The Types of Profile Evolution in Damascene Plating
After demonstrating the first fully integrated devices with Cu interconnects in 1997 [2], a mathematical model for damascene plating was proposed by Andricacos et al. in 1998 [7]. In their work, they proposed three possible ways for the profile of plated Cu to evolve with time. They are shown in Fig. 2-1. For a defect-free filling, the deposition rate at the trench bottom should be higher than that on sidewall of the feature, leading to a superconformal deposition. This phenomenon is also called
“superfilling.” The superfilling is attributed to the inhibiting additives in copper electroplating bath consumed on the wafer surface that suppresses the kinetics of Cu deposition. Since the interior regime of the trench is less-populated by the additive, we expect to see a higher deposition rate within the trench.
The Concentrations of Cupric Ion and Sulfuric Acid
The concentrations of Cu2+ and H2SO4 not only greatly affect the diffusion coefficient of the Cu2+ but also the filling performance. The effects of Cu2+, H2SO4
concentration, and temperature have been discussed thoroughly by Moats et al. [8].
Some important results are presented in Table 2-1 and 2-2. It can be concluded that the diffusion coefficient of Cu2+ decreases as the concentrations for CuSO4 and H2SO4
increase. Moreover, the diffusion coefficient improves with increasing temperature as expected.
Figure 2-1. Three possible ways of profile evolution in plated copper; subconformal, conformal, and superconformal deposition [7].
Table 2-1. Effect of CuSO4 and H2SO4 concentrations on the diffusion coefficient of
Table 2-2. Effect of temperature on the Cu2+ on the diffusion coefficient in 40 g/L CuSO4 and 160 g/L H2SO4 electrolyte [8].
The effects of Cu2+ and H2SO4 were also discussed by J. Reid in 2001 [9]. He proposed that the electroplating bath should balance the advantages of higher and lower acid solutions to reach a compromised value. A mathematical model discussing fluid flow, transport by diffusion, migration, convection, and multiple species in Cu electroplating bath was proposed by Georgidadou et al. [10]. In their work, they simulated the shape evolution during electroplating in different width and aspect-ratio trenches by testing two different H2SO4 concentrations. A similar simulation was also conducted by A.C. West in 2000 [11]. In his study, he defined two dimensionless parameters which corresponded to the concentration of Cu2+ and additives to discuss the filling performance. Another important parameter defined in his work is:
⎟⎟
Where ibottom and itop are the current density at the trench bottom and trench top, espectively. The current density difference between the trench top and bottom is used to evaluate the filling performance.
The Current Density
Different current densities during electroplating often lead to considerable differences in resistivity and surface roughness. These effects were discussed by Chang et al. previously [12-13]. In their work, the surface roughness and resistivity for the deposited Cu film were lower at specific current density range. This specific current density range varied contingent on the electroplating bath selected.
The Rotating Rate
In the electroplating of commercial wafer, a fountain flow type equipment is widely used. In laboratory experiments, the electroplating is often carried out with stirring or air bubbling to ensure necessary mass transport. It is understood that the agitation of the electroplating bath also makes a difference in the convection rate between the trench open and bottom. This convection-dependent adsorption is to be discussed later. Unfortunately, previous authors did not disclose relevant stirring rates in their work. In the experiment form J.J. Kelly et al. in 1999 [14], the authors used the threaded rotating cylinder electrode as the working electrode to study the leveling effect. They found that the leveling power was greater when a specific relation was established between the current density and rotation speed, as shown in Fig. 2-2.
Figure 2-2. Effect of the cylinder electrode rotation speed and current density on the leveling power [14].
Roles of Chloride Ion
The roles of Cl- were investigated using scanning electron microscope (SEM), optical microscope (OM), electron paramagnetic resonance (EPR), and galvanostatic measurements by Dow et al. [15-17]. In their work, the authors determined that the Cl- played three distinct rules; the electron bridge for the reduction of Cu2+, the anchor for suppressor, and the promoter for accelerator. The electron transfer bridge made by Cl- between the Cu2+ was deduced form the broadening of the EPR peak, which was due to the delocalization of unpaired electrons. Polarization effects were apparent from the polarization curves in galvanostatic measurements of the electrolyte with
SPS or MPS accelerator but without Cl-. In the SEM observations, cuprous chloride (CuCl) crystals were used as tracers for adsorbed-PEG. It was determined that the CuCl resided mostly at the bottom of the micro-via rather than the external surface.
This phenomenon supported that the desirable bottom-up filling behavior was succeeded by the unique synergistic effect between accelerator and Cl-. The electrons transfer models for the Cu2+ is shown in Fig. 2-3 [17].
The role of Cl- in suppressing the Cu electrodeposition with PEG was widely discussed by J.J. Kelly et al. [18-19]. The authors used a polarization curve to detect the polarization behavior for the electrolyte with or without additives. They found that the polarization curves for the acid electrolyte containing only the PEG were almost identical. However, those polarization curves revealed considerable response with the addition of PEG and Cl-. The frequency difference from quartz crystal microbalance (QCM) was also used to detect the effects for the concentration of Cl- and the molecular weight of PEG. It was used to predict the adsorption behavior of PEG, as shown in Fig. 2-4 [18]. The authors also proposed the inhibiting effect of PEG, as shown in Fig. 2-5 [19]. It displays that the adsorption active sites were blocked by PEG molecules. Fitting between experimental and simulated impedance spectra for different Cl- concentrations also agreed with the steady-state polarization curves, and their results confirmed the hypothesis of a dimensionless surface coverage of PEG. A near complete surface coverage of PEG molecules competed available adsorption sites with Cu2+, forming an inhibiting monolayer composed of spherical packed PEG molecules that were responsible for the polarization effects.
Figure 2-3. Illustration of various outer and inner sphere electron transfer models of Cu2+ complexes [17].
A mathematical model was proposed by K.R. Hebert [20]. This model formulated a relation between the adsorbed PEG and Cl-, in which the PEG coverage was determined by the Cl- adsorption. After fitting with experimental results, the model was applied to predict steady-state current-potential curves. Their results helped elucidating the synergistic effects between PEG and Cl-, as well as the reasons for the optimized Cl- concentrations in Cu electroplating baths.
Suppressor
PEG is extensively used as a suppressor, which is a critical component in the Cu electroplating bath. Although the three-additive system was widely used in many studies for Cu electroplating, some literature reported only two-additive system, that is, accelerator and suppressor. Superfilling can be achieved by combining these two additives at proper ratios, but the “overfill bump” was typically formed above the superfilled features. Recently, some studies demonstrated that using only the PEG as a suppressor with optimum Cl- concentration could also achieve desirable bottom-up superfilling on patterned wafers [21-23]. In our work, we also achieved superfilling with an electrolyte without accelerator. This phenomenon can be explained by the depolarization effect of Cl- for Cu electroplating and the inhibiting effect of PEG. As the ratio between these two components was optimized, a proper spatial distribution of Cl- and PEG at different locations allowed the bottom-up filling.
PEG is a common name for a polymer called polyethylene glycol. The PEG can be obtained with average molecular weight (Mw) from 200 to 35000 g/mol. Recently, Dow et al. discussed the influence of PEG molecular weight on the filling performance in microvias [24]. Their study indicated that the only lager PEG whose Mw exceeded 2000 g/mol could effectively polarize the cathode. Furthermore, the desirable filling would be obtained with PEG ranging between 6000 and 8000 g/mol.
When the Mw of PEG was above this range, a strong convection-dependent adsorption induced a significant drop of filling performance in larger via because the
fluid motion is more active at the bottom of the large via compared to the smaller one.
The morphology for the adsorbed PEG and its potential dependency also attracted many attentions. Yokoi et al. suggested that PEG trapped the cuprous ion (Cu+) on the surface to form an adsorbed inhibiting form [25], Kelly et al. indicated
PEG molecules [18], and Jin et al. used the AFM to determine that the adsorbed particles revealed a cone shape with a bottom radius about 15-25 nm and a height of 2-4 nm [26].
Figure 2-5. A schematic diagram of PEG behavior model [19].
Accelerator
Accelerator is sometimes called brightener because it can increase the brightness of the deposited Cu surface. There have been many studies discussing common accelerators such as SPS and MPS(A). The accelerator could help the reduction of Cu2+ because of its synergistic effects with Cl-. It was well known that the reduction from Cu2+ to Cu+ is the rate-limiting step, and the thiol group belonging to the accelerator can help the creation of Cu+ by the synergistic effects with Cl- [16-17].
However, in the study of superfilling evolution dependence on the aging time of MPSA by Kim et al. [27], the authors reported that the MPS was ineffective in the superconformal deposition of Cu. The UV-vis spectroscopy confirmed that the MPS was converted to SPS after aging for 12 hrs. The SEM observation revealed substantial differences in the filling performance between the MPSA containing electrolytes with different aging time. A schematic illustration for the suggested mechanism is displayed in Fig. 2-6.
In addition to SPS and MPS, some substitutive accelerators were also explored.
Cho et al. presented superfilling using 3-N,N-dimethylaminodithiocarbamoyl -1-propanesulfonic acid (DPS) as an accelerator [28], and discussed the equilibrium of DPS related to its concentration, which affected the acceleration effects.
.
Figure 2-6. A schematic illustration of the proposed mechanism for different filling aspects between MPSA and SPS/aged MPSA [27].
From literature, the principal issuse for the accelerator-suppressor system is the
[31-35]. The authors used a slow scan rate voltammetry, and identified the slope of the voltammetric curves to show acceleration of Cu deposition as a function of concentration of MPSA. The i-η curves was described by the Bulter-Volmer equation, and the time-dependent fractional surface coverage θ(t) was calculated assuming irreversible statistical adsorption. The simulating i(θ)- η(θ) curves also revealed hysteresis as a function of CMPSA which was ascribed to the competitive interaction between the additives. They demonstrated nicely with the experiments, particularly for higher CMPSA. As the interface moved, the local coverage increased on the concave surface and decreased on the convex portions. This model brought general implications for understanding the accelerator and was clearly different from the leveling models, which was based on the diffusion limited accumulation of inhibiting molecules. Curvature-enhanced accelerator coverage mechanism was soon extended as curvature enhanced adsorbate coverage mechanism, because this mechanism is suitably applied for not only accelerator but also other chemical additives [36].
A similar simulation model was proposed by West et al. [29], and the aspect of this model was that the reduction of the surface area available for additive adsorption engendered a temporary decrease in the amount of inhibition. There were three assumptions made in A.C. West’s model [29]:
1. Acceleration at the bottom of a feature is due to a change in surface area as deposition proceeds. The decrease in surface area results in increase in the amount of SPS.
2. An increase of the surface coverage of SPS lowers the surface coverage of PEG.
3. The concentration variations in the electrolyte of all species can be neglected. This assumption becomes increasingly appropriate as feature size decreases and
is relatively straightforward to relax.
surface coverage of PEG and SPS. The change in surface area was estimated from the local angle of interaction of neighboring elements and the local current density. The simulation were applied for different trenches and compared with the SEM observations.
CEAC mechanism and A.C. West’s model both simulated the filling performances based on the same hypothesis, that is, the surface coverage of the additives changed with the surface area. Then the authors connected the surface coverage of additive with the local current density distribution to simulate the Cu depositing evolution and bump formation.
Leveler
Leveler is one kind of additives used to increase the smoothness of the deposited metal films by its inhibiting effects. However, the “leveling effect” does not refer only for the levelers. Definition of leveler is still disputable and ambiguous, and specific function groups of molecules are usually used to distinguish leveler. Chang et al. and Lin et al. reported that the additives with benzyl and amino groups were desirable levelers [12,37]. However, Dow et al. indicated that amine and heterocyclic compounds were common functional groups of levelers, and usually, these levelers contained primary, secondary, tertiary amines, or particularly quaternary ammonium salts. Therefore, these levelers commonly possess one or more positive charges [38]. J.
Reid interpreted that leveler is one kind of current suppressing molecules, and usually added to the plating bath with a low concentration [9]. Kim et al. pointed out that the
(ⅰ) cationic or neutral heteroaromatic; (ⅱ) condensed heteroaromatic; (ⅲ) polymers with aromatic, cyclic or nitrogen containing substituents.
Several levelers were discussed in previous literature and described below.
Benzotriazole (BTA) was known not only as an excellent corrosion inhibitor for Cu surface but an inhibitor for Cu deposition [3-4]. Leung et al. investigated the effects of four substituted benzotriazole compounds for Cu electrodeposition by AFM and secondary ion mass spectroscope (SIMS) [3]. BTA and the effectively substituted compounds of BTA for Cu electroplating were believed to act as an chain like inhibiting adlayer. The structure of the polymeric Cu(I)BTA complex are shown in Fig.
2-7. The authors suggested that the smoothening effect was strongly related to the ability of the additives to form a polymeric complex through the triazole ring and/or the substituent groups.
Figure 2-7. A schematic illustration of the suggested polymeric Cu(I)BTA complex [3].
Kim et al. also studied the effects of BTA in the electroplating solution on the properties of the deposited copper films [4]. Moffat et al. studied the superconformal deposition at trenches in various scales using electroplating bathes with different additives [5]. Some additives, such as thiourea, ammounium peroxydisulfate, 4-mercaptopyridine (4-MP), 2-mercaptopyridine (2-MP), and 2-aminobenzothiazole
(2-ABT) were also employed as levelers in the Cu electroplating bath. The filling performance and inhibiting ability of those levelerst were studied by Lin et al. [37].
The difference in adsorption/desorption ability between 2-MP and 4-MP were compared to elucidate the filling performance difference between them. Fig. 2-8 provides an illustration scheme for their behaviors. Diethyl safranine azo dimethyl aniline (Janus Green B, JGB) was also widely applied and studied in literature [6,14, 38,42-45], and its molecular structure is displayed in Fig. 2-9. Kim et al. investigated the impact of the branched polyethyleneimine (PEI) on Cu electrodeposition by voltammetric curves and SEM observations [39]. Bozzini et al. reported the Cu electrodeposition from acidic electroplating bath containing a promising polymeric leveler, benzyl-phenyl-modified polyethyleneimine (BPPEI) [40].
Convection-dependent adsorption of JGB was proposed by several authors [38, 43-45]. Dow et al. studied the electrochemical and inhibitive behaviors characterized by cyclic linear sweep voltammetry (CLSV) using different rotating speeds of the working electrode (WE) [38]. They found that the inhibition effect of JGB on Cu deposition depended on the applied potential of the WE. As the rotation speed was increased, the inhibiting effect of JGB was correspondingly enhanced. The enhanced inhibiting effect achieved by stronger forced convection was attributed to the diffusion-limited transfer of JGB and the convection transport of Cl-. Sun et al.
calculated the stripping areas of the cyclic linear voltammetry (CVS) at different JGB concentrations. They found the difference between the stripping area of 100 rpm and 2500 rpm revealed the largest value in JGB ranging 20 to 50 mg/L. Hence, they
study inhibiting effect under different rotation speeds of RDE and reached the same conclusion as the literature proposed by Miura and Honma [45]. Based on these studies, Dow and Liu presented a feasible method to evaluate the filling performance of Cu plating formulas [6].
Figure 2-8. Adsorption and desorption ability of 2-MP and 4-MP onto a Cu surface [41].
Figure 2-9. Molecular structure of JGB [38].
Recently, the deactivation effect of leveler to the adsorbed accelerator was widely studied. Although the superfilling can be achieved in electrodeposition by electrolyte containing only accelerator and suppressor, the addition of leveler could control the bump formation and relieved subsequent CMP process. The effects of Cu deposition of PEI and dodecyltrimethylammoniumchloride (DTAC) were studied by Kim et al. [39,46]. They suggested that the addition of cationic polyelectrolyte, PEI, quenched the activity of SPS. This effect was attributed to an ion-pairing interaction between the cationic imine groups of the polyelectrolyte and the anionic tail groups of the adsorbed SPS accelerator [39].
A diffusion-adsorption mechanism was widely used to interpret the leveling effect of both suppressor and leveler [47-49]. Roha and Landau presented a quantitative model for the leveling effect of plating additives [47]. The fundamental assumption for this model was that the coverage of adsorbed additives was controlled by mass transport. They discussed the mass balance through three processes;
Tobias used some assumptions made by previous authors and presented a model based on the diffusion-adsorption mechanism [48]. As shown in Fig. 2-11, because the shorter diffusion distance from the bulk solution to the peak relative to the valley, more inhibitors arrived at the peaks. Therefore, the electrodeposition at the peak was more inhibited, and the profile became smoother. The electrodeposition of Ni into an angular trench in the presence of coumarin, a widely used inhibitor, was simulated using boundary layer approximations for flow parallel and transverse to the grooves in this literature. Cheng and West employed an electrohydrodynamic (EHD) impedance spectroscopy to study the influence of coumarin in Ni electrodeposition [49]. In their study, the determination of the interfacial kinetic and transport parameters relevant to their models of leveling agents was clearly demonstrated.
Figure 2-11. A schematic illustration of diffusion-adsorption mechanism of leveling effect [48].
Three Additive Model
Cao et al. proposed a model describing the effect of SPS (accelerator), PEG (suppressor), and JGB (leveler) on the leveling efficiency at sub-micrometer trenches [50]. Their simulating results were also compared to the filling experiments. The dependence of the derivative of the current density with respect to diffusion layer thickness, as well as the variation of p for three geometries as a function of JGB concentration is shown in Fig. 2-12.
The authors proposed a parameter for convience, that is:
⎟⎟
When p > 0, superfilling should be observed. Only when p ≈ 0, a nearly conformal deposition is expected.
Figure 2-12. The dependence of derivative of current density with respect to the diffusion layer thickness, as well as the variation of p for three geometries as a function of JGB concentration [50].