pubs.acs.org/Langmuir © 2010 American Chemical Society
Kinetics of Electroless Deposition: The Copper-Dimethylamine Borane System Daniela Plana, Andrew I. Campbell, Samson N. Patole, Galyna Shul, and Robert A. W. Dryfe* School of Chemistry, University of Manchester, Oxford Road, Manchester M13 9PL, United Kingdom Received January 27, 2010. Revised Manuscript Received April 30, 2010 A kinetic study of the electroless deposition of copper on gold, using dimethylamine borane (DMAB) as a reducing agent, has been carried out. The copper deposition rate in the electroless bath was determined to be 50 nm min-1, through electrochemical stripping of the copper deposits as well as from direct measurements of the film thickness using atomic force microscopy (AFM). Comparison with a galvanic cell setup, where the two half-reactions were physically separated, yielded a lower deposition rate of 30 nm min-1. An important kinetic effect of the surface on the oxidation of the reducing agent, and thus on the overall process, was therefore revealed. The efficiency of the process was measured over time, revealing the contribution of side reactions in the cathodic half-cell, particularly during the initial stages of the electroless process.
1. Introduction Electroless deposition is used in a wide range of industrial applications, ranging from the preparation of electronic circuitry and thermal or electrical conductors to metallurgy and corrosion prevention; the process has also found simple decorative uses.1-4 Electroless deposition of copper, in particular, has found applications in microelectronics and as a metallic starting point for subsequent metallization by other techniques due to the ease of copper deposition on prepatterned circuits, plastics and other nonconducting surfaces, and semiconductors.5-7 It has also been considered as a replacement for aluminum in ultralarge scale integration (ULSI) techniques due to its lower resistivity and higher resistance to stress-induced voiding and electromigration.8,9 Despite the widespread use of electroless deposition, and the apparent simplicity of the process, in which a metallic ion is reduced to its zerovalent state by the presence of a suitable reducing agent and a catalyzing surface, the reaction mechanisms are still not well understood.6 Electroless deposition occurs via a complex, multistep redox mechanism, involving the diffusion of the metal complex and the reducing agent to the catalytic surface and their subsequent reaction on it; each of these steps could limit the reaction.9,10 The evolution of hydrogen, which frequently accompanies many electroless processes, further complicates kinetic and mechanistic studies by producing microconvection which *Corresponding author: e-mail
[email protected]; Fax þ44 161 275-4734. (1) Balci, S.; Bittner, A. M.; Hahn, K.; Scheu, C.; Knez, M.; Kadri, A.; Wege, C.; Jeske, H.; Kern, K. Electrochim. Acta 2006, 51, 6251. (2) Jagannathan, R.; Krishnan, M. IBM J. Res. Dev. 1993, 37, 117. (3) Sverdlov, Y.; Bogush, V.; Einati, H.; Shacham-Diamand, Y. J. Electrochem. Soc. 2005, 152, C631. (4) Lelental, M. J. Catal. 1974, 32, 429. (5) Henry, J. R. Met. Finish. 2002, 100, 409. (6) O’Sullivan, E. J. Fundamental and Practical Aspects of the Electroless Deposition Reaction. In Advances in Electrochemical Science and Engineering; Alkire, R. C., Kolb, D. M., Eds.; Wiley-VCH: New York, 2002; p 225. (7) Vaskelis, A.; Jaciauskiene, J.; Stalnioniene, I.; Norkus, E. J. Electroanal. Chem. 2007, 600, 6. (8) Patterson, J. C.; O’Reilly, M.; Crean, G. M.; Barrett, J. Microelectron. Eng. 1997, 33, 65. (9) Dubin, V. M.; Shacham-Diamand, Y.; Zhao, B.; Vasudev, P. K.; Ting, C. H. J. Electrochem. Soc. 1997, 144, 898. (10) Schumacher, R.; Pesek, J. J.; Melroy, O. R. J. Phys. Chem. 1985, 89, 4338. (11) Donahue, F. M. J. Electrochem. Soc. 1980, 127, 51.
10334 DOI: 10.1021/la100390x
influences the overall rate of the electroless copper deposition process.11 The instability and complexity of typical deposition baths have made fundamental mechanistic and kinetic studies difficult, producing conflicting results.10,12,13 An intrinsic difficulty encountered when investigating the electroless deposition process arises from the spontaneous nature of the process and the fact that no net current is produced. A number of alternative techniques have been employed to measure or estimate deposition rates for the specific case of copper electroless deposition. Traditionally gravimetric techniques have been used to determine the weight change during deposition; this is frequently done ex-situ so in many cases a limited number of time points are evaluated or average weight gain is determined at a single time point.2,7,14-16 In-situ weight measurements can be made via the quartz-crystal microbalance.9,10,17-19 Surface techniques, such as the use of roughness step testers20 or profilometers,8,21 have been employed on a number of occasions to measure the thickness of the copper deposits. Where time-dependent weight gain has been performed, constant growth rates have been observed, although an initial induction period may be required. The rate of electroless copper deposition is extremely sensitive to the bath composition and operating conditions, with the limiting step for deposition, using formaldehyde as the reducing agent, changing from anodic to cathodic depending on the reagent concentrations.15 A detailed in-situ microbalance study demonstrated that the deposition process was first order in the (12) Homma, T.; Tamaki, A.; Nakai, H.; Osaka, T. J. Electroanal. Chem. 2003, 559, 131. (13) Djokic, S. S. Electroless Deposition of Metals and Alloys. In Modern Aspects of Electrochemistry; Conway, B. E., White, R. E., Eds.; Kluwer Academic: Dordrecht, 2002; Vol. 35. (14) Jiang, H. Y.; Liu, Z. J.; Wang, X. W.; Wang, Z. L. Trans. Inst. Met. Finish. 2007, 85, 103. (15) Mishra, K. G.; Paramguru, R. K. J. Electrochem. Soc. 1996, 143, 510. (16) Paunovic, M.; Vitkavage, D. J. Electrochem. Soc. 1979, 126, 2282. (17) Wiese, H.; Weil, K. G. Ber. Bunsen-Ges. Phys. Chem. Chem. Phys. 1987, 91, 619. (18) Hendricks, T. R.; Lee, I. Thin Solid Films 2006, 515, 2347. (19) Feldman, B. J.; Melroy, O. R. J. Electrochem. Soc. 1989, 136, 640. (20) Aithal, R. K.; Yenamandra, S.; Gunasekaran, R. A.; Coane, P.; Varahramyan, K. Mater. Chem. Phys. 2006, 98, 95. (21) Patterson, J. C.; Dheasuna, C. N.; Barrett, J.; Spalding, T. R.; O’Reilly, M.; Jiang, X.; Crean, G. M. Appl. Surf. Sci. 1995, 91, 124.
Published on Web 05/12/2010
Langmuir 2010, 26(12), 10334–10340
Plana et al.
methylene glycolate concentration (the conjugate base of the hydrated form of formaldehyde) and zeroth order in copper: the observation of a strong deuterium isotope effect led to the conclusion that the cleavage of the C-H bond in the adsorbed glycolate anion was the rate-determining step.10 In contrast, earlier kinetic studies of the copper/formaldehyde system have reported conflicting results, with a fractional order dependence on copper and a low sensitivity to the formaldehyde concentration.22,23 One explanation given for the large differences quoted for reaction kinetics is the variation in reagent concentration between the interface and bulk solution; reagent orders become more uniform if extrapolated interfacial concentrations are used for kinetic studies.10 A further issue is the possible change in limiting process with substrate geometry: for deposition into deep or narrow features (important for via filling), mass transport can become the limiting factor.12 One recurring question in the kinetic study of electroless deposition concerns the interdependence (or otherwise) of the cathodic and anodic processes. If these processes can be treated as independent, then the mixed potential theory holds.15,24 This theory is applied to electrochemical systems where two redox couples, with different standard potentials, are in contact and reach equilibrium by reduction of the more electropositive couple, with the simultaneous oxidation of the other couple. The theory assumes that once this condition is achieved, at a potential between the equilibrium potentials of the two redox systems, the rate of both reactions is equal; hence no net current flows.25,26 Accordingly, physical separation of the cathodic and anodic processes (i.e., a galvanic cell, with connection via a salt bridge) would produce a deposition rate identical to that seen in the conventional electroless bath. There are several limitations to the use of this theory, such as changes in the rate-determining step throughout the potential range covered by each reaction or in the surface kinetics of one or both reactions, due to adsorption of a bath component or (particularly in the case of an electroless process) due to changes in the structure of the substrate arising from metal deposition. Previous kinetic studies have found a specific interaction between the two couples: for example, the deposition rate of copper (using formaldehyde) in a galvanic cell was found to be lower than that seen in the electroless bath.6,15,27,28 As the conditions needed for the application of the mixed potential theory in the study of electroless deposition are not necessarily met, for example in some cases of Ni-P and Cu deposition, more complex methods, such as rapid solution exchange and bipolar configurations, have been applied.17,29,30 Weil et al. studied the copper/formaldehyde system in detail: the solution exchange method (rapid removal of either the copper ions or formaldehyde, while maintaining the substrate at the deposition potential) gave a significant decrease in deposition rate, suggesting that an interdependence of the half-reactions existed.17 Subsequent work by Weil used capacitance measurement, (22) Molennar, A.; Holdrinet, M. F. E.; van Beek, L. K. H. Plating 1974, 61, 238. (23) Dumesic, J.; Koutsky, J. A.; Chapman, T. W. J. Electrochem. Soc. 1974, 121, 1405. (24) Izumi, O.; Osamu, W.; Shiro, H. J. Electrochem. Soc. 1985, 132, 2323. (25) Wagner, C.; Traud, W. Z. Elektrochem. 1938, 44, 391. (26) Bindra, P.; White, J. R. Fundamental Aspects of Electroless Copper Plating. In Electroless Plating - Fundamentals and Applications; Mallory, G. O., Hajdu, J. B., Eds.; William Andrew Publishing: Norwich, NY, 1990; p 289. (27) Spiro, M. A Critique of the Additivity Principle for Mixed Couples. In Modern Aspects of Electrochemistry; Bockris, J. O. M., Conway, B. E., White, R. E., Eds.; Plenum Publishers: New York, 2001; Vol. 34, p 1. (28) Mital, C. K.; Shrivastava, P. B.; Dhaneshwar, R. G. Met. Finish. 1987, 85, 87. (29) Abrantes, L. M.; Correia, J. P. J. Electrochem. Soc. 1994, 141, 2356. (30) Plana, D.; Shul, G.; Stephenson, M. J.; Dryfe, R. A. W. Electrochem. Commun. 2009, 11, 61.
Langmuir 2010, 26(12), 10334–10340
Article
surface-enhanced Raman spectroscopy, and scanning tunneling microscopy to show that a microscopically rough surface was formed on deposition, leading the authors to conclude that the rough surface contained specific copper sites which acted as formaldehyde oxidation catalysts; i.e., the morphology evolved because of the strong interdependence between the cathodic and anodic processes.31-33 By contrast, earlier work on copper deposition with formaldehyde had assumed that the mixed potential approach was valid.16,26,34 Similarly for Ni-P alloy deposition via electroless methods, contradictory literature exists which assumes that the mixed potential theory holds35 or specifically states that it is invalid for this system.29 This contribution focuses on the mechanism of electroless copper deposition: the most widely used reducing agent for copper is formaldehyde; however, environmental and toxicity concerns have more recently led to the adoption of alternative reducing agents such as DMAB.2,6,12,36 The aim of this article is to investigate the rate of copper deposition using DMAB in the electroless bath and in the corresponding galvanic cell and to assess the validity of the mixed potential theory in this case. The copper deposition rate in both cases was determined by a simple method, where the film thickness was determined from the oxidative stripping of the copper; the validity of this approach was confirmed by comparison with film thickness measurements using AFM.
2. Experimental Section All deposition and stripping experiments were performed in aqueous solution, using water (of 18 MΩ cm resistivity) prepared using a reverse osmosis unit (Millipore, Watford, UK) coupled to an Elga “Purelab Ultra” purification system (Veolia Water systems, Marlow, UK). Analytical grade copper(II) sulfate, 1,5,8,12tetraazadodecane, potassium hydroxide, potassium chloride, and agar were obtained from Sigma-Aldrich, DMAB and triethanolamine were obtained from Alfa-Aesar, and sulfuric acid (95þ%) was purchased from Fisher. The deposition substrates were gold disks (2 mm diameter) obtained from IJ Cambria Scientific (Burry Port, UK). Before each deposition, the disks were mechanically polished using diamond suspensions (1 and 1/4 μm grade, supplied by Kemet International, Maidstone, UK) and electrochemically cleaned in 0.1 M sulfuric acid by repeated cyclic voltammetry in a potential range between 0.00 and 1.70 V vs Ag/AgCl. Two typical electroless baths were used; in these cases all the reactants are present in one solution and the baths differ only in the concentration of the reducing agent. These are denoted Bath 1 and Bath 2 (see Schemes 1 and 2, respectively), where DMAB is the reducing agent, triethanolamine is added to control the pH, and 1,5,8,12-tetraazadodecane is employed as a complexing agent for the copper(II) ions to prevent the spontaneous formation of copper hydroxide.2,37 The pH was adjusted to 11.6 using potassium hydroxide and sulfuric acid as needed, while the temperature of deposition was kept constant at 55.0 ( 0.5 °C, using a thermostatic water bath. For baths under hydrodynamic control, an EG&G PAR model 616 rotator was used to rotate the Au disk at defined frequencies (Oak Ridge, TN). All measurements in the (31) Bittner, A.; Wanner, M.; Weil, K. G. Ber. Bunsen-Ges. Phys. Chem. Chem. Phys. 1992, 96, 647. (32) Wanner, M.; Wiese, H.; Weil, K. G. Ber. Bunsen-Ges. Phys. Chem. Chem. Phys. 1988, 92, 736. (33) Weber, C. J.; Pickering, H. W.; Weil, K. G. J. Electrochem. Soc. 1997, 144, 2364. (34) Elraghy, S. M.; Abosalama, A. A. J. Electrochem. Soc. 1979, 126, 171. (35) Kim, Y. S.; Sohn, H. J. J. Electrochem. Soc. 1996, 143, 505. (36) Sargent, A.; Sadik, O. A.; Luis, J. M. J. Electrochem. Soc. 2001, 148, C257. (37) Masahiro, S.; Hideyuki, T.; Makoto, S.; Masayuki, K. Electroless gold plating solution. European Patent Office, 1987.
DOI: 10.1021/la100390x
10335
Article
Plana et al.
Figure 1. Schematic of the galvanic cell. Scheme 1. Bath 1: Typical Electroless Bath Used
Scheme 2. Bath 2: Electroless Bath with Lowered DMAB Concentration (with Respect to Bath 1)
Scheme 3. Cell 1: Galvanic Cell with Similar Chemical Composition to Bath 1
Scheme 4. Cell 2: Galvanic Cell with Lower Concentration of Copper Ions (with Respect to Cell 1)
Scheme 5. Cell 3: Galvanic Cell with Similar Chemical Composition to Bath 1
electroless bath were performed six times, and the average values are presented here. Note that in many applications of electroless deposition baths contain other components (levelers, brighteners, etc.), which are typically surface-active organic molecules added to alter the morphology of the final deposit. Such additives were deliberately excluded from the baths studied here to simplify kinetic analysis. In order to study the interdependence of the two half-reactions which constitute the electroless deposition process, a galvanic cell corresponding to Bath 1 was used. The galvanic cell consisted of two glass cells connected by a salt bridge (glass U-tube filled with agar gel containing 1.0 M KCl), as shown in Figure 1. The gold electrodes were connected externally through a switch. To follow the course of the reaction, a National Instruments (Austin, TX) PCI-4065 digital multimeter card was used; a custom-written LabView program allowed the open circuit potential (OCP) before deposition to be recorded, along with the current transients produced during deposition. At its highest accuracy setting of 22 bits, the PCI-4065 card can probe the reaction at a maximum rate of 10 Hz, with an accuracy of (0.15 μA and (26 μV when measuring dc current and voltage, respectively. Experiments using this setup were performed at least three times, and average values were taken. Two different chemical compositions were used in the galvanic setup described above and are denoted as Cell 1 and Cell 2 (see 10336 DOI: 10.1021/la100390x
Figure 2. Anodic stripping chronoamperomograms for deposits made from immersion in Bath 1 for 2, 5, and 10 min (dashed, dotted, and solid lines, respectively). Schemes 3 and 4, respectively). Temperature and pH were controlled as stated for the electroless baths. Given that the composition of the active surface changes during the electroless process, a copper electrode was introduced to the DMAB-containing half-cell to study the effect of the copper surface on the deposition rate. The Cu electrode was prepared by prior electroless deposition on gold, using Bath 1 for 10 min. The corresponding galvanic cell is denoted as Cell 3 (see Scheme 5). Note that reference electrodes were not present during the deposition experiments (in either the galvanic cell or electroless bath configurations). Possible chloride contamination of the deposit from the salt bridge was minimized by the distance between the bridge and the deposition substrate (2-3 cm): neither elemental analysis nor X-ray diffraction (see below) suggested CuCl formation, nor was any precipitate observed in the deposition solution. Stripping measurements were performed using an Autolab 100 potentiostat (Eco-chemie, Utrecht, The Netherlands) with an Ag/ AgClsat reference electrode made in-house and a platinum gauze as the counter electrode. The copper-covered electrodes were immersed in 0.1 M H2SO4, and a potential of 0.45 V vs Ag/AgCl was applied for 180 s (60 s was sufficient for samples where the deposition time was less than 1 min) to remove all the copper from the surface.38 Representative chronoamperometric responses, obtained during anodic stripping of the copper deposits, can be seen in Figure 2. The charge transferred was used to calculate the number of moles of copper deposited, using the Faraday equation, assuming that a two-electron oxidation occurs. This implicitly assumes that ambient oxidation of the copper deposit is minimal: the validity of this assumption is discussed below. The thickness of the deposits was obtained assuming homogeneous deposits with the properties of metallic copper.39 All AFM measurements were performed in contact mode, in air, with a Quesant 250 AFM supplied by Windsor Scientific Limited (Slough, UK). The samples for AFM were prepared on 1 cm2 gold foils, with thicknesses of either 0.25 or 0.50 mm (Goodfellow, Huntingdon, UK), which were cleaned by submersion in dilute nitric acid and mechanical polishing as stated for the gold disks. For copper film thickness measurements, the copper was electrolessly deposited, as described above, on selected parts of the gold substrates. Nail varnish was applied on half of each gold foil to achieve partial copper deposition: the viscosity of the varnish produces a well-defined edge (to within a micrometer) on drying. The varnish was then removed with acetone prior to AFM (38) Herranen, M.; Carlsson, J. O. Corros. Sci. 2001, 43, 365. (39) CRC Handbook of Chemistry and Physics, 76th ed.; Lide, D. R., Ed.; CRC Press, Inc.: Boca Raton, FL, 1995.
Langmuir 2010, 26(12), 10334–10340
Plana et al.
Article
Figure 3. AFM image of a copper deposit (left) on a gold substrate (right), using Bath 1 for 2 min. Plan view (upper) and cross-section view (lower) are shown. measurement to reveal the edge at the copper/gold interface (see Figure 3). Film thicknesses were measured across the gold/copper interface on at least five points of each sample, and an average value is given. A three-point tilt removing filter was used in all cases. An example of an AFM image, obtained for a 2 min deposit performed as described, is shown in Figure 3. AFM has been used to study the morphology of electroless deposits on a number of previous occasions.8,20,40 In this study, the roughness values were obtained from the root-mean-square values of surface height measurements, using the image analysis software supplied with the instrument. The roughness was measured over an area of 3 μm 3 μm, and all the measurements were taken from the central areas of the image to avoid edge effects. The rms roughness values reported are averages of at least six measurements on different areas across the surface. X-ray diffraction (XRD) was also used to provide structural information on the deposit: an X-pert (Philips) diffractometer was used for this purpose. Samples for diffraction were prepared on the same gold films as used for AFM measurements. The elemental composition of the deposits was determined by inductively coupled plasma optical emission spectroscopy (Fisons Horizon).
3. Results and Discussion Electroless copper deposition was carried out, for varying amounts of time, using Bath 1 in order to determine the deposition rate. Film thickness determination was performed through anodic stripping and AFM measurements as described in the Experimental Section; the results are presented in Figure 4. Using linear fits as shown in the figure, a steady deposition rate of 52 ( 3 nm min-1 was obtained from stripping measurements and a value of 49 ( 7 nm min-1 was found from AFM, for a range of deposition times from 1 min to half an hour. The agreement between the two measurement methods indicated that electrochemical stripping is a suitably simple, yet effective, technique to evaluate the kinetics of electroless deposition; it was consequently used throughout the rest of this work. The good agreement between the two methods, particularly at long times, implies that the fraction occupied by void space (which, if present, would increase the microscopic thickness measurement) in the copper deposit is quite low. The level of agreement also implies that the contribution due to ambient oxidation of the copper surface (which, if present, would decrease the effective thickness determined electrochemically) is low. XRD (40) Sverdlov, Y.; Shacham-Diamand, Y. Microelectron. Eng. 2003, 70, 512.
Langmuir 2010, 26(12), 10334–10340
Figure 4. Thickness of the copper deposits as a function of deposition time in Bath 1, measured using electrochemical stripping methods (circle) and AFM measurements (triangle). Least-squares lines of best fit of each data set are shown.
analysis of the deposit only revealed diffraction peaks attributable to metallic copper and the gold substrate (data not shown). The copper is deposited under strongly reducing conditions; i.e., the presence of the DMAB should minimize oxide formation during the deposition process. On removal from the deposition system, either to the ambient or to the acid solution for stripping measurement, oxidation of the copper surface may occur. Analysis of the elemental composition of the deposits indicated that they contained >97.5 wt % Cu and 0.2% C. The remaining sample weight is therefore attributed to oxygen, which corresponds to an oxygen content by atom of 9%; this value is expected to be considerably higher than that of the films following exposure to the ambient, as the elemental analysis was performed on pulverized electroless copper deposits, with surface to volume ratios a great deal higher than the as-deposited films. No traces of boron were found, in agreement with earlier reports on the electroless deposition of copper using boron-based reducing agents.6 Literature reports state that ambient oxidation of copper is restricted to a thin (
