Abstract
A two-step transient method was developed to electrochemically screen suppressor molecules toward the applications in Cu plating chemistry for narrow trenches. The first potential transient with full amount of suppressor was acquired to simulate the potential applied on the wafer, which is largely determined by the plating of the field region on the wafer. A second current transient experiment with the potential controlled by the transients acquired in the first experiment is used to mimic the plating at different locations across the wafer. When different fractional amounts of suppressor are injected into the second experiment, the current transients acquired thereof simulate the plating rate in the trench structure. A slower super filling rate at higher plating current was well predicted by the transient method and was confirmed by filling experiments. A good qualitative correlation was also observed between the current transients and the filling performance of different suppressors.
Export citation and abstract BibTeX RIS
This is an open access article distributed under the terms of the Creative Commons Attribution Non-Commercial No Derivatives 4.0 License (CC BY-NC-ND, http://creativecommons.org/licenses/by-nc-nd/4.0/), which permits non-commercial reuse, distribution, and reproduction in any medium, provided the original work is not changed in any way and is properly cited. For permission for commercial reuse, please email: oa@electrochem.org.
Cu electroplating has been one of the critical steps in the Cu interconnect technology, which has been used for over a decade in semiconductor and microelectronic devices.1,2 The Cu electroplating is a so called super-conformal filling or super filling process, where the deposition rate at the bottom of a structure is much faster than the rate at the top of the structure so that the structure is filled void free. These super filling processes generally rely on the organic additives in the plating solution, particularly, the suppressor and accelerator. The additive behavior3–18 and the associated properties of the plated Cu films19–25 have been extensively studied. As the dimensions of the state-of-the-art semiconductor devices scale, advanced plating chemistries are demanded to form the structures defect free.
The mechanism of the super conformal filling has been studied extensively. A widely adopted mechanism was proposed by Moffat et al.9–12 and West et al.,13 where an accumulation and crowding of the accelerator was proposed to occur on the Cu surface at the bottom of the structure due to a continuous shrinkage of the Cu growth front surface area during plating. This so-called curvature enhanced accelerator coverage (CEAC) hypothesis is in agreement with the observations that the accelerator floats on top of the Cu surface during plating11 and that the plating rate suppressed by the suppressor is promoted by the accelerator due to a displacement or disruption of the surface passivation film.12 The CEAC model therefore successfully explains a faster local plating rate at the bottom of the feature, resulting in the super filling. The CEAC mechanism also predicts a bump formation on top of dense patterns of small structures due to the locally high surface coverage of accelerator after the structures are filled.10,13 In addition, the CEAC mechanism predicts that a pre-treatment of the Cu substrate in a more concentrated accelerator solution can provide a higher initial coverage of accelerator. Therefore, the curvature enhancement of the accelerator coverage can be achieved, facilitating the super filling, in a plating solution with suppressor only.11
Another widely used model was put forward by Akolkar and Landau16,26 to address the transient effect of the suppressor. Due to the fast adsorption and slow transport of suppressor molecules such as PEG, the fast transport of accelerator molecule, SPS, and the slow displacement of suppressor by accelerator, they predicted a suppressor concentration gradient along the feature depth, which is established immediately after the substrate is immersed into the electrolyte. This concentration gradient of the suppressors, coupled with the interaction with accelerator, was used to explain the faster Cu plating at the bottom of the structure. Although a concentration gradient is typically involved in the leveling mechanism in a plating solution, this is a new perspective of looking at the suppressor behavior in the super filling mechanism.
The impact of the diffusion and distribution of suppressor was demonstrated by Akolkar and Dubin, where they observed a much slower filling rate on the isolated structures compared with the nested structures.27 They further simulated the suppressor distribution and correlated the slower filling rate of the isolated structure with a thinner depletion zone of the suppressor. Another observation of the importance of the suppressor concentration gradient to the super filling is presented in Figures 1. Three substrates with 80 nm wide lines were plated in the same Cu plating solution with different diffusion conditions. Super conformal and defect free filling was observed for the sample plated with hot entry, where the plating starts as soon as the substrate is immersed in the solution. On the other hand, conformal plating was observed when sonication was used during the plating or when a time delay was applied between the immersion and the plating. While the sonication promotes the mass transport of suppressor, a time delay allows more time for suppressor to diffuse into the narrow features. The suppressor concentration gradient is believed to be mitigated in both cases and therefore resulted in conformal plating.
Figure 1. Cross sectional SEM images of 80 nm wide lines after 6 second plating with (a) hot entry, (b) hot entry and sonication during plating, (c) 1 second delay of the plating current after immersing the substrate into the solution.
As the interconnects keep scaling the sub-100 nm trenches are filled within a short period of plating time, generally within 5 seconds. Therefore, suppressors that quickly establish a suppressing state and a big contrast between the plating rates at top and bottom of the trenches are in demand. While an ultimate correlation between the suppressor molecular structure and its filling performance is not available and is difficult to establish at this point, electrochemical screening methods are needed to identify the promising suppressor candidates. In a previous paper, we developed a screening method using potential transients (galvanostatic plating) with injection of the full amount of accelerator and different partial amounts of suppressor.28 Potential transient in presence of full amounts of both accelerator and suppressor represents the plating situation at the opening of a trench. Due to the presence of a suppressor concentration gradient, the plating at different depth of the trench was experimentally simulated by the transients with full amount of accelerator and different fractions of the full amount of suppressor. The difference between the potential transients with full and partial doses of suppressor was used to evaluate the difference in the suppression of the plating between the top and the bottom of the trench.
While a large difference between the potential transients with partial and full doses of suppressor was correlated to a better filling performance, the potential transients do not provide direct information on the super filling rates. A larger difference in the potential does not always guarantee a larger difference in the current, i.e. the plating rate. The later also depends on the polarization behavior of the chemistry. In this paper, we further developed a quasi-potentiostatic plating method for suppressor screening. This method is to better electrochemically simulate the real plating of the patterned substrates and to provide a more direct comparison of the filling rates for different suppressors and for different plating conditions.
Experimental
An AUTOLAB potentiostat-galvanostat electrochemical system was used for the study. Cu plating was carried out using an electrolyte containing 0.63 M CuSO4, 0.1 M H2SO4 and 1.5 mM HCl. Different experimental versions of the CUPUR suppressors from BASF SE were used for suppressor screening development and filling tests. In these studies, the different suppressors were compared at a fixed, vendor recommended dose. The concentration of the accelerator, bis-(sodium sulfopropyl)-disulfide or SPS, was kept constant. Two other commercial additive packages, Chemistry A and C, were used in the study as well with the vendor-recommended doses.
Chronopotentiometry (potential transients) and chronoamperometry (current transients) were acquired to electrochemically characterize the additives. A typical three-compartment cell with a glass frit to separate the anode and cathode was used. Pt mesh was used as the anode in the anode compartment. A saturated mercury sulfate electrode (MSE) was used as the reference. A Pt rotating disk electrode (RDE) was used at 100 rpm as the cathode. The Pt RDE was pre-plated at −20 mA/cm2 for 10 second in the additive-free virgin make-up solution (VMS) before each transient characterization. During a typical transient study, 50 mL VMS premixed with the calculated amount of additives was added into 150 mL VMS, where a magnetic stirrer was used for mixing. In this paper, the different amounts of suppressors are referred to with respect to the recommended dose. For exmple, 1× and 1/5× refer to one portion of and one fifty of recommended dose, respectively.
Filling experiments were performed on 2 cm × 2 cm patterned chips with 20 nm PVD Cu seed. The structures of interest are nested trenches, around 40 nm wide and 140 nm deep. The post-liner-seed aspect ratio of the different trenches varied from 3 to 5. Chips were plated with constant current without agitation. Hot entry was used and the time of the plating was controlled with an electronic timer-switch. The galvanostat was turned on before the sample being immersed, and the electrochemical system was at overload mode. As soon as the sample was being immersed, the potential dropped to the value corresponding to the applied plating current and the timer started ticking. The circuit was opened (switched off) once the plating time was up. This electronic switch allowed an accurate control of the plating time with hot entry. The chips were then cleaved perpendicular to the trenches and coated with a carbon-based ink. The cleaved samples were polished with a focused ion beam (FIB) after a portion of the cleaved edge being protected with Pt (in the FIB tool). After polishing, the prepared samples were imaged with a scanning electron microscope (SEM) at the cross section.
Results and Discussion
As discussed in the introduction, a potential transient method with the injection of full amount of accelerator and different fractions of the full amount of suppressor was developed previously to simulate the different additive adsorption rates across the depth of the trench.28 Figure 2 shows the typical potential transient curves and the method of data extraction from the curves. Upon the injection of the additives in galvanostatic plating, the potential shifts negatively due to the fast adsorption of suppressors onto the Cu surface. A positive shift of the potential follows afterwards as the accelerator displaces the suppressor. The potential response suggests how suppressed or how accelerated the plating is at any moment when there is a fractional amount of suppressors. Since a suppressor concentration gradient is present across the depth of the trench, the difference between these potential curves simulates the suppression of the plating at different locations across the depth of trenches.
Figure 2. (a) Typical potential transients at galvanostatic plating with the injection of full dose of accelerator and (green solid line)1×, (red dash line) 1/5×, (purple dash dot line) 1/20×, and (black long dash line) 1/100× of the full dose of suppressor. (b) Zoom in potential transients with illustrations of data extraction from the curves.
As illustrated by Figure 2b, at Δt after the injection, the plating potential at the location where 1× of suppressor is present in the solution is higher, by ΔE1, than the location where only 1/5× of suppressor is present. This potential difference is ΔE2 between the locations where 1/5× and 1/20× of suppressor are available. A shielding coefficient was introduced to describe the ratio of the suppressor adsorption onto the sidewall and onto the bottom of the trenches. This shielding coefficient tells the equivalent fraction of the suppressor that is available at bottom of the trenches in terms of the suppressor adsorption rate. It can be calculated with numerical simulation or estimated with an approximate equation.28 For trenches with aspect ratio of 3.5, the equivalent suppressor concentration at the bottom of the trench is about 1/20 of the concentration at the top of the trench. Therefore, the larger ΔE1 and ΔE2 are, the more difference is between the plating at the top and at the bottom of the trenches. However, this difference is actually the potential difference between the top and the bottom of the trench if they were plated at the same current.
In the real plating process, the whole wafer, including the trenches as well, is plated with a continuous Cu seed. Therefore the different locations along the depth of a trench are plated at a same potential. While there might be a potential difference across the whole wafer when the Cu seed layer is extremely thin or when more resistive seed materials are used, the potential difference across the trench depth is negligible. At a same potential, the bottom and the top of the trench are plated at different rates as a result from the different additive adsorption behaviors at these different locations. For a state-of-the-art interconnect wafer, structures with different dimensions are present and are plated at the same time. The sub-100 nm features are a small portion of the total wafer area. Therefore, once the wafer is immersed into the solution, the plating potential is determined by the plating behavior of the so-called field region, or the blanket area, and the top of the trenches. This plating potential at the field region changes along the time and this potential profile depends on the adsorption and desorption of the additives on the Cu surface. The plating at the bottom of the trenches is controlled by this potential profile. But it also depends on the additive diffusion and adsorption behavior at the bottom of the trench. Therefore, a current transient measurement controlled by that potential profile with the injection of different doses of the additives is believed to better simulate the plating scenario at the different depths of a trench.
A suppressor screening method was therefore developed consisting of two types of transient experiments. A first experiment was carried out to acquire the potential transient when the full doses of accelerator and suppressor were injected. This potential transient represents the plating potential for the top of the trench and the field region or the whole wafer. The second plating experiment was controlled by this potential transient profile. A full dose of accelerator and a fractional dose of the suppressor were injected together at the same time when the additives were injected in the first experiment. Figure 3 shows the experimental results for the commercial Chemistry C, including a potential transient from the first experiment and the current transients from the second experiment. The additives were injected at t = 50 sec in all experiments, to simulate a wafer plating with the wafer immersed into the solution at t = 50 sec.
Figure 3. (a) Potential transients acquired during the galvanostatic plating experiments with the injection of full doses of accelerator and suppressor; (b) Potential profile applied in the current transient experiments; (c-h) The current transients acquired with the plating controlled by the potential profile: (c) without any additive injection and (d-h) with the injection of 1× accelerator and (d) 1×, (e) 1/4×, (f) 1/10×, (g) 1/20×, (h) 1/100× suppressor.
In Figure 3, a constant current density of −3 mA/cm2 was used in the first experiment. The curve (a) is the potential transient with the injection of 1× suppressor and 1× accelerator at t = 50 sec. Curve (b) is the potential used in the second experiment to acquire the current transients. In the real practice, the AUTOLAB was operated in the so called sweep-and-hold mode. The curve (b) is a segmented linear approximation of the potential profile of curve (a). While the digital-to-analog converted signals would simulate the true potential profile more accurately, the linear approximation is believed to simulate the profile well enough for the studies. A small leading time of 1 sec was purposely added to the curve (b) as compared to the original potential profile, curve (a). In addition, different partial doses of suppressor and full dose of accelerator were injected at t = 50.2 sec (instead of 50 sec) during the second experiment. The total delay of 1.2 sec of the additive injection with respective to the potential profile is to avoid the chance where the additives in the second experiment are added too early due to experimental variation, which will introduce a current dip in the transients. While having additives at the bottom of the trench earlier than at the top is against reality, this 1.2 sec delay is expected to create a small over-estimation in the difference between the plating rates at the bottom and the top of the trenches. Curves (c) and (d) are two current transients acquired with two different additive concentrations with the plating controlled with the potential profile (b).
Curve (c) was acquired in the VMS without adding any additives, representing the highest possible current density that can be achieved at the bottom of the trench, where no suppressor is available. While an injection of accelerator alone would be a more accurate representation, its difference from curve (c) was negligible (not shown). Curve (d) is the current transient with the injection of 1× suppressor and 1× accelerator at t = 50 sec. This curve represents the lowest current density that can be possibly achieved at the bottom of the trench, where the suppressor diffuses fast or the trench is shallow and the suppressor concentration is the same as the bulk. It is evident that the current density in curve (d) stays constant at about −3 mA/cm2 except for an initial current spike. The spike lasted for about 3 seconds and is related to the 1.2-sec leading time of the potential profile (b) with respect to the injection of the additives in the second experiment. However, this short-lasting current spike suggests that the aforementioned 1.2-sec delay in the additive does not affect the current transients beyond a few seconds. In addition, the constant current observed after the spike at about –3 mA/cm2, the same as the current density used in the first experiment, further validates the experiments. This constant current is a result of the mutual annihilation of the effects of two processes. Firstly, the applied potential shifts negatively at t = 50 sec (ignore the 1.2 sec difference) and then changes according to the pre-acquired potential profile. Secondly, the additive adsorbs and/or desorbs at Cu surface, starting at the same t = 50 sec (ignore the 1.2 sec difference). From the first experiment, the potential applied here is the potential required to maintain the constant current density of −3 mA/cm2 when 1× suppressor and 1× accelerator are injected at t = 50 sec. As long as the applied potential profile and the additive behavior at the Cu surface are synchronized in the second experiment, these two effects cancel each other and the current in curve (d) maintains at −3 mA/cm2. In other words, if the suppressor and accelerator concentrations at the bottom of the trench were the same as at the top of the trench, the plating rates at these two locations would have been the same and stays constant.
Current transient curves (e) to (h) represent the different intermediate cases between (c) and (d), where different fractional amounts of suppressor are injected. As discussed above, a concentration gradient of suppressor is established due to the slow diffusion and fast adsorption of suppressor. Due to the small opening and high aspect ratio of the trenches, forced convection is believed to be minimal inside the trenches. While numerical simulation of the suppressor diffusion and adsorption can be used to calculate the suppressor concentration at any point, a simplified equation was previously developed to estimate the equivalent suppressor concentration at the bottom of a trench.28 For a typical trench structure in the state-of-the-art interconnects, the post liner-seed aspect ratio is about 3 to 6. The estimated equivalent suppressor concentration at the bottom of the trench is between 1/17 to 1/40 of the bulk concentration. For a trench structure with non-ideal etch profile and/or PVD liner-seed profile, this ratio can be even bigger. Fractional doses of 1/4 to 1/100 of the recommended dose of suppressor were used in the second experiment.
When 1/4× suppressor was injected, the current transient is similar to the case where 1× was injected, current approximately stays at constant after a short current spike at the beginning. However, when 1/10× or less suppressor was injected, the current further increases beyond the time range of the spike. This current increase is due to a much slower suppressing film formation on the Cu surface at a much lower suppressor concentration. In detail, the suppressors start to adsorb at the bottom of the trench, resulting in a suppressed state of plating. However, the suppressing effect at the top of the trench builds up much faster. The potential change overwhelms the formation of the suppressing film at the bottom of the trench and therefore the current continues to increase. As the suppression effect at the top of the trench saturates the potential stops increasing anymore. Furthermore, due to the presence of the accelerator, the potential starts to shift positively after the saturation. On the other hand, the suppression effect at the bottom of the trench builds up much more slowly and does not saturate until much later. Therefore in a longer term, the suppression effect formation at the trench bottom dominates and a current decrease was observed, confirmed by Figure 3. A peak current was therefore observed, indicating the maximum possible plating rate at the bottom of the trench.
The super filling ratio is defined as the ratio of the plating rate at the trench bottom over the rate at the trench top. In the current transient studies in Figure 3, the plating rate at trench top is the current that was applied during the potential transient experiment, −3 mA/cm2. In an effort of comparing the super filling performance at different current densities, three sets of experiments were carried out at three different plating current densities. Each set of experiments includes a potential transient at the applied current density, and a series of current transients for each of the acquired potential profiles.
Figures 4a – 4c show the current transients normalized with the plating current densities of −3, −5, and −10 mA/cm2. Only three fractional injections of suppressor, 1/4 ×, 1/20 ×, and 1/100 × are presented. Assuming an aspect ratio of 3.5 for the trench structures, the equivalent suppressor dilution at the trench bottom is 1/20. The maximum super filling ratios were about 10, 6.4, and 3.2 for the three current densities used. In other words, the ratio between the plating rate at the trench bottom and the plating rate at the top is much higher at a lower plating current density. Figure 5 shows the cross sectional SEM images of 40 nm wide trenches plated at −3, −5, and −10 mA/cm2. While the total charge of plating was the same for the three cases, the trench was filled much faster at the lower current density. This observation on the 40 nm wide trenches confirms the prediction from the current transients in Figure 4. Due to the super filling in the trenches, the average current density on a patterned substrate is higher than the current density in the field region, or at the top of the trenches. Therefore the observation in Figure 5 is a qualitative rather than a quantitative confirmation of the super filling ratio observed in Figure 4.
Figure 4. (a-c) Normalized current transients acquired during the second experiment, where the (d) potential profile was acquired with a galvanostatic plating at (a, d1) −3, (b, d2) −5, (c, d3) −10 mA/cm2. The suppressor dose fractions are (dashed line) 1/4, (solid line) 1/20, and (long dashed line) 1/100.
Figure 5. Cross sectional SEM images of 40 nm wide trenches plated at (a) −3, (b) −5, (c) −10 mA/cm2 for (a) 3.4, (b) 2, (c) 1 sec.
An interesting observation in the transient curves is that the peak currents were about the same across the three studies in spite of the different applied potentials, shown in Figure 4d. The potential difference between −3 and −5 mA/cm2 remains at a constant of about 50 mV. This 50 mV resulted in a small difference between the peak current in the current transients, −30 and −32 mA/cm2, corresponding to a super filling ratio of 10, and 6.4. On the other hand, while the potential difference between −5 and −10 mA/cm2 is also about 50 mV in VMS, the difference is smaller after the injection of the additives. When the current transient is concerned, the effect of this smaller potential difference was even further diminished. Both potential profiles resulted in a peak current of −32 mA/cm2, corresponding to the super filling ratio of 6.4 and 3.2.
As the interconnect structure continues to scale, the thickness of the liner and Cu seed decreases, too. The dissolution of the Cu seed, especially the air-oxidized Cu seed, becomes more of a concern in the plating. In addition, the thinner Cu seed demands a good nucleation on a possibly partially exposed liner material.29 In general, a higher bias or a higher current density is beneficial for nucleation and seed protection. Therefore, a plating chemistry that enables a good super filling at a high current density is in demand. The transient method described above provides a screening method to compare or even predict the filling performance on trenches of different aspect ratios at different current densities. This method can be further extended to structures of non-ideal geometries as well.
Figure 6 shows the normalized current transients and the filling performance for five different experimental versions of BASF CUPUR suppressors. A plating current of −3 mA/cm2 was used for both the screening transients and the filling experiments. It is evident from the SEM images that the suppressors (ii) and (iv) resulted in a much more conformal filling, or a much lower super filling ratio. Among the other three, the suppressor (v) completely filled the trench and (iii) almost filled the trench with a very nice bottom up growth profile. While suppressor (i) also showed nice super filling ratio as compared with (ii) and (iv), some trench to trench variation was observed on the super filling ratio. All the current transients of these different suppressors showed a maximum super filling ratio from 8 to 10, similar to chemistry C in Figure 4. However, the suppressors differ drastically in how long that peak current lasts in the transient study. All the three suppressors that provide a good filling performance, (i) (iii) and (v), showed a much longer peak current than the two suppressors that don't, (ii) and (iv). The filling experiment was carried out on trenches with straighter side walls and aspect ratio of 5, of which the equivalent dilution factor for suppressor is 1/35. Although this difference in the duration of the peak current is less obvious for 1/20 dilution cases, it is much more pronounced for 1/100 dilution.
Figure 6. Normalized current transients in the second experiments and the filling performance for five different suppressors, (i) to (v), at a plating current density of −3 mA/cm2. The 40 nm wide trenches were plated for 3.4 sec. The suppressor dose fractions are (dashed line) 1/4, (solid line) 1/20, and (long dashed line) 1/100.
Figure 7 shows the normalized current transients and the cross sectional SEM images of plated trenches for two other chemistries, Chemistry A and C, at a higher current density of –5 mA/cm2. Figure 7a is a replot of Figure 4b. Different peak currents and different durations of the peak current were observed for the two chemistries. With 1/100× suppressor injection, Chemistry C reached a super filling rate of 6.4 while chemistry A only reached 5.5. With 1/20× suppressor, this difference becomes even bigger, 6.3 vs. 3.1. In addition, the peak current of Chemistry A lasts much shorter than Chemistry C. These differences are believed to result in the much slower super filling or much more conformal plating for Chemistry A as observed in the SEM images.
Figure 7. Normalized current transients in the second experiments and the filling performance for (a,c) Chemistry C, and (b,d) Chemistry A at a plating current density of −5 mA/cm2. The 40 nm wide trenches were plated for 2 sec. The suppressor dose fractions are (dashed line) 1/4, (solid line) 1/20, and (long dashed line) 1/100.
While the above results demonstrated a method to screen different suppressors, the method is rather qualitative. One main concern of this method is that the transient study spans for more than 100 seconds while the trenches are filled within 5 seconds in real practice. Ideally, in order to accurately characterize the chemistry within a few seconds of plating, the transient experiments described here need to be carried out with the premixed plating solutions to avoid the mixing effect and with electronic trigger to avoid the current spike (discussed in Figure 3). More quantitative comparison and more accurate suppressor screening are expected by using the hot-entry transients into the premixed solutions. The method described in this paper uses transients in a longer time scale, which yet reflect the nature of the suppressors and qualitatively correlate to the behaviors of the suppressors at a shorter time scale. The method is believed to be one of the many steps toward an ultimate method to electrochemically characterize and predict the super filling performance of additives.
Another concern of the screening method described here might be the absence of the leveler. In a real wafer plating process, a third additive, leveler, is generally used to mitigate the surface topography created by the momentum plating, a result of super filling.30–32 It is desired and designed in most commercial chemistry packages that the leveler is neither absent inside the trenches at all or exerts little impact to the behavior of the suppressor and accelerator inside the trenches in real practice. The second possible way of changing the super filling by the leveler is that the presence of leveler in the bulk solution can change the potential bias applied to the wafer, or the potential transients acquired during the first experiment.33 However, this leveler effect is only present in a longer time scale, typically after about 20 seconds,33 as compared with the filling time, typically within 5 seconds. The potential response within a few seconds is independent of the presence of leveler and is only determined by the suppressor and accelerator. Therefore, the transient experiments in this paper were carried out without leveler for the screening of suppressors with regards to their filling performance.
Finally, the transient method discussed above is not able test the CEAC behavior. In other words, it does not provide information on how the suppressors influence the accelerator coverage change in response to the Cu surface area shrinkage. Figure 8 shows the normalized current transients in an attempt to characterize the accelerator concentration effect. In this study, multiplied doses of the accelerator were injected together with fractional doses of suppressor. No effect of the accelerator concentration was observed on either the peak current height or the duration. While the accelerator coverage is expected to change as the Cu surface curvature evolves during plating, this effect on the surface adsorption could not be easily simulated by having more accelerators in the solution. As the surface shrinkage during the super filling of the vias is more pronounced than the filling of the trenches, the CEAC effect is expected to be stronger during the filling of the vias. However, the method described in this paper is believed to be extendable for the filling of the vias as a qualitative screening method for the suppressor molecules. The suppressor concentration gradient in a via can be simulated or computed in a similar way as for trenches.28
Figure 8. Normalized current transients in the second experiments for Chemistry C at a plating current density of −3 mA/cm2 and with the injection of (a) 1/4× (b) 1/20× suppressor and (a1, b1) 1×, (a2, b2) 2×, and (a3, b3) 5× of acceelrator.
Conclusions
In an effort to electrochemically identify suppressor molecules for good super filling performance, a two-step transient method was developed as an improvement from the previous potential transient method.28 A potential transient during galvanostatic plating with full amount of suppressor was acquired firstly to simulate the potential applied on the wafer during the plating, which is largely determined by the plating of the field region on the wafer. While this potential is not a constant, it is used to control the second plating experiment to mimic the plating at different locations across the wafer. Due to the slow diffusion and fast adsorption of suppressor, a concentration gradient is present across the depth of the trenches. Therefore, different fractional amounts of suppressor are injected into the second experiment to simulate plating inside the trench structure. The current transients acquired thereof reflect the plating rates across the depth of the trench. A lower super filling ratio at higher plating current density predicted by the transient method was well confirmed by filling experiments. A good correlation was also observed between the current transients and the filling performance for different suppressor molecules. While this correlation is still qualitative, hot entry transient experiments with improved control is expected to allow more quantitative comparison of the suppressors.
Acknowledgment
Dr. Alexander Fluegel at BASF SE is acknowledged for helpful discussions. The Nano Science and Technology Laboratory (NSTL) at the IBM T.J. Watson Research Center and the IBM Albany Nanotech Center are greatly appreciated for providing the substrates for the study. BASF SE is also acknowledged for supplying the experimental suppressor molecules.