Fig. 4-1 demonstrates representative voltage responses for the Cu plating RDEs with 25μM of 4-amino-2,1,3-benzothiadiazol at rotation speeds of 200 and 430 rpm, respectively. These patterns were consistent with what were observed earlier by Dow and Liu in which the voltage revealed a sudden rise (becoming less negative) and stabilized after 50 sec [6]. Voltages at the initiation of deposition were more polarized, it is because of the rapid adsorption of suppressor and leveler. Similar phenomenon was also observed by Reid et al. in their galvanostatic measurements [51]. The applied voltage then became less negative, it is ascribed to the adsorption of accelerating component, in our cases, Cl-. In “Literature Review” section, the convection-dependent adsorption mechanism was reviewed. A smaller rotating speed, in our studies, 200 rpm, could be used to simulate smoother flow at the trench bottom.
A higher rotating rate could be used to simulate the stronger flow at the trench top.
Thus, the different adsorption abilities of additives versus different convection rate would make considerable leveling effects. Apparently, the voltage curve from 200 rpm was less negative than that of 430 rpm. Since the RDE underwent a galvanostatic Cu deposition, the voltage with less negative reading indicated a lower plating resistance which was attributed to a reduced amount of leveler absorption. From these profiles, we surmised that in plating trenches at identical leveler concentration, there would be stronger absorption of levelers at trench top as opposed to the bottom. As a result, Cu deposition at trench top was retarded relatively to that of trench bottom.
For fair comparison, we took the potential difference at 300 sec and defined Δŋ
Δŋ = Es - Ef
where Es and Ef are the voltages at 200 and 430 rpm respectively. In such way, a positive value of Δŋ inferred there was stronger absorption of additive at trench top relative to that of trench bottom while a negative Δŋ suggested otherwise. Because spatial variation of inhibitor in a patterned substrate is known to determine local current distribution, we could use the Δŋ to estimate current difference between trench top and bottom. In this way, a positive Δŋ indicated a larger current at the trench bottom while a negative Δŋ suggested a larger current at the trench top.
Figure 4-1. Representative voltage responses for galvanostatic RDE measurements at 200 and 430 rpm using a Cu plating bath with 25μM of 4-amino-2,1,3
The evolutions for Δŋ on base electrolyte, PEG containing solution for RDE measurements at 200 and 430 rpm are provided in Fig. 4-2. Furthermore, the evolutions for Δŋ on various leveler concentrations for RDE measurements at 200 and 430 rpm are provided in Fig. 4-3. For convenience, the results of Fig. 4-2 and 4-3 are summarized and demonstrated in Fig. 4-4.
For the base electrolyte and PEG contained solution, the Δŋ values are positive for the base electrolyte but negative for the PEG contained solution, respectively.
Based on previous theories, superfilling would take place at the base electrolyte while voids would form at the PEG containing solution.
Interestingly, both levelers revealed a similar “volcano” pattern with the highest point occurring at medium concentration. The appearance of volcano pattern was predicted by Cao et al. in their theoretic studies of Cu filling in high-aspect ratio trenches [50]. For the 4-amino-2,1,3-benzothiadiazol, the Δŋ value were positive for concentrations between 10 and 50 μM. For concentration outside this range, the Δŋ value became negative. On the other hand, for 6-aminobenzo-thiazole, the Δŋ value were negative for the entire concentration range. According to Cao et al., deposition behaviors of Cu in trenches can be predicted by the numerical value of p in following relation [50],
where ibottom and itop are local current density at the trench bottom and top, respectively.
It was theorized and confirmed experimentally that superfilling was taking place only when p became positive [50]. On the other hand, conformal deposition was likely at p
≈ 0 and unfilled trenches were occurring for p < 0. Therefore, we expected superfilling performance to appear for 4-amino-2,1,3-benzothiadiazol with
Figure 4-3. Representative voltage responses for galvanostatic RDE measurements at
concentration of (a) 1, (b) 10, (c) 25, (d) 50, and (e) 100 μM; as well as 6-aminobenzo-thiazole in (f) 1, (g) 10, (h) 25, (i) 50, and (j) 100 μM.
Figure 4-4. Profiles for potential difference (Δŋ) from galvanostatic RDE
measurements at 200 and 430 rpm using a Cu plating bath with; base electrolyte (▲) , PEG contained solution (◇), 4-amino-2,1,3-benzothiadiazol (▓), and 6-aminobenzo -thiazole (○) at various concentrations.
Fig. 4-5 demonstrates the cross-sectional SEM images for the plating results from base electrolyte and PEG containing solution. As expected from the RDE analysis, considerable filling performance difference revealed between base
clear that the result of PEG containing solution consistent with its RDE analysis since Δŋ values is negative for PEG containing solution. However, for the plating result from base electrolyte, the defects appeared at the sidewall of the trenches. This undesirable result is oppositive to the expecting superfilling performance from RDE experiments.
In the study by Dow et al. [6], the addition of PEG did not show discernible difference from voltages between fast rotating speed and low rotating speed of galvanostatic RDE measurements. Dow et al. also performed CLSV measurements in their study of convection-dependent adsorption mechanism [38]. In their study, no obvious change was detected from CLSV curves between fast rotating speed and low rotating speed for their base electrolyte and PEG containing solution.
Figure 4-5. Cross-sectional SEM images for the Cu plating bath; (a) no additives, (b) PEG.
Fig. 4-6 demonstrates the cross-sectional SEM images for the plating results from Cu electrolytes with the leveler of 4-amino-2,1,3-benzothiadiazol. As shown, when the leveler concentration was 1 μM, seam appeared in every trench, which was attributed to the conformal deposition of Cu. Once the leveler concentration was increased to 10 and 25 μM, desirable superfilling behaviors were occurring. However, at concentrations of 50 μM and above, we observed formation of voids in trenches, which were often resulted from subconformal deposition. Overall, we witnessed filled and unfilled trenches contingent on the concentration of levelers. The trend agreed well with what was predicted in Fig. 4-4.
Cross-sectional SEM images for the plating results from Cu electrolyte with the leveler of 6-aminobenzo-thiazole at identical concentration range are shown in Fig.
4-7. As expected, suitable concentration for desirable superfilling was not identified.
For the leveler concentration at 1, 10, and 25 μM, we observed obvious voids in each trenches. At a concentration of 50 μM, the number of defects was the smallest.
However, the filling turned worse again once the leveler concentration was increased to 100 μM. Microstructural evolution seems to agree with what was predicted in Fig.
4-4.
Figure 4-6. Cross-sectional SEM images for the Cu plating bath with 4-amino-2,1,3 -benzothiadiazol at concentrations of (a) 1, (b) 10, (c) 25, (d) 50, and (e) 100 μM, respectively.
A schematic diagram illustrating the operation mechanism for these two levelers is presented in Fig. 4-8. In general, adsorption capability of levelers on a plating surface was determined by their amine groups. Furthermore, the lone pair of N and S atoms of the levelers promoted complexation with Cu+ that affects leveling power [41]. Therefore, we expected the 4-amino-2,1,3-benzothiadiazol to exhibit a higher absorption ability as opposed to that of 6-aminobenzo-thiazole because the latter would allowed an up or down absorption only while the former permitted additional lateral position. According to the diffusion-adsorption mechanism, the concentration of leveler was predominant on the trench mouth [48]. Because the 4-amino-2,1,3-benzothiadiazol has a higher absorption ability, we expected a normal spatial distribution with significant difference between trench mouth and bottom. On the other hand, the 6-aminobenzo-thiazole exhibited an intrinsically reduced absorption ability so its spatial distribution between trench mouth and bottom was relatively moderate. Similar behaviors were established by Chiu et al., where the concentration gradient of 2-mercaptopyridine between trench top and bottom was more than that of 4-mercaptopyridine because the adsorption ability of former was higher than that of latter [41].
Leung et al. also studied influence of different substitution to the smoothening effects [3]. In their study, the morphology of Cu deposits changes dramatically when the molecular structure of the additive varies slightly. The deposits are smoother when the positions at the triazole nitrogen are not blocked. Their finding suggest that the smoothening effect is strongly related to the ability of additive to form a polymetric complex through the triazole righ and/or the substitutent groups. By detecting our two leveler, it seems that 4-amino-2,1,3-benzothiadiazol would be more effective to form a polymeric complex than 6-aminobenzo-thiazole because of the more closed amino
However, the trend of filling performance with respect to the concentrations of leveler is still not explained. Fig. 4-9 shows the galvanostatic current measurements of 4-amino-2, 1,3-benzothiadiazol and 6-aminobenzo-thiazole. The initial increase in the polarization may be due to the rapid adsorption of suppressor and leveler [51]. As plating continued, the cathodic potential decreased, which was likely due to either morphology changes or increasing adsorption of the accelerator (Cl-). After initial periods of fluctuation, the cathodic polarization stabilized and remained unchanged.
For all these curves, the voltage reading becomes constant after electroplating for about 75 seconds. It is interesting to find that the filling performances prediction and filling performance shown in Fig. 4-4 and 4-6 conformed well to the trend of polarization in Fig. 4-9 (a). Since both suppressor and leveler formed inhibiting complex with the adsorbed Cu+ and Cl- on the substrate, they are likely to have a synergistic inhibiting effect with each other. A schematic illustration is shown in Fig.
4-10. Therefore, there must be uniform surface coverage for both suppressor and leveler. If the concentration ratio of leveler became undesirable larger or smaller, the synergistic inhibiting effect was expected to decrease. According to literature of Dow et al. [24], the binding strength of PEG was proportional to the number of ether groups on the PEG. It is because the ether groups functioned as ligands to coordinate the Cu+ ions, and which are linked with the Cl- ions. Thus, the adsorption ability of PEG is likely to be greater than the levelers because of more ligands on PEG. As the concentration of leveler was in excess, relatively smaller leveler was able to diffuse much quickly than PEG, so the resulting surface coverage for the leveler became larger. The greater surface coverage of leveler has a lower inhibiting ability because the amine groups on the leveler are less than the ether groups on PEG. In addition, it was expected that the leveler had a lower binding strength. On the other hand, if the
smaller. Hence, the superfilling could only be achieved in optimum concentration range.
However, for 6-aminobenzo-thiazole, the polarization was not consistent with the filling performances prediction and filling performance. It is because that the polarization curve was acquired from the wafer fragment, the representing resistance was the combining results of trench mouth and trench bottom. The filling performance was mainly regard to the different resistance between the trench mouth and trench bottom rather than the overall resistance. More studies are still necessary to clarify the meaning of these galvanostatic measurements.
Figure 4-9. Galvanostatic current measurements of (a) 4-amino-2,1,3-benzothiadiazol and (b) 6-aminobenzo-thiazole: 0 (◇), 1 (●), 10 (○), 25 (▽), 50 (★), and 100 μM (▲), respectively.
Figure 4-10. A schematic illustration of the synergistic inhibiting effect between PEG and leveler.
XRD analysis of the deposited Cu films is presented in Fig. 4-11. The Cu film exhibited an obvious (220) peak and much weaker (111) and (200) peaks. PEG addition suppressed the (220) peak and increased the (111) and (200) peaks. However, both the 4-amino-2,1,3-benzothiadiazol and 6-aminobenzo-thiazole did not apparently alter the peak of diffraction peak. In the study of Chang et al. [13], they indicated that (111) plane was the lowest surface energy plane for Cu, and increased plating current density would improve the (111) peak. In our case, altering concentration of our
Figure 4-11. X-Ray diffraction patterns of the electroplating Cu films on the patterned wafer; (A) 4-amino-2, 1, 3-benzothiadiazol and (B) 6-aminobenzo-thiazole.
Surface roughness for the deposited Cu film was also studied. They are demonstrated in Fig. 4-12 and 4-13. The mean roughness of Cu film deposited by base electrolyte was 38.344 nm. In contrast, for PEG-containing electrolyte, the mean roughness was only 9.656 nm. The addition of PEG in the electrolyte not only changed the XRD pattern but also suppressed the deposition rate. However, it revealed an interesting result of reduced surface roughness. As shown in Fig. 4-12 (c) and (d), the surface roughness for the deposited Cu film from 25 and 100 μM 4-amino-2,1,3-benzothiadiazol addition electrolyte were 11.457 and 12.164 nm, respectively. Comparing to the surface roughness on these two concentrations (one revealed superfilling and the other one delivered poor filling performance), there was negligible difference between their surface roughness. It is likely that the concentration of leveler did not exert any influence on the Cu. Similar behavior was also observed from electroplating by 6-aminobenzo-thiazole. Although different concentrations of the leveler greatly affected the filling performance, no apparent variation on the surface roughness for the deposited films was observed.
From the studies of Leung and Kim et al. [3,5], the different concentration would greatly affect the roughness. However, the same behavior was not observed in our experiments. It was possible because a lower concentration or different substrate we chosen in our case.
Figure 4-12. AFM analysis on 2D surface roughness in (a) base electrolyte without additives, (b) PEG, (c) 4-amino-2,1,3-benzothiadiazol at 25 μM, and (d) 4-amino-2,1, 3-benzothiadiazol at 100 μM.
Figure 4-13. AFM analysis on 2D surface roughness of 6-aminobenzo-thiazole in (a) 10, (b) 25, (c) 50, and (d) 100 μM, repectively.