating formula. The occasion of the ⌬␩

Figure 9. Relationships between potential difference 共⌬␩兲 and plating time. Each curve is evaluated by subtracting the potential of the dotted line from that of the solid line that is shown in Fig. 8.

Journal of The Electrochemical Society, 152 共11兲 C769-C775 共2005兲 has been discussed in the previous works,22,23 and a detailed study of the ⌬␩ vs the filling performance will appear in a forthcoming work. Conclusions Microvias can be conditionally filled in the form of bottom-up using a plating formula with large PEG. When the mass concentration of PEG is fixed, the filling performance of plating formula is a function of the PEG Mw. With the increase in the PEG Mw from 200 to 20,000 g/mol, the best filling performance of various plating formulas occurs at the PEG Mw of 6000–8000 g/mol. The inhibiting effect of the suppressor is a function of the projective coverage of the PEG, which is determined by the PEG Mw. Large PEG results in dense and stable surface coverage of the suppressor, while small PEG does not, even though its molar concentration is high. The stable and strong inhibition of large PEG on the copper deposition is attributed to the great number of ether groups of PEG, which interact with the anchors 共i.e., Cu+ ¯ Cl−兲 that underlie the large PEG. Bottom-up filling of microvia is achieved by the synergistically competitive adsorption between the suppressor and accelerator. The competitive adsorption between the suppressor and accelerator is established on the promise that the cathode must be polarized by the suppressor to a certain level that in turn induces the activation of accelerator. The synergistically competitive adsorption between the suppressor and accelerator is not performed if the cathode cannot be polarized by the suppressor to a high enough level. In other words, the accumulated catalysis effect of the accelerator at the via bottom is conditionally effective on the superfilling of microvia. The adsorption behavior of the suppressor depends on forced convection. The accelerator adsorbed on the copper surface can be displaced by the suppressor due to the assistance of forced convection. Based on this specific adsorption behavior, a simple galvanostatic method that involves two different rotation speeds of Cu-RDE is developed to characterize and predict the filling performance of various plating formulas. Acknowledgments The authors thank the National Science Council of Taiwan for financially supporting this research under contract no. NSC 922214-E-224-005. Unicap Electronics Industry Corporation 共Taiwan兲 is appreciated for providing the required PCBs. Rockwood Electrochemicals Asia, Limited, 共Taiwan兲 is also appreciated for allowing us to use its SEM instruments. National Yunlin University of Science and Technology assisted in meeting the publication costs of this article.

References 1. T. Kobayashi, J. Kawasaki, K. Mihara, and H. Honma, Electrochim. Acta, 47, 85

C775

共2001兲. 2. W.-P. Dow, H.-S. Huang, and Z. Lin, Electrochem. Solid-State Lett., 6, C134 共2003兲. 3. W.-P. Dow and H.-H. Chen, Circuit World, 30, 33 共2004兲. 4. T. P. Moffat, D. Wheeler, W. H. Huber, and D. Josell, Electrochem. Solid-State Lett., 4, C26 共2001兲. 5. A. C. West, S. Mayer, and J. Reid, Electrochem. Solid-State Lett., 4, C50 共2001兲. 6. P. C. Andricacos, C. Uzoh, J. O. Dukovic, J. Horkans, and H. Deligianni, IBM J. Res. Dev., 42, 567 共1998兲. 7. M. Yokoi, S. Konishi, and T. Hayashi, Denki Kagaku oyobi Kogyo Butsuri Kagaku, 52, 218 共1984兲. 8. J. D. Reid and A. P. David, Plat. Surf. Finish., 74, 66 共1987兲. 9. J. P. Healy, D. Pletcher, and M. Goodenough, J. Electroanal. Chem., 338, 155 共1992兲. 10. D. Stoychev and C. Tsvetanov, J. Appl. Electrochem., 26, 741 共1996兲. 11. J. J. Kelly and A. C. West, J. Electrochem. Soc., 145, 3472 共1998兲. 12. J. J. Kelly and A. C. West, J. Electrochem. Soc., 145, 3477 共1998兲. 13. K. Doblhofer, S. Wasle, D. M. Soares, K. G. Weil, and G. Ertl, J. Electrochem. Soc., 150, C657 共2003兲. 14. Z. V. Feng, X. Li, and A. A. Gewirth, J. Phys. Chem. B, 107, 9415 共2003兲. 15. T. P. Moffat, J. E. Bonevich, W. H. Huber, A. Stanishevsky, D. R. Kelly, G. R. Stafford, and D. Josell, J. Electrochem. Soc., 147, 4524 共2000兲. 16. P. Taephaisitphongse, Y. Cao, and A. C. West, J. Electrochem. Soc., 148, C492 共2001兲. 17. M. Kang and A. A. Gewirth, J. Electrochem. Soc., 150, C426 共2003兲. 18. K. Kondo, T. Matsumoto, and K. Watanabe, J. Electrochem. Soc., 151, C250 共2004兲. 19. K. Kondo, N. Yamakawa, Z. Tanaka, and K. Hayashi, J. Electroanal. Chem., 559, 137 共2003兲. 20. Y. Cao, P. Taephaisitphongse, R. Chalupa, and A. C. West, J. Electrochem. Soc., 148, C466 共2001兲. 21. W.-P. Dow and H.-S. Huang, J. Electrochem. Soc., 152, C67 共2005兲. 22. W.-P. Dow, H.-S. Huang, M.-Y. Yen, and H.-H. Chen, J. Electrochem. Soc., 152, C77 共2005兲. 23. W.-P. Dow, H.-S. Huang, M.-Y. Yen, and H.-C. Huang, J. Electrochem. Soc., 152, C425 共2005兲. 24. M. Hasegawa, Y. Negishi, T. Nakanishi, and T. Osaka, J. Electrochem. Soc., 152, C221 共2005兲. 25. K. R. Hebert, S. Adhikari, and J. E. Houser, J. Electrochem. Soc., 152, C324 共2005兲. 26. T. P. Moffat, D. Wheeler, M. D. Edelstein, and D. Josell, IBM J. Res. Dev., 49, 19 共2005兲. 27. Z. Nagy, J. P. Blaudeau, N. C. Hung, L. A. Curtiss, and D. J. Zurawski, J. Electrochem. Soc., 142, L87 共1995兲. 28. H. P. Lee, K. Nobe, and A. J. Pearlstein, J. Electrochem. Soc., 132, 1031 共1985兲. 29. A. J. Pearlstein, H. P. Lee, and K. Nobe, J. Electrochem. Soc., 132, 2159 共1985兲. 30. V. D. Jović and B. M. Jović, J. Serb. Chem. Soc., 66, 935 共2001兲. 31. M. Tan and J. N. Harb, J. Electrochem. Soc., 150, C420 共2003兲. 32. T. P. Moffat, D. Wheeler, and D. Josell, J. Electrochem. Soc., 151, C262 共2004兲. 33. K. R. Hebert, J. Electrochem. Soc., 152, C283 共2005兲. 34. G. M. Brown and G. A. Hope, J. Electroanal. Chem., 405, 211 共1996兲. 35. G. M. Brown and G. A. Hope, J. Electroanal. Chem., 413, 153 共1996兲. 36. O. Azzaroni, P. L. Schilardi, and R. C. Salvarezza, Electrochim. Acta, 48, 3107 共2003兲. 37. O. Azzaroni, M. E. Vela, M. Fonticelli, G. Benítez, P. Carro, B. Blum, and R. C. Salvarezza, J. Phys. Chem. B, 107, 13446 共2003兲. 38. J. J. Calvente, Z. Kováčová, M. D. Sanchez, R. Andreu, and W. R. Fawcett, Langmuir, 12, 5696 共1996兲. 39. D.-F. Yang and M. Morin, J. Electroanal. Chem., 441, 173 共1998兲. 40. F. Loglio, M. Schweizer, and D. M. Kolb, Langmuir, 19, 830 共2003兲.