Journal of The Electrochemical Society, 154 共1兲 D13-D20 共2007兲
D13
0013-4651/2006/154共1兲/D13/8/$20.00 © The Electrochemical Society
A Model for Copper Deposition in the Damascene Process Application to the Aging of the Deposition Bath C. Gabrielli,a,* P. Moçotéguy,a H. Perrot,a,* A. Zdunek,b,* and D. Nieto Sanzc a
UPR 15 CNRS, LISE, Université de Pierre et Marie Curie, Paris, France Air Liquide, Chicago Research Center, Countryside, Illinois 60525, USA c Air Liquide, Centre de Recherche Claude Delorme, 78354 Jouy en Josas, France b
Previously, a model of copper deposition in the damascene process was proposed that takes into account the existence of a copper-accelerator complex as well as the competitive adsorption of an inhibitor and an accelerator. The model was verified experimentally by electrochemical impedance spectroscopy and dc voltammetry measurements on fresh solutions of a superfilling, copper deposition plating bath. In this paper, the same model was applied to plating baths during copper deposition where aging and possible relaxation of the bath back to a steady state can occur. The model can provide insight into interpreting the impedance spectra during and after copper deposition. Changes in the Cu-accelerator complex in the bulk solution and the surface coverage of this complex were followed using a parameter based on the low-frequency inductive loop of the electrochemical impedance spectra. Results indicate that the impedance spectroscopy measurements are mainly sensitive to Cu-accelerator complex formation and destruction. © 2006 The Electrochemical Society. 关DOI: 10.1149/1.2364831兴 All rights reserved. Manuscript submitted January 23, 2006; revised manuscript received June 23, 2006. Available electronically November 14, 2006.
The damascene process for fabrication of copper on-chip metal interconnects requires electrodeposition into trenches or vias with width dimensions on the order of 130 nm or lower. To obtain voidfree deposits, superconformal deposition, or superfilling, is necessary. These terms refer to the occurrence of more rapid electrodeposition in the bottom of a trench or via than toward its entrance. Copper electroplating baths usually contain a mixture of H2SO4 and CuSO4 in concentrations close to 1.8 and 0.25 M, respectively, to which chloride ions are added in the range of 1 to 2 mM. However, superfilling is only obtained in the presence of a certain combination of additives in the electroplating bath,1,2 including brighteners/ accelerators, carriers/suppressors, and levellers. Brighteners are usually propane sulfonic acid derivatives, either MPSA 共HSO3–共CH2兲3–SH兲 or SPS 关HSO3–共CH2兲3–S–S–共CH2兲3–SO3H兴. They change the nucleation process by providing growth sites and accelerate the charge-transfer process at the copper interface. Suppressors are often polyalkylene glycol 关HO–共CkH2k–O兲n–H兴. They adsorb evenly at the wafer surface and change the structure of the deposit and increase the overpotential. Accelerators diffuse more easily than inhibitors towards the bottom of the trenches, whereas the inhibitors stay at the mouth, allowing the trenches to be filled from the bottom to the top without voids. Levellers are used to decrease the growth rate in regions of high mass transfer rates, thereby limiting the thickness of the copper deposit above trenches and vias.2 Electrochemical impedance spectroscopy 共EIS兲 has been used to investigate copper deposition without additives, and has revealed, in addition to the expected charge transfer resistance, low frequency relaxations3,4 which have been ascribed to various processes. The low-frequency capacitive features have been interpreted as diffusion of Cu2+ from the solution and reaction intermediate coverage relaxations. The inductive features have been interpreted as the relaxation of the surface concentration of adatoms5 and an activation of the electrode area with increasing potentials. However, it has not been possible to make a clear distinction between a slow increase of the surface area due to the nucleation and the development of growth centers6 and a slow removal of inhibiting species such as anionic species, hydroxides, and organic molecules.4 A previous thorough investigation of copper deposition7 has already shown that an electrode activation process occurs with increasing current density and that the low-frequency features appear to be strongly dependent on the growth mode of deposit. It has been shown that the addition of small amounts of chloride
* Electrochemical Society Active Member.
ions into the copper bath solution modifies the electrode kinetics. Chloride ion addition in a copper sulfate bath stabilizes the cuprous ions by forming complexes, which are dependent on the respective concentrations of chloride and cuprous ions.8 However, due to the low amount of chloride ions added to a typical copper electroplating bath, the only complex that is formed is the poorly soluble CuCl Cu+ + Cl− CuClsolid
关1兴
where KCuCl = Cu+ Cl− = 1.72·10−7M2. In practice, the amount of chloride ions added to the plating baths are adjusted so that CuCl is not formed in the bulk bath. However, chloride ions adsorb at the copper surface above the potential of zero charge 共pzc兲, which is negative to the open-circuit potential 共OCP兲, even at amounts as low as 1 ppm.9,10 When the CuCl film coverage is low, chloride ions enhance copper deposition/ dissolution rate because, above the pzc, they facilitate the access of Cu2+ to the electrode by bridging it to the metal. Soares et al.8 have shown that the formation of CuCl opens a parallel mechanism of Cu2+ reduction to Cu0. In a previous paper by Gabrielli et al.,11 the role of chlorides in the copper deposition mechanism was investigated by kinetic methods based on impedance measurement techniques in acidic copper sulfate baths without organic additives. It was shown that chloride ions adsorb on the copper electrode to form CuCl, which partially blocks the electrode surface, and that this adsorbed CuCl is further reduced to copper metal. This path, via CuCl, is preponderant for copper deposition over the two-step mechanism through Cu+. In addition, at sufficiently high current densities, the model showed that the surface coverage of CuCl is dominant, which supports the major role of the CuCl deposition path. For lower current densities, the effect of an anionic adsorbed species had to be taken into account to fit the model predictions with experimental results—an inductive loop is observed in the lowest frequency range. Finally, it has been shown that the experimentally observed mass transport limitations were due to the diffusion of cupric ions. The demands of copper electrodeposition in the copper interconnect damascene process require a thorough knowledge of the reaction mechanism underlying this phenomenon. The aim of this paper was twofold. First, to use a model that takes into account the effect the organic additives and compares it to experimental results obtained through electrochemical impedance spectroscopy on experiments carried out over long periods of time. Previously, this model was used to compare experimental results for copper deposition in fresh electroplating baths containing various quantities of polyethylene glycol 共PEG兲 as an inhibitor and either MPSA or SPS as the accelerator.12 Then, after successfully testing by changing the concentrations of the electroactive species, in this paper, this reaction
Journal of The Electrochemical Society, 154 共1兲 D13-D20 共2007兲
D14
mechanism is tested for aging of the deposition bath with MPSA as a brightener and PEG as the inhibitor. Second, from the results obtained, we shall try to find practical parameters to characterize the degree of degradation or consumption of the organic additives occurring when the plating bath is aged. Theory Cathodic deposition of copper in sulfuric acid solutions generally occurs by two consecutive charge transfer steps involving the soluble intermediate Cu+ 13-17 k1
Cu2+ + 共ⴱ兲 + e− ——→ Cu+共ⴱ兲
E01 = − 0.493 V vs SMSE 关2兴
as reported in Ref. 17 k2
Cu+共ⴱ兲 + e− ——→ Cu
E02 = − 0.133 V vs SMSE
关3兴
as reported in Ref. 17, where SMSE is the saturated mercurous sulfate reference electrode. If chloride ions are present in the sulfuric acid solution, the twostep reaction pathway with adsorbed CuCl as the intermediate also has to be taken into account 2+
Cu
k3
+ Cl + e + 共ⴱ兲 ——→ CuCl共ⴱ兲 −
−
E0 = − 0.112 V vs SMSE
关4兴
according to Ref. 8 k4
CuCl共ⴱ兲 + e− ——→ Cu0 + Cl− + 共ⴱ兲
E0 = − 0.513 V vs SMSE 关5兴
according to Ref. 8. Moreover, Healy et al.10,18 have shown that, above the pzc, chloride ions bridge the CuI intermediates to PEG molecules, creating a CuI–Cl-PEG complex. This complex forms a film 共⬃1 nm兲 at the copper surface that hinders Cu2+ transport and slows the copper deposition rate. Close to and negative to pzc, Cl− will no longer adsorb and the CuI–Cl-PEG complex is dislodged. Because chloride desorption is totally irreversible, PEG loses its ability to inhibit copper deposition as soon as the deposition potential is decreased below pzc. It will recover this ability when the potential is again increased above the pzc. A decrease in the chloride ion concentration decreases the amount of adsorbed chloride at copper surface and thus weakens the PEG adsorption. From these results the following reactions are proposed k5
PEG + CuCl共ⴱ兲 ——→ PEG-Cl-CuI共ⴱ兲 k6
PEG − Cl − CuI共ⴱ兲 + e− ——→ Cu0 + PEG * + Cl−
关6兴 关7兴
where PEG* means a degraded form of PEG with a lower molecular mass due to the break of the chains of the polymer. It was previously shown that the superfilling behavior of standard copper electroplating baths was due to the substitution of adsorbed inhibitors by accelerator species after an induction period, and subsequently reduction of copper with the adsorbed accelerator can occur.1,19-21 This behavior explains the hysteresis effect observed in cyclic voltammetry 共CV兲 measurements with the accelerator present, however, at present, the accelerator species that substitutes the adsorbed CuI–Cl-PEG complex has not yet been determined. Healy et al.22 have observed that a CuI-thiolate complex acts as a key intermediate for both SPS and MPSA decomposition and that, when no cupric ion is present or can be formed in the solution, a much slower decomposition occurs. They have proposed the following mechanism for complex formation Copper proportionation/ disproportionation reaction
2Cu+ Cu0 + Cu2+
关8兴
With the equilibrium constant K f,Cu+ =
关Cu+兴2 = 5.6 ⫻ 10−7 关Cu2+兴
关9兴
as reported in Ref. 2. Cuprous ions that are generated react with MPSA or SPS to produce a thiolate complex. Frank and Bard23 have demonstrated using HPLC, mass spectrometry and UV-visible spectroscopy that the ratio of MPS to Cu in the Cu-thiolate complex is 2:1 and confirmed that the oxidation state of copper is +1. Thus, the CuI-thiolate compound is either a complex between one CuI and one SPS molecule or between one CuI and two MPSA molecules. They have also determined that SPS, MPSA, and CuI-accelerator complex are oxidized to thiolsulfonate by the air: HSO3–共CH2兲3–SO–SO–共CH2兲3–SO3H. Furthermore, Moffat et al.24 have observed that MPSAcontaining solutions evolve within a few hours, even when no current passes, while SPS containing solutions undergo no changes. Moreover, these solutions exhibit the same electrochemical behavior as the SPS containing solutions. They suggested the following oxidative dimerization mechanism of MPSA to SPS, followed by cupric ion oxidation by dissolved oxygen gas 2MPSA + 2Cu2+ SPS + 2H+ + 2Cu+
关10兴
They have also observed that a new species was generated during the electrolysis of SPS containing solutions, and that when this aged bath was rested overnight without any current or any copper electrode in the solution, it exhibited the same electrochemical characteristics as a freshly made, SPS-containing solution. They assumed that MPSA was formed during the electrolysis and then oxidized back to SPS during the overnight rest period. They finally proposed the following mechanism Reaction between electrogenerated CuI species at the anode and SPS additive 4CuI + SPS → 2CuI共MPS−兲 + 2Cu2+
关11兴
Complex oxidation to MPSA according to the following reaction 4CuI共MPS−兲n + O2 + 共4 + 4n兲H+ → 4Cu2+ + 4nMPSA + 2H2O 关12兴 where the MPS− portion corresponds to the following ion: HSO3–共CH2兲3–S− However, the results of our aging experiments with MPSA and SPS containing baths suggest that the complexes present in each solution are different.25 In addition, in the case of thiourea and its dimer, Suarez and Olson26 have also suggested different CuI complexes. Finally, because the results of Frank and Bard23 have shown that n = 2, the following reactions can be suggested Cu+ + 2MPSA CuI共MPSA兲2
关13兴
with the equilibrium constant K f,CuI共MPSA兲2 =
关CuI共MPSA兲2兴 关Cu+兴关MPSA兴2
关14兴
and the oxidation reaction 4CuI共MPSA兲2 + O2 → 4Cu2+ + 8MPSA + 2H2O
关15兴
In these reactions, it is assumed that the sulfonic groups of the bonded MPSA molecules are in their acidic form. This might not be true, as sulfonic groups are strong acids, but the hypothesis does not affect the accelerator degradation/complexation mechanisms, since it is generally considered that these mechanisms implie the thiol bonds. In previous studies,25,27,28 it was shown that the plating bath evolves differently depending on what type of anode material is used for aging. As shown in Fig. 1, extracted from Ref. 28, when a pure copper anode was used during copper deposition up to
Journal of The Electrochemical Society, 154 共1兲 D13-D20 共2007兲
D15
Figure 1. Impedance spectra evolution during plating bath aging with Goodfellow, 99.99%⫹ 共a兲 and a metal sample phosphorized copper anode 共b兲 containing 0.024 wt % of P. Extracted from Ref. 28.
14 Ah L−1 共a兲, an inductive loop whose size increases with aging time is observed in the impedance diagram. On the other hand, this behavior was not observed when a phosphorized copper anode 共b兲 or a carbon anode was used.25 Copper anodes containing phosphorus are known to prevent CuI release into the plating bath by either favoring the direct two-electron copper reduction reaction or by trapping CuI in black films generated on the anode.17,29,30 It was also shown in the model that the inductive loop was directly associated with the surface coverage of the CuI-accelerator complex, which increases as its concentration in the bath increases, independent of the additive 共PEG, MPSA, or SPS兲 concentrations in the bath.12 Moreover, an increase in the size of the inductive loop was observed when either MPSA or SPS concentration was increased in a freshly made plating bath. The behavior observed with a pure copper anode is then ascribed to the concentration buildup of a CuI-accelerator complex in the bath. The model used here assumes that the CuI共MPSA兲2 complex is already formed in the bulk bath, as all of the compounds needed for its existence are simultaneously present in the plating bath. Indeed, according to Reaction 9, the concentration of free Cu+ can be calculated as to be equal to 3.74 ⫻ 10−4 M when 关Cu2+兴 = 0.25 M. This value is far above the accelerator concentration, typically about 10−5 M in copper electrodeposition baths. Thus, the complex is in equilibrium with its constituents, according to Reaction 13 and is always present in the plating bath even at concentrations so low that it cannot be detected. Finally, it is assumed that the complex CuI共MPSA兲2 adsorbs at the copper surface, according to k7
CuI共MPSA兲2 + 共ⴱ兲 ——→ CuI共MPSA兲2共ⴱ兲
From Reactions 2-7, 13, and 17, the following set of equations, which define the model proposed, can be deduced dCu+ dt
d2 = k 5c 2 1 − k 6 2 dt
关20兴
d3 = k7c3共1 − 1 − 2 − 3 − Cu+兲 − k83 dt
关21兴
关22兴
ZF−1共兲 =
⌬IF 共兲 = Rt−1 + F关共k4 − c0共k1 + k3c1兲兲T1共兲 ⌬E + 共k6 − c0共k1 + k3c1兲兲T2共兲 + 共k8 − c0共k1 + k3c1兲兲T3共兲兴 关23兴
where Rt−1 = F关c0共b1k1 + b3k3c1兲共1 − 1 − 2 − 3 − Cu+兲 + b2k2Cu+ + b 4k 4 1 + b 6k 6 2 + b 8k 8 3兴
关24兴
and
冉
冊
冊
k 7c 3 b 8k 8 3 b 6k 6 2 1+ + j + k6 j + k8 + k7c3 j + k8 + k7c3 k 5c 2 k 7c 3 k 5c 2 j + k3c0c1 + k4 + k5c2 + k3c0c1 + 1+ j + k6 j + k8 + k7c3 j + k6 K 1 + k 3c 0c 1
−
where c0, c1, c2, and c3 are the concentrations of Cu , Cl , PEG, and Cu共MPSA兲2, and 1, 2, and 3 are the surface coverages of CuCl, PEG–Cl–CuI, and Cu共MPSA兲2 on the electrode, respectively. The impedance ZF, was calculated according to Eq. 23. The complete development of the model and the details of the calculation are given in Ref. 12
关17兴
冉
关19兴
2+
where Cu0-MPSA means that the MPSA species are trapped in the copper deposit.
⌬1 共兲 = T1共兲 = ⌬E
d1 = k3c0c1共1 − 1 − 2 − 3 − Cu+兲 − 共k4 + k5c2兲1 dt
+ k 6 2 + k 8 3兴
It is also supposed that, once adsorbed, this complex can be reduced to copper metal according to CuI共MPSA兲2共ⴱ兲 + e− ——→ Cu0 − MPSA + MPSA
关18兴
IF = F关c0共k1 + k3c1兲共1 − 1 − 2 − 3 − Cu+兲 + k2Cu+ + k41
关16兴
k8
= k1c0共1 − 1 − 2 − 3 − Cu+兲 − k2Cu+
冉
冉
冊冊
关25兴
Journal of The Electrochemical Society, 154 共1兲 D13-D20 共2007兲
D16
− b 6k 6 2 + k 5c 2T 1 ⌬2 共兲 = T2共兲 = ⌬E j + k6
关26兴
b8k83 + k7c3共T1 + T2兲 ⌬3 共兲 = T3共兲 = − ⌬E j + k8 + k7c3
关27兴
The impedance can then be calculated for each frequency according to the following equation, which takes into account the double layer capacity, Cdl Z共兲 =
1 ZF−1 + jCdl
关28兴
This reaction mechanism has been already validated in terms of electrochemical impedance for copper deposition in baths of various compositions. Here, it is tested on aged solutions where the concentrations of the electroactive species are modified due to their degradation or consumption during copper deposition over long periods of time.
Figure 2. Influence of anode material on the evolution during plating bath aging of the size of inductive loop, DIL, observed in the impedance spectra. 共䊏兲 Cu0 anode and 共䊐兲 Cu–P anode.
Experimental A copper plating bath with the proven ability to provide superfilling in submicrometer cavities was used for the experiments.1,19 The specific concentrations of the plating bath are 关H2SO4兴 = 1.8 M, 关CuSO4兴 = 0.25 M, 关NaCl兴 = 10−3 M, 关PEG兴 = 88 ⫻ 10−6 M, and 关MPSA兴 = 10−5 M. The PEG used in the electrolyte had a molecular weight of 3400 g mol−1. This bath is hereafter defined as the plating bath in the following discussions. When copper is electrodeposited for a sufficiently long time, the plating bath becomes degraded or aged. This evolution of bath aging was followed during copper deposition by electrochemical impedance spectroscopy. The aging experiments were performed with a 0.7 L solution and a setup containing two independent electrical circuits.31 First, an aging circuit was used which consisted of a saturated mercurous sulfate electrode 共SMSE兲, an anode plate, and a 99.9% industrial-grade copper cathode acting as the deposition substrate 共cathode兲. The anode material was either pure copper metal 共Goodfellow, 99.99%兲 or a phosphorized copper anode from metal sample containing 0.024 wt % of phosphorus. The results regarding the effect of the nature of the anode material on bath aging that are presented in this paper were taken from Ref. 27 and 28. During the experiments, the copper was electrodeposited at a current of 1.22 A 共i.e., 25 mA cm−2兲. The evolution of the plating bath during aging was studied up to 4.28 A h L−1 for the pure copper anode. In addition, the relaxation of the aged plating bath after turning off the deposition current was also followed by EIS. Both the anode and cathode active areas were controlled by using a TFM Electromask green insulating resin 共supplied by Henkel兲 and fixed at 48.75 cm2. The experimental conditions were identical with those used in Ref. 25, 27, and 28. The EIS circuit used in the test setup contained an additional saturated mercurous sulfate electrode, an anode sheet whose material was identical to the one used in the anode aging circuit, and a rotating disk electrode rotating at 2000 rpm as the cathode. For all measurements, the working electrode consisted of a 5 mm diam disk 共Goodfellow, 99.99%⫹ copper rod embedded in an inert and insulating Presi Allylic Glass Fiber resin兲. The disk was polished with a 1200 grade SiC paper and rinsed with deionized water to clean the surface before each measurement, and had an active area of 0.2 cm2. The impedance spectra were measured using a Solartron 1250 frequency response analyzer. Data acquisition was performed using a software program designed by the CNRS/LISE laboratory. The spectra were acquired in a frequency range of 62.5 kHz down to 10 mHz, in galvanostatic mode at a 25 mA cm−2 average deposition current density. Results and Discussion The change of the kinetic behavior of copper deposition in aged plating baths was investigated by electrochemical impedance spec-
troscopy for two conditions—during bath aging occurring by copper deposition over long periods of time, and during the relaxation of the aged plating bath after switching off the deposition current. Plating bath aging, experimental results.— Figure 1 shows the evolution of the impedance diagram during plating bath aging, when the deposition current is applied to the substrate with a pure copper anode 共a兲 and a phosphorized copper anode 共b兲, respectively. The experimental impedance diagrams generally exhibit three capacitive loops and one inductive loop in the low-frequency range. Figure 1a shows that an inductive loop and a small third capacitive loop is formed when bath is aged with the pure copper anode. The size of the inductive loop increases with the aging charge amount up to 9 A h L−1. When the aging charge amount becomes too high, the size of the inductive loop decreases. However, as shown by Fig. 1b, when the bath is aged with a phosphorized copper anode, the inductive loop formation is quickly inhibited. The evolution observed in Fig. 1a at the beginning of the plating bath aging is similar to that observed when the concentration of the accelerator is increased in freshly made plating baths.12 However, the aged baths exhibit a small third capacitive loop that was not observed in fresh baths. Note that the two capacitive loops in the highest frequency range did not significantly change with bath aging, and the size of the loops appears to be independent of the anode material. To better quantify the size of the inductive loop, the following practical parameter is defined DIL = Max关Re共Z兲兴 − Re共Z10
mHz兲
Figure 2 presents the influence of anode material on the evolution of this parameter during bath aging with respect to the electric charge flowing through the aging circuit. Plating bath aging: model calculations.— The parameters used in the model were extracted from the experimental conditions or from the experimental results: c0 = 0.25 M, c1 = 10−3 M,  = 10−9 mol cm−2, c2 = 8.8 ⫻ 10−5 M, Cdl = 1.2 ⫻ 10−5 F. The double-layer capacitance, Cdl, was obtained from the impedance spectra of a freshly made plating bath containing 10−5 M of MPSA and using a proprietary software program, the value was calculated. In the model, the CuI共MPSA兲2 concentration, c3, is directly related to the MPSA concentration through the equilibrium constant defined by Eq. 14. The parameter c3, was used as a variable but, because the Kf,CuI共MPSA兲2 value is not known, the concentration of CuI共MPSA兲2 complex in the bulk is also unknown. However, according to Reaction 13, CuI共MPSA兲2 cannot be higher than the half the initial concentration of MPSA, which would correspond to a complete forward reaction. To determine the best values of the other parameters, the
Journal of The Electrochemical Society, 154 共1兲 D13-D20 共2007兲
D17
Figure 3. Influence of CuI共MPSA兲2 complex concentration for a nominal PEG concentration 共a兲 and PEG concentration when CuI共MPSA兲2 complex concentration is fixed at 2.5 ⫻ 10−6 M 共b兲 on calculated impedance spectra in the plating bath. The frequencies are given in Hz.
calculated spectra were compared with the experimental results. The values of the parameters which lead to the same shape, frequency repartition, and evolution of the impedance calculated from the model as the the impedances measured on the experimental baths were found by a trial and error technique. The best set of parameters was the following: k01 = 3 ⫻ 103 cm s−1; b1 = −10 V−1; k02 = 5 ⫻ 104 cm s−1; b2 = −10 V−1; k03 = 2 ⫻ 103 cm s−1; b3 = −16 V−1; k04 = 0.049 cm s−1; b4 = −10 V−1; k5 = 4 ⫻ 103 cm s−1; k06 = 0.8 cm s−1; b6 = −6 V−1; k7 = 1.7 ⫻ 104 cm s−1; k08 = 2 −5 −1 −1 ⫻ 10 cm s ; b8 = −20 V . The given value of k7 was chosen to fulfill the requirement that the value of c3 would be lower than half the initial concentration of MPSA for all the experimental results. Indeed, in the model, the kinetic constant, k7, and complex concentration, c3, are always associated with each other through the product, k7c3. Hence, none of these values are separately known in the model Figure 3a shows the influence of the CuI共MPSA兲2 complex concentration, c3, in the plating bath, on the calculated impedance spectra, Z共兲, for a bath with a nominal PEG concentration of 88 M. When the copper-complex concentration was increased in the model, the calculated impedance spectra exhibited an increase in the size of the inductive loop, and also a slight increase in the charge transfer resistance. These evolutions are similar to those observed in fresh baths.12 Figure 3b presents the influence of the change of the PEG concentration in the plating bath on the calculated impedance spectra when the CuI共MPSA兲2 complex concentration is relatively high 共2.5 M兲. In this case, a decrease in the size of the third
capacitive loop is observed when PEG concentration decreases while the charge transfer resistance seems to decrease as PEG concentration increase. Figure 4a and b compare experimental and calculated impedance spectra for two different cases; shortly after deposition was begun and after a significant amount of charging had occurred. The experimental spectra showed that the inductive loop increases with increasing charging. Calculated spectra from the model were obtained with the same set of parameters, except for the CuI共MPSA兲2 concentration which was increased ten times from Figure 4a and b. In both cases, a good correlation is obtained between calculated and impedance spectra. Figure 4 shows that there is a close correlation between the size of the inductive loop, DIL, and the copper-complex concentration. Hence, DIL was plotted with respect to the CuI共MPSA兲2 concentration obtained from the calculated impedance spectra given in Fig. 3a 共Fig. 5兲. The size of the inductive loop increases when the copper-complex concentration is increased. Moreover, a slight increase in the size of the inductive loop is observed when the PEG concentration is increased and when the CuI共MPSA兲2 complex is maintained at 2.5 ⫻ 10−6 M 共not represented here兲. The curve of Fig. 5 was fitted with a polynomial function to obtain an empirical relation between DIL and the c3 value. Using this relation, the evolution of the concentration of the CuI共MPSA兲2 complex during bath aging was calculated from the experimental DIL values obtained during plating bath aging 共Fig. 2兲. Figure 6 presents the evolution with bath aging of the complex concentration in the bulk solution, as deduced from the model. A
Figure 4. Comparison between calculated and experimental impedance spectra.
D18
Journal of The Electrochemical Society, 154 共1兲 D13-D20 共2007兲
Figure 5. Influence of CuI共MPSA兲2 concentration on the inductive loop size, DIL, in calculated impedance spectra for c2 = 8.8 ⫻ 10−5 M.
400% increase in the complex concentration is obtained in the first hours of aging when the bath is aged with a pure copper anode while the complex concentration is maintained in the 10−7 M range 共actually at the limit of the accuracy, due to the lack of accuracy in the DIL measurement兲 when the bath is aged with a phosphorized copper anode. In another set of calculations to deduce the effect of CuI共MPSA兲2 surface coverage on the inductive loop diameter, DIL, the concentration c2 in the model was set at its nominal value. The I , was then surface coverage of the CuI共MPSA兲2 complex, Cu共MPSA兲2 calculated and the size of the inductive loop DLth was extracted from the calculated impedance spectra, for different values of c3. In a second set of calculations, the concentration c3 in the model was set at 2.5 ⫻ 10−6 M. The surface coverage of the CuI共MPSA兲2 comI , was calculated and the DLth value was extracted plex, Cu共MPSA兲2 from the calculated impedance spectra, for different values of c2 ranging from 10−6 M to the nominal value. Figure 7 shows the corI and DLth deduced from relation between the values of Cu共MPSA兲2 these two sets of calculations. All of the points are practically on the same curve, whatever the additive concentrations and their nature. Hence, no clear correlation can be established between the size of the inductive loop and the surface coverage of either CuCl or CuI–Cl-PEG. As a consequence, a direct relationship between the size of the inductive loop and the CuI共MPSA兲2 complex surface coverage can be established, indicating that the inductive loop directly depends only on the surface coverage of CuI共MPSA兲2.
Figure 6. Influence of anode material on the evolution during plating bath aging of the CuI共MPSA兲2 concentration as calculated by the model. 共䊏兲 Cu0 anode and 共쎲兲 Cu–P anode.
Figure 7. Correlation between the surface coverage of the CuI共MPSA兲2 complex and the size of the inductive loop DL deduced from the impedance spectra calculated by considering that the concentration of one of the additives is kept constant whereas the other one is changed. 共䉱兲 关PEG兴 = 88 ⫻10−6 M, 共䉮兲, 关CuI共MPSA兲2兴 = 2.5 ⫻ 10−6 M.
Changes after bath aging.— Figure 8 presents the impedance spectra evolution measured on a plating bath aged with a 4.28 A h L−1 charge amount after the current was switched off. Figure 9 compares the evolution of the size of the inductive loop with time after current switch-off, trelax, for two different aging charge amounts. The graphs show that the size of the inductive loop first increases up to a maximum, decreases, and finally disappears altogether. This change in the inductive loop occurs at a faster rate when the copper deposition charge amount before turning off the current is lower 共less deposition leading to less aged plating baths兲. Moreover, the impedance spectra of the aged bath after relaxation was different than the impedance spectra obtained for a fresh bath. This clearly indicates that a plating bath degradation has occurred during copper deposition, i.e., bath aging, in addition to the additive consumption shown previously. Discussion According to Reaction 13, the CuI共MPSA兲2 complex is in equilibrium with free Cu+ and with MPSA. On the other hand, the dis-
Figure 8. Evolution with time after current shut down of the impedance spectra of an aged plating bath.
Journal of The Electrochemical Society, 154 共1兲 D13-D20 共2007兲
D19
Conclusion
Figure 9. Evolution with relaxation time of the size of the inductive loop of the impedance diagram, for two different charge amounts passed through the aged plating bath. 共䊏兲 It/V = 7.18 Ah L−1, 共쎲兲 It/V = 4.28 Ah L−1
solution from the pure copper anode is known to continuously release Cu+ ions into the bath while the phosphorized copper anode inhibits this phenomenon. Thus, when the bath is aged with a pure copper anode, a large amount of Cu+ ions is released in the plating bath and their concentration increases above their equilibrium value as defined by Eq. 9. However, when the bath is aged with the phosphorized copper anode, the increase will be prevented. As a consequence, according to Eq. 14, when the bath is aged with a pure copper anode, the CuI共MPSA兲2 complex concentration in the plating bath increases. If the size of the inductive loop is associated with the amount of CuI共MPSA兲2 complex in the plating bath, the impedance spectra changes observed in Fig. 1a can then be easily explained. Therefore, the size of the inductive loop can be considered as a marker of the copper-accelerator complex concentration. In addition, the Cu+ disproportionation Reaction 8 and the CuI共MPSA兲2 complex formation Reaction 13 are most likely not instantaneous, and their kinetic rates are probably different. Thus, if the size of the inductive loop is associated with the amount of CuI共MPSA兲2 complex in the plating bath, the differences in the reaction kinetics together with the unknown rate at which Cu+ ions are released in the plating bath during pure copper anode dissolution might explain the continuous accumulation of the complex during bath aging. It might also explain why a transient increase in the complex concentration is observed during aged bath relaxation after the current is turned off, i.e., after the Cu+ ion release in the bath is stopped. The decrease of the size of the inductive loop observed when the bath is strongly aged with a pure copper anode can be easily explained by the progressive degradation of either the MPSA or the complex during aging, probably into HSO3–共CH2兲3–SO–SO–共CH2兲3–SO3H as determined by Frank and Bard.23 The instability with time of the aged solution observed through the impedance spectra during bath relaxation when the aging current is turned off can also be correlated with the CuI共MPSA兲2 complex stability. Indeed, so long as the current is maintained, Cu+ ions are released in the bath. After switching off the aging current, no Cu+ is released in the plating bath by pure copper anode dissolution. Thus Cu+ concentration in the plating bath will decrease towards their equilibrium value, as defined by Eq. 9, and the CuI共MPSA兲2 complex, in turn, will decrease according to Eq. 14. Moreover, the longer the bath is aged, the higher the released amount of Cu+. Thus, the higher is the complex stability and the longer it takes to recover back to the equilibrium state.
A model of copper deposition for an acidic copper sulfate plating bath containing an accelerator and an inhibitor was tested. The model is able to describe the impedance spectra evolution during copper deposition and subsequent bath aging. A good correlation between the calculated and experimental EIS spectra was obtained. The model shows that the electrode surface mainly consists of CuCl. In addition, a direct relationship was observed between the copperaccelerator complex surface coverage and the size of the inductive loop. Thus it was possible to explain the changes of the impedance spectra during plating bath aging by the existence of a CuI-accelerator complex in the bulk bath. This complex is in equilibrium with its constituents, Cu+ ions, and MPSA. It was also possible to deduce the evolutions of the complex concentration in the bulk phase and to show that, when the bath is aged with a pure copper anode, the complex concentration in the bath increased fivefold due to the release of Cu+ at the anode in the first hours of aging. The results indicate also that bath aging leads to an additive degradation in addition to the additive consumption. Hence, impedance spectroscopy measurements are mainly sensitive to the Cuaccelerator complex formation and destruction. The copper-accelerator complex seems to be the most limiting additive for superfilling copper deposition by the damascene process. As there is a fairly good correlation between the concentration of the copper-accelerator complex and the size of the inductive loop, impedance spectroscopy in the low frequency range can be a good candidate to be used as a sensor for the monitoring of the aging of this type of copper plating bath. The Centre National de la Recherche Scientifique assisted in meeting the publication costs of this article.
List of Symbols Bulk concentrations of Cu2+, Cl−, PEG, and MPSA, respectively 共mol/L兲 Cdl Double layer capacitance 共F兲 bi Exponential coefficient of Butler-Volmer equation corresponding to reaction nr i b ␣neF/RT 共bi ⬍ 0 at the cathode and bi ⬎ 0 at the anode兲 E Overpotential 共V兲 ki Kinetic rate constant of reaction i ki0 Standard kinetic rate constant of reaction i n Number of mole K1 Constant IF Faradic current Rt Charge transfer resistance f Frequency 共Hz兲 Z Impedance 共⍀兲 ZF Faradic impedance 共⍀兲  Total number of adsorption sites at copper surface 1, 2, and 3 Surface fraction covered by adsorbed CuCl, Cu–Cl-PEG, and CuMPSA, respectively Pulsation 共 = 2/f兲 共 *兲 Free site at copper surface M共 *兲 M specie adsorbed at copper surface
c0, c1, c2, and c3
References 1. T. P. Moffat, D. Wheeler, W. H. Huber, and D. Josell, Electrochem. Solid-State Lett., 4, C26 共2001兲. 2. L. T. Koh, G. Z. You, S. Y. Lim, C. Y. Lim, and P. D. Foo, Microelectron. J., 32, 973 共2001兲. 3. I. Epelboin, F. Lenoir, and R. Wiart, J. Cryst. Growth, 13, 417 共1972兲. 4. I. R. Burrows, J. A. Harrison, and J. Thomson, J. Electroanal. Chem. Interfacial Electrochem., 58, 241 共1975兲. 5. G. Razumney and J. O’M. Bockris, J. Electroanal. Chem. Interfacial Electrochem., 46, 185 共1973兲. 6. J. A. Harrison and H. R. Thirsk, in Electroanalytical Chemistry, A. J. Bard, Editor, p. 67, Marcel Dekker, New York 共1971兲. 7. E. Chassaing and R. Wiart, Electrochim. Acta, 29, 649 共1984兲. 8. D. M. Soares, S. Wasle, K. G. Weil, and K. Doblhofer, J. Electroanal. Chem., 532, 353 共2002兲. 9. S. Goldbach, W. Messing, T. Daenen, and F. Lapicque, Electrochim. Acta, 44, 323 共1998兲. 10. J. P. Healy, D. Pletcher, and M. Goodenough, J. Electroanal. Chem., 338, 155 共1992兲.
D20
Journal of The Electrochemical Society, 154 共1兲 D13-D20 共2007兲
11. C. Gabrielli, Ph. Moçotéguy, H. Perrot, and R. Wiart, J. Electroanal. Chem., 572, 376 共2004兲. 12. C. Gabrielli, P. Moçotéguy, H. Perrot, D. Nieto-Sanz, and A. Zdunek, Electrochim. Acta, 51, 1462 共2006兲. 13. E. Mattson and J. O’M. Bockris, Trans. Faraday Soc., 55, 1586 共1959兲. 14. J. O’M. Bockris and M. Enyo, Trans. Faraday Soc., 58, 1187 共1962兲. 15. O. R. Brown and H. R. Thirsk, Electrochim. Acta, 10, 383 共1965兲. 16. F. Chao and M. Costa, Bull. Soc. Chim. Fr., 10, 4015 共1968兲. 17. S. T. Rashkov and L. Vuchkov, Surf. Technol., 14, 309 共1981兲. 18. J. P. Healy, D. Pletcher, and M. Goodenough, J. Electroanal. Chem., 338, 179 共1992兲. 19. T. P. Moffat, J. E. Bonevich, W. H. Huber, A. Stanishevsky, D. R. Kelly, G. R. Stafford, and D. Josell, J. Electrochem. Soc., 147, 4524 共2000兲. 20. K. R. Hebert, J. Electrochem. Soc., 148, C726 共2001兲. 21. N. Y. Kovarsky, Z-W. Sun, and G. Dixit, in Proceedings of the ULSI XVII Conference 共2002兲. 22. J. P. Healy, D. Pletcher, and M. Goodenough, J. Electroanal. Chem., 338, 167
共1992兲. 23. A. Frank and A. J. Bard, J. Electrochem. Soc., 150, C244 共2003兲. 24. T. P. Moffat, B. Baker, D. Wheeler, and J. D. Josell, Electrochem. Solid-State Lett., 6, C59 共2003兲. 25. C. Gabrielli, P. Moçotéguy, H. Perrot, D. Nieto-Sanz, and A. Zdunek, J. Electrochem. Soc., Submitted. 26. D. F. Suarez and F. A. Olson, J. Appl. Electrochem., 22, 1002 共1992兲. 27. C. Gabrielli, P. Moçotéguy, H. Perrot, A. Zdunek, D. Nieto-Sanz, and M. C. Clech, in Copper Interconnects, Low-k Inter-level Dielectrics, and New Contact Metallurgies/Structure, G. S. Mathad, Editor, PV 2003-10, p. 112, The Electrochemical Society Proceedings Series, Pennington, NJ 共2003兲. 28. P. Moçotéguy, C. Gabrielli, H. Perrot, A. Zdunek, and D. Nieto-Sanz, J. Electrochem. Soc., 153, G1086 共2006兲. 29. J. Sedzimir and W. Gumowska, Hydrometallurgy, 24, 203 共1990兲. 30. L. Mirkova and S. Rashkov, J. Appl. Electrochem., 24, 420 共1994兲. 31. C. Gabrielli, P. Moçoteguy, H. Perrot, A. Zdunek, P. Bouard, and M. Haddix, Electrochem. Solid-State Lett., 7, C31 共2004兲.
