- Email: [email protected]
Structural properties of electrodeposited Cu-Ag alloys
Accepted Manuscript Title: Structural properties of electrodeposited Cu-Ag alloys Authors: Roberto Bernasconi, James L. Hart, Andrew C. Lang, Luca Magagnin, Luca Nobili, Mitra L. Taheri PII: DOI: Reference:
S0013-4686(17)31738-3 http://dx.doi.org/10.1016/j.electacta.2017.08.097 EA 30101
To appear in:
Electrochimica Acta
Received date: Revised date: Accepted date:
21-7-2016 27-7-2017 15-8-2017
Please cite this article as: Roberto Bernasconi, James L.Hart, Andrew C.Lang, Luca Magagnin, Luca Nobili, Mitra L.Taheri, Structural properties of electrodeposited Cu-Ag alloys, Electrochimica Actahttp://dx.doi.org/10.1016/j.electacta.2017.08.097 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Structural properties of electrodeposited Cu-Ag alloys Roberto Bernasconia, James L. Hartb, Andrew C. Langb, Luca Magagnina, Luca Nobilia,*, Mitra L. Taherib a
Dip. Chimica, Materiali e Ing. Chimica Giulio Natta, Politecnico di Milano, Via Mancinelli 7, 20131, Milano, Italy b
Department of Materials Science and Engineering, Drexel University,
LeBow Engineering Hall-344, 3141 Chestnut St., Philadelphia, PA 19104, USA * Corresponding author E-mail addresses: [email protected] (R. Bernasconi), [email protected] (J.L. Hart), [email protected] (A.C. Lang), [email protected] (L. Magagnin), [email protected] (L. Nobili), [email protected] (M. Taheri) Abstract Cu-Ag alloys are promising materials for electrical interconnections due to their combination of high conductivity and superior mechanical strength. In the present work, Cu-Ag alloys were electrodeposited from a cyanide-free electrolytic bath and different Ag percentages, ranging from 3 to 15 at.%, were obtained by controlling the deposition conditions. Electrochemical processes occurring during Cu-Ag deposition were investigated by means of cyclic voltammetry. Structural properties were examined by TEM and XRD and hardness was assessed by microindentation. TEM observations revealed the presence of nano-sized Ag precipitates at low Ag content, and the composition of the Cu-rich and Ag-rich phases in high-Ag alloys was deduced from XRD measurements. As-deposited alloys exhibited high hardness values and the contribution of solid-solution hardening was highlighted. Keywords: Electrodeposition; ; ; ; , Cu-Ag, TEM, XRD, supersaturation
1. Introduction Copper electroplated microstructures are widely used in the contemporary microelectronics industry. The electrical interconnections in printed circuit board (PCB) cards and integrated circuits constitute a large portion of the market, and the properties of pure copper are usually adequate to achieve the desired performances. For some applications like probing, however, where high mechanical properties are required together with good resistance to electromigration, copper possesses limitations. Such phenomenon is significant in structures of small thickness where a high current flows [1], like probing tips. Mechanical
1
strength and resistance to electromigration could be improved by the addition of an alloying element to the electrodeposited copper. Among the possible metals, silver is an attractive option [2], as the enhancement in mechanical properties provided by the addition of small quantities of this metal to copper is acceptably counterbalanced by a minor decrease in the electrical conductivity. The Ag-Cu phase diagram [3] clearly shows that the two metals are almost completely immiscible at room temperature, as a consequence of their positive heat of mixing. Nonetheless, silver and copper are known to form supersaturated solid solutions in particular manufacturing conditions [3, 4]. The methods used to synthesize these alloys consist in producing a metastable solid solution which is kinetically prevented from demixing into the separate metals and include electrodeposition [5, 6], mechanical alloying [1] and sputtering [7]. In particular, supplementary energy required to form the metastable solid solution is relatively small in electrodeposition, typically below 1 eV per atom. The thermodynamic properties of CuAg alloys have been extensively studied in previous works [3, 4], together with the optical and electrical properties [7, 8]. The most remarkable consequence of replacing copper atoms with silver atoms is the strain in the crystal lattice and the resulting increase in mechanical strength. Since supersaturated Cu-Ag alloys are intrinsically unstable, annealing at sufficiently high temperature can lead to the decomposition of the two individual metals, with significant effects on the properties of the material due to possible precipitation hardening. Therefore, the as-deposited alloy and the annealed alloy may exhibit the strength required for employment in probing technology. The existing literature on Cu-Ag electroplating is poor and is mainly focused on the deposition of alloys from cyanide electrolytes [5, 6]. In order to attain acceptable mechanical properties while keeping the fabrication cost at an affordable level, the Ag percentage in the deposit should not exceed 8 at.%. Studies concerning the deposition of alloys with 1-4 at.% Ag have been performed, but the processes described in the associated literature are not reliable for the deposition of coatings with high thickness or the preparation of electroformed parts [2]. These limitations can be overcome by using a pyrophosphate-iodide electrolyte and the deposition of Cu-Ag alloys from this bath was discussed in a previous work [9]. Alloys
2
containing Ag additions as high as 15 at.% can be obtained in a simple way, with a notable improvement of mechanical properties with respect to pure electrodeposited copper. The aim of the present work is the investigation of the structural properties of Cu-Ag alloys obtained from a pyrophosphate-iodide electrolyte, with a particular focus on the deposition kinetics and Ag demixing during deposition and annealing. Some aspects of the electrochemical behavior of the electrolyte are determined by performing cyclic voltammetry on two solutions, each containing one of the two metals together with its complexant. The presence of a real metastable alloy is demonstrated by means of transmission electron microscopy (TEM) coupled with the analysis of peak positions in X-Ray diffraction (XRD) spectra. In addition, the presence of Ag nano-precipitates is revealed by TEM observation.
2. Experimental Electrodeposition of Cu-Ag alloys was performed from a pyrophosphate-iodide electrolyte with the following composition: 50 g/L CuSO4 ∙ 5H2O, 150 g/L K4P2O7, 150 g/L KI, 75 mL/L 0.1M AgNO3, 10 g/L KNO3. Chemicals were purchased from Sigma Aldrich and used in the as-received condition. The pH of the obtained solution was 8.9. Electroplating was done in non-stirred conditions at 50°C, typical operating temperature for pyrophosphate copper solutions. Current densities between 2 mA/cm2 and 30 mA/cm2 were used, following the procedure described in a previous work [9]. The electrochemical cell consisted of a pure copper plate as anode and electrodeposited nickel as cathode. This nickel layer was deposited on carbon steel samples using an industrial solution of nickel sulfamate operated at a current density of 20 mA/cm2 and a temperature of 45°C, with moderate stirring. Steel plates were polished with emery paper to 1200-grit finish before nickel plating. The power supply system was a PSI 8065-05 T generator from ElektroAutomatik GmbH. To study the electrochemical properties of the electrolyte, two solutions were derived from the Cu-Ag bath and cyclic voltammetry experiments were performed on them. The first bath, namely AGSOL, contained only Ag and its complexing agent: 150 g/L KI, 75 ml/L 0.1M AgNO3. The second one, namely CUSOL, contained only Cu and its complexing agent: 50 g/L CuSO4, 150 g/L K4P2O7. Cyclic voltammetry measurements were carried out using an EG&G potentiostat-galvanostat model 273. The cell was in this case composed of glassy carbon (GC) as working electrode, a standard SCE as reference
3
electrode and graphite as counter electrode. Chemical composition of the Cu-Ag alloys was determined by X-Ray Fluorescence (XRF) using a Fischerscope X-Ray XAN instrument. Structural analysis was performed by XRD and diffraction patterns were acquired by a Philips 1830 appliance using Cu K radiation. Annealing of the Cu-Ag alloy deposited at 2 mA/cm2 was carried out in a Carbolite MTF 12/38/250 tubular furnace under nitrogen flow at 400 °C for 2 hours. A freestanding sample of Cu-4.4% Ag alloy was prepared for TEM analysis through a standard focused ion beam (FIB) lift-out procedure with a dual-beam FIB (FEI DB235); final thinning was completed at 5 keV with Ga ions. Selected area electron diffraction (SAED), lowmagnification imaging, and high resolution TEM (HRTEM) was completed with a JEOL 2100 LaB6 microscope equipped with a high-resolution pole-piece, operated at 200 keV. Hardness of the deposited alloys was measured by microindentation using a Fischerscope HCU instrument equipped with a diamond Vickers indenter.
3. Results and discussion 3.1 Electrochemistry of the Cu-Ag electrolyte The voltammetric investigation of the base electrolyte (containing both Cu and Ag) did not provide a good description of the deposition process because the superimposition of current peaks coming from different electrochemical reactions was observed in the most significant potential ranges [9]. For this reason, the electrochemical behavior of the deposition bath was investigated by voltammetry studies of the separate solutions (CUSOL and AGSOL). Cyclic voltammograms of the Cu pyrophosphate electrolyte (CUSOL) acquired with different scan rates are displayed in Fig. 1. The cathodic current starts to increase around -0.3 V and tends towards a plateau as the potential decreases, rather than exhibiting a distinct peak. It is reported that the predominant Cu pyrophosphate complex in alkaline solutions is Cu(P2O7)26- (CuL2), with minor presence of the CuP2O72- (CuL) complex [10, 11]. Equilibrium concentrations of Cu complexes and pyrophosphate ions (P2O74-) in the voltammetry electrolyte were evaluated by using equilibrium data reported in the literature [11, 12] and were found to be 0.2 M (CuL2), 0.15 mM (CuL) and 0.034 M (P2O74-).
4
In addition, the equilibrium potential for Cu reduction reaction resulted to be -0.267 V vs SCE, which is consistent with the value of -0.3 V observed in the voltammogram. According to the model proposed by Konno and Nagayama [11], Cu reduction involves two steps, dissociation of a ligand (reaction 1) and electrochemical reduction of the CuL complex, for [P2O74-] < 0.09 M (reaction 2). Cu(P2O7)26- ↔ CuP2O72- + P2O74-
(1)
CuP2O72- + 2e- Cu + P2O74-
(2)
The cathodic current given by reaction (2) is expected to be limited by the low concentration of the CuL complex (0.15 mM). Under these conditions, a current peak is not discernible in the voltammogram and the magnitude of the current increase is independent of the scan rate [13], as observed in Fig. 1. Pyrophosphate ions would be adsorbed on the electrode surface, unless the electrode potential becomes more negative than a critical value that depends on the solution composition [11]. For the voltammetry electrolyte, the estimated critical potential should be about -0.7 V vs SCE. Below this value, deposition inhibition due to pyrophosphate adsorption would vanish and the cathodic current might increase. Actually, the voltammogram in Fig. 1 shows that the cathodic current rises at -0.9 V and increases further as the potential decreases. Nevertheless, this feature can also be related to hydrogen discharge, which can contribute significantly to the cathodic current in this potential range [14]. In the reverse scan, Cu dissolves showing distinct anodic peaks above -0.3 V (Fig. 1). Cu-Ag alloys were deposited from the pyrophosphate–iodide electrolytes at cathode potentials ranging from -0.85 to -0.65 V vs SCE, where reactions (1) and (2) can describe the Cu reduction process. Cyclic voltammograms obtained from the Ag iodide solution at different scan rates are shown in Fig. 2. For each scan rate, three current peaks are observed in the cathodic sweep. The first peak occurs around -0.1 V and can be assigned to the reduction of I3- to I- [15, 16]. The other two peaks (A, B) arise in the potential range of -0.6 to -1.3 V and are related to Ag reduction reactions in iodide solutions [17, 18]. The more
5
intense current peak (A) occurs in the potential range where Cu-Ag alloys were deposited (-0.85 to -0.65 V). In the voltammogram, anodic peaks of Ag dissolution are identified in the potential range of -0.4 to -0.2 V (Fig. 2). A variety of Ag complexes can form in iodide solutions and their equilibrium concentration in the voltammetry electrolyte was calculated by using the relevant equilibrium constants [19], assuming that activity coefficients of ionic species are unity. The equilibrium concentrations are listed in Table 1. The species with the highest concentrations are AgI43- (4.73 mM) and AgI32- (1.23 mM), the other complex ions (Ag2I64-, Ag3I85-, AgI2-) amount altogether to 0.64 mM. It is noted in Fig. 2 that current peaks A and B are broad and asymmetric and accurate evaluation of their position and width would require mathematical processing. The operation of semi-differentiation could be used to convert such broad peaks in well-defined symmetric peaks [20,21], whose parameters (position, width, and height) can be estimated by employing the sech2 fitting function [22]. Semi-derivatives of the current peaks A and B in Fig. 2 have been calculated and the estimated fitting parameters are reported in Table 2. It is seen that the semi-differentiated peaks are still very broad and their width increases as the scan rate rises. This behavior is typical of electrolytic solutions containing quasi-labile metal complexes, which undergo ligand exchange reactions characterized by rate constants neither very small (inert complexes) nor very large (labile complexes) [23]. The interpretation of the electrodeposition process is very difficult, because all the silver complexes existing in the electrolyte may be involved in the cathodic reactions and many unknown kinetic parameters will play a decisive role in determining the overall behavior of a so complicated system. Nevertheless a qualitative explanation of the occurrence of peaks A and B in the voltammogram can be proposed. According to the criterion that complexes having a smaller number of ligands more likely contribute to metal electrodeposition [24], AgI2- should sustain the dominant cathodic reaction, as suggested in another study [25]. Yet, this complex ion is present with very low concentration (Table 1), therefore its discharge reaction should be supported by fast chemical dissociation of complex species
6
containing a larger number of ligands [24]. Taking into account the quasi-labile nature of complexes in our electrolyte, this situation is unlikely to establish. The most abundant complex species (Table 1) are AgI43- and AgI32-, that can participate in the reduction reactions (3) and (4): AgI43- + e- Ag + 4I-
(3)
AgI32- + e- Ag + 3I-
(4)
Kinetics of these reactions can be described by means of their exchange current density [26,24] and the values of this parameter for the two reactions can differ significantly from one another. It was shown that the position of voltammetric cathodic peaks shifts to lower potential values as the exchange current density decreases [26]. Hence, the appearance of peak B in the voltammogram (Fig. 2) may be associated with the cathodic reaction of the electroactive species characterized by the lower exchange current density and the higher overvoltage values. Another important effect is related to ligand exchange reactions that occur at the cathode surface as Ag reduction reactions proceed. In particular, the relative quantity of mononuclear (AgI43-, AgI32-, AgI2-, AgI) and polynuclear (Ag2I64-, Ag3I85-) complexes is dependent on the total Ag content. Under equilibrium conditions, the relative amount of polynuclear complexes, i.e. ([Ag2I64-]+[ Ag3I85-])/[Ag]total, decreases as the total Ag content in the solution becomes smaller (square brackets represent the molar concentration of the included species). As reactions (3) and (4) proceed, the concentration of AgI43- and AgI32- in the solution near the cathode will decrease, together with the total Ag content, therefore the equilibrium concentration of the polynuclear complexes Ag2I64- and Ag3I85- will diminish greatly and become smaller than the existing value. Consequently, ligand exchange reactions like the following ones may take place: Ag2I64- + 2I- 2AgI43-
(5)
Ag3I85- + 3I- AgI32- + 2AgI43-
(6)
7
Since quasi-labile complexes are involved, these reactions will take a finite time to produce an appreciable increase in the concentration of AgI43- and AgI32-. This supplementary provision of electroactive species may contribute to develop a second current peak after some time, i.e. at more negative potential values (peak B in Fig. 2). Because of the high stability of Ag iodide complexes, the equilibrium concentration of free Ag ions (Ag+) in the electrolyte is as low as 3.3·10-17 M (Table 1) and causes the equilibrium potential for Ag reduction to shift down to -0.509 V vs SCE. Accordingly, Cu-Ag alloys were deposited at potential values lower than -0.65 V vs SCE. 3.2 Cu-Ag microstructure: XRD analysis Cu-Ag alloys with varying Ag content were prepared by changing the current density in the deposition cell. The overall Ag content decreased as the current density increased and an atomic percentage as high as 15 % was attained at 2 mA/cm2 (Table 3). These samples gave the opportunity to analyze the metastable behavior of Cu-Ag alloys in a wide composition range, contrary to other studies reported in the literature [2,27]. The structure of the produced alloys was examined by XRD and the diffraction patterns are reported in Fig. 3. Pure copper deposited from a standard pyrophosphate solution was used as a reference sample and its XRD spectrum is shown in Fig. 3. Both copper and silver have face-centered cubic structure, but the lattice parameters are different. In the XRD patterns of all the Cu-Ag alloys, the reflections of copper are identified. One additional peak appears for 2 approximately equal to 38° in the spectra of the two alloys with higher Ag content. This peak corresponds to the (111) reflection of silver and indicates that a second phase rich in silver is present in these two alloys, together with the Cu rich matrix. Because the intensity of the Ag (111) peak is relatively low, it is argued that the other Ag reflections are too weak to be distinguished from the background signal. The interplanar spacing of Cu (111) planes was calculated from the angular position of the (111) reflection in the XRD patterns. The results are reported in Table 4, together with the difference between the
8
interplanar distance in the alloy and that in pure copper (dalloy - dCu). It is noted that the interplanar spacing increases as the silver content rises, which demonstrates that a solid solution of Cu and Ag exists in the electrodeposited alloys. According to Vegard’s law [28], the lattice parameter of a binary alloys depends linearly on the content of the solute element, but deviations from this law have been observed in many systems. An alternative model developed by Lubarda includes an apparent size of the solute atom to account for the electronic interactions between the solute and solvent atoms [29]. Experimental values of the lattice parameter of splat-cooled Cu-Ag alloys are also available [30]. The variation of the interplanar spacing (dalloy – dCu) is plotted as a function of the Ag atomic fraction in Fig. 4. The results of this work are compared with the data reported by Predel et al. [30] and the predictions of the Vegard’s law and Lubarda’s model [29]. The alloys with higher Ag content (12.9 at.% and 15.4 at.%) exhibit interplanar distances significantly lower than the values of single-phase Cu-Ag alloys with the same overall composition. This is consistent with the presence of the (111) Ag reflection in the XRD patterns and confirms that silver demixing occurs in these alloys. In the XRD spectrum of the Cu-12.9% Ag alloy, the Ag (111) peak is sufficiently well defined to attempt an evaluation of the composition of the two phases. The angular position of this peak (2 = 38.45°) corresponds to an interplanar spacing of 2.339 Å, which means that the Ag rich precipitate would contain a percentage of Cu comprised between 5.4 at.% (Vegard’s law) and 7.0 at.% (ref. [30]). Similarly, the Ag content of the Cu rich matrix can be deduced from the angular position of the Cu (111) reflection; it would be in the range of 5.1 at.% (ref. [30]) to 6.3 at.% (Vegard’s law). Because the reciprocal solubility of Cu and Ag at the deposition temperature (50 °C) and below is negligible [31], the system consists of two supersaturated solid solutions, i.e. silver demixing is not complete. The full width at half maximum of the Ag (111) peak indicates that the average crystallite size of the Ag precipitate should be around 6 nm, according to the Scherrer’s equation. The same reasoning applied to the Cu-15.4% Ag alloy led to estimate the silver percentage in the Cu rich matrix as a value comprised between 6.6 at.% (ref. [30]) and 8.1 at.% (Vegard’s law).
9
As expected, the annealing treatment of the alloy with the highest Ag content (15.4 at.%) enhanced the precipitation of silver, as shown in Fig. 5, where the XRD patterns of the annealed sample and the asdeposited one are compared. It is noted that all the Ag reflections are clearly recognized in the spectrum of the annealed alloy and the peaks are significantly narrower than those of the as-deposited sample. The average crystallite size in the annealed alloy was estimated by the Scherrer’s equation and was found to be 43 nm. For the electrodeposited alloys with low Ag content (up to 4.4 at.%), XRD patterns do not display any reflection demonstrating the existence of a second phase in addition to the Cu rich matrix. Nevertheless, the presence of a Ag rich precipitate cannot be excluded because the XRD method is not sensitive to small fractions of a second phase, especially if it is present in the form of nano-sized particles. Actually, even for low Ag content alloys (3.2-4.4% Ag) observed interplanar spacings are smaller than expected based on Vegard’s law or Lubarda’s model. This result suggests that excess Ag might be hosted in a second phase. 3.3 Cu-Ag microstructure: TEM analysis
To resolve the discrepancy between expected and measured lattice parameters for the low at. % Ag alloys, a Cu-4.4% Ag sample was analyzed by TEM. Figure 6A shows a SAED pattern of the alloy, demonstrating the samples polycrystalline nature. All of the major reflections correspond to Cu; however, at higher magnification, reflections matching expected Ag lattice spacings are observed (Figure 6B). For each strongly diffraction Cu (022) reflection, there is a much weaker and more diffuse reflection corresponding to the Ag (022) reflection. This reciprocal space relationship between Cu and Ag reflections indicates that Ag rich nanoparticles exist as coherent precipitates within larger Cu rich grains. Figure 6C shows a radially averaged intensity profile of the SAED pattern shown in Figures 6A and 6B. The weak Ag (022) reflections marked in Figure 6B produce a noticeable, albeit broad, peak around 7 nm-1. HRTEM analysis of Ag precipitates demonstrates that the precipitates are only a few nm in diameter (Figure 6D and 6E), explaining their absence from XRD measurements. 10
TEM results showed that the assessment of the interplanar distance in the Cu matrix is a reliable criterion for inferring the existence of Ag precipitates (Fig. 4). According to this criterion, Cu-Ag alloys with the lowest Ag percentages (3.2 and 3.5 at.%) are likely to have a very small fraction of Ag nano-precipitates similar to those observed in the Cu-4.4% Ag sample. Hardness values measured by microindentation are reported in Fig. 7. Alloys with low Ag content (3.5 at.% and 4.4 at.%) were found to be significantly harder than pure copper. Such hardness enhancement is to be ascribed to solid-solution strenghtening, with a minor contribution from precipitation hardening given by the small fraction of silver precipitates shown by TEM analysis. Further increase in hardness was observed in high-Ag alloys (12.9 at.% and 15.4 at.%). According to the XRD analysis, these alloys are characterized by higher Ag contents in the Cu matrix (from 5 at.% to 8 at.%) and detectable fractions of Ag rich precipitates. As a result, both solid-solution hardening and precipitation hardening become more intense, thus justifying the high hardness values displayed in Fig.7. In particular, hardness as high as 628 HV was attained in the Cu15.4% Ag alloy (the Vickers hardness HV is measured in kgf/mm2; 1 kgf/mm2 = 9.807 MPa). During the annealing treatment, silver was rejected from the Cu matrix and caused solid-solution hardening to vanish and precipitation hardening to intensify. Because hardness decreased notably after annealing, solidsolution hardening appears to contribute to alloy strengthening more than precipitation hardening. Actually, the annealed alloy was found to be approximately as hard as the Cu-3.5% Ag alloy, whose silver content is largely lower. However, all the alloys were notably harder than electrodeposited pure Cu.
4. Conclusions Voltammetric characterization of the Cu-Ag pyrophosphate-iodide electrolyte was carried out analyzing two separate solutions containing the two metals. Reduction of Cu was found to take place in two steps: dissociation of the CuL2 ligand and electrochemical reduction of the CuL species. Conversely, Ag deposition mainly occurs as consequence of the reduction of mononuclear complexes (AgI43-, AgI32-) to elemental Ag. Cu-Ag alloys with an Ag content ranging from 3 to 15 at.% were prepared by electrodeposition and characterized by TEM and XRD. A supersaturated Cu-Ag solid solution was found in all the alloys, while Ag rich precipitates were detected by XRD only in alloys with high silver content (12.9 at.% or more). It was
11
demonstrated, however, that Ag demixing was incomplete, as the Cu matrix remained significantly supersaturated with Ag. TEM analysis revealed the presence of nano-sized Ag precipitates which were not detected in the XRD pattern in a 4.4 at.% Ag sample, in agreement with matrix supersaturation predicted by XRD analysis. These TEM results, coupled with a comparison of predicted and measured lattice parameter values, indicate that all low at.% Ag samples studied here contain Ag nano-precipitates. Addition of Ag to Cu caused the hardness to increase notably, even at low Ag content. Comparison between as-deposited and annealed samples showed that solid-solution hardening was more effective than precipitation strengthening in increasing the alloy hardness. Acknowledgments Authors J.L.H., A.C.L., and M.L.T. acknowledge support from the Office of Naval Research under contract number N00014-14-1-0058. Authors M.L.T. and J.L.H. also acknowledge the Engineers as Global Leaders in Energy and Sustainability (EAGLES) project, supported by the FIPSE program of the US Department of Education, and the National Science Foundation under grant number 1429661. References [1] J.R. Lloyd, J.J. Clement, Electromigration in copper conductors, Thin Solid Films 262 (1995) 135. [2] S. Strehle, S. Menzel, J.W. Bartha, K. Wetzig, Electroplating of Cu(Ag) thin films for interconnect applications, Microelectron. Eng. 87 (2010) 180. [3] R. Najafabadi, D. J. Srolovitz, E. Ma, M. Atzmon, Thermodynamic properties of metastable Ag-Cu alloys, J. Appl. Phys. 74 (1993) 3144. [4] K. Uenishi, K. F. Kobayashi, K. N. Ishihara, P. H. Shingu, Formation of a super-saturated solid solution in the Ag-Cu system by mechanical alloying, Mater. Sci. Eng. A 134A (1991) 1342. [5] M. J. Kim, H. J. Lee, S. H. Yong, O. J. Kwon, S. K. Kim and J. J. Kim, Facile formation of Cu-Ag film by electrodeposition for the oxidation-resistive metal interconnect, J. Electrochem. Soc. 159 (2012) D253.
12
[6] M. J. Kim, S. H. Yong, H. S. Ko, T. Lim, K. J. Park, O. J. Kwon and J. J. Kim, Superfilling of Cu-Ag using electrodeposition in cyanide-based electrolyte, J. Electrochem. Soc. 159 (2012) D656. [7] J. Song, H. Li, J. Li, S. Wang, S. Zhou, Fabrication and optical properties of metastable Cu-Ag alloys, Appl. Opt. 41 (2002) 5413. [8] S. Strehle, J.W. Bartha, K. Wetzig, Electrical properties of electroplated Cu(Ag) thin films, Thin Solid Films 517 (2009) 3320. [9] R. Bernasconi, L. Nobili, L. Magagnin, Electrodeposition of supersaturated CuAg alloys in pyrophosphate-iodide electrolytes, ECS Trans. 58(32) (2014) 53. [10] G. W. Higgins, P. E. Sturrock, Polarographic Irreversibility of the Copper(II) Pyrophosphate System, Anal. Chem. 41 (1969) 633. [11] H. Konno, M. Nagayama, Mechanism of Electrodeposition of Copper from Cupric Pyrophosphate Solutions, Electrochim. Acta 22 (1977) 353. [12] O. E. Schupp III, P. E. Sturrock, J. I. Watters, A Study of the Stability and Basicity of the Copper(I1) Pyrophosphate Complexes Using the Dropping Amalgam Electrode, Inorg. Chem. 2 (1963) 106. [13] R. S. Nicholson, I. Shain, Theory of Stationary Electrode Polarography, Anal. Chem. 36 (1964) 706. [14] K. Johannsen, D. Page, S. Roy, A systematic investigation of current efficiency during brass deposition from a pyrophosphate electrolyte using RDE, RCE, and QCM, Electrochim. Acta 45 (2000) 3691. [15] J.-Y. Kim, J.-K. Lee, S.-B. Han, Y.-W. Lee, K.-W. Park, Improved Tri-iodide Reduction Reaction of CoTMPP/C as a Non-Pt Counter Electrode in Dye-Sensitized Solar Cells, J. Electrochem. Sci. Tech. 1 (2010) 75. [16] Y. A. Yaraliyev, Oxidation of iodide ions by means of cyclic voltammetry, Electrochim. Acta 29 (1984) 1213.
13
[17] V. Venkatasamy, S. Riemer, I. Tabakovic, Electrodeposition of eutectic Sn96.5Ag3.5 films from iodide– pyrophosphate solution, Electrochim. Acta 56 (2011) 4834. [18] H. H. Hassan, M. A. M. Ibrahim, S. S. Abd El Rehim and M. A. Amin, Comparative Studies of the Electrochemical Behavior of Silver Electrode in Chloride, Bromide and Iodide Aqueous Solutions, Int. J. Electrochem. Sci. 5 (2010) 278. [19] J. N. Butler, Ionic Equilibrium, Addison-Wesley Publishing Company, Massachusetts, 1964. [20] M. Grenness, K. B. Oldham, Semiintegral Electroanalysis: Theory and Verification, Anal. Chem. 44 (1972) 1121. [21] J. J. Toman, S. D. Brown, Peak Resolution by Semiderivative Voltammetry, Anal. Chem. 53 (1981) 1497. [22] D. M. Caster, J. J. Toman, S. D. Brown, Curve Fitting of Semiderivative Linear Scan Voltammetric Responses: Effect of Reaction Reversibility, Anal. Chem. 55 (1983) 2143. [23] J. J. Toman, R. M. Corn, S. D. Brown, Convolution Voltammetry of Metal Complexes, Anal. Chim. Acta, 123 (1981) 187. [24] A.G. Zelinsky, B.Ya. Pirogov, Numerical simulation and experiment in the systems of labile complexes of metals. Silver reduction from the thiourea electrolyte, J. Electroanalyt. Chem. 694 (2013) 68. [25] O. A. Ashiru, J. P. G. Farr, Kinetics of Reduction of Silver Complexes at a Rotating Disk Electrode, J. Electrochem. Soc. 139 (1992) 2806. [26] A. Survila, P. V. Stasiukaitis, Linear potential sweep voltammetry of electroreduction of labile metal complexes-I. Background model, EIectrochim. Acta 42 (1997) 1113. [27] S. Strehle, S. Menzel, K. Wetzig, J.W. Bartha, Microstructure of electroplated Cu(Ag) alloy thin films, Thin Solid Films 519 (2011) 3522. [28] L. Vegard, Die Konstitution der Mischkristallen und die Raumfüllung der Atome, Z. fur Physik 5 (1921) 17.
14
[29] V. A. Lubarda, On the effective lattice parameter of binary alloys, Mech. Mater. 35 (2003) 53. [30] B. Predel, G. Schluckebier, Investigations of the Structure, Thermodynamic Properties and Kinetics of Decomposition of Metastable Silver Copper Solid Solutions Produced by Splat Cooling, Z. Metallkd. 63 (1972) 782. [31] J. L. Murray, Calculations of Stable and Metastable Equilibrium Diagrams of the Ag-Cu and Cd-Zn systems, Metall. Trans. A 15A (1984) 261. Captions to illustrations Figure 1 Cyclic voltammograms of the Cu pyrophosphate solution (CUSOL) obtained with scan rates ranging from 5 to 20 mV s-1. Cathodic currents are plotted along the negative direction. Figure 2 Cyclic voltammograms of the Ag iodide solution (AGSOL) obtained with scan rates ranging from 10 to 50 mV s-1. Cathodic currents are plotted along the negative direction. Figure 3 XRD patterns of Cu-Ag alloys under various plating conditions; spectrum of pure Cu is reported for comparison. Figure 4 Experimental lattice expansion compared with data in the literature. Figure 5 XRD analysis of the sample deposited at 2 mA/cm2 (15.4 at.% Ag) before and after annealing at 400°C for 2 h. Figure 6
15
TEM analysis of the Cu-4.4% Ag sample. A)SAED pattern showing polycrystalline nature of sample. B) Magnified version of A) showing Ag and Cu (022) spots. C) Radially averaged intensity profile of SAED shown in A) and B) alongside expected peaks for Cu and Ag; the broad peak at ≈7 nm-1 corresponding to the Ag (022) reflection. D) and E) HRTEM images showing Ag nanoparticles several nm in diameter. Black arrows indicate the location of a Ag precipitate in each image; however, both images contain many unmarked precipitates. Figure 7 Vickers hardness of as-deposited and annealed Cu-Ag samples.
16
Fig.1
17
Fig.2
18
Fig.3
19
Fig.4
20
Fig.5
21
Fig.6
22
Fig.7
23
Table 1 – Equilibrium concentration of silver-iodide species in the voltammetry electrolyte. ________________________________________________________________________________ Species
Ag+
Concentration 3.3·10-17
AgI(aq) AgI2-
AgI32-
AgI43-
Ag2I64- Ag3I85- I-
3.9·10-9 2.2·10-6 1.2·10-3 4.7·10-3 3.6·10-4 2.7·10-4 8.7·10-1 mol dm-3
________________________________________________________________________________
24
Table 2. Position, height and width of semiderivative peaks calculated from current peaks A and B in Fig. 2. Coefficients of determination R2 are also reported. ________________________________________________________________________________ Peak A
Peak B
________________________________________________________________________________ Scan rate/mV s-1
10
20
50
10
50 Position/V vs SCE
-0.786 -0.819 -0.876
-0.943 -1.029 -1.183
Height/A cm-2 V-1/2
0.0259 0.0276 0.0312
0.0045 0.0083 0.0128
Widtha/V
0.143 0.180 0.248
0.048 0.085 0.157
R2
0.973 0.979 0.981
0.973 0.979 0.981
______________________________________________________________________________ a
Full Width Half Maximum
25
20
Table 3 – Ag content in Cu-Ag alloys at different current densities. Current density Ag (at.%)
j/mA cm-2 30
3.2
20
3.5
10
4.4
5
12.9
2
15.4
Table 4. Interplanar spacing of (111) planes in the Cu matrix and deviation from the value of pure Cu (dalloy – dCu).
Ag (at.%)
Peak position 2θ/°
d111 spacing d/Å
dalloy - dCu d/Å
0
43.43
2.082
0
3.2
43.27
2.089
0.007
3.5
43.25
2.090
0.008
4.4
43.21
2.092
0.010
12.9
43.06
2.099
0.017
15.4
42.95
2.104
0.022
26
S0013-4686(17)31738-3 http://dx.doi.org/10.1016/j.electacta.2017.08.097 EA 30101
To appear in:
Electrochimica Acta
Received date: Revised date: Accepted date:
21-7-2016 27-7-2017 15-8-2017
Please cite this article as: Roberto Bernasconi, James L.Hart, Andrew C.Lang, Luca Magagnin, Luca Nobili, Mitra L.Taheri, Structural properties of electrodeposited Cu-Ag alloys, Electrochimica Actahttp://dx.doi.org/10.1016/j.electacta.2017.08.097 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Structural properties of electrodeposited Cu-Ag alloys Roberto Bernasconia, James L. Hartb, Andrew C. Langb, Luca Magagnina, Luca Nobilia,*, Mitra L. Taherib a
Dip. Chimica, Materiali e Ing. Chimica Giulio Natta, Politecnico di Milano, Via Mancinelli 7, 20131, Milano, Italy b
Department of Materials Science and Engineering, Drexel University,
LeBow Engineering Hall-344, 3141 Chestnut St., Philadelphia, PA 19104, USA * Corresponding author E-mail addresses: [email protected] (R. Bernasconi), [email protected] (J.L. Hart), [email protected] (A.C. Lang), [email protected] (L. Magagnin), [email protected] (L. Nobili), [email protected] (M. Taheri) Abstract Cu-Ag alloys are promising materials for electrical interconnections due to their combination of high conductivity and superior mechanical strength. In the present work, Cu-Ag alloys were electrodeposited from a cyanide-free electrolytic bath and different Ag percentages, ranging from 3 to 15 at.%, were obtained by controlling the deposition conditions. Electrochemical processes occurring during Cu-Ag deposition were investigated by means of cyclic voltammetry. Structural properties were examined by TEM and XRD and hardness was assessed by microindentation. TEM observations revealed the presence of nano-sized Ag precipitates at low Ag content, and the composition of the Cu-rich and Ag-rich phases in high-Ag alloys was deduced from XRD measurements. As-deposited alloys exhibited high hardness values and the contribution of solid-solution hardening was highlighted. Keywords: Electrodeposition; ; ; ; , Cu-Ag, TEM, XRD, supersaturation
1. Introduction Copper electroplated microstructures are widely used in the contemporary microelectronics industry. The electrical interconnections in printed circuit board (PCB) cards and integrated circuits constitute a large portion of the market, and the properties of pure copper are usually adequate to achieve the desired performances. For some applications like probing, however, where high mechanical properties are required together with good resistance to electromigration, copper possesses limitations. Such phenomenon is significant in structures of small thickness where a high current flows [1], like probing tips. Mechanical
1
strength and resistance to electromigration could be improved by the addition of an alloying element to the electrodeposited copper. Among the possible metals, silver is an attractive option [2], as the enhancement in mechanical properties provided by the addition of small quantities of this metal to copper is acceptably counterbalanced by a minor decrease in the electrical conductivity. The Ag-Cu phase diagram [3] clearly shows that the two metals are almost completely immiscible at room temperature, as a consequence of their positive heat of mixing. Nonetheless, silver and copper are known to form supersaturated solid solutions in particular manufacturing conditions [3, 4]. The methods used to synthesize these alloys consist in producing a metastable solid solution which is kinetically prevented from demixing into the separate metals and include electrodeposition [5, 6], mechanical alloying [1] and sputtering [7]. In particular, supplementary energy required to form the metastable solid solution is relatively small in electrodeposition, typically below 1 eV per atom. The thermodynamic properties of CuAg alloys have been extensively studied in previous works [3, 4], together with the optical and electrical properties [7, 8]. The most remarkable consequence of replacing copper atoms with silver atoms is the strain in the crystal lattice and the resulting increase in mechanical strength. Since supersaturated Cu-Ag alloys are intrinsically unstable, annealing at sufficiently high temperature can lead to the decomposition of the two individual metals, with significant effects on the properties of the material due to possible precipitation hardening. Therefore, the as-deposited alloy and the annealed alloy may exhibit the strength required for employment in probing technology. The existing literature on Cu-Ag electroplating is poor and is mainly focused on the deposition of alloys from cyanide electrolytes [5, 6]. In order to attain acceptable mechanical properties while keeping the fabrication cost at an affordable level, the Ag percentage in the deposit should not exceed 8 at.%. Studies concerning the deposition of alloys with 1-4 at.% Ag have been performed, but the processes described in the associated literature are not reliable for the deposition of coatings with high thickness or the preparation of electroformed parts [2]. These limitations can be overcome by using a pyrophosphate-iodide electrolyte and the deposition of Cu-Ag alloys from this bath was discussed in a previous work [9]. Alloys
2
containing Ag additions as high as 15 at.% can be obtained in a simple way, with a notable improvement of mechanical properties with respect to pure electrodeposited copper. The aim of the present work is the investigation of the structural properties of Cu-Ag alloys obtained from a pyrophosphate-iodide electrolyte, with a particular focus on the deposition kinetics and Ag demixing during deposition and annealing. Some aspects of the electrochemical behavior of the electrolyte are determined by performing cyclic voltammetry on two solutions, each containing one of the two metals together with its complexant. The presence of a real metastable alloy is demonstrated by means of transmission electron microscopy (TEM) coupled with the analysis of peak positions in X-Ray diffraction (XRD) spectra. In addition, the presence of Ag nano-precipitates is revealed by TEM observation.
2. Experimental Electrodeposition of Cu-Ag alloys was performed from a pyrophosphate-iodide electrolyte with the following composition: 50 g/L CuSO4 ∙ 5H2O, 150 g/L K4P2O7, 150 g/L KI, 75 mL/L 0.1M AgNO3, 10 g/L KNO3. Chemicals were purchased from Sigma Aldrich and used in the as-received condition. The pH of the obtained solution was 8.9. Electroplating was done in non-stirred conditions at 50°C, typical operating temperature for pyrophosphate copper solutions. Current densities between 2 mA/cm2 and 30 mA/cm2 were used, following the procedure described in a previous work [9]. The electrochemical cell consisted of a pure copper plate as anode and electrodeposited nickel as cathode. This nickel layer was deposited on carbon steel samples using an industrial solution of nickel sulfamate operated at a current density of 20 mA/cm2 and a temperature of 45°C, with moderate stirring. Steel plates were polished with emery paper to 1200-grit finish before nickel plating. The power supply system was a PSI 8065-05 T generator from ElektroAutomatik GmbH. To study the electrochemical properties of the electrolyte, two solutions were derived from the Cu-Ag bath and cyclic voltammetry experiments were performed on them. The first bath, namely AGSOL, contained only Ag and its complexing agent: 150 g/L KI, 75 ml/L 0.1M AgNO3. The second one, namely CUSOL, contained only Cu and its complexing agent: 50 g/L CuSO4, 150 g/L K4P2O7. Cyclic voltammetry measurements were carried out using an EG&G potentiostat-galvanostat model 273. The cell was in this case composed of glassy carbon (GC) as working electrode, a standard SCE as reference
3
electrode and graphite as counter electrode. Chemical composition of the Cu-Ag alloys was determined by X-Ray Fluorescence (XRF) using a Fischerscope X-Ray XAN instrument. Structural analysis was performed by XRD and diffraction patterns were acquired by a Philips 1830 appliance using Cu K radiation. Annealing of the Cu-Ag alloy deposited at 2 mA/cm2 was carried out in a Carbolite MTF 12/38/250 tubular furnace under nitrogen flow at 400 °C for 2 hours. A freestanding sample of Cu-4.4% Ag alloy was prepared for TEM analysis through a standard focused ion beam (FIB) lift-out procedure with a dual-beam FIB (FEI DB235); final thinning was completed at 5 keV with Ga ions. Selected area electron diffraction (SAED), lowmagnification imaging, and high resolution TEM (HRTEM) was completed with a JEOL 2100 LaB6 microscope equipped with a high-resolution pole-piece, operated at 200 keV. Hardness of the deposited alloys was measured by microindentation using a Fischerscope HCU instrument equipped with a diamond Vickers indenter.
3. Results and discussion 3.1 Electrochemistry of the Cu-Ag electrolyte The voltammetric investigation of the base electrolyte (containing both Cu and Ag) did not provide a good description of the deposition process because the superimposition of current peaks coming from different electrochemical reactions was observed in the most significant potential ranges [9]. For this reason, the electrochemical behavior of the deposition bath was investigated by voltammetry studies of the separate solutions (CUSOL and AGSOL). Cyclic voltammograms of the Cu pyrophosphate electrolyte (CUSOL) acquired with different scan rates are displayed in Fig. 1. The cathodic current starts to increase around -0.3 V and tends towards a plateau as the potential decreases, rather than exhibiting a distinct peak. It is reported that the predominant Cu pyrophosphate complex in alkaline solutions is Cu(P2O7)26- (CuL2), with minor presence of the CuP2O72- (CuL) complex [10, 11]. Equilibrium concentrations of Cu complexes and pyrophosphate ions (P2O74-) in the voltammetry electrolyte were evaluated by using equilibrium data reported in the literature [11, 12] and were found to be 0.2 M (CuL2), 0.15 mM (CuL) and 0.034 M (P2O74-).
4
In addition, the equilibrium potential for Cu reduction reaction resulted to be -0.267 V vs SCE, which is consistent with the value of -0.3 V observed in the voltammogram. According to the model proposed by Konno and Nagayama [11], Cu reduction involves two steps, dissociation of a ligand (reaction 1) and electrochemical reduction of the CuL complex, for [P2O74-] < 0.09 M (reaction 2). Cu(P2O7)26- ↔ CuP2O72- + P2O74-
(1)
CuP2O72- + 2e- Cu + P2O74-
(2)
The cathodic current given by reaction (2) is expected to be limited by the low concentration of the CuL complex (0.15 mM). Under these conditions, a current peak is not discernible in the voltammogram and the magnitude of the current increase is independent of the scan rate [13], as observed in Fig. 1. Pyrophosphate ions would be adsorbed on the electrode surface, unless the electrode potential becomes more negative than a critical value that depends on the solution composition [11]. For the voltammetry electrolyte, the estimated critical potential should be about -0.7 V vs SCE. Below this value, deposition inhibition due to pyrophosphate adsorption would vanish and the cathodic current might increase. Actually, the voltammogram in Fig. 1 shows that the cathodic current rises at -0.9 V and increases further as the potential decreases. Nevertheless, this feature can also be related to hydrogen discharge, which can contribute significantly to the cathodic current in this potential range [14]. In the reverse scan, Cu dissolves showing distinct anodic peaks above -0.3 V (Fig. 1). Cu-Ag alloys were deposited from the pyrophosphate–iodide electrolytes at cathode potentials ranging from -0.85 to -0.65 V vs SCE, where reactions (1) and (2) can describe the Cu reduction process. Cyclic voltammograms obtained from the Ag iodide solution at different scan rates are shown in Fig. 2. For each scan rate, three current peaks are observed in the cathodic sweep. The first peak occurs around -0.1 V and can be assigned to the reduction of I3- to I- [15, 16]. The other two peaks (A, B) arise in the potential range of -0.6 to -1.3 V and are related to Ag reduction reactions in iodide solutions [17, 18]. The more
5
intense current peak (A) occurs in the potential range where Cu-Ag alloys were deposited (-0.85 to -0.65 V). In the voltammogram, anodic peaks of Ag dissolution are identified in the potential range of -0.4 to -0.2 V (Fig. 2). A variety of Ag complexes can form in iodide solutions and their equilibrium concentration in the voltammetry electrolyte was calculated by using the relevant equilibrium constants [19], assuming that activity coefficients of ionic species are unity. The equilibrium concentrations are listed in Table 1. The species with the highest concentrations are AgI43- (4.73 mM) and AgI32- (1.23 mM), the other complex ions (Ag2I64-, Ag3I85-, AgI2-) amount altogether to 0.64 mM. It is noted in Fig. 2 that current peaks A and B are broad and asymmetric and accurate evaluation of their position and width would require mathematical processing. The operation of semi-differentiation could be used to convert such broad peaks in well-defined symmetric peaks [20,21], whose parameters (position, width, and height) can be estimated by employing the sech2 fitting function [22]. Semi-derivatives of the current peaks A and B in Fig. 2 have been calculated and the estimated fitting parameters are reported in Table 2. It is seen that the semi-differentiated peaks are still very broad and their width increases as the scan rate rises. This behavior is typical of electrolytic solutions containing quasi-labile metal complexes, which undergo ligand exchange reactions characterized by rate constants neither very small (inert complexes) nor very large (labile complexes) [23]. The interpretation of the electrodeposition process is very difficult, because all the silver complexes existing in the electrolyte may be involved in the cathodic reactions and many unknown kinetic parameters will play a decisive role in determining the overall behavior of a so complicated system. Nevertheless a qualitative explanation of the occurrence of peaks A and B in the voltammogram can be proposed. According to the criterion that complexes having a smaller number of ligands more likely contribute to metal electrodeposition [24], AgI2- should sustain the dominant cathodic reaction, as suggested in another study [25]. Yet, this complex ion is present with very low concentration (Table 1), therefore its discharge reaction should be supported by fast chemical dissociation of complex species
6
containing a larger number of ligands [24]. Taking into account the quasi-labile nature of complexes in our electrolyte, this situation is unlikely to establish. The most abundant complex species (Table 1) are AgI43- and AgI32-, that can participate in the reduction reactions (3) and (4): AgI43- + e- Ag + 4I-
(3)
AgI32- + e- Ag + 3I-
(4)
Kinetics of these reactions can be described by means of their exchange current density [26,24] and the values of this parameter for the two reactions can differ significantly from one another. It was shown that the position of voltammetric cathodic peaks shifts to lower potential values as the exchange current density decreases [26]. Hence, the appearance of peak B in the voltammogram (Fig. 2) may be associated with the cathodic reaction of the electroactive species characterized by the lower exchange current density and the higher overvoltage values. Another important effect is related to ligand exchange reactions that occur at the cathode surface as Ag reduction reactions proceed. In particular, the relative quantity of mononuclear (AgI43-, AgI32-, AgI2-, AgI) and polynuclear (Ag2I64-, Ag3I85-) complexes is dependent on the total Ag content. Under equilibrium conditions, the relative amount of polynuclear complexes, i.e. ([Ag2I64-]+[ Ag3I85-])/[Ag]total, decreases as the total Ag content in the solution becomes smaller (square brackets represent the molar concentration of the included species). As reactions (3) and (4) proceed, the concentration of AgI43- and AgI32- in the solution near the cathode will decrease, together with the total Ag content, therefore the equilibrium concentration of the polynuclear complexes Ag2I64- and Ag3I85- will diminish greatly and become smaller than the existing value. Consequently, ligand exchange reactions like the following ones may take place: Ag2I64- + 2I- 2AgI43-
(5)
Ag3I85- + 3I- AgI32- + 2AgI43-
(6)
7
Since quasi-labile complexes are involved, these reactions will take a finite time to produce an appreciable increase in the concentration of AgI43- and AgI32-. This supplementary provision of electroactive species may contribute to develop a second current peak after some time, i.e. at more negative potential values (peak B in Fig. 2). Because of the high stability of Ag iodide complexes, the equilibrium concentration of free Ag ions (Ag+) in the electrolyte is as low as 3.3·10-17 M (Table 1) and causes the equilibrium potential for Ag reduction to shift down to -0.509 V vs SCE. Accordingly, Cu-Ag alloys were deposited at potential values lower than -0.65 V vs SCE. 3.2 Cu-Ag microstructure: XRD analysis Cu-Ag alloys with varying Ag content were prepared by changing the current density in the deposition cell. The overall Ag content decreased as the current density increased and an atomic percentage as high as 15 % was attained at 2 mA/cm2 (Table 3). These samples gave the opportunity to analyze the metastable behavior of Cu-Ag alloys in a wide composition range, contrary to other studies reported in the literature [2,27]. The structure of the produced alloys was examined by XRD and the diffraction patterns are reported in Fig. 3. Pure copper deposited from a standard pyrophosphate solution was used as a reference sample and its XRD spectrum is shown in Fig. 3. Both copper and silver have face-centered cubic structure, but the lattice parameters are different. In the XRD patterns of all the Cu-Ag alloys, the reflections of copper are identified. One additional peak appears for 2 approximately equal to 38° in the spectra of the two alloys with higher Ag content. This peak corresponds to the (111) reflection of silver and indicates that a second phase rich in silver is present in these two alloys, together with the Cu rich matrix. Because the intensity of the Ag (111) peak is relatively low, it is argued that the other Ag reflections are too weak to be distinguished from the background signal. The interplanar spacing of Cu (111) planes was calculated from the angular position of the (111) reflection in the XRD patterns. The results are reported in Table 4, together with the difference between the
8
interplanar distance in the alloy and that in pure copper (dalloy - dCu). It is noted that the interplanar spacing increases as the silver content rises, which demonstrates that a solid solution of Cu and Ag exists in the electrodeposited alloys. According to Vegard’s law [28], the lattice parameter of a binary alloys depends linearly on the content of the solute element, but deviations from this law have been observed in many systems. An alternative model developed by Lubarda includes an apparent size of the solute atom to account for the electronic interactions between the solute and solvent atoms [29]. Experimental values of the lattice parameter of splat-cooled Cu-Ag alloys are also available [30]. The variation of the interplanar spacing (dalloy – dCu) is plotted as a function of the Ag atomic fraction in Fig. 4. The results of this work are compared with the data reported by Predel et al. [30] and the predictions of the Vegard’s law and Lubarda’s model [29]. The alloys with higher Ag content (12.9 at.% and 15.4 at.%) exhibit interplanar distances significantly lower than the values of single-phase Cu-Ag alloys with the same overall composition. This is consistent with the presence of the (111) Ag reflection in the XRD patterns and confirms that silver demixing occurs in these alloys. In the XRD spectrum of the Cu-12.9% Ag alloy, the Ag (111) peak is sufficiently well defined to attempt an evaluation of the composition of the two phases. The angular position of this peak (2 = 38.45°) corresponds to an interplanar spacing of 2.339 Å, which means that the Ag rich precipitate would contain a percentage of Cu comprised between 5.4 at.% (Vegard’s law) and 7.0 at.% (ref. [30]). Similarly, the Ag content of the Cu rich matrix can be deduced from the angular position of the Cu (111) reflection; it would be in the range of 5.1 at.% (ref. [30]) to 6.3 at.% (Vegard’s law). Because the reciprocal solubility of Cu and Ag at the deposition temperature (50 °C) and below is negligible [31], the system consists of two supersaturated solid solutions, i.e. silver demixing is not complete. The full width at half maximum of the Ag (111) peak indicates that the average crystallite size of the Ag precipitate should be around 6 nm, according to the Scherrer’s equation. The same reasoning applied to the Cu-15.4% Ag alloy led to estimate the silver percentage in the Cu rich matrix as a value comprised between 6.6 at.% (ref. [30]) and 8.1 at.% (Vegard’s law).
9
As expected, the annealing treatment of the alloy with the highest Ag content (15.4 at.%) enhanced the precipitation of silver, as shown in Fig. 5, where the XRD patterns of the annealed sample and the asdeposited one are compared. It is noted that all the Ag reflections are clearly recognized in the spectrum of the annealed alloy and the peaks are significantly narrower than those of the as-deposited sample. The average crystallite size in the annealed alloy was estimated by the Scherrer’s equation and was found to be 43 nm. For the electrodeposited alloys with low Ag content (up to 4.4 at.%), XRD patterns do not display any reflection demonstrating the existence of a second phase in addition to the Cu rich matrix. Nevertheless, the presence of a Ag rich precipitate cannot be excluded because the XRD method is not sensitive to small fractions of a second phase, especially if it is present in the form of nano-sized particles. Actually, even for low Ag content alloys (3.2-4.4% Ag) observed interplanar spacings are smaller than expected based on Vegard’s law or Lubarda’s model. This result suggests that excess Ag might be hosted in a second phase. 3.3 Cu-Ag microstructure: TEM analysis
To resolve the discrepancy between expected and measured lattice parameters for the low at. % Ag alloys, a Cu-4.4% Ag sample was analyzed by TEM. Figure 6A shows a SAED pattern of the alloy, demonstrating the samples polycrystalline nature. All of the major reflections correspond to Cu; however, at higher magnification, reflections matching expected Ag lattice spacings are observed (Figure 6B). For each strongly diffraction Cu (022) reflection, there is a much weaker and more diffuse reflection corresponding to the Ag (022) reflection. This reciprocal space relationship between Cu and Ag reflections indicates that Ag rich nanoparticles exist as coherent precipitates within larger Cu rich grains. Figure 6C shows a radially averaged intensity profile of the SAED pattern shown in Figures 6A and 6B. The weak Ag (022) reflections marked in Figure 6B produce a noticeable, albeit broad, peak around 7 nm-1. HRTEM analysis of Ag precipitates demonstrates that the precipitates are only a few nm in diameter (Figure 6D and 6E), explaining their absence from XRD measurements. 10
TEM results showed that the assessment of the interplanar distance in the Cu matrix is a reliable criterion for inferring the existence of Ag precipitates (Fig. 4). According to this criterion, Cu-Ag alloys with the lowest Ag percentages (3.2 and 3.5 at.%) are likely to have a very small fraction of Ag nano-precipitates similar to those observed in the Cu-4.4% Ag sample. Hardness values measured by microindentation are reported in Fig. 7. Alloys with low Ag content (3.5 at.% and 4.4 at.%) were found to be significantly harder than pure copper. Such hardness enhancement is to be ascribed to solid-solution strenghtening, with a minor contribution from precipitation hardening given by the small fraction of silver precipitates shown by TEM analysis. Further increase in hardness was observed in high-Ag alloys (12.9 at.% and 15.4 at.%). According to the XRD analysis, these alloys are characterized by higher Ag contents in the Cu matrix (from 5 at.% to 8 at.%) and detectable fractions of Ag rich precipitates. As a result, both solid-solution hardening and precipitation hardening become more intense, thus justifying the high hardness values displayed in Fig.7. In particular, hardness as high as 628 HV was attained in the Cu15.4% Ag alloy (the Vickers hardness HV is measured in kgf/mm2; 1 kgf/mm2 = 9.807 MPa). During the annealing treatment, silver was rejected from the Cu matrix and caused solid-solution hardening to vanish and precipitation hardening to intensify. Because hardness decreased notably after annealing, solidsolution hardening appears to contribute to alloy strengthening more than precipitation hardening. Actually, the annealed alloy was found to be approximately as hard as the Cu-3.5% Ag alloy, whose silver content is largely lower. However, all the alloys were notably harder than electrodeposited pure Cu.
4. Conclusions Voltammetric characterization of the Cu-Ag pyrophosphate-iodide electrolyte was carried out analyzing two separate solutions containing the two metals. Reduction of Cu was found to take place in two steps: dissociation of the CuL2 ligand and electrochemical reduction of the CuL species. Conversely, Ag deposition mainly occurs as consequence of the reduction of mononuclear complexes (AgI43-, AgI32-) to elemental Ag. Cu-Ag alloys with an Ag content ranging from 3 to 15 at.% were prepared by electrodeposition and characterized by TEM and XRD. A supersaturated Cu-Ag solid solution was found in all the alloys, while Ag rich precipitates were detected by XRD only in alloys with high silver content (12.9 at.% or more). It was
11
demonstrated, however, that Ag demixing was incomplete, as the Cu matrix remained significantly supersaturated with Ag. TEM analysis revealed the presence of nano-sized Ag precipitates which were not detected in the XRD pattern in a 4.4 at.% Ag sample, in agreement with matrix supersaturation predicted by XRD analysis. These TEM results, coupled with a comparison of predicted and measured lattice parameter values, indicate that all low at.% Ag samples studied here contain Ag nano-precipitates. Addition of Ag to Cu caused the hardness to increase notably, even at low Ag content. Comparison between as-deposited and annealed samples showed that solid-solution hardening was more effective than precipitation strengthening in increasing the alloy hardness. Acknowledgments Authors J.L.H., A.C.L., and M.L.T. acknowledge support from the Office of Naval Research under contract number N00014-14-1-0058. Authors M.L.T. and J.L.H. also acknowledge the Engineers as Global Leaders in Energy and Sustainability (EAGLES) project, supported by the FIPSE program of the US Department of Education, and the National Science Foundation under grant number 1429661. References [1] J.R. Lloyd, J.J. Clement, Electromigration in copper conductors, Thin Solid Films 262 (1995) 135. [2] S. Strehle, S. Menzel, J.W. Bartha, K. Wetzig, Electroplating of Cu(Ag) thin films for interconnect applications, Microelectron. Eng. 87 (2010) 180. [3] R. Najafabadi, D. J. Srolovitz, E. Ma, M. Atzmon, Thermodynamic properties of metastable Ag-Cu alloys, J. Appl. Phys. 74 (1993) 3144. [4] K. Uenishi, K. F. Kobayashi, K. N. Ishihara, P. H. Shingu, Formation of a super-saturated solid solution in the Ag-Cu system by mechanical alloying, Mater. Sci. Eng. A 134A (1991) 1342. [5] M. J. Kim, H. J. Lee, S. H. Yong, O. J. Kwon, S. K. Kim and J. J. Kim, Facile formation of Cu-Ag film by electrodeposition for the oxidation-resistive metal interconnect, J. Electrochem. Soc. 159 (2012) D253.
12
[6] M. J. Kim, S. H. Yong, H. S. Ko, T. Lim, K. J. Park, O. J. Kwon and J. J. Kim, Superfilling of Cu-Ag using electrodeposition in cyanide-based electrolyte, J. Electrochem. Soc. 159 (2012) D656. [7] J. Song, H. Li, J. Li, S. Wang, S. Zhou, Fabrication and optical properties of metastable Cu-Ag alloys, Appl. Opt. 41 (2002) 5413. [8] S. Strehle, J.W. Bartha, K. Wetzig, Electrical properties of electroplated Cu(Ag) thin films, Thin Solid Films 517 (2009) 3320. [9] R. Bernasconi, L. Nobili, L. Magagnin, Electrodeposition of supersaturated CuAg alloys in pyrophosphate-iodide electrolytes, ECS Trans. 58(32) (2014) 53. [10] G. W. Higgins, P. E. Sturrock, Polarographic Irreversibility of the Copper(II) Pyrophosphate System, Anal. Chem. 41 (1969) 633. [11] H. Konno, M. Nagayama, Mechanism of Electrodeposition of Copper from Cupric Pyrophosphate Solutions, Electrochim. Acta 22 (1977) 353. [12] O. E. Schupp III, P. E. Sturrock, J. I. Watters, A Study of the Stability and Basicity of the Copper(I1) Pyrophosphate Complexes Using the Dropping Amalgam Electrode, Inorg. Chem. 2 (1963) 106. [13] R. S. Nicholson, I. Shain, Theory of Stationary Electrode Polarography, Anal. Chem. 36 (1964) 706. [14] K. Johannsen, D. Page, S. Roy, A systematic investigation of current efficiency during brass deposition from a pyrophosphate electrolyte using RDE, RCE, and QCM, Electrochim. Acta 45 (2000) 3691. [15] J.-Y. Kim, J.-K. Lee, S.-B. Han, Y.-W. Lee, K.-W. Park, Improved Tri-iodide Reduction Reaction of CoTMPP/C as a Non-Pt Counter Electrode in Dye-Sensitized Solar Cells, J. Electrochem. Sci. Tech. 1 (2010) 75. [16] Y. A. Yaraliyev, Oxidation of iodide ions by means of cyclic voltammetry, Electrochim. Acta 29 (1984) 1213.
13
[17] V. Venkatasamy, S. Riemer, I. Tabakovic, Electrodeposition of eutectic Sn96.5Ag3.5 films from iodide– pyrophosphate solution, Electrochim. Acta 56 (2011) 4834. [18] H. H. Hassan, M. A. M. Ibrahim, S. S. Abd El Rehim and M. A. Amin, Comparative Studies of the Electrochemical Behavior of Silver Electrode in Chloride, Bromide and Iodide Aqueous Solutions, Int. J. Electrochem. Sci. 5 (2010) 278. [19] J. N. Butler, Ionic Equilibrium, Addison-Wesley Publishing Company, Massachusetts, 1964. [20] M. Grenness, K. B. Oldham, Semiintegral Electroanalysis: Theory and Verification, Anal. Chem. 44 (1972) 1121. [21] J. J. Toman, S. D. Brown, Peak Resolution by Semiderivative Voltammetry, Anal. Chem. 53 (1981) 1497. [22] D. M. Caster, J. J. Toman, S. D. Brown, Curve Fitting of Semiderivative Linear Scan Voltammetric Responses: Effect of Reaction Reversibility, Anal. Chem. 55 (1983) 2143. [23] J. J. Toman, R. M. Corn, S. D. Brown, Convolution Voltammetry of Metal Complexes, Anal. Chim. Acta, 123 (1981) 187. [24] A.G. Zelinsky, B.Ya. Pirogov, Numerical simulation and experiment in the systems of labile complexes of metals. Silver reduction from the thiourea electrolyte, J. Electroanalyt. Chem. 694 (2013) 68. [25] O. A. Ashiru, J. P. G. Farr, Kinetics of Reduction of Silver Complexes at a Rotating Disk Electrode, J. Electrochem. Soc. 139 (1992) 2806. [26] A. Survila, P. V. Stasiukaitis, Linear potential sweep voltammetry of electroreduction of labile metal complexes-I. Background model, EIectrochim. Acta 42 (1997) 1113. [27] S. Strehle, S. Menzel, K. Wetzig, J.W. Bartha, Microstructure of electroplated Cu(Ag) alloy thin films, Thin Solid Films 519 (2011) 3522. [28] L. Vegard, Die Konstitution der Mischkristallen und die Raumfüllung der Atome, Z. fur Physik 5 (1921) 17.
14
[29] V. A. Lubarda, On the effective lattice parameter of binary alloys, Mech. Mater. 35 (2003) 53. [30] B. Predel, G. Schluckebier, Investigations of the Structure, Thermodynamic Properties and Kinetics of Decomposition of Metastable Silver Copper Solid Solutions Produced by Splat Cooling, Z. Metallkd. 63 (1972) 782. [31] J. L. Murray, Calculations of Stable and Metastable Equilibrium Diagrams of the Ag-Cu and Cd-Zn systems, Metall. Trans. A 15A (1984) 261. Captions to illustrations Figure 1 Cyclic voltammograms of the Cu pyrophosphate solution (CUSOL) obtained with scan rates ranging from 5 to 20 mV s-1. Cathodic currents are plotted along the negative direction. Figure 2 Cyclic voltammograms of the Ag iodide solution (AGSOL) obtained with scan rates ranging from 10 to 50 mV s-1. Cathodic currents are plotted along the negative direction. Figure 3 XRD patterns of Cu-Ag alloys under various plating conditions; spectrum of pure Cu is reported for comparison. Figure 4 Experimental lattice expansion compared with data in the literature. Figure 5 XRD analysis of the sample deposited at 2 mA/cm2 (15.4 at.% Ag) before and after annealing at 400°C for 2 h. Figure 6
15
TEM analysis of the Cu-4.4% Ag sample. A)SAED pattern showing polycrystalline nature of sample. B) Magnified version of A) showing Ag and Cu (022) spots. C) Radially averaged intensity profile of SAED shown in A) and B) alongside expected peaks for Cu and Ag; the broad peak at ≈7 nm-1 corresponding to the Ag (022) reflection. D) and E) HRTEM images showing Ag nanoparticles several nm in diameter. Black arrows indicate the location of a Ag precipitate in each image; however, both images contain many unmarked precipitates. Figure 7 Vickers hardness of as-deposited and annealed Cu-Ag samples.
16
Fig.1
17
Fig.2
18
Fig.3
19
Fig.4
20
Fig.5
21
Fig.6
22
Fig.7
23
Table 1 – Equilibrium concentration of silver-iodide species in the voltammetry electrolyte. ________________________________________________________________________________ Species
Ag+
Concentration 3.3·10-17
AgI(aq) AgI2-
AgI32-
AgI43-
Ag2I64- Ag3I85- I-
3.9·10-9 2.2·10-6 1.2·10-3 4.7·10-3 3.6·10-4 2.7·10-4 8.7·10-1 mol dm-3
________________________________________________________________________________
24
Table 2. Position, height and width of semiderivative peaks calculated from current peaks A and B in Fig. 2. Coefficients of determination R2 are also reported. ________________________________________________________________________________ Peak A
Peak B
________________________________________________________________________________ Scan rate/mV s-1
10
20
50
10
50 Position/V vs SCE
-0.786 -0.819 -0.876
-0.943 -1.029 -1.183
Height/A cm-2 V-1/2
0.0259 0.0276 0.0312
0.0045 0.0083 0.0128
Widtha/V
0.143 0.180 0.248
0.048 0.085 0.157
R2
0.973 0.979 0.981
0.973 0.979 0.981
______________________________________________________________________________ a
Full Width Half Maximum
25
20
Table 3 – Ag content in Cu-Ag alloys at different current densities. Current density Ag (at.%)
j/mA cm-2 30
3.2
20
3.5
10
4.4
5
12.9
2
15.4
Table 4. Interplanar spacing of (111) planes in the Cu matrix and deviation from the value of pure Cu (dalloy – dCu).
Ag (at.%)
Peak position 2θ/°
d111 spacing d/Å
dalloy - dCu d/Å
0
43.43
2.082
0
3.2
43.27
2.089
0.007
3.5
43.25
2.090
0.008
4.4
43.21
2.092
0.010
12.9
43.06
2.099
0.017
15.4
42.95
2.104
0.022
26